diff options
Diffstat (limited to '')
74 files changed, 98015 insertions, 0 deletions
diff --git a/netwerk/sctp/datachannel/DataChannel.cpp b/netwerk/sctp/datachannel/DataChannel.cpp new file mode 100644 index 0000000000..864a0b79ac --- /dev/null +++ b/netwerk/sctp/datachannel/DataChannel.cpp @@ -0,0 +1,3438 @@ +/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */ +/* vim: set ts=2 et sw=2 tw=80: */ +/* This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this file, + * You can obtain one at http://mozilla.org/MPL/2.0/. */ + +#include <algorithm> +#include <stdio.h> +#include <stdlib.h> +#if !defined(__Userspace_os_Windows) +# include <arpa/inet.h> +#endif +// usrsctp.h expects to have errno definitions prior to its inclusion. +#include <errno.h> + +#define SCTP_DEBUG 1 +#define SCTP_STDINT_INCLUDE <stdint.h> + +#ifdef _MSC_VER +// Disable "warning C4200: nonstandard extension used : zero-sized array in +// struct/union" +// ...which the third-party file usrsctp.h runs afoul of. +# pragma warning(push) +# pragma warning(disable : 4200) +#endif + +#include "usrsctp.h" + +#ifdef _MSC_VER +# pragma warning(pop) +#endif + +#include "nsServiceManagerUtils.h" +#include "nsIInputStream.h" +#include "nsIPrefBranch.h" +#include "nsIPrefService.h" +#include "mozilla/Services.h" +#include "mozilla/Sprintf.h" +#include "nsProxyRelease.h" +#include "nsThread.h" +#include "nsThreadUtils.h" +#include "nsNetUtil.h" +#include "nsNetCID.h" +#include "mozilla/RandomNum.h" +#include "mozilla/StaticMutex.h" +#include "mozilla/UniquePtrExtensions.h" +#include "mozilla/Unused.h" +#include "mozilla/dom/RTCDataChannelBinding.h" +#include "mozilla/dom/RTCStatsReportBinding.h" +#include "mozilla/media/MediaUtils.h" +#ifdef MOZ_PEERCONNECTION +# include "transport/runnable_utils.h" +# include "jsapi/MediaTransportHandler.h" +# include "mediapacket.h" +#endif + +#include "DataChannel.h" +#include "DataChannelProtocol.h" + +// Let us turn on and off important assertions in non-debug builds +#ifdef DEBUG +# define ASSERT_WEBRTC(x) MOZ_ASSERT((x)) +#elif defined(MOZ_WEBRTC_ASSERT_ALWAYS) +# define ASSERT_WEBRTC(x) \ + do { \ + if (!(x)) { \ + MOZ_CRASH(); \ + } \ + } while (0) +#endif + +namespace mozilla { + +LazyLogModule gDataChannelLog("DataChannel"); +static LazyLogModule gSCTPLog("SCTP"); + +#define SCTP_LOG(args) \ + MOZ_LOG(mozilla::gSCTPLog, mozilla::LogLevel::Debug, args) + +static void debug_printf(const char* format, ...) { + va_list ap; + char buffer[1024]; + + if (MOZ_LOG_TEST(gSCTPLog, LogLevel::Debug)) { + va_start(ap, format); +#ifdef _WIN32 + if (vsnprintf_s(buffer, sizeof(buffer), _TRUNCATE, format, ap) > 0) { +#else + if (VsprintfLiteral(buffer, format, ap) > 0) { +#endif + SCTP_LOG(("%s", buffer)); + } + va_end(ap); + } +} + +class DataChannelRegistry { + public: + NS_INLINE_DECL_THREADSAFE_REFCOUNTING(DataChannelRegistry) + + static uintptr_t Register(DataChannelConnection* aConnection) { + StaticMutexAutoLock lock(sInstanceMutex); + uintptr_t result = EnsureInstance()->RegisterImpl(aConnection); + DC_DEBUG( + ("Registering connection %p as ulp %p", aConnection, (void*)result)); + return result; + } + + static void Deregister(uintptr_t aId) { + RefPtr<DataChannelRegistry> maybeTrash; + + { + StaticMutexAutoLock lock(sInstanceMutex); + DC_DEBUG(("Deregistering connection ulp = %p", (void*)aId)); + if (NS_WARN_IF(!Instance())) { + return; + } + Instance()->DeregisterImpl(aId); + if (Instance()->Empty()) { + // Unset singleton inside mutex lock, but don't call Shutdown until we + // unlock, since that involves calling into libusrsctp, which invites + // deadlock. + maybeTrash = Instance().forget(); + } + } + } + + static RefPtr<DataChannelConnection> Lookup(uintptr_t aId) { + StaticMutexAutoLock lock(sInstanceMutex); + if (NS_WARN_IF(!Instance())) { + return nullptr; + } + return Instance()->LookupImpl(aId); + } + + private: + // This is a singleton class, so don't let just anyone create one of these + DataChannelRegistry() { + ASSERT_WEBRTC(NS_IsMainThread()); + mShutdownBlocker = media::ShutdownBlockingTicket::Create( + u"DataChannelRegistry::mShutdownBlocker"_ns, + NS_LITERAL_STRING_FROM_CSTRING(__FILE__), __LINE__); + InitUsrSctp(); + } + + static RefPtr<DataChannelRegistry>& Instance() { + static RefPtr<DataChannelRegistry> sRegistry; + return sRegistry; + } + + static RefPtr<DataChannelRegistry>& EnsureInstance() { + ASSERT_WEBRTC(NS_IsMainThread()); + if (!Instance()) { + Instance() = new DataChannelRegistry(); + } + return Instance(); + } + + uintptr_t RegisterImpl(DataChannelConnection* aConnection) { + ASSERT_WEBRTC(NS_IsMainThread()); + mConnections.emplace(mNextId, aConnection); + return mNextId++; + } + + void DeregisterImpl(uintptr_t aId) { + ASSERT_WEBRTC(NS_IsMainThread()); + mConnections.erase(aId); + } + + bool Empty() const { return mConnections.empty(); } + + RefPtr<DataChannelConnection> LookupImpl(uintptr_t aId) { + auto it = mConnections.find(aId); + if (NS_WARN_IF(it == mConnections.end())) { + DC_DEBUG(("Can't find connection ulp %p", (void*)aId)); + return nullptr; + } + return it->second; + } + + virtual ~DataChannelRegistry() { + ASSERT_WEBRTC(NS_IsMainThread()); + + if (NS_WARN_IF(!mConnections.empty())) { + MOZ_ASSERT(false); + mConnections.clear(); + } + + DeinitUsrSctp(); + } + +#ifdef SCTP_DTLS_SUPPORTED + static int SctpDtlsOutput(void* addr, void* buffer, size_t length, + uint8_t tos, uint8_t set_df) { + uintptr_t id = reinterpret_cast<uintptr_t>(addr); + RefPtr<DataChannelConnection> connection = DataChannelRegistry::Lookup(id); + if (NS_WARN_IF(!connection) || connection->InShutdown()) { + return 0; + } + return connection->SctpDtlsOutput(addr, buffer, length, tos, set_df); + } +#endif + + void InitUsrSctp() { +#ifndef MOZ_PEERCONNECTION + MOZ_CRASH("Trying to use SCTP/DTLS without dom/media/webrtc/transport"); +#endif + + DC_DEBUG(("Calling usrsctp_init %p", this)); + + usrsctp_init(0, DataChannelRegistry::SctpDtlsOutput, debug_printf); + + // Set logging to SCTP:LogLevel::Debug to get SCTP debugs + if (MOZ_LOG_TEST(gSCTPLog, LogLevel::Debug)) { + usrsctp_sysctl_set_sctp_debug_on(SCTP_DEBUG_ALL); + } + + // Do not send ABORTs in response to INITs (1). + // Do not send ABORTs for received Out of the Blue packets (2). + usrsctp_sysctl_set_sctp_blackhole(2); + + // Disable the Explicit Congestion Notification extension (currently not + // supported by the Firefox code) + usrsctp_sysctl_set_sctp_ecn_enable(0); + + // Enable interleaving messages for different streams (incoming) + // See: https://tools.ietf.org/html/rfc6458#section-8.1.20 + usrsctp_sysctl_set_sctp_default_frag_interleave(2); + + // Disabling authentication and dynamic address reconfiguration as neither + // of them are used for data channel and only result in additional code + // paths being used. + usrsctp_sysctl_set_sctp_asconf_enable(0); + usrsctp_sysctl_set_sctp_auth_enable(0); + } + + void DeinitUsrSctp() { + DC_DEBUG(("Calling usrsctp_finish %p", this)); + usrsctp_finish(); + } + + uintptr_t mNextId = 1; + std::map<uintptr_t, RefPtr<DataChannelConnection>> mConnections; + UniquePtr<media::ShutdownBlockingTicket> mShutdownBlocker; + static StaticMutex sInstanceMutex MOZ_UNANNOTATED; +}; + +StaticMutex DataChannelRegistry::sInstanceMutex; + +OutgoingMsg::OutgoingMsg(struct sctp_sendv_spa& info, const uint8_t* data, + size_t length) + : mLength(length), mData(data) { + mInfo = &info; + mPos = 0; +} + +void OutgoingMsg::Advance(size_t offset) { + mPos += offset; + if (mPos > mLength) { + mPos = mLength; + } +} + +BufferedOutgoingMsg::BufferedOutgoingMsg(OutgoingMsg& msg) { + size_t length = msg.GetLeft(); + auto* tmp = new uint8_t[length]; // infallible malloc! + memcpy(tmp, msg.GetData(), length); + mLength = length; + mData = tmp; + mInfo = new sctp_sendv_spa; + *mInfo = msg.GetInfo(); + mPos = 0; +} + +BufferedOutgoingMsg::~BufferedOutgoingMsg() { + delete mInfo; + delete[] mData; +} + +static int receive_cb(struct socket* sock, union sctp_sockstore addr, + void* data, size_t datalen, struct sctp_rcvinfo rcv, + int flags, void* ulp_info) { + DC_DEBUG(("In receive_cb, ulp_info=%p", ulp_info)); + uintptr_t id = reinterpret_cast<uintptr_t>(ulp_info); + RefPtr<DataChannelConnection> connection = DataChannelRegistry::Lookup(id); + if (!connection) { + // Unfortunately, we can get callbacks after calling + // usrsctp_close(socket), so we need to simply ignore them if we've + // already killed the DataChannelConnection object + DC_DEBUG(( + "Ignoring receive callback for terminated Connection ulp=%p, %zu bytes", + ulp_info, datalen)); + return 0; + } + return connection->ReceiveCallback(sock, data, datalen, rcv, flags); +} + +static RefPtr<DataChannelConnection> GetConnectionFromSocket( + struct socket* sock) { + struct sockaddr* addrs = nullptr; + int naddrs = usrsctp_getladdrs(sock, 0, &addrs); + if (naddrs <= 0 || addrs[0].sa_family != AF_CONN) { + return nullptr; + } + // usrsctp_getladdrs() returns the addresses bound to this socket, which + // contains the SctpDataMediaChannel* as sconn_addr. Read the pointer, + // then free the list of addresses once we have the pointer. We only open + // AF_CONN sockets, and they should all have the sconn_addr set to the + // pointer that created them, so [0] is as good as any other. + struct sockaddr_conn* sconn = + reinterpret_cast<struct sockaddr_conn*>(&addrs[0]); + uintptr_t id = reinterpret_cast<uintptr_t>(sconn->sconn_addr); + RefPtr<DataChannelConnection> connection = DataChannelRegistry::Lookup(id); + usrsctp_freeladdrs(addrs); + + return connection; +} + +// Called when the buffer empties to the threshold value. This is called +// from SctpDtlsInput() through the sctp stack. SctpDtlsInput() calls +// usrsctp_conninput() under lock +static int threshold_event(struct socket* sock, uint32_t sb_free, + void* ulp_info) { + RefPtr<DataChannelConnection> connection = GetConnectionFromSocket(sock); + connection->mLock.AssertCurrentThreadOwns(); + if (connection) { + connection->SendDeferredMessages(); + } else { + DC_ERROR(("Can't find connection for socket %p", sock)); + } + return 0; +} + +DataChannelConnection::~DataChannelConnection() { + DC_DEBUG(("Deleting DataChannelConnection %p", (void*)this)); + // This may die on the MainThread, or on the STS thread, or on an + // sctp thread if we were in a callback when the DOM side shut things down. + ASSERT_WEBRTC(mState == CLOSED); + MOZ_ASSERT(!mMasterSocket); + MOZ_ASSERT(mPending.GetSize() == 0); + + if (!IsSTSThread()) { + // We may be on MainThread *or* on an sctp thread (being called from + // receive_cb() or SctpDtlsOutput()) + if (mInternalIOThread) { + // Avoid spinning the event thread from here (which if we're mainthread + // is in the event loop already) + nsCOMPtr<nsIRunnable> r = WrapRunnable( + nsCOMPtr<nsIThread>(mInternalIOThread), &nsIThread::AsyncShutdown); + Dispatch(r.forget()); + } + } else { + // on STS, safe to call shutdown + if (mInternalIOThread) { + mInternalIOThread->Shutdown(); + } + } +} + +void DataChannelConnection::Destroy() { + // Though it's probably ok to do this and close the sockets; + // if we really want it to do true clean shutdowns it can + // create a dependant Internal object that would remain around + // until the network shut down the association or timed out. + DC_DEBUG(("Destroying DataChannelConnection %p", (void*)this)); + ASSERT_WEBRTC(NS_IsMainThread()); + CloseAll(); + + MutexAutoLock lock(mLock); + // If we had a pending reset, we aren't waiting for it - clear the list so + // we can deregister this DataChannelConnection without leaking. + ClearResets(); + + MOZ_ASSERT(mSTS); + ASSERT_WEBRTC(NS_IsMainThread()); + mListener = nullptr; + // Finish Destroy on STS thread to avoid bug 876167 - once that's fixed, + // the usrsctp_close() calls can move back here (and just proxy the + // disconnect_all()) + RUN_ON_THREAD(mSTS, + WrapRunnable(RefPtr<DataChannelConnection>(this), + &DataChannelConnection::DestroyOnSTS, mSocket, + mMasterSocket), + NS_DISPATCH_NORMAL); + + // These will be released on STS + mSocket = nullptr; + mMasterSocket = nullptr; // also a flag that we've Destroyed this connection + + // We can't get any more *new* callbacks from the SCTP library + + // All existing callbacks have refs to DataChannelConnection - however, + // we need to handle their destroying the object off mainthread/STS + + // nsDOMDataChannel objects have refs to DataChannels that have refs to us +} + +void DataChannelConnection::DestroyOnSTS(struct socket* aMasterSocket, + struct socket* aSocket) { + if (aSocket && aSocket != aMasterSocket) usrsctp_close(aSocket); + if (aMasterSocket) usrsctp_close(aMasterSocket); + + usrsctp_deregister_address(reinterpret_cast<void*>(mId)); + DC_DEBUG( + ("Deregistered %p from the SCTP stack.", reinterpret_cast<void*>(mId))); +#ifdef MOZ_DIAGNOSTIC_ASSERT_ENABLED + mShutdown = true; + DC_DEBUG(("Shutting down connection %p, id %p", this, (void*)mId)); +#endif + + disconnect_all(); + mTransportHandler = nullptr; + GetMainThreadSerialEventTarget()->Dispatch(NS_NewRunnableFunction( + "DataChannelConnection::Destroy", + [id = mId]() { DataChannelRegistry::Deregister(id); })); +} + +Maybe<RefPtr<DataChannelConnection>> DataChannelConnection::Create( + DataChannelConnection::DataConnectionListener* aListener, + nsISerialEventTarget* aTarget, MediaTransportHandler* aHandler, + const uint16_t aLocalPort, const uint16_t aNumStreams, + const Maybe<uint64_t>& aMaxMessageSize) { + ASSERT_WEBRTC(NS_IsMainThread()); + + RefPtr<DataChannelConnection> connection = new DataChannelConnection( + aListener, aTarget, aHandler); // Walks into a bar + return connection->Init(aLocalPort, aNumStreams, aMaxMessageSize) + ? Some(connection) + : Nothing(); +} + +DataChannelConnection::DataChannelConnection( + DataChannelConnection::DataConnectionListener* aListener, + nsISerialEventTarget* aTarget, MediaTransportHandler* aHandler) + : NeckoTargetHolder(aTarget), + mLock("netwerk::sctp::DataChannelConnection"), + mListener(aListener), + mTransportHandler(aHandler) { + DC_VERBOSE(("Constructor DataChannelConnection=%p, listener=%p", this, + mListener.get())); +#ifdef MOZ_DIAGNOSTIC_ASSERT_ENABLED + mShutdown = false; +#endif +} + +bool DataChannelConnection::Init(const uint16_t aLocalPort, + const uint16_t aNumStreams, + const Maybe<uint64_t>& aMaxMessageSize) { + ASSERT_WEBRTC(NS_IsMainThread()); + + struct sctp_initmsg initmsg; + struct sctp_assoc_value av; + struct sctp_event event; + socklen_t len; + + uint16_t event_types[] = { + SCTP_ASSOC_CHANGE, SCTP_PEER_ADDR_CHANGE, + SCTP_REMOTE_ERROR, SCTP_SHUTDOWN_EVENT, + SCTP_ADAPTATION_INDICATION, SCTP_PARTIAL_DELIVERY_EVENT, + SCTP_SEND_FAILED_EVENT, SCTP_STREAM_RESET_EVENT, + SCTP_STREAM_CHANGE_EVENT}; + { + // MutexAutoLock lock(mLock); Not needed since we're on mainthread always + mLocalPort = aLocalPort; + SetMaxMessageSize(aMaxMessageSize.isSome(), aMaxMessageSize.valueOr(0)); + } + + mId = DataChannelRegistry::Register(this); + + // XXX FIX! make this a global we get once + // Find the STS thread + nsresult rv; + mSTS = do_GetService(NS_SOCKETTRANSPORTSERVICE_CONTRACTID, &rv); + MOZ_ASSERT(NS_SUCCEEDED(rv)); + + // Open sctp with a callback + if ((mMasterSocket = usrsctp_socket( + AF_CONN, SOCK_STREAM, IPPROTO_SCTP, receive_cb, threshold_event, + usrsctp_sysctl_get_sctp_sendspace() / 2, + reinterpret_cast<void*>(mId))) == nullptr) { + return false; + } + + int buf_size = 1024 * 1024; + + if (usrsctp_setsockopt(mMasterSocket, SOL_SOCKET, SO_RCVBUF, + (const void*)&buf_size, sizeof(buf_size)) < 0) { + DC_ERROR(("Couldn't change receive buffer size on SCTP socket")); + goto error_cleanup; + } + if (usrsctp_setsockopt(mMasterSocket, SOL_SOCKET, SO_SNDBUF, + (const void*)&buf_size, sizeof(buf_size)) < 0) { + DC_ERROR(("Couldn't change send buffer size on SCTP socket")); + goto error_cleanup; + } + + // Make non-blocking for bind/connect. SCTP over UDP defaults to non-blocking + // in associations for normal IO + if (usrsctp_set_non_blocking(mMasterSocket, 1) < 0) { + DC_ERROR(("Couldn't set non_blocking on SCTP socket")); + // We can't handle connect() safely if it will block, not that this will + // even happen. + goto error_cleanup; + } + + // Make sure when we close the socket, make sure it doesn't call us back + // again! This would cause it try to use an invalid DataChannelConnection + // pointer + struct linger l; + l.l_onoff = 1; + l.l_linger = 0; + if (usrsctp_setsockopt(mMasterSocket, SOL_SOCKET, SO_LINGER, (const void*)&l, + (socklen_t)sizeof(struct linger)) < 0) { + DC_ERROR(("Couldn't set SO_LINGER on SCTP socket")); + // unsafe to allow it to continue if this fails + goto error_cleanup; + } + + // XXX Consider disabling this when we add proper SDP negotiation. + // We may want to leave enabled for supporting 'cloning' of SDP offers, which + // implies re-use of the same pseudo-port number, or forcing a renegotiation. + { + const int option_value = 1; + if (usrsctp_setsockopt(mMasterSocket, IPPROTO_SCTP, SCTP_REUSE_PORT, + (const void*)&option_value, + (socklen_t)sizeof(option_value)) < 0) { + DC_WARN(("Couldn't set SCTP_REUSE_PORT on SCTP socket")); + } + if (usrsctp_setsockopt(mMasterSocket, IPPROTO_SCTP, SCTP_NODELAY, + (const void*)&option_value, + (socklen_t)sizeof(option_value)) < 0) { + DC_WARN(("Couldn't set SCTP_NODELAY on SCTP socket")); + } + } + + // Set explicit EOR + { + const int option_value = 1; + if (usrsctp_setsockopt(mMasterSocket, IPPROTO_SCTP, SCTP_EXPLICIT_EOR, + (const void*)&option_value, + (socklen_t)sizeof(option_value)) < 0) { + DC_ERROR(("*** failed to enable explicit EOR mode %d", errno)); + goto error_cleanup; + } + } + + // Enable ndata + // TODO: Bug 1381145, enable this once ndata has been deployed +#if 0 + av.assoc_id = SCTP_FUTURE_ASSOC; + av.assoc_value = 1; + if (usrsctp_setsockopt(mMasterSocket, IPPROTO_SCTP, SCTP_INTERLEAVING_SUPPORTED, &av, + (socklen_t)sizeof(struct sctp_assoc_value)) < 0) { + DC_ERROR(("*** failed enable ndata errno %d", errno)); + goto error_cleanup; + } +#endif + + av.assoc_id = SCTP_ALL_ASSOC; + av.assoc_value = SCTP_ENABLE_RESET_STREAM_REQ | SCTP_ENABLE_CHANGE_ASSOC_REQ; + if (usrsctp_setsockopt(mMasterSocket, IPPROTO_SCTP, SCTP_ENABLE_STREAM_RESET, + &av, (socklen_t)sizeof(struct sctp_assoc_value)) < 0) { + DC_ERROR(("*** failed enable stream reset errno %d", errno)); + goto error_cleanup; + } + + /* Enable the events of interest. */ + memset(&event, 0, sizeof(event)); + event.se_assoc_id = SCTP_ALL_ASSOC; + event.se_on = 1; + for (unsigned short event_type : event_types) { + event.se_type = event_type; + if (usrsctp_setsockopt(mMasterSocket, IPPROTO_SCTP, SCTP_EVENT, &event, + sizeof(event)) < 0) { + DC_ERROR(("*** failed setsockopt SCTP_EVENT errno %d", errno)); + goto error_cleanup; + } + } + + memset(&initmsg, 0, sizeof(initmsg)); + len = sizeof(initmsg); + if (usrsctp_getsockopt(mMasterSocket, IPPROTO_SCTP, SCTP_INITMSG, &initmsg, + &len) < 0) { + DC_ERROR(("*** failed getsockopt SCTP_INITMSG")); + goto error_cleanup; + } + DC_DEBUG(("Setting number of SCTP streams to %u, was %u/%u", aNumStreams, + initmsg.sinit_num_ostreams, initmsg.sinit_max_instreams)); + initmsg.sinit_num_ostreams = aNumStreams; + initmsg.sinit_max_instreams = MAX_NUM_STREAMS; + if (usrsctp_setsockopt(mMasterSocket, IPPROTO_SCTP, SCTP_INITMSG, &initmsg, + (socklen_t)sizeof(initmsg)) < 0) { + DC_ERROR(("*** failed setsockopt SCTP_INITMSG, errno %d", errno)); + goto error_cleanup; + } + + mSocket = nullptr; + mSTS->Dispatch( + NS_NewRunnableFunction("DataChannelConnection::Init", [id = mId]() { + usrsctp_register_address(reinterpret_cast<void*>(id)); + DC_DEBUG(("Registered %p within the SCTP stack.", + reinterpret_cast<void*>(id))); + })); + + return true; + +error_cleanup: + usrsctp_close(mMasterSocket); + mMasterSocket = nullptr; + return false; +} + +// Only called on MainThread, mMaxMessageSize is read on other threads +void DataChannelConnection::SetMaxMessageSize(bool aMaxMessageSizeSet, + uint64_t aMaxMessageSize) { + ASSERT_WEBRTC(NS_IsMainThread()); + MutexAutoLock lock(mLock); + + if (mMaxMessageSizeSet && !aMaxMessageSizeSet) { + // Don't overwrite already set MMS with default values + return; + } + + mMaxMessageSizeSet = aMaxMessageSizeSet; + mMaxMessageSize = aMaxMessageSize; + + nsresult rv; + nsCOMPtr<nsIPrefService> prefs = + do_GetService("@mozilla.org/preferences-service;1", &rv); + if (!NS_WARN_IF(NS_FAILED(rv))) { + nsCOMPtr<nsIPrefBranch> branch = do_QueryInterface(prefs); + + if (branch) { + int32_t temp; + if (!NS_FAILED(branch->GetIntPref( + "media.peerconnection.sctp.force_maximum_message_size", &temp))) { + if (temp >= 0) { + mMaxMessageSize = (uint64_t)temp; + } + } + } + } + + // Fix remote MMS. This code exists, so future implementations of + // RTCSctpTransport.maxMessageSize can simply provide that value from + // GetMaxMessageSize. + + // TODO: Bug 1382779, once resolved, can be increased to + // min(Uint8ArrayMaxSize, UINT32_MAX) + // TODO: Bug 1381146, once resolved, can be increased to whatever we support + // then (hopefully + // SIZE_MAX) + if (mMaxMessageSize == 0 || + mMaxMessageSize > WEBRTC_DATACHANNEL_MAX_MESSAGE_SIZE_REMOTE) { + mMaxMessageSize = WEBRTC_DATACHANNEL_MAX_MESSAGE_SIZE_REMOTE; + } + + DC_DEBUG(("Maximum message size (outgoing data): %" PRIu64 + " (set=%s, enforced=%s)", + mMaxMessageSize, mMaxMessageSizeSet ? "yes" : "no", + aMaxMessageSize != mMaxMessageSize ? "yes" : "no")); +} + +uint64_t DataChannelConnection::GetMaxMessageSize() { + MutexAutoLock lock(mLock); + return mMaxMessageSize; +} + +void DataChannelConnection::AppendStatsToReport( + const UniquePtr<dom::RTCStatsCollection>& aReport, + const DOMHighResTimeStamp aTimestamp) const { + ASSERT_WEBRTC(NS_IsMainThread()); + nsString temp; + for (const RefPtr<DataChannel>& chan : mChannels.GetAll()) { + // If channel is empty, ignore + if (!chan) { + continue; + } + mozilla::dom::RTCDataChannelStats stats; + nsString id = u"dc"_ns; + id.AppendInt(chan->GetStream()); + stats.mId.Construct(id); + chan->GetLabel(temp); + stats.mTimestamp.Construct(aTimestamp); + stats.mType.Construct(mozilla::dom::RTCStatsType::Data_channel); + stats.mLabel.Construct(temp); + chan->GetProtocol(temp); + stats.mProtocol.Construct(temp); + stats.mDataChannelIdentifier.Construct(chan->GetStream()); + { + using State = mozilla::dom::RTCDataChannelState; + State state; + switch (chan->GetReadyState()) { + case CONNECTING: + state = State::Connecting; + break; + case OPEN: + state = State::Open; + break; + case CLOSING: + state = State::Closing; + break; + case CLOSED: + state = State::Closed; + break; + default: + MOZ_ASSERT(false, "Unknown DataChannel state"); + continue; + }; + stats.mState.Construct(state); + } + auto counters = chan->GetTrafficCounters(); + stats.mMessagesSent.Construct(counters.mMessagesSent); + stats.mBytesSent.Construct(counters.mBytesSent); + stats.mMessagesReceived.Construct(counters.mMessagesReceived); + stats.mBytesReceived.Construct(counters.mBytesReceived); + if (!aReport->mDataChannelStats.AppendElement(stats, fallible)) { + mozalloc_handle_oom(0); + } + } +} + +#ifdef MOZ_PEERCONNECTION + +bool DataChannelConnection::ConnectToTransport(const std::string& aTransportId, + const bool aClient, + const uint16_t aLocalPort, + const uint16_t aRemotePort) { + MutexAutoLock lock(mLock); + + MOZ_ASSERT(mMasterSocket, + "SCTP wasn't initialized before ConnectToTransport!"); + static const auto paramString = + [](const std::string& tId, const Maybe<bool>& client, + const uint16_t localPort, const uint16_t remotePort) -> std::string { + std::ostringstream stream; + stream << "Transport ID: '" << tId << "', Role: '" + << (client ? (client.value() ? "client" : "server") : "") + << "', Local Port: '" << localPort << "', Remote Port: '" + << remotePort << "'"; + return stream.str(); + }; + + const auto params = + paramString(aTransportId, Some(aClient), aLocalPort, aRemotePort); + DC_DEBUG(("ConnectToTransport connecting DTLS transport with parameters: %s", + params.c_str())); + + const auto currentReadyState = GetReadyState(); + if (currentReadyState == OPEN) { + if (aTransportId == mTransportId && mAllocateEven.isSome() && + mAllocateEven.value() == aClient && mLocalPort == aLocalPort && + mRemotePort == aRemotePort) { + DC_WARN( + ("Skipping attempt to connect to an already OPEN transport with " + "identical parameters.")); + return true; + } + DC_WARN( + ("Attempting to connect to an already OPEN transport, because " + "different parameters were provided.")); + DC_WARN(("Original transport parameters: %s", + paramString(mTransportId, mAllocateEven, mLocalPort, aRemotePort) + .c_str())); + DC_WARN(("New transport parameters: %s", params.c_str())); + } + if (NS_WARN_IF(aTransportId.empty())) { + return false; + } + + mLocalPort = aLocalPort; + mRemotePort = aRemotePort; + SetReadyState(CONNECTING); + mAllocateEven = Some(aClient); + + // Could be faster. Probably doesn't matter. + while (auto channel = mChannels.Get(INVALID_STREAM)) { + mChannels.Remove(channel); + channel->mStream = FindFreeStream(); + if (channel->mStream != INVALID_STREAM) { + mChannels.Insert(channel); + } + } + RUN_ON_THREAD(mSTS, + WrapRunnable(RefPtr<DataChannelConnection>(this), + &DataChannelConnection::SetSignals, aTransportId), + NS_DISPATCH_NORMAL); + return true; +} + +void DataChannelConnection::SetSignals(const std::string& aTransportId) { + ASSERT_WEBRTC(IsSTSThread()); + { + MutexAutoLock lock(mLock); + mTransportId = aTransportId; + } + mTransportHandler->SignalPacketReceived.connect( + this, &DataChannelConnection::SctpDtlsInput); + mTransportHandler->SignalStateChange.connect( + this, &DataChannelConnection::TransportStateChange); + // SignalStateChange() doesn't call you with the initial state + if (mTransportHandler->GetState(mTransportId, false) == + TransportLayer::TS_OPEN) { + DC_DEBUG(("Setting transport signals, dtls already open")); + CompleteConnect(); + } else { + DC_DEBUG(("Setting transport signals, dtls not open yet")); + } +} + +void DataChannelConnection::TransportStateChange( + const std::string& aTransportId, TransportLayer::State aState) { + ASSERT_WEBRTC(IsSTSThread()); + if (aTransportId == mTransportId) { + if (aState == TransportLayer::TS_OPEN) { + DC_DEBUG(("Transport is open!")); + CompleteConnect(); + } else if (aState == TransportLayer::TS_CLOSED || + aState == TransportLayer::TS_NONE || + aState == TransportLayer::TS_ERROR) { + DC_DEBUG(("Transport is closed!")); + Stop(); + } + } +} + +void DataChannelConnection::CompleteConnect() { + MutexAutoLock lock(mLock); + + DC_DEBUG(("dtls open")); + ASSERT_WEBRTC(IsSTSThread()); + if (!mMasterSocket) { + return; + } + + struct sockaddr_conn addr; + memset(&addr, 0, sizeof(addr)); + addr.sconn_family = AF_CONN; +# if defined(__Userspace_os_Darwin) + addr.sconn_len = sizeof(addr); +# endif + addr.sconn_port = htons(mLocalPort); + addr.sconn_addr = reinterpret_cast<void*>(mId); + + DC_DEBUG(("Calling usrsctp_bind")); + int r = usrsctp_bind(mMasterSocket, reinterpret_cast<struct sockaddr*>(&addr), + sizeof(addr)); + if (r < 0) { + DC_ERROR(("usrsctp_bind failed: %d", r)); + } else { + // This is the remote addr + addr.sconn_port = htons(mRemotePort); + DC_DEBUG(("Calling usrsctp_connect")); + r = usrsctp_connect( + mMasterSocket, reinterpret_cast<struct sockaddr*>(&addr), sizeof(addr)); + if (r >= 0 || errno == EINPROGRESS) { + struct sctp_paddrparams paddrparams; + socklen_t opt_len; + + memset(&paddrparams, 0, sizeof(struct sctp_paddrparams)); + memcpy(&paddrparams.spp_address, &addr, sizeof(struct sockaddr_conn)); + opt_len = (socklen_t)sizeof(struct sctp_paddrparams); + r = usrsctp_getsockopt(mMasterSocket, IPPROTO_SCTP, SCTP_PEER_ADDR_PARAMS, + &paddrparams, &opt_len); + if (r < 0) { + DC_ERROR(("usrsctp_getsockopt failed: %d", r)); + } else { + // This field is misnamed. |spp_pathmtu| represents the maximum + // _payload_ size in libusrsctp. So: + // 1280 (a reasonable IPV6 MTU according to RFC 8831) + // -12 (sctp header) + // -24 (GCM sipher) + // -13 (DTLS record header) + // -8 (UDP header) + // -4 (TURN ChannelData) + // -40 (IPV6 header) + // = 1179 + // We could further restrict this, because RFC 8831 suggests a starting + // IPV4 path MTU of 1200, which would lead to a value of 1115. + // I suspect that in practice the path MTU for IPV4 is substantially + // larger than 1200. + paddrparams.spp_pathmtu = 1179; + paddrparams.spp_flags &= ~SPP_PMTUD_ENABLE; + paddrparams.spp_flags |= SPP_PMTUD_DISABLE; + opt_len = (socklen_t)sizeof(struct sctp_paddrparams); + r = usrsctp_setsockopt(mMasterSocket, IPPROTO_SCTP, + SCTP_PEER_ADDR_PARAMS, &paddrparams, opt_len); + if (r < 0) { + DC_ERROR(("usrsctp_getsockopt failed: %d", r)); + } else { + DC_ERROR(("usrsctp: PMTUD disabled, MTU set to %u", + paddrparams.spp_pathmtu)); + } + } + } + if (r < 0) { + if (errno == EINPROGRESS) { + // non-blocking + return; + } + DC_ERROR(("usrsctp_connect failed: %d", errno)); + SetReadyState(CLOSED); + } else { + // We fire ON_CONNECTION via SCTP_COMM_UP when we get that + return; + } + } + // Note: currently this doesn't actually notify the application + Dispatch(do_AddRef(new DataChannelOnMessageAvailable( + DataChannelOnMessageAvailable::ON_CONNECTION, this))); +} + +// Process any pending Opens +void DataChannelConnection::ProcessQueuedOpens() { + // The nsDeque holds channels with an AddRef applied. Another reference + // (may) be held by the DOMDataChannel, unless it's been GC'd. No other + // references should exist. + + // Can't copy nsDeque's. Move into temp array since any that fail will + // go back to mPending + nsRefPtrDeque<DataChannel> temp; + RefPtr<DataChannel> temp_channel; + while (nullptr != (temp_channel = mPending.PopFront())) { + temp.Push(temp_channel.forget()); + } + + RefPtr<DataChannel> channel; + + while (nullptr != (channel = temp.PopFront())) { + if (channel->mFlags & DATA_CHANNEL_FLAGS_FINISH_OPEN) { + DC_DEBUG(("Processing queued open for %p (%u)", channel.get(), + channel->mStream)); + channel->mFlags &= ~DATA_CHANNEL_FLAGS_FINISH_OPEN; + // OpenFinish returns a reference itself, so we need to take it can + // Release it + channel = OpenFinish(channel.forget()); // may reset the flag and re-push + } else { + NS_ASSERTION( + false, + "How did a DataChannel get queued without the FINISH_OPEN flag?"); + } + } +} + +void DataChannelConnection::SctpDtlsInput(const std::string& aTransportId, + const MediaPacket& packet) { + MutexAutoLock lock(mLock); + if ((packet.type() != MediaPacket::SCTP) || (mTransportId != aTransportId)) { + return; + } + + if (MOZ_LOG_TEST(gSCTPLog, LogLevel::Debug)) { + char* buf; + + if ((buf = usrsctp_dumppacket((void*)packet.data(), packet.len(), + SCTP_DUMP_INBOUND)) != nullptr) { + SCTP_LOG(("%s", buf)); + usrsctp_freedumpbuffer(buf); + } + } + // Pass the data to SCTP + usrsctp_conninput(reinterpret_cast<void*>(mId), packet.data(), packet.len(), + 0); +} + +void DataChannelConnection::SendPacket(std::unique_ptr<MediaPacket>&& packet) { + mSTS->Dispatch(NS_NewRunnableFunction( + "DataChannelConnection::SendPacket", + [this, self = RefPtr<DataChannelConnection>(this), + packet = std::move(packet)]() mutable { + // DC_DEBUG(("%p: SCTP/DTLS sent %ld bytes", this, len)); + if (!mTransportId.empty() && mTransportHandler) { + mTransportHandler->SendPacket(mTransportId, std::move(*packet)); + } + })); +} + +int DataChannelConnection::SctpDtlsOutput(void* addr, void* buffer, + size_t length, uint8_t tos, + uint8_t set_df) { + if (MOZ_LOG_TEST(gSCTPLog, LogLevel::Debug)) { + char* buf; + + if ((buf = usrsctp_dumppacket(buffer, length, SCTP_DUMP_OUTBOUND)) != + nullptr) { + SCTP_LOG(("%s", buf)); + usrsctp_freedumpbuffer(buf); + } + } + + // We're async proxying even if on the STSThread because this is called + // with internal SCTP locks held in some cases (such as in usrsctp_connect()). + // SCTP has an option for Apple, on IP connections only, to release at least + // one of the locks before calling a packet output routine; with changes to + // the underlying SCTP stack this might remove the need to use an async proxy. + std::unique_ptr<MediaPacket> packet(new MediaPacket); + packet->SetType(MediaPacket::SCTP); + packet->Copy(static_cast<const uint8_t*>(buffer), length); + + if (NS_IsMainThread() && mDeferSend) { + mDeferredSend.emplace_back(std::move(packet)); + return 0; + } + + SendPacket(std::move(packet)); + return 0; // cheat! Packets can always be dropped later anyways +} +#endif + +#ifdef ALLOW_DIRECT_SCTP_LISTEN_CONNECT +// listen for incoming associations +// Blocks! - Don't call this from main thread! + +bool DataChannelConnection::Listen(unsigned short port) { + struct sockaddr_in addr; + socklen_t addr_len; + + NS_WARNING_ASSERTION(!NS_IsMainThread(), + "Blocks, do not call from main thread!!!"); + + /* Acting as the 'server' */ + memset((void*)&addr, 0, sizeof(addr)); +# ifdef HAVE_SIN_LEN + addr.sin_len = sizeof(struct sockaddr_in); +# endif + addr.sin_family = AF_INET; + addr.sin_port = htons(port); + addr.sin_addr.s_addr = htonl(INADDR_ANY); + DC_DEBUG(("Waiting for connections on port %u", ntohs(addr.sin_port))); + { + MutexAutoLock lock(mLock); + SetReadyState(CONNECTING); + } + if (usrsctp_bind(mMasterSocket, reinterpret_cast<struct sockaddr*>(&addr), + sizeof(struct sockaddr_in)) < 0) { + DC_ERROR(("***Failed userspace_bind")); + return false; + } + if (usrsctp_listen(mMasterSocket, 1) < 0) { + DC_ERROR(("***Failed userspace_listen")); + return false; + } + + DC_DEBUG(("Accepting connection")); + addr_len = 0; + if ((mSocket = usrsctp_accept(mMasterSocket, nullptr, &addr_len)) == + nullptr) { + DC_ERROR(("***Failed accept")); + return false; + } + + { + MutexAutoLock lock(mLock); + SetReadyState(OPEN); + } + + struct linger l; + l.l_onoff = 1; + l.l_linger = 0; + if (usrsctp_setsockopt(mSocket, SOL_SOCKET, SO_LINGER, (const void*)&l, + (socklen_t)sizeof(struct linger)) < 0) { + DC_WARN(("Couldn't set SO_LINGER on SCTP socket")); + } + + // Notify Connection open + // XXX We need to make sure connection sticks around until the message is + // delivered + DC_DEBUG(("%s: sending ON_CONNECTION for %p", __FUNCTION__, this)); + Dispatch(do_AddRef(new DataChannelOnMessageAvailable( + DataChannelOnMessageAvailable::ON_CONNECTION, this, + (DataChannel*)nullptr))); + return true; +} + +// Blocks! - Don't call this from main thread! +bool DataChannelConnection::Connect(const char* addr, unsigned short port) { + struct sockaddr_in addr4; + struct sockaddr_in6 addr6; + + NS_WARNING_ASSERTION(!NS_IsMainThread(), + "Blocks, do not call from main thread!!!"); + + /* Acting as the connector */ + DC_DEBUG(("Connecting to %s, port %u", addr, port)); + memset((void*)&addr4, 0, sizeof(struct sockaddr_in)); + memset((void*)&addr6, 0, sizeof(struct sockaddr_in6)); +# ifdef HAVE_SIN_LEN + addr4.sin_len = sizeof(struct sockaddr_in); +# endif +# ifdef HAVE_SIN6_LEN + addr6.sin6_len = sizeof(struct sockaddr_in6); +# endif + addr4.sin_family = AF_INET; + addr6.sin6_family = AF_INET6; + addr4.sin_port = htons(port); + addr6.sin6_port = htons(port); + { + MutexAutoLock lock(mLock); + SetReadyState(CONNECTING); + } +# if !defined(__Userspace_os_Windows) + if (inet_pton(AF_INET6, addr, &addr6.sin6_addr) == 1) { + if (usrsctp_connect(mMasterSocket, + reinterpret_cast<struct sockaddr*>(&addr6), + sizeof(struct sockaddr_in6)) < 0) { + DC_ERROR(("*** Failed userspace_connect")); + return false; + } + } else if (inet_pton(AF_INET, addr, &addr4.sin_addr) == 1) { + if (usrsctp_connect(mMasterSocket, + reinterpret_cast<struct sockaddr*>(&addr4), + sizeof(struct sockaddr_in)) < 0) { + DC_ERROR(("*** Failed userspace_connect")); + return false; + } + } else { + DC_ERROR(("*** Illegal destination address.")); + } +# else + { + struct sockaddr_storage ss; + int sslen = sizeof(ss); + + if (!WSAStringToAddressA(const_cast<char*>(addr), AF_INET6, nullptr, + (struct sockaddr*)&ss, &sslen)) { + addr6.sin6_addr = + (reinterpret_cast<struct sockaddr_in6*>(&ss))->sin6_addr; + if (usrsctp_connect(mMasterSocket, + reinterpret_cast<struct sockaddr*>(&addr6), + sizeof(struct sockaddr_in6)) < 0) { + DC_ERROR(("*** Failed userspace_connect")); + return false; + } + } else if (!WSAStringToAddressA(const_cast<char*>(addr), AF_INET, nullptr, + (struct sockaddr*)&ss, &sslen)) { + addr4.sin_addr = (reinterpret_cast<struct sockaddr_in*>(&ss))->sin_addr; + if (usrsctp_connect(mMasterSocket, + reinterpret_cast<struct sockaddr*>(&addr4), + sizeof(struct sockaddr_in)) < 0) { + DC_ERROR(("*** Failed userspace_connect")); + return false; + } + } else { + DC_ERROR(("*** Illegal destination address.")); + } + } +# endif + + mSocket = mMasterSocket; + + DC_DEBUG(("connect() succeeded! Entering connected mode")); + { + MutexAutoLock lock(mLock); + SetReadyState(OPEN); + } + // Notify Connection open + // XXX We need to make sure connection sticks around until the message is + // delivered + DC_DEBUG(("%s: sending ON_CONNECTION for %p", __FUNCTION__, this)); + Dispatch(do_AddRef(new DataChannelOnMessageAvailable( + DataChannelOnMessageAvailable::ON_CONNECTION, this, + (DataChannel*)nullptr))); + return true; +} +#endif + +DataChannel* DataChannelConnection::FindChannelByStream(uint16_t stream) { + return mChannels.Get(stream).get(); +} + +uint16_t DataChannelConnection::FindFreeStream() { + ASSERT_WEBRTC(NS_IsMainThread()); + uint16_t i, limit; + + limit = MAX_NUM_STREAMS; + + MOZ_ASSERT(mAllocateEven.isSome()); + for (i = (*mAllocateEven ? 0 : 1); i < limit; i += 2) { + if (mChannels.Get(i)) { + continue; + } + + // Verify it's not still in the process of closing + size_t j; + for (j = 0; j < mStreamsResetting.Length(); ++j) { + if (mStreamsResetting[j] == i) { + break; + } + } + + if (j == mStreamsResetting.Length()) { + return i; + } + } + return INVALID_STREAM; +} + +uint32_t DataChannelConnection::UpdateCurrentStreamIndex() { + RefPtr<DataChannel> channel = mChannels.GetNextChannel(mCurrentStream); + if (!channel) { + mCurrentStream = 0; + } else { + mCurrentStream = channel->mStream; + } + return mCurrentStream; +} + +uint32_t DataChannelConnection::GetCurrentStreamIndex() { + if (!mChannels.Get(mCurrentStream)) { + // The stream muse have been removed, reset + DC_DEBUG(("Reset mCurrentChannel")); + mCurrentStream = 0; + } + return mCurrentStream; +} + +bool DataChannelConnection::RequestMoreStreams(int32_t aNeeded) { + struct sctp_status status; + struct sctp_add_streams sas; + uint32_t outStreamsNeeded; + socklen_t len; + + if (aNeeded + mNegotiatedIdLimit > MAX_NUM_STREAMS) { + aNeeded = MAX_NUM_STREAMS - mNegotiatedIdLimit; + } + if (aNeeded <= 0) { + return false; + } + + len = (socklen_t)sizeof(struct sctp_status); + if (usrsctp_getsockopt(mMasterSocket, IPPROTO_SCTP, SCTP_STATUS, &status, + &len) < 0) { + DC_ERROR(("***failed: getsockopt SCTP_STATUS")); + return false; + } + outStreamsNeeded = aNeeded; // number to add + + // Note: if multiple channel opens happen when we don't have enough space, + // we'll call RequestMoreStreams() multiple times + memset(&sas, 0, sizeof(sas)); + sas.sas_instrms = 0; + sas.sas_outstrms = (uint16_t)outStreamsNeeded; /* XXX error handling */ + // Doesn't block, we get an event when it succeeds or fails + if (usrsctp_setsockopt(mMasterSocket, IPPROTO_SCTP, SCTP_ADD_STREAMS, &sas, + (socklen_t)sizeof(struct sctp_add_streams)) < 0) { + if (errno == EALREADY) { + DC_DEBUG(("Already have %u output streams", outStreamsNeeded)); + return true; + } + + DC_ERROR(("***failed: setsockopt ADD errno=%d", errno)); + return false; + } + DC_DEBUG(("Requested %u more streams", outStreamsNeeded)); + // We add to mNegotiatedIdLimit when we get a SCTP_STREAM_CHANGE_EVENT and the + // values are larger than mNegotiatedIdLimit + return true; +} + +// Returns a POSIX error code. +int DataChannelConnection::SendControlMessage(const uint8_t* data, uint32_t len, + uint16_t stream) { + struct sctp_sendv_spa info = {0}; + + // General flags + info.sendv_flags = SCTP_SEND_SNDINFO_VALID; + + // Set stream identifier, protocol identifier and flags + info.sendv_sndinfo.snd_sid = stream; + info.sendv_sndinfo.snd_flags = SCTP_EOR; + info.sendv_sndinfo.snd_ppid = htonl(DATA_CHANNEL_PPID_CONTROL); + + // Create message instance and send + // Note: Main-thread IO, but doesn't block +#if (UINT32_MAX > SIZE_MAX) + if (len > SIZE_MAX) { + return EMSGSIZE; + } +#endif + OutgoingMsg msg(info, data, (size_t)len); + bool buffered; + int error = SendMsgInternalOrBuffer(mBufferedControl, msg, buffered, nullptr); + + // Set pending type (if buffered) + if (!error && buffered && !mPendingType) { + mPendingType = PENDING_DCEP; + } + return error; +} + +// Returns a POSIX error code. +int DataChannelConnection::SendOpenAckMessage(uint16_t stream) { + struct rtcweb_datachannel_ack ack; + + memset(&ack, 0, sizeof(struct rtcweb_datachannel_ack)); + ack.msg_type = DATA_CHANNEL_ACK; + + return SendControlMessage((const uint8_t*)&ack, sizeof(ack), stream); +} + +// Returns a POSIX error code. +int DataChannelConnection::SendOpenRequestMessage( + const nsACString& label, const nsACString& protocol, uint16_t stream, + bool unordered, uint16_t prPolicy, uint32_t prValue) { + const int label_len = label.Length(); // not including nul + const int proto_len = protocol.Length(); // not including nul + // careful - request struct include one char for the label + const int req_size = sizeof(struct rtcweb_datachannel_open_request) - 1 + + label_len + proto_len; + UniqueFreePtr<struct rtcweb_datachannel_open_request> req( + (struct rtcweb_datachannel_open_request*)moz_xmalloc(req_size)); + + memset(req.get(), 0, req_size); + req->msg_type = DATA_CHANNEL_OPEN_REQUEST; + switch (prPolicy) { + case SCTP_PR_SCTP_NONE: + req->channel_type = DATA_CHANNEL_RELIABLE; + break; + case SCTP_PR_SCTP_TTL: + req->channel_type = DATA_CHANNEL_PARTIAL_RELIABLE_TIMED; + break; + case SCTP_PR_SCTP_RTX: + req->channel_type = DATA_CHANNEL_PARTIAL_RELIABLE_REXMIT; + break; + default: + return EINVAL; + } + if (unordered) { + // Per the current types, all differ by 0x80 between ordered and unordered + req->channel_type |= + 0x80; // NOTE: be careful if new types are added in the future + } + + req->reliability_param = htonl(prValue); + req->priority = htons(0); /* XXX: add support */ + req->label_length = htons(label_len); + req->protocol_length = htons(proto_len); + memcpy(&req->label[0], PromiseFlatCString(label).get(), label_len); + memcpy(&req->label[label_len], PromiseFlatCString(protocol).get(), proto_len); + + // TODO: req_size is an int... that looks hairy + int error = SendControlMessage((const uint8_t*)req.get(), req_size, stream); + return error; +} + +// XXX This should use a separate thread (outbound queue) which should +// select() to know when to *try* to send data to the socket again. +// Alternatively, it can use a timeout, but that's guaranteed to be wrong +// (just not sure in what direction). We could re-implement NSPR's +// PR_POLL_WRITE/etc handling... with a lot of work. + +// Better yet, use the SCTP stack's notifications on buffer state to avoid +// filling the SCTP's buffers. + +// returns if we're still blocked (true) +bool DataChannelConnection::SendDeferredMessages() { + RefPtr<DataChannel> channel; // we may null out the refs to this + + // This may block while something is modifying channels, but should not block + // for IO + ASSERT_WEBRTC(!NS_IsMainThread()); + mLock.AssertCurrentThreadOwns(); + + DC_DEBUG(("SendDeferredMessages called, pending type: %d", mPendingType)); + if (!mPendingType) { + return false; + } + + // Send pending control messages + // Note: If ndata is not active, check if DCEP messages are currently + // outstanding. These need to + // be sent first before other streams can be used for sending. + if (!mBufferedControl.IsEmpty() && + (mSendInterleaved || mPendingType == PENDING_DCEP)) { + if (SendBufferedMessages(mBufferedControl, nullptr)) { + return true; + } + + // Note: There may or may not be pending data messages + mPendingType = PENDING_DATA; + } + + bool blocked = false; + uint32_t i = GetCurrentStreamIndex(); + uint32_t end = i; + do { + channel = mChannels.Get(i); + // Should already be cleared if closing/closed + if (!channel || channel->mBufferedData.IsEmpty()) { + i = UpdateCurrentStreamIndex(); + continue; + } + + // Send buffered data messages + // Warning: This will fail in case ndata is inactive and a previously + // deallocated data channel has not been closed properly. If you + // ever see that no messages can be sent on any channel, this is + // likely the cause (an explicit EOR message partially sent whose + // remaining chunks are still being waited for). + size_t written = 0; + mDeferSend = true; + blocked = SendBufferedMessages(channel->mBufferedData, &written); + mDeferSend = false; + if (written) { + channel->DecrementBufferedAmount(written); + } + + for (auto&& packet : mDeferredSend) { + MOZ_ASSERT(written); + SendPacket(std::move(packet)); + } + mDeferredSend.clear(); + + // Update current stream index + // Note: If ndata is not active, the outstanding data messages on this + // stream need to be sent first before other streams can be used for + // sending. + if (mSendInterleaved || !blocked) { + i = UpdateCurrentStreamIndex(); + } + } while (!blocked && i != end); + + if (!blocked) { + mPendingType = mBufferedControl.IsEmpty() ? PENDING_NONE : PENDING_DCEP; + } + return blocked; +} + +// Called with mLock locked! +// buffer MUST have at least one item! +// returns if we're still blocked (true) +bool DataChannelConnection::SendBufferedMessages( + nsTArray<UniquePtr<BufferedOutgoingMsg>>& buffer, size_t* aWritten) { + do { + // Re-send message + int error = SendMsgInternal(*buffer[0], aWritten); + switch (error) { + case 0: + buffer.RemoveElementAt(0); + break; + case EAGAIN: +#if (EAGAIN != EWOULDBLOCK) + case EWOULDBLOCK: +#endif + return true; + default: + buffer.RemoveElementAt(0); + DC_ERROR(("error on sending: %d", error)); + break; + } + } while (!buffer.IsEmpty()); + + return false; +} + +// Caller must ensure that length <= SIZE_MAX +void DataChannelConnection::HandleOpenRequestMessage( + const struct rtcweb_datachannel_open_request* req, uint32_t length, + uint16_t stream) { + RefPtr<DataChannel> channel; + uint32_t prValue; + uint16_t prPolicy; + + ASSERT_WEBRTC(!NS_IsMainThread()); + mLock.AssertCurrentThreadOwns(); + + const size_t requiredLength = (sizeof(*req) - 1) + ntohs(req->label_length) + + ntohs(req->protocol_length); + if (((size_t)length) != requiredLength) { + if (((size_t)length) < requiredLength) { + DC_ERROR( + ("%s: insufficient length: %u, should be %zu. Unable to continue.", + __FUNCTION__, length, requiredLength)); + return; + } + DC_WARN(("%s: Inconsistent length: %u, should be %zu", __FUNCTION__, length, + requiredLength)); + } + + DC_DEBUG(("%s: length %u, sizeof(*req) = %zu", __FUNCTION__, length, + sizeof(*req))); + + switch (req->channel_type) { + case DATA_CHANNEL_RELIABLE: + case DATA_CHANNEL_RELIABLE_UNORDERED: + prPolicy = SCTP_PR_SCTP_NONE; + break; + case DATA_CHANNEL_PARTIAL_RELIABLE_REXMIT: + case DATA_CHANNEL_PARTIAL_RELIABLE_REXMIT_UNORDERED: + prPolicy = SCTP_PR_SCTP_RTX; + break; + case DATA_CHANNEL_PARTIAL_RELIABLE_TIMED: + case DATA_CHANNEL_PARTIAL_RELIABLE_TIMED_UNORDERED: + prPolicy = SCTP_PR_SCTP_TTL; + break; + default: + DC_ERROR(("Unknown channel type %d", req->channel_type)); + /* XXX error handling */ + return; + } + prValue = ntohl(req->reliability_param); + bool ordered = !(req->channel_type & 0x80); + + if ((channel = FindChannelByStream(stream))) { + if (!channel->mNegotiated) { + DC_ERROR( + ("HandleOpenRequestMessage: channel for pre-existing stream " + "%u that was not externally negotiated. JS is lying to us, or " + "there's an id collision.", + stream)); + /* XXX: some error handling */ + } else { + DC_DEBUG(("Open for externally negotiated channel %u", stream)); + // XXX should also check protocol, maybe label + if (prPolicy != channel->mPrPolicy || prValue != channel->mPrValue || + ordered != channel->mOrdered) { + DC_WARN( + ("external negotiation mismatch with OpenRequest:" + "channel %u, policy %u/%u, value %u/%u, ordered %d/%d", + stream, prPolicy, channel->mPrPolicy, prValue, channel->mPrValue, + static_cast<int>(ordered), static_cast<int>(channel->mOrdered))); + } + } + return; + } + if (stream >= mNegotiatedIdLimit) { + DC_ERROR(("%s: stream %u out of bounds (%zu)", __FUNCTION__, stream, + mNegotiatedIdLimit)); + return; + } + + nsCString label( + nsDependentCSubstring(&req->label[0], ntohs(req->label_length))); + nsCString protocol(nsDependentCSubstring( + &req->label[ntohs(req->label_length)], ntohs(req->protocol_length))); + + channel = + new DataChannel(this, stream, DataChannel::OPEN, label, protocol, + prPolicy, prValue, ordered, false, nullptr, nullptr); + mChannels.Insert(channel); + + DC_DEBUG(("%s: sending ON_CHANNEL_CREATED for %s/%s: %u", __FUNCTION__, + channel->mLabel.get(), channel->mProtocol.get(), stream)); + Dispatch(do_AddRef(new DataChannelOnMessageAvailable( + DataChannelOnMessageAvailable::ON_CHANNEL_CREATED, this, channel))); + + DC_DEBUG(("%s: deferring sending ON_CHANNEL_OPEN for %p", __FUNCTION__, + channel.get())); + channel->AnnounceOpen(); + + // Note that any message can be buffered; SendOpenAckMessage may + // error later than this check. + const auto error = SendOpenAckMessage(channel->mStream); + if (error) { + DC_ERROR(("SendOpenRequest failed, error = %d", error)); + Dispatch(NS_NewRunnableFunction( + "DataChannelConnection::HandleOpenRequestMessage", + [channel, connection = RefPtr<DataChannelConnection>(this)]() { + // Close the channel on failure + connection->Close(channel); + })); + return; + } + DeliverQueuedData(channel->mStream); +} + +// NOTE: the updated spec from the IETF says we should set in-order until we +// receive an ACK. That would make this code moot. Keep it for now for +// backwards compatibility. +void DataChannelConnection::DeliverQueuedData(uint16_t stream) { + mLock.AssertCurrentThreadOwns(); + + mQueuedData.RemoveElementsBy([stream, this](const auto& dataItem) { + mLock.AssertCurrentThreadOwns(); + const bool match = dataItem->mStream == stream; + if (match) { + DC_DEBUG(("Delivering queued data for stream %u, length %u", stream, + dataItem->mLength)); + // Deliver the queued data + HandleDataMessage(dataItem->mData, dataItem->mLength, dataItem->mPpid, + dataItem->mStream, dataItem->mFlags); + } + return match; + }); +} + +// Caller must ensure that length <= SIZE_MAX +void DataChannelConnection::HandleOpenAckMessage( + const struct rtcweb_datachannel_ack* ack, uint32_t length, + uint16_t stream) { + DataChannel* channel; + + mLock.AssertCurrentThreadOwns(); + + channel = FindChannelByStream(stream); + if (NS_WARN_IF(!channel)) { + return; + } + + DC_DEBUG(("OpenAck received for stream %u, waiting=%d", stream, + (channel->mFlags & DATA_CHANNEL_FLAGS_WAITING_ACK) ? 1 : 0)); + + channel->mFlags &= ~DATA_CHANNEL_FLAGS_WAITING_ACK; +} + +// Caller must ensure that length <= SIZE_MAX +void DataChannelConnection::HandleUnknownMessage(uint32_t ppid, uint32_t length, + uint16_t stream) { + /* XXX: Send an error message? */ + DC_ERROR(("unknown DataChannel message received: %u, len %u on stream %d", + ppid, length, stream)); + // XXX Log to JS error console if possible +} + +uint8_t DataChannelConnection::BufferMessage(nsACString& recvBuffer, + const void* data, uint32_t length, + uint32_t ppid, int flags) { + const char* buffer = (const char*)data; + uint8_t bufferFlags = 0; + + if ((flags & MSG_EOR) && ppid != DATA_CHANNEL_PPID_BINARY_PARTIAL && + ppid != DATA_CHANNEL_PPID_DOMSTRING_PARTIAL) { + bufferFlags |= DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_COMPLETE; + + // Return directly if nothing has been buffered + if (recvBuffer.IsEmpty()) { + return bufferFlags; + } + } + + // Ensure it doesn't blow up our buffer + // TODO: Change 'WEBRTC_DATACHANNEL_MAX_MESSAGE_SIZE_LOCAL' to whatever the + // new buffer is capable of holding. + if (((uint64_t)recvBuffer.Length()) + ((uint64_t)length) > + WEBRTC_DATACHANNEL_MAX_MESSAGE_SIZE_LOCAL) { + bufferFlags |= DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_TOO_LARGE; + return bufferFlags; + } + + // Copy & add to receive buffer + recvBuffer.Append(buffer, length); + bufferFlags |= DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_BUFFERED; + return bufferFlags; +} + +void DataChannelConnection::HandleDataMessage(const void* data, size_t length, + uint32_t ppid, uint16_t stream, + int flags) { + DataChannel* channel; + const char* buffer = (const char*)data; + + mLock.AssertCurrentThreadOwns(); + channel = FindChannelByStream(stream); + + // Note: Until we support SIZE_MAX sized messages, we need this check +#if (SIZE_MAX > UINT32_MAX) + if (length > UINT32_MAX) { + DC_ERROR(("DataChannel: Cannot handle message of size %zu (max=%" PRIu32 + ")", + length, UINT32_MAX)); + CloseLocked(channel); + return; + } +#endif + uint32_t data_length = (uint32_t)length; + + // XXX A closed channel may trip this... check + // NOTE: the updated spec from the IETF says we should set in-order until we + // receive an ACK. That would make this code moot. Keep it for now for + // backwards compatibility. + if (!channel) { + // In the updated 0-RTT open case, the sender can send data immediately + // after Open, and doesn't set the in-order bit (since we don't have a + // response or ack). Also, with external negotiation, data can come in + // before we're told about the external negotiation. We need to buffer + // data until either a) Open comes in, if the ordering get messed up, + // or b) the app tells us this channel was externally negotiated. When + // these occur, we deliver the data. + + // Since this is rare and non-performance, keep a single list of queued + // data messages to deliver once the channel opens. + DC_DEBUG(("Queuing data for stream %u, length %u", stream, data_length)); + // Copies data + mQueuedData.AppendElement( + new QueuedDataMessage(stream, ppid, flags, data, data_length)); + return; + } + + bool is_binary = true; + uint8_t bufferFlags; + int32_t type; + const char* info = ""; + + if (ppid == DATA_CHANNEL_PPID_DOMSTRING_PARTIAL || + ppid == DATA_CHANNEL_PPID_DOMSTRING || + ppid == DATA_CHANNEL_PPID_DOMSTRING_EMPTY) { + is_binary = false; + } + if (is_binary != channel->mIsRecvBinary && !channel->mRecvBuffer.IsEmpty()) { + NS_WARNING("DataChannel message aborted by fragment type change!"); + // TODO: Maybe closing would be better as this is a hard to detect protocol + // violation? + channel->mRecvBuffer.Truncate(0); + } + channel->mIsRecvBinary = is_binary; + + // Remaining chunks of previously truncated message (due to the buffer being + // full)? + if (channel->mFlags & DATA_CHANNEL_FLAGS_CLOSING_TOO_LARGE) { + DC_ERROR( + ("DataChannel: Ignoring partial message of length %u, buffer full and " + "closing", + data_length)); + // Only unblock if unordered + if (!channel->mOrdered && (flags & MSG_EOR)) { + channel->mFlags &= ~DATA_CHANNEL_FLAGS_CLOSING_TOO_LARGE; + } + } + + // Buffer message until complete + bufferFlags = + BufferMessage(channel->mRecvBuffer, buffer, data_length, ppid, flags); + if (bufferFlags & DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_TOO_LARGE) { + DC_ERROR( + ("DataChannel: Buffered message would become too large to handle, " + "closing channel")); + channel->mRecvBuffer.Truncate(0); + channel->mFlags |= DATA_CHANNEL_FLAGS_CLOSING_TOO_LARGE; + CloseLocked(channel); + return; + } + if (!(bufferFlags & DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_COMPLETE)) { + DC_DEBUG( + ("DataChannel: Partial %s message of length %u (total %zu) on channel " + "id %u", + is_binary ? "binary" : "string", data_length, + channel->mRecvBuffer.Length(), channel->mStream)); + return; // Not ready to notify application + } + if (bufferFlags & DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_BUFFERED) { + data_length = channel->mRecvBuffer.Length(); + } + + // Complain about large messages (only complain - we can handle it) + if (data_length > WEBRTC_DATACHANNEL_MAX_MESSAGE_SIZE_LOCAL) { + DC_WARN( + ("DataChannel: Received message of length %u is > announced maximum " + "message size (%u)", + data_length, WEBRTC_DATACHANNEL_MAX_MESSAGE_SIZE_LOCAL)); + } + + bool is_empty = false; + + switch (ppid) { + case DATA_CHANNEL_PPID_DOMSTRING: + DC_DEBUG( + ("DataChannel: Received string message of length %u on channel %u", + data_length, channel->mStream)); + type = DataChannelOnMessageAvailable::ON_DATA_STRING; + if (bufferFlags & DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_BUFFERED) { + info = " (string fragmented)"; + } + // else send using recvData normally + + // WebSockets checks IsUTF8() here; we can try to deliver it + break; + + case DATA_CHANNEL_PPID_DOMSTRING_EMPTY: + DC_DEBUG( + ("DataChannel: Received empty string message of length %u on channel " + "%u", + data_length, channel->mStream)); + type = DataChannelOnMessageAvailable::ON_DATA_STRING; + if (bufferFlags & DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_BUFFERED) { + info = " (string fragmented)"; + } + is_empty = true; + break; + + case DATA_CHANNEL_PPID_BINARY: + DC_DEBUG( + ("DataChannel: Received binary message of length %u on channel id %u", + data_length, channel->mStream)); + type = DataChannelOnMessageAvailable::ON_DATA_BINARY; + if (bufferFlags & DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_BUFFERED) { + info = " (binary fragmented)"; + } + + // else send using recvData normally + break; + + case DATA_CHANNEL_PPID_BINARY_EMPTY: + DC_DEBUG( + ("DataChannel: Received empty binary message of length %u on channel " + "id %u", + data_length, channel->mStream)); + type = DataChannelOnMessageAvailable::ON_DATA_BINARY; + if (bufferFlags & DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_BUFFERED) { + info = " (binary fragmented)"; + } + is_empty = true; + break; + + default: + NS_ERROR("Unknown data PPID"); + DC_ERROR(("Unknown data PPID %" PRIu32, ppid)); + return; + } + + channel->WithTrafficCounters( + [&data_length](DataChannel::TrafficCounters& counters) { + counters.mMessagesReceived++; + counters.mBytesReceived += data_length; + }); + + // Notify onmessage + DC_DEBUG(("%s: sending ON_DATA_%s%s for %p", __FUNCTION__, + (type == DataChannelOnMessageAvailable::ON_DATA_STRING) ? "STRING" + : "BINARY", + info, channel)); + if (bufferFlags & DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_BUFFERED) { + channel->SendOrQueue(new DataChannelOnMessageAvailable( + type, this, channel, channel->mRecvBuffer)); + channel->mRecvBuffer.Truncate(0); + } else { + nsAutoCString recvData(is_empty ? "" : buffer, + is_empty ? 0 : data_length); // allocates >64 + channel->SendOrQueue( + new DataChannelOnMessageAvailable(type, this, channel, recvData)); + } +} + +void DataChannelConnection::HandleDCEPMessage(const void* buffer, size_t length, + uint32_t ppid, uint16_t stream, + int flags) { + const struct rtcweb_datachannel_open_request* req; + const struct rtcweb_datachannel_ack* ack; + + // Note: Until we support SIZE_MAX sized messages, we need this check +#if (SIZE_MAX > UINT32_MAX) + if (length > UINT32_MAX) { + DC_ERROR(("DataChannel: Cannot handle message of size %zu (max=%u)", length, + UINT32_MAX)); + Stop(); + return; + } +#endif + uint32_t data_length = (uint32_t)length; + + mLock.AssertCurrentThreadOwns(); + + // Buffer message until complete + const uint8_t bufferFlags = + BufferMessage(mRecvBuffer, buffer, data_length, ppid, flags); + if (bufferFlags & DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_TOO_LARGE) { + DC_ERROR( + ("DataChannel: Buffered message would become too large to handle, " + "closing connection")); + mRecvBuffer.Truncate(0); + Stop(); + return; + } + if (!(bufferFlags & DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_COMPLETE)) { + DC_DEBUG(("Buffered partial DCEP message of length %u", data_length)); + return; + } + if (bufferFlags & DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_BUFFERED) { + buffer = reinterpret_cast<const void*>(mRecvBuffer.BeginReading()); + data_length = mRecvBuffer.Length(); + } + + req = static_cast<const struct rtcweb_datachannel_open_request*>(buffer); + DC_DEBUG(("Handling DCEP message of length %u", data_length)); + + // Ensure minimum message size (ack is the smallest DCEP message) + if ((size_t)data_length < sizeof(*ack)) { + DC_WARN(("Ignored invalid DCEP message (too short)")); + return; + } + + switch (req->msg_type) { + case DATA_CHANNEL_OPEN_REQUEST: + // structure includes a possibly-unused char label[1] (in a packed + // structure) + if (NS_WARN_IF((size_t)data_length < sizeof(*req) - 1)) { + return; + } + + HandleOpenRequestMessage(req, data_length, stream); + break; + case DATA_CHANNEL_ACK: + // >= sizeof(*ack) checked above + + ack = static_cast<const struct rtcweb_datachannel_ack*>(buffer); + HandleOpenAckMessage(ack, data_length, stream); + break; + default: + HandleUnknownMessage(ppid, data_length, stream); + break; + } + + // Reset buffer + mRecvBuffer.Truncate(0); +} + +// Called with mLock locked! +void DataChannelConnection::HandleMessage(const void* buffer, size_t length, + uint32_t ppid, uint16_t stream, + int flags) { + mLock.AssertCurrentThreadOwns(); + + switch (ppid) { + case DATA_CHANNEL_PPID_CONTROL: + HandleDCEPMessage(buffer, length, ppid, stream, flags); + break; + case DATA_CHANNEL_PPID_DOMSTRING_PARTIAL: + case DATA_CHANNEL_PPID_DOMSTRING: + case DATA_CHANNEL_PPID_DOMSTRING_EMPTY: + case DATA_CHANNEL_PPID_BINARY_PARTIAL: + case DATA_CHANNEL_PPID_BINARY: + case DATA_CHANNEL_PPID_BINARY_EMPTY: + HandleDataMessage(buffer, length, ppid, stream, flags); + break; + default: + DC_ERROR(( + "Unhandled message of length %zu PPID %u on stream %u received (%s).", + length, ppid, stream, (flags & MSG_EOR) ? "complete" : "partial")); + break; + } +} + +void DataChannelConnection::HandleAssociationChangeEvent( + const struct sctp_assoc_change* sac) { + mLock.AssertCurrentThreadOwns(); + + uint32_t i, n; + const auto readyState = GetReadyState(); + switch (sac->sac_state) { + case SCTP_COMM_UP: + DC_DEBUG(("Association change: SCTP_COMM_UP")); + if (readyState == CONNECTING) { + mSocket = mMasterSocket; + SetReadyState(OPEN); + + DC_DEBUG(("Negotiated number of incoming streams: %" PRIu16, + sac->sac_inbound_streams)); + DC_DEBUG(("Negotiated number of outgoing streams: %" PRIu16, + sac->sac_outbound_streams)); + mNegotiatedIdLimit = + std::max(mNegotiatedIdLimit, + static_cast<size_t>(std::max(sac->sac_outbound_streams, + sac->sac_inbound_streams))); + + Dispatch(do_AddRef(new DataChannelOnMessageAvailable( + DataChannelOnMessageAvailable::ON_CONNECTION, this))); + DC_DEBUG(("DTLS connect() succeeded! Entering connected mode")); + + // Open any streams pending... + ProcessQueuedOpens(); + + } else if (readyState == OPEN) { + DC_DEBUG(("DataConnection Already OPEN")); + } else { + DC_ERROR(("Unexpected state: %d", readyState)); + } + break; + case SCTP_COMM_LOST: + DC_DEBUG(("Association change: SCTP_COMM_LOST")); + // This association is toast, so also close all the channels -- from + // mainthread! + Stop(); + break; + case SCTP_RESTART: + DC_DEBUG(("Association change: SCTP_RESTART")); + break; + case SCTP_SHUTDOWN_COMP: + DC_DEBUG(("Association change: SCTP_SHUTDOWN_COMP")); + Stop(); + break; + case SCTP_CANT_STR_ASSOC: + DC_DEBUG(("Association change: SCTP_CANT_STR_ASSOC")); + break; + default: + DC_DEBUG(("Association change: UNKNOWN")); + break; + } + DC_DEBUG(("Association change: streams (in/out) = (%u/%u)", + sac->sac_inbound_streams, sac->sac_outbound_streams)); + + if (NS_WARN_IF(!sac)) { + return; + } + + n = sac->sac_length - sizeof(*sac); + if ((sac->sac_state == SCTP_COMM_UP) || (sac->sac_state == SCTP_RESTART)) { + if (n > 0) { + for (i = 0; i < n; ++i) { + switch (sac->sac_info[i]) { + case SCTP_ASSOC_SUPPORTS_PR: + DC_DEBUG(("Supports: PR")); + break; + case SCTP_ASSOC_SUPPORTS_AUTH: + DC_DEBUG(("Supports: AUTH")); + break; + case SCTP_ASSOC_SUPPORTS_ASCONF: + DC_DEBUG(("Supports: ASCONF")); + break; + case SCTP_ASSOC_SUPPORTS_MULTIBUF: + DC_DEBUG(("Supports: MULTIBUF")); + break; + case SCTP_ASSOC_SUPPORTS_RE_CONFIG: + DC_DEBUG(("Supports: RE-CONFIG")); + break; +#if defined(SCTP_ASSOC_SUPPORTS_INTERLEAVING) + case SCTP_ASSOC_SUPPORTS_INTERLEAVING: + DC_DEBUG(("Supports: NDATA")); + // TODO: This should probably be set earlier above in 'case + // SCTP_COMM_UP' but we also need this for 'SCTP_RESTART'. + mSendInterleaved = true; + break; +#endif + default: + DC_ERROR(("Supports: UNKNOWN(0x%02x)", sac->sac_info[i])); + break; + } + } + } + } else if (((sac->sac_state == SCTP_COMM_LOST) || + (sac->sac_state == SCTP_CANT_STR_ASSOC)) && + (n > 0)) { + DC_DEBUG(("Association: ABORT =")); + for (i = 0; i < n; ++i) { + DC_DEBUG((" 0x%02x", sac->sac_info[i])); + } + } + if ((sac->sac_state == SCTP_CANT_STR_ASSOC) || + (sac->sac_state == SCTP_SHUTDOWN_COMP) || + (sac->sac_state == SCTP_COMM_LOST)) { + return; + } +} + +void DataChannelConnection::HandlePeerAddressChangeEvent( + const struct sctp_paddr_change* spc) { + const char* addr = ""; +#if !defined(__Userspace_os_Windows) + char addr_buf[INET6_ADDRSTRLEN]; + struct sockaddr_in* sin; + struct sockaddr_in6* sin6; +#endif + + switch (spc->spc_aaddr.ss_family) { + case AF_INET: +#if !defined(__Userspace_os_Windows) + sin = (struct sockaddr_in*)&spc->spc_aaddr; + addr = inet_ntop(AF_INET, &sin->sin_addr, addr_buf, INET6_ADDRSTRLEN); +#endif + break; + case AF_INET6: +#if !defined(__Userspace_os_Windows) + sin6 = (struct sockaddr_in6*)&spc->spc_aaddr; + addr = inet_ntop(AF_INET6, &sin6->sin6_addr, addr_buf, INET6_ADDRSTRLEN); +#endif + break; + case AF_CONN: + addr = "DTLS connection"; + break; + default: + break; + } + DC_DEBUG(("Peer address %s is now ", addr)); + switch (spc->spc_state) { + case SCTP_ADDR_AVAILABLE: + DC_DEBUG(("SCTP_ADDR_AVAILABLE")); + break; + case SCTP_ADDR_UNREACHABLE: + DC_DEBUG(("SCTP_ADDR_UNREACHABLE")); + break; + case SCTP_ADDR_REMOVED: + DC_DEBUG(("SCTP_ADDR_REMOVED")); + break; + case SCTP_ADDR_ADDED: + DC_DEBUG(("SCTP_ADDR_ADDED")); + break; + case SCTP_ADDR_MADE_PRIM: + DC_DEBUG(("SCTP_ADDR_MADE_PRIM")); + break; + case SCTP_ADDR_CONFIRMED: + DC_DEBUG(("SCTP_ADDR_CONFIRMED")); + break; + default: + DC_ERROR(("UNKNOWN SCP STATE")); + break; + } + if (spc->spc_error) { + DC_ERROR((" (error = 0x%08x).\n", spc->spc_error)); + } +} + +void DataChannelConnection::HandleRemoteErrorEvent( + const struct sctp_remote_error* sre) { + size_t i, n; + + n = sre->sre_length - sizeof(struct sctp_remote_error); + DC_WARN(("Remote Error (error = 0x%04x): ", sre->sre_error)); + for (i = 0; i < n; ++i) { + DC_WARN((" 0x%02x", sre->sre_data[i])); + } +} + +void DataChannelConnection::HandleShutdownEvent( + const struct sctp_shutdown_event* sse) { + DC_DEBUG(("Shutdown event.")); + /* XXX: notify all channels. */ + // Attempts to actually send anything will fail +} + +void DataChannelConnection::HandleAdaptationIndication( + const struct sctp_adaptation_event* sai) { + DC_DEBUG(("Adaptation indication: %x.", sai->sai_adaptation_ind)); +} + +void DataChannelConnection::HandlePartialDeliveryEvent( + const struct sctp_pdapi_event* spde) { + // Note: Be aware that stream and sequence number being u32 instead of u16 is + // a bug in the SCTP API. This may change in the future. + + DC_DEBUG(("Partial delivery event: ")); + switch (spde->pdapi_indication) { + case SCTP_PARTIAL_DELIVERY_ABORTED: + DC_DEBUG(("delivery aborted ")); + break; + default: + DC_ERROR(("??? ")); + break; + } + DC_DEBUG(("(flags = %x), stream = %" PRIu32 ", sn = %" PRIu32, + spde->pdapi_flags, spde->pdapi_stream, spde->pdapi_seq)); + + // Validate stream ID + if (spde->pdapi_stream >= UINT16_MAX) { + DC_ERROR(("Invalid stream id in partial delivery event: %" PRIu32 "\n", + spde->pdapi_stream)); + return; + } + + // Find channel and reset buffer + DataChannel* channel = FindChannelByStream((uint16_t)spde->pdapi_stream); + if (channel) { + DC_WARN(("Abort partially delivered message of %zu bytes\n", + channel->mRecvBuffer.Length())); + channel->mRecvBuffer.Truncate(0); + } +} + +void DataChannelConnection::HandleSendFailedEvent( + const struct sctp_send_failed_event* ssfe) { + size_t i, n; + + if (ssfe->ssfe_flags & SCTP_DATA_UNSENT) { + DC_DEBUG(("Unsent ")); + } + if (ssfe->ssfe_flags & SCTP_DATA_SENT) { + DC_DEBUG(("Sent ")); + } + if (ssfe->ssfe_flags & ~(SCTP_DATA_SENT | SCTP_DATA_UNSENT)) { + DC_DEBUG(("(flags = %x) ", ssfe->ssfe_flags)); + } +#ifdef XP_WIN +# define PRIPPID "lu" +#else +# define PRIPPID "u" +#endif + DC_DEBUG(("message with PPID = %" PRIPPID + ", SID = %d, flags: 0x%04x due to error = 0x%08x", + ntohl(ssfe->ssfe_info.snd_ppid), ssfe->ssfe_info.snd_sid, + ssfe->ssfe_info.snd_flags, ssfe->ssfe_error)); +#undef PRIPPID + n = ssfe->ssfe_length - sizeof(struct sctp_send_failed_event); + for (i = 0; i < n; ++i) { + DC_DEBUG((" 0x%02x", ssfe->ssfe_data[i])); + } +} + +void DataChannelConnection::ClearResets() { + // Clear all pending resets + if (!mStreamsResetting.IsEmpty()) { + DC_DEBUG(("Clearing resets for %zu streams", mStreamsResetting.Length())); + } + + for (uint32_t i = 0; i < mStreamsResetting.Length(); ++i) { + RefPtr<DataChannel> channel; + channel = FindChannelByStream(mStreamsResetting[i]); + if (channel) { + DC_DEBUG(("Forgetting channel %u (%p) with pending reset", + channel->mStream, channel.get())); + // TODO: Do we _really_ want to remove this? Are we allowed to reuse the + // id? + mChannels.Remove(channel); + } + } + mStreamsResetting.Clear(); +} + +void DataChannelConnection::ResetOutgoingStream(uint16_t stream) { + uint32_t i; + + mLock.AssertCurrentThreadOwns(); + DC_DEBUG( + ("Connection %p: Resetting outgoing stream %u", (void*)this, stream)); + // Rarely has more than a couple items and only for a short time + for (i = 0; i < mStreamsResetting.Length(); ++i) { + if (mStreamsResetting[i] == stream) { + return; + } + } + mStreamsResetting.AppendElement(stream); +} + +void DataChannelConnection::SendOutgoingStreamReset() { + struct sctp_reset_streams* srs; + uint32_t i; + size_t len; + + DC_DEBUG(("Connection %p: Sending outgoing stream reset for %zu streams", + (void*)this, mStreamsResetting.Length())); + mLock.AssertCurrentThreadOwns(); + if (mStreamsResetting.IsEmpty()) { + DC_DEBUG(("No streams to reset")); + return; + } + len = sizeof(sctp_assoc_t) + + (2 + mStreamsResetting.Length()) * sizeof(uint16_t); + srs = static_cast<struct sctp_reset_streams*>( + moz_xmalloc(len)); // infallible malloc + memset(srs, 0, len); + srs->srs_flags = SCTP_STREAM_RESET_OUTGOING; + srs->srs_number_streams = mStreamsResetting.Length(); + for (i = 0; i < mStreamsResetting.Length(); ++i) { + srs->srs_stream_list[i] = mStreamsResetting[i]; + } + if (usrsctp_setsockopt(mMasterSocket, IPPROTO_SCTP, SCTP_RESET_STREAMS, srs, + (socklen_t)len) < 0) { + DC_ERROR(("***failed: setsockopt RESET, errno %d", errno)); + // if errno == EALREADY, this is normal - we can't send another reset + // with one pending. + // When we get an incoming reset (which may be a response to our + // outstanding one), see if we have any pending outgoing resets and + // send them + } else { + mStreamsResetting.Clear(); + } + free(srs); +} + +void DataChannelConnection::HandleStreamResetEvent( + const struct sctp_stream_reset_event* strrst) { + uint32_t n, i; + RefPtr<DataChannel> channel; // since we may null out the ref to the channel + + if (!(strrst->strreset_flags & SCTP_STREAM_RESET_DENIED) && + !(strrst->strreset_flags & SCTP_STREAM_RESET_FAILED)) { + n = (strrst->strreset_length - sizeof(struct sctp_stream_reset_event)) / + sizeof(uint16_t); + for (i = 0; i < n; ++i) { + if (strrst->strreset_flags & SCTP_STREAM_RESET_INCOMING_SSN) { + channel = FindChannelByStream(strrst->strreset_stream_list[i]); + if (channel) { + // The other side closed the channel + // We could be in three states: + // 1. Normal state (input and output streams (OPEN) + // Notify application, send a RESET in response on our + // outbound channel. Go to CLOSED + // 2. We sent our own reset (CLOSING); either they crossed on the + // wire, or this is a response to our Reset. + // Go to CLOSED + // 3. We've sent a open but haven't gotten a response yet (CONNECTING) + // I believe this is impossible, as we don't have an input stream + // yet. + + DC_DEBUG(("Incoming: Channel %u closed", channel->mStream)); + if (mChannels.Remove(channel)) { + // Mark the stream for reset (the reset is sent below) + ResetOutgoingStream(channel->mStream); + } + + DC_DEBUG(("Disconnected DataChannel %p from connection %p", + (void*)channel.get(), (void*)channel->mConnection.get())); + channel->StreamClosedLocked(); + } else { + DC_WARN(("Can't find incoming channel %d", i)); + } + } + } + } + + // Process any pending resets now: + if (!mStreamsResetting.IsEmpty()) { + DC_DEBUG(("Sending %zu pending resets", mStreamsResetting.Length())); + SendOutgoingStreamReset(); + } +} + +void DataChannelConnection::HandleStreamChangeEvent( + const struct sctp_stream_change_event* strchg) { + ASSERT_WEBRTC(!NS_IsMainThread()); + if (strchg->strchange_flags == SCTP_STREAM_CHANGE_DENIED) { + DC_ERROR(("*** Failed increasing number of streams from %zu (%u/%u)", + mNegotiatedIdLimit, strchg->strchange_instrms, + strchg->strchange_outstrms)); + // XXX FIX! notify pending opens of failure + return; + } + if (strchg->strchange_instrms > mNegotiatedIdLimit) { + DC_DEBUG(("Other side increased streams from %zu to %u", mNegotiatedIdLimit, + strchg->strchange_instrms)); + } + uint16_t old_limit = mNegotiatedIdLimit; + uint16_t new_limit = + std::max(strchg->strchange_outstrms, strchg->strchange_instrms); + if (new_limit > mNegotiatedIdLimit) { + DC_DEBUG(("Increasing number of streams from %u to %u - adding %u (in: %u)", + old_limit, new_limit, new_limit - old_limit, + strchg->strchange_instrms)); + // make sure both are the same length + mNegotiatedIdLimit = new_limit; + DC_DEBUG(("New length = %zu (was %d)", mNegotiatedIdLimit, old_limit)); + // Re-process any channels waiting for streams. + // Linear search, but we don't increase channels often and + // the array would only get long in case of an app error normally + + // Make sure we request enough streams if there's a big jump in streams + // Could make a more complex API for OpenXxxFinish() and avoid this loop + auto channels = mChannels.GetAll(); + size_t num_needed = + channels.Length() ? (channels.LastElement()->mStream + 1) : 0; + MOZ_ASSERT(num_needed != INVALID_STREAM); + if (num_needed > new_limit) { + int32_t more_needed = num_needed - ((int32_t)mNegotiatedIdLimit) + 16; + DC_DEBUG(("Not enough new streams, asking for %d more", more_needed)); + // TODO: parameter is an int32_t but we pass size_t + RequestMoreStreams(more_needed); + } else if (strchg->strchange_outstrms < strchg->strchange_instrms) { + DC_DEBUG(("Requesting %d output streams to match partner", + strchg->strchange_instrms - strchg->strchange_outstrms)); + RequestMoreStreams(strchg->strchange_instrms - + strchg->strchange_outstrms); + } + + ProcessQueuedOpens(); + } + // else probably not a change in # of streams + + if ((strchg->strchange_flags & SCTP_STREAM_CHANGE_DENIED) || + (strchg->strchange_flags & SCTP_STREAM_CHANGE_FAILED)) { + // Other side denied our request. Need to AnnounceClosed some stuff. + for (auto& channel : mChannels.GetAll()) { + if (channel->mStream >= mNegotiatedIdLimit) { + /* XXX: Signal to the other end. */ + channel->AnnounceClosed(); + // maybe fire onError (bug 843625) + } + } + } +} + +// Called with mLock locked! +void DataChannelConnection::HandleNotification( + const union sctp_notification* notif, size_t n) { + mLock.AssertCurrentThreadOwns(); + if (notif->sn_header.sn_length != (uint32_t)n) { + return; + } + switch (notif->sn_header.sn_type) { + case SCTP_ASSOC_CHANGE: + HandleAssociationChangeEvent(&(notif->sn_assoc_change)); + break; + case SCTP_PEER_ADDR_CHANGE: + HandlePeerAddressChangeEvent(&(notif->sn_paddr_change)); + break; + case SCTP_REMOTE_ERROR: + HandleRemoteErrorEvent(&(notif->sn_remote_error)); + break; + case SCTP_SHUTDOWN_EVENT: + HandleShutdownEvent(&(notif->sn_shutdown_event)); + break; + case SCTP_ADAPTATION_INDICATION: + HandleAdaptationIndication(&(notif->sn_adaptation_event)); + break; + case SCTP_AUTHENTICATION_EVENT: + DC_DEBUG(("SCTP_AUTHENTICATION_EVENT")); + break; + case SCTP_SENDER_DRY_EVENT: + // DC_DEBUG(("SCTP_SENDER_DRY_EVENT")); + break; + case SCTP_NOTIFICATIONS_STOPPED_EVENT: + DC_DEBUG(("SCTP_NOTIFICATIONS_STOPPED_EVENT")); + break; + case SCTP_PARTIAL_DELIVERY_EVENT: + HandlePartialDeliveryEvent(&(notif->sn_pdapi_event)); + break; + case SCTP_SEND_FAILED_EVENT: + HandleSendFailedEvent(&(notif->sn_send_failed_event)); + break; + case SCTP_STREAM_RESET_EVENT: + HandleStreamResetEvent(&(notif->sn_strreset_event)); + break; + case SCTP_ASSOC_RESET_EVENT: + DC_DEBUG(("SCTP_ASSOC_RESET_EVENT")); + break; + case SCTP_STREAM_CHANGE_EVENT: + HandleStreamChangeEvent(&(notif->sn_strchange_event)); + break; + default: + DC_ERROR(("unknown SCTP event: %u", (uint32_t)notif->sn_header.sn_type)); + break; + } +} + +int DataChannelConnection::ReceiveCallback( + struct socket* sock, void* data, size_t datalen, struct sctp_rcvinfo rcv, + int flags) MOZ_NO_THREAD_SAFETY_ANALYSIS { + ASSERT_WEBRTC(!NS_IsMainThread()); + DC_DEBUG(("In ReceiveCallback")); + + // libusrsctp just went reentrant on us. Put a stop to this. + mSTS->Dispatch(NS_NewRunnableFunction( + "DataChannelConnection::ReceiveCallback", + [data, datalen, rcv, flags, this, + self = RefPtr<DataChannelConnection>(this)]() mutable { + if (!data) { + DC_DEBUG(("ReceiveCallback: SCTP has finished shutting down")); + } else { + mLock.Lock(); + if (flags & MSG_NOTIFICATION) { + HandleNotification(static_cast<union sctp_notification*>(data), + datalen); + } else { + HandleMessage(data, datalen, ntohl(rcv.rcv_ppid), rcv.rcv_sid, + flags); + } + mLock.Unlock(); + // sctp allocates 'data' with malloc(), and expects the receiver to + // free it (presumably with free). + // XXX future optimization: try to deliver messages without an + // internal alloc/copy, and if so delay the free until later. + free(data); + } + })); + + // usrsctp defines the callback as returning an int, but doesn't use it + return 1; +} + +already_AddRefed<DataChannel> DataChannelConnection::Open( + const nsACString& label, const nsACString& protocol, Type type, + bool inOrder, uint32_t prValue, DataChannelListener* aListener, + nsISupports* aContext, bool aExternalNegotiated, uint16_t aStream) { + ASSERT_WEBRTC(NS_IsMainThread()); + MutexAutoLock lock(mLock); // OpenFinish assumes this + if (!aExternalNegotiated) { + if (mAllocateEven.isSome()) { + aStream = FindFreeStream(); + if (aStream == INVALID_STREAM) { + return nullptr; + } + } else { + // We do not yet know whether we are client or server, and an id has not + // been chosen for us. We will need to choose later. + aStream = INVALID_STREAM; + } + } + uint16_t prPolicy = SCTP_PR_SCTP_NONE; + + DC_DEBUG( + ("DC Open: label %s/%s, type %u, inorder %d, prValue %u, listener %p, " + "context %p, external: %s, stream %u", + PromiseFlatCString(label).get(), PromiseFlatCString(protocol).get(), + type, inOrder, prValue, aListener, aContext, + aExternalNegotiated ? "true" : "false", aStream)); + switch (type) { + case DATA_CHANNEL_RELIABLE: + prPolicy = SCTP_PR_SCTP_NONE; + break; + case DATA_CHANNEL_PARTIAL_RELIABLE_REXMIT: + prPolicy = SCTP_PR_SCTP_RTX; + break; + case DATA_CHANNEL_PARTIAL_RELIABLE_TIMED: + prPolicy = SCTP_PR_SCTP_TTL; + break; + default: + DC_ERROR(("unsupported channel type: %u", type)); + MOZ_ASSERT(false); + return nullptr; + } + if ((prPolicy == SCTP_PR_SCTP_NONE) && (prValue != 0)) { + return nullptr; + } + + if (aStream != INVALID_STREAM && mChannels.Get(aStream)) { + DC_ERROR(("external negotiation of already-open channel %u", aStream)); + return nullptr; + } + + RefPtr<DataChannel> channel(new DataChannel( + this, aStream, DataChannel::CONNECTING, label, protocol, prPolicy, + prValue, inOrder, aExternalNegotiated, aListener, aContext)); + mChannels.Insert(channel); + + return OpenFinish(channel.forget()); +} + +// Separate routine so we can also call it to finish up from pending opens +already_AddRefed<DataChannel> DataChannelConnection::OpenFinish( + already_AddRefed<DataChannel>&& aChannel) { + RefPtr<DataChannel> channel(aChannel); // takes the reference passed in + // Normally 1 reference if called from ::Open(), or 2 if called from + // ProcessQueuedOpens() unless the DOMDataChannel was gc'd + const uint16_t stream = channel->mStream; + + mLock.AssertCurrentThreadOwns(); + + // Cases we care about: + // Pre-negotiated: + // Not Open: + // Doesn't fit: + // -> change initial ask or renegotiate after open + // -> queue open + // Open: + // Doesn't fit: + // -> RequestMoreStreams && queue + // Does fit: + // -> open + // Not negotiated: + // Not Open: + // -> queue open + // Open: + // -> Try to get a stream + // Doesn't fit: + // -> RequestMoreStreams && queue + // Does fit: + // -> open + // So the Open cases are basically the same + // Not Open cases are simply queue for non-negotiated, and + // either change the initial ask or possibly renegotiate after open. + const auto readyState = GetReadyState(); + if (readyState != OPEN || stream >= mNegotiatedIdLimit) { + if (readyState == OPEN) { + MOZ_ASSERT(stream != INVALID_STREAM); + // RequestMoreStreams() limits to MAX_NUM_STREAMS -- allocate extra + // streams to avoid going back immediately for more if the ask to N, N+1, + // etc + int32_t more_needed = stream - ((int32_t)mNegotiatedIdLimit) + 16; + if (!RequestMoreStreams(more_needed)) { + // Something bad happened... we're done + goto request_error_cleanup; + } + } + DC_DEBUG(("Queuing channel %p (%u) to finish open", channel.get(), stream)); + // Also serves to mark we told the app + channel->mFlags |= DATA_CHANNEL_FLAGS_FINISH_OPEN; + mPending.Push(channel); + return channel.forget(); + } + + MOZ_ASSERT(stream != INVALID_STREAM); + MOZ_ASSERT(stream < mNegotiatedIdLimit); + +#ifdef TEST_QUEUED_DATA + // It's painful to write a test for this... + channel->AnnounceOpen(); + SendDataMsgInternalOrBuffer(channel, "Help me!", 8, + DATA_CHANNEL_PPID_DOMSTRING); +#endif + + if (!channel->mOrdered) { + // Don't send unordered until this gets cleared + channel->mFlags |= DATA_CHANNEL_FLAGS_WAITING_ACK; + } + + if (!channel->mNegotiated) { + int error = SendOpenRequestMessage(channel->mLabel, channel->mProtocol, + stream, !channel->mOrdered, + channel->mPrPolicy, channel->mPrValue); + if (error) { + DC_ERROR(("SendOpenRequest failed, error = %d", error)); + if (channel->mFlags & DATA_CHANNEL_FLAGS_FINISH_OPEN) { + // We already returned the channel to the app. + NS_ERROR("Failed to send open request"); + channel->AnnounceClosed(); + } + // If we haven't returned the channel yet, it will get destroyed when we + // exit this function. + mChannels.Remove(channel); + // we'll be destroying the channel + return nullptr; + /* NOTREACHED */ + } + } + + // Either externally negotiated or we sent Open + // FIX? Move into DOMDataChannel? I don't think we can send it yet here + channel->AnnounceOpen(); + + return channel.forget(); + +request_error_cleanup: + if (channel->mFlags & DATA_CHANNEL_FLAGS_FINISH_OPEN) { + // We already returned the channel to the app. + NS_ERROR("Failed to request more streams"); + channel->AnnounceClosed(); + return channel.forget(); + } + // we'll be destroying the channel, but it never really got set up + // Alternative would be to RUN_ON_THREAD(channel.forget(),::Destroy,...) and + // Dispatch it to ourselves + return nullptr; +} + +// Requires mLock to be locked! +// Returns a POSIX error code directly instead of setting errno. +int DataChannelConnection::SendMsgInternal(OutgoingMsg& msg, size_t* aWritten) { + auto& info = msg.GetInfo().sendv_sndinfo; + int error; + + // EOR set? + bool eor_set = info.snd_flags & SCTP_EOR ? true : false; + + // Send until buffer is empty + size_t left = msg.GetLeft(); + do { + size_t length; + + // Carefully chunk the buffer + if (left > DATA_CHANNEL_MAX_BINARY_FRAGMENT) { + length = DATA_CHANNEL_MAX_BINARY_FRAGMENT; + + // Unset EOR flag + info.snd_flags &= ~SCTP_EOR; + } else { + length = left; + + // Set EOR flag + if (eor_set) { + info.snd_flags |= SCTP_EOR; + } + } + + // Send (or try at least) + // SCTP will return EMSGSIZE if the message is bigger than the buffer + // size (or EAGAIN if there isn't space). However, we can avoid EMSGSIZE + // by carefully crafting small enough message chunks. + ssize_t written = usrsctp_sendv( + mSocket, msg.GetData(), length, nullptr, 0, (void*)&msg.GetInfo(), + (socklen_t)sizeof(struct sctp_sendv_spa), SCTP_SENDV_SPA, 0); + + if (written < 0) { + error = errno; + goto out; + } + + if (aWritten) { + *aWritten += written; + } + DC_DEBUG(("Sent buffer (written=%zu, len=%zu, left=%zu)", (size_t)written, + length, left - (size_t)written)); + + // TODO: Remove once resolved + // (https://github.com/sctplab/usrsctp/issues/132) + if (written == 0) { + DC_ERROR(("@tuexen: usrsctp_sendv returned 0")); + error = EAGAIN; + goto out; + } + + // If not all bytes have been written, this obviously means that usrsctp's + // buffer is full and we need to try again later. + if ((size_t)written < length) { + msg.Advance((size_t)written); + error = EAGAIN; + goto out; + } + + // Update buffer position + msg.Advance((size_t)written); + + // Get amount of bytes left in the buffer + left = msg.GetLeft(); + } while (left > 0); + + // Done + error = 0; + +out: + // Reset EOR flag + if (eor_set) { + info.snd_flags |= SCTP_EOR; + } + + return error; +} + +// Requires mLock to be locked! +// Returns a POSIX error code directly instead of setting errno. +// IMPORTANT: Ensure that the buffer passed is guarded by mLock! +int DataChannelConnection::SendMsgInternalOrBuffer( + nsTArray<UniquePtr<BufferedOutgoingMsg>>& buffer, OutgoingMsg& msg, + bool& buffered, size_t* aWritten) { + NS_WARNING_ASSERTION(msg.GetLength() > 0, "Length is 0?!"); + + int error = 0; + bool need_buffering = false; + + // Note: Main-thread IO, but doesn't block! + // XXX FIX! to deal with heavy overruns of JS trying to pass data in + // (more than the buffersize) queue data onto another thread to do the + // actual sends. See netwerk/protocol/websocket/WebSocketChannel.cpp + + // Avoid a race between buffer-full-failure (where we have to add the + // packet to the buffered-data queue) and the buffer-now-only-half-full + // callback, which happens on a different thread. Otherwise we might + // fail here, then before we add it to the queue get the half-full + // callback, find nothing to do, then on this thread add it to the + // queue - which would sit there. Also, if we later send more data, it + // would arrive ahead of the buffered message, but if the buffer ever + // got to 1/2 full, the message would get sent - but at a semi-random + // time, after other data it was supposed to be in front of. + + // Must lock before empty check for similar reasons! + mLock.AssertCurrentThreadOwns(); + if (buffer.IsEmpty() && (mSendInterleaved || !mPendingType)) { + error = SendMsgInternal(msg, aWritten); + switch (error) { + case 0: + break; + case EAGAIN: +#if (EAGAIN != EWOULDBLOCK) + case EWOULDBLOCK: +#endif + need_buffering = true; + break; + default: + DC_ERROR(("error %d on sending", error)); + break; + } + } else { + need_buffering = true; + } + + if (need_buffering) { + // queue data for resend! And queue any further data for the stream until + // it is... + auto* bufferedMsg = new BufferedOutgoingMsg(msg); // infallible malloc + buffer.AppendElement(bufferedMsg); // owned by mBufferedData array + DC_DEBUG(("Queued %zu buffers (left=%zu, total=%zu)", buffer.Length(), + msg.GetLeft(), msg.GetLength())); + buffered = true; + return 0; + } + + buffered = false; + return error; +} + +// Caller must ensure that length <= UINT32_MAX +// Returns a POSIX error code. +int DataChannelConnection::SendDataMsgInternalOrBuffer(DataChannel& channel, + const uint8_t* data, + size_t len, + uint32_t ppid) { + if (NS_WARN_IF(channel.GetReadyState() != OPEN)) { + return EINVAL; // TODO: Find a better error code + } + + struct sctp_sendv_spa info = {0}; + + // General flags + info.sendv_flags = SCTP_SEND_SNDINFO_VALID; + + // Set stream identifier, protocol identifier and flags + info.sendv_sndinfo.snd_sid = channel.mStream; + info.sendv_sndinfo.snd_flags = SCTP_EOR; + info.sendv_sndinfo.snd_ppid = htonl(ppid); + + // Unordered? + // To avoid problems where an in-order OPEN is lost and an + // out-of-order data message "beats" it, require data to be in-order + // until we get an ACK. + if (!channel.mOrdered && !(channel.mFlags & DATA_CHANNEL_FLAGS_WAITING_ACK)) { + info.sendv_sndinfo.snd_flags |= SCTP_UNORDERED; + } + + // Partial reliability policy + if (channel.mPrPolicy != SCTP_PR_SCTP_NONE) { + info.sendv_prinfo.pr_policy = channel.mPrPolicy; + info.sendv_prinfo.pr_value = channel.mPrValue; + info.sendv_flags |= SCTP_SEND_PRINFO_VALID; + } + + // Create message instance and send + OutgoingMsg msg(info, data, len); + bool buffered; + size_t written = 0; + mDeferSend = true; + int error = + SendMsgInternalOrBuffer(channel.mBufferedData, msg, buffered, &written); + mDeferSend = false; + if (written && ppid != DATA_CHANNEL_PPID_DOMSTRING_EMPTY && + ppid != DATA_CHANNEL_PPID_BINARY_EMPTY) { + channel.DecrementBufferedAmount(written); + } + + for (auto&& packet : mDeferredSend) { + MOZ_ASSERT(written); + SendPacket(std::move(packet)); + } + mDeferredSend.clear(); + + // Set pending type and stream index (if buffered) + if (!error && buffered && !mPendingType) { + mPendingType = PENDING_DATA; + mCurrentStream = channel.mStream; + } + return error; +} + +// Caller must ensure that length <= UINT32_MAX +// Returns a POSIX error code. +int DataChannelConnection::SendDataMsg(DataChannel& channel, + const uint8_t* data, size_t len, + uint32_t ppidPartial, + uint32_t ppidFinal) { + // We *really* don't want to do this from main thread! - and + // SendDataMsgInternalOrBuffer avoids blocking. + mLock.AssertCurrentThreadOwns(); + + if (mMaxMessageSize != 0 && len > mMaxMessageSize) { + DC_ERROR(("Message rejected, too large (%zu > %" PRIu64 ")", len, + mMaxMessageSize)); + return EMSGSIZE; + } + + // This will use EOR-based fragmentation if the message is too large (> 64 + // KiB) + return SendDataMsgInternalOrBuffer(channel, data, len, ppidFinal); +} + +class ReadBlobRunnable : public Runnable { + public: + ReadBlobRunnable(DataChannelConnection* aConnection, uint16_t aStream, + nsIInputStream* aBlob) + : Runnable("ReadBlobRunnable"), + mConnection(aConnection), + mStream(aStream), + mBlob(aBlob) {} + + NS_IMETHOD Run() override { + // ReadBlob() is responsible to releasing the reference + DataChannelConnection* self = mConnection; + self->ReadBlob(mConnection.forget(), mStream, mBlob); + return NS_OK; + } + + private: + // Make sure the Connection doesn't die while there are jobs outstanding. + // Let it die (if released by PeerConnectionImpl while we're running) + // when we send our runnable back to MainThread. Then ~DataChannelConnection + // can send the IOThread to MainThread to die in a runnable, avoiding + // unsafe event loop recursion. Evil. + RefPtr<DataChannelConnection> mConnection; + uint16_t mStream; + // Use RefCount for preventing the object is deleted when SendBlob returns. + RefPtr<nsIInputStream> mBlob; +}; + +// Returns a POSIX error code. +int DataChannelConnection::SendBlob(uint16_t stream, nsIInputStream* aBlob) { + MutexAutoLock lock(mLock); + RefPtr<DataChannel> channel = mChannels.Get(stream); + if (NS_WARN_IF(!channel)) { + return EINVAL; // TODO: Find a better error code + } + + // Spawn a thread to send the data + if (!mInternalIOThread) { + nsresult rv = + NS_NewNamedThread("DataChannel IO", getter_AddRefs(mInternalIOThread)); + if (NS_FAILED(rv)) { + return EINVAL; // TODO: Find a better error code + } + } + + mInternalIOThread->Dispatch( + do_AddRef(new ReadBlobRunnable(this, stream, aBlob)), NS_DISPATCH_NORMAL); + return 0; +} + +class DataChannelBlobSendRunnable : public Runnable { + public: + DataChannelBlobSendRunnable( + already_AddRefed<DataChannelConnection>& aConnection, uint16_t aStream) + : Runnable("DataChannelBlobSendRunnable"), + mConnection(aConnection), + mStream(aStream) {} + + ~DataChannelBlobSendRunnable() override { + if (!NS_IsMainThread() && mConnection) { + MOZ_ASSERT(false); + // explicitly leak the connection if destroyed off mainthread + Unused << mConnection.forget().take(); + } + } + + NS_IMETHOD Run() override { + ASSERT_WEBRTC(NS_IsMainThread()); + + mConnection->SendBinaryMsg(mStream, mData); + mConnection = nullptr; + return NS_OK; + } + + // explicitly public so we can avoid allocating twice and copying + nsCString mData; + + private: + // Note: we can be destroyed off the target thread, so be careful not to let + // this get Released()ed on the temp thread! + RefPtr<DataChannelConnection> mConnection; + uint16_t mStream; +}; + +static auto readyStateToCStr(const uint16_t state) -> const char* { + switch (state) { + case DataChannelConnection::CONNECTING: + return "CONNECTING"; + case DataChannelConnection::OPEN: + return "OPEN"; + case DataChannelConnection::CLOSING: + return "CLOSING"; + case DataChannelConnection::CLOSED: + return "CLOSED"; + default: { + MOZ_ASSERT(false); + return "UNKNOWW"; + } + } +}; + +void DataChannelConnection::SetReadyState(const uint16_t aState) { + mLock.AssertCurrentThreadOwns(); + + DC_DEBUG( + ("DataChannelConnection labeled %s (%p) switching connection state %s -> " + "%s", + mTransportId.c_str(), this, readyStateToCStr(mState), + readyStateToCStr(aState))); + + mState = aState; +} + +void DataChannelConnection::ReadBlob( + already_AddRefed<DataChannelConnection> aThis, uint16_t aStream, + nsIInputStream* aBlob) { + // NOTE: 'aThis' has been forgotten by the caller to avoid releasing + // it off mainthread; if PeerConnectionImpl has released then we want + // ~DataChannelConnection() to run on MainThread + + // XXX to do this safely, we must enqueue these atomically onto the + // output socket. We need a sender thread(s?) to enqueue data into the + // socket and to avoid main-thread IO that might block. Even on a + // background thread, we may not want to block on one stream's data. + // I.e. run non-blocking and service multiple channels. + + // Must not let Dispatching it cause the DataChannelConnection to get + // released on the wrong thread. Using + // WrapRunnable(RefPtr<DataChannelConnection>(aThis),... will occasionally + // cause aThis to get released on this thread. Also, an explicit Runnable + // lets us avoid copying the blob data an extra time. + RefPtr<DataChannelBlobSendRunnable> runnable = + new DataChannelBlobSendRunnable(aThis, aStream); + // avoid copying the blob data by passing the mData from the runnable + if (NS_FAILED(NS_ReadInputStreamToString(aBlob, runnable->mData, -1))) { + // Bug 966602: Doesn't return an error to the caller via onerror. + // We must release DataChannelConnection on MainThread to avoid issues (bug + // 876167) aThis is now owned by the runnable; release it there + NS_ReleaseOnMainThread("DataChannelBlobSendRunnable", runnable.forget()); + return; + } + aBlob->Close(); + Dispatch(runnable.forget()); +} + +// Returns a POSIX error code. +int DataChannelConnection::SendDataMsgCommon(uint16_t stream, + const nsACString& aMsg, + bool isBinary) { + ASSERT_WEBRTC(NS_IsMainThread()); + // We really could allow this from other threads, so long as we deal with + // asynchronosity issues with channels closing, in particular access to + // mChannels, and issues with the association closing (access to mSocket). + + const uint8_t* data = (const uint8_t*)aMsg.BeginReading(); + uint32_t len = aMsg.Length(); +#if (UINT32_MAX > SIZE_MAX) + if (len > SIZE_MAX) { + return EMSGSIZE; + } +#endif + + DC_DEBUG(("Sending %sto stream %u: %u bytes", isBinary ? "binary " : "", + stream, len)); + // XXX if we want more efficiency, translate flags once at open time + RefPtr<DataChannel> channelPtr = mChannels.Get(stream); + if (NS_WARN_IF(!channelPtr)) { + return EINVAL; // TODO: Find a better error code + } + bool is_empty = len == 0; + const uint8_t byte = 0; + if (is_empty) { + data = &byte; + len = 1; + } + auto& channel = *channelPtr; + int err = 0; + MutexAutoLock lock(mLock); + if (isBinary) { + err = SendDataMsg( + channel, data, len, DATA_CHANNEL_PPID_BINARY_PARTIAL, + is_empty ? DATA_CHANNEL_PPID_BINARY_EMPTY : DATA_CHANNEL_PPID_BINARY); + } else { + err = SendDataMsg(channel, data, len, DATA_CHANNEL_PPID_DOMSTRING_PARTIAL, + is_empty ? DATA_CHANNEL_PPID_DOMSTRING_EMPTY + : DATA_CHANNEL_PPID_DOMSTRING); + } + if (!err) { + channel.WithTrafficCounters([&len](DataChannel::TrafficCounters& counters) { + counters.mMessagesSent++; + counters.mBytesSent += len; + }); + } + + return err; +} + +void DataChannelConnection::Stop() { + // Note: This will call 'CloseAll' from the main thread + Dispatch(do_AddRef(new DataChannelOnMessageAvailable( + DataChannelOnMessageAvailable::ON_DISCONNECTED, this))); +} + +void DataChannelConnection::Close(DataChannel* aChannel) { + MutexAutoLock lock(mLock); + CloseLocked(aChannel); +} + +// So we can call Close() with the lock already held +// Called from someone who holds a ref via ::Close(), or from ~DataChannel +void DataChannelConnection::CloseLocked(DataChannel* aChannel) { + MOZ_ASSERT(aChannel); + RefPtr<DataChannel> channel(aChannel); // make sure it doesn't go away on us + + mLock.AssertCurrentThreadOwns(); + DC_DEBUG(("Connection %p/Channel %p: Closing stream %u", + channel->mConnection.get(), channel.get(), channel->mStream)); + + aChannel->mBufferedData.Clear(); + if (GetReadyState() == CLOSED) { + // If we're CLOSING, we might leave this in place until we can send a + // reset. + mChannels.Remove(channel); + } + + // This is supposed to only be accessed from Main thread, but this has + // been accessed here from the STS thread for a long time now. + // See Bug 1586475 + auto channelState = aChannel->mReadyState; + // re-test since it may have closed before the lock was grabbed + if (channelState == CLOSED || channelState == CLOSING) { + DC_DEBUG(("Channel already closing/closed (%u)", channelState)); + return; + } + + if (channel->mStream != INVALID_STREAM) { + ResetOutgoingStream(channel->mStream); + if (GetReadyState() != CLOSED) { + // Individual channel is being closed, send reset now. + SendOutgoingStreamReset(); + } + } + aChannel->SetReadyState(CLOSING); + if (GetReadyState() == CLOSED) { + // we're not going to hang around waiting + channel->StreamClosedLocked(); + } + // At this point when we leave here, the object is a zombie held alive only by + // the DOM object +} + +void DataChannelConnection::CloseAll() { + DC_DEBUG(("Closing all channels (connection %p)", (void*)this)); + + // Make sure no more channels will be opened + MutexAutoLock lock(mLock); + SetReadyState(CLOSED); + + // Close current channels + // If there are runnables, they hold a strong ref and keep the channel + // and/or connection alive (even if in a CLOSED state) + for (auto& channel : mChannels.GetAll()) { + MutexAutoUnlock lock(mLock); + channel->Close(); + } + + // Clean up any pending opens for channels + RefPtr<DataChannel> channel; + while (nullptr != (channel = mPending.PopFront())) { + DC_DEBUG(("closing pending channel %p, stream %u", channel.get(), + channel->mStream)); + MutexAutoUnlock lock(mLock); + channel->Close(); // also releases the ref on each iteration + } + // It's more efficient to let the Resets queue in shutdown and then + // SendOutgoingStreamReset() here. + SendOutgoingStreamReset(); +} + +bool DataChannelConnection::Channels::IdComparator::Equals( + const RefPtr<DataChannel>& aChannel, uint16_t aId) const { + return aChannel->mStream == aId; +} + +bool DataChannelConnection::Channels::IdComparator::LessThan( + const RefPtr<DataChannel>& aChannel, uint16_t aId) const { + return aChannel->mStream < aId; +} + +bool DataChannelConnection::Channels::IdComparator::Equals( + const RefPtr<DataChannel>& a1, const RefPtr<DataChannel>& a2) const { + return Equals(a1, a2->mStream); +} + +bool DataChannelConnection::Channels::IdComparator::LessThan( + const RefPtr<DataChannel>& a1, const RefPtr<DataChannel>& a2) const { + return LessThan(a1, a2->mStream); +} + +void DataChannelConnection::Channels::Insert( + const RefPtr<DataChannel>& aChannel) { + DC_DEBUG(("Inserting channel %u : %p", aChannel->mStream, aChannel.get())); + MutexAutoLock lock(mMutex); + if (aChannel->mStream != INVALID_STREAM) { + MOZ_ASSERT(!mChannels.ContainsSorted(aChannel, IdComparator())); + } + + MOZ_ASSERT(!mChannels.Contains(aChannel)); + + mChannels.InsertElementSorted(aChannel, IdComparator()); +} + +bool DataChannelConnection::Channels::Remove( + const RefPtr<DataChannel>& aChannel) { + DC_DEBUG(("Removing channel %u : %p", aChannel->mStream, aChannel.get())); + MutexAutoLock lock(mMutex); + if (aChannel->mStream == INVALID_STREAM) { + return mChannels.RemoveElement(aChannel); + } + + return mChannels.RemoveElementSorted(aChannel, IdComparator()); +} + +RefPtr<DataChannel> DataChannelConnection::Channels::Get(uint16_t aId) const { + MutexAutoLock lock(mMutex); + auto index = mChannels.BinaryIndexOf(aId, IdComparator()); + if (index == ChannelArray::NoIndex) { + return nullptr; + } + return mChannels[index]; +} + +RefPtr<DataChannel> DataChannelConnection::Channels::GetNextChannel( + uint16_t aCurrentId) const { + MutexAutoLock lock(mMutex); + if (mChannels.IsEmpty()) { + return nullptr; + } + + auto index = mChannels.IndexOfFirstElementGt(aCurrentId, IdComparator()); + if (index == mChannels.Length()) { + index = 0; + } + return mChannels[index]; +} + +DataChannel::~DataChannel() { + // NS_ASSERTION since this is more "I think I caught all the cases that + // can cause this" than a true kill-the-program assertion. If this is + // wrong, nothing bad happens. A worst it's a leak. + NS_ASSERTION(mReadyState == CLOSED || mReadyState == CLOSING, + "unexpected state in ~DataChannel"); +} + +void DataChannel::Close() { + if (mConnection) { + // ensure we don't get deleted + RefPtr<DataChannelConnection> connection(mConnection); + connection->Close(this); + } +} + +// Used when disconnecting from the DataChannelConnection +void DataChannel::StreamClosedLocked() { + mConnection->mLock.AssertCurrentThreadOwns(); + ENSURE_DATACONNECTION; + + DC_DEBUG(("Destroying Data channel %u", mStream)); + MOZ_ASSERT_IF(mStream != INVALID_STREAM, + !mConnection->FindChannelByStream(mStream)); + AnnounceClosed(); + // We leave mConnection live until the DOM releases us, to avoid races +} + +void DataChannel::ReleaseConnection() { + ASSERT_WEBRTC(NS_IsMainThread()); + mConnection = nullptr; +} + +void DataChannel::SetListener(DataChannelListener* aListener, + nsISupports* aContext) { + ASSERT_WEBRTC(NS_IsMainThread()); + mContext = aContext; + mListener = aListener; +} + +void DataChannel::SendErrnoToErrorResult(int error, size_t aMessageSize, + ErrorResult& aRv) { + switch (error) { + case 0: + break; + case EMSGSIZE: { + nsPrintfCString err("Message size (%zu) exceeds maxMessageSize", + aMessageSize); + aRv.ThrowTypeError(err); + break; + } + default: + aRv.Throw(NS_ERROR_DOM_OPERATION_ERR); + break; + } +} + +void DataChannel::IncrementBufferedAmount(uint32_t aSize, ErrorResult& aRv) { + ASSERT_WEBRTC(NS_IsMainThread()); + if (mBufferedAmount > UINT32_MAX - aSize) { + aRv.Throw(NS_ERROR_FILE_TOO_BIG); + return; + } + + mBufferedAmount += aSize; +} + +void DataChannel::DecrementBufferedAmount(uint32_t aSize) { + mMainThreadEventTarget->Dispatch(NS_NewRunnableFunction( + "DataChannel::DecrementBufferedAmount", + [this, self = RefPtr<DataChannel>(this), aSize] { + MOZ_ASSERT(aSize <= mBufferedAmount); + bool wasLow = mBufferedAmount <= mBufferedThreshold; + mBufferedAmount -= aSize; + if (!wasLow && mBufferedAmount <= mBufferedThreshold) { + DC_DEBUG(("%s: sending BUFFER_LOW_THRESHOLD for %s/%s: %u", + __FUNCTION__, mLabel.get(), mProtocol.get(), mStream)); + mListener->OnBufferLow(mContext); + } + if (mBufferedAmount == 0) { + DC_DEBUG(("%s: sending NO_LONGER_BUFFERED for %s/%s: %u", + __FUNCTION__, mLabel.get(), mProtocol.get(), mStream)); + mListener->NotBuffered(mContext); + } + })); +} + +void DataChannel::AnnounceOpen() { + mMainThreadEventTarget->Dispatch(NS_NewRunnableFunction( + "DataChannel::AnnounceOpen", [this, self = RefPtr<DataChannel>(this)] { + auto state = GetReadyState(); + // Special-case; spec says to put brand-new remote-created DataChannel + // in "open", but queue the firing of the "open" event. + if (state != CLOSING && state != CLOSED) { + if (!mEverOpened && mConnection && mConnection->mListener) { + mEverOpened = true; + mConnection->mListener->NotifyDataChannelOpen(this); + } + SetReadyState(OPEN); + DC_DEBUG(("%s: sending ON_CHANNEL_OPEN for %s/%s: %u", __FUNCTION__, + mLabel.get(), mProtocol.get(), mStream)); + if (mListener) { + mListener->OnChannelConnected(mContext); + } + } + })); +} + +void DataChannel::AnnounceClosed() { + mMainThreadEventTarget->Dispatch(NS_NewRunnableFunction( + "DataChannel::AnnounceClosed", [this, self = RefPtr<DataChannel>(this)] { + if (GetReadyState() == CLOSED) { + return; + } + if (mEverOpened && mConnection && mConnection->mListener) { + mConnection->mListener->NotifyDataChannelClosed(this); + } + SetReadyState(CLOSED); + mBufferedData.Clear(); + if (mListener) { + DC_DEBUG(("%s: sending ON_CHANNEL_CLOSED for %s/%s: %u", __FUNCTION__, + mLabel.get(), mProtocol.get(), mStream)); + mListener->OnChannelClosed(mContext); + } + })); +} + +// Set ready state +void DataChannel::SetReadyState(const uint16_t aState) { + MOZ_ASSERT(NS_IsMainThread()); + + DC_DEBUG( + ("DataChannelConnection labeled %s(%p) (stream %d) changing ready state " + "%s -> %s", + mLabel.get(), this, mStream, readyStateToCStr(mReadyState), + readyStateToCStr(aState))); + + mReadyState = aState; +} + +void DataChannel::SendMsg(const nsACString& aMsg, ErrorResult& aRv) { + if (!EnsureValidStream(aRv)) { + return; + } + + SendErrnoToErrorResult(mConnection->SendMsg(mStream, aMsg), aMsg.Length(), + aRv); + if (!aRv.Failed()) { + IncrementBufferedAmount(aMsg.Length(), aRv); + } +} + +void DataChannel::SendBinaryMsg(const nsACString& aMsg, ErrorResult& aRv) { + if (!EnsureValidStream(aRv)) { + return; + } + + SendErrnoToErrorResult(mConnection->SendBinaryMsg(mStream, aMsg), + aMsg.Length(), aRv); + if (!aRv.Failed()) { + IncrementBufferedAmount(aMsg.Length(), aRv); + } +} + +void DataChannel::SendBinaryBlob(dom::Blob& aBlob, ErrorResult& aRv) { + if (!EnsureValidStream(aRv)) { + return; + } + + uint64_t msgLength = aBlob.GetSize(aRv); + if (aRv.Failed()) { + return; + } + + if (msgLength > UINT32_MAX) { + aRv.Throw(NS_ERROR_FILE_TOO_BIG); + return; + } + + // We convert to an nsIInputStream here, because Blob is not threadsafe, and + // we don't convert it earlier because we need to know how large this is so we + // can update bufferedAmount. + nsCOMPtr<nsIInputStream> msgStream; + aBlob.CreateInputStream(getter_AddRefs(msgStream), aRv); + if (NS_WARN_IF(aRv.Failed())) { + return; + } + + SendErrnoToErrorResult(mConnection->SendBlob(mStream, msgStream), msgLength, + aRv); + if (!aRv.Failed()) { + IncrementBufferedAmount(msgLength, aRv); + } +} + +dom::Nullable<uint16_t> DataChannel::GetMaxPacketLifeTime() const { + if (mPrPolicy == SCTP_PR_SCTP_TTL) { + return dom::Nullable<uint16_t>(mPrValue); + } + return dom::Nullable<uint16_t>(); +} + +dom::Nullable<uint16_t> DataChannel::GetMaxRetransmits() const { + if (mPrPolicy == SCTP_PR_SCTP_RTX) { + return dom::Nullable<uint16_t>(mPrValue); + } + return dom::Nullable<uint16_t>(); +} + +uint32_t DataChannel::GetBufferedAmountLowThreshold() const { + return mBufferedThreshold; +} + +// Never fire immediately, as it's defined to fire on transitions, not state +void DataChannel::SetBufferedAmountLowThreshold(uint32_t aThreshold) { + mBufferedThreshold = aThreshold; +} + +// Called with mLock locked! +void DataChannel::SendOrQueue(DataChannelOnMessageAvailable* aMessage) { + nsCOMPtr<nsIRunnable> runnable = aMessage; + mMainThreadEventTarget->Dispatch(runnable.forget()); +} + +DataChannel::TrafficCounters DataChannel::GetTrafficCounters() const { + MutexAutoLock lock(mStatsLock); + return mTrafficCounters; +} + +bool DataChannel::EnsureValidStream(ErrorResult& aRv) { + MOZ_ASSERT(mConnection); + if (mConnection && mStream != INVALID_STREAM) { + return true; + } + aRv.Throw(NS_ERROR_DOM_INVALID_STATE_ERR); + return false; +} + +void DataChannel::WithTrafficCounters( + const std::function<void(TrafficCounters&)>& aFn) { + MutexAutoLock lock(mStatsLock); + aFn(mTrafficCounters); +} + +} // namespace mozilla diff --git a/netwerk/sctp/datachannel/DataChannel.h b/netwerk/sctp/datachannel/DataChannel.h new file mode 100644 index 0000000000..f985e87a5b --- /dev/null +++ b/netwerk/sctp/datachannel/DataChannel.h @@ -0,0 +1,711 @@ +/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */ +/* vim: set ts=2 et sw=2 tw=80: */ +/* This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this file, + * You can obtain one at http://mozilla.org/MPL/2.0/. */ + +#ifndef NETWERK_SCTP_DATACHANNEL_DATACHANNEL_H_ +#define NETWERK_SCTP_DATACHANNEL_DATACHANNEL_H_ + +#ifdef MOZ_WEBRTC_SIGNALING +# define SCTP_DTLS_SUPPORTED 1 +#endif + +#include <memory> +#include <string> +#include <vector> +#include <errno.h> +#include "nsISupports.h" +#include "nsCOMPtr.h" +#include "mozilla/UniquePtr.h" +#include "mozilla/WeakPtr.h" +#include "nsString.h" +#include "nsThreadUtils.h" +#include "nsTArray.h" +#include "nsDeque.h" +#include "mozilla/dom/Blob.h" +#include "mozilla/Mutex.h" +#include "DataChannelProtocol.h" +#include "DataChannelListener.h" +#include "mozilla/net/NeckoTargetHolder.h" +#include "DataChannelLog.h" + +#ifdef SCTP_DTLS_SUPPORTED +# include "transport/sigslot.h" +# include "transport/transportlayer.h" // For TransportLayer::State +#endif + +#ifndef EALREADY +# define EALREADY WSAEALREADY +#endif + +extern "C" { +struct socket; +struct sctp_rcvinfo; +} + +namespace mozilla { + +class DataChannelConnection; +class DataChannel; +class DataChannelOnMessageAvailable; +class MediaPacket; +class MediaTransportHandler; +namespace dom { +struct RTCStatsCollection; +}; + +// For sending outgoing messages. +// This class only holds a reference to the data and the info structure but does +// not copy it. +class OutgoingMsg { + public: + OutgoingMsg(struct sctp_sendv_spa& info, const uint8_t* data, size_t length); + ~OutgoingMsg() = default; + ; + void Advance(size_t offset); + struct sctp_sendv_spa& GetInfo() { return *mInfo; }; + size_t GetLength() const { return mLength; }; + size_t GetLeft() const { return mLength - mPos; }; + const uint8_t* GetData() { return (const uint8_t*)(mData + mPos); }; + + protected: + OutgoingMsg() // Use this for inheritance only + : mLength(0), mData(nullptr), mInfo(nullptr), mPos(0){}; + size_t mLength; + const uint8_t* mData; + struct sctp_sendv_spa* mInfo; + size_t mPos; +}; + +// For queuing outgoing messages +// This class copies data of an outgoing message. +class BufferedOutgoingMsg : public OutgoingMsg { + public: + explicit BufferedOutgoingMsg(OutgoingMsg& message); + ~BufferedOutgoingMsg(); +}; + +// for queuing incoming data messages before the Open or +// external negotiation is indicated to us +class QueuedDataMessage { + public: + QueuedDataMessage(uint16_t stream, uint32_t ppid, int flags, const void* data, + uint32_t length) + : mStream(stream), mPpid(ppid), mFlags(flags), mLength(length) { + mData = static_cast<uint8_t*>(moz_xmalloc((size_t)length)); // infallible + memcpy(mData, data, (size_t)length); + } + + ~QueuedDataMessage() { free(mData); } + + uint16_t mStream; + uint32_t mPpid; + int mFlags; + uint32_t mLength; + uint8_t* mData; +}; + +// One per PeerConnection +class DataChannelConnection final : public net::NeckoTargetHolder +#ifdef SCTP_DTLS_SUPPORTED + , + public sigslot::has_slots<> +#endif +{ + friend class DataChannel; + friend class DataChannelOnMessageAvailable; + friend class DataChannelConnectRunnable; + + virtual ~DataChannelConnection(); + + public: + enum { + PENDING_NONE = 0U, // No outgoing messages are pending + PENDING_DCEP = 1U, // Outgoing DCEP messages are pending + PENDING_DATA = 2U, // Outgoing data channel messages are pending + }; + + NS_INLINE_DECL_THREADSAFE_REFCOUNTING(DataChannelConnection) + + class DataConnectionListener : public SupportsWeakPtr { + public: + virtual ~DataConnectionListener() = default; + + // Called when a new DataChannel has been opened by the other side. + virtual void NotifyDataChannel(already_AddRefed<DataChannel> channel) = 0; + + // Called when a DataChannel transitions to state open + virtual void NotifyDataChannelOpen(DataChannel* aChannel) = 0; + + // Called when a DataChannel (that was open at some point in the past) + // transitions to state closed + virtual void NotifyDataChannelClosed(DataChannel* aChannel) = 0; + + // Called when SCTP connects + virtual void NotifySctpConnected() = 0; + + // Called when SCTP closes + virtual void NotifySctpClosed() = 0; + }; + + // Create a new DataChannel Connection + // Must be called on Main thread + static Maybe<RefPtr<DataChannelConnection>> Create( + DataConnectionListener* aListener, nsISerialEventTarget* aTarget, + MediaTransportHandler* aHandler, const uint16_t aLocalPort, + const uint16_t aNumStreams, const Maybe<uint64_t>& aMaxMessageSize); + + void Destroy(); // So we can spawn refs tied to runnables in shutdown + // Finish Destroy on STS to avoid SCTP race condition with ABORT from far end + void DestroyOnSTS(struct socket* aMasterSocket, struct socket* aSocket); + void DestroyOnSTSFinal(); + + void SetMaxMessageSize(bool aMaxMessageSizeSet, uint64_t aMaxMessageSize); + uint64_t GetMaxMessageSize(); + + void AppendStatsToReport(const UniquePtr<dom::RTCStatsCollection>& aReport, + const DOMHighResTimeStamp aTimestamp) const; +#ifdef ALLOW_DIRECT_SCTP_LISTEN_CONNECT + // These block; they require something to decide on listener/connector + // (though you can do simultaneous Connect()). Do not call these from + // the main thread! + bool Listen(unsigned short port); + bool Connect(const char* addr, unsigned short port); +#endif + +#ifdef SCTP_DTLS_SUPPORTED + bool ConnectToTransport(const std::string& aTransportId, const bool aClient, + const uint16_t aLocalPort, + const uint16_t aRemotePort); + void TransportStateChange(const std::string& aTransportId, + TransportLayer::State aState); + void CompleteConnect(); + void SetSignals(const std::string& aTransportId); +#endif + + typedef enum { + RELIABLE = 0, + PARTIAL_RELIABLE_REXMIT = 1, + PARTIAL_RELIABLE_TIMED = 2 + } Type; + + [[nodiscard]] already_AddRefed<DataChannel> Open( + const nsACString& label, const nsACString& protocol, Type type, + bool inOrder, uint32_t prValue, DataChannelListener* aListener, + nsISupports* aContext, bool aExternalNegotiated, uint16_t aStream); + + void Stop(); + void Close(DataChannel* aChannel); + void CloseLocked(DataChannel* aChannel) MOZ_REQUIRES(mLock); + void CloseAll(); + + // Returns a POSIX error code. + int SendMsg(uint16_t stream, const nsACString& aMsg) { + return SendDataMsgCommon(stream, aMsg, false); + } + + // Returns a POSIX error code. + int SendBinaryMsg(uint16_t stream, const nsACString& aMsg) { + return SendDataMsgCommon(stream, aMsg, true); + } + + // Returns a POSIX error code. + int SendBlob(uint16_t stream, nsIInputStream* aBlob); + + // Called on data reception from the SCTP library + // must(?) be public so my c->c++ trampoline can call it + // May be called with (STS thread) or without the lock + int ReceiveCallback(struct socket* sock, void* data, size_t datalen, + struct sctp_rcvinfo rcv, int flags); + + // Find out state + enum { CONNECTING = 0U, OPEN = 1U, CLOSING = 2U, CLOSED = 3U }; + + Mutex mLock; + + void ReadBlob(already_AddRefed<DataChannelConnection> aThis, uint16_t aStream, + nsIInputStream* aBlob); + + bool SendDeferredMessages() MOZ_REQUIRES(mLock); + +#ifdef SCTP_DTLS_SUPPORTED + int SctpDtlsOutput(void* addr, void* buffer, size_t length, uint8_t tos, + uint8_t set_df); +#endif + + bool InShutdown() const { +#ifdef MOZ_DIAGNOSTIC_ASSERT_ENABLED + return mShutdown; +#else + return false; +#endif + } + + protected: + // Avoid cycles with PeerConnectionImpl + // Use from main thread only as WeakPtr is not threadsafe + WeakPtr<DataConnectionListener> mListener; + + private: + DataChannelConnection(DataConnectionListener* aListener, + nsISerialEventTarget* aTarget, + MediaTransportHandler* aHandler); + + bool Init(const uint16_t aLocalPort, const uint16_t aNumStreams, + const Maybe<uint64_t>& aMaxMessageSize); + + // Caller must hold mLock + uint16_t GetReadyState() const MOZ_REQUIRES(mLock) { + mLock.AssertCurrentThreadOwns(); + + return mState; + } + + // Caller must hold mLock + void SetReadyState(const uint16_t aState) MOZ_REQUIRES(mLock); + +#ifdef SCTP_DTLS_SUPPORTED + static void DTLSConnectThread(void* data); + void SendPacket(std::unique_ptr<MediaPacket>&& packet); + void SctpDtlsInput(const std::string& aTransportId, + const MediaPacket& packet); +#endif + DataChannel* FindChannelByStream(uint16_t stream) MOZ_REQUIRES(mLock); + uint16_t FindFreeStream() MOZ_REQUIRES(mLock); + bool RequestMoreStreams(int32_t aNeeded = 16) MOZ_REQUIRES(mLock); + uint32_t UpdateCurrentStreamIndex() MOZ_REQUIRES(mLock); + uint32_t GetCurrentStreamIndex() MOZ_REQUIRES(mLock); + int SendControlMessage(const uint8_t* data, uint32_t len, uint16_t stream) + MOZ_REQUIRES(mLock); + int SendOpenAckMessage(uint16_t stream) MOZ_REQUIRES(mLock); + int SendOpenRequestMessage(const nsACString& label, + const nsACString& protocol, uint16_t stream, + bool unordered, uint16_t prPolicy, + uint32_t prValue) MOZ_REQUIRES(mLock); + bool SendBufferedMessages(nsTArray<UniquePtr<BufferedOutgoingMsg>>& buffer, + size_t* aWritten); + int SendMsgInternal(OutgoingMsg& msg, size_t* aWritten); + int SendMsgInternalOrBuffer(nsTArray<UniquePtr<BufferedOutgoingMsg>>& buffer, + OutgoingMsg& msg, bool& buffered, + size_t* aWritten) MOZ_REQUIRES(mLock); + int SendDataMsgInternalOrBuffer(DataChannel& channel, const uint8_t* data, + size_t len, uint32_t ppid) + MOZ_REQUIRES(mLock); + int SendDataMsg(DataChannel& channel, const uint8_t* data, size_t len, + uint32_t ppidPartial, uint32_t ppidFinal) MOZ_REQUIRES(mLock); + int SendDataMsgCommon(uint16_t stream, const nsACString& aMsg, bool isBinary); + + void DeliverQueuedData(uint16_t stream) MOZ_REQUIRES(mLock); + + already_AddRefed<DataChannel> OpenFinish( + already_AddRefed<DataChannel>&& aChannel) MOZ_REQUIRES(mLock); + + void ProcessQueuedOpens() MOZ_REQUIRES(mLock); + void ClearResets() MOZ_REQUIRES(mLock); + void SendOutgoingStreamReset() MOZ_REQUIRES(mLock); + void ResetOutgoingStream(uint16_t stream) MOZ_REQUIRES(mLock); + void HandleOpenRequestMessage( + const struct rtcweb_datachannel_open_request* req, uint32_t length, + uint16_t stream) MOZ_REQUIRES(mLock); + void HandleOpenAckMessage(const struct rtcweb_datachannel_ack* ack, + uint32_t length, uint16_t stream); + void HandleUnknownMessage(uint32_t ppid, uint32_t length, uint16_t stream) + MOZ_REQUIRES(mLock); + uint8_t BufferMessage(nsACString& recvBuffer, const void* data, + uint32_t length, uint32_t ppid, int flags); + void HandleDataMessage(const void* buffer, size_t length, uint32_t ppid, + uint16_t stream, int flags) MOZ_REQUIRES(mLock); + void HandleDCEPMessage(const void* buffer, size_t length, uint32_t ppid, + uint16_t stream, int flags) MOZ_REQUIRES(mLock); + void HandleMessage(const void* buffer, size_t length, uint32_t ppid, + uint16_t stream, int flags) MOZ_REQUIRES(mLock); + void HandleAssociationChangeEvent(const struct sctp_assoc_change* sac) + MOZ_REQUIRES(mLock); + void HandlePeerAddressChangeEvent(const struct sctp_paddr_change* spc) + MOZ_REQUIRES(mLock); + void HandleRemoteErrorEvent(const struct sctp_remote_error* sre) + MOZ_REQUIRES(mLock); + void HandleShutdownEvent(const struct sctp_shutdown_event* sse) + MOZ_REQUIRES(mLock); + void HandleAdaptationIndication(const struct sctp_adaptation_event* sai) + MOZ_REQUIRES(mLock); + void HandlePartialDeliveryEvent(const struct sctp_pdapi_event* spde) + MOZ_REQUIRES(mLock); + void HandleSendFailedEvent(const struct sctp_send_failed_event* ssfe) + MOZ_REQUIRES(mLock); + void HandleStreamResetEvent(const struct sctp_stream_reset_event* strrst) + MOZ_REQUIRES(mLock); + void HandleStreamChangeEvent(const struct sctp_stream_change_event* strchg) + MOZ_REQUIRES(mLock); + void HandleNotification(const union sctp_notification* notif, size_t n) + MOZ_REQUIRES(mLock); + +#ifdef SCTP_DTLS_SUPPORTED + bool IsSTSThread() const { + bool on = false; + if (mSTS) { + mSTS->IsOnCurrentThread(&on); + } + return on; + } +#endif + + class Channels { + public: + Channels() : mMutex("DataChannelConnection::Channels::mMutex") {} + void Insert(const RefPtr<DataChannel>& aChannel); + bool Remove(const RefPtr<DataChannel>& aChannel); + RefPtr<DataChannel> Get(uint16_t aId) const; + typedef AutoTArray<RefPtr<DataChannel>, 16> ChannelArray; + ChannelArray GetAll() const { + MutexAutoLock lock(mMutex); + return mChannels.Clone(); + } + RefPtr<DataChannel> GetNextChannel(uint16_t aCurrentId) const; + + private: + struct IdComparator { + bool Equals(const RefPtr<DataChannel>& aChannel, uint16_t aId) const; + bool LessThan(const RefPtr<DataChannel>& aChannel, uint16_t aId) const; + bool Equals(const RefPtr<DataChannel>& a1, + const RefPtr<DataChannel>& a2) const; + bool LessThan(const RefPtr<DataChannel>& a1, + const RefPtr<DataChannel>& a2) const; + }; + mutable Mutex mMutex; + ChannelArray mChannels MOZ_GUARDED_BY(mMutex); + }; + + bool mSendInterleaved MOZ_GUARDED_BY(mLock) = false; + // MainThread only + bool mMaxMessageSizeSet = false; + // mMaxMessageSize is only set on MainThread, but read off-main-thread + uint64_t mMaxMessageSize MOZ_GUARDED_BY(mLock) = 0; + // Main thread only + Maybe<bool> mAllocateEven; + // Data: + // NOTE: while this container will auto-expand, increases in the number of + // channels available from the stack must be negotiated! + // Accessed from both main and sts, API is threadsafe + Channels mChannels; + // STS only + uint32_t mCurrentStream = 0; + nsRefPtrDeque<DataChannel> mPending; + // STS and main + size_t mNegotiatedIdLimit MOZ_GUARDED_BY(mLock) = 0; + uint8_t mPendingType MOZ_GUARDED_BY(mLock) = PENDING_NONE; + // holds data that's come in before a channel is open + nsTArray<UniquePtr<QueuedDataMessage>> mQueuedData MOZ_GUARDED_BY(mLock); + // holds outgoing control messages + nsTArray<UniquePtr<BufferedOutgoingMsg>> mBufferedControl + MOZ_GUARDED_BY(mLock); + + // Streams pending reset. Accessed from main and STS. + AutoTArray<uint16_t, 4> mStreamsResetting MOZ_GUARDED_BY(mLock); + // accessed from STS thread + struct socket* mMasterSocket = nullptr; + // cloned from mMasterSocket on successful Connect on STS thread + struct socket* mSocket = nullptr; + uint16_t mState MOZ_GUARDED_BY(mLock) = CLOSED; // Protected with mLock + +#ifdef SCTP_DTLS_SUPPORTED + std::string mTransportId; + RefPtr<MediaTransportHandler> mTransportHandler; + nsCOMPtr<nsIEventTarget> mSTS; +#endif + uint16_t mLocalPort = 0; // Accessed from connect thread + uint16_t mRemotePort = 0; + + nsCOMPtr<nsIThread> mInternalIOThread = nullptr; + nsCString mRecvBuffer; + + // Workaround to prevent a message from being received on main before the + // sender sees the decrease in bufferedAmount. + bool mDeferSend = false; + std::vector<std::unique_ptr<MediaPacket>> mDeferredSend; + +#ifdef MOZ_DIAGNOSTIC_ASSERT_ENABLED + bool mShutdown; +#endif + uintptr_t mId = 0; +}; + +#define ENSURE_DATACONNECTION \ + do { \ + MOZ_ASSERT(mConnection); \ + if (!mConnection) { \ + return; \ + } \ + } while (0) + +class DataChannel { + friend class DataChannelOnMessageAvailable; + friend class DataChannelConnection; + + public: + enum { CONNECTING = 0U, OPEN = 1U, CLOSING = 2U, CLOSED = 3U }; + + DataChannel(DataChannelConnection* connection, uint16_t stream, + uint16_t state, const nsACString& label, + const nsACString& protocol, uint16_t policy, uint32_t value, + bool ordered, bool negotiated, DataChannelListener* aListener, + nsISupports* aContext) + : mListener(aListener), + mContext(aContext), + mConnection(connection), + mLabel(label), + mProtocol(protocol), + mReadyState(state), + mStream(stream), + mPrPolicy(policy), + mPrValue(value), + mNegotiated(negotiated), + mOrdered(ordered), + mFlags(0), + mIsRecvBinary(false), + mBufferedThreshold(0), // default from spec + mBufferedAmount(0), + mMainThreadEventTarget(connection->GetNeckoTarget()), + mStatsLock("netwer::sctp::DataChannel::mStatsLock") { + NS_ASSERTION(mConnection, "NULL connection"); + } + + private: + ~DataChannel(); + + public: + NS_INLINE_DECL_THREADSAFE_REFCOUNTING(DataChannel) + + // when we disconnect from the connection after stream RESET + void StreamClosedLocked(); + + // Complete dropping of the link between DataChannel and the connection. + // After this, except for a few methods below listed to be safe, you can't + // call into DataChannel. + void ReleaseConnection(); + + // Close this DataChannel. Can be called multiple times. MUST be called + // before destroying the DataChannel (state must be CLOSED or CLOSING). + void Close(); + + // Set the listener (especially for channels created from the other side) + void SetListener(DataChannelListener* aListener, nsISupports* aContext); + + // Helper for send methods that converts POSIX error codes to an ErrorResult. + static void SendErrnoToErrorResult(int error, size_t aMessageSize, + ErrorResult& aRv); + + // Send a string + void SendMsg(const nsACString& aMsg, ErrorResult& aRv); + + // Send a binary message (TypedArray) + void SendBinaryMsg(const nsACString& aMsg, ErrorResult& aRv); + + // Send a binary blob + void SendBinaryBlob(dom::Blob& aBlob, ErrorResult& aRv); + + uint16_t GetType() const { return mPrPolicy; } + + dom::Nullable<uint16_t> GetMaxPacketLifeTime() const; + + dom::Nullable<uint16_t> GetMaxRetransmits() const; + + bool GetNegotiated() const { return mNegotiated; } + + bool GetOrdered() const { return mOrdered; } + + void IncrementBufferedAmount(uint32_t aSize, ErrorResult& aRv); + void DecrementBufferedAmount(uint32_t aSize); + + // Amount of data buffered to send + uint32_t GetBufferedAmount() const { + MOZ_ASSERT(NS_IsMainThread()); + return mBufferedAmount; + } + + // Trigger amount for generating BufferedAmountLow events + uint32_t GetBufferedAmountLowThreshold() const; + void SetBufferedAmountLowThreshold(uint32_t aThreshold); + + void AnnounceOpen(); + // TODO(bug 843625): Optionally pass an error here. + void AnnounceClosed(); + + // Find out state + uint16_t GetReadyState() const { + MOZ_ASSERT(NS_IsMainThread()); + return mReadyState; + } + + // Set ready state + void SetReadyState(const uint16_t aState); + + void GetLabel(nsAString& aLabel) { CopyUTF8toUTF16(mLabel, aLabel); } + void GetProtocol(nsAString& aProtocol) { + CopyUTF8toUTF16(mProtocol, aProtocol); + } + uint16_t GetStream() const { return mStream; } + + void SendOrQueue(DataChannelOnMessageAvailable* aMessage); + + struct TrafficCounters { + uint32_t mMessagesSent = 0; + uint64_t mBytesSent = 0; + uint32_t mMessagesReceived = 0; + uint64_t mBytesReceived = 0; + }; + + TrafficCounters GetTrafficCounters() const; + + protected: + // These are both mainthread only + DataChannelListener* mListener; + nsCOMPtr<nsISupports> mContext; + + private: + nsresult AddDataToBinaryMsg(const char* data, uint32_t size); + bool EnsureValidStream(ErrorResult& aRv); + void WithTrafficCounters(const std::function<void(TrafficCounters&)>&); + + RefPtr<DataChannelConnection> mConnection; + // mainthread only + bool mEverOpened = false; + nsCString mLabel; + nsCString mProtocol; + // This is mainthread only + uint16_t mReadyState; + uint16_t mStream; + uint16_t mPrPolicy; + uint32_t mPrValue; + // Accessed on main and STS + const bool mNegotiated; + const bool mOrdered; + uint32_t mFlags; + bool mIsRecvBinary; + size_t mBufferedThreshold; + // Read/written on main only. Decremented via message-passing, because the + // spec requires us to queue a task for this. + size_t mBufferedAmount; + nsCString mRecvBuffer; + nsTArray<UniquePtr<BufferedOutgoingMsg>> + mBufferedData; // MOZ_GUARDED_BY(mConnection->mLock) + nsCOMPtr<nsISerialEventTarget> mMainThreadEventTarget; + mutable Mutex mStatsLock; // protects mTrafficCounters + TrafficCounters mTrafficCounters MOZ_GUARDED_BY(mStatsLock); +}; + +// used to dispatch notifications of incoming data to the main thread +// Patterned on CallOnMessageAvailable in WebSockets +// Also used to proxy other items to MainThread +class DataChannelOnMessageAvailable : public Runnable { + public: + enum { + ON_CONNECTION, + ON_DISCONNECTED, + ON_CHANNEL_CREATED, + ON_DATA_STRING, + ON_DATA_BINARY, + }; /* types */ + + DataChannelOnMessageAvailable( + int32_t aType, DataChannelConnection* aConnection, DataChannel* aChannel, + nsCString& aData) // XXX this causes inefficiency + : Runnable("DataChannelOnMessageAvailable"), + mType(aType), + mChannel(aChannel), + mConnection(aConnection), + mData(aData) {} + + DataChannelOnMessageAvailable(int32_t aType, DataChannel* aChannel) + : Runnable("DataChannelOnMessageAvailable"), + mType(aType), + mChannel(aChannel) {} + // XXX is it safe to leave mData uninitialized? This should only be + // used for notifications that don't use them, but I'd like more + // bulletproof compile-time checking. + + DataChannelOnMessageAvailable(int32_t aType, + DataChannelConnection* aConnection, + DataChannel* aChannel) + : Runnable("DataChannelOnMessageAvailable"), + mType(aType), + mChannel(aChannel), + mConnection(aConnection) {} + + // for ON_CONNECTION/ON_DISCONNECTED + DataChannelOnMessageAvailable(int32_t aType, + DataChannelConnection* aConnection) + : Runnable("DataChannelOnMessageAvailable"), + mType(aType), + mConnection(aConnection) {} + + NS_IMETHOD Run() override { + MOZ_ASSERT(NS_IsMainThread()); + + // Note: calling the listeners can indirectly cause the listeners to be + // made available for GC (by removing event listeners), especially for + // OnChannelClosed(). We hold a ref to the Channel and the listener + // while calling this. + switch (mType) { + case ON_DATA_STRING: + case ON_DATA_BINARY: + if (!mChannel->mListener) { + DC_ERROR(("DataChannelOnMessageAvailable (%d) with null Listener!", + mType)); + return NS_OK; + } + + if (mChannel->GetReadyState() == DataChannel::CLOSED || + mChannel->GetReadyState() == DataChannel::CLOSING) { + // Closed by JS, probably + return NS_OK; + } + + if (mType == ON_DATA_STRING) { + mChannel->mListener->OnMessageAvailable(mChannel->mContext, mData); + } else { + mChannel->mListener->OnBinaryMessageAvailable(mChannel->mContext, + mData); + } + break; + case ON_DISCONNECTED: + // If we've disconnected, make sure we close all the streams - from + // mainthread! + if (mConnection->mListener) { + mConnection->mListener->NotifySctpClosed(); + } + mConnection->CloseAll(); + break; + case ON_CHANNEL_CREATED: + if (!mConnection->mListener) { + DC_ERROR(("DataChannelOnMessageAvailable (%d) with null Listener!", + mType)); + return NS_OK; + } + + // important to give it an already_AddRefed pointer! + mConnection->mListener->NotifyDataChannel(mChannel.forget()); + break; + case ON_CONNECTION: + if (mConnection->mListener) { + mConnection->mListener->NotifySctpConnected(); + } + break; + } + return NS_OK; + } + + private: + ~DataChannelOnMessageAvailable() = default; + + int32_t mType; + // XXX should use union + RefPtr<DataChannel> mChannel; + RefPtr<DataChannelConnection> mConnection; + nsCString mData; +}; + +} // namespace mozilla + +#endif // NETWERK_SCTP_DATACHANNEL_DATACHANNEL_H_ diff --git a/netwerk/sctp/datachannel/DataChannelListener.h b/netwerk/sctp/datachannel/DataChannelListener.h new file mode 100644 index 0000000000..06564b6cc7 --- /dev/null +++ b/netwerk/sctp/datachannel/DataChannelListener.h @@ -0,0 +1,45 @@ +/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */ +/* vim: set ts=2 et sw=2 tw=80: */ +/* This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this file, + * You can obtain one at http://mozilla.org/MPL/2.0/. */ + +#ifndef NETWERK_SCTP_DATACHANNEL_DATACHANNELLISTENER_H_ +#define NETWERK_SCTP_DATACHANNEL_DATACHANNELLISTENER_H_ + +#include "nsISupports.h" +#include "nsString.h" + +namespace mozilla { + +// Implemented by consumers of a Channel to receive messages. +// Can't nest it in DataChannelConnection because C++ doesn't allow forward +// refs to embedded classes +class DataChannelListener { + public: + virtual ~DataChannelListener() = default; + + // Called when a DOMString message is received. + virtual nsresult OnMessageAvailable(nsISupports* aContext, + const nsACString& message) = 0; + + // Called when a binary message is received. + virtual nsresult OnBinaryMessageAvailable(nsISupports* aContext, + const nsACString& message) = 0; + + // Called when the channel is connected + virtual nsresult OnChannelConnected(nsISupports* aContext) = 0; + + // Called when the channel is closed + virtual nsresult OnChannelClosed(nsISupports* aContext) = 0; + + // Called when the BufferedAmount drops below the BufferedAmountLowThreshold + virtual nsresult OnBufferLow(nsISupports* aContext) = 0; + + // Called when the BufferedAmount drops to 0 + virtual nsresult NotBuffered(nsISupports* aContext) = 0; +}; + +} // namespace mozilla + +#endif diff --git a/netwerk/sctp/datachannel/DataChannelLog.h b/netwerk/sctp/datachannel/DataChannelLog.h new file mode 100644 index 0000000000..c77fe2e2d8 --- /dev/null +++ b/netwerk/sctp/datachannel/DataChannelLog.h @@ -0,0 +1,32 @@ +/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */ +/* vim: set ts=2 et sw=2 tw=80: */ +/* This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this file, + * You can obtain one at http://mozilla.org/MPL/2.0/. */ + +#ifndef DataChannelLog_h +#define DataChannelLog_h + +#include "base/basictypes.h" +#include "mozilla/Logging.h" + +namespace mozilla { +extern mozilla::LazyLogModule gDataChannelLog; +} + +#define DC_ERROR(args) \ + MOZ_LOG(mozilla::gDataChannelLog, mozilla::LogLevel::Error, args) + +#define DC_WARN(args) \ + MOZ_LOG(mozilla::gDataChannelLog, mozilla::LogLevel::Warning, args) + +#define DC_INFO(args) \ + MOZ_LOG(mozilla::gDataChannelLog, mozilla::LogLevel::Info, args) + +#define DC_DEBUG(args) \ + MOZ_LOG(mozilla::gDataChannelLog, mozilla::LogLevel::Debug, args) + +#define DC_VERBOSE(args) \ + MOZ_LOG(mozilla::gDataChannelLog, mozilla::LogLevel::Verbose, args) + +#endif diff --git a/netwerk/sctp/datachannel/DataChannelProtocol.h b/netwerk/sctp/datachannel/DataChannelProtocol.h new file mode 100644 index 0000000000..ca44e0961d --- /dev/null +++ b/netwerk/sctp/datachannel/DataChannelProtocol.h @@ -0,0 +1,89 @@ +/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */ +/* vim: set ts=2 et sw=2 tw=80: */ +/* This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this file, + * You can obtain one at http://mozilla.org/MPL/2.0/. */ + +#ifndef NETWERK_SCTP_DATACHANNEL_DATACHANNELPROTOCOL_H_ +#define NETWERK_SCTP_DATACHANNEL_DATACHANNELPROTOCOL_H_ + +#if defined(__GNUC__) +# define SCTP_PACKED __attribute__((packed)) +#elif defined(_MSC_VER) +# pragma pack(push, 1) +# define SCTP_PACKED +#else +# error "Unsupported compiler" +#endif + +#define WEBRTC_DATACHANNEL_STREAMS_DEFAULT 256 +// Do not change this value! +#define WEBRTC_DATACHANNEL_PORT_DEFAULT 5000 +// TODO: Bug 1381146, change once we resolve the nsCString limitation +#define WEBRTC_DATACHANNEL_MAX_MESSAGE_SIZE_LOCAL 1073741823 +#define WEBRTC_DATACHANNEL_MAX_MESSAGE_SIZE_REMOTE_DEFAULT 65536 +// TODO: Bug 1382779, once resolved, can be increased to min(Uint8ArrayMaxSize, +// UINT32_MAX) +// TODO: Bug 1381146, once resolved, can be increased to whatever we support +// then (hopefully SIZE_MAX) or be removed +#define WEBRTC_DATACHANNEL_MAX_MESSAGE_SIZE_REMOTE 2147483637 + +#define DATA_CHANNEL_PPID_CONTROL 50 +#define DATA_CHANNEL_PPID_BINARY_PARTIAL 52 +#define DATA_CHANNEL_PPID_BINARY 53 +#define DATA_CHANNEL_PPID_DOMSTRING_PARTIAL 54 +#define DATA_CHANNEL_PPID_DOMSTRING 51 +#define DATA_CHANNEL_PPID_DOMSTRING_EMPTY 56 +#define DATA_CHANNEL_PPID_BINARY_EMPTY 57 + +#define DATA_CHANNEL_MAX_BINARY_FRAGMENT 0x4000 + +#define DATA_CHANNEL_FLAGS_FINISH_OPEN 0x00000004 +#define DATA_CHANNEL_FLAGS_WAITING_ACK 0x00000010 +#define DATA_CHANNEL_FLAGS_CLOSING_TOO_LARGE 0x00000020 + +#define DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_TOO_LARGE 0x01 +#define DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_BUFFERED 0x02 +#define DATA_CHANNEL_BUFFER_MESSAGE_FLAGS_COMPLETE 0x04 + +#define INVALID_STREAM (0xFFFF) +// max is 0xFFFF: Streams 0 to 0xFFFE = 0xFFFF streams +#define MAX_NUM_STREAMS (2048) + +struct rtcweb_datachannel_open_request { + uint8_t msg_type; // DATA_CHANNEL_OPEN + uint8_t channel_type; + int16_t priority; + uint32_t reliability_param; + uint16_t label_length; + uint16_t protocol_length; + char label[1]; // (and protocol) keep VC++ happy... +} SCTP_PACKED; + +struct rtcweb_datachannel_ack { + uint8_t msg_type; // DATA_CHANNEL_ACK +} SCTP_PACKED; + +/* msg_type values: */ +/* 0-1 were used in an early version of the protocol with 3-way handshakes */ +#define DATA_CHANNEL_ACK 2 +#define DATA_CHANNEL_OPEN_REQUEST 3 + +/* channel_type values: */ +#define DATA_CHANNEL_RELIABLE 0x00 +#define DATA_CHANNEL_PARTIAL_RELIABLE_REXMIT 0x01 +#define DATA_CHANNEL_PARTIAL_RELIABLE_TIMED 0x02 + +#define DATA_CHANNEL_RELIABLE_UNORDERED 0x80 +#define DATA_CHANNEL_PARTIAL_RELIABLE_REXMIT_UNORDERED 0x81 +#define DATA_CHANNEL_PARTIAL_RELIABLE_TIMED_UNORDERED 0x82 + +#define ERR_DATA_CHANNEL_ALREADY_OPEN 1 +#define ERR_DATA_CHANNEL_NONE_AVAILABLE 2 + +#if defined(_MSC_VER) +# pragma pack(pop) +# undef SCTP_PACKED +#endif + +#endif diff --git a/netwerk/sctp/datachannel/moz.build b/netwerk/sctp/datachannel/moz.build new file mode 100644 index 0000000000..94656f084e --- /dev/null +++ b/netwerk/sctp/datachannel/moz.build @@ -0,0 +1,36 @@ +# -*- Mode: python; indent-tabs-mode: nil; tab-width: 40 -*- +# vim: set filetype=python: +# This Source Code Form is subject to the terms of the Mozilla Public +# License, v. 2.0. If a copy of the MPL was not distributed with this +# file, You can obtain one at http://mozilla.org/MPL/2.0/. + +EXPORTS.mozilla.net += [ + "DataChannel.h", + "DataChannelListener.h", + "DataChannelLog.h", + "DataChannelProtocol.h", +] + +SOURCES += [ + "DataChannel.cpp", +] + +include("/ipc/chromium/chromium-config.mozbuild") + +FINAL_LIBRARY = "xul" + +LOCAL_INCLUDES += [ + "/dom/media/webrtc", + "/dom/media/webrtc/transport", + "/media/webrtc", + "/netwerk/sctp/src", + "/third_party/libwebrtc", + "/third_party/libwebrtc/third_party/abseil-cpp", +] + +DEFINES["SCTP_DEBUG"] = 1 + +if CONFIG["OS_TARGET"] == "WINNT": + DEFINES["__Userspace_os_Windows"] = 1 +else: + DEFINES["__Userspace_os_%s" % CONFIG["OS_TARGET"]] = 1 diff --git a/netwerk/sctp/src/LICENSE.md b/netwerk/sctp/src/LICENSE.md new file mode 100644 index 0000000000..a2d1f989cf --- /dev/null +++ b/netwerk/sctp/src/LICENSE.md @@ -0,0 +1,27 @@ +Copyright (c) 2015, Randall Stewart and Michael Tuexen +All rights reserved. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are met: + +* Redistributions of source code must retain the above copyright notice, this + list of conditions and the following disclaimer. + +* Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the documentation + and/or other materials provided with the distribution. + +* Neither the name of usrsctp nor the names of its + contributors may be used to endorse or promote products derived from + this software without specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE +FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. diff --git a/netwerk/sctp/src/README.md b/netwerk/sctp/src/README.md new file mode 100644 index 0000000000..3d5ba38f35 --- /dev/null +++ b/netwerk/sctp/src/README.md @@ -0,0 +1,10 @@ +# usrsctp +[](https://scan.coverity.com/projects/usrsctp) + + + +This is a userland SCTP stack supporting FreeBSD, OpenBSD, Linux, Mac OS X and Windows. + +See [manual](Manual.md) for more information. + +The status of continuous integration testing is available from our [Buildbot](http://buildbot.nplab.de:18010/#/console). diff --git a/netwerk/sctp/src/moz.build b/netwerk/sctp/src/moz.build new file mode 100644 index 0000000000..9fa9641276 --- /dev/null +++ b/netwerk/sctp/src/moz.build @@ -0,0 +1,58 @@ +# -*- Mode: python; indent-tabs-mode: nil; tab-width: 40 -*- +# vim: set filetype=python: +# This Source Code Form is subject to the terms of the Mozilla Public +# License, v. 2.0. If a copy of the MPL was not distributed with this +# file, You can obtain one at http://mozilla.org/MPL/2.0/. + +EXPORTS.mozilla.net += [ + 'usrsctp.h', +] + +UNIFIED_SOURCES += [ + 'netinet/sctp_asconf.c', + 'netinet/sctp_auth.c', + 'netinet/sctp_bsd_addr.c', + 'netinet/sctp_callout.c', + 'netinet/sctp_cc_functions.c', + 'netinet/sctp_crc32.c', + 'netinet/sctp_indata.c', + 'netinet/sctp_input.c', + 'netinet/sctp_output.c', + 'netinet/sctp_pcb.c', + 'netinet/sctp_peeloff.c', + 'netinet/sctp_sha1.c', + 'netinet/sctp_ss_functions.c', + 'netinet/sctp_sysctl.c', + 'netinet/sctp_timer.c', + 'netinet/sctp_userspace.c', + 'netinet/sctp_usrreq.c', + 'netinet/sctputil.c', + 'netinet6/sctp6_usrreq.c', + 'user_environment.c', + 'user_mbuf.c', + 'user_recv_thread.c', + 'user_socket.c', +] + +Library('nksctp_s') + +# We allow warnings for third-party code that can be updated from upstream. +AllowCompilerWarnings() + +FINAL_LIBRARY = 'xul' + +LOCAL_INCLUDES += [ + '/dom/base', + '/netwerk/base', +] + +for var in ('__Userspace__', + 'SCTP_SIMPLE_ALLOCATOR', + 'SCTP_PROCESS_LEVEL_LOCKS', + 'SCTP_DEBUG'): + DEFINES[var] = 1 + +if CONFIG['OS_TARGET'] in ('Linux', 'Android'): + DEFINES['_GNU_SOURCE'] = 1 +elif CONFIG['OS_TARGET'] == 'Darwin': + DEFINES['__APPLE_USE_RFC_2292'] = 1 diff --git a/netwerk/sctp/src/moz.yaml b/netwerk/sctp/src/moz.yaml new file mode 100644 index 0000000000..4e1008665d --- /dev/null +++ b/netwerk/sctp/src/moz.yaml @@ -0,0 +1,55 @@ +schema: 1 + +bugzilla: + product: Core + component: Networking + +origin: + name: sctp + description: userland SCTP stack + + url: https://github.com/sctplab/usrsctp/blob/master/Manual.md + + release: f9f95023816b61a2f257d2fb77658dceaea7213f (2023-03-18T22:38:54Z). + revision: f9f95023816b61a2f257d2fb77658dceaea7213f + + license: BSD-3-Clause + +vendoring: + url: https://github.com/sctplab/usrsctp + source-hosting: github + tracking: commit + + exclude: + - "meson*" + - "Makefile.*" + - "Manual.*" + - "CMake*" + - ".*" + - "*.py" + - "*.in" + - "*.ac" + - bootstrap + - cmake/ + - fuzzer/ + - programs/ + - 'usrsctplib/Makefile.*' + - 'usrsctplib/meson.*' + - 'usrsctplib/CMake*' + - 'usrsctplib/netinet*/meson.*' + + keep: + - moz.build + - restore_mod.sh + + update-actions: + - action: move-dir + from: '{vendor_dir}/usrsctplib' + to: '{vendor_dir}' + - action: run-script + script: '{yaml_dir}/restore_mod.sh' + cwd: '{yaml_dir}' + + patches: + - win32-rands.patch + - win32-free.patch diff --git a/netwerk/sctp/src/netinet/sctp.h b/netwerk/sctp/src/netinet/sctp.h new file mode 100644 index 0000000000..203f6f1f6a --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp.h @@ -0,0 +1,667 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_H_ +#define _NETINET_SCTP_H_ + +#if defined(__APPLE__) || defined(__linux__) +#include <stdint.h> +#endif +#include <sys/types.h> + +#if !defined(_WIN32) +#define SCTP_PACKED __attribute__((packed)) +#else +#pragma pack (push, 1) +#define SCTP_PACKED +#endif + +/* + * SCTP protocol - RFC4960. + */ +struct sctphdr { + uint16_t src_port; /* source port */ + uint16_t dest_port; /* destination port */ + uint32_t v_tag; /* verification tag of packet */ + uint32_t checksum; /* CRC32C checksum */ + /* chunks follow... */ +} SCTP_PACKED; + +/* + * SCTP Chunks + */ +struct sctp_chunkhdr { + uint8_t chunk_type; /* chunk type */ + uint8_t chunk_flags; /* chunk flags */ + uint16_t chunk_length; /* chunk length */ + /* optional params follow */ +} SCTP_PACKED; + +/* + * SCTP chunk parameters + */ +struct sctp_paramhdr { + uint16_t param_type; /* parameter type */ + uint16_t param_length; /* parameter length */ +} SCTP_PACKED; + +/* + * user socket options: socket API defined + */ +/* + * read-write options + */ +#define SCTP_RTOINFO 0x00000001 +#define SCTP_ASSOCINFO 0x00000002 +#define SCTP_INITMSG 0x00000003 +#define SCTP_NODELAY 0x00000004 +#define SCTP_AUTOCLOSE 0x00000005 +#define SCTP_SET_PEER_PRIMARY_ADDR 0x00000006 +#define SCTP_PRIMARY_ADDR 0x00000007 +#define SCTP_ADAPTATION_LAYER 0x00000008 +/* same as above */ +#define SCTP_ADAPTION_LAYER 0x00000008 +#define SCTP_DISABLE_FRAGMENTS 0x00000009 +#define SCTP_PEER_ADDR_PARAMS 0x0000000a +#define SCTP_DEFAULT_SEND_PARAM 0x0000000b +/* ancillary data/notification interest options */ +#define SCTP_EVENTS 0x0000000c /* deprecated */ +/* Without this applied we will give V4 and V6 addresses on a V6 socket */ +#define SCTP_I_WANT_MAPPED_V4_ADDR 0x0000000d +#define SCTP_MAXSEG 0x0000000e +#define SCTP_DELAYED_SACK 0x0000000f +#define SCTP_FRAGMENT_INTERLEAVE 0x00000010 +#define SCTP_PARTIAL_DELIVERY_POINT 0x00000011 +/* authentication support */ +#define SCTP_AUTH_CHUNK 0x00000012 +#define SCTP_AUTH_KEY 0x00000013 +#define SCTP_HMAC_IDENT 0x00000014 +#define SCTP_AUTH_ACTIVE_KEY 0x00000015 +#define SCTP_AUTH_DELETE_KEY 0x00000016 +#define SCTP_USE_EXT_RCVINFO 0x00000017 +#define SCTP_AUTO_ASCONF 0x00000018 /* rw */ +#define SCTP_MAXBURST 0x00000019 /* rw */ +#define SCTP_MAX_BURST 0x00000019 /* rw */ +/* assoc level context */ +#define SCTP_CONTEXT 0x0000001a /* rw */ +/* explicit EOR signalling */ +#define SCTP_EXPLICIT_EOR 0x0000001b +#define SCTP_REUSE_PORT 0x0000001c /* rw */ +#define SCTP_AUTH_DEACTIVATE_KEY 0x0000001d +#define SCTP_EVENT 0x0000001e +#define SCTP_RECVRCVINFO 0x0000001f +#define SCTP_RECVNXTINFO 0x00000020 +#define SCTP_DEFAULT_SNDINFO 0x00000021 +#define SCTP_DEFAULT_PRINFO 0x00000022 +#define SCTP_PEER_ADDR_THLDS 0x00000023 +#define SCTP_REMOTE_UDP_ENCAPS_PORT 0x00000024 +#define SCTP_ECN_SUPPORTED 0x00000025 +#define SCTP_PR_SUPPORTED 0x00000026 +#define SCTP_AUTH_SUPPORTED 0x00000027 +#define SCTP_ASCONF_SUPPORTED 0x00000028 +#define SCTP_RECONFIG_SUPPORTED 0x00000029 +#define SCTP_NRSACK_SUPPORTED 0x00000030 +#define SCTP_PKTDROP_SUPPORTED 0x00000031 +#define SCTP_MAX_CWND 0x00000032 +#define SCTP_ACCEPT_ZERO_CHECKSUM 0x00000033 + +/* + * read-only options + */ +#define SCTP_STATUS 0x00000100 +#define SCTP_GET_PEER_ADDR_INFO 0x00000101 +/* authentication support */ +#define SCTP_PEER_AUTH_CHUNKS 0x00000102 +#define SCTP_LOCAL_AUTH_CHUNKS 0x00000103 +#define SCTP_GET_ASSOC_NUMBER 0x00000104 /* ro */ +#define SCTP_GET_ASSOC_ID_LIST 0x00000105 /* ro */ +#define SCTP_TIMEOUTS 0x00000106 +#define SCTP_PR_STREAM_STATUS 0x00000107 +#define SCTP_PR_ASSOC_STATUS 0x00000108 + +/* + * user socket options: BSD implementation specific + */ +/* + * Blocking I/O is enabled on any TCP type socket by default. For the UDP + * model if this is turned on then the socket buffer is shared for send + * resources amongst all associations. The default for the UDP model is that + * is SS_NBIO is set. Which means all associations have a separate send + * limit BUT they will NOT ever BLOCK instead you will get an error back + * EAGAIN if you try to send too much. If you want the blocking semantics you + * set this option at the cost of sharing one socket send buffer size amongst + * all associations. Peeled off sockets turn this option off and block. But + * since both TCP and peeled off sockets have only one assoc per socket this + * is fine. It probably does NOT make sense to set this on SS_NBIO on a TCP + * model OR peeled off UDP model, but we do allow you to do so. You just use + * the normal syscall to toggle SS_NBIO the way you want. + * + * Blocking I/O is controlled by the SS_NBIO flag on the socket state so_state + * field. + */ + +#define SCTP_ENABLE_STREAM_RESET 0x00000900 /* struct sctp_assoc_value */ +#define SCTP_RESET_STREAMS 0x00000901 /* struct sctp_reset_streams */ +#define SCTP_RESET_ASSOC 0x00000902 /* sctp_assoc_t */ +#define SCTP_ADD_STREAMS 0x00000903 /* struct sctp_add_streams */ + +/* For enable stream reset */ +#define SCTP_ENABLE_RESET_STREAM_REQ 0x00000001 +#define SCTP_ENABLE_RESET_ASSOC_REQ 0x00000002 +#define SCTP_ENABLE_CHANGE_ASSOC_REQ 0x00000004 +#define SCTP_ENABLE_VALUE_MASK 0x00000007 +/* For reset streams */ +#define SCTP_STREAM_RESET_INCOMING 0x00000001 +#define SCTP_STREAM_RESET_OUTGOING 0x00000002 + +/* here on down are more implementation specific */ +#define SCTP_SET_DEBUG_LEVEL 0x00001005 +#define SCTP_CLR_STAT_LOG 0x00001007 +/* CMT ON/OFF socket option */ +#define SCTP_CMT_ON_OFF 0x00001200 +#define SCTP_CMT_USE_DAC 0x00001201 +/* JRS - Pluggable Congestion Control Socket option */ +#define SCTP_PLUGGABLE_CC 0x00001202 +/* RS - Pluggable Stream Scheduling Socket option */ +#define SCTP_PLUGGABLE_SS 0x00001203 +#define SCTP_SS_VALUE 0x00001204 +#define SCTP_CC_OPTION 0x00001205 /* Options for CC modules */ +/* For I-DATA */ +#define SCTP_INTERLEAVING_SUPPORTED 0x00001206 + +/* read only */ +#define SCTP_GET_SNDBUF_USE 0x00001101 +#define SCTP_GET_STAT_LOG 0x00001103 +#define SCTP_PCB_STATUS 0x00001104 +#define SCTP_GET_NONCE_VALUES 0x00001105 + +/* Special hook for dynamically setting primary for all assoc's, + * this is a write only option that requires root privilege. + */ +#define SCTP_SET_DYNAMIC_PRIMARY 0x00002001 + +/* VRF (virtual router feature) and multi-VRF support + * options. VRF's provide splits within a router + * that give the views of multiple routers. A + * standard host, without VRF support, is just + * a single VRF. If VRF's are supported then + * the transport must be VRF aware. This means + * that every socket call coming in must be directed + * within the endpoint to one of the VRF's it belongs + * to. The endpoint, before binding, may select + * the "default" VRF it is in by using a set socket + * option with SCTP_VRF_ID. This will also + * get propagated to the default VRF. Once the + * endpoint binds an address then it CANNOT add + * additional VRF's to become a Multi-VRF endpoint. + * + * Before BINDING additional VRF's can be added with + * the SCTP_ADD_VRF_ID call or deleted with + * SCTP_DEL_VRF_ID. + * + * Associations are ALWAYS contained inside a single + * VRF. They cannot reside in two (or more) VRF's. Incoming + * packets, assuming the router is VRF aware, can always + * tell us what VRF they arrived on. A host not supporting + * any VRF's will find that the packets always arrived on the + * single VRF that the host has. + * + */ + +#define SCTP_VRF_ID 0x00003001 +#define SCTP_ADD_VRF_ID 0x00003002 +#define SCTP_GET_VRF_IDS 0x00003003 +#define SCTP_GET_ASOC_VRF 0x00003004 +#define SCTP_DEL_VRF_ID 0x00003005 + +/* + * If you enable packet logging you can get + * a poor mans ethereal output in binary + * form. Note this is a compile option to + * the kernel, SCTP_PACKET_LOGGING, and + * without it in your kernel you + * will get a EOPNOTSUPP + */ +#define SCTP_GET_PACKET_LOG 0x00004001 + +/* + * hidden implementation specific options these are NOT user visible (should + * move out of sctp.h) + */ +/* sctp_bindx() flags as hidden socket options */ +#define SCTP_BINDX_ADD_ADDR 0x00008001 +#define SCTP_BINDX_REM_ADDR 0x00008002 +/* Hidden socket option that gets the addresses */ +#define SCTP_GET_PEER_ADDRESSES 0x00008003 +#define SCTP_GET_LOCAL_ADDRESSES 0x00008004 +/* return the total count in bytes needed to hold all local addresses bound */ +#define SCTP_GET_LOCAL_ADDR_SIZE 0x00008005 +/* Return the total count in bytes needed to hold the remote address */ +#define SCTP_GET_REMOTE_ADDR_SIZE 0x00008006 +/* hidden option for connectx */ +#define SCTP_CONNECT_X 0x00008007 +/* hidden option for connectx_delayed, part of sendx */ +#define SCTP_CONNECT_X_DELAYED 0x00008008 +#define SCTP_CONNECT_X_COMPLETE 0x00008009 +/* hidden socket option based sctp_peeloff */ +#define SCTP_PEELOFF 0x0000800a +/* the real worker for sctp_getaddrlen() */ +#define SCTP_GET_ADDR_LEN 0x0000800b +#if defined(__APPLE__) && !defined(__Userspace__) +/* temporary workaround for Apple listen() issue, no args used */ +#define SCTP_LISTEN_FIX 0x0000800c +#endif +#if defined(_WIN32) && !defined(__Userspace__) +/* workaround for Cygwin on Windows: returns the SOCKET handle */ +#define SCTP_GET_HANDLE 0x0000800d +#endif +/* Debug things that need to be purged */ +#define SCTP_SET_INITIAL_DBG_SEQ 0x00009f00 + +/* JRS - Supported congestion control modules for pluggable + * congestion control + */ +/* Standard TCP Congestion Control */ +#define SCTP_CC_RFC2581 0x00000000 +/* High Speed TCP Congestion Control (Floyd) */ +#define SCTP_CC_HSTCP 0x00000001 +/* HTCP Congestion Control */ +#define SCTP_CC_HTCP 0x00000002 +/* RTCC Congestion Control - RFC2581 plus */ +#define SCTP_CC_RTCC 0x00000003 + +#define SCTP_CC_OPT_RTCC_SETMODE 0x00002000 +#define SCTP_CC_OPT_USE_DCCC_ECN 0x00002001 +#define SCTP_CC_OPT_STEADY_STEP 0x00002002 + +#define SCTP_CMT_OFF 0 +#define SCTP_CMT_BASE 1 +#define SCTP_CMT_RPV1 2 +#define SCTP_CMT_RPV2 3 +#define SCTP_CMT_MPTCP 4 +#define SCTP_CMT_MAX SCTP_CMT_MPTCP + +/* RS - Supported stream scheduling modules for pluggable + * stream scheduling + */ +/* Default simple round-robin */ +#define SCTP_SS_DEFAULT 0x00000000 +/* Real round-robin */ +#define SCTP_SS_ROUND_ROBIN 0x00000001 +/* Real round-robin per packet */ +#define SCTP_SS_ROUND_ROBIN_PACKET 0x00000002 +/* Priority */ +#define SCTP_SS_PRIORITY 0x00000003 +/* Fair Bandwidth */ +#define SCTP_SS_FAIR_BANDWITH 0x00000004 +/* First-come, first-serve */ +#define SCTP_SS_FIRST_COME 0x00000005 + +/* fragment interleave constants + * setting must be one of these or + * EINVAL returned. + */ +#define SCTP_FRAG_LEVEL_0 0x00000000 +#define SCTP_FRAG_LEVEL_1 0x00000001 +#define SCTP_FRAG_LEVEL_2 0x00000002 + +/* + * user state values + */ +#define SCTP_CLOSED 0x0000 +#define SCTP_BOUND 0x1000 +#define SCTP_LISTEN 0x2000 +#define SCTP_COOKIE_WAIT 0x0002 +#define SCTP_COOKIE_ECHOED 0x0004 +#define SCTP_ESTABLISHED 0x0008 +#define SCTP_SHUTDOWN_SENT 0x0010 +#define SCTP_SHUTDOWN_RECEIVED 0x0020 +#define SCTP_SHUTDOWN_ACK_SENT 0x0040 +#define SCTP_SHUTDOWN_PENDING 0x0080 + +/* + * SCTP operational error codes (user visible) + */ +#define SCTP_CAUSE_NO_ERROR 0x0000 +#define SCTP_CAUSE_INVALID_STREAM 0x0001 +#define SCTP_CAUSE_MISSING_PARAM 0x0002 +#define SCTP_CAUSE_STALE_COOKIE 0x0003 +#define SCTP_CAUSE_OUT_OF_RESC 0x0004 +#define SCTP_CAUSE_UNRESOLVABLE_ADDR 0x0005 +#define SCTP_CAUSE_UNRECOG_CHUNK 0x0006 +#define SCTP_CAUSE_INVALID_PARAM 0x0007 +#define SCTP_CAUSE_UNRECOG_PARAM 0x0008 +#define SCTP_CAUSE_NO_USER_DATA 0x0009 +#define SCTP_CAUSE_COOKIE_IN_SHUTDOWN 0x000a +#define SCTP_CAUSE_RESTART_W_NEWADDR 0x000b +#define SCTP_CAUSE_USER_INITIATED_ABT 0x000c +#define SCTP_CAUSE_PROTOCOL_VIOLATION 0x000d + +/* Error causes from RFC5061 */ +#define SCTP_CAUSE_DELETING_LAST_ADDR 0x00a0 +#define SCTP_CAUSE_RESOURCE_SHORTAGE 0x00a1 +#define SCTP_CAUSE_DELETING_SRC_ADDR 0x00a2 +#define SCTP_CAUSE_ILLEGAL_ASCONF_ACK 0x00a3 +#define SCTP_CAUSE_REQUEST_REFUSED 0x00a4 + +/* Error causes from nat-draft */ +#define SCTP_CAUSE_NAT_COLLIDING_STATE 0x00b0 +#define SCTP_CAUSE_NAT_MISSING_STATE 0x00b1 + +/* Error causes from RFC4895 */ +#define SCTP_CAUSE_UNSUPPORTED_HMACID 0x0105 + +/* + * error cause parameters (user visible) + */ +struct sctp_gen_error_cause { + uint16_t code; + uint16_t length; + uint8_t info[]; +} SCTP_PACKED; + +struct sctp_error_cause { + uint16_t code; + uint16_t length; + /* optional cause-specific info may follow */ +} SCTP_PACKED; + +struct sctp_error_invalid_stream { + struct sctp_error_cause cause; /* code=SCTP_CAUSE_INVALID_STREAM */ + uint16_t stream_id; /* stream id of the DATA in error */ + uint16_t reserved; +} SCTP_PACKED; + +struct sctp_error_missing_param { + struct sctp_error_cause cause; /* code=SCTP_CAUSE_MISSING_PARAM */ + uint32_t num_missing_params; /* number of missing parameters */ + uint16_t type[]; +} SCTP_PACKED; + +struct sctp_error_stale_cookie { + struct sctp_error_cause cause; /* code=SCTP_CAUSE_STALE_COOKIE */ + uint32_t stale_time; /* time in usec of staleness */ +} SCTP_PACKED; + +struct sctp_error_out_of_resource { + struct sctp_error_cause cause; /* code=SCTP_CAUSE_OUT_OF_RESOURCES */ +} SCTP_PACKED; + +struct sctp_error_unresolv_addr { + struct sctp_error_cause cause; /* code=SCTP_CAUSE_UNRESOLVABLE_ADDR */ +} SCTP_PACKED; + +struct sctp_error_unrecognized_chunk { + struct sctp_error_cause cause; /* code=SCTP_CAUSE_UNRECOG_CHUNK */ + struct sctp_chunkhdr ch;/* header from chunk in error */ +} SCTP_PACKED; + +struct sctp_error_no_user_data { + struct sctp_error_cause cause; /* code=SCTP_CAUSE_NO_USER_DATA */ + uint32_t tsn; /* TSN of the empty data chunk */ +} SCTP_PACKED; + +struct sctp_error_auth_invalid_hmac { + struct sctp_error_cause cause; /* code=SCTP_CAUSE_UNSUPPORTED_HMACID */ + uint16_t hmac_id; +} SCTP_PACKED; + +/* + * Main SCTP chunk types we place these here so natd and f/w's in user land + * can find them. + */ +/************0x00 series ***********/ +#define SCTP_DATA 0x00 +#define SCTP_INITIATION 0x01 +#define SCTP_INITIATION_ACK 0x02 +#define SCTP_SELECTIVE_ACK 0x03 +#define SCTP_HEARTBEAT_REQUEST 0x04 +#define SCTP_HEARTBEAT_ACK 0x05 +#define SCTP_ABORT_ASSOCIATION 0x06 +#define SCTP_SHUTDOWN 0x07 +#define SCTP_SHUTDOWN_ACK 0x08 +#define SCTP_OPERATION_ERROR 0x09 +#define SCTP_COOKIE_ECHO 0x0a +#define SCTP_COOKIE_ACK 0x0b +#define SCTP_ECN_ECHO 0x0c +#define SCTP_ECN_CWR 0x0d +#define SCTP_SHUTDOWN_COMPLETE 0x0e +/* RFC4895 */ +#define SCTP_AUTHENTICATION 0x0f +/* EY nr_sack chunk id*/ +#define SCTP_NR_SELECTIVE_ACK 0x10 +/************0x40 series ***********/ +#define SCTP_IDATA 0x40 +/************0x80 series ***********/ +/* RFC5061 */ +#define SCTP_ASCONF_ACK 0x80 +/* draft-ietf-stewart-pktdrpsctp */ +#define SCTP_PACKET_DROPPED 0x81 +/* draft-ietf-stewart-strreset-xxx */ +#define SCTP_STREAM_RESET 0x82 + +/* RFC4820 */ +#define SCTP_PAD_CHUNK 0x84 +/************0xc0 series ***********/ +/* RFC3758 */ +#define SCTP_FORWARD_CUM_TSN 0xc0 +/* RFC5061 */ +#define SCTP_ASCONF 0xc1 +#define SCTP_IFORWARD_CUM_TSN 0xc2 + +/* ABORT and SHUTDOWN COMPLETE FLAG */ +#define SCTP_HAD_NO_TCB 0x01 + +/* Packet dropped flags */ +#define SCTP_FROM_MIDDLE_BOX SCTP_HAD_NO_TCB +#define SCTP_BADCRC 0x02 +#define SCTP_PACKET_TRUNCATED 0x04 + +/* Flag for ECN -CWR */ +#define SCTP_CWR_REDUCE_OVERRIDE 0x01 +#define SCTP_CWR_IN_SAME_WINDOW 0x02 + +#define SCTP_SAT_NETWORK_MIN 400 /* min ms for RTT to set satellite + * time */ +#define SCTP_SAT_NETWORK_BURST_INCR 2 /* how many times to multiply maxburst + * in sat */ + +/* Data Chuck Specific Flags */ +#define SCTP_DATA_FRAG_MASK 0x03 +#define SCTP_DATA_MIDDLE_FRAG 0x00 +#define SCTP_DATA_LAST_FRAG 0x01 +#define SCTP_DATA_FIRST_FRAG 0x02 +#define SCTP_DATA_NOT_FRAG 0x03 +#define SCTP_DATA_UNORDERED 0x04 +#define SCTP_DATA_SACK_IMMEDIATELY 0x08 +/* ECN Nonce: SACK Chunk Specific Flags */ +#define SCTP_SACK_NONCE_SUM 0x01 + +/* CMT DAC algorithm SACK flag */ +#define SCTP_SACK_CMT_DAC 0x80 + +/* + * PCB flags (in sctp_flags bitmask). + * Note the features and flags are meant + * for use by netstat. + */ +#define SCTP_PCB_FLAGS_UDPTYPE 0x00000001 +#define SCTP_PCB_FLAGS_TCPTYPE 0x00000002 +#define SCTP_PCB_FLAGS_BOUNDALL 0x00000004 +#define SCTP_PCB_FLAGS_ACCEPTING 0x00000008 +#define SCTP_PCB_FLAGS_UNBOUND 0x00000010 +#define SCTP_PCB_FLAGS_SND_ITERATOR_UP 0x00000020 +#define SCTP_PCB_FLAGS_CLOSE_IP 0x00040000 +#define SCTP_PCB_FLAGS_WAS_CONNECTED 0x00080000 +#define SCTP_PCB_FLAGS_WAS_ABORTED 0x00100000 +/* TCP model support */ + +#define SCTP_PCB_FLAGS_CONNECTED 0x00200000 +#define SCTP_PCB_FLAGS_IN_TCPPOOL 0x00400000 +#define SCTP_PCB_FLAGS_DONT_WAKE 0x00800000 +#define SCTP_PCB_FLAGS_WAKEOUTPUT 0x01000000 +#define SCTP_PCB_FLAGS_WAKEINPUT 0x02000000 +#define SCTP_PCB_FLAGS_BOUND_V6 0x04000000 +#define SCTP_PCB_FLAGS_BLOCKING_IO 0x08000000 +#define SCTP_PCB_FLAGS_SOCKET_GONE 0x10000000 +#define SCTP_PCB_FLAGS_SOCKET_ALLGONE 0x20000000 +#define SCTP_PCB_FLAGS_SOCKET_CANT_READ 0x40000000 +#if defined(__Userspace__) +#define SCTP_PCB_FLAGS_BOUND_CONN 0x80000000 + +/* flags to copy to new PCB */ +#define SCTP_PCB_COPY_FLAGS (SCTP_PCB_FLAGS_BOUNDALL|\ + SCTP_PCB_FLAGS_WAKEINPUT|\ + SCTP_PCB_FLAGS_BOUND_V6|\ + SCTP_PCB_FLAGS_BOUND_CONN) +#else + +/* flags to copy to new PCB */ +#define SCTP_PCB_COPY_FLAGS (SCTP_PCB_FLAGS_BOUNDALL|\ + SCTP_PCB_FLAGS_WAKEINPUT|\ + SCTP_PCB_FLAGS_BOUND_V6) +#endif + +/* + * PCB Features (in sctp_features bitmask) + */ +#define SCTP_PCB_FLAGS_DO_NOT_PMTUD 0x0000000000000001 +#define SCTP_PCB_FLAGS_EXT_RCVINFO 0x0000000000000002 /* deprecated */ +#define SCTP_PCB_FLAGS_DONOT_HEARTBEAT 0x0000000000000004 +#define SCTP_PCB_FLAGS_FRAG_INTERLEAVE 0x0000000000000008 +#define SCTP_PCB_FLAGS_INTERLEAVE_STRMS 0x0000000000000010 +#define SCTP_PCB_FLAGS_DO_ASCONF 0x0000000000000020 +#define SCTP_PCB_FLAGS_AUTO_ASCONF 0x0000000000000040 +/* socket options */ +#define SCTP_PCB_FLAGS_NODELAY 0x0000000000000100 +#define SCTP_PCB_FLAGS_AUTOCLOSE 0x0000000000000200 +#define SCTP_PCB_FLAGS_RECVDATAIOEVNT 0x0000000000000400 /* deprecated */ +#define SCTP_PCB_FLAGS_RECVASSOCEVNT 0x0000000000000800 +#define SCTP_PCB_FLAGS_RECVPADDREVNT 0x0000000000001000 +#define SCTP_PCB_FLAGS_RECVPEERERR 0x0000000000002000 +#define SCTP_PCB_FLAGS_RECVSENDFAILEVNT 0x0000000000004000 /* deprecated */ +#define SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT 0x0000000000008000 +#define SCTP_PCB_FLAGS_ADAPTATIONEVNT 0x0000000000010000 +#define SCTP_PCB_FLAGS_PDAPIEVNT 0x0000000000020000 +#define SCTP_PCB_FLAGS_AUTHEVNT 0x0000000000040000 +#define SCTP_PCB_FLAGS_STREAM_RESETEVNT 0x0000000000080000 +#define SCTP_PCB_FLAGS_NO_FRAGMENT 0x0000000000100000 +#define SCTP_PCB_FLAGS_EXPLICIT_EOR 0x0000000000400000 +#define SCTP_PCB_FLAGS_NEEDS_MAPPED_V4 0x0000000000800000 +#define SCTP_PCB_FLAGS_MULTIPLE_ASCONFS 0x0000000001000000 +#define SCTP_PCB_FLAGS_PORTREUSE 0x0000000002000000 +#define SCTP_PCB_FLAGS_DRYEVNT 0x0000000004000000 +#define SCTP_PCB_FLAGS_RECVRCVINFO 0x0000000008000000 +#define SCTP_PCB_FLAGS_RECVNXTINFO 0x0000000010000000 +#define SCTP_PCB_FLAGS_ASSOC_RESETEVNT 0x0000000020000000 +#define SCTP_PCB_FLAGS_STREAM_CHANGEEVNT 0x0000000040000000 +#define SCTP_PCB_FLAGS_RECVNSENDFAILEVNT 0x0000000080000000 + +/*- + * mobility_features parameters (by micchie).Note + * these features are applied against the + * sctp_mobility_features flags.. not the sctp_features + * flags. + */ +#define SCTP_MOBILITY_BASE 0x00000001 +#define SCTP_MOBILITY_FASTHANDOFF 0x00000002 +#define SCTP_MOBILITY_PRIM_DELETED 0x00000004 + +/* Smallest PMTU allowed when disabling PMTU discovery */ +#define SCTP_SMALLEST_PMTU 512 +/* Largest PMTU allowed when disabling PMTU discovery */ +#define SCTP_LARGEST_PMTU 65536 + +#if defined(_WIN32) +#pragma pack(pop) +#endif +#undef SCTP_PACKED + +#include <netinet/sctp_uio.h> + +/* This dictates the size of the packet + * collection buffer. This only applies + * if SCTP_PACKET_LOGGING is enabled in + * your config. + */ +#define SCTP_PACKET_LOG_SIZE 65536 + +/* Maximum delays and such a user can set for options that + * take ms. + */ +#define SCTP_MAX_SACK_DELAY 500 /* per RFC4960 */ +#define SCTP_MAX_HB_INTERVAL 14400000 /* 4 hours in ms */ +#define SCTP_MIN_COOKIE_LIFE 1000 /* 1 second in ms */ +#define SCTP_MAX_COOKIE_LIFE 3600000 /* 1 hour in ms */ + +/* Types of logging/KTR tracing that can be enabled via the + * sysctl net.inet.sctp.sctp_logging. You must also enable + * SUBSYS tracing. + * Note that you must have the SCTP option in the kernel + * to enable these as well. + */ +#define SCTP_BLK_LOGGING_ENABLE 0x00000001 +#define SCTP_CWND_MONITOR_ENABLE 0x00000002 +#define SCTP_CWND_LOGGING_ENABLE 0x00000004 +#define SCTP_FLIGHT_LOGGING_ENABLE 0x00000020 +#define SCTP_FR_LOGGING_ENABLE 0x00000040 +#define SCTP_LOCK_LOGGING_ENABLE 0x00000080 +#define SCTP_MAP_LOGGING_ENABLE 0x00000100 +#define SCTP_MBCNT_LOGGING_ENABLE 0x00000200 +#define SCTP_MBUF_LOGGING_ENABLE 0x00000400 +#define SCTP_NAGLE_LOGGING_ENABLE 0x00000800 +#define SCTP_RECV_RWND_LOGGING_ENABLE 0x00001000 +#define SCTP_RTTVAR_LOGGING_ENABLE 0x00002000 +#define SCTP_SACK_LOGGING_ENABLE 0x00004000 +#define SCTP_SACK_RWND_LOGGING_ENABLE 0x00008000 +#define SCTP_SB_LOGGING_ENABLE 0x00010000 +#define SCTP_STR_LOGGING_ENABLE 0x00020000 +#define SCTP_WAKE_LOGGING_ENABLE 0x00040000 +#define SCTP_LOG_MAXBURST_ENABLE 0x00080000 +#define SCTP_LOG_RWND_ENABLE 0x00100000 +#define SCTP_LOG_SACK_ARRIVALS_ENABLE 0x00200000 +#define SCTP_LTRACE_CHUNK_ENABLE 0x00400000 +#define SCTP_LTRACE_ERROR_ENABLE 0x00800000 +#define SCTP_LAST_PACKET_TRACING 0x01000000 +#define SCTP_THRESHOLD_LOGGING 0x02000000 +#define SCTP_LOG_AT_SEND_2_SCTP 0x04000000 +#define SCTP_LOG_AT_SEND_2_OUTQ 0x08000000 +#define SCTP_LOG_TRY_ADVANCE 0x10000000 + +#endif /* !_NETINET_SCTP_H_ */ diff --git a/netwerk/sctp/src/netinet/sctp_asconf.c b/netwerk/sctp/src/netinet/sctp_asconf.c new file mode 100644 index 0000000000..6c5d296893 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_asconf.c @@ -0,0 +1,3539 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(_WIN32) +// Needed for unified build so that rand_s is available to all unified +// sources. +#if !defined(_CRT_RAND_S) && !defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) +#define _CRT_RAND_S +#endif +#endif + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#include <netinet/sctp_var.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctp_header.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_asconf.h> +#include <netinet/sctp_timer.h> + +/* + * debug flags: + * SCTP_DEBUG_ASCONF1: protocol info, general info and errors + * SCTP_DEBUG_ASCONF2: detailed info + */ + +/* + * RFC 5061 + * + * An ASCONF parameter queue exists per asoc which holds the pending address + * operations. Lists are updated upon receipt of ASCONF-ACK. + * + * A restricted_addrs list exists per assoc to hold local addresses that are + * not (yet) usable by the assoc as a source address. These addresses are + * either pending an ASCONF operation (and exist on the ASCONF parameter + * queue), or they are permanently restricted (the peer has returned an + * ERROR indication to an ASCONF(ADD), or the peer does not support ASCONF). + * + * Deleted addresses are always immediately removed from the lists as they will + * (shortly) no longer exist in the kernel. We send ASCONFs as a courtesy, + * only if allowed. + */ + +/* + * ASCONF parameter processing. + * response_required: set if a reply is required (eg. SUCCESS_REPORT). + * returns a mbuf to an "error" response parameter or NULL/"success" if ok. + * FIX: allocating this many mbufs on the fly is pretty inefficient... + */ +static struct mbuf * +sctp_asconf_success_response(uint32_t id) +{ + struct mbuf *m_reply = NULL; + struct sctp_asconf_paramhdr *aph; + + m_reply = sctp_get_mbuf_for_msg(sizeof(struct sctp_asconf_paramhdr), + 0, M_NOWAIT, 1, MT_DATA); + if (m_reply == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, + "asconf_success_response: couldn't get mbuf!\n"); + return (NULL); + } + aph = mtod(m_reply, struct sctp_asconf_paramhdr *); + aph->correlation_id = id; + aph->ph.param_type = htons(SCTP_SUCCESS_REPORT); + aph->ph.param_length = sizeof(struct sctp_asconf_paramhdr); + SCTP_BUF_LEN(m_reply) = aph->ph.param_length; + aph->ph.param_length = htons(aph->ph.param_length); + + return (m_reply); +} + +static struct mbuf * +sctp_asconf_error_response(uint32_t id, uint16_t cause, uint8_t *error_tlv, + uint16_t tlv_length) +{ + struct mbuf *m_reply = NULL; + struct sctp_asconf_paramhdr *aph; + struct sctp_error_cause *error; + uint32_t buf_len; + uint16_t i, param_length, cause_length, padding_length; + uint8_t *tlv; + + if (error_tlv == NULL) { + tlv_length = 0; + } + cause_length = sizeof(struct sctp_error_cause) + tlv_length; + param_length = sizeof(struct sctp_asconf_paramhdr) + cause_length; + padding_length = tlv_length % 4; + if (padding_length != 0) { + padding_length = 4 - padding_length; + } + buf_len = param_length + padding_length; + if (buf_len > MLEN) { + SCTPDBG(SCTP_DEBUG_ASCONF1, + "asconf_error_response: tlv_length (%xh) too big\n", + tlv_length); + return (NULL); + } + m_reply = sctp_get_mbuf_for_msg(buf_len, 0, M_NOWAIT, 1, MT_DATA); + if (m_reply == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, + "asconf_error_response: couldn't get mbuf!\n"); + return (NULL); + } + aph = mtod(m_reply, struct sctp_asconf_paramhdr *); + aph->ph.param_type = htons(SCTP_ERROR_CAUSE_IND); + aph->ph.param_length = htons(param_length); + aph->correlation_id = id; + error = (struct sctp_error_cause *)(aph + 1); + error->code = htons(cause); + error->length = htons(cause_length); + if (error_tlv != NULL) { + tlv = (uint8_t *) (error + 1); + memcpy(tlv, error_tlv, tlv_length); + for (i = 0; i < padding_length; i++) { + tlv[tlv_length + i] = 0; + } + } + SCTP_BUF_LEN(m_reply) = buf_len; + return (m_reply); +} + +static struct mbuf * +sctp_process_asconf_add_ip(struct sockaddr *src, struct sctp_asconf_paramhdr *aph, + struct sctp_tcb *stcb, int send_hb, int response_required) +{ + struct sctp_nets *net; + struct mbuf *m_reply = NULL; + union sctp_sockstore store; + struct sctp_paramhdr *ph; + uint16_t param_type, aparam_length; +#if defined(INET) || defined(INET6) + uint16_t param_length; +#endif + struct sockaddr *sa; + int zero_address = 0; + int bad_address = 0; +#ifdef INET + struct sockaddr_in *sin; + struct sctp_ipv4addr_param *v4addr; +#endif +#ifdef INET6 + struct sockaddr_in6 *sin6; + struct sctp_ipv6addr_param *v6addr; +#endif + + aparam_length = ntohs(aph->ph.param_length); + if (aparam_length < sizeof(struct sctp_asconf_paramhdr) + sizeof(struct sctp_paramhdr)) { + return (NULL); + } + ph = (struct sctp_paramhdr *)(aph + 1); + param_type = ntohs(ph->param_type); +#if defined(INET) || defined(INET6) + param_length = ntohs(ph->param_length); + if (param_length + sizeof(struct sctp_asconf_paramhdr) != aparam_length) { + return (NULL); + } +#endif + sa = &store.sa; + switch (param_type) { +#ifdef INET + case SCTP_IPV4_ADDRESS: + if (param_length != sizeof(struct sctp_ipv4addr_param)) { + /* invalid param size */ + return (NULL); + } + v4addr = (struct sctp_ipv4addr_param *)ph; + sin = &store.sin; + memset(sin, 0, sizeof(*sin)); + sin->sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + sin->sin_len = sizeof(struct sockaddr_in); +#endif + sin->sin_port = stcb->rport; + sin->sin_addr.s_addr = v4addr->addr; + if ((sin->sin_addr.s_addr == INADDR_BROADCAST) || + IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { + bad_address = 1; + } + if (sin->sin_addr.s_addr == INADDR_ANY) + zero_address = 1; + SCTPDBG(SCTP_DEBUG_ASCONF1, "process_asconf_add_ip: adding "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, sa); + break; +#endif +#ifdef INET6 + case SCTP_IPV6_ADDRESS: + if (param_length != sizeof(struct sctp_ipv6addr_param)) { + /* invalid param size */ + return (NULL); + } + v6addr = (struct sctp_ipv6addr_param *)ph; + sin6 = &store.sin6; + memset(sin6, 0, sizeof(*sin6)); + sin6->sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6->sin6_len = sizeof(struct sockaddr_in6); +#endif + sin6->sin6_port = stcb->rport; + memcpy((caddr_t)&sin6->sin6_addr, v6addr->addr, + sizeof(struct in6_addr)); + if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { + bad_address = 1; + } + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) + zero_address = 1; + SCTPDBG(SCTP_DEBUG_ASCONF1, "process_asconf_add_ip: adding "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, sa); + break; +#endif + default: + m_reply = sctp_asconf_error_response(aph->correlation_id, + SCTP_CAUSE_INVALID_PARAM, (uint8_t *) aph, + aparam_length); + return (m_reply); + } /* end switch */ + + /* if 0.0.0.0/::0, add the source address instead */ + if (zero_address && SCTP_BASE_SYSCTL(sctp_nat_friendly)) { + sa = src; + SCTPDBG(SCTP_DEBUG_ASCONF1, + "process_asconf_add_ip: using source addr "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, src); + } + net = NULL; + /* add the address */ + if (bad_address) { + m_reply = sctp_asconf_error_response(aph->correlation_id, + SCTP_CAUSE_INVALID_PARAM, (uint8_t *) aph, + aparam_length); + } else if (sctp_add_remote_addr(stcb, sa, &net, stcb->asoc.port, + SCTP_DONOT_SETSCOPE, + SCTP_ADDR_DYNAMIC_ADDED) != 0) { + SCTPDBG(SCTP_DEBUG_ASCONF1, + "process_asconf_add_ip: error adding address\n"); + m_reply = sctp_asconf_error_response(aph->correlation_id, + SCTP_CAUSE_RESOURCE_SHORTAGE, (uint8_t *) aph, + aparam_length); + } else { + if (response_required) { + m_reply = + sctp_asconf_success_response(aph->correlation_id); + } + if (net != NULL) { + /* notify upper layer */ + sctp_ulp_notify(SCTP_NOTIFY_ASCONF_ADD_IP, stcb, 0, sa, SCTP_SO_NOT_LOCKED); + sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, stcb->sctp_ep, stcb, net); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, + stcb, net); + if (send_hb) { + sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED); + } + } + } + return (m_reply); +} + +static int +sctp_asconf_del_remote_addrs_except(struct sctp_tcb *stcb, struct sockaddr *src) +{ + struct sctp_nets *src_net, *net, *nnet; + + /* make sure the source address exists as a destination net */ + src_net = sctp_findnet(stcb, src); + if (src_net == NULL) { + /* not found */ + return (-1); + } + + /* delete all destination addresses except the source */ + TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) { + if (net != src_net) { + /* delete this address */ + SCTPDBG(SCTP_DEBUG_ASCONF1, + "asconf_del_remote_addrs_except: deleting "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, + (struct sockaddr *)&net->ro._l_addr); + /* notify upper layer */ + sctp_ulp_notify(SCTP_NOTIFY_ASCONF_DELETE_IP, stcb, 0, + (struct sockaddr *)&net->ro._l_addr, SCTP_SO_NOT_LOCKED); + sctp_remove_net(stcb, net); + } + } + return (0); +} + +static struct mbuf * +sctp_process_asconf_delete_ip(struct sockaddr *src, + struct sctp_asconf_paramhdr *aph, + struct sctp_tcb *stcb, int response_required) +{ + struct mbuf *m_reply = NULL; + union sctp_sockstore store; + struct sctp_paramhdr *ph; + uint16_t param_type, aparam_length; +#if defined(INET) || defined(INET6) + uint16_t param_length; +#endif + struct sockaddr *sa; + int zero_address = 0; + int result; +#ifdef INET + struct sockaddr_in *sin; + struct sctp_ipv4addr_param *v4addr; +#endif +#ifdef INET6 + struct sockaddr_in6 *sin6; + struct sctp_ipv6addr_param *v6addr; +#endif + + aparam_length = ntohs(aph->ph.param_length); + if (aparam_length < sizeof(struct sctp_asconf_paramhdr) + sizeof(struct sctp_paramhdr)) { + return (NULL); + } + ph = (struct sctp_paramhdr *)(aph + 1); + param_type = ntohs(ph->param_type); +#if defined(INET) || defined(INET6) + param_length = ntohs(ph->param_length); + if (param_length + sizeof(struct sctp_asconf_paramhdr) != aparam_length) { + return (NULL); + } +#endif + sa = &store.sa; + switch (param_type) { +#ifdef INET + case SCTP_IPV4_ADDRESS: + if (param_length != sizeof(struct sctp_ipv4addr_param)) { + /* invalid param size */ + return (NULL); + } + v4addr = (struct sctp_ipv4addr_param *)ph; + sin = &store.sin; + memset(sin, 0, sizeof(*sin)); + sin->sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + sin->sin_len = sizeof(struct sockaddr_in); +#endif + sin->sin_port = stcb->rport; + sin->sin_addr.s_addr = v4addr->addr; + if (sin->sin_addr.s_addr == INADDR_ANY) + zero_address = 1; + SCTPDBG(SCTP_DEBUG_ASCONF1, + "process_asconf_delete_ip: deleting "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, sa); + break; +#endif +#ifdef INET6 + case SCTP_IPV6_ADDRESS: + if (param_length != sizeof(struct sctp_ipv6addr_param)) { + /* invalid param size */ + return (NULL); + } + v6addr = (struct sctp_ipv6addr_param *)ph; + sin6 = &store.sin6; + memset(sin6, 0, sizeof(*sin6)); + sin6->sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6->sin6_len = sizeof(struct sockaddr_in6); +#endif + sin6->sin6_port = stcb->rport; + memcpy(&sin6->sin6_addr, v6addr->addr, + sizeof(struct in6_addr)); + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) + zero_address = 1; + SCTPDBG(SCTP_DEBUG_ASCONF1, + "process_asconf_delete_ip: deleting "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, sa); + break; +#endif + default: + m_reply = sctp_asconf_error_response(aph->correlation_id, + SCTP_CAUSE_UNRESOLVABLE_ADDR, (uint8_t *) aph, + aparam_length); + return (m_reply); + } + + /* make sure the source address is not being deleted */ + if (sctp_cmpaddr(sa, src)) { + /* trying to delete the source address! */ + SCTPDBG(SCTP_DEBUG_ASCONF1, "process_asconf_delete_ip: tried to delete source addr\n"); + m_reply = sctp_asconf_error_response(aph->correlation_id, + SCTP_CAUSE_DELETING_SRC_ADDR, (uint8_t *) aph, + aparam_length); + return (m_reply); + } + + /* if deleting 0.0.0.0/::0, delete all addresses except src addr */ + if (zero_address && SCTP_BASE_SYSCTL(sctp_nat_friendly)) { + result = sctp_asconf_del_remote_addrs_except(stcb, src); + + if (result) { + /* src address did not exist? */ + SCTPDBG(SCTP_DEBUG_ASCONF1, "process_asconf_delete_ip: src addr does not exist?\n"); + /* what error to reply with?? */ + m_reply = + sctp_asconf_error_response(aph->correlation_id, + SCTP_CAUSE_REQUEST_REFUSED, (uint8_t *) aph, + aparam_length); + } else if (response_required) { + m_reply = + sctp_asconf_success_response(aph->correlation_id); + } + return (m_reply); + } + + /* delete the address */ + result = sctp_del_remote_addr(stcb, sa); + /* + * note if result == -2, the address doesn't exist in the asoc but + * since it's being deleted anyways, we just ack the delete -- but + * this probably means something has already gone awry + */ + if (result == -1) { + /* only one address in the asoc */ + SCTPDBG(SCTP_DEBUG_ASCONF1, "process_asconf_delete_ip: tried to delete last IP addr!\n"); + m_reply = sctp_asconf_error_response(aph->correlation_id, + SCTP_CAUSE_DELETING_LAST_ADDR, (uint8_t *) aph, + aparam_length); + } else { + if (response_required) { + m_reply = sctp_asconf_success_response(aph->correlation_id); + } + /* notify upper layer */ + sctp_ulp_notify(SCTP_NOTIFY_ASCONF_DELETE_IP, stcb, 0, sa, SCTP_SO_NOT_LOCKED); + } + return (m_reply); +} + +static struct mbuf * +sctp_process_asconf_set_primary(struct sockaddr *src, + struct sctp_asconf_paramhdr *aph, + struct sctp_tcb *stcb, int response_required) +{ + struct mbuf *m_reply = NULL; + union sctp_sockstore store; + struct sctp_paramhdr *ph; + uint16_t param_type, aparam_length; +#if defined(INET) || defined(INET6) + uint16_t param_length; +#endif + struct sockaddr *sa; + int zero_address = 0; +#ifdef INET + struct sockaddr_in *sin; + struct sctp_ipv4addr_param *v4addr; +#endif +#ifdef INET6 + struct sockaddr_in6 *sin6; + struct sctp_ipv6addr_param *v6addr; +#endif + + aparam_length = ntohs(aph->ph.param_length); + if (aparam_length < sizeof(struct sctp_asconf_paramhdr) + sizeof(struct sctp_paramhdr)) { + return (NULL); + } + ph = (struct sctp_paramhdr *)(aph + 1); + param_type = ntohs(ph->param_type); +#if defined(INET) || defined(INET6) + param_length = ntohs(ph->param_length); + if (param_length + sizeof(struct sctp_asconf_paramhdr) != aparam_length) { + return (NULL); + } +#endif + sa = &store.sa; + switch (param_type) { +#ifdef INET + case SCTP_IPV4_ADDRESS: + if (param_length != sizeof(struct sctp_ipv4addr_param)) { + /* invalid param size */ + return (NULL); + } + v4addr = (struct sctp_ipv4addr_param *)ph; + sin = &store.sin; + memset(sin, 0, sizeof(*sin)); + sin->sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + sin->sin_len = sizeof(struct sockaddr_in); +#endif + sin->sin_addr.s_addr = v4addr->addr; + if (sin->sin_addr.s_addr == INADDR_ANY) + zero_address = 1; + SCTPDBG(SCTP_DEBUG_ASCONF1, "process_asconf_set_primary: "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, sa); + break; +#endif +#ifdef INET6 + case SCTP_IPV6_ADDRESS: + if (param_length != sizeof(struct sctp_ipv6addr_param)) { + /* invalid param size */ + return (NULL); + } + v6addr = (struct sctp_ipv6addr_param *)ph; + sin6 = &store.sin6; + memset(sin6, 0, sizeof(*sin6)); + sin6->sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6->sin6_len = sizeof(struct sockaddr_in6); +#endif + memcpy((caddr_t)&sin6->sin6_addr, v6addr->addr, + sizeof(struct in6_addr)); + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) + zero_address = 1; + SCTPDBG(SCTP_DEBUG_ASCONF1, "process_asconf_set_primary: "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, sa); + break; +#endif + default: + m_reply = sctp_asconf_error_response(aph->correlation_id, + SCTP_CAUSE_UNRESOLVABLE_ADDR, (uint8_t *) aph, + aparam_length); + return (m_reply); + } + + /* if 0.0.0.0/::0, use the source address instead */ + if (zero_address && SCTP_BASE_SYSCTL(sctp_nat_friendly)) { + sa = src; + SCTPDBG(SCTP_DEBUG_ASCONF1, + "process_asconf_set_primary: using source addr "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, src); + } + /* set the primary address */ + if (sctp_set_primary_addr(stcb, sa, NULL) == 0) { + SCTPDBG(SCTP_DEBUG_ASCONF1, + "process_asconf_set_primary: primary address set\n"); + /* notify upper layer */ + sctp_ulp_notify(SCTP_NOTIFY_ASCONF_SET_PRIMARY, stcb, 0, sa, SCTP_SO_NOT_LOCKED); + if ((stcb->asoc.primary_destination->dest_state & SCTP_ADDR_REACHABLE) && + ((stcb->asoc.primary_destination->dest_state & SCTP_ADDR_PF) == 0) && + (stcb->asoc.alternate != NULL)) { + sctp_free_remote_addr(stcb->asoc.alternate); + stcb->asoc.alternate = NULL; + } + if (response_required) { + m_reply = sctp_asconf_success_response(aph->correlation_id); + } + /* Mobility adaptation. + Ideally, when the reception of SET PRIMARY with DELETE IP + ADDRESS of the previous primary destination, unacknowledged + DATA are retransmitted immediately to the new primary + destination for seamless handover. + If the destination is UNCONFIRMED and marked to REQ_PRIM, + The retransmission occur when reception of the + HEARTBEAT-ACK. (See sctp_handle_heartbeat_ack in + sctp_input.c) + Also, when change of the primary destination, it is better + that all subsequent new DATA containing already queued DATA + are transmitted to the new primary destination. (by micchie) + */ + if ((sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_BASE) || + sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_FASTHANDOFF)) && + sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_PRIM_DELETED) && + (stcb->asoc.primary_destination->dest_state & + SCTP_ADDR_UNCONFIRMED) == 0) { + sctp_timer_stop(SCTP_TIMER_TYPE_PRIM_DELETED, + stcb->sctp_ep, stcb, NULL, + SCTP_FROM_SCTP_ASCONF + SCTP_LOC_1); + if (sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_FASTHANDOFF)) { + sctp_assoc_immediate_retrans(stcb, + stcb->asoc.primary_destination); + } + if (sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_BASE)) { + sctp_move_chunks_from_net(stcb, + stcb->asoc.deleted_primary); + } + sctp_delete_prim_timer(stcb->sctp_ep, stcb); + } + } else { + /* couldn't set the requested primary address! */ + SCTPDBG(SCTP_DEBUG_ASCONF1, + "process_asconf_set_primary: set primary failed!\n"); + /* must have been an invalid address, so report */ + m_reply = sctp_asconf_error_response(aph->correlation_id, + SCTP_CAUSE_UNRESOLVABLE_ADDR, (uint8_t *) aph, + aparam_length); + } + + return (m_reply); +} + +/* + * handles an ASCONF chunk. + * if all parameters are processed ok, send a plain (empty) ASCONF-ACK + */ +void +sctp_handle_asconf(struct mbuf *m, unsigned int offset, + struct sockaddr *src, + struct sctp_asconf_chunk *cp, struct sctp_tcb *stcb, + int first) +{ + struct sctp_association *asoc; + uint32_t serial_num; + struct mbuf *n, *m_ack, *m_result, *m_tail; + struct sctp_asconf_ack_chunk *ack_cp; + struct sctp_asconf_paramhdr *aph; + struct sctp_ipv6addr_param *p_addr; + unsigned int asconf_limit, cnt; + int error = 0; /* did an error occur? */ + + /* asconf param buffer */ + uint8_t aparam_buf[SCTP_PARAM_BUFFER_SIZE]; + struct sctp_asconf_ack *ack, *ack_next; + + /* verify minimum length */ + if (ntohs(cp->ch.chunk_length) < sizeof(struct sctp_asconf_chunk)) { + SCTPDBG(SCTP_DEBUG_ASCONF1, + "handle_asconf: chunk too small = %xh\n", + ntohs(cp->ch.chunk_length)); + return; + } + asoc = &stcb->asoc; + serial_num = ntohl(cp->serial_number); + + if (SCTP_TSN_GE(asoc->asconf_seq_in, serial_num)) { + /* got a duplicate ASCONF */ + SCTPDBG(SCTP_DEBUG_ASCONF1, + "handle_asconf: got duplicate serial number = %xh\n", + serial_num); + return; + } else if (serial_num != (asoc->asconf_seq_in + 1)) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "handle_asconf: incorrect serial number = %xh (expected next = %xh)\n", + serial_num, asoc->asconf_seq_in + 1); + return; + } + + /* it's the expected "next" sequence number, so process it */ + asoc->asconf_seq_in = serial_num; /* update sequence */ + /* get length of all the param's in the ASCONF */ + asconf_limit = offset + ntohs(cp->ch.chunk_length); + SCTPDBG(SCTP_DEBUG_ASCONF1, + "handle_asconf: asconf_limit=%u, sequence=%xh\n", + asconf_limit, serial_num); + + if (first) { + /* delete old cache */ + SCTPDBG(SCTP_DEBUG_ASCONF1,"handle_asconf: Now processing first ASCONF. Try to delete old cache\n"); + + TAILQ_FOREACH_SAFE(ack, &asoc->asconf_ack_sent, next, ack_next) { + if (ack->serial_number == serial_num) + break; + SCTPDBG(SCTP_DEBUG_ASCONF1,"handle_asconf: delete old(%u) < first(%u)\n", + ack->serial_number, serial_num); + TAILQ_REMOVE(&asoc->asconf_ack_sent, ack, next); + if (ack->data != NULL) { + sctp_m_freem(ack->data); + } + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), ack); + } + } + + m_ack = sctp_get_mbuf_for_msg(sizeof(struct sctp_asconf_ack_chunk), 0, + M_NOWAIT, 1, MT_DATA); + if (m_ack == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, + "handle_asconf: couldn't get mbuf!\n"); + return; + } + m_tail = m_ack; /* current reply chain's tail */ + + /* fill in ASCONF-ACK header */ + ack_cp = mtod(m_ack, struct sctp_asconf_ack_chunk *); + ack_cp->ch.chunk_type = SCTP_ASCONF_ACK; + ack_cp->ch.chunk_flags = 0; + ack_cp->serial_number = htonl(serial_num); + /* set initial lengths (eg. just an ASCONF-ACK), ntohx at the end! */ + SCTP_BUF_LEN(m_ack) = sizeof(struct sctp_asconf_ack_chunk); + ack_cp->ch.chunk_length = sizeof(struct sctp_asconf_ack_chunk); + + /* skip the lookup address parameter */ + offset += sizeof(struct sctp_asconf_chunk); + p_addr = (struct sctp_ipv6addr_param *)sctp_m_getptr(m, offset, sizeof(struct sctp_paramhdr), (uint8_t *)&aparam_buf); + if (p_addr == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, + "handle_asconf: couldn't get lookup addr!\n"); + /* respond with a missing/invalid mandatory parameter error */ + sctp_m_freem(m_ack); + return; + } + /* skip lookup addr */ + offset += SCTP_SIZE32(ntohs(p_addr->ph.param_length)); + /* get pointer to first asconf param in ASCONF */ + aph = (struct sctp_asconf_paramhdr *)sctp_m_getptr(m, offset, sizeof(struct sctp_asconf_paramhdr), (uint8_t *)&aparam_buf); + if (aph == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "Empty ASCONF received?\n"); + goto send_reply; + } + /* process through all parameters */ + cnt = 0; + while (aph != NULL) { + unsigned int param_length, param_type; + + param_type = ntohs(aph->ph.param_type); + param_length = ntohs(aph->ph.param_length); + if (offset + param_length > asconf_limit) { + /* parameter goes beyond end of chunk! */ + sctp_m_freem(m_ack); + return; + } + m_result = NULL; + + if (param_length > sizeof(aparam_buf)) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "handle_asconf: param length (%u) larger than buffer size!\n", param_length); + sctp_m_freem(m_ack); + return; + } + if (param_length < sizeof(struct sctp_asconf_paramhdr)) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "handle_asconf: param length (%u) too short\n", param_length); + sctp_m_freem(m_ack); + return; + } + /* get the entire parameter */ + aph = (struct sctp_asconf_paramhdr *)sctp_m_getptr(m, offset, param_length, aparam_buf); + if (aph == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "handle_asconf: couldn't get entire param\n"); + sctp_m_freem(m_ack); + return; + } + switch (param_type) { + case SCTP_ADD_IP_ADDRESS: + m_result = sctp_process_asconf_add_ip(src, aph, stcb, + (cnt < SCTP_BASE_SYSCTL(sctp_hb_maxburst)), error); + cnt++; + break; + case SCTP_DEL_IP_ADDRESS: + m_result = sctp_process_asconf_delete_ip(src, aph, stcb, + error); + break; + case SCTP_ERROR_CAUSE_IND: + /* not valid in an ASCONF chunk */ + break; + case SCTP_SET_PRIM_ADDR: + m_result = sctp_process_asconf_set_primary(src, aph, + stcb, error); + break; + case SCTP_NAT_VTAGS: + SCTPDBG(SCTP_DEBUG_ASCONF1, "handle_asconf: sees a NAT VTAG state parameter\n"); + break; + case SCTP_SUCCESS_REPORT: + /* not valid in an ASCONF chunk */ + break; + case SCTP_ULP_ADAPTATION: + /* FIX */ + break; + default: + if ((param_type & 0x8000) == 0) { + /* Been told to STOP at this param */ + asconf_limit = offset; + /* + * FIX FIX - We need to call + * sctp_arethere_unrecognized_parameters() + * to get a operr and send it for any + * param's with the 0x4000 bit set OR do it + * here ourselves... note we still must STOP + * if the 0x8000 bit is clear. + */ + } + /* unknown/invalid param type */ + break; + } /* switch */ + + /* add any (error) result to the reply mbuf chain */ + if (m_result != NULL) { + SCTP_BUF_NEXT(m_tail) = m_result; + m_tail = m_result; + ack_cp->ch.chunk_length += SCTP_BUF_LEN(m_result); + /* set flag to force success reports */ + error = 1; + } + offset += SCTP_SIZE32(param_length); + /* update remaining ASCONF message length to process */ + if (offset >= asconf_limit) { + /* no more data in the mbuf chain */ + break; + } + /* get pointer to next asconf param */ + aph = (struct sctp_asconf_paramhdr *)sctp_m_getptr(m, offset, + sizeof(struct sctp_asconf_paramhdr), + (uint8_t *)&aparam_buf); + if (aph == NULL) { + /* can't get an asconf paramhdr */ + SCTPDBG(SCTP_DEBUG_ASCONF1, "handle_asconf: can't get asconf param hdr!\n"); + /* FIX ME - add error here... */ + } + } + + send_reply: + ack_cp->ch.chunk_length = htons(ack_cp->ch.chunk_length); + /* save the ASCONF-ACK reply */ + ack = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asconf_ack), + struct sctp_asconf_ack); + if (ack == NULL) { + sctp_m_freem(m_ack); + return; + } + ack->serial_number = serial_num; + ack->last_sent_to = NULL; + ack->data = m_ack; + ack->len = 0; + for (n = m_ack; n != NULL; n = SCTP_BUF_NEXT(n)) { + ack->len += SCTP_BUF_LEN(n); + } + TAILQ_INSERT_TAIL(&stcb->asoc.asconf_ack_sent, ack, next); + + /* see if last_control_chunk_from is set properly (use IP src addr) */ + if (stcb->asoc.last_control_chunk_from == NULL) { + /* + * this could happen if the source address was just newly + * added + */ + SCTPDBG(SCTP_DEBUG_ASCONF1, "handle_asconf: looking up net for IP source address\n"); + SCTPDBG(SCTP_DEBUG_ASCONF1, "Looking for IP source: "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, src); + /* look up the from address */ + stcb->asoc.last_control_chunk_from = sctp_findnet(stcb, src); +#ifdef SCTP_DEBUG + if (stcb->asoc.last_control_chunk_from == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "handle_asconf: IP source address not found?!\n"); + } +#endif + } +} + +/* + * does the address match? returns 0 if not, 1 if so + */ +static uint32_t +sctp_asconf_addr_match(struct sctp_asconf_addr *aa, struct sockaddr *sa) +{ + switch (sa->sa_family) { +#ifdef INET6 + case AF_INET6: + { + /* XXX scopeid */ + struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa; + + if ((aa->ap.addrp.ph.param_type == SCTP_IPV6_ADDRESS) && + (memcmp(&aa->ap.addrp.addr, &sin6->sin6_addr, + sizeof(struct in6_addr)) == 0)) { + return (1); + } + break; + } +#endif +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin = (struct sockaddr_in *)sa; + + if ((aa->ap.addrp.ph.param_type == SCTP_IPV4_ADDRESS) && + (memcmp(&aa->ap.addrp.addr, &sin->sin_addr, + sizeof(struct in_addr)) == 0)) { + return (1); + } + break; + } +#endif + default: + break; + } + return (0); +} + +/* + * does the address match? returns 0 if not, 1 if so + */ +static uint32_t +sctp_addr_match(struct sctp_paramhdr *ph, struct sockaddr *sa) +{ +#if defined(INET) || defined(INET6) + uint16_t param_type, param_length; + + param_type = ntohs(ph->param_type); + param_length = ntohs(ph->param_length); +#endif + switch (sa->sa_family) { +#ifdef INET6 + case AF_INET6: + { + /* XXX scopeid */ + struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa; + struct sctp_ipv6addr_param *v6addr; + + v6addr = (struct sctp_ipv6addr_param *)ph; + if ((param_type == SCTP_IPV6_ADDRESS) && + (param_length == sizeof(struct sctp_ipv6addr_param)) && + (memcmp(&v6addr->addr, &sin6->sin6_addr, + sizeof(struct in6_addr)) == 0)) { + return (1); + } + break; + } +#endif +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin = (struct sockaddr_in *)sa; + struct sctp_ipv4addr_param *v4addr; + + v4addr = (struct sctp_ipv4addr_param *)ph; + if ((param_type == SCTP_IPV4_ADDRESS) && + (param_length == sizeof(struct sctp_ipv4addr_param)) && + (memcmp(&v4addr->addr, &sin->sin_addr, + sizeof(struct in_addr)) == 0)) { + return (1); + } + break; + } +#endif + default: + break; + } + return (0); +} +/* + * Cleanup for non-responded/OP ERR'd ASCONF + */ +void +sctp_asconf_cleanup(struct sctp_tcb *stcb) +{ + /* + * clear out any existing asconfs going out + */ + sctp_timer_stop(SCTP_TIMER_TYPE_ASCONF, stcb->sctp_ep, stcb, NULL, + SCTP_FROM_SCTP_ASCONF + SCTP_LOC_2); + stcb->asoc.asconf_seq_out_acked = stcb->asoc.asconf_seq_out; + /* remove the old ASCONF on our outbound queue */ + sctp_toss_old_asconf(stcb); +} + +/* + * cleanup any cached source addresses that may be topologically + * incorrect after a new address has been added to this interface. + */ +static void +sctp_asconf_nets_cleanup(struct sctp_tcb *stcb, struct sctp_ifn *ifn) +{ + struct sctp_nets *net; + + /* + * Ideally, we want to only clear cached routes and source addresses + * that are topologically incorrect. But since there is no easy way + * to know whether the newly added address on the ifn would cause a + * routing change (i.e. a new egress interface would be chosen) + * without doing a new routing lookup and source address selection, + * we will (for now) just flush any cached route using a different + * ifn (and cached source addrs) and let output re-choose them during + * the next send on that net. + */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + /* + * clear any cached route (and cached source address) if the + * route's interface is NOT the same as the address change. + * If it's the same interface, just clear the cached source + * address. + */ + if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro) && + ((ifn == NULL) || + (SCTP_GET_IF_INDEX_FROM_ROUTE(&net->ro) != ifn->ifn_index))) { + /* clear any cached route */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + RO_NHFREE(&net->ro); +#else + RTFREE(net->ro.ro_rt); + net->ro.ro_rt = NULL; +#endif + } + /* clear any cached source address */ + if (net->src_addr_selected) { + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + net->src_addr_selected = 0; + } + } +} + +void +sctp_assoc_immediate_retrans(struct sctp_tcb *stcb, struct sctp_nets *dstnet) +{ + int error; + + if (dstnet->dest_state & SCTP_ADDR_UNCONFIRMED) { + return; + } + if (stcb->asoc.deleted_primary == NULL) { + return; + } + + if (!TAILQ_EMPTY(&stcb->asoc.sent_queue)) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "assoc_immediate_retrans: Deleted primary is "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa); + SCTPDBG(SCTP_DEBUG_ASCONF1, "Current Primary is "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.primary_destination->ro._l_addr.sa); + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, + stcb->asoc.deleted_primary, + SCTP_FROM_SCTP_ASCONF + SCTP_LOC_3); + stcb->asoc.num_send_timers_up--; + if (stcb->asoc.num_send_timers_up < 0) { + stcb->asoc.num_send_timers_up = 0; + } + SCTP_TCB_LOCK_ASSERT(stcb); + error = sctp_t3rxt_timer(stcb->sctp_ep, stcb, + stcb->asoc.deleted_primary); + if (error) { + SCTP_INP_DECR_REF(stcb->sctp_ep); + return; + } + SCTP_TCB_LOCK_ASSERT(stcb); +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(4, stcb->sctp_ep, stcb, stcb->asoc.deleted_primary); +#endif + sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); + if ((stcb->asoc.num_send_timers_up == 0) && + (stcb->asoc.sent_queue_cnt > 0)) { + struct sctp_tmit_chunk *chk; + + TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { + if (chk->whoTo != NULL) { + break; + } + } + if (chk != NULL) { + sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, chk->whoTo); + } + } + } + return; +} + +#if defined(__FreeBSD__) || defined(__APPLE__) || defined(__Userspace__) +static int +sctp_asconf_queue_mgmt(struct sctp_tcb *, struct sctp_ifa *, uint16_t); + +void +sctp_net_immediate_retrans(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + struct sctp_tmit_chunk *chk; + + SCTPDBG(SCTP_DEBUG_ASCONF1, "net_immediate_retrans: RTO is %d\n", net->RTO); + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net, + SCTP_FROM_SCTP_ASCONF + SCTP_LOC_4); + stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net); + net->error_count = 0; + TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { + if (chk->whoTo == net) { + if (chk->sent < SCTP_DATAGRAM_RESEND) { + chk->sent = SCTP_DATAGRAM_RESEND; + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + sctp_flight_size_decrease(chk); + sctp_total_flight_decrease(stcb, chk); + net->marked_retrans++; + stcb->asoc.marked_retrans++; + } + } + } + if (net->marked_retrans) { + sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); + } +} + +static void +sctp_path_check_and_react(struct sctp_tcb *stcb, struct sctp_ifa *newifa) +{ + struct sctp_nets *net; + int addrnum, changed; + + /* If number of local valid addresses is 1, the valid address is + probably newly added address. + Several valid addresses in this association. A source address + may not be changed. Additionally, they can be configured on a + same interface as "alias" addresses. (by micchie) + */ + addrnum = sctp_local_addr_count(stcb); + SCTPDBG(SCTP_DEBUG_ASCONF1, "p_check_react(): %d local addresses\n", + addrnum); + if (addrnum == 1) { + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + /* clear any cached route and source address */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + RO_NHFREE(&net->ro); +#else + if (net->ro.ro_rt) { + RTFREE(net->ro.ro_rt); + net->ro.ro_rt = NULL; + } +#endif + if (net->src_addr_selected) { + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + net->src_addr_selected = 0; + } + /* Retransmit unacknowledged DATA chunks immediately */ + if (sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_FASTHANDOFF)) { + sctp_net_immediate_retrans(stcb, net); + } + /* also, SET PRIMARY is maybe already sent */ + } + return; + } + + /* Multiple local addresses exist in the association. */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + /* clear any cached route and source address */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + RO_NHFREE(&net->ro); +#else + if (net->ro.ro_rt) { + RTFREE(net->ro.ro_rt); + net->ro.ro_rt = NULL; + } +#endif + if (net->src_addr_selected) { + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + net->src_addr_selected = 0; + } + /* Check if the nexthop is corresponding to the new address. + If the new address is corresponding to the current nexthop, + the path will be changed. + If the new address is NOT corresponding to the current + nexthop, the path will not be changed. + */ + SCTP_RTALLOC((sctp_route_t *)&net->ro, + stcb->sctp_ep->def_vrf_id, + stcb->sctp_ep->fibnum); +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (net->ro.ro_nh == NULL) +#else + if (net->ro.ro_rt == NULL) +#endif + continue; + + changed = 0; + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + if (sctp_v4src_match_nexthop(newifa, (sctp_route_t *)&net->ro)) { + changed = 1; + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (sctp_v6src_match_nexthop( + &newifa->address.sin6, (sctp_route_t *)&net->ro)) { + changed = 1; + } + break; +#endif + default: + break; + } + /* if the newly added address does not relate routing + information, we skip. + */ + if (changed == 0) + continue; + /* Retransmit unacknowledged DATA chunks immediately */ + if (sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_FASTHANDOFF)) { + sctp_net_immediate_retrans(stcb, net); + } + /* Send SET PRIMARY for this new address */ + if (net == stcb->asoc.primary_destination) { + (void)sctp_asconf_queue_mgmt(stcb, newifa, + SCTP_SET_PRIM_ADDR); + } + } +} +#endif + +/* + * process an ADD/DELETE IP ack from peer. + * addr: corresponding sctp_ifa to the address being added/deleted. + * type: SCTP_ADD_IP_ADDRESS or SCTP_DEL_IP_ADDRESS. + * flag: 1=success, 0=failure. + */ +static void +sctp_asconf_addr_mgmt_ack(struct sctp_tcb *stcb, struct sctp_ifa *addr, uint32_t flag) +{ + /* + * do the necessary asoc list work- if we get a failure indication, + * leave the address on the assoc's restricted list. If we get a + * success indication, remove the address from the restricted list. + */ + /* + * Note: this will only occur for ADD_IP_ADDRESS, since + * DEL_IP_ADDRESS is never actually added to the list... + */ + if (flag) { + /* success case, so remove from the restricted list */ + sctp_del_local_addr_restricted(stcb, addr); + +#if defined(__FreeBSD__) || defined(__APPLE__) || defined(__Userspace__) + if (sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_BASE) || + sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_FASTHANDOFF)) { + sctp_path_check_and_react(stcb, addr); + return; + } +#endif + /* clear any cached/topologically incorrect source addresses */ + sctp_asconf_nets_cleanup(stcb, addr->ifn_p); + } + /* else, leave it on the list */ +} + +/* + * add an asconf add/delete/set primary IP address parameter to the queue. + * type = SCTP_ADD_IP_ADDRESS, SCTP_DEL_IP_ADDRESS, SCTP_SET_PRIM_ADDR. + * returns 0 if queued, -1 if not queued/removed. + * NOTE: if adding, but a delete for the same address is already scheduled + * (and not yet sent out), simply remove it from queue. Same for deleting + * an address already scheduled for add. If a duplicate operation is found, + * ignore the new one. + */ +static int +sctp_asconf_queue_mgmt(struct sctp_tcb *stcb, struct sctp_ifa *ifa, + uint16_t type) +{ + struct sctp_asconf_addr *aa, *aa_next; + + /* make sure the request isn't already in the queue */ + TAILQ_FOREACH_SAFE(aa, &stcb->asoc.asconf_queue, next, aa_next) { + /* address match? */ + if (sctp_asconf_addr_match(aa, &ifa->address.sa) == 0) + continue; + /* is the request already in queue but not sent? + * pass the request already sent in order to resolve the following case: + * 1. arrival of ADD, then sent + * 2. arrival of DEL. we can't remove the ADD request already sent + * 3. arrival of ADD + */ + if (aa->ap.aph.ph.param_type == type && aa->sent == 0) { + return (-1); + } + /* is the negative request already in queue, and not sent */ + if ((aa->sent == 0) && (type == SCTP_ADD_IP_ADDRESS) && + (aa->ap.aph.ph.param_type == SCTP_DEL_IP_ADDRESS)) { + /* add requested, delete already queued */ + TAILQ_REMOVE(&stcb->asoc.asconf_queue, aa, next); + /* remove the ifa from the restricted list */ + sctp_del_local_addr_restricted(stcb, ifa); + /* free the asconf param */ + SCTP_FREE(aa, SCTP_M_ASC_ADDR); + SCTPDBG(SCTP_DEBUG_ASCONF2, "asconf_queue_mgmt: add removes queued entry\n"); + return (-1); + } + if ((aa->sent == 0) && (type == SCTP_DEL_IP_ADDRESS) && + (aa->ap.aph.ph.param_type == SCTP_ADD_IP_ADDRESS)) { + /* delete requested, add already queued */ + TAILQ_REMOVE(&stcb->asoc.asconf_queue, aa, next); + /* remove the aa->ifa from the restricted list */ + sctp_del_local_addr_restricted(stcb, aa->ifa); + /* free the asconf param */ + SCTP_FREE(aa, SCTP_M_ASC_ADDR); + SCTPDBG(SCTP_DEBUG_ASCONF2, "asconf_queue_mgmt: delete removes queued entry\n"); + return (-1); + } + } /* for each aa */ + + /* adding new request to the queue */ + SCTP_MALLOC(aa, struct sctp_asconf_addr *, sizeof(*aa), + SCTP_M_ASC_ADDR); + if (aa == NULL) { + /* didn't get memory */ + SCTPDBG(SCTP_DEBUG_ASCONF1, "asconf_queue_mgmt: failed to get memory!\n"); + return (-1); + } + aa->special_del = 0; + /* fill in asconf address parameter fields */ + /* top level elements are "networked" during send */ + aa->ap.aph.ph.param_type = type; + aa->ifa = ifa; + atomic_add_int(&ifa->refcount, 1); + /* correlation_id filled in during send routine later... */ + switch (ifa->address.sa.sa_family) { +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6; + + sin6 = &ifa->address.sin6; + aa->ap.addrp.ph.param_type = SCTP_IPV6_ADDRESS; + aa->ap.addrp.ph.param_length = (sizeof(struct sctp_ipv6addr_param)); + aa->ap.aph.ph.param_length = sizeof(struct sctp_asconf_paramhdr) + + sizeof(struct sctp_ipv6addr_param); + memcpy(&aa->ap.addrp.addr, &sin6->sin6_addr, + sizeof(struct in6_addr)); + break; + } +#endif +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin; + + sin = &ifa->address.sin; + aa->ap.addrp.ph.param_type = SCTP_IPV4_ADDRESS; + aa->ap.addrp.ph.param_length = (sizeof(struct sctp_ipv4addr_param)); + aa->ap.aph.ph.param_length = sizeof(struct sctp_asconf_paramhdr) + + sizeof(struct sctp_ipv4addr_param); + memcpy(&aa->ap.addrp.addr, &sin->sin_addr, + sizeof(struct in_addr)); + break; + } +#endif + default: + /* invalid family! */ + SCTP_FREE(aa, SCTP_M_ASC_ADDR); + sctp_free_ifa(ifa); + return (-1); + } + aa->sent = 0; /* clear sent flag */ + + TAILQ_INSERT_TAIL(&stcb->asoc.asconf_queue, aa, next); +#ifdef SCTP_DEBUG + if (SCTP_BASE_SYSCTL(sctp_debug_on) & SCTP_DEBUG_ASCONF2) { + if (type == SCTP_ADD_IP_ADDRESS) { + SCTP_PRINTF("asconf_queue_mgmt: inserted asconf ADD_IP_ADDRESS: "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF2, &ifa->address.sa); + } else if (type == SCTP_DEL_IP_ADDRESS) { + SCTP_PRINTF("asconf_queue_mgmt: appended asconf DEL_IP_ADDRESS: "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF2, &ifa->address.sa); + } else { + SCTP_PRINTF("asconf_queue_mgmt: appended asconf SET_PRIM_ADDR: "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF2, &ifa->address.sa); + } + } +#endif + + return (0); +} + +/* + * add an asconf operation for the given ifa and type. + * type = SCTP_ADD_IP_ADDRESS, SCTP_DEL_IP_ADDRESS, SCTP_SET_PRIM_ADDR. + * returns 0 if completed, -1 if not completed, 1 if immediate send is + * advisable. + */ +static int +sctp_asconf_queue_add(struct sctp_tcb *stcb, struct sctp_ifa *ifa, + uint16_t type) +{ + uint32_t status; + int pending_delete_queued = 0; + int last; + + /* see if peer supports ASCONF */ + if (stcb->asoc.asconf_supported == 0) { + return (-1); + } + + /* + * if this is deleting the last address from the assoc, mark it as + * pending. + */ + if ((type == SCTP_DEL_IP_ADDRESS) && !stcb->asoc.asconf_del_pending) { + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + last = (sctp_local_addr_count(stcb) == 0); + } else { + last = (sctp_local_addr_count(stcb) == 1); + } + if (last) { + /* set the pending delete info only */ + stcb->asoc.asconf_del_pending = 1; + stcb->asoc.asconf_addr_del_pending = ifa; + atomic_add_int(&ifa->refcount, 1); + SCTPDBG(SCTP_DEBUG_ASCONF2, + "asconf_queue_add: mark delete last address pending\n"); + return (-1); + } + } + + /* queue an asconf parameter */ + status = sctp_asconf_queue_mgmt(stcb, ifa, type); + + /* + * if this is an add, and there is a delete also pending (i.e. the + * last local address is being changed), queue the pending delete too. + */ + if ((type == SCTP_ADD_IP_ADDRESS) && stcb->asoc.asconf_del_pending && (status == 0)) { + /* queue in the pending delete */ + if (sctp_asconf_queue_mgmt(stcb, + stcb->asoc.asconf_addr_del_pending, + SCTP_DEL_IP_ADDRESS) == 0) { + SCTPDBG(SCTP_DEBUG_ASCONF2, "asconf_queue_add: queuing pending delete\n"); + pending_delete_queued = 1; + /* clear out the pending delete info */ + stcb->asoc.asconf_del_pending = 0; + sctp_free_ifa(stcb->asoc.asconf_addr_del_pending); + stcb->asoc.asconf_addr_del_pending = NULL; + } + } + + if (pending_delete_queued) { + struct sctp_nets *net; + /* + * since we know that the only/last address is now being + * changed in this case, reset the cwnd/rto on all nets to + * start as a new address and path. Also clear the error + * counts to give the assoc the best chance to complete the + * address change. + */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, + net); + net->RTO = 0; + net->error_count = 0; + } + stcb->asoc.overall_error_count = 0; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { + sctp_misc_ints(SCTP_THRESHOLD_CLEAR, + stcb->asoc.overall_error_count, + 0, + SCTP_FROM_SCTP_ASCONF, + __LINE__); + } + + /* queue in an advisory set primary too */ + (void)sctp_asconf_queue_mgmt(stcb, ifa, SCTP_SET_PRIM_ADDR); + /* let caller know we should send this out immediately */ + status = 1; + } + return (status); +} + +/*- + * add an asconf delete IP address parameter to the queue by sockaddr and + * possibly with no sctp_ifa available. This is only called by the routine + * that checks the addresses in an INIT-ACK against the current address list. + * returns 0 if completed, non-zero if not completed. + * NOTE: if an add is already scheduled (and not yet sent out), simply + * remove it from queue. If a duplicate operation is found, ignore the + * new one. + */ +static int +sctp_asconf_queue_sa_delete(struct sctp_tcb *stcb, struct sockaddr *sa) +{ + struct sctp_ifa *ifa; + struct sctp_asconf_addr *aa, *aa_next; + + if (stcb == NULL) { + return (-1); + } + /* see if peer supports ASCONF */ + if (stcb->asoc.asconf_supported == 0) { + return (-1); + } + /* make sure the request isn't already in the queue */ + TAILQ_FOREACH_SAFE(aa, &stcb->asoc.asconf_queue, next, aa_next) { + /* address match? */ + if (sctp_asconf_addr_match(aa, sa) == 0) + continue; + /* is the request already in queue (sent or not) */ + if (aa->ap.aph.ph.param_type == SCTP_DEL_IP_ADDRESS) { + return (-1); + } + /* is the negative request already in queue, and not sent */ + if (aa->sent == 1) + continue; + if (aa->ap.aph.ph.param_type == SCTP_ADD_IP_ADDRESS) { + /* add already queued, so remove existing entry */ + TAILQ_REMOVE(&stcb->asoc.asconf_queue, aa, next); + sctp_del_local_addr_restricted(stcb, aa->ifa); + /* free the entry */ + SCTP_FREE(aa, SCTP_M_ASC_ADDR); + return (-1); + } + } /* for each aa */ + + /* find any existing ifa-- NOTE ifa CAN be allowed to be NULL */ + ifa = sctp_find_ifa_by_addr(sa, stcb->asoc.vrf_id, SCTP_ADDR_NOT_LOCKED); + + /* adding new request to the queue */ + SCTP_MALLOC(aa, struct sctp_asconf_addr *, sizeof(*aa), + SCTP_M_ASC_ADDR); + if (aa == NULL) { + /* didn't get memory */ + SCTPDBG(SCTP_DEBUG_ASCONF1, + "sctp_asconf_queue_sa_delete: failed to get memory!\n"); + return (-1); + } + aa->special_del = 0; + /* fill in asconf address parameter fields */ + /* top level elements are "networked" during send */ + aa->ap.aph.ph.param_type = SCTP_DEL_IP_ADDRESS; + aa->ifa = ifa; + if (ifa) + atomic_add_int(&ifa->refcount, 1); + /* correlation_id filled in during send routine later... */ + switch (sa->sa_family) { +#ifdef INET6 + case AF_INET6: + { + /* IPv6 address */ + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)sa; + aa->ap.addrp.ph.param_type = SCTP_IPV6_ADDRESS; + aa->ap.addrp.ph.param_length = (sizeof(struct sctp_ipv6addr_param)); + aa->ap.aph.ph.param_length = sizeof(struct sctp_asconf_paramhdr) + sizeof(struct sctp_ipv6addr_param); + memcpy(&aa->ap.addrp.addr, &sin6->sin6_addr, + sizeof(struct in6_addr)); + break; + } +#endif +#ifdef INET + case AF_INET: + { + /* IPv4 address */ + struct sockaddr_in *sin = (struct sockaddr_in *)sa; + + aa->ap.addrp.ph.param_type = SCTP_IPV4_ADDRESS; + aa->ap.addrp.ph.param_length = (sizeof(struct sctp_ipv4addr_param)); + aa->ap.aph.ph.param_length = sizeof(struct sctp_asconf_paramhdr) + sizeof(struct sctp_ipv4addr_param); + memcpy(&aa->ap.addrp.addr, &sin->sin_addr, + sizeof(struct in_addr)); + break; + } +#endif + default: + /* invalid family! */ + SCTP_FREE(aa, SCTP_M_ASC_ADDR); + if (ifa) + sctp_free_ifa(ifa); + return (-1); + } + aa->sent = 0; /* clear sent flag */ + + /* delete goes to the back of the queue */ + TAILQ_INSERT_TAIL(&stcb->asoc.asconf_queue, aa, next); + + /* sa_ignore MEMLEAK {memory is put on the tailq} */ + return (0); +} + +/* + * find a specific asconf param on our "sent" queue + */ +static struct sctp_asconf_addr * +sctp_asconf_find_param(struct sctp_tcb *stcb, uint32_t correlation_id) +{ + struct sctp_asconf_addr *aa; + + TAILQ_FOREACH(aa, &stcb->asoc.asconf_queue, next) { + if (aa->ap.aph.correlation_id == correlation_id && + aa->sent == 1) { + /* found it */ + return (aa); + } + } + /* didn't find it */ + return (NULL); +} + +/* + * process an SCTP_ERROR_CAUSE_IND for a ASCONF-ACK parameter and do + * notifications based on the error response + */ +static void +sctp_asconf_process_error(struct sctp_tcb *stcb SCTP_UNUSED, + struct sctp_asconf_paramhdr *aph) +{ + struct sctp_error_cause *eh; + struct sctp_paramhdr *ph; + uint16_t param_type; + uint16_t error_code; + + eh = (struct sctp_error_cause *)(aph + 1); + ph = (struct sctp_paramhdr *)(eh + 1); + /* validate lengths */ + if (htons(eh->length) + sizeof(struct sctp_error_cause) > + htons(aph->ph.param_length)) { + /* invalid error cause length */ + SCTPDBG(SCTP_DEBUG_ASCONF1, + "asconf_process_error: cause element too long\n"); + return; + } + if (htons(ph->param_length) + sizeof(struct sctp_paramhdr) > + htons(eh->length)) { + /* invalid included TLV length */ + SCTPDBG(SCTP_DEBUG_ASCONF1, + "asconf_process_error: included TLV too long\n"); + return; + } + /* which error code ? */ + error_code = ntohs(eh->code); + param_type = ntohs(aph->ph.param_type); + /* FIX: this should go back up the REMOTE_ERROR ULP notify */ + switch (error_code) { + case SCTP_CAUSE_RESOURCE_SHORTAGE: + /* we allow ourselves to "try again" for this error */ + break; + default: + /* peer can't handle it... */ + switch (param_type) { + case SCTP_ADD_IP_ADDRESS: + case SCTP_DEL_IP_ADDRESS: + case SCTP_SET_PRIM_ADDR: + break; + default: + break; + } + } +} + +/* + * process an asconf queue param. + * aparam: parameter to process, will be removed from the queue. + * flag: 1=success case, 0=failure case + */ +static void +sctp_asconf_process_param_ack(struct sctp_tcb *stcb, + struct sctp_asconf_addr *aparam, uint32_t flag) +{ + uint16_t param_type; + + /* process this param */ + param_type = aparam->ap.aph.ph.param_type; + switch (param_type) { + case SCTP_ADD_IP_ADDRESS: + SCTPDBG(SCTP_DEBUG_ASCONF1, + "process_param_ack: added IP address\n"); + sctp_asconf_addr_mgmt_ack(stcb, aparam->ifa, flag); + break; + case SCTP_DEL_IP_ADDRESS: + SCTPDBG(SCTP_DEBUG_ASCONF1, + "process_param_ack: deleted IP address\n"); + /* nothing really to do... lists already updated */ + break; + case SCTP_SET_PRIM_ADDR: + SCTPDBG(SCTP_DEBUG_ASCONF1, + "process_param_ack: set primary IP address\n"); + /* nothing to do... peer may start using this addr */ + break; + default: + /* should NEVER happen */ + break; + } + + /* remove the param and free it */ + TAILQ_REMOVE(&stcb->asoc.asconf_queue, aparam, next); + if (aparam->ifa) + sctp_free_ifa(aparam->ifa); + SCTP_FREE(aparam, SCTP_M_ASC_ADDR); +} + +/* + * cleanup from a bad asconf ack parameter + */ +static void +sctp_asconf_ack_clear(struct sctp_tcb *stcb SCTP_UNUSED) +{ + /* assume peer doesn't really know how to do asconfs */ + /* XXX we could free the pending queue here */ + +} + +void +sctp_handle_asconf_ack(struct mbuf *m, int offset, + struct sctp_asconf_ack_chunk *cp, struct sctp_tcb *stcb, + struct sctp_nets *net, int *abort_no_unlock) +{ + struct sctp_association *asoc; + uint32_t serial_num; + uint16_t ack_length; + struct sctp_asconf_paramhdr *aph; + struct sctp_asconf_addr *aa, *aa_next; + uint32_t last_error_id = 0; /* last error correlation id */ + uint32_t id; + struct sctp_asconf_addr *ap; + + /* asconf param buffer */ + uint8_t aparam_buf[SCTP_PARAM_BUFFER_SIZE]; + + /* verify minimum length */ + if (ntohs(cp->ch.chunk_length) < sizeof(struct sctp_asconf_ack_chunk)) { + SCTPDBG(SCTP_DEBUG_ASCONF1, + "handle_asconf_ack: chunk too small = %xh\n", + ntohs(cp->ch.chunk_length)); + return; + } + asoc = &stcb->asoc; + serial_num = ntohl(cp->serial_number); + + /* + * NOTE: we may want to handle this differently- currently, we will + * abort when we get an ack for the expected serial number + 1 (eg. + * we didn't send it), process an ack normally if it is the expected + * serial number, and re-send the previous ack for *ALL* other + * serial numbers + */ + + /* + * if the serial number is the next expected, but I didn't send it, + * abort the asoc, since someone probably just hijacked us... + */ + if (serial_num == (asoc->asconf_seq_out + 1)) { + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + SCTPDBG(SCTP_DEBUG_ASCONF1, "handle_asconf_ack: got unexpected next serial number! Aborting asoc!\n"); + SCTP_SNPRINTF(msg, sizeof(msg), "Never sent serial number %8.8x", serial_num); + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + *abort_no_unlock = 1; + return; + } + if (serial_num != asoc->asconf_seq_out_acked + 1) { + /* got a duplicate/unexpected ASCONF-ACK */ + SCTPDBG(SCTP_DEBUG_ASCONF1, "handle_asconf_ack: got duplicate/unexpected serial number = %xh (expected = %xh)\n", + serial_num, asoc->asconf_seq_out_acked + 1); + return; + } + + if (serial_num == asoc->asconf_seq_out - 1) { + /* stop our timer */ + sctp_timer_stop(SCTP_TIMER_TYPE_ASCONF, stcb->sctp_ep, stcb, NULL, + SCTP_FROM_SCTP_ASCONF + SCTP_LOC_5); + } + + /* process the ASCONF-ACK contents */ + ack_length = ntohs(cp->ch.chunk_length) - + sizeof(struct sctp_asconf_ack_chunk); + offset += sizeof(struct sctp_asconf_ack_chunk); + /* process through all parameters */ + while (ack_length >= sizeof(struct sctp_asconf_paramhdr)) { + unsigned int param_length, param_type; + + /* get pointer to next asconf parameter */ + aph = (struct sctp_asconf_paramhdr *)sctp_m_getptr(m, offset, + sizeof(struct sctp_asconf_paramhdr), aparam_buf); + if (aph == NULL) { + /* can't get an asconf paramhdr */ + sctp_asconf_ack_clear(stcb); + return; + } + param_type = ntohs(aph->ph.param_type); + param_length = ntohs(aph->ph.param_length); + if (param_length > ack_length) { + sctp_asconf_ack_clear(stcb); + return; + } + if (param_length < sizeof(struct sctp_asconf_paramhdr)) { + sctp_asconf_ack_clear(stcb); + return; + } + /* get the complete parameter... */ + if (param_length > sizeof(aparam_buf)) { + SCTPDBG(SCTP_DEBUG_ASCONF1, + "param length (%u) larger than buffer size!\n", param_length); + sctp_asconf_ack_clear(stcb); + return; + } + aph = (struct sctp_asconf_paramhdr *)sctp_m_getptr(m, offset, param_length, aparam_buf); + if (aph == NULL) { + sctp_asconf_ack_clear(stcb); + return; + } + /* correlation_id is transparent to peer, no ntohl needed */ + id = aph->correlation_id; + + switch (param_type) { + case SCTP_ERROR_CAUSE_IND: + last_error_id = id; + /* find the corresponding asconf param in our queue */ + ap = sctp_asconf_find_param(stcb, id); + if (ap == NULL) { + /* hmm... can't find this in our queue! */ + break; + } + /* process the parameter, failed flag */ + sctp_asconf_process_param_ack(stcb, ap, 0); + /* process the error response */ + sctp_asconf_process_error(stcb, aph); + break; + case SCTP_SUCCESS_REPORT: + /* find the corresponding asconf param in our queue */ + ap = sctp_asconf_find_param(stcb, id); + if (ap == NULL) { + /* hmm... can't find this in our queue! */ + break; + } + /* process the parameter, success flag */ + sctp_asconf_process_param_ack(stcb, ap, 1); + break; + default: + break; + } /* switch */ + + /* update remaining ASCONF-ACK message length to process */ + if (ack_length > SCTP_SIZE32(param_length)) { + ack_length -= SCTP_SIZE32(param_length); + } else { + break; + } + offset += SCTP_SIZE32(param_length); + } /* while */ + + /* + * if there are any "sent" params still on the queue, these are + * implicitly "success", or "failed" (if we got an error back) ... + * so process these appropriately + * + * we assume that the correlation_id's are monotonically increasing + * beginning from 1 and that we don't have *that* many outstanding + * at any given time + */ + if (last_error_id == 0) + last_error_id--; /* set to "max" value */ + TAILQ_FOREACH_SAFE(aa, &stcb->asoc.asconf_queue, next, aa_next) { + if (aa->sent == 1) { + /* + * implicitly successful or failed if correlation_id + * < last_error_id, then success else, failure + */ + if (aa->ap.aph.correlation_id < last_error_id) + sctp_asconf_process_param_ack(stcb, aa, 1); + else + sctp_asconf_process_param_ack(stcb, aa, 0); + } else { + /* + * since we always process in order (FIFO queue) if + * we reach one that hasn't been sent, the rest + * should not have been sent either. so, we're + * done... + */ + break; + } + } + + /* update the next sequence number to use */ + asoc->asconf_seq_out_acked++; + /* remove the old ASCONF on our outbound queue */ + sctp_toss_old_asconf(stcb); + if (!TAILQ_EMPTY(&stcb->asoc.asconf_queue)) { +#ifdef SCTP_TIMER_BASED_ASCONF + /* we have more params, so restart our timer */ + sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, stcb->sctp_ep, + stcb, net); +#else + /* we have more params, so send out more */ + sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED); +#endif + } +} + +#ifdef INET6 +static uint32_t +sctp_is_scopeid_in_nets(struct sctp_tcb *stcb, struct sockaddr *sa) +{ + struct sockaddr_in6 *sin6, *net6; + struct sctp_nets *net; + + if (sa->sa_family != AF_INET6) { + /* wrong family */ + return (0); + } + sin6 = (struct sockaddr_in6 *)sa; + if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) == 0) { + /* not link local address */ + return (0); + } + /* hunt through our destination nets list for this scope_id */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if (((struct sockaddr *)(&net->ro._l_addr))->sa_family != + AF_INET6) + continue; + net6 = (struct sockaddr_in6 *)&net->ro._l_addr; + if (IN6_IS_ADDR_LINKLOCAL(&net6->sin6_addr) == 0) + continue; + if (sctp_is_same_scope(sin6, net6)) { + /* found one */ + return (1); + } + } + /* didn't find one */ + return (0); +} +#endif + +/* + * address management functions + */ +static void +sctp_addr_mgmt_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct sctp_ifa *ifa, uint16_t type, int addr_locked) +{ + int status; + + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0 || + sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) { + /* subset bound, no ASCONF allowed case, so ignore */ + return; + } + /* + * note: we know this is not the subset bound, no ASCONF case eg. + * this is boundall or subset bound w/ASCONF allowed + */ + + /* first, make sure that the address is IPv4 or IPv6 and not jailed */ + switch (ifa->address.sa.sa_family) { +#ifdef INET6 + case AF_INET6: +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &ifa->address.sin6.sin6_addr) != 0) { + return; + } +#endif + break; +#endif +#ifdef INET + case AF_INET: +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &ifa->address.sin.sin_addr) != 0) { + return; + } +#endif + break; +#endif + default: + return; + } +#ifdef INET6 + /* make sure we're "allowed" to add this type of addr */ + if (ifa->address.sa.sa_family == AF_INET6) { + /* invalid if we're not a v6 endpoint */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) + return; + /* is the v6 addr really valid ? */ + if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { + return; + } + } +#endif + /* put this address on the "pending/do not use yet" list */ + sctp_add_local_addr_restricted(stcb, ifa); + /* + * check address scope if address is out of scope, don't queue + * anything... note: this would leave the address on both inp and + * asoc lists + */ + switch (ifa->address.sa.sa_family) { +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6; + + sin6 = &ifa->address.sin6; + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + /* we skip unspecified addresses */ + return; + } + if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { + if (stcb->asoc.scope.local_scope == 0) { + return; + } + /* is it the right link local scope? */ + if (sctp_is_scopeid_in_nets(stcb, &ifa->address.sa) == 0) { + return; + } + } + if (stcb->asoc.scope.site_scope == 0 && + IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) { + return; + } + break; + } +#endif +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin; + + /* invalid if we are a v6 only endpoint */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(inp)) + return; + + sin = &ifa->address.sin; + if (sin->sin_addr.s_addr == 0) { + /* we skip unspecified addresses */ + return; + } + if (stcb->asoc.scope.ipv4_local_scope == 0 && + IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { + return; + } + break; + } +#endif + default: + /* else, not AF_INET or AF_INET6, so skip */ + return; + } + + /* queue an asconf for this address add/delete */ + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF)) { + /* does the peer do asconf? */ + if (stcb->asoc.asconf_supported) { + /* queue an asconf for this addr */ + status = sctp_asconf_queue_add(stcb, ifa, type); + + /* + * if queued ok, and in the open state, send out the + * ASCONF. If in the non-open state, these will be + * sent when the state goes open. + */ + if (status == 0 && + ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED))) { +#ifdef SCTP_TIMER_BASED_ASCONF + sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, + stcb, stcb->asoc.primary_destination); +#else + sctp_send_asconf(stcb, NULL, addr_locked); +#endif + } + } + } +} + +int +sctp_asconf_iterator_ep(struct sctp_inpcb *inp, void *ptr, uint32_t val SCTP_UNUSED) +{ + struct sctp_asconf_iterator *asc; + struct sctp_ifa *ifa; + struct sctp_laddr *l; + int cnt_invalid = 0; + + asc = (struct sctp_asconf_iterator *)ptr; + LIST_FOREACH(l, &asc->list_of_work, sctp_nxt_addr) { + ifa = l->ifa; + switch (ifa->address.sa.sa_family) { +#ifdef INET6 + case AF_INET6: + /* invalid if we're not a v6 endpoint */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { + cnt_invalid++; + if (asc->cnt == cnt_invalid) + return (1); + } + break; +#endif +#ifdef INET + case AF_INET: + { + /* invalid if we are a v6 only endpoint */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(inp)) { + cnt_invalid++; + if (asc->cnt == cnt_invalid) + return (1); + } + break; + } +#endif + default: + /* invalid address family */ + cnt_invalid++; + if (asc->cnt == cnt_invalid) + return (1); + } + } + return (0); +} + +static int +sctp_asconf_iterator_ep_end(struct sctp_inpcb *inp, void *ptr, uint32_t val SCTP_UNUSED) +{ + struct sctp_ifa *ifa; + struct sctp_asconf_iterator *asc; + struct sctp_laddr *laddr, *nladdr, *l; + + /* Only for specific case not bound all */ + asc = (struct sctp_asconf_iterator *)ptr; + LIST_FOREACH(l, &asc->list_of_work, sctp_nxt_addr) { + ifa = l->ifa; + if (l->action == SCTP_ADD_IP_ADDRESS) { + LIST_FOREACH(laddr, &inp->sctp_addr_list, + sctp_nxt_addr) { + if (laddr->ifa == ifa) { + laddr->action = 0; + break; + } + } + } else if (l->action == SCTP_DEL_IP_ADDRESS) { + LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) { + /* remove only after all guys are done */ + if (laddr->ifa == ifa) { + sctp_del_local_addr_ep(inp, ifa); + } + } + } + } + return (0); +} + +void +sctp_asconf_iterator_stcb(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + void *ptr, uint32_t val SCTP_UNUSED) +{ + struct sctp_asconf_iterator *asc; + struct sctp_ifa *ifa; + struct sctp_laddr *l; + int cnt_invalid = 0; + int type, status; + int num_queued = 0; + + asc = (struct sctp_asconf_iterator *)ptr; + LIST_FOREACH(l, &asc->list_of_work, sctp_nxt_addr) { + ifa = l->ifa; + type = l->action; + + /* address's vrf_id must be the vrf_id of the assoc */ + if (ifa->vrf_id != stcb->asoc.vrf_id) { + continue; + } + + /* Same checks again for assoc */ + switch (ifa->address.sa.sa_family) { +#ifdef INET6 + case AF_INET6: + { + /* invalid if we're not a v6 endpoint */ + struct sockaddr_in6 *sin6; + + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { + cnt_invalid++; + if (asc->cnt == cnt_invalid) + return; + else + continue; + } + sin6 = &ifa->address.sin6; + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + /* we skip unspecified addresses */ + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sin6->sin6_addr) != 0) { + continue; + } +#endif + if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { + if (stcb->asoc.scope.local_scope == 0) { + continue; + } + /* is it the right link local scope? */ + if (sctp_is_scopeid_in_nets(stcb, &ifa->address.sa) == 0) { + continue; + } + } + break; + } +#endif +#ifdef INET + case AF_INET: + { + /* invalid if we are a v6 only endpoint */ + struct sockaddr_in *sin; + + /* invalid if we are a v6 only endpoint */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(inp)) + continue; + + sin = &ifa->address.sin; + if (sin->sin_addr.s_addr == 0) { + /* we skip unspecified addresses */ + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sin->sin_addr) != 0) { + continue; + } +#endif + if (stcb->asoc.scope.ipv4_local_scope == 0 && + IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { + continue; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(inp)) { + cnt_invalid++; + if (asc->cnt == cnt_invalid) + return; + else + continue; + } + break; + } +#endif + default: + /* invalid address family */ + cnt_invalid++; + if (asc->cnt == cnt_invalid) + return; + else + continue; + } + + if (type == SCTP_ADD_IP_ADDRESS) { + /* prevent this address from being used as a source */ + sctp_add_local_addr_restricted(stcb, ifa); + } else if (type == SCTP_DEL_IP_ADDRESS) { + struct sctp_nets *net; + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + /* delete this address if cached */ + if (net->ro._s_addr == ifa) { + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + net->src_addr_selected = 0; +#if defined(__FreeBSD__) && !defined(__Userspace__) + RO_NHFREE(&net->ro); +#else + if (net->ro.ro_rt) { + RTFREE(net->ro.ro_rt); + net->ro.ro_rt = NULL; + } +#endif + /* + * Now we deleted our src address, + * should we not also now reset the + * cwnd/rto to start as if its a new + * address? + */ + stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net); + net->RTO = 0; + } + } + } else if (type == SCTP_SET_PRIM_ADDR) { + if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { + /* must validate the ifa is in the ep */ + if (sctp_is_addr_in_ep(stcb->sctp_ep, ifa) == 0) { + continue; + } + } else { + /* Need to check scopes for this guy */ + if (sctp_is_address_in_scope(ifa, &stcb->asoc.scope, 0) == 0) { + continue; + } + } + } + /* queue an asconf for this address add/delete */ + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF) && + stcb->asoc.asconf_supported == 1) { + /* queue an asconf for this addr */ + status = sctp_asconf_queue_add(stcb, ifa, type); + /* + * if queued ok, and in the open state, update the + * count of queued params. If in the non-open state, + * these get sent when the assoc goes open. + */ + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + if (status >= 0) { + num_queued++; + } + } + } + } + /* + * If we have queued params in the open state, send out an ASCONF. + */ + if (num_queued > 0) { + sctp_send_asconf(stcb, NULL, SCTP_ADDR_NOT_LOCKED); + } +} + +void +sctp_asconf_iterator_end(void *ptr, uint32_t val SCTP_UNUSED) +{ + struct sctp_asconf_iterator *asc; + struct sctp_ifa *ifa; + struct sctp_laddr *l, *nl; + + asc = (struct sctp_asconf_iterator *)ptr; + LIST_FOREACH_SAFE(l, &asc->list_of_work, sctp_nxt_addr, nl) { + ifa = l->ifa; + if (l->action == SCTP_ADD_IP_ADDRESS) { + /* Clear the defer use flag */ + ifa->localifa_flags &= ~SCTP_ADDR_DEFER_USE; + } + sctp_free_ifa(ifa); + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), l); + SCTP_DECR_LADDR_COUNT(); + } + SCTP_FREE(asc, SCTP_M_ASC_IT); +} + +/* + * sa is the sockaddr to ask the peer to set primary to. + * returns: 0 = completed, -1 = error + */ +int32_t +sctp_set_primary_ip_address_sa(struct sctp_tcb *stcb, struct sockaddr *sa) +{ + uint32_t vrf_id; + struct sctp_ifa *ifa; + + /* find the ifa for the desired set primary */ + vrf_id = stcb->asoc.vrf_id; + ifa = sctp_find_ifa_by_addr(sa, vrf_id, SCTP_ADDR_NOT_LOCKED); + if (ifa == NULL) { + /* Invalid address */ + return (-1); + } + + /* queue an ASCONF:SET_PRIM_ADDR to be sent */ + if (!sctp_asconf_queue_add(stcb, ifa, SCTP_SET_PRIM_ADDR)) { + /* set primary queuing succeeded */ + SCTPDBG(SCTP_DEBUG_ASCONF1, + "set_primary_ip_address_sa: queued on tcb=%p, ", + (void *)stcb); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, sa); + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { +#ifdef SCTP_TIMER_BASED_ASCONF + sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, + stcb->sctp_ep, stcb, + stcb->asoc.primary_destination); +#else + sctp_send_asconf(stcb, NULL, SCTP_ADDR_NOT_LOCKED); +#endif + } + } else { + SCTPDBG(SCTP_DEBUG_ASCONF1, "set_primary_ip_address_sa: failed to add to queue on tcb=%p, ", + (void *)stcb); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, sa); + return (-1); + } + return (0); +} + +int +sctp_is_addr_pending(struct sctp_tcb *stcb, struct sctp_ifa *sctp_ifa) +{ + struct sctp_tmit_chunk *chk, *nchk; + unsigned int offset, asconf_limit; + struct sctp_asconf_chunk *acp; + struct sctp_asconf_paramhdr *aph; + uint8_t aparam_buf[SCTP_PARAM_BUFFER_SIZE]; + struct sctp_paramhdr *ph; + int add_cnt, del_cnt; + uint16_t last_param_type; + + add_cnt = del_cnt = 0; + last_param_type = 0; + TAILQ_FOREACH_SAFE(chk, &stcb->asoc.asconf_send_queue, sctp_next, nchk) { + if (chk->data == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "is_addr_pending: No mbuf data?\n"); + continue; + } + offset = 0; + acp = mtod(chk->data, struct sctp_asconf_chunk *); + offset += sizeof(struct sctp_asconf_chunk); + asconf_limit = ntohs(acp->ch.chunk_length); + ph = (struct sctp_paramhdr *)sctp_m_getptr(chk->data, offset, sizeof(struct sctp_paramhdr), aparam_buf); + if (ph == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "is_addr_pending: couldn't get lookup addr!\n"); + continue; + } + offset += ntohs(ph->param_length); + + aph = (struct sctp_asconf_paramhdr *)sctp_m_getptr(chk->data, offset, sizeof(struct sctp_asconf_paramhdr), aparam_buf); + if (aph == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "is_addr_pending: Empty ASCONF will be sent?\n"); + continue; + } + while (aph != NULL) { + unsigned int param_length, param_type; + + param_type = ntohs(aph->ph.param_type); + param_length = ntohs(aph->ph.param_length); + if (offset + param_length > asconf_limit) { + /* parameter goes beyond end of chunk! */ + break; + } + if (param_length > sizeof(aparam_buf)) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "is_addr_pending: param length (%u) larger than buffer size!\n", param_length); + break; + } + if (param_length <= sizeof(struct sctp_paramhdr)) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "is_addr_pending: param length(%u) too short\n", param_length); + break; + } + + aph = (struct sctp_asconf_paramhdr *)sctp_m_getptr(chk->data, offset, param_length, aparam_buf); + if (aph == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "is_addr_pending: couldn't get entire param\n"); + break; + } + + ph = (struct sctp_paramhdr *)(aph + 1); + if (sctp_addr_match(ph, &sctp_ifa->address.sa) != 0) { + switch (param_type) { + case SCTP_ADD_IP_ADDRESS: + add_cnt++; + break; + case SCTP_DEL_IP_ADDRESS: + del_cnt++; + break; + default: + break; + } + last_param_type = param_type; + } + + offset += SCTP_SIZE32(param_length); + if (offset >= asconf_limit) { + /* no more data in the mbuf chain */ + break; + } + /* get pointer to next asconf param */ + aph = (struct sctp_asconf_paramhdr *)sctp_m_getptr(chk->data, offset, sizeof(struct sctp_asconf_paramhdr), aparam_buf); + } + } + + /* we want to find the sequences which consist of ADD -> DEL -> ADD or DEL -> ADD */ + if (add_cnt > del_cnt || + (add_cnt == del_cnt && last_param_type == SCTP_ADD_IP_ADDRESS)) { + return (1); + } + return (0); +} + +static struct sockaddr * +sctp_find_valid_localaddr(struct sctp_tcb *stcb, int addr_locked) +{ + struct sctp_vrf *vrf = NULL; + struct sctp_ifn *sctp_ifn; + struct sctp_ifa *sctp_ifa; + + if (addr_locked == SCTP_ADDR_NOT_LOCKED) + SCTP_IPI_ADDR_RLOCK(); + vrf = sctp_find_vrf(stcb->asoc.vrf_id); + if (vrf == NULL) { + if (addr_locked == SCTP_ADDR_NOT_LOCKED) + SCTP_IPI_ADDR_RUNLOCK(); + return (NULL); + } + LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { + if (stcb->asoc.scope.loopback_scope == 0 && + SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { + /* Skip if loopback_scope not set */ + continue; + } + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { + switch (sctp_ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: + if (stcb->asoc.scope.ipv4_addr_legal) { + struct sockaddr_in *sin; + + sin = &sctp_ifa->address.sin; + if (sin->sin_addr.s_addr == 0) { + /* skip unspecified addresses */ + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred, + &sin->sin_addr) != 0) { + continue; + } +#endif + if (stcb->asoc.scope.ipv4_local_scope == 0 && + IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) + continue; + + if (sctp_is_addr_restricted(stcb, sctp_ifa) && + (!sctp_is_addr_pending(stcb, sctp_ifa))) + continue; + /* found a valid local v4 address to use */ + if (addr_locked == SCTP_ADDR_NOT_LOCKED) + SCTP_IPI_ADDR_RUNLOCK(); + return (&sctp_ifa->address.sa); + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (stcb->asoc.scope.ipv6_addr_legal) { + struct sockaddr_in6 *sin6; + + if (sctp_ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { + continue; + } + + sin6 = &sctp_ifa->address.sin6; + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + /* we skip unspecified addresses */ + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred, + &sin6->sin6_addr) != 0) { + continue; + } +#endif + if (stcb->asoc.scope.local_scope == 0 && + IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) + continue; + if (stcb->asoc.scope.site_scope == 0 && + IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) + continue; + + if (sctp_is_addr_restricted(stcb, sctp_ifa) && + (!sctp_is_addr_pending(stcb, sctp_ifa))) + continue; + /* found a valid local v6 address to use */ + if (addr_locked == SCTP_ADDR_NOT_LOCKED) + SCTP_IPI_ADDR_RUNLOCK(); + return (&sctp_ifa->address.sa); + } + break; +#endif + default: + break; + } + } + } + /* no valid addresses found */ + if (addr_locked == SCTP_ADDR_NOT_LOCKED) + SCTP_IPI_ADDR_RUNLOCK(); + return (NULL); +} + +static struct sockaddr * +sctp_find_valid_localaddr_ep(struct sctp_tcb *stcb) +{ + struct sctp_laddr *laddr; + + LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa == NULL) { + continue; + } + /* is the address restricted ? */ + if (sctp_is_addr_restricted(stcb, laddr->ifa) && + (!sctp_is_addr_pending(stcb, laddr->ifa))) + continue; + + /* found a valid local address to use */ + return (&laddr->ifa->address.sa); + } + /* no valid addresses found */ + return (NULL); +} + +/* + * builds an ASCONF chunk from queued ASCONF params. + * returns NULL on error (no mbuf, no ASCONF params queued, etc). + */ +struct mbuf * +sctp_compose_asconf(struct sctp_tcb *stcb, int *retlen, int addr_locked) +{ + struct mbuf *m_asconf, *m_asconf_chk; + struct sctp_asconf_addr *aa; + struct sctp_asconf_chunk *acp; + struct sctp_asconf_paramhdr *aph; + struct sctp_asconf_addr_param *aap; + uint32_t p_length, overhead; + uint32_t correlation_id = 1; /* 0 is reserved... */ + caddr_t ptr, lookup_ptr; + uint8_t lookup_used = 0; + + /* are there any asconf params to send? */ + TAILQ_FOREACH(aa, &stcb->asoc.asconf_queue, next) { + if (aa->sent == 0) + break; + } + if (aa == NULL) + return (NULL); + + /* Consider IP header and SCTP common header. */ + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + overhead = SCTP_MIN_OVERHEAD; + } else { + overhead = SCTP_MIN_V4_OVERHEAD; + } + /* Consider ASONF chunk. */ + overhead += sizeof(struct sctp_asconf_chunk); + /* Consider AUTH chunk. */ + overhead += sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); + if (stcb->asoc.smallest_mtu <= overhead) { + /* MTU too small. */ + return (NULL); + } + /* + * get a chunk header mbuf and a cluster for the asconf params since + * it's simpler to fill in the asconf chunk header lookup address on + * the fly + */ + m_asconf_chk = sctp_get_mbuf_for_msg(sizeof(struct sctp_asconf_chunk), 0, M_NOWAIT, 1, MT_DATA); + if (m_asconf_chk == NULL) { + /* no mbuf's */ + SCTPDBG(SCTP_DEBUG_ASCONF1, + "sctp_compose_asconf: couldn't get chunk mbuf!\n"); + return (NULL); + } + m_asconf = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_NOWAIT, 1, MT_DATA); + if (m_asconf == NULL) { + /* no mbuf's */ + SCTPDBG(SCTP_DEBUG_ASCONF1, + "sctp_compose_asconf: couldn't get mbuf!\n"); + sctp_m_freem(m_asconf_chk); + return (NULL); + } + SCTP_BUF_LEN(m_asconf_chk) = sizeof(struct sctp_asconf_chunk); + SCTP_BUF_LEN(m_asconf) = 0; + acp = mtod(m_asconf_chk, struct sctp_asconf_chunk *); + memset(acp, 0, sizeof(struct sctp_asconf_chunk)); + /* save pointers to lookup address and asconf params */ + lookup_ptr = (caddr_t)(acp + 1); /* after the header */ + ptr = mtod(m_asconf, caddr_t); /* beginning of cluster */ + + /* fill in chunk header info */ + acp->ch.chunk_type = SCTP_ASCONF; + acp->ch.chunk_flags = 0; + acp->serial_number = htonl(stcb->asoc.asconf_seq_out); + stcb->asoc.asconf_seq_out++; + + /* add parameters... up to smallest MTU allowed */ + TAILQ_FOREACH(aa, &stcb->asoc.asconf_queue, next) { + if (aa->sent) + continue; + /* get the parameter length */ + p_length = SCTP_SIZE32(aa->ap.aph.ph.param_length); + /* will it fit in current chunk? */ + if ((SCTP_BUF_LEN(m_asconf) + p_length > stcb->asoc.smallest_mtu - overhead) || + (SCTP_BUF_LEN(m_asconf) + p_length > MCLBYTES)) { + /* won't fit, so we're done with this chunk */ + break; + } + /* assign (and store) a correlation id */ + aa->ap.aph.correlation_id = correlation_id++; + + /* + * fill in address if we're doing a delete this is a simple + * way for us to fill in the correlation address, which + * should only be used by the peer if we're deleting our + * source address and adding a new address (e.g. renumbering + * case) + */ + if (lookup_used == 0 && + (aa->special_del == 0) && + aa->ap.aph.ph.param_type == SCTP_DEL_IP_ADDRESS) { + struct sctp_ipv6addr_param *lookup; + uint16_t p_size, addr_size; + + lookup = (struct sctp_ipv6addr_param *)lookup_ptr; + lookup->ph.param_type = + htons(aa->ap.addrp.ph.param_type); + if (aa->ap.addrp.ph.param_type == SCTP_IPV6_ADDRESS) { + /* copy IPv6 address */ + p_size = sizeof(struct sctp_ipv6addr_param); + addr_size = sizeof(struct in6_addr); + } else { + /* copy IPv4 address */ + p_size = sizeof(struct sctp_ipv4addr_param); + addr_size = sizeof(struct in_addr); + } + lookup->ph.param_length = htons(SCTP_SIZE32(p_size)); + memcpy(lookup->addr, &aa->ap.addrp.addr, addr_size); + SCTP_BUF_LEN(m_asconf_chk) += SCTP_SIZE32(p_size); + lookup_used = 1; + } + /* copy into current space */ + memcpy(ptr, &aa->ap, p_length); + + /* network elements and update lengths */ + aph = (struct sctp_asconf_paramhdr *)ptr; + aap = (struct sctp_asconf_addr_param *)ptr; + /* correlation_id is transparent to peer, no htonl needed */ + aph->ph.param_type = htons(aph->ph.param_type); + aph->ph.param_length = htons(aph->ph.param_length); + aap->addrp.ph.param_type = htons(aap->addrp.ph.param_type); + aap->addrp.ph.param_length = htons(aap->addrp.ph.param_length); + + SCTP_BUF_LEN(m_asconf) += SCTP_SIZE32(p_length); + ptr += SCTP_SIZE32(p_length); + + /* + * these params are removed off the pending list upon + * getting an ASCONF-ACK back from the peer, just set flag + */ + aa->sent = 1; + } + /* check to see if the lookup addr has been populated yet */ + if (lookup_used == 0) { + /* NOTE: if the address param is optional, can skip this... */ + /* add any valid (existing) address... */ + struct sctp_ipv6addr_param *lookup; + uint16_t p_size, addr_size; + struct sockaddr *found_addr; + caddr_t addr_ptr; + + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) + found_addr = sctp_find_valid_localaddr(stcb, + addr_locked); + else + found_addr = sctp_find_valid_localaddr_ep(stcb); + + lookup = (struct sctp_ipv6addr_param *)lookup_ptr; + if (found_addr != NULL) { + switch (found_addr->sa_family) { +#ifdef INET6 + case AF_INET6: + /* copy IPv6 address */ + lookup->ph.param_type = + htons(SCTP_IPV6_ADDRESS); + p_size = sizeof(struct sctp_ipv6addr_param); + addr_size = sizeof(struct in6_addr); + addr_ptr = (caddr_t)&((struct sockaddr_in6 *) + found_addr)->sin6_addr; + break; +#endif +#ifdef INET + case AF_INET: + /* copy IPv4 address */ + lookup->ph.param_type = + htons(SCTP_IPV4_ADDRESS); + p_size = sizeof(struct sctp_ipv4addr_param); + addr_size = sizeof(struct in_addr); + addr_ptr = (caddr_t)&((struct sockaddr_in *) + found_addr)->sin_addr; + break; +#endif + default: + SCTPDBG(SCTP_DEBUG_ASCONF1, + "sctp_compose_asconf: no usable lookup addr (family = %d)!\n", + found_addr->sa_family); + sctp_m_freem(m_asconf_chk); + sctp_m_freem(m_asconf); + return (NULL); + } + lookup->ph.param_length = htons(SCTP_SIZE32(p_size)); + memcpy(lookup->addr, addr_ptr, addr_size); + SCTP_BUF_LEN(m_asconf_chk) += SCTP_SIZE32(p_size); + } else { + /* uh oh... don't have any address?? */ + SCTPDBG(SCTP_DEBUG_ASCONF1, + "sctp_compose_asconf: no lookup addr!\n"); + sctp_m_freem(m_asconf_chk); + sctp_m_freem(m_asconf); + return (NULL); + } + } + /* chain it all together */ + SCTP_BUF_NEXT(m_asconf_chk) = m_asconf; + *retlen = SCTP_BUF_LEN(m_asconf_chk) + SCTP_BUF_LEN(m_asconf); + acp->ch.chunk_length = htons(*retlen); + + return (m_asconf_chk); +} + +/* + * section to handle address changes before an association is up eg. changes + * during INIT/INIT-ACK/COOKIE-ECHO handshake + */ + +/* + * processes the (local) addresses in the INIT-ACK chunk + */ +static void +sctp_process_initack_addresses(struct sctp_tcb *stcb, struct mbuf *m, + unsigned int offset, unsigned int length) +{ + struct sctp_paramhdr tmp_param, *ph; + uint16_t plen, ptype; + struct sctp_ifa *sctp_ifa; + union sctp_sockstore store; +#ifdef INET6 + struct sctp_ipv6addr_param addr6_store; +#endif +#ifdef INET + struct sctp_ipv4addr_param addr4_store; +#endif + + SCTPDBG(SCTP_DEBUG_ASCONF2, "processing init-ack addresses\n"); + if (stcb == NULL) /* Un-needed check for SA */ + return; + + /* convert to upper bound */ + length += offset; + + if ((offset + sizeof(struct sctp_paramhdr)) > length) { + return; + } + /* go through the addresses in the init-ack */ + ph = (struct sctp_paramhdr *) + sctp_m_getptr(m, offset, sizeof(struct sctp_paramhdr), + (uint8_t *)&tmp_param); + while (ph != NULL) { + ptype = ntohs(ph->param_type); + plen = ntohs(ph->param_length); + switch (ptype) { +#ifdef INET6 + case SCTP_IPV6_ADDRESS: + { + struct sctp_ipv6addr_param *a6p; + + /* get the entire IPv6 address param */ + a6p = (struct sctp_ipv6addr_param *) + sctp_m_getptr(m, offset, + sizeof(struct sctp_ipv6addr_param), + (uint8_t *)&addr6_store); + if (plen != sizeof(struct sctp_ipv6addr_param) || + a6p == NULL) { + return; + } + memset(&store, 0, sizeof(union sctp_sockstore)); + store.sin6.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + store.sin6.sin6_len = sizeof(struct sockaddr_in6); +#endif + store.sin6.sin6_port = stcb->rport; + memcpy(&store.sin6.sin6_addr, a6p->addr, sizeof(struct in6_addr)); + break; + } +#endif +#ifdef INET + case SCTP_IPV4_ADDRESS: + { + struct sctp_ipv4addr_param *a4p; + + /* get the entire IPv4 address param */ + a4p = (struct sctp_ipv4addr_param *)sctp_m_getptr(m, offset, + sizeof(struct sctp_ipv4addr_param), + (uint8_t *)&addr4_store); + if (plen != sizeof(struct sctp_ipv4addr_param) || + a4p == NULL) { + return; + } + memset(&store, 0, sizeof(union sctp_sockstore)); + store.sin.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + store.sin.sin_len = sizeof(struct sockaddr_in); +#endif + store.sin.sin_port = stcb->rport; + store.sin.sin_addr.s_addr = a4p->addr; + break; + } +#endif + default: + goto next_addr; + } + + /* see if this address really (still) exists */ + sctp_ifa = sctp_find_ifa_by_addr(&store.sa, stcb->asoc.vrf_id, + SCTP_ADDR_NOT_LOCKED); + if (sctp_ifa == NULL) { + /* address doesn't exist anymore */ + int status; + + /* are ASCONFs allowed ? */ + if ((sctp_is_feature_on(stcb->sctp_ep, + SCTP_PCB_FLAGS_DO_ASCONF)) && + stcb->asoc.asconf_supported) { + /* queue an ASCONF DEL_IP_ADDRESS */ + status = sctp_asconf_queue_sa_delete(stcb, &store.sa); + /* + * if queued ok, and in correct state, send + * out the ASCONF. + */ + if (status == 0 && + SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) { +#ifdef SCTP_TIMER_BASED_ASCONF + sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, + stcb->sctp_ep, stcb, + stcb->asoc.primary_destination); +#else + sctp_send_asconf(stcb, NULL, SCTP_ADDR_NOT_LOCKED); +#endif + } + } + } + +next_addr: + /* + * Sanity check: Make sure the length isn't 0, otherwise + * we'll be stuck in this loop for a long time... + */ + if (SCTP_SIZE32(plen) == 0) { + SCTP_PRINTF("process_initack_addrs: bad len (%d) type=%xh\n", + plen, ptype); + return; + } + /* get next parameter */ + offset += SCTP_SIZE32(plen); + if ((offset + sizeof(struct sctp_paramhdr)) > length) + return; + ph = (struct sctp_paramhdr *)sctp_m_getptr(m, offset, + sizeof(struct sctp_paramhdr), (uint8_t *)&tmp_param); + } /* while */ +} + +/* FIX ME: need to verify return result for v6 address type if v6 disabled */ +/* + * checks to see if a specific address is in the initack address list returns + * 1 if found, 0 if not + */ +static uint32_t +sctp_addr_in_initack(struct mbuf *m, uint32_t offset, uint32_t length, struct sockaddr *sa) +{ + struct sctp_paramhdr tmp_param, *ph; + uint16_t plen, ptype; +#ifdef INET + struct sockaddr_in *sin; + struct sctp_ipv4addr_param *a4p; + struct sctp_ipv6addr_param addr4_store; +#endif +#ifdef INET6 + struct sockaddr_in6 *sin6; + struct sctp_ipv6addr_param *a6p; + struct sctp_ipv6addr_param addr6_store; +#ifdef SCTP_EMBEDDED_V6_SCOPE + struct sockaddr_in6 sin6_tmp; +#endif +#endif + + switch (sa->sa_family) { +#ifdef INET + case AF_INET: + break; +#endif +#ifdef INET6 + case AF_INET6: + break; +#endif + default: + return (0); + } + + SCTPDBG(SCTP_DEBUG_ASCONF2, "find_initack_addr: starting search for "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF2, sa); + /* convert to upper bound */ + length += offset; + + if ((offset + sizeof(struct sctp_paramhdr)) > length) { + SCTPDBG(SCTP_DEBUG_ASCONF1, + "find_initack_addr: invalid offset?\n"); + return (0); + } + /* go through the addresses in the init-ack */ + ph = (struct sctp_paramhdr *)sctp_m_getptr(m, offset, + sizeof(struct sctp_paramhdr), (uint8_t *) & tmp_param); + while (ph != NULL) { + ptype = ntohs(ph->param_type); + plen = ntohs(ph->param_length); + switch (ptype) { +#ifdef INET6 + case SCTP_IPV6_ADDRESS: + if (sa->sa_family == AF_INET6) { + /* get the entire IPv6 address param */ + if (plen != sizeof(struct sctp_ipv6addr_param)) { + break; + } + /* get the entire IPv6 address param */ + a6p = (struct sctp_ipv6addr_param *) + sctp_m_getptr(m, offset, + sizeof(struct sctp_ipv6addr_param), + (uint8_t *)&addr6_store); + if (a6p == NULL) { + return (0); + } + sin6 = (struct sockaddr_in6 *)sa; +#ifdef SCTP_EMBEDDED_V6_SCOPE + if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) { + /* create a copy and clear scope */ + memcpy(&sin6_tmp, sin6, + sizeof(struct sockaddr_in6)); + sin6 = &sin6_tmp; + in6_clearscope(&sin6->sin6_addr); + } +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + if (memcmp(&sin6->sin6_addr, a6p->addr, + sizeof(struct in6_addr)) == 0) { + /* found it */ + return (1); + } + } + break; +#endif /* INET6 */ +#ifdef INET + case SCTP_IPV4_ADDRESS: + if (sa->sa_family == AF_INET) { + if (plen != sizeof(struct sctp_ipv4addr_param)) { + break; + } + /* get the entire IPv4 address param */ + a4p = (struct sctp_ipv4addr_param *) + sctp_m_getptr(m, offset, + sizeof(struct sctp_ipv4addr_param), + (uint8_t *)&addr4_store); + if (a4p == NULL) { + return (0); + } + sin = (struct sockaddr_in *)sa; + if (sin->sin_addr.s_addr == a4p->addr) { + /* found it */ + return (1); + } + } + break; +#endif + default: + break; + } + /* get next parameter */ + offset += SCTP_SIZE32(plen); + if (offset + sizeof(struct sctp_paramhdr) > length) { + return (0); + } + ph = (struct sctp_paramhdr *) + sctp_m_getptr(m, offset, sizeof(struct sctp_paramhdr), + (uint8_t *) & tmp_param); + } /* while */ + /* not found! */ + return (0); +} + +/* + * makes sure that the current endpoint local addr list is consistent with + * the new association (eg. subset bound, asconf allowed) adds addresses as + * necessary + */ +static void +sctp_check_address_list_ep(struct sctp_tcb *stcb, struct mbuf *m, int offset, + int length, struct sockaddr *init_addr) +{ + struct sctp_laddr *laddr; + + /* go through the endpoint list */ + LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) { + /* be paranoid and validate the laddr */ + if (laddr->ifa == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, + "check_addr_list_ep: laddr->ifa is NULL"); + continue; + } + /* do i have it implicitly? */ + if (sctp_cmpaddr(&laddr->ifa->address.sa, init_addr)) { + continue; + } + /* check to see if in the init-ack */ + if (!sctp_addr_in_initack(m, offset, length, &laddr->ifa->address.sa)) { + /* try to add it */ + sctp_addr_mgmt_assoc(stcb->sctp_ep, stcb, laddr->ifa, + SCTP_ADD_IP_ADDRESS, SCTP_ADDR_NOT_LOCKED); + } + } +} + +/* + * makes sure that the current kernel address list is consistent with the new + * association (with all addrs bound) adds addresses as necessary + */ +static void +sctp_check_address_list_all(struct sctp_tcb *stcb, struct mbuf *m, int offset, + int length, struct sockaddr *init_addr, + uint16_t local_scope, uint16_t site_scope, + uint16_t ipv4_scope, uint16_t loopback_scope) +{ + struct sctp_vrf *vrf = NULL; + struct sctp_ifn *sctp_ifn; + struct sctp_ifa *sctp_ifa; + uint32_t vrf_id; +#ifdef INET + struct sockaddr_in *sin; +#endif +#ifdef INET6 + struct sockaddr_in6 *sin6; +#endif + + if (stcb) { + vrf_id = stcb->asoc.vrf_id; + } else { + return; + } + SCTP_IPI_ADDR_RLOCK(); + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) { + SCTP_IPI_ADDR_RUNLOCK(); + return; + } + /* go through all our known interfaces */ + LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { + if (loopback_scope == 0 && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { + /* skip loopback interface */ + continue; + } + /* go through each interface address */ + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { + /* do i have it implicitly? */ + if (sctp_cmpaddr(&sctp_ifa->address.sa, init_addr)) { + continue; + } + switch (sctp_ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: + sin = &sctp_ifa->address.sin; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred, + &sin->sin_addr) != 0) { + continue; + } +#endif + if ((ipv4_scope == 0) && + (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { + /* private address not in scope */ + continue; + } + break; +#endif +#ifdef INET6 + case AF_INET6: + sin6 = &sctp_ifa->address.sin6; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred, + &sin6->sin6_addr) != 0) { + continue; + } +#endif + if ((local_scope == 0) && + (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr))) { + continue; + } + if ((site_scope == 0) && + (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { + continue; + } + break; +#endif + default: + break; + } + /* check to see if in the init-ack */ + if (!sctp_addr_in_initack(m, offset, length, &sctp_ifa->address.sa)) { + /* try to add it */ + sctp_addr_mgmt_assoc(stcb->sctp_ep, stcb, + sctp_ifa, SCTP_ADD_IP_ADDRESS, + SCTP_ADDR_LOCKED); + } + } /* end foreach ifa */ + } /* end foreach ifn */ + SCTP_IPI_ADDR_RUNLOCK(); +} + +/* + * validates an init-ack chunk (from a cookie-echo) with current addresses + * adds addresses from the init-ack into our local address list, if needed + * queues asconf adds/deletes addresses as needed and makes appropriate list + * changes for source address selection m, offset: points to the start of the + * address list in an init-ack chunk length: total length of the address + * params only init_addr: address where my INIT-ACK was sent from + */ +void +sctp_check_address_list(struct sctp_tcb *stcb, struct mbuf *m, int offset, + int length, struct sockaddr *init_addr, + uint16_t local_scope, uint16_t site_scope, + uint16_t ipv4_scope, uint16_t loopback_scope) +{ + /* process the local addresses in the initack */ + sctp_process_initack_addresses(stcb, m, offset, length); + + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + /* bound all case */ + sctp_check_address_list_all(stcb, m, offset, length, init_addr, + local_scope, site_scope, ipv4_scope, loopback_scope); + } else { + /* subset bound case */ + if (sctp_is_feature_on(stcb->sctp_ep, + SCTP_PCB_FLAGS_DO_ASCONF)) { + /* asconf's allowed */ + sctp_check_address_list_ep(stcb, m, offset, length, + init_addr); + } + /* else, no asconfs allowed, so what we sent is what we get */ + } +} + +/* + * sctp_bindx() support + */ +uint32_t +sctp_addr_mgmt_ep_sa(struct sctp_inpcb *inp, struct sockaddr *sa, + uint32_t type, uint32_t vrf_id) +{ + struct sctp_ifa *ifa; + struct sctp_laddr *laddr, *nladdr; + +#ifdef HAVE_SA_LEN + if (sa->sa_len == 0) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_ASCONF, EINVAL); + return (EINVAL); + } +#endif + if (type == SCTP_ADD_IP_ADDRESS) { + /* For an add the address MUST be on the system */ + ifa = sctp_find_ifa_by_addr(sa, vrf_id, SCTP_ADDR_NOT_LOCKED); + } else if (type == SCTP_DEL_IP_ADDRESS) { + /* For a delete we need to find it in the inp */ + ifa = sctp_find_ifa_in_ep(inp, sa, SCTP_ADDR_NOT_LOCKED); + } else { + ifa = NULL; + } + if (ifa != NULL) { + if (type == SCTP_ADD_IP_ADDRESS) { + sctp_add_local_addr_ep(inp, ifa, type); + } else if (type == SCTP_DEL_IP_ADDRESS) { + if (inp->laddr_count < 2) { + /* can't delete the last local address */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_ASCONF, EINVAL); + return (EINVAL); + } + LIST_FOREACH(laddr, &inp->sctp_addr_list, + sctp_nxt_addr) { + if (ifa == laddr->ifa) { + /* Mark in the delete */ + laddr->action = type; + } + } + } + if (LIST_EMPTY(&inp->sctp_asoc_list)) { + /* + * There is no need to start the iterator if + * the inp has no associations. + */ + if (type == SCTP_DEL_IP_ADDRESS) { + LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) { + if (laddr->ifa == ifa) { + sctp_del_local_addr_ep(inp, ifa); + } + } + } + } else { + struct sctp_asconf_iterator *asc; + struct sctp_laddr *wi; + int ret; + + SCTP_MALLOC(asc, struct sctp_asconf_iterator *, + sizeof(struct sctp_asconf_iterator), + SCTP_M_ASC_IT); + if (asc == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_ASCONF, ENOMEM); + return (ENOMEM); + } + wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); + if (wi == NULL) { + SCTP_FREE(asc, SCTP_M_ASC_IT); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_ASCONF, ENOMEM); + return (ENOMEM); + } + LIST_INIT(&asc->list_of_work); + asc->cnt = 1; + SCTP_INCR_LADDR_COUNT(); + wi->ifa = ifa; + wi->action = type; + atomic_add_int(&ifa->refcount, 1); + LIST_INSERT_HEAD(&asc->list_of_work, wi, sctp_nxt_addr); + ret = sctp_initiate_iterator(sctp_asconf_iterator_ep, + sctp_asconf_iterator_stcb, + sctp_asconf_iterator_ep_end, + SCTP_PCB_ANY_FLAGS, + SCTP_PCB_ANY_FEATURES, + SCTP_ASOC_ANY_STATE, + (void *)asc, 0, + sctp_asconf_iterator_end, inp, 0); + if (ret) { + SCTP_PRINTF("Failed to initiate iterator for addr_mgmt_ep_sa\n"); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_ASCONF, EFAULT); + sctp_asconf_iterator_end(asc, 0); + return (EFAULT); + } + } + return (0); + } else { + /* invalid address! */ + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_ASCONF, EADDRNOTAVAIL); + return (EADDRNOTAVAIL); + } +} + +void +sctp_asconf_send_nat_state_update(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + struct sctp_asconf_addr *aa_vtag, *aa_add, *aa_del; + struct sctp_ifa *sctp_ifap; + struct sctp_asconf_tag_param *vtag; +#ifdef INET + struct sockaddr_in *to; +#endif +#ifdef INET6 + struct sockaddr_in6 *to6; +#endif + + if (net == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "sctp_asconf_send_nat_state_update: Missing net\n"); + return; + } + if (stcb == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "sctp_asconf_send_nat_state_update: Missing stcb\n"); + return; + } + /* Need to have in the ASCONF: + * - VTAG(my_vtag/peer_vtag) + * - ADD(wildcard) + * - DEL(wildcard) + * - ADD(Any global addresses) + */ + SCTP_MALLOC(aa_vtag, struct sctp_asconf_addr *, sizeof(struct sctp_asconf_addr), SCTP_M_ASC_ADDR); + SCTP_MALLOC(aa_add, struct sctp_asconf_addr *, sizeof(struct sctp_asconf_addr), SCTP_M_ASC_ADDR); + SCTP_MALLOC(aa_del, struct sctp_asconf_addr *, sizeof(struct sctp_asconf_addr), SCTP_M_ASC_ADDR); + + if ((aa_vtag == NULL) || (aa_add == NULL) || (aa_del == NULL)) { + /* Didn't get memory */ + SCTPDBG(SCTP_DEBUG_ASCONF1, "sctp_asconf_send_nat_state_update: failed to get memory!\n"); +out: + if (aa_vtag != NULL) { + SCTP_FREE(aa_vtag, SCTP_M_ASC_ADDR); + } + if (aa_add != NULL) { + SCTP_FREE(aa_add, SCTP_M_ASC_ADDR); + } + if (aa_del != NULL) { + SCTP_FREE(aa_del, SCTP_M_ASC_ADDR); + } + return; + } + memset(aa_vtag, 0, sizeof(struct sctp_asconf_addr)); + aa_vtag->special_del = 0; + /* Fill in ASCONF address parameter fields. */ + /* Top level elements are "networked" during send. */ + aa_vtag->ifa = NULL; + aa_vtag->sent = 0; /* clear sent flag */ + vtag = (struct sctp_asconf_tag_param *)&aa_vtag->ap.aph; + vtag->aph.ph.param_type = SCTP_NAT_VTAGS; + vtag->aph.ph.param_length = sizeof(struct sctp_asconf_tag_param); + vtag->local_vtag = htonl(stcb->asoc.my_vtag); + vtag->remote_vtag = htonl(stcb->asoc.peer_vtag); + + memset(aa_add, 0, sizeof(struct sctp_asconf_addr)); + memset(aa_del, 0, sizeof(struct sctp_asconf_addr)); + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + aa_add->ap.aph.ph.param_type = SCTP_ADD_IP_ADDRESS; + aa_add->ap.aph.ph.param_length = sizeof(struct sctp_asconf_addrv4_param); + aa_add->ap.addrp.ph.param_type = SCTP_IPV4_ADDRESS; + aa_add->ap.addrp.ph.param_length = sizeof (struct sctp_ipv4addr_param); + /* No need to fill the address, we are using 0.0.0.0 */ + aa_del->ap.aph.ph.param_type = SCTP_DEL_IP_ADDRESS; + aa_del->ap.aph.ph.param_length = sizeof(struct sctp_asconf_addrv4_param); + aa_del->ap.addrp.ph.param_type = SCTP_IPV4_ADDRESS; + aa_del->ap.addrp.ph.param_length = sizeof (struct sctp_ipv4addr_param); + /* No need to fill the address, we are using 0.0.0.0 */ + break; +#endif +#ifdef INET6 + case AF_INET6: + aa_add->ap.aph.ph.param_type = SCTP_ADD_IP_ADDRESS; + aa_add->ap.aph.ph.param_length = sizeof(struct sctp_asconf_addr_param); + aa_add->ap.addrp.ph.param_type = SCTP_IPV6_ADDRESS; + aa_add->ap.addrp.ph.param_length = sizeof (struct sctp_ipv6addr_param); + /* No need to fill the address, we are using ::0 */ + aa_del->ap.aph.ph.param_type = SCTP_DEL_IP_ADDRESS; + aa_del->ap.aph.ph.param_length = sizeof(struct sctp_asconf_addr_param); + aa_del->ap.addrp.ph.param_type = SCTP_IPV6_ADDRESS; + aa_del->ap.addrp.ph.param_length = sizeof (struct sctp_ipv6addr_param); + /* No need to fill the address, we are using ::0 */ + break; +#endif + default: + SCTPDBG(SCTP_DEBUG_ASCONF1, + "sctp_asconf_send_nat_state_update: unknown address family %d\n", + net->ro._l_addr.sa.sa_family); + goto out; + } + TAILQ_INSERT_TAIL(&stcb->asoc.asconf_queue, aa_vtag, next); + TAILQ_INSERT_TAIL(&stcb->asoc.asconf_queue, aa_add, next); + TAILQ_INSERT_TAIL(&stcb->asoc.asconf_queue, aa_del, next); + + /* Now we must hunt the addresses and add all global addresses */ + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + struct sctp_vrf *vrf = NULL; + struct sctp_ifn *sctp_ifnp; + uint32_t vrf_id; + + vrf_id = stcb->sctp_ep->def_vrf_id; + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) { + goto skip_rest; + } + + SCTP_IPI_ADDR_RLOCK(); + LIST_FOREACH(sctp_ifnp, &vrf->ifnlist, next_ifn) { + LIST_FOREACH(sctp_ifap, &sctp_ifnp->ifalist, next_ifa) { + switch (sctp_ifap->address.sa.sa_family) { +#ifdef INET + case AF_INET: + to = &sctp_ifap->address.sin; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred, + &to->sin_addr) != 0) { + continue; + } +#endif + if (IN4_ISPRIVATE_ADDRESS(&to->sin_addr)) { + continue; + } + if (IN4_ISLOOPBACK_ADDRESS(&to->sin_addr)) { + continue; + } + break; +#endif +#ifdef INET6 + case AF_INET6: + to6 = &sctp_ifap->address.sin6; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred, + &to6->sin6_addr) != 0) { + continue; + } +#endif + if (IN6_IS_ADDR_LOOPBACK(&to6->sin6_addr)) { + continue; + } + if (IN6_IS_ADDR_LINKLOCAL(&to6->sin6_addr)) { + continue; + } + break; +#endif + default: + continue; + } + sctp_asconf_queue_mgmt(stcb, sctp_ifap, SCTP_ADD_IP_ADDRESS); + } + } + SCTP_IPI_ADDR_RUNLOCK(); + } else { + struct sctp_laddr *laddr; + + LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa == NULL) { + continue; + } + if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) + /* Address being deleted by the system, dont + * list. + */ + continue; + if (laddr->action == SCTP_DEL_IP_ADDRESS) { + /* Address being deleted on this ep + * don't list. + */ + continue; + } + sctp_ifap = laddr->ifa; + switch (sctp_ifap->address.sa.sa_family) { +#ifdef INET + case AF_INET: + to = &sctp_ifap->address.sin; + if (IN4_ISPRIVATE_ADDRESS(&to->sin_addr)) { + continue; + } + if (IN4_ISLOOPBACK_ADDRESS(&to->sin_addr)) { + continue; + } + break; +#endif +#ifdef INET6 + case AF_INET6: + to6 = &sctp_ifap->address.sin6; + if (IN6_IS_ADDR_LOOPBACK(&to6->sin6_addr)) { + continue; + } + if (IN6_IS_ADDR_LINKLOCAL(&to6->sin6_addr)) { + continue; + } + break; +#endif + default: + continue; + } + sctp_asconf_queue_mgmt(stcb, sctp_ifap, SCTP_ADD_IP_ADDRESS); + } + } + skip_rest: + /* Now we must send the asconf into the queue */ + sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED); +} diff --git a/netwerk/sctp/src/netinet/sctp_asconf.h b/netwerk/sctp/src/netinet/sctp_asconf.h new file mode 100644 index 0000000000..318b26b168 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_asconf.h @@ -0,0 +1,94 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_ASCONF_H_ +#define _NETINET_SCTP_ASCONF_H_ + +#if defined(_KERNEL) || defined(__Userspace__) + +/* + * function prototypes + */ +extern void sctp_asconf_cleanup(struct sctp_tcb *); + +extern struct mbuf *sctp_compose_asconf(struct sctp_tcb *, int *, int); + +extern void +sctp_handle_asconf(struct mbuf *, unsigned int, struct sockaddr *, + struct sctp_asconf_chunk *, struct sctp_tcb *, int); + +extern void +sctp_handle_asconf_ack(struct mbuf *, int, struct sctp_asconf_ack_chunk *, + struct sctp_tcb *, struct sctp_nets *, int *); + +extern uint32_t +sctp_addr_mgmt_ep_sa(struct sctp_inpcb *, struct sockaddr *, uint32_t, + uint32_t); + +extern int sctp_asconf_iterator_ep(struct sctp_inpcb *inp, void *ptr, + uint32_t val); +extern void sctp_asconf_iterator_stcb(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + void *ptr, uint32_t type); +extern void sctp_asconf_iterator_end(void *ptr, uint32_t val); + +extern int32_t +sctp_set_primary_ip_address_sa(struct sctp_tcb *, + struct sockaddr *); + +extern void +sctp_check_address_list(struct sctp_tcb *, struct mbuf *, int, int, + struct sockaddr *, uint16_t, uint16_t, uint16_t, uint16_t); + +extern void +sctp_assoc_immediate_retrans(struct sctp_tcb *, struct sctp_nets *); +#if defined(__FreeBSD__) || defined(__APPLE__) || defined(__Userspace__) +extern void +sctp_net_immediate_retrans(struct sctp_tcb *, struct sctp_nets *); +#endif + +extern void +sctp_asconf_send_nat_state_update(struct sctp_tcb *stcb, + struct sctp_nets *net); + +extern int +sctp_is_addr_pending(struct sctp_tcb *, struct sctp_ifa *); +#endif /* _KERNEL */ + +#endif /* !_NETINET_SCTP_ASCONF_H_ */ diff --git a/netwerk/sctp/src/netinet/sctp_auth.c b/netwerk/sctp/src/netinet/sctp_auth.c new file mode 100644 index 0000000000..39a0f5014a --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_auth.c @@ -0,0 +1,2302 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#include <netinet/sctp.h> +#include <netinet/sctp_header.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctp_var.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_indata.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_auth.h> + +#ifdef SCTP_DEBUG +#define SCTP_AUTH_DEBUG (SCTP_BASE_SYSCTL(sctp_debug_on) & SCTP_DEBUG_AUTH1) +#define SCTP_AUTH_DEBUG2 (SCTP_BASE_SYSCTL(sctp_debug_on) & SCTP_DEBUG_AUTH2) +#endif /* SCTP_DEBUG */ + +void +sctp_clear_chunklist(sctp_auth_chklist_t *chklist) +{ + memset(chklist, 0, sizeof(*chklist)); + /* chklist->num_chunks = 0; */ +} + +sctp_auth_chklist_t * +sctp_alloc_chunklist(void) +{ + sctp_auth_chklist_t *chklist; + + SCTP_MALLOC(chklist, sctp_auth_chklist_t *, sizeof(*chklist), + SCTP_M_AUTH_CL); + if (chklist == NULL) { + SCTPDBG(SCTP_DEBUG_AUTH1, "sctp_alloc_chunklist: failed to get memory!\n"); + } else { + sctp_clear_chunklist(chklist); + } + return (chklist); +} + +void +sctp_free_chunklist(sctp_auth_chklist_t *list) +{ + if (list != NULL) + SCTP_FREE(list, SCTP_M_AUTH_CL); +} + +sctp_auth_chklist_t * +sctp_copy_chunklist(sctp_auth_chklist_t *list) +{ + sctp_auth_chklist_t *new_list; + + if (list == NULL) + return (NULL); + + /* get a new list */ + new_list = sctp_alloc_chunklist(); + if (new_list == NULL) + return (NULL); + /* copy it */ + memcpy(new_list, list, sizeof(*new_list)); + + return (new_list); +} + +/* + * add a chunk to the required chunks list + */ +int +sctp_auth_add_chunk(uint8_t chunk, sctp_auth_chklist_t *list) +{ + if (list == NULL) + return (-1); + + /* is chunk restricted? */ + if ((chunk == SCTP_INITIATION) || + (chunk == SCTP_INITIATION_ACK) || + (chunk == SCTP_SHUTDOWN_COMPLETE) || + (chunk == SCTP_AUTHENTICATION)) { + return (-1); + } + if (list->chunks[chunk] == 0) { + list->chunks[chunk] = 1; + list->num_chunks++; + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP: added chunk %u (0x%02x) to Auth list\n", + chunk, chunk); + } + return (0); +} + +/* + * delete a chunk from the required chunks list + */ +int +sctp_auth_delete_chunk(uint8_t chunk, sctp_auth_chklist_t *list) +{ + if (list == NULL) + return (-1); + + if (list->chunks[chunk] == 1) { + list->chunks[chunk] = 0; + list->num_chunks--; + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP: deleted chunk %u (0x%02x) from Auth list\n", + chunk, chunk); + } + return (0); +} + +size_t +sctp_auth_get_chklist_size(const sctp_auth_chklist_t *list) +{ + if (list == NULL) + return (0); + else + return (list->num_chunks); +} + +/* + * return the current number and list of required chunks caller must + * guarantee ptr has space for up to 256 bytes + */ +int +sctp_serialize_auth_chunks(const sctp_auth_chklist_t *list, uint8_t *ptr) +{ + int i, count = 0; + + if (list == NULL) + return (0); + + for (i = 0; i < 256; i++) { + if (list->chunks[i] != 0) { + *ptr++ = i; + count++; + } + } + return (count); +} + +int +sctp_pack_auth_chunks(const sctp_auth_chklist_t *list, uint8_t *ptr) +{ + int i, size = 0; + + if (list == NULL) + return (0); + + if (list->num_chunks <= 32) { + /* just list them, one byte each */ + for (i = 0; i < 256; i++) { + if (list->chunks[i] != 0) { + *ptr++ = i; + size++; + } + } + } else { + int index, offset; + + /* pack into a 32 byte bitfield */ + for (i = 0; i < 256; i++) { + if (list->chunks[i] != 0) { + index = i / 8; + offset = i % 8; + ptr[index] |= (1 << offset); + } + } + size = 32; + } + return (size); +} + +int +sctp_unpack_auth_chunks(const uint8_t *ptr, uint8_t num_chunks, + sctp_auth_chklist_t *list) +{ + int i; + int size; + + if (list == NULL) + return (0); + + if (num_chunks <= 32) { + /* just pull them, one byte each */ + for (i = 0; i < num_chunks; i++) { + (void)sctp_auth_add_chunk(*ptr++, list); + } + size = num_chunks; + } else { + int index, offset; + + /* unpack from a 32 byte bitfield */ + for (index = 0; index < 32; index++) { + for (offset = 0; offset < 8; offset++) { + if (ptr[index] & (1 << offset)) { + (void)sctp_auth_add_chunk((index * 8) + offset, list); + } + } + } + size = 32; + } + return (size); +} + +/* + * allocate structure space for a key of length keylen + */ +sctp_key_t * +sctp_alloc_key(uint32_t keylen) +{ + sctp_key_t *new_key; + + SCTP_MALLOC(new_key, sctp_key_t *, sizeof(*new_key) + keylen, + SCTP_M_AUTH_KY); + if (new_key == NULL) { + /* out of memory */ + return (NULL); + } + new_key->keylen = keylen; + return (new_key); +} + +void +sctp_free_key(sctp_key_t *key) +{ + if (key != NULL) + SCTP_FREE(key,SCTP_M_AUTH_KY); +} + +void +sctp_print_key(sctp_key_t *key, const char *str) +{ + uint32_t i; + + if (key == NULL) { + SCTP_PRINTF("%s: [Null key]\n", str); + return; + } + SCTP_PRINTF("%s: len %u, ", str, key->keylen); + if (key->keylen) { + for (i = 0; i < key->keylen; i++) + SCTP_PRINTF("%02x", key->key[i]); + SCTP_PRINTF("\n"); + } else { + SCTP_PRINTF("[Null key]\n"); + } +} + +void +sctp_show_key(sctp_key_t *key, const char *str) +{ + uint32_t i; + + if (key == NULL) { + SCTP_PRINTF("%s: [Null key]\n", str); + return; + } + SCTP_PRINTF("%s: len %u, ", str, key->keylen); + if (key->keylen) { + for (i = 0; i < key->keylen; i++) + SCTP_PRINTF("%02x", key->key[i]); + SCTP_PRINTF("\n"); + } else { + SCTP_PRINTF("[Null key]\n"); + } +} + +static uint32_t +sctp_get_keylen(sctp_key_t *key) +{ + if (key != NULL) + return (key->keylen); + else + return (0); +} + +/* + * generate a new random key of length 'keylen' + */ +sctp_key_t * +sctp_generate_random_key(uint32_t keylen) +{ + sctp_key_t *new_key; + + new_key = sctp_alloc_key(keylen); + if (new_key == NULL) { + /* out of memory */ + return (NULL); + } + SCTP_READ_RANDOM(new_key->key, keylen); + new_key->keylen = keylen; + return (new_key); +} + +sctp_key_t * +sctp_set_key(uint8_t *key, uint32_t keylen) +{ + sctp_key_t *new_key; + + new_key = sctp_alloc_key(keylen); + if (new_key == NULL) { + /* out of memory */ + return (NULL); + } + memcpy(new_key->key, key, keylen); + return (new_key); +} + +/*- + * given two keys of variable size, compute which key is "larger/smaller" + * returns: 1 if key1 > key2 + * -1 if key1 < key2 + * 0 if key1 = key2 + */ +static int +sctp_compare_key(sctp_key_t *key1, sctp_key_t *key2) +{ + uint32_t maxlen; + uint32_t i; + uint32_t key1len, key2len; + uint8_t *key_1, *key_2; + uint8_t val1, val2; + + /* sanity/length check */ + key1len = sctp_get_keylen(key1); + key2len = sctp_get_keylen(key2); + if ((key1len == 0) && (key2len == 0)) + return (0); + else if (key1len == 0) + return (-1); + else if (key2len == 0) + return (1); + + if (key1len < key2len) { + maxlen = key2len; + } else { + maxlen = key1len; + } + key_1 = key1->key; + key_2 = key2->key; + /* check for numeric equality */ + for (i = 0; i < maxlen; i++) { + /* left-pad with zeros */ + val1 = (i < (maxlen - key1len)) ? 0 : *(key_1++); + val2 = (i < (maxlen - key2len)) ? 0 : *(key_2++); + if (val1 > val2) { + return (1); + } else if (val1 < val2) { + return (-1); + } + } + /* keys are equal value, so check lengths */ + if (key1len == key2len) + return (0); + else if (key1len < key2len) + return (-1); + else + return (1); +} + +/* + * generate the concatenated keying material based on the two keys and the + * shared key (if available). draft-ietf-tsvwg-auth specifies the specific + * order for concatenation + */ +sctp_key_t * +sctp_compute_hashkey(sctp_key_t *key1, sctp_key_t *key2, sctp_key_t *shared) +{ + uint32_t keylen; + sctp_key_t *new_key; + uint8_t *key_ptr; + + keylen = sctp_get_keylen(key1) + sctp_get_keylen(key2) + + sctp_get_keylen(shared); + + if (keylen > 0) { + /* get space for the new key */ + new_key = sctp_alloc_key(keylen); + if (new_key == NULL) { + /* out of memory */ + return (NULL); + } + new_key->keylen = keylen; + key_ptr = new_key->key; + } else { + /* all keys empty/null?! */ + return (NULL); + } + + /* concatenate the keys */ + if (sctp_compare_key(key1, key2) <= 0) { + /* key is shared + key1 + key2 */ + if (sctp_get_keylen(shared)) { + memcpy(key_ptr, shared->key, shared->keylen); + key_ptr += shared->keylen; + } + if (sctp_get_keylen(key1)) { + memcpy(key_ptr, key1->key, key1->keylen); + key_ptr += key1->keylen; + } + if (sctp_get_keylen(key2)) { + memcpy(key_ptr, key2->key, key2->keylen); + } + } else { + /* key is shared + key2 + key1 */ + if (sctp_get_keylen(shared)) { + memcpy(key_ptr, shared->key, shared->keylen); + key_ptr += shared->keylen; + } + if (sctp_get_keylen(key2)) { + memcpy(key_ptr, key2->key, key2->keylen); + key_ptr += key2->keylen; + } + if (sctp_get_keylen(key1)) { + memcpy(key_ptr, key1->key, key1->keylen); + } + } + return (new_key); +} + +sctp_sharedkey_t * +sctp_alloc_sharedkey(void) +{ + sctp_sharedkey_t *new_key; + + SCTP_MALLOC(new_key, sctp_sharedkey_t *, sizeof(*new_key), + SCTP_M_AUTH_KY); + if (new_key == NULL) { + /* out of memory */ + return (NULL); + } + new_key->keyid = 0; + new_key->key = NULL; + new_key->refcount = 1; + new_key->deactivated = 0; + return (new_key); +} + +void +sctp_free_sharedkey(sctp_sharedkey_t *skey) +{ + if (skey == NULL) + return; + + if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&skey->refcount)) { + if (skey->key != NULL) + sctp_free_key(skey->key); + SCTP_FREE(skey, SCTP_M_AUTH_KY); + } +} + +sctp_sharedkey_t * +sctp_find_sharedkey(struct sctp_keyhead *shared_keys, uint16_t key_id) +{ + sctp_sharedkey_t *skey; + + LIST_FOREACH(skey, shared_keys, next) { + if (skey->keyid == key_id) + return (skey); + } + return (NULL); +} + +int +sctp_insert_sharedkey(struct sctp_keyhead *shared_keys, + sctp_sharedkey_t *new_skey) +{ + sctp_sharedkey_t *skey; + + if ((shared_keys == NULL) || (new_skey == NULL)) + return (EINVAL); + + /* insert into an empty list? */ + if (LIST_EMPTY(shared_keys)) { + LIST_INSERT_HEAD(shared_keys, new_skey, next); + return (0); + } + /* insert into the existing list, ordered by key id */ + LIST_FOREACH(skey, shared_keys, next) { + if (new_skey->keyid < skey->keyid) { + /* insert it before here */ + LIST_INSERT_BEFORE(skey, new_skey, next); + return (0); + } else if (new_skey->keyid == skey->keyid) { + /* replace the existing key */ + /* verify this key *can* be replaced */ + if ((skey->deactivated) || (skey->refcount > 1)) { + SCTPDBG(SCTP_DEBUG_AUTH1, + "can't replace shared key id %u\n", + new_skey->keyid); + return (EBUSY); + } + SCTPDBG(SCTP_DEBUG_AUTH1, + "replacing shared key id %u\n", + new_skey->keyid); + LIST_INSERT_BEFORE(skey, new_skey, next); + LIST_REMOVE(skey, next); + sctp_free_sharedkey(skey); + return (0); + } + if (LIST_NEXT(skey, next) == NULL) { + /* belongs at the end of the list */ + LIST_INSERT_AFTER(skey, new_skey, next); + return (0); + } + } + /* shouldn't reach here */ + return (EINVAL); +} + +void +sctp_auth_key_acquire(struct sctp_tcb *stcb, uint16_t key_id) +{ + sctp_sharedkey_t *skey; + + /* find the shared key */ + skey = sctp_find_sharedkey(&stcb->asoc.shared_keys, key_id); + + /* bump the ref count */ + if (skey) { + atomic_add_int(&skey->refcount, 1); + SCTPDBG(SCTP_DEBUG_AUTH2, + "%s: stcb %p key %u refcount acquire to %d\n", + __func__, (void *)stcb, key_id, skey->refcount); + } +} + +void +sctp_auth_key_release(struct sctp_tcb *stcb, uint16_t key_id, int so_locked) +{ + sctp_sharedkey_t *skey; + + /* find the shared key */ + skey = sctp_find_sharedkey(&stcb->asoc.shared_keys, key_id); + + /* decrement the ref count */ + if (skey) { + SCTPDBG(SCTP_DEBUG_AUTH2, + "%s: stcb %p key %u refcount release to %d\n", + __func__, (void *)stcb, key_id, skey->refcount); + + /* see if a notification should be generated */ + if ((skey->refcount <= 2) && (skey->deactivated)) { + /* notify ULP that key is no longer used */ + sctp_ulp_notify(SCTP_NOTIFY_AUTH_FREE_KEY, stcb, + key_id, 0, so_locked); + SCTPDBG(SCTP_DEBUG_AUTH2, + "%s: stcb %p key %u no longer used, %d\n", + __func__, (void *)stcb, key_id, skey->refcount); + } + sctp_free_sharedkey(skey); + } +} + +static sctp_sharedkey_t * +sctp_copy_sharedkey(const sctp_sharedkey_t *skey) +{ + sctp_sharedkey_t *new_skey; + + if (skey == NULL) + return (NULL); + new_skey = sctp_alloc_sharedkey(); + if (new_skey == NULL) + return (NULL); + if (skey->key != NULL) + new_skey->key = sctp_set_key(skey->key->key, skey->key->keylen); + else + new_skey->key = NULL; + new_skey->keyid = skey->keyid; + return (new_skey); +} + +int +sctp_copy_skeylist(const struct sctp_keyhead *src, struct sctp_keyhead *dest) +{ + sctp_sharedkey_t *skey, *new_skey; + int count = 0; + + if ((src == NULL) || (dest == NULL)) + return (0); + LIST_FOREACH(skey, src, next) { + new_skey = sctp_copy_sharedkey(skey); + if (new_skey != NULL) { + if (sctp_insert_sharedkey(dest, new_skey)) { + sctp_free_sharedkey(new_skey); + } else { + count++; + } + } + } + return (count); +} + +sctp_hmaclist_t * +sctp_alloc_hmaclist(uint16_t num_hmacs) +{ + sctp_hmaclist_t *new_list; + int alloc_size; + + alloc_size = sizeof(*new_list) + num_hmacs * sizeof(new_list->hmac[0]); + SCTP_MALLOC(new_list, sctp_hmaclist_t *, alloc_size, + SCTP_M_AUTH_HL); + if (new_list == NULL) { + /* out of memory */ + return (NULL); + } + new_list->max_algo = num_hmacs; + new_list->num_algo = 0; + return (new_list); +} + +void +sctp_free_hmaclist(sctp_hmaclist_t *list) +{ + if (list != NULL) { + SCTP_FREE(list,SCTP_M_AUTH_HL); + } +} + +int +sctp_auth_add_hmacid(sctp_hmaclist_t *list, uint16_t hmac_id) +{ + int i; + if (list == NULL) + return (-1); + if (list->num_algo == list->max_algo) { + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP: HMAC id list full, ignoring add %u\n", hmac_id); + return (-1); + } +#if defined(SCTP_SUPPORT_HMAC_SHA256) + if ((hmac_id != SCTP_AUTH_HMAC_ID_SHA1) && + (hmac_id != SCTP_AUTH_HMAC_ID_SHA256)) { +#else + if (hmac_id != SCTP_AUTH_HMAC_ID_SHA1) { +#endif + return (-1); + } + /* Now is it already in the list */ + for (i = 0; i < list->num_algo; i++) { + if (list->hmac[i] == hmac_id) { + /* already in list */ + return (-1); + } + } + SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: add HMAC id %u to list\n", hmac_id); + list->hmac[list->num_algo++] = hmac_id; + return (0); +} + +sctp_hmaclist_t * +sctp_copy_hmaclist(sctp_hmaclist_t *list) +{ + sctp_hmaclist_t *new_list; + int i; + + if (list == NULL) + return (NULL); + /* get a new list */ + new_list = sctp_alloc_hmaclist(list->max_algo); + if (new_list == NULL) + return (NULL); + /* copy it */ + new_list->max_algo = list->max_algo; + new_list->num_algo = list->num_algo; + for (i = 0; i < list->num_algo; i++) + new_list->hmac[i] = list->hmac[i]; + return (new_list); +} + +sctp_hmaclist_t * +sctp_default_supported_hmaclist(void) +{ + sctp_hmaclist_t *new_list; + +#if defined(SCTP_SUPPORT_HMAC_SHA256) + new_list = sctp_alloc_hmaclist(2); +#else + new_list = sctp_alloc_hmaclist(1); +#endif + if (new_list == NULL) + return (NULL); +#if defined(SCTP_SUPPORT_HMAC_SHA256) + /* We prefer SHA256, so list it first */ + (void)sctp_auth_add_hmacid(new_list, SCTP_AUTH_HMAC_ID_SHA256); +#endif + (void)sctp_auth_add_hmacid(new_list, SCTP_AUTH_HMAC_ID_SHA1); + return (new_list); +} + +/*- + * HMAC algos are listed in priority/preference order + * find the best HMAC id to use for the peer based on local support + */ +uint16_t +sctp_negotiate_hmacid(sctp_hmaclist_t *peer, sctp_hmaclist_t *local) +{ + int i, j; + + if ((local == NULL) || (peer == NULL)) + return (SCTP_AUTH_HMAC_ID_RSVD); + + for (i = 0; i < peer->num_algo; i++) { + for (j = 0; j < local->num_algo; j++) { + if (peer->hmac[i] == local->hmac[j]) { + /* found the "best" one */ + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP: negotiated peer HMAC id %u\n", + peer->hmac[i]); + return (peer->hmac[i]); + } + } + } + /* didn't find one! */ + return (SCTP_AUTH_HMAC_ID_RSVD); +} + +/*- + * serialize the HMAC algo list and return space used + * caller must guarantee ptr has appropriate space + */ +int +sctp_serialize_hmaclist(sctp_hmaclist_t *list, uint8_t *ptr) +{ + int i; + uint16_t hmac_id; + + if (list == NULL) + return (0); + + for (i = 0; i < list->num_algo; i++) { + hmac_id = htons(list->hmac[i]); + memcpy(ptr, &hmac_id, sizeof(hmac_id)); + ptr += sizeof(hmac_id); + } + return (list->num_algo * sizeof(hmac_id)); +} + +int +sctp_verify_hmac_param (struct sctp_auth_hmac_algo *hmacs, uint32_t num_hmacs) +{ + uint32_t i; + + for (i = 0; i < num_hmacs; i++) { + if (ntohs(hmacs->hmac_ids[i]) == SCTP_AUTH_HMAC_ID_SHA1) { + return (0); + } + } + return (-1); +} + +sctp_authinfo_t * +sctp_alloc_authinfo(void) +{ + sctp_authinfo_t *new_authinfo; + + SCTP_MALLOC(new_authinfo, sctp_authinfo_t *, sizeof(*new_authinfo), + SCTP_M_AUTH_IF); + + if (new_authinfo == NULL) { + /* out of memory */ + return (NULL); + } + memset(new_authinfo, 0, sizeof(*new_authinfo)); + return (new_authinfo); +} + +void +sctp_free_authinfo(sctp_authinfo_t *authinfo) +{ + if (authinfo == NULL) + return; + + if (authinfo->random != NULL) + sctp_free_key(authinfo->random); + if (authinfo->peer_random != NULL) + sctp_free_key(authinfo->peer_random); + if (authinfo->assoc_key != NULL) + sctp_free_key(authinfo->assoc_key); + if (authinfo->recv_key != NULL) + sctp_free_key(authinfo->recv_key); + + /* We are NOT dynamically allocating authinfo's right now... */ + /* SCTP_FREE(authinfo, SCTP_M_AUTH_??); */ +} + +uint32_t +sctp_get_auth_chunk_len(uint16_t hmac_algo) +{ + int size; + + size = sizeof(struct sctp_auth_chunk) + sctp_get_hmac_digest_len(hmac_algo); + return (SCTP_SIZE32(size)); +} + +uint32_t +sctp_get_hmac_digest_len(uint16_t hmac_algo) +{ + switch (hmac_algo) { + case SCTP_AUTH_HMAC_ID_SHA1: + return (SCTP_AUTH_DIGEST_LEN_SHA1); +#if defined(SCTP_SUPPORT_HMAC_SHA256) + case SCTP_AUTH_HMAC_ID_SHA256: + return (SCTP_AUTH_DIGEST_LEN_SHA256); +#endif + default: + /* unknown HMAC algorithm: can't do anything */ + return (0); + } /* end switch */ +} + +static inline int +sctp_get_hmac_block_len(uint16_t hmac_algo) +{ + switch (hmac_algo) { + case SCTP_AUTH_HMAC_ID_SHA1: + return (64); +#if defined(SCTP_SUPPORT_HMAC_SHA256) + case SCTP_AUTH_HMAC_ID_SHA256: + return (64); +#endif + case SCTP_AUTH_HMAC_ID_RSVD: + default: + /* unknown HMAC algorithm: can't do anything */ + return (0); + } /* end switch */ +} + +#if defined(__Userspace__) +/* __Userspace__ SHA1_Init is defined in libcrypto.a (libssl-dev on Ubuntu) */ +#endif +static void +sctp_hmac_init(uint16_t hmac_algo, sctp_hash_context_t *ctx) +{ + switch (hmac_algo) { + case SCTP_AUTH_HMAC_ID_SHA1: + SCTP_SHA1_INIT(&ctx->sha1); + break; +#if defined(SCTP_SUPPORT_HMAC_SHA256) + case SCTP_AUTH_HMAC_ID_SHA256: + SCTP_SHA256_INIT(&ctx->sha256); + break; +#endif + case SCTP_AUTH_HMAC_ID_RSVD: + default: + /* unknown HMAC algorithm: can't do anything */ + return; + } /* end switch */ +} + +static void +sctp_hmac_update(uint16_t hmac_algo, sctp_hash_context_t *ctx, + uint8_t *text, uint32_t textlen) +{ + switch (hmac_algo) { + case SCTP_AUTH_HMAC_ID_SHA1: + SCTP_SHA1_UPDATE(&ctx->sha1, text, textlen); + break; +#if defined(SCTP_SUPPORT_HMAC_SHA256) + case SCTP_AUTH_HMAC_ID_SHA256: + SCTP_SHA256_UPDATE(&ctx->sha256, text, textlen); + break; +#endif + case SCTP_AUTH_HMAC_ID_RSVD: + default: + /* unknown HMAC algorithm: can't do anything */ + return; + } /* end switch */ +} + +static void +sctp_hmac_final(uint16_t hmac_algo, sctp_hash_context_t *ctx, + uint8_t *digest) +{ + switch (hmac_algo) { + case SCTP_AUTH_HMAC_ID_SHA1: + SCTP_SHA1_FINAL(digest, &ctx->sha1); + break; +#if defined(SCTP_SUPPORT_HMAC_SHA256) + case SCTP_AUTH_HMAC_ID_SHA256: + SCTP_SHA256_FINAL(digest, &ctx->sha256); + break; +#endif + case SCTP_AUTH_HMAC_ID_RSVD: + default: + /* unknown HMAC algorithm: can't do anything */ + return; + } /* end switch */ +} + +/*- + * Keyed-Hashing for Message Authentication: FIPS 198 (RFC 2104) + * + * Compute the HMAC digest using the desired hash key, text, and HMAC + * algorithm. Resulting digest is placed in 'digest' and digest length + * is returned, if the HMAC was performed. + * + * WARNING: it is up to the caller to supply sufficient space to hold the + * resultant digest. + */ +uint32_t +sctp_hmac(uint16_t hmac_algo, uint8_t *key, uint32_t keylen, + uint8_t *text, uint32_t textlen, uint8_t *digest) +{ + uint32_t digestlen; + uint32_t blocklen; + sctp_hash_context_t ctx; + uint8_t ipad[128], opad[128]; /* keyed hash inner/outer pads */ + uint8_t temp[SCTP_AUTH_DIGEST_LEN_MAX]; + uint32_t i; + + /* sanity check the material and length */ + if ((key == NULL) || (keylen == 0) || (text == NULL) || + (textlen == 0) || (digest == NULL)) { + /* can't do HMAC with empty key or text or digest store */ + return (0); + } + /* validate the hmac algo and get the digest length */ + digestlen = sctp_get_hmac_digest_len(hmac_algo); + if (digestlen == 0) + return (0); + + /* hash the key if it is longer than the hash block size */ + blocklen = sctp_get_hmac_block_len(hmac_algo); + if (keylen > blocklen) { + sctp_hmac_init(hmac_algo, &ctx); + sctp_hmac_update(hmac_algo, &ctx, key, keylen); + sctp_hmac_final(hmac_algo, &ctx, temp); + /* set the hashed key as the key */ + keylen = digestlen; + key = temp; + } + /* initialize the inner/outer pads with the key and "append" zeroes */ + memset(ipad, 0, blocklen); + memset(opad, 0, blocklen); + memcpy(ipad, key, keylen); + memcpy(opad, key, keylen); + + /* XOR the key with ipad and opad values */ + for (i = 0; i < blocklen; i++) { + ipad[i] ^= 0x36; + opad[i] ^= 0x5c; + } + + /* perform inner hash */ + sctp_hmac_init(hmac_algo, &ctx); + sctp_hmac_update(hmac_algo, &ctx, ipad, blocklen); + sctp_hmac_update(hmac_algo, &ctx, text, textlen); + sctp_hmac_final(hmac_algo, &ctx, temp); + + /* perform outer hash */ + sctp_hmac_init(hmac_algo, &ctx); + sctp_hmac_update(hmac_algo, &ctx, opad, blocklen); + sctp_hmac_update(hmac_algo, &ctx, temp, digestlen); + sctp_hmac_final(hmac_algo, &ctx, digest); + + return (digestlen); +} + +/* mbuf version */ +uint32_t +sctp_hmac_m(uint16_t hmac_algo, uint8_t *key, uint32_t keylen, + struct mbuf *m, uint32_t m_offset, uint8_t *digest, uint32_t trailer) +{ + uint32_t digestlen; + uint32_t blocklen; + sctp_hash_context_t ctx; + uint8_t ipad[128], opad[128]; /* keyed hash inner/outer pads */ + uint8_t temp[SCTP_AUTH_DIGEST_LEN_MAX]; + uint32_t i; + struct mbuf *m_tmp; + + /* sanity check the material and length */ + if ((key == NULL) || (keylen == 0) || (m == NULL) || (digest == NULL)) { + /* can't do HMAC with empty key or text or digest store */ + return (0); + } + /* validate the hmac algo and get the digest length */ + digestlen = sctp_get_hmac_digest_len(hmac_algo); + if (digestlen == 0) + return (0); + + /* hash the key if it is longer than the hash block size */ + blocklen = sctp_get_hmac_block_len(hmac_algo); + if (keylen > blocklen) { + sctp_hmac_init(hmac_algo, &ctx); + sctp_hmac_update(hmac_algo, &ctx, key, keylen); + sctp_hmac_final(hmac_algo, &ctx, temp); + /* set the hashed key as the key */ + keylen = digestlen; + key = temp; + } + /* initialize the inner/outer pads with the key and "append" zeroes */ + memset(ipad, 0, blocklen); + memset(opad, 0, blocklen); + memcpy(ipad, key, keylen); + memcpy(opad, key, keylen); + + /* XOR the key with ipad and opad values */ + for (i = 0; i < blocklen; i++) { + ipad[i] ^= 0x36; + opad[i] ^= 0x5c; + } + + /* perform inner hash */ + sctp_hmac_init(hmac_algo, &ctx); + sctp_hmac_update(hmac_algo, &ctx, ipad, blocklen); + /* find the correct starting mbuf and offset (get start of text) */ + m_tmp = m; + while ((m_tmp != NULL) && (m_offset >= (uint32_t) SCTP_BUF_LEN(m_tmp))) { + m_offset -= SCTP_BUF_LEN(m_tmp); + m_tmp = SCTP_BUF_NEXT(m_tmp); + } + /* now use the rest of the mbuf chain for the text */ + while (m_tmp != NULL) { + if ((SCTP_BUF_NEXT(m_tmp) == NULL) && trailer) { + sctp_hmac_update(hmac_algo, &ctx, mtod(m_tmp, uint8_t *) + m_offset, + SCTP_BUF_LEN(m_tmp) - (trailer+m_offset)); + } else { + sctp_hmac_update(hmac_algo, &ctx, mtod(m_tmp, uint8_t *) + m_offset, + SCTP_BUF_LEN(m_tmp) - m_offset); + } + + /* clear the offset since it's only for the first mbuf */ + m_offset = 0; + m_tmp = SCTP_BUF_NEXT(m_tmp); + } + sctp_hmac_final(hmac_algo, &ctx, temp); + + /* perform outer hash */ + sctp_hmac_init(hmac_algo, &ctx); + sctp_hmac_update(hmac_algo, &ctx, opad, blocklen); + sctp_hmac_update(hmac_algo, &ctx, temp, digestlen); + sctp_hmac_final(hmac_algo, &ctx, digest); + + return (digestlen); +} + +/* + * computes the requested HMAC using a key struct (which may be modified if + * the keylen exceeds the HMAC block len). + */ +uint32_t +sctp_compute_hmac(uint16_t hmac_algo, sctp_key_t *key, uint8_t *text, + uint32_t textlen, uint8_t *digest) +{ + uint32_t digestlen; + uint32_t blocklen; + sctp_hash_context_t ctx; + uint8_t temp[SCTP_AUTH_DIGEST_LEN_MAX]; + + /* sanity check */ + if ((key == NULL) || (text == NULL) || (textlen == 0) || + (digest == NULL)) { + /* can't do HMAC with empty key or text or digest store */ + return (0); + } + /* validate the hmac algo and get the digest length */ + digestlen = sctp_get_hmac_digest_len(hmac_algo); + if (digestlen == 0) + return (0); + + /* hash the key if it is longer than the hash block size */ + blocklen = sctp_get_hmac_block_len(hmac_algo); + if (key->keylen > blocklen) { + sctp_hmac_init(hmac_algo, &ctx); + sctp_hmac_update(hmac_algo, &ctx, key->key, key->keylen); + sctp_hmac_final(hmac_algo, &ctx, temp); + /* save the hashed key as the new key */ + key->keylen = digestlen; + memcpy(key->key, temp, key->keylen); + } + return (sctp_hmac(hmac_algo, key->key, key->keylen, text, textlen, + digest)); +} + +/* mbuf version */ +uint32_t +sctp_compute_hmac_m(uint16_t hmac_algo, sctp_key_t *key, struct mbuf *m, + uint32_t m_offset, uint8_t *digest) +{ + uint32_t digestlen; + uint32_t blocklen; + sctp_hash_context_t ctx; + uint8_t temp[SCTP_AUTH_DIGEST_LEN_MAX]; + + /* sanity check */ + if ((key == NULL) || (m == NULL) || (digest == NULL)) { + /* can't do HMAC with empty key or text or digest store */ + return (0); + } + /* validate the hmac algo and get the digest length */ + digestlen = sctp_get_hmac_digest_len(hmac_algo); + if (digestlen == 0) + return (0); + + /* hash the key if it is longer than the hash block size */ + blocklen = sctp_get_hmac_block_len(hmac_algo); + if (key->keylen > blocklen) { + sctp_hmac_init(hmac_algo, &ctx); + sctp_hmac_update(hmac_algo, &ctx, key->key, key->keylen); + sctp_hmac_final(hmac_algo, &ctx, temp); + /* save the hashed key as the new key */ + key->keylen = digestlen; + memcpy(key->key, temp, key->keylen); + } + return (sctp_hmac_m(hmac_algo, key->key, key->keylen, m, m_offset, digest, 0)); +} + +int +sctp_auth_is_supported_hmac(sctp_hmaclist_t *list, uint16_t id) +{ + int i; + + if ((list == NULL) || (id == SCTP_AUTH_HMAC_ID_RSVD)) + return (0); + + for (i = 0; i < list->num_algo; i++) + if (list->hmac[i] == id) + return (1); + + /* not in the list */ + return (0); +} + +/*- + * clear any cached key(s) if they match the given key id on an association. + * the cached key(s) will be recomputed and re-cached at next use. + * ASSUMES TCB_LOCK is already held + */ +void +sctp_clear_cachedkeys(struct sctp_tcb *stcb, uint16_t keyid) +{ + if (stcb == NULL) + return; + + if (keyid == stcb->asoc.authinfo.assoc_keyid) { + sctp_free_key(stcb->asoc.authinfo.assoc_key); + stcb->asoc.authinfo.assoc_key = NULL; + } + if (keyid == stcb->asoc.authinfo.recv_keyid) { + sctp_free_key(stcb->asoc.authinfo.recv_key); + stcb->asoc.authinfo.recv_key = NULL; + } +} + +/*- + * clear any cached key(s) if they match the given key id for all assocs on + * an endpoint. + * ASSUMES INP_WLOCK is already held + */ +void +sctp_clear_cachedkeys_ep(struct sctp_inpcb *inp, uint16_t keyid) +{ + struct sctp_tcb *stcb; + + if (inp == NULL) + return; + + /* clear the cached keys on all assocs on this instance */ + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + sctp_clear_cachedkeys(stcb, keyid); + SCTP_TCB_UNLOCK(stcb); + } +} + +/*- + * delete a shared key from an association + * ASSUMES TCB_LOCK is already held + */ +int +sctp_delete_sharedkey(struct sctp_tcb *stcb, uint16_t keyid) +{ + sctp_sharedkey_t *skey; + + if (stcb == NULL) + return (-1); + + /* is the keyid the assoc active sending key */ + if (keyid == stcb->asoc.authinfo.active_keyid) + return (-1); + + /* does the key exist? */ + skey = sctp_find_sharedkey(&stcb->asoc.shared_keys, keyid); + if (skey == NULL) + return (-1); + + /* are there other refcount holders on the key? */ + if (skey->refcount > 1) + return (-1); + + /* remove it */ + LIST_REMOVE(skey, next); + sctp_free_sharedkey(skey); /* frees skey->key as well */ + + /* clear any cached keys */ + sctp_clear_cachedkeys(stcb, keyid); + return (0); +} + +/*- + * deletes a shared key from the endpoint + * ASSUMES INP_WLOCK is already held + */ +int +sctp_delete_sharedkey_ep(struct sctp_inpcb *inp, uint16_t keyid) +{ + sctp_sharedkey_t *skey; + + if (inp == NULL) + return (-1); + + /* is the keyid the active sending key on the endpoint */ + if (keyid == inp->sctp_ep.default_keyid) + return (-1); + + /* does the key exist? */ + skey = sctp_find_sharedkey(&inp->sctp_ep.shared_keys, keyid); + if (skey == NULL) + return (-1); + + /* endpoint keys are not refcounted */ + + /* remove it */ + LIST_REMOVE(skey, next); + sctp_free_sharedkey(skey); /* frees skey->key as well */ + + /* clear any cached keys */ + sctp_clear_cachedkeys_ep(inp, keyid); + return (0); +} + +/*- + * set the active key on an association + * ASSUMES TCB_LOCK is already held + */ +int +sctp_auth_setactivekey(struct sctp_tcb *stcb, uint16_t keyid) +{ + sctp_sharedkey_t *skey = NULL; + + /* find the key on the assoc */ + skey = sctp_find_sharedkey(&stcb->asoc.shared_keys, keyid); + if (skey == NULL) { + /* that key doesn't exist */ + return (-1); + } + if ((skey->deactivated) && (skey->refcount > 1)) { + /* can't reactivate a deactivated key with other refcounts */ + return (-1); + } + + /* set the (new) active key */ + stcb->asoc.authinfo.active_keyid = keyid; + /* reset the deactivated flag */ + skey->deactivated = 0; + + return (0); +} + +/*- + * set the active key on an endpoint + * ASSUMES INP_WLOCK is already held + */ +int +sctp_auth_setactivekey_ep(struct sctp_inpcb *inp, uint16_t keyid) +{ + sctp_sharedkey_t *skey; + + /* find the key */ + skey = sctp_find_sharedkey(&inp->sctp_ep.shared_keys, keyid); + if (skey == NULL) { + /* that key doesn't exist */ + return (-1); + } + inp->sctp_ep.default_keyid = keyid; + return (0); +} + +/*- + * deactivates a shared key from the association + * ASSUMES INP_WLOCK is already held + */ +int +sctp_deact_sharedkey(struct sctp_tcb *stcb, uint16_t keyid) +{ + sctp_sharedkey_t *skey; + + if (stcb == NULL) + return (-1); + + /* is the keyid the assoc active sending key */ + if (keyid == stcb->asoc.authinfo.active_keyid) + return (-1); + + /* does the key exist? */ + skey = sctp_find_sharedkey(&stcb->asoc.shared_keys, keyid); + if (skey == NULL) + return (-1); + + /* are there other refcount holders on the key? */ + if (skey->refcount == 1) { + /* no other users, send a notification for this key */ + sctp_ulp_notify(SCTP_NOTIFY_AUTH_FREE_KEY, stcb, keyid, 0, + SCTP_SO_LOCKED); + } + + /* mark the key as deactivated */ + skey->deactivated = 1; + + return (0); +} + +/*- + * deactivates a shared key from the endpoint + * ASSUMES INP_WLOCK is already held + */ +int +sctp_deact_sharedkey_ep(struct sctp_inpcb *inp, uint16_t keyid) +{ + sctp_sharedkey_t *skey; + + if (inp == NULL) + return (-1); + + /* is the keyid the active sending key on the endpoint */ + if (keyid == inp->sctp_ep.default_keyid) + return (-1); + + /* does the key exist? */ + skey = sctp_find_sharedkey(&inp->sctp_ep.shared_keys, keyid); + if (skey == NULL) + return (-1); + + /* endpoint keys are not refcounted */ + + /* remove it */ + LIST_REMOVE(skey, next); + sctp_free_sharedkey(skey); /* frees skey->key as well */ + + return (0); +} + +/* + * get local authentication parameters from cookie (from INIT-ACK) + */ +void +sctp_auth_get_cookie_params(struct sctp_tcb *stcb, struct mbuf *m, + uint32_t offset, uint32_t length) +{ + struct sctp_paramhdr *phdr, tmp_param; + uint16_t plen, ptype; + uint8_t random_store[SCTP_PARAM_BUFFER_SIZE]; + struct sctp_auth_random *p_random = NULL; + uint16_t random_len = 0; + uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE]; + struct sctp_auth_hmac_algo *hmacs = NULL; + uint16_t hmacs_len = 0; + uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE]; + struct sctp_auth_chunk_list *chunks = NULL; + uint16_t num_chunks = 0; + sctp_key_t *new_key; + uint32_t keylen; + + /* convert to upper bound */ + length += offset; + + phdr = (struct sctp_paramhdr *)sctp_m_getptr(m, offset, + sizeof(struct sctp_paramhdr), (uint8_t *)&tmp_param); + while (phdr != NULL) { + ptype = ntohs(phdr->param_type); + plen = ntohs(phdr->param_length); + + if ((plen < sizeof(struct sctp_paramhdr)) || + (offset + plen > length)) + break; + + if (ptype == SCTP_RANDOM) { + if (plen > sizeof(random_store)) + break; + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)random_store, plen); + if (phdr == NULL) + return; + /* save the random and length for the key */ + p_random = (struct sctp_auth_random *)phdr; + random_len = plen - sizeof(*p_random); + } else if (ptype == SCTP_HMAC_LIST) { + uint16_t num_hmacs; + uint16_t i; + + if (plen > sizeof(hmacs_store)) + break; + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)hmacs_store, plen); + if (phdr == NULL) + return; + /* save the hmacs list and num for the key */ + hmacs = (struct sctp_auth_hmac_algo *)phdr; + hmacs_len = plen - sizeof(*hmacs); + num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]); + if (stcb->asoc.local_hmacs != NULL) + sctp_free_hmaclist(stcb->asoc.local_hmacs); + stcb->asoc.local_hmacs = sctp_alloc_hmaclist(num_hmacs); + if (stcb->asoc.local_hmacs != NULL) { + for (i = 0; i < num_hmacs; i++) { + (void)sctp_auth_add_hmacid(stcb->asoc.local_hmacs, + ntohs(hmacs->hmac_ids[i])); + } + } + } else if (ptype == SCTP_CHUNK_LIST) { + int i; + + if (plen > sizeof(chunks_store)) + break; + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)chunks_store, plen); + if (phdr == NULL) + return; + chunks = (struct sctp_auth_chunk_list *)phdr; + num_chunks = plen - sizeof(*chunks); + /* save chunks list and num for the key */ + if (stcb->asoc.local_auth_chunks != NULL) + sctp_clear_chunklist(stcb->asoc.local_auth_chunks); + else + stcb->asoc.local_auth_chunks = sctp_alloc_chunklist(); + for (i = 0; i < num_chunks; i++) { + (void)sctp_auth_add_chunk(chunks->chunk_types[i], + stcb->asoc.local_auth_chunks); + } + } + /* get next parameter */ + offset += SCTP_SIZE32(plen); + if (offset + sizeof(struct sctp_paramhdr) > length) + break; + phdr = (struct sctp_paramhdr *)sctp_m_getptr(m, offset, sizeof(struct sctp_paramhdr), + (uint8_t *)&tmp_param); + } + /* concatenate the full random key */ + keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len; + if (chunks != NULL) { + keylen += sizeof(*chunks) + num_chunks; + } + new_key = sctp_alloc_key(keylen); + if (new_key != NULL) { + /* copy in the RANDOM */ + if (p_random != NULL) { + keylen = sizeof(*p_random) + random_len; + memcpy(new_key->key, p_random, keylen); + } else { + keylen = 0; + } + /* append in the AUTH chunks */ + if (chunks != NULL) { + memcpy(new_key->key + keylen, chunks, + sizeof(*chunks) + num_chunks); + keylen += sizeof(*chunks) + num_chunks; + } + /* append in the HMACs */ + if (hmacs != NULL) { + memcpy(new_key->key + keylen, hmacs, + sizeof(*hmacs) + hmacs_len); + } + } + if (stcb->asoc.authinfo.random != NULL) + sctp_free_key(stcb->asoc.authinfo.random); + stcb->asoc.authinfo.random = new_key; + stcb->asoc.authinfo.random_len = random_len; + sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid); + sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid); + + /* negotiate what HMAC to use for the peer */ + stcb->asoc.peer_hmac_id = sctp_negotiate_hmacid(stcb->asoc.peer_hmacs, + stcb->asoc.local_hmacs); + + /* copy defaults from the endpoint */ + /* FIX ME: put in cookie? */ + stcb->asoc.authinfo.active_keyid = stcb->sctp_ep->sctp_ep.default_keyid; + /* copy out the shared key list (by reference) from the endpoint */ + (void)sctp_copy_skeylist(&stcb->sctp_ep->sctp_ep.shared_keys, + &stcb->asoc.shared_keys); +} + +/* + * compute and fill in the HMAC digest for a packet + */ +void +sctp_fill_hmac_digest_m(struct mbuf *m, uint32_t auth_offset, + struct sctp_auth_chunk *auth, struct sctp_tcb *stcb, uint16_t keyid) +{ + uint32_t digestlen; + sctp_sharedkey_t *skey; + sctp_key_t *key; + + if ((stcb == NULL) || (auth == NULL)) + return; + + /* zero the digest + chunk padding */ + digestlen = sctp_get_hmac_digest_len(stcb->asoc.peer_hmac_id); + memset(auth->hmac, 0, SCTP_SIZE32(digestlen)); + + /* is the desired key cached? */ + if ((keyid != stcb->asoc.authinfo.assoc_keyid) || + (stcb->asoc.authinfo.assoc_key == NULL)) { + if (stcb->asoc.authinfo.assoc_key != NULL) { + /* free the old cached key */ + sctp_free_key(stcb->asoc.authinfo.assoc_key); + } + skey = sctp_find_sharedkey(&stcb->asoc.shared_keys, keyid); + /* the only way skey is NULL is if null key id 0 is used */ + if (skey != NULL) + key = skey->key; + else + key = NULL; + /* compute a new assoc key and cache it */ + stcb->asoc.authinfo.assoc_key = + sctp_compute_hashkey(stcb->asoc.authinfo.random, + stcb->asoc.authinfo.peer_random, key); + stcb->asoc.authinfo.assoc_keyid = keyid; + SCTPDBG(SCTP_DEBUG_AUTH1, "caching key id %u\n", + stcb->asoc.authinfo.assoc_keyid); +#ifdef SCTP_DEBUG + if (SCTP_AUTH_DEBUG) + sctp_print_key(stcb->asoc.authinfo.assoc_key, + "Assoc Key"); +#endif + } + + /* set in the active key id */ + auth->shared_key_id = htons(keyid); + + /* compute and fill in the digest */ + (void)sctp_compute_hmac_m(stcb->asoc.peer_hmac_id, stcb->asoc.authinfo.assoc_key, + m, auth_offset, auth->hmac); +} + +static void +sctp_zero_m(struct mbuf *m, uint32_t m_offset, uint32_t size) +{ + struct mbuf *m_tmp; + uint8_t *data; + + /* sanity check */ + if (m == NULL) + return; + + /* find the correct starting mbuf and offset (get start position) */ + m_tmp = m; + while ((m_tmp != NULL) && (m_offset >= (uint32_t) SCTP_BUF_LEN(m_tmp))) { + m_offset -= SCTP_BUF_LEN(m_tmp); + m_tmp = SCTP_BUF_NEXT(m_tmp); + } + /* now use the rest of the mbuf chain */ + while ((m_tmp != NULL) && (size > 0)) { + data = mtod(m_tmp, uint8_t *) + m_offset; + if (size > (uint32_t)(SCTP_BUF_LEN(m_tmp) - m_offset)) { + memset(data, 0, SCTP_BUF_LEN(m_tmp) - m_offset); + size -= SCTP_BUF_LEN(m_tmp) - m_offset; + } else { + memset(data, 0, size); + size = 0; + } + /* clear the offset since it's only for the first mbuf */ + m_offset = 0; + m_tmp = SCTP_BUF_NEXT(m_tmp); + } +} + +/*- + * process the incoming Authentication chunk + * return codes: + * -1 on any authentication error + * 0 on authentication verification + */ +int +sctp_handle_auth(struct sctp_tcb *stcb, struct sctp_auth_chunk *auth, + struct mbuf *m, uint32_t offset) +{ + uint16_t chunklen; + uint16_t shared_key_id; + uint16_t hmac_id; + sctp_sharedkey_t *skey; + uint32_t digestlen; + uint8_t digest[SCTP_AUTH_DIGEST_LEN_MAX]; + uint8_t computed_digest[SCTP_AUTH_DIGEST_LEN_MAX]; + + /* auth is checked for NULL by caller */ + chunklen = ntohs(auth->ch.chunk_length); + if (chunklen < sizeof(*auth)) { + SCTP_STAT_INCR(sctps_recvauthfailed); + return (-1); + } + SCTP_STAT_INCR(sctps_recvauth); + + /* get the auth params */ + shared_key_id = ntohs(auth->shared_key_id); + hmac_id = ntohs(auth->hmac_id); + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP AUTH Chunk: shared key %u, HMAC id %u\n", + shared_key_id, hmac_id); + +#if defined(__Userspace__) && defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) + return (0); +#endif + /* is the indicated HMAC supported? */ + if (!sctp_auth_is_supported_hmac(stcb->asoc.local_hmacs, hmac_id)) { + struct mbuf *op_err; + struct sctp_error_auth_invalid_hmac *cause; + + SCTP_STAT_INCR(sctps_recvivalhmacid); + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP Auth: unsupported HMAC id %u\n", + hmac_id); + /* + * report this in an Error Chunk: Unsupported HMAC + * Identifier + */ + op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_error_auth_invalid_hmac), + 0, M_NOWAIT, 1, MT_HEADER); + if (op_err != NULL) { + /* pre-reserve some space */ + SCTP_BUF_RESV_UF(op_err, sizeof(struct sctp_chunkhdr)); + /* fill in the error */ + cause = mtod(op_err, struct sctp_error_auth_invalid_hmac *); + cause->cause.code = htons(SCTP_CAUSE_UNSUPPORTED_HMACID); + cause->cause.length = htons(sizeof(struct sctp_error_auth_invalid_hmac)); + cause->hmac_id = ntohs(hmac_id); + SCTP_BUF_LEN(op_err) = sizeof(struct sctp_error_auth_invalid_hmac); + /* queue it */ + sctp_queue_op_err(stcb, op_err); + } + return (-1); + } + /* get the indicated shared key, if available */ + if ((stcb->asoc.authinfo.recv_key == NULL) || + (stcb->asoc.authinfo.recv_keyid != shared_key_id)) { + /* find the shared key on the assoc first */ + skey = sctp_find_sharedkey(&stcb->asoc.shared_keys, + shared_key_id); + /* if the shared key isn't found, discard the chunk */ + if (skey == NULL) { + SCTP_STAT_INCR(sctps_recvivalkeyid); + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP Auth: unknown key id %u\n", + shared_key_id); + return (-1); + } + /* generate a notification if this is a new key id */ + if (stcb->asoc.authinfo.recv_keyid != shared_key_id) + /* + * sctp_ulp_notify(SCTP_NOTIFY_AUTH_NEW_KEY, stcb, + * shared_key_id, (void + * *)stcb->asoc.authinfo.recv_keyid); + */ + sctp_notify_authentication(stcb, SCTP_AUTH_NEW_KEY, + shared_key_id, stcb->asoc.authinfo.recv_keyid, + SCTP_SO_NOT_LOCKED); + /* compute a new recv assoc key and cache it */ + if (stcb->asoc.authinfo.recv_key != NULL) + sctp_free_key(stcb->asoc.authinfo.recv_key); + stcb->asoc.authinfo.recv_key = + sctp_compute_hashkey(stcb->asoc.authinfo.random, + stcb->asoc.authinfo.peer_random, skey->key); + stcb->asoc.authinfo.recv_keyid = shared_key_id; +#ifdef SCTP_DEBUG + if (SCTP_AUTH_DEBUG) + sctp_print_key(stcb->asoc.authinfo.recv_key, "Recv Key"); +#endif + } + /* validate the digest length */ + digestlen = sctp_get_hmac_digest_len(hmac_id); + if (chunklen < (sizeof(*auth) + digestlen)) { + /* invalid digest length */ + SCTP_STAT_INCR(sctps_recvauthfailed); + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP Auth: chunk too short for HMAC\n"); + return (-1); + } + /* save a copy of the digest, zero the pseudo header, and validate */ + memcpy(digest, auth->hmac, digestlen); + sctp_zero_m(m, offset + sizeof(*auth), SCTP_SIZE32(digestlen)); + (void)sctp_compute_hmac_m(hmac_id, stcb->asoc.authinfo.recv_key, + m, offset, computed_digest); + + /* compare the computed digest with the one in the AUTH chunk */ + if (timingsafe_bcmp(digest, computed_digest, digestlen) != 0) { + SCTP_STAT_INCR(sctps_recvauthfailed); + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP Auth: HMAC digest check failed\n"); + return (-1); + } + return (0); +} + +/* + * Generate NOTIFICATION + */ +void +sctp_notify_authentication(struct sctp_tcb *stcb, uint32_t indication, + uint16_t keyid, uint16_t alt_keyid, int so_locked) +{ + struct mbuf *m_notify; + struct sctp_authkey_event *auth; + struct sctp_queued_to_read *control; + + if ((stcb == NULL) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) + ) { + /* If the socket is gone we are out of here */ + return; + } + + if (sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_AUTHEVNT)) + /* event not enabled */ + return; + + m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_authkey_event), + 0, M_NOWAIT, 1, MT_HEADER); + if (m_notify == NULL) + /* no space left */ + return; + + SCTP_BUF_LEN(m_notify) = 0; + auth = mtod(m_notify, struct sctp_authkey_event *); + memset(auth, 0, sizeof(struct sctp_authkey_event)); + auth->auth_type = SCTP_AUTHENTICATION_EVENT; + auth->auth_flags = 0; + auth->auth_length = sizeof(*auth); + auth->auth_keynumber = keyid; + auth->auth_altkeynumber = alt_keyid; + auth->auth_indication = indication; + auth->auth_assoc_id = sctp_get_associd(stcb); + + SCTP_BUF_LEN(m_notify) = sizeof(*auth); + SCTP_BUF_NEXT(m_notify) = NULL; + + /* append to socket */ + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, m_notify); + if (control == NULL) { + /* no memory */ + sctp_m_freem(m_notify); + return; + } + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sctp_add_to_readq(stcb->sctp_ep, stcb, control, + &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, so_locked); +} + +/*- + * validates the AUTHentication related parameters in an INIT/INIT-ACK + * Note: currently only used for INIT as INIT-ACK is handled inline + * with sctp_load_addresses_from_init() + */ +int +sctp_validate_init_auth_params(struct mbuf *m, int offset, int limit) +{ + struct sctp_paramhdr *phdr, param_buf; + uint16_t ptype, plen; + int peer_supports_asconf = 0; + int peer_supports_auth = 0; + int got_random = 0, got_hmacs = 0, got_chklist = 0; + uint8_t saw_asconf = 0; + uint8_t saw_asconf_ack = 0; + + /* go through each of the params. */ + phdr = sctp_get_next_param(m, offset, ¶m_buf, sizeof(param_buf)); + while (phdr) { + ptype = ntohs(phdr->param_type); + plen = ntohs(phdr->param_length); + + if (offset + plen > limit) { + break; + } + if (plen < sizeof(struct sctp_paramhdr)) { + break; + } + if (ptype == SCTP_SUPPORTED_CHUNK_EXT) { + /* A supported extension chunk */ + struct sctp_supported_chunk_types_param *pr_supported; + uint8_t local_store[SCTP_SMALL_CHUNK_STORE]; + int num_ent, i; + + if (plen > sizeof(local_store)) { + break; + } + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)&local_store, + plen); + if (phdr == NULL) { + return (-1); + } + pr_supported = (struct sctp_supported_chunk_types_param *)phdr; + num_ent = plen - sizeof(struct sctp_paramhdr); + for (i = 0; i < num_ent; i++) { + switch (pr_supported->chunk_types[i]) { + case SCTP_ASCONF: + case SCTP_ASCONF_ACK: + peer_supports_asconf = 1; + break; + default: + /* one we don't care about */ + break; + } + } + } else if (ptype == SCTP_RANDOM) { + /* enforce the random length */ + if (plen != (sizeof(struct sctp_auth_random) + + SCTP_AUTH_RANDOM_SIZE_REQUIRED)) { + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP: invalid RANDOM len\n"); + return (-1); + } + got_random = 1; + } else if (ptype == SCTP_HMAC_LIST) { + struct sctp_auth_hmac_algo *hmacs; + uint8_t store[SCTP_PARAM_BUFFER_SIZE]; + int num_hmacs; + + if (plen > sizeof(store)) { + break; + } + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)store, + plen); + if (phdr == NULL) { + return (-1); + } + hmacs = (struct sctp_auth_hmac_algo *)phdr; + num_hmacs = (plen - sizeof(*hmacs)) / sizeof(hmacs->hmac_ids[0]); + /* validate the hmac list */ + if (sctp_verify_hmac_param(hmacs, num_hmacs)) { + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP: invalid HMAC param\n"); + return (-1); + } + got_hmacs = 1; + } else if (ptype == SCTP_CHUNK_LIST) { + struct sctp_auth_chunk_list *chunks; + uint8_t chunks_store[SCTP_SMALL_CHUNK_STORE]; + int i, num_chunks; + + if (plen > sizeof(chunks_store)) { + break; + } + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)chunks_store, + plen); + if (phdr == NULL) { + return (-1); + } + /*- + * Flip through the list and mark that the + * peer supports asconf/asconf_ack. + */ + chunks = (struct sctp_auth_chunk_list *)phdr; + num_chunks = plen - sizeof(*chunks); + for (i = 0; i < num_chunks; i++) { + /* record asconf/asconf-ack if listed */ + if (chunks->chunk_types[i] == SCTP_ASCONF) + saw_asconf = 1; + if (chunks->chunk_types[i] == SCTP_ASCONF_ACK) + saw_asconf_ack = 1; + } + if (num_chunks) + got_chklist = 1; + } + + offset += SCTP_SIZE32(plen); + if (offset >= limit) { + break; + } + phdr = sctp_get_next_param(m, offset, ¶m_buf, + sizeof(param_buf)); + } + /* validate authentication required parameters */ + if (got_random && got_hmacs) { + peer_supports_auth = 1; + } else { + peer_supports_auth = 0; + } + if (!peer_supports_auth && got_chklist) { + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP: peer sent chunk list w/o AUTH\n"); + return (-1); + } + if (peer_supports_asconf && !peer_supports_auth) { + SCTPDBG(SCTP_DEBUG_AUTH1, + "SCTP: peer supports ASCONF but not AUTH\n"); + return (-1); + } else if ((peer_supports_asconf) && (peer_supports_auth) && + ((saw_asconf == 0) || (saw_asconf_ack == 0))) { + return (-2); + } + return (0); +} + +void +sctp_initialize_auth_params(struct sctp_inpcb *inp, struct sctp_tcb *stcb) +{ + uint16_t chunks_len = 0; + uint16_t hmacs_len = 0; + uint16_t random_len = SCTP_AUTH_RANDOM_SIZE_DEFAULT; + sctp_key_t *new_key; + uint16_t keylen; + + /* initialize hmac list from endpoint */ + stcb->asoc.local_hmacs = sctp_copy_hmaclist(inp->sctp_ep.local_hmacs); + if (stcb->asoc.local_hmacs != NULL) { + hmacs_len = stcb->asoc.local_hmacs->num_algo * + sizeof(stcb->asoc.local_hmacs->hmac[0]); + } + /* initialize auth chunks list from endpoint */ + stcb->asoc.local_auth_chunks = + sctp_copy_chunklist(inp->sctp_ep.local_auth_chunks); + if (stcb->asoc.local_auth_chunks != NULL) { + int i; + for (i = 0; i < 256; i++) { + if (stcb->asoc.local_auth_chunks->chunks[i]) + chunks_len++; + } + } + /* copy defaults from the endpoint */ + stcb->asoc.authinfo.active_keyid = inp->sctp_ep.default_keyid; + + /* copy out the shared key list (by reference) from the endpoint */ + (void)sctp_copy_skeylist(&inp->sctp_ep.shared_keys, + &stcb->asoc.shared_keys); + + /* now set the concatenated key (random + chunks + hmacs) */ + /* key includes parameter headers */ + keylen = (3 * sizeof(struct sctp_paramhdr)) + random_len + chunks_len + + hmacs_len; + new_key = sctp_alloc_key(keylen); + if (new_key != NULL) { + struct sctp_paramhdr *ph; + int plen; + /* generate and copy in the RANDOM */ + ph = (struct sctp_paramhdr *)new_key->key; + ph->param_type = htons(SCTP_RANDOM); + plen = sizeof(*ph) + random_len; + ph->param_length = htons(plen); + SCTP_READ_RANDOM(new_key->key + sizeof(*ph), random_len); + keylen = plen; + + /* append in the AUTH chunks */ + /* NOTE: currently we always have chunks to list */ + ph = (struct sctp_paramhdr *)(new_key->key + keylen); + ph->param_type = htons(SCTP_CHUNK_LIST); + plen = sizeof(*ph) + chunks_len; + ph->param_length = htons(plen); + keylen += sizeof(*ph); + if (stcb->asoc.local_auth_chunks) { + int i; + for (i = 0; i < 256; i++) { + if (stcb->asoc.local_auth_chunks->chunks[i]) + new_key->key[keylen++] = i; + } + } + + /* append in the HMACs */ + ph = (struct sctp_paramhdr *)(new_key->key + keylen); + ph->param_type = htons(SCTP_HMAC_LIST); + plen = sizeof(*ph) + hmacs_len; + ph->param_length = htons(plen); + keylen += sizeof(*ph); + (void)sctp_serialize_hmaclist(stcb->asoc.local_hmacs, + new_key->key + keylen); + } + if (stcb->asoc.authinfo.random != NULL) + sctp_free_key(stcb->asoc.authinfo.random); + stcb->asoc.authinfo.random = new_key; + stcb->asoc.authinfo.random_len = random_len; +} + + +#ifdef SCTP_HMAC_TEST +/* + * HMAC and key concatenation tests + */ +static void +sctp_print_digest(uint8_t *digest, uint32_t digestlen, const char *str) +{ + uint32_t i; + + SCTP_PRINTF("\n%s: 0x", str); + if (digest == NULL) + return; + + for (i = 0; i < digestlen; i++) + SCTP_PRINTF("%02x", digest[i]); +} + +static int +sctp_test_hmac(const char *str, uint16_t hmac_id, uint8_t *key, + uint32_t keylen, uint8_t *text, uint32_t textlen, + uint8_t *digest, uint32_t digestlen) +{ + uint8_t computed_digest[SCTP_AUTH_DIGEST_LEN_MAX]; + + SCTP_PRINTF("\n%s:", str); + sctp_hmac(hmac_id, key, keylen, text, textlen, computed_digest); + sctp_print_digest(digest, digestlen, "Expected digest"); + sctp_print_digest(computed_digest, digestlen, "Computed digest"); + if (memcmp(digest, computed_digest, digestlen) != 0) { + SCTP_PRINTF("\nFAILED"); + return (-1); + } else { + SCTP_PRINTF("\nPASSED"); + return (0); + } +} + + +/* + * RFC 2202: HMAC-SHA1 test cases + */ +void +sctp_test_hmac_sha1(void) +{ + uint8_t *digest; + uint8_t key[128]; + uint32_t keylen; + uint8_t text[128]; + uint32_t textlen; + uint32_t digestlen = 20; + int failed = 0; + + /*- + * test_case = 1 + * key = 0x0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b + * key_len = 20 + * data = "Hi There" + * data_len = 8 + * digest = 0xb617318655057264e28bc0b6fb378c8ef146be00 + */ + keylen = 20; + memset(key, 0x0b, keylen); + textlen = 8; + strcpy(text, "Hi There"); + digest = "\xb6\x17\x31\x86\x55\x05\x72\x64\xe2\x8b\xc0\xb6\xfb\x37\x8c\x8e\xf1\x46\xbe\x00"; + if (sctp_test_hmac("SHA1 test case 1", SCTP_AUTH_HMAC_ID_SHA1, key, keylen, + text, textlen, digest, digestlen) < 0) + failed++; + + /*- + * test_case = 2 + * key = "Jefe" + * key_len = 4 + * data = "what do ya want for nothing?" + * data_len = 28 + * digest = 0xeffcdf6ae5eb2fa2d27416d5f184df9c259a7c79 + */ + keylen = 4; + strcpy(key, "Jefe"); + textlen = 28; + strcpy(text, "what do ya want for nothing?"); + digest = "\xef\xfc\xdf\x6a\xe5\xeb\x2f\xa2\xd2\x74\x16\xd5\xf1\x84\xdf\x9c\x25\x9a\x7c\x79"; + if (sctp_test_hmac("SHA1 test case 2", SCTP_AUTH_HMAC_ID_SHA1, key, keylen, + text, textlen, digest, digestlen) < 0) + failed++; + + /*- + * test_case = 3 + * key = 0xaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa + * key_len = 20 + * data = 0xdd repeated 50 times + * data_len = 50 + * digest = 0x125d7342b9ac11cd91a39af48aa17b4f63f175d3 + */ + keylen = 20; + memset(key, 0xaa, keylen); + textlen = 50; + memset(text, 0xdd, textlen); + digest = "\x12\x5d\x73\x42\xb9\xac\x11\xcd\x91\xa3\x9a\xf4\x8a\xa1\x7b\x4f\x63\xf1\x75\xd3"; + if (sctp_test_hmac("SHA1 test case 3", SCTP_AUTH_HMAC_ID_SHA1, key, keylen, + text, textlen, digest, digestlen) < 0) + failed++; + + /*- + * test_case = 4 + * key = 0x0102030405060708090a0b0c0d0e0f10111213141516171819 + * key_len = 25 + * data = 0xcd repeated 50 times + * data_len = 50 + * digest = 0x4c9007f4026250c6bc8414f9bf50c86c2d7235da + */ + keylen = 25; + memcpy(key, "\x01\x02\x03\x04\x05\x06\x07\x08\x09\x0a\x0b\x0c\x0d\x0e\x0f\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19", keylen); + textlen = 50; + memset(text, 0xcd, textlen); + digest = "\x4c\x90\x07\xf4\x02\x62\x50\xc6\xbc\x84\x14\xf9\xbf\x50\xc8\x6c\x2d\x72\x35\xda"; + if (sctp_test_hmac("SHA1 test case 4", SCTP_AUTH_HMAC_ID_SHA1, key, keylen, + text, textlen, digest, digestlen) < 0) + failed++; + + /*- + * test_case = 5 + * key = 0x0c0c0c0c0c0c0c0c0c0c0c0c0c0c0c0c0c0c0c0c + * key_len = 20 + * data = "Test With Truncation" + * data_len = 20 + * digest = 0x4c1a03424b55e07fe7f27be1d58bb9324a9a5a04 + * digest-96 = 0x4c1a03424b55e07fe7f27be1 + */ + keylen = 20; + memset(key, 0x0c, keylen); + textlen = 20; + strcpy(text, "Test With Truncation"); + digest = "\x4c\x1a\x03\x42\x4b\x55\xe0\x7f\xe7\xf2\x7b\xe1\xd5\x8b\xb9\x32\x4a\x9a\x5a\x04"; + if (sctp_test_hmac("SHA1 test case 5", SCTP_AUTH_HMAC_ID_SHA1, key, keylen, + text, textlen, digest, digestlen) < 0) + failed++; + + /*- + * test_case = 6 + * key = 0xaa repeated 80 times + * key_len = 80 + * data = "Test Using Larger Than Block-Size Key - Hash Key First" + * data_len = 54 + * digest = 0xaa4ae5e15272d00e95705637ce8a3b55ed402112 + */ + keylen = 80; + memset(key, 0xaa, keylen); + textlen = 54; + strcpy(text, "Test Using Larger Than Block-Size Key - Hash Key First"); + digest = "\xaa\x4a\xe5\xe1\x52\x72\xd0\x0e\x95\x70\x56\x37\xce\x8a\x3b\x55\xed\x40\x21\x12"; + if (sctp_test_hmac("SHA1 test case 6", SCTP_AUTH_HMAC_ID_SHA1, key, keylen, + text, textlen, digest, digestlen) < 0) + failed++; + + /*- + * test_case = 7 + * key = 0xaa repeated 80 times + * key_len = 80 + * data = "Test Using Larger Than Block-Size Key and Larger Than One Block-Size Data" + * data_len = 73 + * digest = 0xe8e99d0f45237d786d6bbaa7965c7808bbff1a91 + */ + keylen = 80; + memset(key, 0xaa, keylen); + textlen = 73; + strcpy(text, "Test Using Larger Than Block-Size Key and Larger Than One Block-Size Data"); + digest = "\xe8\xe9\x9d\x0f\x45\x23\x7d\x78\x6d\x6b\xba\xa7\x96\x5c\x78\x08\xbb\xff\x1a\x91"; + if (sctp_test_hmac("SHA1 test case 7", SCTP_AUTH_HMAC_ID_SHA1, key, keylen, + text, textlen, digest, digestlen) < 0) + failed++; + + /* done with all tests */ + if (failed) + SCTP_PRINTF("\nSHA1 test results: %d cases failed", failed); + else + SCTP_PRINTF("\nSHA1 test results: all test cases passed"); +} + +/* + * test assoc key concatenation + */ +static int +sctp_test_key_concatenation(sctp_key_t *key1, sctp_key_t *key2, + sctp_key_t *expected_key) +{ + sctp_key_t *key; + int ret_val; + + sctp_show_key(key1, "\nkey1"); + sctp_show_key(key2, "\nkey2"); + key = sctp_compute_hashkey(key1, key2, NULL); + sctp_show_key(expected_key, "\nExpected"); + sctp_show_key(key, "\nComputed"); + if (memcmp(key, expected_key, expected_key->keylen) != 0) { + SCTP_PRINTF("\nFAILED"); + ret_val = -1; + } else { + SCTP_PRINTF("\nPASSED"); + ret_val = 0; + } + sctp_free_key(key1); + sctp_free_key(key2); + sctp_free_key(expected_key); + sctp_free_key(key); + return (ret_val); +} + + +void +sctp_test_authkey(void) +{ + sctp_key_t *key1, *key2, *expected_key; + int failed = 0; + + /* test case 1 */ + key1 = sctp_set_key("\x01\x01\x01\x01", 4); + key2 = sctp_set_key("\x01\x02\x03\x04", 4); + expected_key = sctp_set_key("\x01\x01\x01\x01\x01\x02\x03\x04", 8); + if (sctp_test_key_concatenation(key1, key2, expected_key) < 0) + failed++; + + /* test case 2 */ + key1 = sctp_set_key("\x00\x00\x00\x01", 4); + key2 = sctp_set_key("\x02", 1); + expected_key = sctp_set_key("\x00\x00\x00\x01\x02", 5); + if (sctp_test_key_concatenation(key1, key2, expected_key) < 0) + failed++; + + /* test case 3 */ + key1 = sctp_set_key("\x01", 1); + key2 = sctp_set_key("\x00\x00\x00\x02", 4); + expected_key = sctp_set_key("\x01\x00\x00\x00\x02", 5); + if (sctp_test_key_concatenation(key1, key2, expected_key) < 0) + failed++; + + /* test case 4 */ + key1 = sctp_set_key("\x00\x00\x00\x01", 4); + key2 = sctp_set_key("\x01", 1); + expected_key = sctp_set_key("\x01\x00\x00\x00\x01", 5); + if (sctp_test_key_concatenation(key1, key2, expected_key) < 0) + failed++; + + /* test case 5 */ + key1 = sctp_set_key("\x01", 1); + key2 = sctp_set_key("\x00\x00\x00\x01", 4); + expected_key = sctp_set_key("\x01\x00\x00\x00\x01", 5); + if (sctp_test_key_concatenation(key1, key2, expected_key) < 0) + failed++; + + /* test case 6 */ + key1 = sctp_set_key("\x00\x00\x00\x00\x01\x02\x03\x04\x05\x06\x07", 11); + key2 = sctp_set_key("\x00\x00\x00\x00\x01\x02\x03\x04\x05\x06\x08", 11); + expected_key = sctp_set_key("\x00\x00\x00\x00\x01\x02\x03\x04\x05\x06\x07\x00\x00\x00\x00\x01\x02\x03\x04\x05\x06\x08", 22); + if (sctp_test_key_concatenation(key1, key2, expected_key) < 0) + failed++; + + /* test case 7 */ + key1 = sctp_set_key("\x00\x00\x00\x00\x01\x02\x03\x04\x05\x06\x08", 11); + key2 = sctp_set_key("\x00\x00\x00\x00\x01\x02\x03\x04\x05\x06\x07", 11); + expected_key = sctp_set_key("\x00\x00\x00\x00\x01\x02\x03\x04\x05\x06\x07\x00\x00\x00\x00\x01\x02\x03\x04\x05\x06\x08", 22); + if (sctp_test_key_concatenation(key1, key2, expected_key) < 0) + failed++; + + /* done with all tests */ + if (failed) + SCTP_PRINTF("\nKey concatenation test results: %d cases failed", failed); + else + SCTP_PRINTF("\nKey concatenation test results: all test cases passed"); +} + + +#if defined(STANDALONE_HMAC_TEST) +int +main(void) +{ + sctp_test_hmac_sha1(); + sctp_test_authkey(); +} + +#endif /* STANDALONE_HMAC_TEST */ + +#endif /* SCTP_HMAC_TEST */ diff --git a/netwerk/sctp/src/netinet/sctp_auth.h b/netwerk/sctp/src/netinet/sctp_auth.h new file mode 100644 index 0000000000..718b2f5ba9 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_auth.h @@ -0,0 +1,213 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_AUTH_H_ +#define _NETINET_SCTP_AUTH_H_ + +#include <netinet/sctp_os.h> + +/* digest lengths */ +#define SCTP_AUTH_DIGEST_LEN_SHA1 20 +#define SCTP_AUTH_DIGEST_LEN_SHA256 32 +#define SCTP_AUTH_DIGEST_LEN_MAX SCTP_AUTH_DIGEST_LEN_SHA256 + +/* random sizes */ +#define SCTP_AUTH_RANDOM_SIZE_DEFAULT 32 +#define SCTP_AUTH_RANDOM_SIZE_REQUIRED 32 + +/* union of all supported HMAC algorithm contexts */ +typedef union sctp_hash_context { + SCTP_SHA1_CTX sha1; +#if defined(SCTP_SUPPORT_HMAC_SHA256) + SCTP_SHA256_CTX sha256; +#endif +} sctp_hash_context_t; + +typedef struct sctp_key { + uint32_t keylen; + uint8_t key[]; +} sctp_key_t; + +typedef struct sctp_shared_key { + LIST_ENTRY(sctp_shared_key) next; + sctp_key_t *key; /* key text */ + uint32_t refcount; /* reference count */ + uint16_t keyid; /* shared key ID */ + uint8_t deactivated; /* key is deactivated */ +} sctp_sharedkey_t; + +LIST_HEAD(sctp_keyhead, sctp_shared_key); + +/* authentication chunks list */ +typedef struct sctp_auth_chklist { + uint8_t chunks[256]; + uint8_t num_chunks; +} sctp_auth_chklist_t; + +/* hmac algos supported list */ +typedef struct sctp_hmaclist { + uint16_t max_algo; /* max algorithms allocated */ + uint16_t num_algo; /* num algorithms used */ + uint16_t hmac[]; +} sctp_hmaclist_t; + +/* authentication info */ +typedef struct sctp_authinformation { + sctp_key_t *random; /* local random key (concatenated) */ + uint32_t random_len; /* local random number length for param */ + sctp_key_t *peer_random;/* peer's random key (concatenated) */ + sctp_key_t *assoc_key; /* cached concatenated send key */ + sctp_key_t *recv_key; /* cached concatenated recv key */ + uint16_t active_keyid; /* active send keyid */ + uint16_t assoc_keyid; /* current send keyid (cached) */ + uint16_t recv_keyid; /* last recv keyid (cached) */ +} sctp_authinfo_t; + +/* + * Macros + */ +#define sctp_auth_is_required_chunk(chunk, list) ((list == NULL) ? (0) : (list->chunks[chunk] != 0)) + +/* + * function prototypes + */ + +/* socket option api functions */ +extern sctp_auth_chklist_t *sctp_alloc_chunklist(void); +extern void sctp_free_chunklist(sctp_auth_chklist_t *chklist); +extern void sctp_clear_chunklist(sctp_auth_chklist_t *chklist); +extern sctp_auth_chklist_t *sctp_copy_chunklist(sctp_auth_chklist_t *chklist); +extern int sctp_auth_add_chunk(uint8_t chunk, sctp_auth_chklist_t *list); +extern int sctp_auth_delete_chunk(uint8_t chunk, sctp_auth_chklist_t *list); +extern size_t sctp_auth_get_chklist_size(const sctp_auth_chklist_t *list); +extern int sctp_serialize_auth_chunks(const sctp_auth_chklist_t *list, + uint8_t *ptr); +extern int sctp_pack_auth_chunks(const sctp_auth_chklist_t *list, + uint8_t *ptr); +extern int sctp_unpack_auth_chunks(const uint8_t *ptr, uint8_t num_chunks, + sctp_auth_chklist_t *list); + +/* key handling */ +extern sctp_key_t *sctp_alloc_key(uint32_t keylen); +extern void sctp_free_key(sctp_key_t *key); +extern void sctp_print_key(sctp_key_t *key, const char *str); +extern void sctp_show_key(sctp_key_t *key, const char *str); +extern sctp_key_t *sctp_generate_random_key(uint32_t keylen); +extern sctp_key_t *sctp_set_key(uint8_t *key, uint32_t keylen); +extern sctp_key_t *sctp_compute_hashkey(sctp_key_t *key1, sctp_key_t *key2, + sctp_key_t *shared); + +/* shared key handling */ +extern sctp_sharedkey_t *sctp_alloc_sharedkey(void); +extern void sctp_free_sharedkey(sctp_sharedkey_t *skey); +extern sctp_sharedkey_t *sctp_find_sharedkey(struct sctp_keyhead *shared_keys, + uint16_t key_id); +extern int sctp_insert_sharedkey(struct sctp_keyhead *shared_keys, + sctp_sharedkey_t *new_skey); +extern int sctp_copy_skeylist(const struct sctp_keyhead *src, + struct sctp_keyhead *dest); +/* ref counts on shared keys, by key id */ +extern void sctp_auth_key_acquire(struct sctp_tcb *stcb, uint16_t keyid); +extern void sctp_auth_key_release(struct sctp_tcb *stcb, uint16_t keyid, + int so_locked); + +/* hmac list handling */ +extern sctp_hmaclist_t *sctp_alloc_hmaclist(uint16_t num_hmacs); +extern void sctp_free_hmaclist(sctp_hmaclist_t *list); +extern int sctp_auth_add_hmacid(sctp_hmaclist_t *list, uint16_t hmac_id); +extern sctp_hmaclist_t *sctp_copy_hmaclist(sctp_hmaclist_t *list); +extern sctp_hmaclist_t *sctp_default_supported_hmaclist(void); +extern uint16_t sctp_negotiate_hmacid(sctp_hmaclist_t *peer, + sctp_hmaclist_t *local); +extern int sctp_serialize_hmaclist(sctp_hmaclist_t *list, uint8_t *ptr); +extern int sctp_verify_hmac_param(struct sctp_auth_hmac_algo *hmacs, + uint32_t num_hmacs); + +extern sctp_authinfo_t *sctp_alloc_authinfo(void); +extern void sctp_free_authinfo(sctp_authinfo_t *authinfo); + +/* keyed-HMAC functions */ +extern uint32_t sctp_get_auth_chunk_len(uint16_t hmac_algo); +extern uint32_t sctp_get_hmac_digest_len(uint16_t hmac_algo); +extern uint32_t sctp_hmac(uint16_t hmac_algo, uint8_t *key, uint32_t keylen, + uint8_t *text, uint32_t textlen, uint8_t *digest); +extern uint32_t sctp_compute_hmac(uint16_t hmac_algo, sctp_key_t *key, + uint8_t *text, uint32_t textlen, uint8_t *digest); +extern int sctp_auth_is_supported_hmac(sctp_hmaclist_t *list, uint16_t id); + +/* mbuf versions */ +extern uint32_t sctp_hmac_m(uint16_t hmac_algo, uint8_t *key, uint32_t keylen, + struct mbuf *m, uint32_t m_offset, uint8_t *digest, uint32_t trailer); +extern uint32_t sctp_compute_hmac_m(uint16_t hmac_algo, sctp_key_t *key, + struct mbuf *m, uint32_t m_offset, uint8_t *digest); + +/* + * authentication routines + */ +extern void sctp_clear_cachedkeys(struct sctp_tcb *stcb, uint16_t keyid); +extern void sctp_clear_cachedkeys_ep(struct sctp_inpcb *inp, uint16_t keyid); +extern int sctp_delete_sharedkey(struct sctp_tcb *stcb, uint16_t keyid); +extern int sctp_delete_sharedkey_ep(struct sctp_inpcb *inp, uint16_t keyid); +extern int sctp_auth_setactivekey(struct sctp_tcb *stcb, uint16_t keyid); +extern int sctp_auth_setactivekey_ep(struct sctp_inpcb *inp, uint16_t keyid); +extern int sctp_deact_sharedkey(struct sctp_tcb *stcb, uint16_t keyid); +extern int sctp_deact_sharedkey_ep(struct sctp_inpcb *inp, uint16_t keyid); + +extern void sctp_auth_get_cookie_params(struct sctp_tcb *stcb, struct mbuf *m, + uint32_t offset, uint32_t length); +extern void sctp_fill_hmac_digest_m(struct mbuf *m, uint32_t auth_offset, + struct sctp_auth_chunk *auth, struct sctp_tcb *stcb, uint16_t key_id); +extern struct mbuf *sctp_add_auth_chunk(struct mbuf *m, struct mbuf **m_end, + struct sctp_auth_chunk **auth_ret, uint32_t *offset, + struct sctp_tcb *stcb, uint8_t chunk); +extern int sctp_handle_auth(struct sctp_tcb *stcb, struct sctp_auth_chunk *ch, + struct mbuf *m, uint32_t offset); +extern void sctp_notify_authentication(struct sctp_tcb *stcb, + uint32_t indication, uint16_t keyid, uint16_t alt_keyid, int so_locked); +extern int sctp_validate_init_auth_params(struct mbuf *m, int offset, + int limit); +extern void sctp_initialize_auth_params(struct sctp_inpcb *inp, + struct sctp_tcb *stcb); + +/* test functions */ +#ifdef SCTP_HMAC_TEST +extern void sctp_test_hmac_sha1(void); +extern void sctp_test_authkey(void); +#endif +#endif /* __SCTP_AUTH_H__ */ diff --git a/netwerk/sctp/src/netinet/sctp_bsd_addr.c b/netwerk/sctp/src/netinet/sctp_bsd_addr.c new file mode 100644 index 0000000000..7b40fc4084 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_bsd_addr.c @@ -0,0 +1,982 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#include <netinet/sctp_var.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctp_header.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_bsd_addr.h> +#include <netinet/sctp_uio.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_timer.h> +#include <netinet/sctp_asconf.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_indata.h> +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/unistd.h> +#endif + +/* Declare all of our malloc named types */ +MALLOC_DEFINE(SCTP_M_MAP, "sctp_map", "sctp asoc map descriptor"); +MALLOC_DEFINE(SCTP_M_STRMI, "sctp_stri", "sctp stream in array"); +MALLOC_DEFINE(SCTP_M_STRMO, "sctp_stro", "sctp stream out array"); +MALLOC_DEFINE(SCTP_M_ASC_ADDR, "sctp_aadr", "sctp asconf address"); +MALLOC_DEFINE(SCTP_M_ASC_IT, "sctp_a_it", "sctp asconf iterator"); +MALLOC_DEFINE(SCTP_M_AUTH_CL, "sctp_atcl", "sctp auth chunklist"); +MALLOC_DEFINE(SCTP_M_AUTH_KY, "sctp_atky", "sctp auth key"); +MALLOC_DEFINE(SCTP_M_AUTH_HL, "sctp_athm", "sctp auth hmac list"); +MALLOC_DEFINE(SCTP_M_AUTH_IF, "sctp_athi", "sctp auth info"); +MALLOC_DEFINE(SCTP_M_STRESET, "sctp_stre", "sctp stream reset"); +MALLOC_DEFINE(SCTP_M_CMSG, "sctp_cmsg", "sctp CMSG buffer"); +MALLOC_DEFINE(SCTP_M_COPYAL, "sctp_cpal", "sctp copy all"); +MALLOC_DEFINE(SCTP_M_VRF, "sctp_vrf", "sctp vrf struct"); +MALLOC_DEFINE(SCTP_M_IFA, "sctp_ifa", "sctp ifa struct"); +MALLOC_DEFINE(SCTP_M_IFN, "sctp_ifn", "sctp ifn struct"); +MALLOC_DEFINE(SCTP_M_TIMW, "sctp_timw", "sctp time block"); +MALLOC_DEFINE(SCTP_M_MVRF, "sctp_mvrf", "sctp mvrf pcb list"); +MALLOC_DEFINE(SCTP_M_ITER, "sctp_iter", "sctp iterator control"); +MALLOC_DEFINE(SCTP_M_SOCKOPT, "sctp_socko", "sctp socket option"); +MALLOC_DEFINE(SCTP_M_MCORE, "sctp_mcore", "sctp mcore queue"); + +/* Global NON-VNET structure that controls the iterator */ +struct iterator_control sctp_it_ctl; +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + +static void +sctp_cleanup_itqueue(void) +{ + struct sctp_iterator *it, *nit; + + TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) { + if (it->function_atend != NULL) { + (*it->function_atend) (it->pointer, it->val); + } + TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); + SCTP_FREE(it, SCTP_M_ITER); + } +} +#endif +#if defined(__Userspace__) +/*__Userspace__ TODO if we use thread based iterator + * then the implementation of wakeup will need to change. + * Currently we are using timeo_cond for ident so_timeo + * but that is not sufficient if we need to use another ident + * like wakeup(&sctppcbinfo.iterator_running); + */ +#endif + +void +sctp_wakeup_iterator(void) +{ +#if defined(SCTP_PROCESS_LEVEL_LOCKS) +#if defined(_WIN32) + WakeAllConditionVariable(&sctp_it_ctl.iterator_wakeup); +#else + pthread_cond_broadcast(&sctp_it_ctl.iterator_wakeup); +#endif +#else + wakeup(&sctp_it_ctl.iterator_running); +#endif +} + +#if defined(__Userspace__) +static void * +#else +static void +#endif +sctp_iterator_thread(void *v SCTP_UNUSED) +{ +#if defined(__Userspace__) + sctp_userspace_set_threadname("SCTP iterator"); +#endif + SCTP_IPI_ITERATOR_WQ_LOCK(); + /* In FreeBSD this thread never terminates. */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + for (;;) { +#else + while ((sctp_it_ctl.iterator_flags & SCTP_ITERATOR_MUST_EXIT) == 0) { +#endif +#if !defined(__Userspace__) + msleep(&sctp_it_ctl.iterator_running, +#if defined(__FreeBSD__) + &sctp_it_ctl.ipi_iterator_wq_mtx, +#elif defined(__APPLE__) + sctp_it_ctl.ipi_iterator_wq_mtx, +#endif + 0, "waiting_for_work", 0); +#else +#if defined(_WIN32) + SleepConditionVariableCS(&sctp_it_ctl.iterator_wakeup, &sctp_it_ctl.ipi_iterator_wq_mtx, INFINITE); +#else + pthread_cond_wait(&sctp_it_ctl.iterator_wakeup, &sctp_it_ctl.ipi_iterator_wq_mtx); +#endif +#endif +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + if (sctp_it_ctl.iterator_flags & SCTP_ITERATOR_MUST_EXIT) { + break; + } +#endif + sctp_iterator_worker(); + } +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + /* Now this thread needs to be terminated */ + sctp_cleanup_itqueue(); + sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_EXITED; + SCTP_IPI_ITERATOR_WQ_UNLOCK(); +#if defined(__Userspace__) + sctp_wakeup_iterator(); + return (NULL); +#else + wakeup(&sctp_it_ctl.iterator_flags); + thread_terminate(current_thread()); +#ifdef INVARIANTS + panic("Hmm. thread_terminate() continues..."); +#endif +#endif +#endif +} + +void +sctp_startup_iterator(void) +{ + if (sctp_it_ctl.thread_proc) { + /* You only get one */ + return; + } + /* Initialize global locks here, thus only once. */ + SCTP_ITERATOR_LOCK_INIT(); + SCTP_IPI_ITERATOR_WQ_INIT(); + TAILQ_INIT(&sctp_it_ctl.iteratorhead); +#if defined(__Userspace__) + if (sctp_userspace_thread_create(&sctp_it_ctl.thread_proc, &sctp_iterator_thread)) { + SCTP_PRINTF("ERROR: Creating sctp_iterator_thread failed.\n"); + } else { + SCTP_BASE_VAR(iterator_thread_started) = 1; + } +#elif defined(__FreeBSD__) + kproc_create(sctp_iterator_thread, + (void *)NULL, + &sctp_it_ctl.thread_proc, + 0, + SCTP_KTHREAD_PAGES, + SCTP_KTRHEAD_NAME); +#elif defined(__APPLE__) + kernel_thread_start((thread_continue_t)sctp_iterator_thread, NULL, &sctp_it_ctl.thread_proc); +#endif +} + +#ifdef INET6 + +#if defined(__Userspace__) +/* __Userspace__ TODO. struct in6_ifaddr is defined in sys/netinet6/in6_var.h + ip6_use_deprecated is defined as int ip6_use_deprecated = 1; in /src/sys/netinet6/in6_proto.c + */ +void +sctp_gather_internal_ifa_flags(struct sctp_ifa *ifa) +{ + return; /* stub */ +} +#else +void +sctp_gather_internal_ifa_flags(struct sctp_ifa *ifa) +{ + struct in6_ifaddr *ifa6; + + ifa6 = (struct in6_ifaddr *)ifa->ifa; + ifa->flags = ifa6->ia6_flags; + if (!MODULE_GLOBAL(ip6_use_deprecated)) { + if (ifa->flags & + IN6_IFF_DEPRECATED) { + ifa->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE; + } else { + ifa->localifa_flags &= ~SCTP_ADDR_IFA_UNUSEABLE; + } + } else { + ifa->localifa_flags &= ~SCTP_ADDR_IFA_UNUSEABLE; + } + if (ifa->flags & + (IN6_IFF_DETACHED | + IN6_IFF_ANYCAST | + IN6_IFF_NOTREADY)) { + ifa->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE; + } else { + ifa->localifa_flags &= ~SCTP_ADDR_IFA_UNUSEABLE; + } +} +#endif /* __Userspace__ */ +#endif /* INET6 */ + +#if !defined(__Userspace__) +static uint32_t +sctp_is_desired_interface_type(struct ifnet *ifn) +{ + int result; + + /* check the interface type to see if it's one we care about */ +#if defined(__APPLE__) && !defined(__Userspace__) + switch(ifnet_type(ifn)) { +#else + switch (ifn->if_type) { +#endif + case IFT_ETHER: + case IFT_ISO88023: + case IFT_ISO88024: + case IFT_ISO88025: + case IFT_ISO88026: + case IFT_STARLAN: + case IFT_P10: + case IFT_P80: + case IFT_HY: + case IFT_FDDI: + case IFT_XETHER: + case IFT_ISDNBASIC: + case IFT_ISDNPRIMARY: + case IFT_PTPSERIAL: + case IFT_OTHER: + case IFT_PPP: + case IFT_LOOP: + case IFT_SLIP: + case IFT_GIF: + case IFT_L2VLAN: + case IFT_STF: +#if !(defined(__APPLE__) && !defined(__Userspace__)) + case IFT_IP: + case IFT_IPOVERCDLC: + case IFT_IPOVERCLAW: + case IFT_PROPVIRTUAL: /* NetGraph Virtual too */ + case IFT_VIRTUALIPADDRESS: +#endif + result = 1; + break; + default: + result = 0; + } + + return (result); +} +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + +int +sctp_is_vmware_interface(struct ifnet *ifn) +{ + return (strncmp(ifnet_name(ifn), "vmnet", 5) == 0); +} + +#endif + +#if defined(_WIN32) && defined(__Userspace__) +#define SCTP_BSD_FREE(x) HeapFree(GetProcessHeap(), 0, (x)) +static void +sctp_init_ifns_for_vrf(int vrfid) +{ +#if defined(INET) || defined(INET6) + struct sctp_ifa *sctp_ifa; + DWORD Err, AdapterAddrsSize; + PIP_ADAPTER_ADDRESSES pAdapterAddrs, pAdapt; + PIP_ADAPTER_UNICAST_ADDRESS pUnicast; +#endif + +#ifdef INET + AdapterAddrsSize = 0; + + if ((Err = GetAdaptersAddresses(AF_INET, 0, NULL, NULL, &AdapterAddrsSize)) != 0) { + if ((Err != ERROR_BUFFER_OVERFLOW) && (Err != ERROR_INSUFFICIENT_BUFFER)) { + SCTP_PRINTF("GetAdaptersV4Addresses() sizing failed with error code %d\n", Err); + SCTP_PRINTF("err = %d; AdapterAddrsSize = %d\n", Err, AdapterAddrsSize); + return; + } + } + + /* Allocate memory from sizing information */ + if ((pAdapterAddrs = (PIP_ADAPTER_ADDRESSES) GlobalAlloc(GPTR, AdapterAddrsSize)) == NULL) { + SCTP_PRINTF("Memory allocation error!\n"); + return; + } + /* Get actual adapter information */ + if ((Err = GetAdaptersAddresses(AF_INET, 0, NULL, pAdapterAddrs, &AdapterAddrsSize)) != ERROR_SUCCESS) { + SCTP_PRINTF("GetAdaptersV4Addresses() failed with error code %d\n", Err); + SCTP_BSD_FREE(pAdapterAddrs); + return; + } + /* Enumerate through each returned adapter and save its information */ + for (pAdapt = pAdapterAddrs; pAdapt; pAdapt = pAdapt->Next) { + if (pAdapt->IfType == IF_TYPE_IEEE80211 || pAdapt->IfType == IF_TYPE_ETHERNET_CSMACD) { + for (pUnicast = pAdapt->FirstUnicastAddress; pUnicast; pUnicast = pUnicast->Next) { + if (IN4_ISLINKLOCAL_ADDRESS(&(((struct sockaddr_in *)(pUnicast->Address.lpSockaddr))->sin_addr))) { + continue; + } + sctp_ifa = sctp_add_addr_to_vrf(0, + NULL, + pAdapt->IfIndex, + (pAdapt->IfType == IF_TYPE_IEEE80211)?MIB_IF_TYPE_ETHERNET:pAdapt->IfType, + pAdapt->AdapterName, + NULL, + pUnicast->Address.lpSockaddr, + pAdapt->Flags, + 0); + if (sctp_ifa) { + sctp_ifa->localifa_flags &= ~SCTP_ADDR_DEFER_USE; + } + } + } + } + SCTP_BSD_FREE(pAdapterAddrs); +#endif +#ifdef INET6 + AdapterAddrsSize = 0; + + if ((Err = GetAdaptersAddresses(AF_INET6, 0, NULL, NULL, &AdapterAddrsSize)) != 0) { + if ((Err != ERROR_BUFFER_OVERFLOW) && (Err != ERROR_INSUFFICIENT_BUFFER)) { + SCTP_PRINTF("GetAdaptersV6Addresses() sizing failed with error code %d\n", Err); + SCTP_PRINTF("err = %d; AdapterAddrsSize = %d\n", Err, AdapterAddrsSize); + return; + } + } + /* Allocate memory from sizing information */ + if ((pAdapterAddrs = (PIP_ADAPTER_ADDRESSES) GlobalAlloc(GPTR, AdapterAddrsSize)) == NULL) { + SCTP_PRINTF("Memory allocation error!\n"); + return; + } + /* Get actual adapter information */ + if ((Err = GetAdaptersAddresses(AF_INET6, 0, NULL, pAdapterAddrs, &AdapterAddrsSize)) != ERROR_SUCCESS) { + SCTP_PRINTF("GetAdaptersV6Addresses() failed with error code %d\n", Err); + SCTP_BSD_FREE(pAdapterAddrs); + return; + } + /* Enumerate through each returned adapter and save its information */ + for (pAdapt = pAdapterAddrs; pAdapt; pAdapt = pAdapt->Next) { + if (pAdapt->IfType == IF_TYPE_IEEE80211 || pAdapt->IfType == IF_TYPE_ETHERNET_CSMACD) { + for (pUnicast = pAdapt->FirstUnicastAddress; pUnicast; pUnicast = pUnicast->Next) { + sctp_ifa = sctp_add_addr_to_vrf(0, + NULL, + pAdapt->Ipv6IfIndex, + (pAdapt->IfType == IF_TYPE_IEEE80211)?MIB_IF_TYPE_ETHERNET:pAdapt->IfType, + pAdapt->AdapterName, + NULL, + pUnicast->Address.lpSockaddr, + pAdapt->Flags, + 0); + if (sctp_ifa) { + sctp_ifa->localifa_flags &= ~SCTP_ADDR_DEFER_USE; + } + } + } + } + SCTP_BSD_FREE(pAdapterAddrs); +#endif +} +#elif defined(__Userspace__) +static void +sctp_init_ifns_for_vrf(int vrfid) +{ +#if defined(INET) || defined(INET6) + int rc; + struct ifaddrs *ifa, *ifas; + struct sctp_ifa *sctp_ifa; + uint32_t ifa_flags; + + rc = getifaddrs(&ifas); + if (rc != 0) { + return; + } + for (ifa = ifas; ifa; ifa = ifa->ifa_next) { + if (ifa->ifa_addr == NULL) { + continue; + } +#if !defined(INET) + if (ifa->ifa_addr->sa_family != AF_INET6) { + /* non inet6 skip */ + continue; + } +#elif !defined(INET6) + if (ifa->ifa_addr->sa_family != AF_INET) { + /* non inet skip */ + continue; + } +#else + if ((ifa->ifa_addr->sa_family != AF_INET) && (ifa->ifa_addr->sa_family != AF_INET6)) { + /* non inet/inet6 skip */ + continue; + } +#endif +#if defined(INET6) + if ((ifa->ifa_addr->sa_family == AF_INET6) && + IN6_IS_ADDR_UNSPECIFIED(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr)) { + /* skip unspecified addresses */ + continue; + } +#endif +#if defined(INET) + if (ifa->ifa_addr->sa_family == AF_INET && + ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr == 0) { + continue; + } +#endif + ifa_flags = 0; + sctp_ifa = sctp_add_addr_to_vrf(vrfid, + NULL, + if_nametoindex(ifa->ifa_name), + 0, + ifa->ifa_name, + NULL, + ifa->ifa_addr, + ifa_flags, + 0); + if (sctp_ifa) { + sctp_ifa->localifa_flags &= ~SCTP_ADDR_DEFER_USE; + } + } + freeifaddrs(ifas); +#endif +} +#endif +#if defined(__APPLE__) && !defined(__Userspace__) +static void +sctp_init_ifns_for_vrf(int vrfid) +{ + /* Here we must apply ANY locks needed by the + * IFN we access and also make sure we lock + * any IFA that exists as we float through the + * list of IFA's + */ + struct ifnet **ifnetlist; + uint32_t i, j, count; + char name[SCTP_IFNAMSIZ]; + struct ifnet *ifn; + struct ifaddr **ifaddrlist; + struct ifaddr *ifa; + struct in6_ifaddr *ifa6; + struct sctp_ifa *sctp_ifa; + uint32_t ifa_flags; + + if (ifnet_list_get(IFNET_FAMILY_ANY, &ifnetlist, &count) != 0) { + return; + } + for (i = 0; i < count; i++) { + ifn = ifnetlist[i]; + if (SCTP_BASE_SYSCTL(sctp_ignore_vmware_interfaces) && sctp_is_vmware_interface(ifn)) { + continue; + } + if (sctp_is_desired_interface_type(ifn) == 0) { + /* non desired type */ + continue; + } + if (ifnet_get_address_list(ifn, &ifaddrlist) != 0) { + continue; + } + for (j = 0; ifaddrlist[j] != NULL; j++) { + ifa = ifaddrlist[j]; + if (ifa->ifa_addr == NULL) { + continue; + } + if ((ifa->ifa_addr->sa_family != AF_INET) && (ifa->ifa_addr->sa_family != AF_INET6)) { + /* non inet/inet6 skip */ + continue; + } + if (ifa->ifa_addr->sa_family == AF_INET6) { + if (IN6_IS_ADDR_UNSPECIFIED(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr)) { + /* skip unspecified addresses */ + continue; + } + } else { + if (((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr == INADDR_ANY) { + continue; + } + } + if (ifa->ifa_addr->sa_family == AF_INET6) { + ifa6 = (struct in6_ifaddr *)ifa; + ifa_flags = ifa6->ia6_flags; + } else { + ifa_flags = 0; + } + SCTP_SNPRINTF(name, SCTP_IFNAMSIZ, "%s%d", ifnet_name(ifn), ifnet_unit(ifn)); + sctp_ifa = sctp_add_addr_to_vrf(vrfid, + (void *)ifn, /* XXX */ + ifnet_index(ifn), + ifnet_type(ifn), + name, + (void *)ifa, /* XXX */ + ifa->ifa_addr, + ifa_flags, + 0); + if (sctp_ifa) { + sctp_ifa->localifa_flags &= ~SCTP_ADDR_DEFER_USE; + } + } + ifnet_free_address_list(ifaddrlist); + } + ifnet_list_free(ifnetlist); +} +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) +static void +sctp_init_ifns_for_vrf(int vrfid) +{ + /* Here we must apply ANY locks needed by the + * IFN we access and also make sure we lock + * any IFA that exists as we float through the + * list of IFA's + */ + struct epoch_tracker et; + struct ifnet *ifn; + struct ifaddr *ifa; + struct sctp_ifa *sctp_ifa; + uint32_t ifa_flags; +#ifdef INET6 + struct in6_ifaddr *ifa6; +#endif + + IFNET_RLOCK(); + NET_EPOCH_ENTER(et); + CK_STAILQ_FOREACH(ifn, &MODULE_GLOBAL(ifnet), if_link) { + if (sctp_is_desired_interface_type(ifn) == 0) { + /* non desired type */ + continue; + } + CK_STAILQ_FOREACH(ifa, &ifn->if_addrhead, ifa_link) { + if (ifa->ifa_addr == NULL) { + continue; + } + switch (ifa->ifa_addr->sa_family) { +#ifdef INET + case AF_INET: + if (((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr == 0) { + continue; + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (IN6_IS_ADDR_UNSPECIFIED(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr)) { + /* skip unspecified addresses */ + continue; + } + break; +#endif + default: + continue; + } + switch (ifa->ifa_addr->sa_family) { +#ifdef INET + case AF_INET: + ifa_flags = 0; + break; +#endif +#ifdef INET6 + case AF_INET6: + ifa6 = (struct in6_ifaddr *)ifa; + ifa_flags = ifa6->ia6_flags; + break; +#endif + default: + ifa_flags = 0; + break; + } + sctp_ifa = sctp_add_addr_to_vrf(vrfid, + (void *)ifn, + ifn->if_index, + ifn->if_type, + ifn->if_xname, + (void *)ifa, + ifa->ifa_addr, + ifa_flags, + 0); + if (sctp_ifa) { + sctp_ifa->localifa_flags &= ~SCTP_ADDR_DEFER_USE; + } + } + } + NET_EPOCH_EXIT(et); + IFNET_RUNLOCK(); +} +#endif + +void +sctp_init_vrf_list(int vrfid) +{ + if (vrfid > SCTP_MAX_VRF_ID) + /* can't do that */ + return; + + /* Don't care about return here */ + (void)sctp_allocate_vrf(vrfid); + + /* Now we need to build all the ifn's + * for this vrf and there addresses + */ + sctp_init_ifns_for_vrf(vrfid); +} + +void +sctp_addr_change(struct ifaddr *ifa, int cmd) +{ +#if defined(__Userspace__) + return; +#else + uint32_t ifa_flags = 0; + + if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) { + return; + } + /* BSD only has one VRF, if this changes + * we will need to hook in the right + * things here to get the id to pass to + * the address management routine. + */ + if (SCTP_BASE_VAR(first_time) == 0) { + /* Special test to see if my ::1 will showup with this */ + SCTP_BASE_VAR(first_time) = 1; + sctp_init_ifns_for_vrf(SCTP_DEFAULT_VRFID); + } + + if ((cmd != RTM_ADD) && (cmd != RTM_DELETE)) { + /* don't know what to do with this */ + return; + } + + if (ifa->ifa_addr == NULL) { + return; + } + if (sctp_is_desired_interface_type(ifa->ifa_ifp) == 0) { + /* non desired type */ + return; + } + switch (ifa->ifa_addr->sa_family) { +#ifdef INET + case AF_INET: + if (((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr == 0) { + return; + } + break; +#endif +#ifdef INET6 + case AF_INET6: + ifa_flags = ((struct in6_ifaddr *)ifa)->ia6_flags; + if (IN6_IS_ADDR_UNSPECIFIED(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr)) { + /* skip unspecified addresses */ + return; + } + break; +#endif + default: + /* non inet/inet6 skip */ + return; + } + if (cmd == RTM_ADD) { + (void)sctp_add_addr_to_vrf(SCTP_DEFAULT_VRFID, (void *)ifa->ifa_ifp, +#if defined(__APPLE__) && !defined(__Userspace__) + ifnet_index(ifa->ifa_ifp), ifnet_type(ifa->ifa_ifp), ifnet_name(ifa->ifa_ifp), +#else + ifa->ifa_ifp->if_index, ifa->ifa_ifp->if_type, ifa->ifa_ifp->if_xname, +#endif + (void *)ifa, ifa->ifa_addr, ifa_flags, 1); + } else { + sctp_del_addr_from_vrf(SCTP_DEFAULT_VRFID, ifa->ifa_addr, +#if defined(__APPLE__) && !defined(__Userspace__) + ifnet_index(ifa->ifa_ifp), + ifnet_name(ifa->ifa_ifp)); +#else + ifa->ifa_ifp->if_index, + ifa->ifa_ifp->if_xname); +#endif + + /* We don't bump refcount here so when it completes + * the final delete will happen. + */ + } +#endif +} + +#if defined(__FreeBSD__) && !defined(__Userspace__) +void +sctp_addr_change_event_handler(void *arg __unused, struct ifaddr *ifa, int cmd) { + sctp_addr_change(ifa, cmd); +} +#endif +#if defined(__APPLE__) && !defined(__Userspace__) +void +sctp_add_or_del_interfaces(int (*pred)(struct ifnet *), int add) +{ + struct ifnet **ifnetlist; + struct ifaddr **ifaddrlist; + uint32_t i, j, count; + + if (ifnet_list_get(IFNET_FAMILY_ANY, &ifnetlist, &count) != 0) { + return; + } + for (i = 0; i < count; i++) { + if (!(*pred)(ifnetlist[i])) { + continue; + } + if (ifnet_get_address_list(ifnetlist[i], &ifaddrlist) != 0) { + continue; + } + for (j = 0; ifaddrlist[j] != NULL; j++) { + sctp_addr_change(ifaddrlist[j], add ? RTM_ADD : RTM_DELETE); + } + ifnet_free_address_list(ifaddrlist); + } + ifnet_list_free(ifnetlist); + return; +} +#endif + +struct mbuf * +sctp_get_mbuf_for_msg(unsigned int space_needed, int want_header, + int how, int allonebuf, int type) +{ + struct mbuf *m = NULL; +#if defined(__FreeBSD__) || defined(__Userspace__) +#if defined(__Userspace__) + m = m_getm2(NULL, space_needed, how, type, want_header ? M_PKTHDR : 0, allonebuf); +#else + m = m_getm2(NULL, space_needed, how, type, want_header ? M_PKTHDR : 0); +#endif + if (m == NULL) { + /* bad, no memory */ + return (m); + } +#if !defined(__Userspace__) + if (allonebuf) { + if (SCTP_BUF_SIZE(m) < space_needed) { + m_freem(m); + return (NULL); + } + KASSERT(SCTP_BUF_NEXT(m) == NULL, ("%s: no chain allowed", __func__)); + } +#endif +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mb(m, SCTP_MBUF_IALLOC); + } +#endif +#else + int mbuf_threshold; + unsigned int size; + + if (want_header) { + MGETHDR(m, how, type); + size = MHLEN; + } else { + MGET(m, how, type); + size = MLEN; + } + if (m == NULL) { + return (NULL); + } + if (allonebuf == 0) { + mbuf_threshold = SCTP_BASE_SYSCTL(sctp_mbuf_threshold_count); + } else { + mbuf_threshold = 1; + } + + if (space_needed > (unsigned int)(((mbuf_threshold - 1) * MLEN) + MHLEN)) { + MCLGET(m, how); + if (m == NULL) { + return (NULL); + } + if (SCTP_BUF_IS_EXTENDED(m) == 0) { + sctp_m_freem(m); + return (NULL); + } + size = SCTP_BUF_EXTEND_SIZE(m); + } + if (allonebuf != 0 && size < space_needed) { + m_freem(m); + return (NULL); + } + SCTP_BUF_LEN(m) = 0; + SCTP_BUF_NEXT(m) = SCTP_BUF_NEXT_PKT(m) = NULL; +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mb(m, SCTP_MBUF_IALLOC); + } +#endif +#endif + return (m); +} + +#ifdef SCTP_PACKET_LOGGING +void +sctp_packet_log(struct mbuf *m) +{ + int *lenat, thisone; + void *copyto; + uint32_t *tick_tock; + int length; + int total_len; + int grabbed_lock = 0; + int value, newval, thisend, thisbegin; + /* + * Buffer layout. + * -sizeof this entry (total_len) + * -previous end (value) + * -ticks of log (ticks) + * o -ip packet + * o -as logged + * - where this started (thisbegin) + * x <--end points here + */ + length = SCTP_HEADER_LEN(m); + total_len = SCTP_SIZE32((length + (4 * sizeof(int)))); + /* Log a packet to the buffer. */ + if (total_len> SCTP_PACKET_LOG_SIZE) { + /* Can't log this packet I have not a buffer big enough */ + return; + } + if (length < (int)(SCTP_MIN_V4_OVERHEAD + sizeof(struct sctp_cookie_ack_chunk))) { + return; + } + atomic_add_int(&SCTP_BASE_VAR(packet_log_writers), 1); + try_again: + if (SCTP_BASE_VAR(packet_log_writers) > SCTP_PKTLOG_WRITERS_NEED_LOCK) { + SCTP_IP_PKTLOG_LOCK(); + grabbed_lock = 1; + again_locked: + value = SCTP_BASE_VAR(packet_log_end); + newval = SCTP_BASE_VAR(packet_log_end) + total_len; + if (newval >= SCTP_PACKET_LOG_SIZE) { + /* we wrapped */ + thisbegin = 0; + thisend = total_len; + } else { + thisbegin = SCTP_BASE_VAR(packet_log_end); + thisend = newval; + } + if (!(atomic_cmpset_int(&SCTP_BASE_VAR(packet_log_end), value, thisend))) { + goto again_locked; + } + } else { + value = SCTP_BASE_VAR(packet_log_end); + newval = SCTP_BASE_VAR(packet_log_end) + total_len; + if (newval >= SCTP_PACKET_LOG_SIZE) { + /* we wrapped */ + thisbegin = 0; + thisend = total_len; + } else { + thisbegin = SCTP_BASE_VAR(packet_log_end); + thisend = newval; + } + if (!(atomic_cmpset_int(&SCTP_BASE_VAR(packet_log_end), value, thisend))) { + goto try_again; + } + } + /* Sanity check */ + if (thisend >= SCTP_PACKET_LOG_SIZE) { + SCTP_PRINTF("Insanity stops a log thisbegin:%d thisend:%d writers:%d lock:%d end:%d\n", + thisbegin, + thisend, + SCTP_BASE_VAR(packet_log_writers), + grabbed_lock, + SCTP_BASE_VAR(packet_log_end)); + SCTP_BASE_VAR(packet_log_end) = 0; + goto no_log; + } + lenat = (int *)&SCTP_BASE_VAR(packet_log_buffer)[thisbegin]; + *lenat = total_len; + lenat++; + *lenat = value; + lenat++; + tick_tock = (uint32_t *)lenat; + lenat++; + *tick_tock = sctp_get_tick_count(); + copyto = (void *)lenat; + thisone = thisend - sizeof(int); + lenat = (int *)&SCTP_BASE_VAR(packet_log_buffer)[thisone]; + *lenat = thisbegin; + if (grabbed_lock) { + SCTP_IP_PKTLOG_UNLOCK(); + grabbed_lock = 0; + } + m_copydata(m, 0, length, (caddr_t)copyto); + no_log: + if (grabbed_lock) { + SCTP_IP_PKTLOG_UNLOCK(); + } + atomic_subtract_int(&SCTP_BASE_VAR(packet_log_writers), 1); +} + +int +sctp_copy_out_packet_log(uint8_t *target, int length) +{ + /* We wind through the packet log starting at + * start copying up to length bytes out. + * We return the number of bytes copied. + */ + int this_copy; + int *lenat; + int did_delay = 0; + + if (length < (int)(2 * sizeof(int))) { + /* not enough room */ + return (0); + } + if (SCTP_PKTLOG_WRITERS_NEED_LOCK) { + atomic_add_int(&SCTP_BASE_VAR(packet_log_writers), SCTP_PKTLOG_WRITERS_NEED_LOCK); + again: + if ((did_delay == 0) && (SCTP_BASE_VAR(packet_log_writers) != SCTP_PKTLOG_WRITERS_NEED_LOCK)) { + /* we delay here for just a moment hoping the writer(s) that were + * present when we entered will have left and we only have + * locking ones that will contend with us for the lock. This + * does not assure 100% access, but its good enough for + * a logging facility like this. + */ + did_delay = 1; + DELAY(10); + goto again; + } + } + SCTP_IP_PKTLOG_LOCK(); + lenat = (int *)target; + *lenat = SCTP_BASE_VAR(packet_log_end); + lenat++; + this_copy = min((length - sizeof(int)), SCTP_PACKET_LOG_SIZE); + memcpy((void *)lenat, (void *)SCTP_BASE_VAR(packet_log_buffer), this_copy); + if (SCTP_PKTLOG_WRITERS_NEED_LOCK) { + atomic_subtract_int(&SCTP_BASE_VAR(packet_log_writers), + SCTP_PKTLOG_WRITERS_NEED_LOCK); + } + SCTP_IP_PKTLOG_UNLOCK(); + return (this_copy + sizeof(int)); +} + +#endif diff --git a/netwerk/sctp/src/netinet/sctp_bsd_addr.h b/netwerk/sctp/src/netinet/sctp_bsd_addr.h new file mode 100644 index 0000000000..bbbc933adc --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_bsd_addr.h @@ -0,0 +1,72 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_BSD_ADDR_H_ +#define _NETINET_SCTP_BSD_ADDR_H_ + +#include <netinet/sctp_pcb.h> + +#if defined(_KERNEL) || defined(__Userspace__) + +extern struct iterator_control sctp_it_ctl; +void sctp_wakeup_iterator(void); + +void sctp_startup_iterator(void); + +#ifdef INET6 +void sctp_gather_internal_ifa_flags(struct sctp_ifa *ifa); +#endif + +#ifdef SCTP_PACKET_LOGGING + +void sctp_packet_log(struct mbuf *m); +int sctp_copy_out_packet_log(uint8_t *target, int length); + +#endif + +void sctp_addr_change(struct ifaddr *ifa, int cmd); +#if defined(__FreeBSD__) && !defined(__Userspace__) + +void sctp_addr_change_event_handler(void *, struct ifaddr *, int); +#endif + +void sctp_add_or_del_interfaces(int (*pred)(struct ifnet *), int add); + +#endif +#endif diff --git a/netwerk/sctp/src/netinet/sctp_callout.c b/netwerk/sctp/src/netinet/sctp_callout.c new file mode 100644 index 0000000000..ee6cd4cddd --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_callout.c @@ -0,0 +1,249 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__Userspace__) +#include <sys/types.h> +#if !defined(_WIN32) +#include <sys/wait.h> +#include <unistd.h> +#include <pthread.h> +#endif +#if defined(__native_client__) +#include <sys/select.h> +#endif +#include <stdlib.h> +#include <string.h> +#include <stdio.h> +#include <errno.h> +#include <user_atomic.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_pcb.h> +#else +#include <netinet/sctp_os.h> +#include <netinet/sctp_callout.h> +#include <netinet/sctp_pcb.h> +#endif +#include <netinet/sctputil.h> + +/* + * Callout/Timer routines for OS that doesn't have them + */ +#if defined(__APPLE__) || defined(__Userspace__) +static uint32_t ticks = 0; +#else +extern int ticks; +#endif + +uint32_t sctp_get_tick_count(void) { + uint32_t ret; + + SCTP_TIMERQ_LOCK(); + ret = ticks; + SCTP_TIMERQ_UNLOCK(); + return ret; +} + +/* + * SCTP_TIMERQ_LOCK protects: + * - SCTP_BASE_INFO(callqueue) + * - sctp_os_timer_next: next timer to check + */ +static sctp_os_timer_t *sctp_os_timer_next = NULL; + +void +sctp_os_timer_init(sctp_os_timer_t *c) +{ + memset(c, 0, sizeof(*c)); +} + +int +sctp_os_timer_start(sctp_os_timer_t *c, uint32_t to_ticks, void (*ftn) (void *), + void *arg) +{ + int ret = 0; + + /* paranoia */ + if ((c == NULL) || (ftn == NULL)) + return (ret); + + SCTP_TIMERQ_LOCK(); + /* check to see if we're rescheduling a timer */ + if (c->c_flags & SCTP_CALLOUT_PENDING) { + ret = 1; + if (c == sctp_os_timer_next) { + sctp_os_timer_next = TAILQ_NEXT(c, tqe); + } + TAILQ_REMOVE(&SCTP_BASE_INFO(callqueue), c, tqe); + /* + * part of the normal "stop a pending callout" process + * is to clear the CALLOUT_ACTIVE and CALLOUT_PENDING + * flags. We don't bother since we are setting these + * below and we still hold the lock. + */ + } + + /* + * We could unlock/splx here and lock/spl at the TAILQ_INSERT_TAIL, + * but there's no point since doing this setup doesn't take much time. + */ + if (to_ticks == 0) + to_ticks = 1; + + c->c_arg = arg; + c->c_flags = (SCTP_CALLOUT_ACTIVE | SCTP_CALLOUT_PENDING); + c->c_func = ftn; + c->c_time = ticks + to_ticks; + TAILQ_INSERT_TAIL(&SCTP_BASE_INFO(callqueue), c, tqe); + SCTP_TIMERQ_UNLOCK(); + return (ret); +} + +int +sctp_os_timer_stop(sctp_os_timer_t *c) +{ + SCTP_TIMERQ_LOCK(); + /* + * Don't attempt to delete a callout that's not on the queue. + */ + if ((c->c_flags & SCTP_CALLOUT_PENDING) == 0) { + c->c_flags &= ~SCTP_CALLOUT_ACTIVE; + SCTP_TIMERQ_UNLOCK(); + return (0); + } + c->c_flags &= ~(SCTP_CALLOUT_ACTIVE | SCTP_CALLOUT_PENDING); + if (c == sctp_os_timer_next) { + sctp_os_timer_next = TAILQ_NEXT(c, tqe); + } + TAILQ_REMOVE(&SCTP_BASE_INFO(callqueue), c, tqe); + SCTP_TIMERQ_UNLOCK(); + return (1); +} + +void +sctp_handle_tick(uint32_t elapsed_ticks) +{ + sctp_os_timer_t *c; + void (*c_func)(void *); + void *c_arg; + + SCTP_TIMERQ_LOCK(); + /* update our tick count */ + ticks += elapsed_ticks; + c = TAILQ_FIRST(&SCTP_BASE_INFO(callqueue)); + while (c) { + if (SCTP_UINT32_GE(ticks, c->c_time)) { + sctp_os_timer_next = TAILQ_NEXT(c, tqe); + TAILQ_REMOVE(&SCTP_BASE_INFO(callqueue), c, tqe); + c_func = c->c_func; + c_arg = c->c_arg; + c->c_flags &= ~SCTP_CALLOUT_PENDING; + SCTP_TIMERQ_UNLOCK(); + c_func(c_arg); + SCTP_TIMERQ_LOCK(); + c = sctp_os_timer_next; + } else { + c = TAILQ_NEXT(c, tqe); + } + } + sctp_os_timer_next = NULL; + SCTP_TIMERQ_UNLOCK(); +} + +#if defined(__APPLE__) && !defined(__Userspace__) +void +sctp_timeout(void *arg SCTP_UNUSED) +{ + sctp_handle_tick(SCTP_BASE_VAR(sctp_main_timer_ticks)); + sctp_start_main_timer(); +} +#endif + +#if defined(__Userspace__) +#define TIMEOUT_INTERVAL 10 + +void * +user_sctp_timer_iterate(void *arg) +{ + sctp_userspace_set_threadname("SCTP timer"); + for (;;) { +#if defined(_WIN32) + Sleep(TIMEOUT_INTERVAL); +#else + struct timespec amount, remaining; + + remaining.tv_sec = 0; + remaining.tv_nsec = TIMEOUT_INTERVAL * 1000 * 1000; + do { + amount = remaining; + } while (nanosleep(&amount, &remaining) == -1); +#endif + if (atomic_cmpset_int(&SCTP_BASE_VAR(timer_thread_should_exit), 1, 1)) { + break; + } + sctp_handle_tick(sctp_msecs_to_ticks(TIMEOUT_INTERVAL)); + } + return (NULL); +} + +void +sctp_start_timer_thread(void) +{ + /* + * No need to do SCTP_TIMERQ_LOCK_INIT(); + * here, it is being done in sctp_pcb_init() + */ + int rc; + + rc = sctp_userspace_thread_create(&SCTP_BASE_VAR(timer_thread), user_sctp_timer_iterate); + if (rc) { + SCTP_PRINTF("ERROR; return code from sctp_thread_create() is %d\n", rc); + } else { + SCTP_BASE_VAR(timer_thread_started) = 1; + } +} + +void +sctp_stop_timer_thread(void) +{ + atomic_cmpset_int(&SCTP_BASE_VAR(timer_thread_should_exit), 0, 1); + if (SCTP_BASE_VAR(timer_thread_started)) { +#if defined(_WIN32) + WaitForSingleObject(SCTP_BASE_VAR(timer_thread), INFINITE); + CloseHandle(SCTP_BASE_VAR(timer_thread)); +#else + pthread_join(SCTP_BASE_VAR(timer_thread), NULL); +#endif + } +} +#endif diff --git a/netwerk/sctp/src/netinet/sctp_callout.h b/netwerk/sctp/src/netinet/sctp_callout.h new file mode 100644 index 0000000000..81fd8530d1 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_callout.h @@ -0,0 +1,119 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the project nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_CALLOUT_ +#define _NETINET_SCTP_CALLOUT_ + +/* + * NOTE: the following MACROS are required for locking the callout + * queue along with a lock/mutex in the OS specific headers and + * implementation files:: + * - SCTP_TIMERQ_LOCK() + * - SCTP_TIMERQ_UNLOCK() + * - SCTP_TIMERQ_LOCK_INIT() + * - SCTP_TIMERQ_LOCK_DESTROY() + */ + +#define _SCTP_NEEDS_CALLOUT_ 1 + +#define SCTP_TICKS_PER_FASTTIMO 20 /* called about every 20ms */ + +#if defined(__Userspace__) +#if defined(_WIN32) +#define SCTP_TIMERQ_LOCK() EnterCriticalSection(&SCTP_BASE_VAR(timer_mtx)) +#define SCTP_TIMERQ_UNLOCK() LeaveCriticalSection(&SCTP_BASE_VAR(timer_mtx)) +#define SCTP_TIMERQ_LOCK_INIT() InitializeCriticalSection(&SCTP_BASE_VAR(timer_mtx)) +#define SCTP_TIMERQ_LOCK_DESTROY() DeleteCriticalSection(&SCTP_BASE_VAR(timer_mtx)) +#else +#ifdef INVARIANTS +#define SCTP_TIMERQ_LOCK() KASSERT(pthread_mutex_lock(&SCTP_BASE_VAR(timer_mtx)) == 0, ("%s: timer_mtx already locked", __func__)) +#define SCTP_TIMERQ_UNLOCK() KASSERT(pthread_mutex_unlock(&SCTP_BASE_VAR(timer_mtx)) == 0, ("%s: timer_mtx not locked", __func__)) +#else +#define SCTP_TIMERQ_LOCK() (void)pthread_mutex_lock(&SCTP_BASE_VAR(timer_mtx)) +#define SCTP_TIMERQ_UNLOCK() (void)pthread_mutex_unlock(&SCTP_BASE_VAR(timer_mtx)) +#endif +#define SCTP_TIMERQ_LOCK_INIT() (void)pthread_mutex_init(&SCTP_BASE_VAR(timer_mtx), &SCTP_BASE_VAR(mtx_attr)) +#define SCTP_TIMERQ_LOCK_DESTROY() (void)pthread_mutex_destroy(&SCTP_BASE_VAR(timer_mtx)) +#endif +#endif + +uint32_t sctp_get_tick_count(void); + +TAILQ_HEAD(calloutlist, sctp_callout); + +struct sctp_callout { + TAILQ_ENTRY(sctp_callout) tqe; + uint32_t c_time; /* ticks to the event */ + void *c_arg; /* function argument */ + void (*c_func)(void *); /* function to call */ + int c_flags; /* state of this entry */ +}; +typedef struct sctp_callout sctp_os_timer_t; + +#define SCTP_CALLOUT_ACTIVE 0x0002 /* callout is currently active */ +#define SCTP_CALLOUT_PENDING 0x0004 /* callout is waiting for timeout */ + +void sctp_os_timer_init(sctp_os_timer_t *tmr); +/* Returns 1 if pending timer was rescheduled, 0 otherwise. */ +int sctp_os_timer_start(sctp_os_timer_t *, uint32_t, void (*)(void *), void *); +/* Returns 1 if pending timer was stopped, 0 otherwise. */ +int sctp_os_timer_stop(sctp_os_timer_t *); +void sctp_handle_tick(uint32_t); + +#define SCTP_OS_TIMER_INIT sctp_os_timer_init +/* + * NOTE: The next two shouldn't be called directly outside of sctp_timer_start() + * and sctp_timer_stop(), since they don't handle incrementing/decrementing + * relevant reference counts. + */ +#define SCTP_OS_TIMER_START sctp_os_timer_start +#define SCTP_OS_TIMER_STOP sctp_os_timer_stop +/* MT FIXME: Is the following correct? */ +#define SCTP_OS_TIMER_STOP_DRAIN SCTP_OS_TIMER_STOP +#define SCTP_OS_TIMER_PENDING(tmr) ((tmr)->c_flags & SCTP_CALLOUT_PENDING) +#define SCTP_OS_TIMER_ACTIVE(tmr) ((tmr)->c_flags & SCTP_CALLOUT_ACTIVE) +#define SCTP_OS_TIMER_DEACTIVATE(tmr) ((tmr)->c_flags &= ~SCTP_CALLOUT_ACTIVE) + +#if defined(__Userspace__) +void sctp_start_timer_thread(void); +void sctp_stop_timer_thread(void); +#endif +#if defined(__APPLE__) && !defined(__Userspace__) +void sctp_timeout(void *); +#endif + +#endif diff --git a/netwerk/sctp/src/netinet/sctp_cc_functions.c b/netwerk/sctp/src/netinet/sctp_cc_functions.c new file mode 100644 index 0000000000..bf2ba8ab54 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_cc_functions.c @@ -0,0 +1,2493 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#include <netinet/sctp_var.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctp_header.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_input.h> +#include <netinet/sctp_indata.h> +#include <netinet/sctp_uio.h> +#include <netinet/sctp_timer.h> +#include <netinet/sctp_auth.h> +#include <netinet/sctp_asconf.h> +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <netinet/sctp_kdtrace.h> +#endif + +#define SHIFT_MPTCP_MULTI_N 40 +#define SHIFT_MPTCP_MULTI_Z 16 +#define SHIFT_MPTCP_MULTI 8 + +static void +sctp_enforce_cwnd_limit(struct sctp_association *assoc, struct sctp_nets *net) +{ + if ((assoc->max_cwnd > 0) && + (net->cwnd > assoc->max_cwnd) && + (net->cwnd > (net->mtu - sizeof(struct sctphdr)))) { + net->cwnd = assoc->max_cwnd; + if (net->cwnd < (net->mtu - sizeof(struct sctphdr))) { + net->cwnd = net->mtu - sizeof(struct sctphdr); + } + } +} + +static void +sctp_set_initial_cc_param(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + struct sctp_association *assoc; + uint32_t cwnd_in_mtu; + + assoc = &stcb->asoc; + cwnd_in_mtu = SCTP_BASE_SYSCTL(sctp_initial_cwnd); + if (cwnd_in_mtu == 0) { + /* Using 0 means that the value of RFC 4960 is used. */ + net->cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND)); + } else { + /* + * We take the minimum of the burst limit and the + * initial congestion window. + */ + if ((assoc->max_burst > 0) && (cwnd_in_mtu > assoc->max_burst)) + cwnd_in_mtu = assoc->max_burst; + net->cwnd = (net->mtu - sizeof(struct sctphdr)) * cwnd_in_mtu; + } + if ((stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV1) || + (stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV2)) { + /* In case of resource pooling initialize appropriately */ + net->cwnd /= assoc->numnets; + if (net->cwnd < (net->mtu - sizeof(struct sctphdr))) { + net->cwnd = net->mtu - sizeof(struct sctphdr); + } + } + sctp_enforce_cwnd_limit(assoc, net); + net->ssthresh = assoc->peers_rwnd; +#if defined(__FreeBSD__) && !defined(__Userspace__) + SDT_PROBE5(sctp, cwnd, net, init, + stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net, + 0, net->cwnd); +#endif + if (SCTP_BASE_SYSCTL(sctp_logging_level) & + (SCTP_CWND_MONITOR_ENABLE|SCTP_CWND_LOGGING_ENABLE)) { + sctp_log_cwnd(stcb, net, 0, SCTP_CWND_INITIALIZATION); + } +} + +static void +sctp_cwnd_update_after_fr(struct sctp_tcb *stcb, + struct sctp_association *asoc) +{ + struct sctp_nets *net; + uint32_t t_ssthresh, t_cwnd; + uint64_t t_ucwnd_sbw; + + /* MT FIXME: Don't compute this over and over again */ + t_ssthresh = 0; + t_cwnd = 0; + t_ucwnd_sbw = 0; + if ((asoc->sctp_cmt_on_off == SCTP_CMT_RPV1) || + (asoc->sctp_cmt_on_off == SCTP_CMT_RPV2)) { + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + t_ssthresh += net->ssthresh; + t_cwnd += net->cwnd; + if (net->lastsa > 0) { + t_ucwnd_sbw += (uint64_t)net->cwnd / (uint64_t)net->lastsa; + } + } + if (t_ucwnd_sbw == 0) { + t_ucwnd_sbw = 1; + } + } + + /*- + * CMT fast recovery code. Need to debug. ((sctp_cmt_on_off > 0) && + * (net->fast_retran_loss_recovery == 0))) + */ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if ((asoc->fast_retran_loss_recovery == 0) || + (asoc->sctp_cmt_on_off > 0)) { + /* out of a RFC2582 Fast recovery window? */ + if (net->net_ack > 0) { + /* + * per section 7.2.3, are there any + * destinations that had a fast retransmit + * to them. If so what we need to do is + * adjust ssthresh and cwnd. + */ + struct sctp_tmit_chunk *lchk; + int old_cwnd = net->cwnd; + + if ((asoc->sctp_cmt_on_off == SCTP_CMT_RPV1) || + (asoc->sctp_cmt_on_off == SCTP_CMT_RPV2)) { + if (asoc->sctp_cmt_on_off == SCTP_CMT_RPV1) { + net->ssthresh = (uint32_t)(((uint64_t)4 * + (uint64_t)net->mtu * + (uint64_t)net->ssthresh) / + (uint64_t)t_ssthresh); + } + if (asoc->sctp_cmt_on_off == SCTP_CMT_RPV2) { + uint32_t srtt; + + srtt = net->lastsa; + /* lastsa>>3; we don't need to devide ...*/ + if (srtt == 0) { + srtt = 1; + } + /* Short Version => Equal to Contel Version MBe */ + net->ssthresh = (uint32_t) (((uint64_t)4 * + (uint64_t)net->mtu * + (uint64_t)net->cwnd) / + ((uint64_t)srtt * + t_ucwnd_sbw)); + /* INCREASE FACTOR */; + } + if ((net->cwnd > t_cwnd / 2) && + (net->ssthresh < net->cwnd - t_cwnd / 2)) { + net->ssthresh = net->cwnd - t_cwnd / 2; + } + if (net->ssthresh < net->mtu) { + net->ssthresh = net->mtu; + } + } else { + net->ssthresh = net->cwnd / 2; + if (net->ssthresh < (net->mtu * 2)) { + net->ssthresh = 2 * net->mtu; + } + } + net->cwnd = net->ssthresh; + sctp_enforce_cwnd_limit(asoc, net); +#if defined(__FreeBSD__) && !defined(__Userspace__) + SDT_PROBE5(sctp, cwnd, net, fr, + stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net, + old_cwnd, net->cwnd); +#endif + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), + SCTP_CWND_LOG_FROM_FR); + } + lchk = TAILQ_FIRST(&asoc->send_queue); + + net->partial_bytes_acked = 0; + /* Turn on fast recovery window */ + asoc->fast_retran_loss_recovery = 1; + if (lchk == NULL) { + /* Mark end of the window */ + asoc->fast_recovery_tsn = asoc->sending_seq - 1; + } else { + asoc->fast_recovery_tsn = lchk->rec.data.tsn - 1; + } + + /* + * CMT fast recovery -- per destination + * recovery variable. + */ + net->fast_retran_loss_recovery = 1; + + if (lchk == NULL) { + /* Mark end of the window */ + net->fast_recovery_tsn = asoc->sending_seq - 1; + } else { + net->fast_recovery_tsn = lchk->rec.data.tsn - 1; + } + + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, + stcb->sctp_ep, stcb, net, + SCTP_FROM_SCTP_CC_FUNCTIONS + SCTP_LOC_1); + sctp_timer_start(SCTP_TIMER_TYPE_SEND, + stcb->sctp_ep, stcb, net); + } + } else if (net->net_ack > 0) { + /* + * Mark a peg that we WOULD have done a cwnd + * reduction but RFC2582 prevented this action. + */ + SCTP_STAT_INCR(sctps_fastretransinrtt); + } + } +} + +/* Defines for instantaneous bw decisions */ +#define SCTP_INST_LOOSING 1 /* Losing to other flows */ +#define SCTP_INST_NEUTRAL 2 /* Neutral, no indication */ +#define SCTP_INST_GAINING 3 /* Gaining, step down possible */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +static int +cc_bw_same(struct sctp_tcb *stcb, struct sctp_nets *net, uint64_t nbw, + uint64_t rtt_offset, uint64_t vtag, uint8_t inst_ind) +#else +static int +cc_bw_same(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_nets *net, uint64_t nbw, + uint64_t rtt_offset, uint8_t inst_ind) +#endif +{ +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint64_t oth, probepoint; +#endif + +#if defined(__FreeBSD__) && !defined(__Userspace__) + probepoint = (((uint64_t)net->cwnd) << 32); +#endif + if (net->rtt > net->cc_mod.rtcc.lbw_rtt + rtt_offset) { + /* + * rtt increased + * we don't update bw.. so we don't + * update the rtt either. + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* Probe point 5 */ + probepoint |= ((5 << 16) | 1); + SDT_PROBE5(sctp, cwnd, net, rttvar, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + net->flight_size, + probepoint); +#endif + if ((net->cc_mod.rtcc.steady_step) && (inst_ind != SCTP_INST_LOOSING)) { + if (net->cc_mod.rtcc.last_step_state == 5) + net->cc_mod.rtcc.step_cnt++; + else + net->cc_mod.rtcc.step_cnt = 1; + net->cc_mod.rtcc.last_step_state = 5; + if ((net->cc_mod.rtcc.step_cnt == net->cc_mod.rtcc.steady_step) || + ((net->cc_mod.rtcc.step_cnt > net->cc_mod.rtcc.steady_step) && + ((net->cc_mod.rtcc.step_cnt % net->cc_mod.rtcc.steady_step) == 0))) { + /* Try a step down */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + oth = net->cc_mod.rtcc.vol_reduce; + oth <<= 16; + oth |= net->cc_mod.rtcc.step_cnt; + oth <<= 16; + oth |= net->cc_mod.rtcc.last_step_state; + SDT_PROBE5(sctp, cwnd, net, rttstep, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + oth, + probepoint); +#endif + if (net->cwnd > (4 * net->mtu)) { + net->cwnd -= net->mtu; + net->cc_mod.rtcc.vol_reduce++; + } else { + net->cc_mod.rtcc.step_cnt = 0; + } + } + } + return (1); + } + if (net->rtt < net->cc_mod.rtcc.lbw_rtt-rtt_offset) { + /* + * rtt decreased, there could be more room. + * we update both the bw and the rtt here to + * lock this in as a good step down. + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* Probe point 6 */ + probepoint |= ((6 << 16) | 0); + SDT_PROBE5(sctp, cwnd, net, rttvar, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + net->flight_size, + probepoint); +#endif + if (net->cc_mod.rtcc.steady_step) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + oth = net->cc_mod.rtcc.vol_reduce; + oth <<= 16; + oth |= net->cc_mod.rtcc.step_cnt; + oth <<= 16; + oth |= net->cc_mod.rtcc.last_step_state; + SDT_PROBE5(sctp, cwnd, net, rttstep, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + oth, + probepoint); +#endif + if ((net->cc_mod.rtcc.last_step_state == 5) && + (net->cc_mod.rtcc.step_cnt > net->cc_mod.rtcc.steady_step)) { + /* Step down worked */ + net->cc_mod.rtcc.step_cnt = 0; + return (1); + } else { + net->cc_mod.rtcc.last_step_state = 6; + net->cc_mod.rtcc.step_cnt = 0; + } + } + net->cc_mod.rtcc.lbw = nbw; + net->cc_mod.rtcc.lbw_rtt = net->rtt; + net->cc_mod.rtcc.cwnd_at_bw_set = net->cwnd; + if (inst_ind == SCTP_INST_GAINING) + return (1); + else if (inst_ind == SCTP_INST_NEUTRAL) + return (1); + else + return (0); + } + /* Ok bw and rtt remained the same .. no update to any + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* Probe point 7 */ + probepoint |= ((7 << 16) | net->cc_mod.rtcc.ret_from_eq); + SDT_PROBE5(sctp, cwnd, net, rttvar, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + net->flight_size, + probepoint); +#endif + if ((net->cc_mod.rtcc.steady_step) && (inst_ind != SCTP_INST_LOOSING)) { + if (net->cc_mod.rtcc.last_step_state == 5) + net->cc_mod.rtcc.step_cnt++; + else + net->cc_mod.rtcc.step_cnt = 1; + net->cc_mod.rtcc.last_step_state = 5; + if ((net->cc_mod.rtcc.step_cnt == net->cc_mod.rtcc.steady_step) || + ((net->cc_mod.rtcc.step_cnt > net->cc_mod.rtcc.steady_step) && + ((net->cc_mod.rtcc.step_cnt % net->cc_mod.rtcc.steady_step) == 0))) { + /* Try a step down */ + if (net->cwnd > (4 * net->mtu)) { + net->cwnd -= net->mtu; + net->cc_mod.rtcc.vol_reduce++; + return (1); + } else { + net->cc_mod.rtcc.step_cnt = 0; + } + } + } + if (inst_ind == SCTP_INST_GAINING) + return (1); + else if (inst_ind == SCTP_INST_NEUTRAL) + return (1); + else + return ((int)net->cc_mod.rtcc.ret_from_eq); +} + +#if defined(__FreeBSD__) && !defined(__Userspace__) +static int +cc_bw_decrease(struct sctp_tcb *stcb, struct sctp_nets *net, uint64_t nbw, uint64_t rtt_offset, + uint64_t vtag, uint8_t inst_ind) +#else +static int +cc_bw_decrease(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_nets *net, uint64_t nbw, uint64_t rtt_offset, + uint8_t inst_ind) +#endif +{ +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint64_t oth, probepoint; +#endif + + /* Bandwidth decreased.*/ +#if defined(__FreeBSD__) && !defined(__Userspace__) + probepoint = (((uint64_t)net->cwnd) << 32); +#endif + if (net->rtt > net->cc_mod.rtcc.lbw_rtt+rtt_offset) { + /* rtt increased */ + /* Did we add more */ + if ((net->cwnd > net->cc_mod.rtcc.cwnd_at_bw_set) && + (inst_ind != SCTP_INST_LOOSING)) { + /* We caused it maybe.. back off? */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* PROBE POINT 1 */ + probepoint |= ((1 << 16) | 1); + SDT_PROBE5(sctp, cwnd, net, rttvar, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + net->flight_size, + probepoint); +#endif + if (net->cc_mod.rtcc.ret_from_eq) { + /* Switch over to CA if we are less aggressive */ + net->ssthresh = net->cwnd-1; + net->partial_bytes_acked = 0; + } + return (1); + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* Probe point 2 */ + probepoint |= ((2 << 16) | 0); + SDT_PROBE5(sctp, cwnd, net, rttvar, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + net->flight_size, + probepoint); +#endif + /* Someone else - fight for more? */ + if (net->cc_mod.rtcc.steady_step) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + oth = net->cc_mod.rtcc.vol_reduce; + oth <<= 16; + oth |= net->cc_mod.rtcc.step_cnt; + oth <<= 16; + oth |= net->cc_mod.rtcc.last_step_state; + SDT_PROBE5(sctp, cwnd, net, rttstep, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + oth, + probepoint); +#endif + /* Did we voluntarily give up some? if so take + * one back please + */ + if ((net->cc_mod.rtcc.vol_reduce) && + (inst_ind != SCTP_INST_GAINING)) { + net->cwnd += net->mtu; + sctp_enforce_cwnd_limit(&stcb->asoc, net); + net->cc_mod.rtcc.vol_reduce--; + } + net->cc_mod.rtcc.last_step_state = 2; + net->cc_mod.rtcc.step_cnt = 0; + } + goto out_decision; + } else if (net->rtt < net->cc_mod.rtcc.lbw_rtt-rtt_offset) { + /* bw & rtt decreased */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* Probe point 3 */ + probepoint |= ((3 << 16) | 0); + SDT_PROBE5(sctp, cwnd, net, rttvar, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + net->flight_size, + probepoint); +#endif + if (net->cc_mod.rtcc.steady_step) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + oth = net->cc_mod.rtcc.vol_reduce; + oth <<= 16; + oth |= net->cc_mod.rtcc.step_cnt; + oth <<= 16; + oth |= net->cc_mod.rtcc.last_step_state; + SDT_PROBE5(sctp, cwnd, net, rttstep, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + oth, + probepoint); +#endif + if ((net->cc_mod.rtcc.vol_reduce) && + (inst_ind != SCTP_INST_GAINING)) { + net->cwnd += net->mtu; + sctp_enforce_cwnd_limit(&stcb->asoc, net); + net->cc_mod.rtcc.vol_reduce--; + } + net->cc_mod.rtcc.last_step_state = 3; + net->cc_mod.rtcc.step_cnt = 0; + } + goto out_decision; + } + /* The bw decreased but rtt stayed the same */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* Probe point 4 */ + probepoint |= ((4 << 16) | 0); + SDT_PROBE5(sctp, cwnd, net, rttvar, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + net->flight_size, + probepoint); +#endif + if (net->cc_mod.rtcc.steady_step) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + oth = net->cc_mod.rtcc.vol_reduce; + oth <<= 16; + oth |= net->cc_mod.rtcc.step_cnt; + oth <<= 16; + oth |= net->cc_mod.rtcc.last_step_state; + SDT_PROBE5(sctp, cwnd, net, rttstep, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + oth, + probepoint); +#endif + if ((net->cc_mod.rtcc.vol_reduce) && + (inst_ind != SCTP_INST_GAINING)) { + net->cwnd += net->mtu; + sctp_enforce_cwnd_limit(&stcb->asoc, net); + net->cc_mod.rtcc.vol_reduce--; + } + net->cc_mod.rtcc.last_step_state = 4; + net->cc_mod.rtcc.step_cnt = 0; + } +out_decision: + net->cc_mod.rtcc.lbw = nbw; + net->cc_mod.rtcc.lbw_rtt = net->rtt; + net->cc_mod.rtcc.cwnd_at_bw_set = net->cwnd; + if (inst_ind == SCTP_INST_GAINING) { + return (1); + } else { + return (0); + } +} + +#if defined(__FreeBSD__) && !defined(__Userspace__) +static int +cc_bw_increase(struct sctp_tcb *stcb, struct sctp_nets *net, uint64_t nbw, uint64_t vtag) +#else +static int +cc_bw_increase(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_nets *net, uint64_t nbw) +#endif +{ +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint64_t oth, probepoint; + +#endif + /* BW increased, so update and + * return 0, since all actions in + * our table say to do the normal CC + * update. Note that we pay no attention to + * the inst_ind since our overall sum is increasing. + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* PROBE POINT 0 */ + probepoint = (((uint64_t)net->cwnd) << 32); + SDT_PROBE5(sctp, cwnd, net, rttvar, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + net->flight_size, + probepoint); +#endif + if (net->cc_mod.rtcc.steady_step) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + oth = net->cc_mod.rtcc.vol_reduce; + oth <<= 16; + oth |= net->cc_mod.rtcc.step_cnt; + oth <<= 16; + oth |= net->cc_mod.rtcc.last_step_state; + SDT_PROBE5(sctp, cwnd, net, rttstep, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | nbw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + oth, + probepoint); +#endif + net->cc_mod.rtcc.last_step_state = 0; + net->cc_mod.rtcc.step_cnt = 0; + net->cc_mod.rtcc.vol_reduce = 0; + } + net->cc_mod.rtcc.lbw = nbw; + net->cc_mod.rtcc.lbw_rtt = net->rtt; + net->cc_mod.rtcc.cwnd_at_bw_set = net->cwnd; + return (0); +} + +/* RTCC Algorithm to limit growth of cwnd, return + * true if you want to NOT allow cwnd growth + */ +static int +cc_bw_limit(struct sctp_tcb *stcb, struct sctp_nets *net, uint64_t nbw) +{ + uint64_t bw_offset, rtt_offset; +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint64_t probepoint, rtt, vtag; +#endif + uint64_t bytes_for_this_rtt, inst_bw; + uint64_t div, inst_off; + int bw_shift; + uint8_t inst_ind; + int ret; + /*- + * Here we need to see if we want + * to limit cwnd growth due to increase + * in overall rtt but no increase in bw. + * We use the following table to figure + * out what we should do. When we return + * 0, cc update goes on as planned. If we + * return 1, then no cc update happens and cwnd + * stays where it is at. + * ---------------------------------- + * BW | RTT | Action + * ********************************* + * INC | INC | return 0 + * ---------------------------------- + * INC | SAME | return 0 + * ---------------------------------- + * INC | DECR | return 0 + * ---------------------------------- + * SAME | INC | return 1 + * ---------------------------------- + * SAME | SAME | return 1 + * ---------------------------------- + * SAME | DECR | return 0 + * ---------------------------------- + * DECR | INC | return 0 or 1 based on if we caused. + * ---------------------------------- + * DECR | SAME | return 0 + * ---------------------------------- + * DECR | DECR | return 0 + * ---------------------------------- + * + * We are a bit fuzz on what an increase or + * decrease is. For BW it is the same if + * it did not change within 1/64th. For + * RTT it stayed the same if it did not + * change within 1/32nd + */ + bw_shift = SCTP_BASE_SYSCTL(sctp_rttvar_bw); +#if defined(__FreeBSD__) && !defined(__Userspace__) + rtt = stcb->asoc.my_vtag; + vtag = (rtt << 32) | (((uint32_t)(stcb->sctp_ep->sctp_lport)) << 16) | (stcb->rport); + probepoint = (((uint64_t)net->cwnd) << 32); + rtt = net->rtt; +#endif + if (net->cc_mod.rtcc.rtt_set_this_sack) { + net->cc_mod.rtcc.rtt_set_this_sack = 0; + bytes_for_this_rtt = net->cc_mod.rtcc.bw_bytes - net->cc_mod.rtcc.bw_bytes_at_last_rttc; + net->cc_mod.rtcc.bw_bytes_at_last_rttc = net->cc_mod.rtcc.bw_bytes; + if (net->rtt) { + div = net->rtt / 1000; + if (div) { + inst_bw = bytes_for_this_rtt / div; + inst_off = inst_bw >> bw_shift; + if (inst_bw > nbw) + inst_ind = SCTP_INST_GAINING; + else if ((inst_bw+inst_off) < nbw) + inst_ind = SCTP_INST_LOOSING; + else + inst_ind = SCTP_INST_NEUTRAL; +#if defined(__FreeBSD__) && !defined(__Userspace__) + probepoint |= ((0xb << 16) | inst_ind); +#endif + } else { + inst_ind = net->cc_mod.rtcc.last_inst_ind; +#if defined(__FreeBSD__) && !defined(__Userspace__) + inst_bw = bytes_for_this_rtt / (uint64_t)(net->rtt); + /* Can't determine do not change */ + probepoint |= ((0xc << 16) | inst_ind); +#endif + } + } else { + inst_ind = net->cc_mod.rtcc.last_inst_ind; +#if defined(__FreeBSD__) && !defined(__Userspace__) + inst_bw = bytes_for_this_rtt; + /* Can't determine do not change */ + probepoint |= ((0xd << 16) | inst_ind); +#endif + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + SDT_PROBE5(sctp, cwnd, net, rttvar, + vtag, + ((nbw << 32) | inst_bw), + ((net->cc_mod.rtcc.lbw_rtt << 32) | rtt), + net->flight_size, + probepoint); +#endif + } else { + /* No rtt measurement, use last one */ + inst_ind = net->cc_mod.rtcc.last_inst_ind; + } + bw_offset = net->cc_mod.rtcc.lbw >> bw_shift; + if (nbw > net->cc_mod.rtcc.lbw + bw_offset) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + ret = cc_bw_increase(stcb, net, nbw, vtag); +#else + ret = cc_bw_increase(stcb, net, nbw); +#endif + goto out; + } + rtt_offset = net->cc_mod.rtcc.lbw_rtt >> SCTP_BASE_SYSCTL(sctp_rttvar_rtt); + if (nbw < net->cc_mod.rtcc.lbw - bw_offset) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + ret = cc_bw_decrease(stcb, net, nbw, rtt_offset, vtag, inst_ind); +#else + ret = cc_bw_decrease(stcb, net, nbw, rtt_offset, inst_ind); +#endif + goto out; + } + /* If we reach here then + * we are in a situation where + * the bw stayed the same. + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + ret = cc_bw_same(stcb, net, nbw, rtt_offset, vtag, inst_ind); +#else + ret = cc_bw_same(stcb, net, nbw, rtt_offset, inst_ind); +#endif +out: + net->cc_mod.rtcc.last_inst_ind = inst_ind; + return (ret); +} + +static void +sctp_cwnd_update_after_sack_common(struct sctp_tcb *stcb, + struct sctp_association *asoc, + int accum_moved, int reneged_all SCTP_UNUSED, int will_exit, int use_rtcc) +{ + struct sctp_nets *net; +#if defined(__FreeBSD__) && !defined(__Userspace__) + int old_cwnd; +#endif + uint32_t t_ssthresh, incr; + uint64_t t_ucwnd_sbw; + uint64_t t_path_mptcp; + uint64_t mptcp_like_alpha; + uint32_t srtt; + uint64_t max_path; + + /* MT FIXME: Don't compute this over and over again */ + t_ssthresh = 0; + t_ucwnd_sbw = 0; + t_path_mptcp = 0; + mptcp_like_alpha = 1; + if ((stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV1) || + (stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV2) || + (stcb->asoc.sctp_cmt_on_off == SCTP_CMT_MPTCP)) { + max_path = 0; + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + t_ssthresh += net->ssthresh; + /* lastsa>>3; we don't need to devide ...*/ + srtt = net->lastsa; + if (srtt > 0) { + uint64_t tmp; + + t_ucwnd_sbw += (uint64_t)net->cwnd / (uint64_t)srtt; + t_path_mptcp += (((uint64_t)net->cwnd) << SHIFT_MPTCP_MULTI_Z) / + (((uint64_t)net->mtu) * (uint64_t)srtt); + tmp = (((uint64_t)net->cwnd) << SHIFT_MPTCP_MULTI_N) / + ((uint64_t)net->mtu * (uint64_t)(srtt * srtt)); + if (tmp > max_path) { + max_path = tmp; + } + } + } + if (t_path_mptcp > 0) { + mptcp_like_alpha = max_path / (t_path_mptcp * t_path_mptcp); + } else { + mptcp_like_alpha = 1; + } + } + if (t_ssthresh == 0) { + t_ssthresh = 1; + } + if (t_ucwnd_sbw == 0) { + t_ucwnd_sbw = 1; + } + /******************************/ + /* update cwnd and Early FR */ + /******************************/ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { +#ifdef JANA_CMT_FAST_RECOVERY + /* + * CMT fast recovery code. Need to debug. + */ + if (net->fast_retran_loss_recovery && net->new_pseudo_cumack) { + if (SCTP_TSN_GE(asoc->last_acked_seq, net->fast_recovery_tsn) || + SCTP_TSN_GE(net->pseudo_cumack,net->fast_recovery_tsn)) { + net->will_exit_fast_recovery = 1; + } + } +#endif + /* if nothing was acked on this destination skip it */ + if (net->net_ack == 0) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FROM_SACK); + } + continue; + } +#ifdef JANA_CMT_FAST_RECOVERY + /* CMT fast recovery code + */ + /* + if (sctp_cmt_on_off > 0 && net->fast_retran_loss_recovery && net->will_exit_fast_recovery == 0) { + @@@ Do something + } + else if (sctp_cmt_on_off == 0 && asoc->fast_retran_loss_recovery && will_exit == 0) { + */ +#endif + + if (asoc->fast_retran_loss_recovery && + (will_exit == 0) && + (asoc->sctp_cmt_on_off == 0)) { + /* + * If we are in loss recovery we skip any cwnd + * update + */ + return; + } + /* + * Did any measurements go on for this network? + */ + if (use_rtcc && (net->cc_mod.rtcc.tls_needs_set > 0)) { + uint64_t nbw; + /* + * At this point our bw_bytes has been updated + * by incoming sack information. + * + * But our bw may not yet be set. + * + */ + if ((net->cc_mod.rtcc.new_tot_time/1000) > 0) { + nbw = net->cc_mod.rtcc.bw_bytes/(net->cc_mod.rtcc.new_tot_time/1000); + } else { + nbw = net->cc_mod.rtcc.bw_bytes; + } + if (net->cc_mod.rtcc.lbw) { + if (cc_bw_limit(stcb, net, nbw)) { + /* Hold here, no update */ + continue; + } + } else { +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint64_t vtag, probepoint; + + probepoint = (((uint64_t)net->cwnd) << 32); + probepoint |= ((0xa << 16) | 0); + vtag = (net->rtt << 32) | + (((uint32_t)(stcb->sctp_ep->sctp_lport)) << 16) | + (stcb->rport); + + SDT_PROBE5(sctp, cwnd, net, rttvar, + vtag, + nbw, + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + net->flight_size, + probepoint); +#endif + net->cc_mod.rtcc.lbw = nbw; + net->cc_mod.rtcc.lbw_rtt = net->rtt; + if (net->cc_mod.rtcc.rtt_set_this_sack) { + net->cc_mod.rtcc.rtt_set_this_sack = 0; + net->cc_mod.rtcc.bw_bytes_at_last_rttc = net->cc_mod.rtcc.bw_bytes; + } + } + } + /* + * CMT: CUC algorithm. Update cwnd if pseudo-cumack has + * moved. + */ + if (accum_moved || + ((asoc->sctp_cmt_on_off > 0) && net->new_pseudo_cumack)) { + /* If the cumulative ack moved we can proceed */ + if (net->cwnd <= net->ssthresh) { + /* We are in slow start */ + if (net->flight_size + net->net_ack >= net->cwnd) { + uint32_t limit; + +#if defined(__FreeBSD__) && !defined(__Userspace__) + old_cwnd = net->cwnd; +#endif + switch (asoc->sctp_cmt_on_off) { + case SCTP_CMT_RPV1: + limit = (uint32_t)(((uint64_t)net->mtu * + (uint64_t)SCTP_BASE_SYSCTL(sctp_L2_abc_variable) * + (uint64_t)net->ssthresh) / + (uint64_t)t_ssthresh); + incr = (uint32_t)(((uint64_t)net->net_ack * + (uint64_t)net->ssthresh) / + (uint64_t)t_ssthresh); + if (incr > limit) { + incr = limit; + } + if (incr == 0) { + incr = 1; + } + break; + case SCTP_CMT_RPV2: + /* lastsa>>3; we don't need to divide ...*/ + srtt = net->lastsa; + if (srtt == 0) { + srtt = 1; + } + limit = (uint32_t)(((uint64_t)net->mtu * + (uint64_t)SCTP_BASE_SYSCTL(sctp_L2_abc_variable) * + (uint64_t)net->cwnd) / + ((uint64_t)srtt * t_ucwnd_sbw)); + /* INCREASE FACTOR */ + incr = (uint32_t)(((uint64_t)net->net_ack * + (uint64_t)net->cwnd) / + ((uint64_t)srtt * t_ucwnd_sbw)); + /* INCREASE FACTOR */ + if (incr > limit) { + incr = limit; + } + if (incr == 0) { + incr = 1; + } + break; + case SCTP_CMT_MPTCP: + limit = (uint32_t)(((uint64_t)net->mtu * + mptcp_like_alpha * + (uint64_t)SCTP_BASE_SYSCTL(sctp_L2_abc_variable)) >> + SHIFT_MPTCP_MULTI); + incr = (uint32_t)(((uint64_t)net->net_ack * + mptcp_like_alpha) >> + SHIFT_MPTCP_MULTI); + if (incr > limit) { + incr = limit; + } + if (incr > net->net_ack) { + incr = net->net_ack; + } + if (incr > net->mtu) { + incr = net->mtu; + } + break; + default: + incr = net->net_ack; + if (incr > net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable)) { + incr = net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable); + } + break; + } + net->cwnd += incr; + sctp_enforce_cwnd_limit(asoc, net); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, incr, + SCTP_CWND_LOG_FROM_SS); + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + SDT_PROBE5(sctp, cwnd, net, ack, + stcb->asoc.my_vtag, + ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), + net, + old_cwnd, net->cwnd); +#endif + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, net->net_ack, + SCTP_CWND_LOG_NOADV_SS); + } + } + } else { + /* We are in congestion avoidance */ + /* + * Add to pba + */ + net->partial_bytes_acked += net->net_ack; + + if ((net->flight_size + net->net_ack >= net->cwnd) && + (net->partial_bytes_acked >= net->cwnd)) { + net->partial_bytes_acked -= net->cwnd; +#if defined(__FreeBSD__) && !defined(__Userspace__) + old_cwnd = net->cwnd; +#endif + switch (asoc->sctp_cmt_on_off) { + case SCTP_CMT_RPV1: + incr = (uint32_t)(((uint64_t)net->mtu * + (uint64_t)net->ssthresh) / + (uint64_t)t_ssthresh); + if (incr == 0) { + incr = 1; + } + break; + case SCTP_CMT_RPV2: + /* lastsa>>3; we don't need to divide ... */ + srtt = net->lastsa; + if (srtt == 0) { + srtt = 1; + } + incr = (uint32_t)((uint64_t)net->mtu * + (uint64_t)net->cwnd / + ((uint64_t)srtt * + t_ucwnd_sbw)); + /* INCREASE FACTOR */ + if (incr == 0) { + incr = 1; + } + break; + case SCTP_CMT_MPTCP: + incr = (uint32_t)((mptcp_like_alpha * + (uint64_t) net->cwnd) >> + SHIFT_MPTCP_MULTI); + if (incr > net->mtu) { + incr = net->mtu; + } + break; + default: + incr = net->mtu; + break; + } + net->cwnd += incr; + sctp_enforce_cwnd_limit(asoc, net); +#if defined(__FreeBSD__) && !defined(__Userspace__) + SDT_PROBE5(sctp, cwnd, net, ack, + stcb->asoc.my_vtag, + ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), + net, + old_cwnd, net->cwnd); +#endif + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, net->mtu, + SCTP_CWND_LOG_FROM_CA); + } + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, net->net_ack, + SCTP_CWND_LOG_NOADV_CA); + } + } + } + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, net->mtu, + SCTP_CWND_LOG_NO_CUMACK); + } + } + } +} + +#if defined(__FreeBSD__) && !defined(__Userspace__) +static void +sctp_cwnd_update_exit_pf_common(struct sctp_tcb *stcb, struct sctp_nets *net) +#else +static void +sctp_cwnd_update_exit_pf_common(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_nets *net) +#endif +{ +#if defined(__FreeBSD__) && !defined(__Userspace__) + int old_cwnd; + + old_cwnd = net->cwnd; +#endif + net->cwnd = net->mtu; +#if defined(__FreeBSD__) && !defined(__Userspace__) + SDT_PROBE5(sctp, cwnd, net, ack, + stcb->asoc.my_vtag, ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), net, + old_cwnd, net->cwnd); +#endif + SCTPDBG(SCTP_DEBUG_INDATA1, "Destination %p moved from PF to reachable with cwnd %d.\n", + (void *)net, net->cwnd); +} + +static void +sctp_cwnd_update_after_timeout(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + int old_cwnd = net->cwnd; + uint32_t t_ssthresh, t_cwnd; + uint64_t t_ucwnd_sbw; + + /* MT FIXME: Don't compute this over and over again */ + t_ssthresh = 0; + t_cwnd = 0; + if ((stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV1) || + (stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV2)) { + struct sctp_nets *lnet; + uint32_t srtt; + + t_ucwnd_sbw = 0; + TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { + t_ssthresh += lnet->ssthresh; + t_cwnd += lnet->cwnd; + srtt = lnet->lastsa; + /* lastsa>>3; we don't need to divide ... */ + if (srtt > 0) { + t_ucwnd_sbw += (uint64_t)lnet->cwnd / (uint64_t)srtt; + } + } + if (t_ssthresh < 1) { + t_ssthresh = 1; + } + if (t_ucwnd_sbw < 1) { + t_ucwnd_sbw = 1; + } + if (stcb->asoc.sctp_cmt_on_off == SCTP_CMT_RPV1) { + net->ssthresh = (uint32_t)(((uint64_t)4 * + (uint64_t)net->mtu * + (uint64_t)net->ssthresh) / + (uint64_t)t_ssthresh); + } else { + uint64_t cc_delta; + + srtt = net->lastsa; + /* lastsa>>3; we don't need to divide ... */ + if (srtt == 0) { + srtt = 1; + } + cc_delta = t_ucwnd_sbw * (uint64_t)srtt / 2; + if (cc_delta < t_cwnd) { + net->ssthresh = (uint32_t)((uint64_t)t_cwnd - cc_delta); + } else { + net->ssthresh = net->mtu; + } + } + if ((net->cwnd > t_cwnd / 2) && + (net->ssthresh < net->cwnd - t_cwnd / 2)) { + net->ssthresh = net->cwnd - t_cwnd / 2; + } + if (net->ssthresh < net->mtu) { + net->ssthresh = net->mtu; + } + } else { + net->ssthresh = max(net->cwnd / 2, 4 * net->mtu); + } + net->cwnd = net->mtu; + net->partial_bytes_acked = 0; +#if defined(__FreeBSD__) && !defined(__Userspace__) + SDT_PROBE5(sctp, cwnd, net, to, + stcb->asoc.my_vtag, + ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), + net, + old_cwnd, net->cwnd); +#endif + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, net->cwnd - old_cwnd, SCTP_CWND_LOG_FROM_RTX); + } +} + +static void +sctp_cwnd_update_after_ecn_echo_common(struct sctp_tcb *stcb, struct sctp_nets *net, + int in_window, int num_pkt_lost, int use_rtcc) +{ + int old_cwnd = net->cwnd; + if ((use_rtcc) && (net->lan_type == SCTP_LAN_LOCAL) && (net->cc_mod.rtcc.use_dccc_ecn)) { + /* Data center Congestion Control */ + if (in_window == 0) { + /* Go to CA with the cwnd at the point we sent + * the TSN that was marked with a CE. + */ + if (net->ecn_prev_cwnd < net->cwnd) { + /* Restore to prev cwnd */ + net->cwnd = net->ecn_prev_cwnd - (net->mtu * num_pkt_lost); + } else { + /* Just cut in 1/2 */ + net->cwnd /= 2; + } + /* Drop to CA */ + net->ssthresh = net->cwnd - (num_pkt_lost * net->mtu); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT); + } + } else { + /* Further tuning down required over the drastic original cut */ + net->ssthresh -= (net->mtu * num_pkt_lost); + net->cwnd -= (net->mtu * num_pkt_lost); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT); + } + } + SCTP_STAT_INCR(sctps_ecnereducedcwnd); + } else { + if (in_window == 0) { + SCTP_STAT_INCR(sctps_ecnereducedcwnd); + net->ssthresh = net->cwnd / 2; + if (net->ssthresh < net->mtu) { + net->ssthresh = net->mtu; + /* here back off the timer as well, to slow us down */ + net->RTO <<= 1; + } + net->cwnd = net->ssthresh; +#if defined(__FreeBSD__) && !defined(__Userspace__) + SDT_PROBE5(sctp, cwnd, net, ecn, + stcb->asoc.my_vtag, + ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), + net, + old_cwnd, net->cwnd); +#endif + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT); + } + } + } + +} + +static void +sctp_cwnd_update_after_packet_dropped(struct sctp_tcb *stcb, + struct sctp_nets *net, struct sctp_pktdrop_chunk *cp, + uint32_t *bottle_bw, uint32_t *on_queue) +{ + uint32_t bw_avail; + unsigned int incr; + int old_cwnd = net->cwnd; + + /* get bottle neck bw */ + *bottle_bw = ntohl(cp->bottle_bw); + /* and whats on queue */ + *on_queue = ntohl(cp->current_onq); + /* + * adjust the on-queue if our flight is more it could be + * that the router has not yet gotten data "in-flight" to it + */ + if (*on_queue < net->flight_size) { + *on_queue = net->flight_size; + } + /* rtt is measured in micro seconds, bottle_bw in bytes per second */ + bw_avail = (uint32_t)(((uint64_t)(*bottle_bw) * net->rtt) / (uint64_t)1000000); + if (bw_avail > *bottle_bw) { + /* + * Cap the growth to no more than the bottle neck. + * This can happen as RTT slides up due to queues. + * It also means if you have more than a 1 second + * RTT with a empty queue you will be limited to the + * bottle_bw per second no matter if other points + * have 1/2 the RTT and you could get more out... + */ + bw_avail = *bottle_bw; + } + if (*on_queue > bw_avail) { + /* + * No room for anything else don't allow anything + * else to be "added to the fire". + */ + int seg_inflight, seg_onqueue, my_portion; + + net->partial_bytes_acked = 0; + /* how much are we over queue size? */ + incr = *on_queue - bw_avail; + if (stcb->asoc.seen_a_sack_this_pkt) { + /* + * undo any cwnd adjustment that the sack + * might have made + */ + net->cwnd = net->prev_cwnd; + } + /* Now how much of that is mine? */ + seg_inflight = net->flight_size / net->mtu; + seg_onqueue = *on_queue / net->mtu; + my_portion = (incr * seg_inflight) / seg_onqueue; + + /* Have I made an adjustment already */ + if (net->cwnd > net->flight_size) { + /* + * for this flight I made an adjustment we + * need to decrease the portion by a share + * our previous adjustment. + */ + int diff_adj; + + diff_adj = net->cwnd - net->flight_size; + if (diff_adj > my_portion) + my_portion = 0; + else + my_portion -= diff_adj; + } + /* + * back down to the previous cwnd (assume we have + * had a sack before this packet). minus what ever + * portion of the overage is my fault. + */ + net->cwnd -= my_portion; + + /* we will NOT back down more than 1 MTU */ + if (net->cwnd <= net->mtu) { + net->cwnd = net->mtu; + } + /* force into CA */ + net->ssthresh = net->cwnd - 1; + } else { + /* + * Take 1/4 of the space left or max burst up .. + * whichever is less. + */ + incr = (bw_avail - *on_queue) >> 2; + if ((stcb->asoc.max_burst > 0) && + (stcb->asoc.max_burst * net->mtu < incr)) { + incr = stcb->asoc.max_burst * net->mtu; + } + net->cwnd += incr; + } + if (net->cwnd > bw_avail) { + /* We can't exceed the pipe size */ + net->cwnd = bw_avail; + } + if (net->cwnd < net->mtu) { + /* We always have 1 MTU */ + net->cwnd = net->mtu; + } + sctp_enforce_cwnd_limit(&stcb->asoc, net); + if (net->cwnd - old_cwnd != 0) { + /* log only changes */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + SDT_PROBE5(sctp, cwnd, net, pd, + stcb->asoc.my_vtag, + ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), + net, + old_cwnd, net->cwnd); +#endif + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), + SCTP_CWND_LOG_FROM_SAT); + } + } +} + +static void +sctp_cwnd_update_after_output(struct sctp_tcb *stcb, + struct sctp_nets *net, int burst_limit) +{ + int old_cwnd = net->cwnd; + + if (net->ssthresh < net->cwnd) + net->ssthresh = net->cwnd; + if (burst_limit) { + net->cwnd = (net->flight_size + (burst_limit * net->mtu)); + sctp_enforce_cwnd_limit(&stcb->asoc, net); +#if defined(__FreeBSD__) && !defined(__Userspace__) + SDT_PROBE5(sctp, cwnd, net, bl, + stcb->asoc.my_vtag, + ((stcb->sctp_ep->sctp_lport << 16) | (stcb->rport)), + net, + old_cwnd, net->cwnd); +#endif + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_BRST); + } + } +} + +static void +sctp_cwnd_update_after_sack(struct sctp_tcb *stcb, + struct sctp_association *asoc, + int accum_moved, int reneged_all, int will_exit) +{ + /* Passing a zero argument in last disables the rtcc algorithm */ + sctp_cwnd_update_after_sack_common(stcb, asoc, accum_moved, reneged_all, will_exit, 0); +} + +static void +sctp_cwnd_update_after_ecn_echo(struct sctp_tcb *stcb, struct sctp_nets *net, + int in_window, int num_pkt_lost) +{ + /* Passing a zero argument in last disables the rtcc algorithm */ + sctp_cwnd_update_after_ecn_echo_common(stcb, net, in_window, num_pkt_lost, 0); +} + +/* Here starts the RTCCVAR type CC invented by RRS which + * is a slight mod to RFC2581. We reuse a common routine or + * two since these algorithms are so close and need to + * remain the same. + */ +static void +sctp_cwnd_update_rtcc_after_ecn_echo(struct sctp_tcb *stcb, struct sctp_nets *net, + int in_window, int num_pkt_lost) +{ + sctp_cwnd_update_after_ecn_echo_common(stcb, net, in_window, num_pkt_lost, 1); +} + +static void sctp_cwnd_update_rtcc_tsn_acknowledged(struct sctp_nets *net, + struct sctp_tmit_chunk *tp1) +{ + net->cc_mod.rtcc.bw_bytes += tp1->send_size; +} + +static void +sctp_cwnd_prepare_rtcc_net_for_sack(struct sctp_tcb *stcb SCTP_UNUSED, + struct sctp_nets *net) +{ + if (net->cc_mod.rtcc.tls_needs_set > 0) { + /* We had a bw measurement going on */ + struct timeval ltls; + SCTP_GETPTIME_TIMEVAL(<ls); + timevalsub(<ls, &net->cc_mod.rtcc.tls); + net->cc_mod.rtcc.new_tot_time = (ltls.tv_sec * 1000000) + ltls.tv_usec; + } +} + +static void +sctp_cwnd_new_rtcc_transmission_begins(struct sctp_tcb *stcb, + struct sctp_nets *net) +{ +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint64_t vtag, probepoint; + +#endif + if (net->cc_mod.rtcc.lbw) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* Clear the old bw.. we went to 0 in-flight */ + vtag = (net->rtt << 32) | (((uint32_t)(stcb->sctp_ep->sctp_lport)) << 16) | + (stcb->rport); + probepoint = (((uint64_t)net->cwnd) << 32); + /* Probe point 8 */ + probepoint |= ((8 << 16) | 0); + SDT_PROBE5(sctp, cwnd, net, rttvar, + vtag, + ((net->cc_mod.rtcc.lbw << 32) | 0), + ((net->cc_mod.rtcc.lbw_rtt << 32) | net->rtt), + net->flight_size, + probepoint); +#endif + net->cc_mod.rtcc.lbw_rtt = 0; + net->cc_mod.rtcc.cwnd_at_bw_set = 0; + net->cc_mod.rtcc.lbw = 0; + net->cc_mod.rtcc.bw_bytes_at_last_rttc = 0; + net->cc_mod.rtcc.vol_reduce = 0; + net->cc_mod.rtcc.bw_tot_time = 0; + net->cc_mod.rtcc.bw_bytes = 0; + net->cc_mod.rtcc.tls_needs_set = 0; + if (net->cc_mod.rtcc.steady_step) { + net->cc_mod.rtcc.vol_reduce = 0; + net->cc_mod.rtcc.step_cnt = 0; + net->cc_mod.rtcc.last_step_state = 0; + } + if (net->cc_mod.rtcc.ret_from_eq) { + /* less aggressive one - reset cwnd too */ + uint32_t cwnd_in_mtu, cwnd; + + cwnd_in_mtu = SCTP_BASE_SYSCTL(sctp_initial_cwnd); + if (cwnd_in_mtu == 0) { + /* Using 0 means that the value of RFC 4960 is used. */ + cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND)); + } else { + /* + * We take the minimum of the burst limit and the + * initial congestion window. + */ + if ((stcb->asoc.max_burst > 0) && (cwnd_in_mtu > stcb->asoc.max_burst)) + cwnd_in_mtu = stcb->asoc.max_burst; + cwnd = (net->mtu - sizeof(struct sctphdr)) * cwnd_in_mtu; + } + if (net->cwnd > cwnd) { + /* Only set if we are not a timeout (i.e. down to 1 mtu) */ + net->cwnd = cwnd; + } + } + } +} + +static void +sctp_set_rtcc_initial_cc_param(struct sctp_tcb *stcb, + struct sctp_nets *net) +{ +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint64_t vtag, probepoint; + +#endif + sctp_set_initial_cc_param(stcb, net); + stcb->asoc.use_precise_time = 1; +#if defined(__FreeBSD__) && !defined(__Userspace__) + probepoint = (((uint64_t)net->cwnd) << 32); + probepoint |= ((9 << 16) | 0); + vtag = (net->rtt << 32) | + (((uint32_t)(stcb->sctp_ep->sctp_lport)) << 16) | + (stcb->rport); + SDT_PROBE5(sctp, cwnd, net, rttvar, + vtag, + 0, + 0, + 0, + probepoint); +#endif + net->cc_mod.rtcc.lbw_rtt = 0; + net->cc_mod.rtcc.cwnd_at_bw_set = 0; + net->cc_mod.rtcc.vol_reduce = 0; + net->cc_mod.rtcc.lbw = 0; + net->cc_mod.rtcc.vol_reduce = 0; + net->cc_mod.rtcc.bw_bytes_at_last_rttc = 0; + net->cc_mod.rtcc.bw_tot_time = 0; + net->cc_mod.rtcc.bw_bytes = 0; + net->cc_mod.rtcc.tls_needs_set = 0; + net->cc_mod.rtcc.ret_from_eq = SCTP_BASE_SYSCTL(sctp_rttvar_eqret); + net->cc_mod.rtcc.steady_step = SCTP_BASE_SYSCTL(sctp_steady_step); + net->cc_mod.rtcc.use_dccc_ecn = SCTP_BASE_SYSCTL(sctp_use_dccc_ecn); + net->cc_mod.rtcc.step_cnt = 0; + net->cc_mod.rtcc.last_step_state = 0; +} + +static int +sctp_cwnd_rtcc_socket_option(struct sctp_tcb *stcb, int setorget, + struct sctp_cc_option *cc_opt) +{ + struct sctp_nets *net; + + if (setorget == 1) { + /* a set */ + if (cc_opt->option == SCTP_CC_OPT_RTCC_SETMODE) { + if ((cc_opt->aid_value.assoc_value != 0) && + (cc_opt->aid_value.assoc_value != 1)) { + return (EINVAL); + } + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + net->cc_mod.rtcc.ret_from_eq = cc_opt->aid_value.assoc_value; + } + } else if (cc_opt->option == SCTP_CC_OPT_USE_DCCC_ECN) { + if ((cc_opt->aid_value.assoc_value != 0) && + (cc_opt->aid_value.assoc_value != 1)) { + return (EINVAL); + } + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + net->cc_mod.rtcc.use_dccc_ecn = cc_opt->aid_value.assoc_value; + } + } else if (cc_opt->option == SCTP_CC_OPT_STEADY_STEP) { + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + net->cc_mod.rtcc.steady_step = cc_opt->aid_value.assoc_value; + } + } else { + return (EINVAL); + } + } else { + /* a get */ + if (cc_opt->option == SCTP_CC_OPT_RTCC_SETMODE) { + net = TAILQ_FIRST(&stcb->asoc.nets); + if (net == NULL) { + return (EFAULT); + } + cc_opt->aid_value.assoc_value = net->cc_mod.rtcc.ret_from_eq; + } else if (cc_opt->option == SCTP_CC_OPT_USE_DCCC_ECN) { + net = TAILQ_FIRST(&stcb->asoc.nets); + if (net == NULL) { + return (EFAULT); + } + cc_opt->aid_value.assoc_value = net->cc_mod.rtcc.use_dccc_ecn; + } else if (cc_opt->option == SCTP_CC_OPT_STEADY_STEP) { + net = TAILQ_FIRST(&stcb->asoc.nets); + if (net == NULL) { + return (EFAULT); + } + cc_opt->aid_value.assoc_value = net->cc_mod.rtcc.steady_step; + } else { + return (EINVAL); + } + } + return (0); +} + +static void +sctp_cwnd_update_rtcc_packet_transmitted(struct sctp_tcb *stcb SCTP_UNUSED, + struct sctp_nets *net) +{ + if (net->cc_mod.rtcc.tls_needs_set == 0) { + SCTP_GETPTIME_TIMEVAL(&net->cc_mod.rtcc.tls); + net->cc_mod.rtcc.tls_needs_set = 2; + } +} + +static void +sctp_cwnd_update_rtcc_after_sack(struct sctp_tcb *stcb, + struct sctp_association *asoc, + int accum_moved, int reneged_all, int will_exit) +{ + /* Passing a one argument at the last enables the rtcc algorithm */ + sctp_cwnd_update_after_sack_common(stcb, asoc, accum_moved, reneged_all, will_exit, 1); +} + +static void +sctp_rtt_rtcc_calculated(struct sctp_tcb *stcb SCTP_UNUSED, + struct sctp_nets *net, + struct timeval *now SCTP_UNUSED) +{ + net->cc_mod.rtcc.rtt_set_this_sack = 1; +} + +/* Here starts Sally Floyds HS-TCP */ + +struct sctp_hs_raise_drop { + int32_t cwnd; + int8_t increase; + int8_t drop_percent; +}; + +#define SCTP_HS_TABLE_SIZE 73 + +static const struct sctp_hs_raise_drop sctp_cwnd_adjust[SCTP_HS_TABLE_SIZE] = { + {38, 1, 50}, /* 0 */ + {118, 2, 44}, /* 1 */ + {221, 3, 41}, /* 2 */ + {347, 4, 38}, /* 3 */ + {495, 5, 37}, /* 4 */ + {663, 6, 35}, /* 5 */ + {851, 7, 34}, /* 6 */ + {1058, 8, 33}, /* 7 */ + {1284, 9, 32}, /* 8 */ + {1529, 10, 31}, /* 9 */ + {1793, 11, 30}, /* 10 */ + {2076, 12, 29}, /* 11 */ + {2378, 13, 28}, /* 12 */ + {2699, 14, 28}, /* 13 */ + {3039, 15, 27}, /* 14 */ + {3399, 16, 27}, /* 15 */ + {3778, 17, 26}, /* 16 */ + {4177, 18, 26}, /* 17 */ + {4596, 19, 25}, /* 18 */ + {5036, 20, 25}, /* 19 */ + {5497, 21, 24}, /* 20 */ + {5979, 22, 24}, /* 21 */ + {6483, 23, 23}, /* 22 */ + {7009, 24, 23}, /* 23 */ + {7558, 25, 22}, /* 24 */ + {8130, 26, 22}, /* 25 */ + {8726, 27, 22}, /* 26 */ + {9346, 28, 21}, /* 27 */ + {9991, 29, 21}, /* 28 */ + {10661, 30, 21}, /* 29 */ + {11358, 31, 20}, /* 30 */ + {12082, 32, 20}, /* 31 */ + {12834, 33, 20}, /* 32 */ + {13614, 34, 19}, /* 33 */ + {14424, 35, 19}, /* 34 */ + {15265, 36, 19}, /* 35 */ + {16137, 37, 19}, /* 36 */ + {17042, 38, 18}, /* 37 */ + {17981, 39, 18}, /* 38 */ + {18955, 40, 18}, /* 39 */ + {19965, 41, 17}, /* 40 */ + {21013, 42, 17}, /* 41 */ + {22101, 43, 17}, /* 42 */ + {23230, 44, 17}, /* 43 */ + {24402, 45, 16}, /* 44 */ + {25618, 46, 16}, /* 45 */ + {26881, 47, 16}, /* 46 */ + {28193, 48, 16}, /* 47 */ + {29557, 49, 15}, /* 48 */ + {30975, 50, 15}, /* 49 */ + {32450, 51, 15}, /* 50 */ + {33986, 52, 15}, /* 51 */ + {35586, 53, 14}, /* 52 */ + {37253, 54, 14}, /* 53 */ + {38992, 55, 14}, /* 54 */ + {40808, 56, 14}, /* 55 */ + {42707, 57, 13}, /* 56 */ + {44694, 58, 13}, /* 57 */ + {46776, 59, 13}, /* 58 */ + {48961, 60, 13}, /* 59 */ + {51258, 61, 13}, /* 60 */ + {53677, 62, 12}, /* 61 */ + {56230, 63, 12}, /* 62 */ + {58932, 64, 12}, /* 63 */ + {61799, 65, 12}, /* 64 */ + {64851, 66, 11}, /* 65 */ + {68113, 67, 11}, /* 66 */ + {71617, 68, 11}, /* 67 */ + {75401, 69, 10}, /* 68 */ + {79517, 70, 10}, /* 69 */ + {84035, 71, 10}, /* 70 */ + {89053, 72, 10}, /* 71 */ + {94717, 73, 9} /* 72 */ +}; + +static void +sctp_hs_cwnd_increase(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + int cur_val, i, indx, incr; + int old_cwnd = net->cwnd; + + cur_val = net->cwnd >> 10; + indx = SCTP_HS_TABLE_SIZE - 1; + + if (cur_val < sctp_cwnd_adjust[0].cwnd) { + /* normal mode */ + if (net->net_ack > net->mtu) { + net->cwnd += net->mtu; + } else { + net->cwnd += net->net_ack; + } + } else { + for (i = net->last_hs_used; i < SCTP_HS_TABLE_SIZE; i++) { + if (cur_val < sctp_cwnd_adjust[i].cwnd) { + indx = i; + break; + } + } + net->last_hs_used = indx; + incr = (((int32_t)sctp_cwnd_adjust[indx].increase) << 10); + net->cwnd += incr; + } + sctp_enforce_cwnd_limit(&stcb->asoc, net); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SS); + } +} + +static void +sctp_hs_cwnd_decrease(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + int cur_val, i, indx; + int old_cwnd = net->cwnd; + + cur_val = net->cwnd >> 10; + if (cur_val < sctp_cwnd_adjust[0].cwnd) { + /* normal mode */ + net->ssthresh = net->cwnd / 2; + if (net->ssthresh < (net->mtu * 2)) { + net->ssthresh = 2 * net->mtu; + } + net->cwnd = net->ssthresh; + } else { + /* drop by the proper amount */ + net->ssthresh = net->cwnd - (int)((net->cwnd / 100) * + (int32_t)sctp_cwnd_adjust[net->last_hs_used].drop_percent); + net->cwnd = net->ssthresh; + /* now where are we */ + indx = net->last_hs_used; + cur_val = net->cwnd >> 10; + /* reset where we are in the table */ + if (cur_val < sctp_cwnd_adjust[0].cwnd) { + /* feel out of hs */ + net->last_hs_used = 0; + } else { + for (i = indx; i >= 1; i--) { + if (cur_val > sctp_cwnd_adjust[i - 1].cwnd) { + break; + } + } + net->last_hs_used = indx; + } + } + sctp_enforce_cwnd_limit(&stcb->asoc, net); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_FR); + } +} + +static void +sctp_hs_cwnd_update_after_fr(struct sctp_tcb *stcb, + struct sctp_association *asoc) +{ + struct sctp_nets *net; + /* + * CMT fast recovery code. Need to debug. ((sctp_cmt_on_off > 0) && + * (net->fast_retran_loss_recovery == 0))) + */ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if ((asoc->fast_retran_loss_recovery == 0) || + (asoc->sctp_cmt_on_off > 0)) { + /* out of a RFC2582 Fast recovery window? */ + if (net->net_ack > 0) { + /* + * per section 7.2.3, are there any + * destinations that had a fast retransmit + * to them. If so what we need to do is + * adjust ssthresh and cwnd. + */ + struct sctp_tmit_chunk *lchk; + + sctp_hs_cwnd_decrease(stcb, net); + + lchk = TAILQ_FIRST(&asoc->send_queue); + + net->partial_bytes_acked = 0; + /* Turn on fast recovery window */ + asoc->fast_retran_loss_recovery = 1; + if (lchk == NULL) { + /* Mark end of the window */ + asoc->fast_recovery_tsn = asoc->sending_seq - 1; + } else { + asoc->fast_recovery_tsn = lchk->rec.data.tsn - 1; + } + + /* + * CMT fast recovery -- per destination + * recovery variable. + */ + net->fast_retran_loss_recovery = 1; + + if (lchk == NULL) { + /* Mark end of the window */ + net->fast_recovery_tsn = asoc->sending_seq - 1; + } else { + net->fast_recovery_tsn = lchk->rec.data.tsn - 1; + } + + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, + stcb->sctp_ep, stcb, net, + SCTP_FROM_SCTP_CC_FUNCTIONS + SCTP_LOC_2); + sctp_timer_start(SCTP_TIMER_TYPE_SEND, + stcb->sctp_ep, stcb, net); + } + } else if (net->net_ack > 0) { + /* + * Mark a peg that we WOULD have done a cwnd + * reduction but RFC2582 prevented this action. + */ + SCTP_STAT_INCR(sctps_fastretransinrtt); + } + } +} + +static void +sctp_hs_cwnd_update_after_sack(struct sctp_tcb *stcb, + struct sctp_association *asoc, + int accum_moved, int reneged_all SCTP_UNUSED, int will_exit) +{ + struct sctp_nets *net; + /******************************/ + /* update cwnd and Early FR */ + /******************************/ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { +#ifdef JANA_CMT_FAST_RECOVERY + /* + * CMT fast recovery code. Need to debug. + */ + if (net->fast_retran_loss_recovery && net->new_pseudo_cumack) { + if (SCTP_TSN_GE(asoc->last_acked_seq, net->fast_recovery_tsn) || + SCTP_TSN_GE(net->pseudo_cumack,net->fast_recovery_tsn)) { + net->will_exit_fast_recovery = 1; + } + } +#endif + /* if nothing was acked on this destination skip it */ + if (net->net_ack == 0) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FROM_SACK); + } + continue; + } +#ifdef JANA_CMT_FAST_RECOVERY + /* CMT fast recovery code + */ + /* + if (sctp_cmt_on_off > 0 && net->fast_retran_loss_recovery && net->will_exit_fast_recovery == 0) { + @@@ Do something + } + else if (sctp_cmt_on_off == 0 && asoc->fast_retran_loss_recovery && will_exit == 0) { + */ +#endif + + if (asoc->fast_retran_loss_recovery && + (will_exit == 0) && + (asoc->sctp_cmt_on_off == 0)) { + /* + * If we are in loss recovery we skip any cwnd + * update + */ + return; + } + /* + * CMT: CUC algorithm. Update cwnd if pseudo-cumack has + * moved. + */ + if (accum_moved || + ((asoc->sctp_cmt_on_off > 0) && net->new_pseudo_cumack)) { + /* If the cumulative ack moved we can proceed */ + if (net->cwnd <= net->ssthresh) { + /* We are in slow start */ + if (net->flight_size + net->net_ack >= net->cwnd) { + sctp_hs_cwnd_increase(stcb, net); + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, net->net_ack, + SCTP_CWND_LOG_NOADV_SS); + } + } + } else { + /* We are in congestion avoidance */ + net->partial_bytes_acked += net->net_ack; + if ((net->flight_size + net->net_ack >= net->cwnd) && + (net->partial_bytes_acked >= net->cwnd)) { + net->partial_bytes_acked -= net->cwnd; + net->cwnd += net->mtu; + sctp_enforce_cwnd_limit(asoc, net); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, net->mtu, + SCTP_CWND_LOG_FROM_CA); + } + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, net->net_ack, + SCTP_CWND_LOG_NOADV_CA); + } + } + } + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, net->mtu, + SCTP_CWND_LOG_NO_CUMACK); + } + } + } +} + +/* + * H-TCP congestion control. The algorithm is detailed in: + * R.N.Shorten, D.J.Leith: + * "H-TCP: TCP for high-speed and long-distance networks" + * Proc. PFLDnet, Argonne, 2004. + * http://www.hamilton.ie/net/htcp3.pdf + */ + +static int use_rtt_scaling = 1; +static int use_bandwidth_switch = 1; + +static inline int +between(uint32_t seq1, uint32_t seq2, uint32_t seq3) +{ + return (seq3 - seq2 >= seq1 - seq2); +} + +static inline uint32_t +htcp_cong_time(struct htcp *ca) +{ + return (sctp_get_tick_count() - ca->last_cong); +} + +static inline uint32_t +htcp_ccount(struct htcp *ca) +{ + return (ca->minRTT == 0 ? htcp_cong_time(ca) : htcp_cong_time(ca)/ca->minRTT); +} + +static inline void +htcp_reset(struct htcp *ca) +{ + ca->undo_last_cong = ca->last_cong; + ca->undo_maxRTT = ca->maxRTT; + ca->undo_old_maxB = ca->old_maxB; + ca->last_cong = sctp_get_tick_count(); +} + +#ifdef SCTP_NOT_USED + +static uint32_t +htcp_cwnd_undo(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + net->cc_mod.htcp_ca.last_cong = net->cc_mod.htcp_ca.undo_last_cong; + net->cc_mod.htcp_ca.maxRTT = net->cc_mod.htcp_ca.undo_maxRTT; + net->cc_mod.htcp_ca.old_maxB = net->cc_mod.htcp_ca.undo_old_maxB; + return (max(net->cwnd, ((net->ssthresh/net->mtu<<7)/net->cc_mod.htcp_ca.beta)*net->mtu)); +} + +#endif + +static inline void +measure_rtt(struct sctp_nets *net) +{ + uint32_t srtt = net->lastsa>>SCTP_RTT_SHIFT; + + /* keep track of minimum RTT seen so far, minRTT is zero at first */ + if (net->cc_mod.htcp_ca.minRTT > srtt || !net->cc_mod.htcp_ca.minRTT) + net->cc_mod.htcp_ca.minRTT = srtt; + + /* max RTT */ + if (net->fast_retran_ip == 0 && net->ssthresh < 0xFFFF && htcp_ccount(&net->cc_mod.htcp_ca) > 3) { + if (net->cc_mod.htcp_ca.maxRTT < net->cc_mod.htcp_ca.minRTT) + net->cc_mod.htcp_ca.maxRTT = net->cc_mod.htcp_ca.minRTT; + if (net->cc_mod.htcp_ca.maxRTT < srtt && srtt <= net->cc_mod.htcp_ca.maxRTT+sctp_msecs_to_ticks(20)) + net->cc_mod.htcp_ca.maxRTT = srtt; + } +} + +static void +measure_achieved_throughput(struct sctp_nets *net) +{ + uint32_t now = sctp_get_tick_count(); + + if (net->fast_retran_ip == 0) + net->cc_mod.htcp_ca.bytes_acked = net->net_ack; + + if (!use_bandwidth_switch) + return; + + /* achieved throughput calculations */ + /* JRS - not 100% sure of this statement */ + if (net->fast_retran_ip == 1) { + net->cc_mod.htcp_ca.bytecount = 0; + net->cc_mod.htcp_ca.lasttime = now; + return; + } + + net->cc_mod.htcp_ca.bytecount += net->net_ack; + if ((net->cc_mod.htcp_ca.bytecount >= net->cwnd - (((net->cc_mod.htcp_ca.alpha >> 7) ? (net->cc_mod.htcp_ca.alpha >> 7) : 1) * net->mtu)) && + (now - net->cc_mod.htcp_ca.lasttime >= net->cc_mod.htcp_ca.minRTT) && + (net->cc_mod.htcp_ca.minRTT > 0)) { + uint32_t cur_Bi = net->cc_mod.htcp_ca.bytecount/net->mtu*hz/(now - net->cc_mod.htcp_ca.lasttime); + + if (htcp_ccount(&net->cc_mod.htcp_ca) <= 3) { + /* just after backoff */ + net->cc_mod.htcp_ca.minB = net->cc_mod.htcp_ca.maxB = net->cc_mod.htcp_ca.Bi = cur_Bi; + } else { + net->cc_mod.htcp_ca.Bi = (3*net->cc_mod.htcp_ca.Bi + cur_Bi)/4; + if (net->cc_mod.htcp_ca.Bi > net->cc_mod.htcp_ca.maxB) + net->cc_mod.htcp_ca.maxB = net->cc_mod.htcp_ca.Bi; + if (net->cc_mod.htcp_ca.minB > net->cc_mod.htcp_ca.maxB) + net->cc_mod.htcp_ca.minB = net->cc_mod.htcp_ca.maxB; + } + net->cc_mod.htcp_ca.bytecount = 0; + net->cc_mod.htcp_ca.lasttime = now; + } +} + +static inline void +htcp_beta_update(struct htcp *ca, uint32_t minRTT, uint32_t maxRTT) +{ + if (use_bandwidth_switch) { + uint32_t maxB = ca->maxB; + uint32_t old_maxB = ca->old_maxB; + ca->old_maxB = ca->maxB; + + if (!between(5*maxB, 4*old_maxB, 6*old_maxB)) { + ca->beta = BETA_MIN; + ca->modeswitch = 0; + return; + } + } + + if (ca->modeswitch && minRTT > sctp_msecs_to_ticks(10) && maxRTT) { + ca->beta = (minRTT<<7)/maxRTT; + if (ca->beta < BETA_MIN) + ca->beta = BETA_MIN; + else if (ca->beta > BETA_MAX) + ca->beta = BETA_MAX; + } else { + ca->beta = BETA_MIN; + ca->modeswitch = 1; + } +} + +static inline void +htcp_alpha_update(struct htcp *ca) +{ + uint32_t minRTT = ca->minRTT; + uint32_t factor = 1; + uint32_t diff = htcp_cong_time(ca); + + if (diff > (uint32_t)hz) { + diff -= hz; + factor = 1+ (10 * diff + ((diff / 2) * (diff / 2) / hz)) / hz; + } + + if (use_rtt_scaling && minRTT) { + uint32_t scale = (hz << 3) / (10 * minRTT); + scale = min(max(scale, 1U << 2), 10U << 3); /* clamping ratio to interval [0.5,10]<<3 */ + factor = (factor << 3) / scale; + if (factor != 0) + factor = 1; + } + + ca->alpha = 2 * factor * ((1 << 7) - ca->beta); + if (ca->alpha != 0) + ca->alpha = ALPHA_BASE; +} + +/* After we have the rtt data to calculate beta, we'd still prefer to wait one + * rtt before we adjust our beta to ensure we are working from a consistent + * data. + * + * This function should be called when we hit a congestion event since only at + * that point do we really have a real sense of maxRTT (the queues en route + * were getting just too full now). + */ +static void +htcp_param_update(struct sctp_nets *net) +{ + uint32_t minRTT = net->cc_mod.htcp_ca.minRTT; + uint32_t maxRTT = net->cc_mod.htcp_ca.maxRTT; + + htcp_beta_update(&net->cc_mod.htcp_ca, minRTT, maxRTT); + htcp_alpha_update(&net->cc_mod.htcp_ca); + + /* add slowly fading memory for maxRTT to accommodate routing changes etc */ + if (minRTT > 0 && maxRTT > minRTT) + net->cc_mod.htcp_ca.maxRTT = minRTT + ((maxRTT-minRTT)*95)/100; +} + +static uint32_t +htcp_recalc_ssthresh(struct sctp_nets *net) +{ + htcp_param_update(net); + return (max(((net->cwnd/net->mtu * net->cc_mod.htcp_ca.beta) >> 7)*net->mtu, 2U*net->mtu)); +} + +static void +htcp_cong_avoid(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + /*- + * How to handle these functions? + * if (!tcp_is_cwnd_limited(sk, in_flight)) RRS - good question. + * return; + */ + if (net->cwnd <= net->ssthresh) { + /* We are in slow start */ + if (net->flight_size + net->net_ack >= net->cwnd) { + if (net->net_ack > (net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable))) { + net->cwnd += (net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable)); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, net->mtu, + SCTP_CWND_LOG_FROM_SS); + } + + } else { + net->cwnd += net->net_ack; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, net->net_ack, + SCTP_CWND_LOG_FROM_SS); + } + } + sctp_enforce_cwnd_limit(&stcb->asoc, net); + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, net->net_ack, + SCTP_CWND_LOG_NOADV_SS); + } + } + } else { + measure_rtt(net); + + /* In dangerous area, increase slowly. + * In theory this is net->cwnd += alpha / net->cwnd + */ + /* What is snd_cwnd_cnt?? */ + if (((net->partial_bytes_acked/net->mtu * net->cc_mod.htcp_ca.alpha) >> 7)*net->mtu >= net->cwnd) { + /*- + * Does SCTP have a cwnd clamp? + * if (net->snd_cwnd < net->snd_cwnd_clamp) - Nope (RRS). + */ + net->cwnd += net->mtu; + net->partial_bytes_acked = 0; + sctp_enforce_cwnd_limit(&stcb->asoc, net); + htcp_alpha_update(&net->cc_mod.htcp_ca); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, net->mtu, + SCTP_CWND_LOG_FROM_CA); + } + } else { + net->partial_bytes_acked += net->net_ack; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, net->net_ack, + SCTP_CWND_LOG_NOADV_CA); + } + } + + net->cc_mod.htcp_ca.bytes_acked = net->mtu; + } +} + +#ifdef SCTP_NOT_USED +/* Lower bound on congestion window. */ +static uint32_t +htcp_min_cwnd(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + return (net->ssthresh); +} +#endif + +static void +htcp_init(struct sctp_nets *net) +{ + memset(&net->cc_mod.htcp_ca, 0, sizeof(struct htcp)); + net->cc_mod.htcp_ca.alpha = ALPHA_BASE; + net->cc_mod.htcp_ca.beta = BETA_MIN; + net->cc_mod.htcp_ca.bytes_acked = net->mtu; + net->cc_mod.htcp_ca.last_cong = sctp_get_tick_count(); +} + +static void +sctp_htcp_set_initial_cc_param(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + /* + * We take the max of the burst limit times a MTU or the + * INITIAL_CWND. We then limit this to 4 MTU's of sending. + */ + net->cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND)); + net->ssthresh = stcb->asoc.peers_rwnd; + sctp_enforce_cwnd_limit(&stcb->asoc, net); + htcp_init(net); + + if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_CWND_MONITOR_ENABLE|SCTP_CWND_LOGGING_ENABLE)) { + sctp_log_cwnd(stcb, net, 0, SCTP_CWND_INITIALIZATION); + } +} + +static void +sctp_htcp_cwnd_update_after_sack(struct sctp_tcb *stcb, + struct sctp_association *asoc, + int accum_moved, int reneged_all SCTP_UNUSED, int will_exit) +{ + struct sctp_nets *net; + + /******************************/ + /* update cwnd and Early FR */ + /******************************/ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { +#ifdef JANA_CMT_FAST_RECOVERY + /* + * CMT fast recovery code. Need to debug. + */ + if (net->fast_retran_loss_recovery && net->new_pseudo_cumack) { + if (SCTP_TSN_GE(asoc->last_acked_seq, net->fast_recovery_tsn) || + SCTP_TSN_GE(net->pseudo_cumack,net->fast_recovery_tsn)) { + net->will_exit_fast_recovery = 1; + } + } +#endif + /* if nothing was acked on this destination skip it */ + if (net->net_ack == 0) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FROM_SACK); + } + continue; + } +#ifdef JANA_CMT_FAST_RECOVERY + /* CMT fast recovery code + */ + /* + if (sctp_cmt_on_off > 0 && net->fast_retran_loss_recovery && net->will_exit_fast_recovery == 0) { + @@@ Do something + } + else if (sctp_cmt_on_off == 0 && asoc->fast_retran_loss_recovery && will_exit == 0) { + */ +#endif + + if (asoc->fast_retran_loss_recovery && + will_exit == 0 && + (asoc->sctp_cmt_on_off == 0)) { + /* + * If we are in loss recovery we skip any cwnd + * update + */ + return; + } + /* + * CMT: CUC algorithm. Update cwnd if pseudo-cumack has + * moved. + */ + if (accum_moved || + ((asoc->sctp_cmt_on_off > 0) && net->new_pseudo_cumack)) { + htcp_cong_avoid(stcb, net); + measure_achieved_throughput(net); + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, net->mtu, + SCTP_CWND_LOG_NO_CUMACK); + } + } + } +} + +static void +sctp_htcp_cwnd_update_after_fr(struct sctp_tcb *stcb, + struct sctp_association *asoc) +{ + struct sctp_nets *net; + /* + * CMT fast recovery code. Need to debug. ((sctp_cmt_on_off > 0) && + * (net->fast_retran_loss_recovery == 0))) + */ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if ((asoc->fast_retran_loss_recovery == 0) || + (asoc->sctp_cmt_on_off > 0)) { + /* out of a RFC2582 Fast recovery window? */ + if (net->net_ack > 0) { + /* + * per section 7.2.3, are there any + * destinations that had a fast retransmit + * to them. If so what we need to do is + * adjust ssthresh and cwnd. + */ + struct sctp_tmit_chunk *lchk; + int old_cwnd = net->cwnd; + + /* JRS - reset as if state were changed */ + htcp_reset(&net->cc_mod.htcp_ca); + net->ssthresh = htcp_recalc_ssthresh(net); + net->cwnd = net->ssthresh; + sctp_enforce_cwnd_limit(asoc, net); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), + SCTP_CWND_LOG_FROM_FR); + } + lchk = TAILQ_FIRST(&asoc->send_queue); + + net->partial_bytes_acked = 0; + /* Turn on fast recovery window */ + asoc->fast_retran_loss_recovery = 1; + if (lchk == NULL) { + /* Mark end of the window */ + asoc->fast_recovery_tsn = asoc->sending_seq - 1; + } else { + asoc->fast_recovery_tsn = lchk->rec.data.tsn - 1; + } + + /* + * CMT fast recovery -- per destination + * recovery variable. + */ + net->fast_retran_loss_recovery = 1; + + if (lchk == NULL) { + /* Mark end of the window */ + net->fast_recovery_tsn = asoc->sending_seq - 1; + } else { + net->fast_recovery_tsn = lchk->rec.data.tsn - 1; + } + + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, + stcb->sctp_ep, stcb, net, + SCTP_FROM_SCTP_CC_FUNCTIONS + SCTP_LOC_3); + sctp_timer_start(SCTP_TIMER_TYPE_SEND, + stcb->sctp_ep, stcb, net); + } + } else if (net->net_ack > 0) { + /* + * Mark a peg that we WOULD have done a cwnd + * reduction but RFC2582 prevented this action. + */ + SCTP_STAT_INCR(sctps_fastretransinrtt); + } + } +} + +static void +sctp_htcp_cwnd_update_after_timeout(struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + int old_cwnd = net->cwnd; + + /* JRS - reset as if the state were being changed to timeout */ + htcp_reset(&net->cc_mod.htcp_ca); + net->ssthresh = htcp_recalc_ssthresh(net); + net->cwnd = net->mtu; + net->partial_bytes_acked = 0; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, net->cwnd - old_cwnd, SCTP_CWND_LOG_FROM_RTX); + } +} + +static void +sctp_htcp_cwnd_update_after_ecn_echo(struct sctp_tcb *stcb, + struct sctp_nets *net, int in_window, int num_pkt_lost SCTP_UNUSED) +{ + int old_cwnd; + old_cwnd = net->cwnd; + + /* JRS - reset hctp as if state changed */ + if (in_window == 0) { + htcp_reset(&net->cc_mod.htcp_ca); + SCTP_STAT_INCR(sctps_ecnereducedcwnd); + net->ssthresh = htcp_recalc_ssthresh(net); + if (net->ssthresh < net->mtu) { + net->ssthresh = net->mtu; + /* here back off the timer as well, to slow us down */ + net->RTO <<= 1; + } + net->cwnd = net->ssthresh; + sctp_enforce_cwnd_limit(&stcb->asoc, net); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT); + } + } +} + +const struct sctp_cc_functions sctp_cc_functions[] = { +{ +#if defined(_WIN32) && !defined(__MINGW32__) + sctp_set_initial_cc_param, + sctp_cwnd_update_after_sack, + sctp_cwnd_update_exit_pf_common, + sctp_cwnd_update_after_fr, + sctp_cwnd_update_after_timeout, + sctp_cwnd_update_after_ecn_echo, + sctp_cwnd_update_after_packet_dropped, + sctp_cwnd_update_after_output, +#else + .sctp_set_initial_cc_param = sctp_set_initial_cc_param, + .sctp_cwnd_update_after_sack = sctp_cwnd_update_after_sack, + .sctp_cwnd_update_exit_pf = sctp_cwnd_update_exit_pf_common, + .sctp_cwnd_update_after_fr = sctp_cwnd_update_after_fr, + .sctp_cwnd_update_after_timeout = sctp_cwnd_update_after_timeout, + .sctp_cwnd_update_after_ecn_echo = sctp_cwnd_update_after_ecn_echo, + .sctp_cwnd_update_after_packet_dropped = sctp_cwnd_update_after_packet_dropped, + .sctp_cwnd_update_after_output = sctp_cwnd_update_after_output, +#endif +}, +{ +#if defined(_WIN32) && !defined(__MINGW32__) + sctp_set_initial_cc_param, + sctp_hs_cwnd_update_after_sack, + sctp_cwnd_update_exit_pf_common, + sctp_hs_cwnd_update_after_fr, + sctp_cwnd_update_after_timeout, + sctp_cwnd_update_after_ecn_echo, + sctp_cwnd_update_after_packet_dropped, + sctp_cwnd_update_after_output, +#else + .sctp_set_initial_cc_param = sctp_set_initial_cc_param, + .sctp_cwnd_update_after_sack = sctp_hs_cwnd_update_after_sack, + .sctp_cwnd_update_exit_pf = sctp_cwnd_update_exit_pf_common, + .sctp_cwnd_update_after_fr = sctp_hs_cwnd_update_after_fr, + .sctp_cwnd_update_after_timeout = sctp_cwnd_update_after_timeout, + .sctp_cwnd_update_after_ecn_echo = sctp_cwnd_update_after_ecn_echo, + .sctp_cwnd_update_after_packet_dropped = sctp_cwnd_update_after_packet_dropped, + .sctp_cwnd_update_after_output = sctp_cwnd_update_after_output, +#endif +}, +{ +#if defined(_WIN32) && !defined(__MINGW32__) + sctp_htcp_set_initial_cc_param, + sctp_htcp_cwnd_update_after_sack, + sctp_cwnd_update_exit_pf_common, + sctp_htcp_cwnd_update_after_fr, + sctp_htcp_cwnd_update_after_timeout, + sctp_htcp_cwnd_update_after_ecn_echo, + sctp_cwnd_update_after_packet_dropped, + sctp_cwnd_update_after_output, +#else + .sctp_set_initial_cc_param = sctp_htcp_set_initial_cc_param, + .sctp_cwnd_update_after_sack = sctp_htcp_cwnd_update_after_sack, + .sctp_cwnd_update_exit_pf = sctp_cwnd_update_exit_pf_common, + .sctp_cwnd_update_after_fr = sctp_htcp_cwnd_update_after_fr, + .sctp_cwnd_update_after_timeout = sctp_htcp_cwnd_update_after_timeout, + .sctp_cwnd_update_after_ecn_echo = sctp_htcp_cwnd_update_after_ecn_echo, + .sctp_cwnd_update_after_packet_dropped = sctp_cwnd_update_after_packet_dropped, + .sctp_cwnd_update_after_output = sctp_cwnd_update_after_output, +#endif +}, +{ +#if defined(_WIN32) && !defined(__MINGW32__) + sctp_set_rtcc_initial_cc_param, + sctp_cwnd_update_rtcc_after_sack, + sctp_cwnd_update_exit_pf_common, + sctp_cwnd_update_after_fr, + sctp_cwnd_update_after_timeout, + sctp_cwnd_update_rtcc_after_ecn_echo, + sctp_cwnd_update_after_packet_dropped, + sctp_cwnd_update_after_output, + sctp_cwnd_update_rtcc_packet_transmitted, + sctp_cwnd_update_rtcc_tsn_acknowledged, + sctp_cwnd_new_rtcc_transmission_begins, + sctp_cwnd_prepare_rtcc_net_for_sack, + sctp_cwnd_rtcc_socket_option, + sctp_rtt_rtcc_calculated +#else + .sctp_set_initial_cc_param = sctp_set_rtcc_initial_cc_param, + .sctp_cwnd_update_after_sack = sctp_cwnd_update_rtcc_after_sack, + .sctp_cwnd_update_exit_pf = sctp_cwnd_update_exit_pf_common, + .sctp_cwnd_update_after_fr = sctp_cwnd_update_after_fr, + .sctp_cwnd_update_after_timeout = sctp_cwnd_update_after_timeout, + .sctp_cwnd_update_after_ecn_echo = sctp_cwnd_update_rtcc_after_ecn_echo, + .sctp_cwnd_update_after_packet_dropped = sctp_cwnd_update_after_packet_dropped, + .sctp_cwnd_update_after_output = sctp_cwnd_update_after_output, + .sctp_cwnd_update_packet_transmitted = sctp_cwnd_update_rtcc_packet_transmitted, + .sctp_cwnd_update_tsn_acknowledged = sctp_cwnd_update_rtcc_tsn_acknowledged, + .sctp_cwnd_new_transmission_begins = sctp_cwnd_new_rtcc_transmission_begins, + .sctp_cwnd_prepare_net_for_sack = sctp_cwnd_prepare_rtcc_net_for_sack, + .sctp_cwnd_socket_option = sctp_cwnd_rtcc_socket_option, + .sctp_rtt_calculated = sctp_rtt_rtcc_calculated +#endif +} +}; diff --git a/netwerk/sctp/src/netinet/sctp_constants.h b/netwerk/sctp/src/netinet/sctp_constants.h new file mode 100644 index 0000000000..ca2ed70b59 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_constants.h @@ -0,0 +1,1068 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_CONSTANTS_H_ +#define _NETINET_SCTP_CONSTANTS_H_ + +#if defined(_WIN32) && defined(__Userspace__) +extern void getwintimeofday(struct timeval *tv); + +#endif +/* IANA assigned port number for SCTP over UDP encapsulation */ +#define SCTP_OVER_UDP_TUNNELING_PORT 9899 + +/* Number of packets to get before sack sent by default */ +#define SCTP_DEFAULT_SACK_FREQ 2 + +/* Address limit - This variable is calculated + * based on an 65535 byte max ip packet. We take out 100 bytes + * for the cookie, 40 bytes for a v6 header and 32 + * bytes for the init structure. A second init structure + * for the init-ack and then finally a third one for the + * imbedded init. This yeilds 100+40+(3 * 32) = 236 bytes. + * This leaves 65299 bytes for addresses. We throw out the 299 bytes. + * Now whatever we send in the INIT() we need to allow to get back in the + * INIT-ACK plus all the values from INIT and INIT-ACK + * listed in the cookie. Plus we need some overhead for + * maybe copied parameters in the COOKIE. If we + * allow 1080 addresses, and each side has 1080 V6 addresses + * that will be 21600 bytes. In the INIT-ACK we will + * see the INIT-ACK 21600 + 43200 in the cookie. This leaves + * about 500 bytes slack for misc things in the cookie. + */ +#define SCTP_ADDRESS_LIMIT 1080 + +/* We need at least 2k of space for us, inits + * larger than that lets abort. + */ +#define SCTP_LARGEST_INIT_ACCEPTED (65535 - 2048) + +/* Largest length of a chunk */ +#define SCTP_MAX_CHUNK_LENGTH 0xffff +/* Largest length of an error cause */ +#define SCTP_MAX_CAUSE_LENGTH 0xffff +/* Number of addresses where we just skip the counting */ +#define SCTP_COUNT_LIMIT 40 + +#define SCTP_ZERO_COPY_TICK_DELAY (((100 * hz) + 999) / 1000) +#define SCTP_ZERO_COPY_SENDQ_TICK_DELAY (((100 * hz) + 999) / 1000) + +/* Number of ticks to delay before running + * iterator on an address change. + */ +#define SCTP_ADDRESS_TICK_DELAY 2 + +#define SCTP_VERSION_STRING "KAME-BSD 1.1" +/* #define SCTP_AUDITING_ENABLED 1 used for debug/auditing */ +#define SCTP_AUDIT_SIZE 256 + +#define SCTP_KTRHEAD_NAME "sctp_iterator" +#define SCTP_KTHREAD_PAGES 0 + +#define SCTP_MCORE_NAME "sctp_core_worker" + +/* If you support Multi-VRF how big to + * make the initial array of VRF's to. + */ +#define SCTP_DEFAULT_VRF_SIZE 4 + +/* JRS - Values defined for the HTCP algorithm */ +#define ALPHA_BASE (1<<7) /* 1.0 with shift << 7 */ +#define BETA_MIN (1<<6) /* 0.5 with shift << 7 */ +#define BETA_MAX 102 /* 0.8 with shift << 7 */ + +/* Places that CWND log can happen from */ +#define SCTP_CWND_LOG_FROM_FR 1 +#define SCTP_CWND_LOG_FROM_RTX 2 +#define SCTP_CWND_LOG_FROM_BRST 3 +#define SCTP_CWND_LOG_FROM_SS 4 +#define SCTP_CWND_LOG_FROM_CA 5 +#define SCTP_CWND_LOG_FROM_SAT 6 +#define SCTP_BLOCK_LOG_INTO_BLK 7 +#define SCTP_BLOCK_LOG_OUTOF_BLK 8 +#define SCTP_BLOCK_LOG_CHECK 9 +#define SCTP_STR_LOG_FROM_INTO_STRD 10 +#define SCTP_STR_LOG_FROM_IMMED_DEL 11 +#define SCTP_STR_LOG_FROM_INSERT_HD 12 +#define SCTP_STR_LOG_FROM_INSERT_MD 13 +#define SCTP_STR_LOG_FROM_INSERT_TL 14 +#define SCTP_STR_LOG_FROM_MARK_TSN 15 +#define SCTP_STR_LOG_FROM_EXPRS_DEL 16 +#define SCTP_FR_LOG_BIGGEST_TSNS 17 +#define SCTP_FR_LOG_STRIKE_TEST 18 +#define SCTP_FR_LOG_STRIKE_CHUNK 19 +#define SCTP_FR_T3_TIMEOUT 20 +#define SCTP_MAP_PREPARE_SLIDE 21 +#define SCTP_MAP_SLIDE_FROM 22 +#define SCTP_MAP_SLIDE_RESULT 23 +#define SCTP_MAP_SLIDE_CLEARED 24 +#define SCTP_MAP_SLIDE_NONE 25 +#define SCTP_FR_T3_MARK_TIME 26 +#define SCTP_FR_T3_MARKED 27 +#define SCTP_FR_T3_STOPPED 28 +#define SCTP_FR_MARKED 30 +#define SCTP_CWND_LOG_NOADV_SS 31 +#define SCTP_CWND_LOG_NOADV_CA 32 +#define SCTP_MAX_BURST_APPLIED 33 +#define SCTP_MAX_IFP_APPLIED 34 +#define SCTP_MAX_BURST_ERROR_STOP 35 +#define SCTP_INCREASE_PEER_RWND 36 +#define SCTP_DECREASE_PEER_RWND 37 +#define SCTP_SET_PEER_RWND_VIA_SACK 38 +#define SCTP_LOG_MBCNT_INCREASE 39 +#define SCTP_LOG_MBCNT_DECREASE 40 +#define SCTP_LOG_MBCNT_CHKSET 41 +#define SCTP_LOG_NEW_SACK 42 +#define SCTP_LOG_TSN_ACKED 43 +#define SCTP_LOG_TSN_REVOKED 44 +#define SCTP_LOG_LOCK_TCB 45 +#define SCTP_LOG_LOCK_INP 46 +#define SCTP_LOG_LOCK_SOCK 47 +#define SCTP_LOG_LOCK_SOCKBUF_R 48 +#define SCTP_LOG_LOCK_SOCKBUF_S 49 +#define SCTP_LOG_LOCK_CREATE 50 +#define SCTP_LOG_INITIAL_RTT 51 +#define SCTP_LOG_RTTVAR 52 +#define SCTP_LOG_SBALLOC 53 +#define SCTP_LOG_SBFREE 54 +#define SCTP_LOG_SBRESULT 55 +#define SCTP_FR_DUPED 56 +#define SCTP_FR_MARKED_EARLY 57 +#define SCTP_FR_CWND_REPORT 58 +#define SCTP_FR_CWND_REPORT_START 59 +#define SCTP_FR_CWND_REPORT_STOP 60 +#define SCTP_CWND_LOG_FROM_SEND 61 +#define SCTP_CWND_INITIALIZATION 62 +#define SCTP_CWND_LOG_FROM_T3 63 +#define SCTP_CWND_LOG_FROM_SACK 64 +#define SCTP_CWND_LOG_NO_CUMACK 65 +#define SCTP_CWND_LOG_FROM_RESEND 66 +#define SCTP_FR_LOG_CHECK_STRIKE 67 +#define SCTP_SEND_NOW_COMPLETES 68 +#define SCTP_CWND_LOG_FILL_OUTQ_CALLED 69 +#define SCTP_CWND_LOG_FILL_OUTQ_FILLS 70 +#define SCTP_LOG_FREE_SENT 71 +#define SCTP_NAGLE_APPLIED 72 +#define SCTP_NAGLE_SKIPPED 73 +#define SCTP_WAKESND_FROM_SACK 74 +#define SCTP_WAKESND_FROM_FWDTSN 75 +#define SCTP_NOWAKE_FROM_SACK 76 +#define SCTP_CWNDLOG_PRESEND 77 +#define SCTP_CWNDLOG_ENDSEND 78 +#define SCTP_AT_END_OF_SACK 79 +#define SCTP_REASON_FOR_SC 80 +#define SCTP_BLOCK_LOG_INTO_BLKA 81 +#define SCTP_ENTER_USER_RECV 82 +#define SCTP_USER_RECV_SACKS 83 +#define SCTP_SORECV_BLOCKSA 84 +#define SCTP_SORECV_BLOCKSB 85 +#define SCTP_SORECV_DONE 86 +#define SCTP_SACK_RWND_UPDATE 87 +#define SCTP_SORECV_ENTER 88 +#define SCTP_SORECV_ENTERPL 89 +#define SCTP_MBUF_INPUT 90 +#define SCTP_MBUF_IALLOC 91 +#define SCTP_MBUF_IFREE 92 +#define SCTP_MBUF_ICOPY 93 +#define SCTP_MBUF_SPLIT 94 +#define SCTP_SORCV_FREECTL 95 +#define SCTP_SORCV_DOESCPY 96 +#define SCTP_SORCV_DOESLCK 97 +#define SCTP_SORCV_DOESADJ 98 +#define SCTP_SORCV_BOTWHILE 99 +#define SCTP_SORCV_PASSBF 100 +#define SCTP_SORCV_ADJD 101 +#define SCTP_UNKNOWN_MAX 102 +#define SCTP_RANDY_STUFF 103 +#define SCTP_RANDY_STUFF1 104 +#define SCTP_STRMOUT_LOG_ASSIGN 105 +#define SCTP_STRMOUT_LOG_SEND 106 +#define SCTP_FLIGHT_LOG_DOWN_CA 107 +#define SCTP_FLIGHT_LOG_UP 108 +#define SCTP_FLIGHT_LOG_DOWN_GAP 109 +#define SCTP_FLIGHT_LOG_DOWN_RSND 110 +#define SCTP_FLIGHT_LOG_UP_RSND 111 +#define SCTP_FLIGHT_LOG_DOWN_RSND_TO 112 +#define SCTP_FLIGHT_LOG_DOWN_WP 113 +#define SCTP_FLIGHT_LOG_UP_REVOKE 114 +#define SCTP_FLIGHT_LOG_DOWN_PDRP 115 +#define SCTP_FLIGHT_LOG_DOWN_PMTU 116 +#define SCTP_SACK_LOG_NORMAL 117 +#define SCTP_SACK_LOG_EXPRESS 118 +#define SCTP_MAP_TSN_ENTERS 119 +#define SCTP_THRESHOLD_CLEAR 120 +#define SCTP_THRESHOLD_INCR 121 +#define SCTP_FLIGHT_LOG_DWN_WP_FWD 122 +#define SCTP_FWD_TSN_CHECK 123 +#define SCTP_LOG_MAX_TYPES 124 +/* + * To turn on various logging, you must first enable 'options KTR' and + * you might want to bump the entires 'options KTR_ENTRIES=80000'. + * To get something to log you define one of the logging defines. + * (see LINT). + * + * This gets the compile in place, but you still need to turn the + * logging flag on too in the sysctl (see in sctp.h). + */ + +#define SCTP_LOG_EVENT_UNKNOWN 0 +#define SCTP_LOG_EVENT_CWND 1 +#define SCTP_LOG_EVENT_BLOCK 2 +#define SCTP_LOG_EVENT_STRM 3 +#define SCTP_LOG_EVENT_FR 4 +#define SCTP_LOG_EVENT_MAP 5 +#define SCTP_LOG_EVENT_MAXBURST 6 +#define SCTP_LOG_EVENT_RWND 7 +#define SCTP_LOG_EVENT_MBCNT 8 +#define SCTP_LOG_EVENT_SACK 9 +#define SCTP_LOG_LOCK_EVENT 10 +#define SCTP_LOG_EVENT_RTT 11 +#define SCTP_LOG_EVENT_SB 12 +#define SCTP_LOG_EVENT_NAGLE 13 +#define SCTP_LOG_EVENT_WAKE 14 +#define SCTP_LOG_MISC_EVENT 15 +#define SCTP_LOG_EVENT_CLOSE 16 +#define SCTP_LOG_EVENT_MBUF 17 +#define SCTP_LOG_CHUNK_PROC 18 +#define SCTP_LOG_ERROR_RET 19 + +#define SCTP_LOG_MAX_EVENT 20 + +#define SCTP_LOCK_UNKNOWN 2 + +/* number of associations by default for zone allocation */ +#define SCTP_MAX_NUM_OF_ASOC 40000 +/* how many addresses per assoc remote and local */ +#define SCTP_SCALE_FOR_ADDR 2 + +/* default MULTIPLE_ASCONF mode enable(1)/disable(0) value (sysctl) */ +#define SCTP_DEFAULT_MULTIPLE_ASCONFS 0 + +/* + * Threshold for rwnd updates, we have to read (sb_hiwat >> + * SCTP_RWND_HIWAT_SHIFT) before we will look to see if we need to send a + * window update sack. When we look, we compare the last rwnd we sent vs the + * current rwnd. It too must be greater than this value. Using 3 divdes the + * hiwat by 8, so for 200k rwnd we need to read 24k. For a 64k rwnd we need + * to read 8k. This seems about right.. I hope :-D.. we do set a + * min of a MTU on it so if the rwnd is real small we will insist + * on a full MTU of 1500 bytes. + */ +#define SCTP_RWND_HIWAT_SHIFT 3 + +/* How much of the rwnd must the + * message be taking up to start partial delivery. + * We calculate this by shifing the hi_water (recv_win) + * left the following .. set to 1, when a message holds + * 1/2 the rwnd. If we set it to 2 when a message holds + * 1/4 the rwnd...etc.. + */ + +#define SCTP_PARTIAL_DELIVERY_SHIFT 1 + +/* + * default HMAC for cookies, etc... use one of the AUTH HMAC id's + * SCTP_HMAC is the HMAC_ID to use + * SCTP_SIGNATURE_SIZE is the digest length + */ +#define SCTP_HMAC SCTP_AUTH_HMAC_ID_SHA1 +#define SCTP_SIGNATURE_SIZE SCTP_AUTH_DIGEST_LEN_SHA1 +#define SCTP_SIGNATURE_ALOC_SIZE SCTP_SIGNATURE_SIZE + +/* + * the SCTP protocol signature this includes the version number encoded in + * the last 4 bits of the signature. + */ +#define PROTO_SIGNATURE_A 0x30000000 +#define SCTP_VERSION_NUMBER 0x3 + +#define MAX_TSN 0xffffffff + +/* how many executions every N tick's */ +#define SCTP_ITERATOR_MAX_AT_ONCE 20 + +/* number of clock ticks between iterator executions */ +#define SCTP_ITERATOR_TICKS 1 + +/* + * option: If you comment out the following you will receive the old behavior + * of obeying cwnd for the fast retransmit algorithm. With this defined a FR + * happens right away with-out waiting for the flightsize to drop below the + * cwnd value (which is reduced by the FR to 1/2 the inflight packets). + */ +#define SCTP_IGNORE_CWND_ON_FR 1 + +/* + * Adds implementors guide behavior to only use newest highest update in SACK + * gap ack's to figure out if you need to stroke a chunk for FR. + */ +#define SCTP_NO_FR_UNLESS_SEGMENT_SMALLER 1 + +/* default max I can burst out after a fast retransmit, 0 disables it */ +#define SCTP_DEF_MAX_BURST 4 +#define SCTP_DEF_HBMAX_BURST 4 +#define SCTP_DEF_FRMAX_BURST 4 + +/* RTO calculation flag to say if it + * is safe to determine local lan or not. + */ +#define SCTP_RTT_FROM_NON_DATA 0 +#define SCTP_RTT_FROM_DATA 1 + +#define PR_SCTP_UNORDERED_FLAG 0x0001 + +/* IP hdr (20/40) + 12+2+2 (enet) + sctp common 12 */ +#define SCTP_FIRST_MBUF_RESV 68 +/* Packet transmit states in the sent field */ +#define SCTP_DATAGRAM_UNSENT 0 +#define SCTP_DATAGRAM_SENT 1 +#define SCTP_DATAGRAM_RESEND1 2 /* not used (in code, but may + * hit this value) */ +#define SCTP_DATAGRAM_RESEND2 3 /* not used (in code, but may + * hit this value) */ +#define SCTP_DATAGRAM_RESEND 4 +#define SCTP_DATAGRAM_ACKED 10010 +#define SCTP_DATAGRAM_MARKED 20010 +#define SCTP_FORWARD_TSN_SKIP 30010 +#define SCTP_DATAGRAM_NR_ACKED 40010 + +/* chunk output send from locations */ +#define SCTP_OUTPUT_FROM_USR_SEND 0 +#define SCTP_OUTPUT_FROM_T3 1 +#define SCTP_OUTPUT_FROM_INPUT_ERROR 2 +#define SCTP_OUTPUT_FROM_CONTROL_PROC 3 +#define SCTP_OUTPUT_FROM_SACK_TMR 4 +#define SCTP_OUTPUT_FROM_SHUT_TMR 5 +#define SCTP_OUTPUT_FROM_HB_TMR 6 +#define SCTP_OUTPUT_FROM_SHUT_ACK_TMR 7 +#define SCTP_OUTPUT_FROM_ASCONF_TMR 8 +#define SCTP_OUTPUT_FROM_STRRST_TMR 9 +#define SCTP_OUTPUT_FROM_AUTOCLOSE_TMR 10 +#define SCTP_OUTPUT_FROM_EARLY_FR_TMR 11 +#define SCTP_OUTPUT_FROM_STRRST_REQ 12 +#define SCTP_OUTPUT_FROM_USR_RCVD 13 +#define SCTP_OUTPUT_FROM_COOKIE_ACK 14 +#define SCTP_OUTPUT_FROM_DRAIN 15 +#define SCTP_OUTPUT_FROM_CLOSING 16 +#define SCTP_OUTPUT_FROM_SOCKOPT 17 + +/* SCTP chunk types are moved sctp.h for application (NAT, FW) use */ + +/* align to 32-bit sizes */ +#define SCTP_SIZE32(x) ((((x) + 3) >> 2) << 2) + +#define IS_SCTP_CONTROL(a) (((a)->chunk_type != SCTP_DATA) && ((a)->chunk_type != SCTP_IDATA)) +#define IS_SCTP_DATA(a) (((a)->chunk_type == SCTP_DATA) || ((a)->chunk_type == SCTP_IDATA)) + +/* SCTP parameter types */ +/*************0x0000 series*************/ +#define SCTP_HEARTBEAT_INFO 0x0001 +#if defined(__Userspace__) +#define SCTP_CONN_ADDRESS 0x0004 +#endif +#define SCTP_IPV4_ADDRESS 0x0005 +#define SCTP_IPV6_ADDRESS 0x0006 +#define SCTP_STATE_COOKIE 0x0007 +#define SCTP_UNRECOG_PARAM 0x0008 +#define SCTP_COOKIE_PRESERVE 0x0009 +#define SCTP_HOSTNAME_ADDRESS 0x000b +#define SCTP_SUPPORTED_ADDRTYPE 0x000c + +/* RFC 6525 */ +#define SCTP_STR_RESET_OUT_REQUEST 0x000d +#define SCTP_STR_RESET_IN_REQUEST 0x000e +#define SCTP_STR_RESET_TSN_REQUEST 0x000f +#define SCTP_STR_RESET_RESPONSE 0x0010 +#define SCTP_STR_RESET_ADD_OUT_STREAMS 0x0011 +#define SCTP_STR_RESET_ADD_IN_STREAMS 0x0012 + +#define SCTP_MAX_RESET_PARAMS 2 +#define SCTP_STREAM_RESET_TSN_DELTA 0x1000 + +/*************0x4000 series*************/ + +/*************0x8000 series*************/ +#define SCTP_ECN_CAPABLE 0x8000 +#define SCTP_ZERO_CHECKSUM_ACCEPTABLE 0x8001 +/* RFC 4895 */ +#define SCTP_RANDOM 0x8002 +#define SCTP_CHUNK_LIST 0x8003 +#define SCTP_HMAC_LIST 0x8004 +/* RFC 4820 */ +#define SCTP_PAD 0x8005 +/* RFC 5061 */ +#define SCTP_SUPPORTED_CHUNK_EXT 0x8008 + +/*************0xC000 series*************/ +#define SCTP_PRSCTP_SUPPORTED 0xc000 +/* RFC 5061 */ +#define SCTP_ADD_IP_ADDRESS 0xc001 +#define SCTP_DEL_IP_ADDRESS 0xc002 +#define SCTP_ERROR_CAUSE_IND 0xc003 +#define SCTP_SET_PRIM_ADDR 0xc004 +#define SCTP_SUCCESS_REPORT 0xc005 +#define SCTP_ULP_ADAPTATION 0xc006 +/* behave-nat-draft */ +#define SCTP_HAS_NAT_SUPPORT 0xc007 +#define SCTP_NAT_VTAGS 0xc008 + +/* bits for TOS field */ +#define SCTP_ECT0_BIT 0x02 +#define SCTP_ECT1_BIT 0x01 +#define SCTP_CE_BITS 0x03 + +/* below turns off above */ +#define SCTP_FLEXIBLE_ADDRESS 0x20 +#define SCTP_NO_HEARTBEAT 0x40 + +/* mask to get sticky */ +#define SCTP_STICKY_OPTIONS_MASK 0x0c + +/* + * SCTP states for internal state machine + */ +#define SCTP_STATE_EMPTY 0x0000 +#define SCTP_STATE_INUSE 0x0001 +#define SCTP_STATE_COOKIE_WAIT 0x0002 +#define SCTP_STATE_COOKIE_ECHOED 0x0004 +#define SCTP_STATE_OPEN 0x0008 +#define SCTP_STATE_SHUTDOWN_SENT 0x0010 +#define SCTP_STATE_SHUTDOWN_RECEIVED 0x0020 +#define SCTP_STATE_SHUTDOWN_ACK_SENT 0x0040 +#define SCTP_STATE_SHUTDOWN_PENDING 0x0080 +#define SCTP_STATE_CLOSED_SOCKET 0x0100 +#define SCTP_STATE_ABOUT_TO_BE_FREED 0x0200 +#define SCTP_STATE_PARTIAL_MSG_LEFT 0x0400 +#define SCTP_STATE_WAS_ABORTED 0x0800 +#define SCTP_STATE_IN_ACCEPT_QUEUE 0x1000 +#define SCTP_STATE_MASK 0x007f + +#define SCTP_GET_STATE(_stcb) \ + ((_stcb)->asoc.state & SCTP_STATE_MASK) +#define SCTP_SET_STATE(_stcb, _state) \ + sctp_set_state(_stcb, _state) +#define SCTP_CLEAR_SUBSTATE(_stcb, _substate) \ + (_stcb)->asoc.state &= ~(_substate) +#define SCTP_ADD_SUBSTATE(_stcb, _substate) \ + sctp_add_substate(_stcb, _substate) + +/* SCTP reachability state for each address */ +#define SCTP_ADDR_REACHABLE 0x001 +#define SCTP_ADDR_NO_PMTUD 0x002 +#define SCTP_ADDR_NOHB 0x004 +#define SCTP_ADDR_BEING_DELETED 0x008 +#define SCTP_ADDR_NOT_IN_ASSOC 0x010 +#define SCTP_ADDR_OUT_OF_SCOPE 0x080 +#define SCTP_ADDR_UNCONFIRMED 0x200 +#define SCTP_ADDR_REQ_PRIMARY 0x400 +/* JRS 5/13/07 - Added potentially failed state for CMT PF */ +#define SCTP_ADDR_PF 0x800 + +/* bound address types (e.g. valid address types to allow) */ +#define SCTP_BOUND_V6 0x01 +#define SCTP_BOUND_V4 0x02 + +/* + * what is the default number of mbufs in a chain I allow before switching to + * a cluster + */ +#define SCTP_DEFAULT_MBUFS_IN_CHAIN 5 + +/* How long a cookie lives in milli-seconds */ +#define SCTP_DEFAULT_COOKIE_LIFE 60000 + +/* Maximum the mapping array will grow to (TSN mapping array) */ +#define SCTP_MAPPING_ARRAY 512 + +/* size of the initial malloc on the mapping array */ +#define SCTP_INITIAL_MAPPING_ARRAY 16 +/* how much we grow the mapping array each call */ +#define SCTP_MAPPING_ARRAY_INCR 32 + +/* + * Here we define the timer types used by the implementation as arguments in + * the set/get timer type calls. + */ +#define SCTP_TIMER_INIT 0 +#define SCTP_TIMER_RECV 1 +#define SCTP_TIMER_SEND 2 +#define SCTP_TIMER_HEARTBEAT 3 +#define SCTP_TIMER_PMTU 4 +#define SCTP_TIMER_MAXSHUTDOWN 5 +#define SCTP_TIMER_SIGNATURE 6 +/* + * number of timer types in the base SCTP structure used in the set/get and + * has the base default. + */ +#define SCTP_NUM_TMRS 7 + +/* timer types */ +#define SCTP_TIMER_TYPE_NONE 0 +#define SCTP_TIMER_TYPE_SEND 1 +#define SCTP_TIMER_TYPE_INIT 2 +#define SCTP_TIMER_TYPE_RECV 3 +#define SCTP_TIMER_TYPE_SHUTDOWN 4 +#define SCTP_TIMER_TYPE_HEARTBEAT 5 +#define SCTP_TIMER_TYPE_COOKIE 6 +#define SCTP_TIMER_TYPE_NEWCOOKIE 7 +#define SCTP_TIMER_TYPE_PATHMTURAISE 8 +#define SCTP_TIMER_TYPE_SHUTDOWNACK 9 +#define SCTP_TIMER_TYPE_ASCONF 10 +#define SCTP_TIMER_TYPE_SHUTDOWNGUARD 11 +#define SCTP_TIMER_TYPE_AUTOCLOSE 12 +#define SCTP_TIMER_TYPE_STRRESET 13 +#define SCTP_TIMER_TYPE_INPKILL 14 +#define SCTP_TIMER_TYPE_ASOCKILL 15 +#define SCTP_TIMER_TYPE_ADDR_WQ 16 +#define SCTP_TIMER_TYPE_PRIM_DELETED 17 +/* add new timers here - and increment LAST */ +#define SCTP_TIMER_TYPE_LAST 18 + +#define SCTP_IS_TIMER_TYPE_VALID(t) (((t) > SCTP_TIMER_TYPE_NONE) && \ + ((t) < SCTP_TIMER_TYPE_LAST)) + +#if defined(__APPLE__) && !defined(__Userspace__) +/* Number of ticks to run the main timer at in msec */ +#define SCTP_MAIN_TIMER_DEFAULT 10 + +#endif +/* max number of TSN's dup'd that I will hold */ +#define SCTP_MAX_DUP_TSNS 20 + +/* + * Here we define the types used when setting the retry amounts. + */ +/* How many drop re-attempts we make on INIT/COOKIE-ECHO */ +#define SCTP_RETRY_DROPPED_THRESH 4 + +/* + * Maxmium number of chunks a single association can have on it. Note that + * this is a squishy number since the count can run over this if the user + * sends a large message down .. the fragmented chunks don't count until + * AFTER the message is on queue.. it would be the next send that blocks + * things. This number will get tuned up at boot in the sctp_init and use the + * number of clusters as a base. This way high bandwidth environments will + * not get impacted by the lower bandwidth sending a bunch of 1 byte chunks + */ +#define SCTP_ASOC_MAX_CHUNKS_ON_QUEUE 512 + +/* + * Basically the minimum amount of time before I do a early FR. Making this + * value to low will cause duplicate retransmissions. + */ +#define SCTP_MINFR_MSEC_TIMER 250 +/* The floor this value is allowed to fall to when starting a timer. */ +#define SCTP_MINFR_MSEC_FLOOR 20 + +/* init timer def = 1 sec */ +#define SCTP_INIT_SEC 1 + +/* send timer def = 1 seconds */ +#define SCTP_SEND_SEC 1 + +/* recv timer def = 200ms */ +#define SCTP_RECV_MSEC 200 + +/* 30 seconds + RTO (in ms) */ +#define SCTP_HB_DEFAULT_MSEC 30000 + +/* + * This is how long a secret lives, NOT how long a cookie lives how many + * ticks the current secret will live. + */ +#define SCTP_DEFAULT_SECRET_LIFE_SEC 3600 + +#define SCTP_RTO_UPPER_BOUND (60000) /* 60 sec in ms */ +#define SCTP_RTO_LOWER_BOUND (1000) /* 1 sec is ms */ +#define SCTP_RTO_INITIAL (1000) /* 1 sec in ms */ + +#define SCTP_INP_KILL_TIMEOUT 20 /* number of ms to retry kill of inpcb */ +#define SCTP_ASOC_KILL_TIMEOUT 10 /* number of ms to retry kill of inpcb */ + +#define SCTP_DEF_MAX_INIT 8 +#define SCTP_DEF_MAX_SEND 10 +#define SCTP_DEF_MAX_PATH_RTX 5 +#define SCTP_DEF_PATH_PF_THRESHOLD SCTP_DEF_MAX_PATH_RTX + +#define SCTP_DEF_PMTU_RAISE_SEC 600 /* 10 min between raise attempts */ + +/* How many streams I request initially by default */ +#define SCTP_OSTREAM_INITIAL 10 +#define SCTP_ISTREAM_INITIAL 2048 + +/* + * How many smallest_mtu's need to increase before a window update sack is + * sent (should be a power of 2). + */ +/* Send window update (incr * this > hiwat). Should be a power of 2 */ +#define SCTP_MINIMAL_RWND (4096) /* minimal rwnd */ + +#define SCTP_ADDRMAX 16 + +/* SCTP DEBUG Switch parameters */ +#define SCTP_DEBUG_TIMER1 0x00000001 +#define SCTP_DEBUG_TIMER2 0x00000002 /* unused */ +#define SCTP_DEBUG_TIMER3 0x00000004 /* unused */ +#define SCTP_DEBUG_TIMER4 0x00000008 +#define SCTP_DEBUG_OUTPUT1 0x00000010 +#define SCTP_DEBUG_OUTPUT2 0x00000020 +#define SCTP_DEBUG_OUTPUT3 0x00000040 +#define SCTP_DEBUG_OUTPUT4 0x00000080 +#define SCTP_DEBUG_UTIL1 0x00000100 +#define SCTP_DEBUG_UTIL2 0x00000200 /* unused */ +#define SCTP_DEBUG_AUTH1 0x00000400 +#define SCTP_DEBUG_AUTH2 0x00000800 /* unused */ +#define SCTP_DEBUG_INPUT1 0x00001000 +#define SCTP_DEBUG_INPUT2 0x00002000 +#define SCTP_DEBUG_INPUT3 0x00004000 +#define SCTP_DEBUG_INPUT4 0x00008000 /* unused */ +#define SCTP_DEBUG_ASCONF1 0x00010000 +#define SCTP_DEBUG_ASCONF2 0x00020000 +#define SCTP_DEBUG_OUTPUT5 0x00040000 /* unused */ +#define SCTP_DEBUG_XXX 0x00080000 /* unused */ +#define SCTP_DEBUG_PCB1 0x00100000 +#define SCTP_DEBUG_PCB2 0x00200000 /* unused */ +#define SCTP_DEBUG_PCB3 0x00400000 +#define SCTP_DEBUG_PCB4 0x00800000 +#define SCTP_DEBUG_INDATA1 0x01000000 +#define SCTP_DEBUG_INDATA2 0x02000000 /* unused */ +#define SCTP_DEBUG_INDATA3 0x04000000 /* unused */ +#define SCTP_DEBUG_CRCOFFLOAD 0x08000000 /* unused */ +#define SCTP_DEBUG_USRREQ1 0x10000000 /* unused */ +#define SCTP_DEBUG_USRREQ2 0x20000000 /* unused */ +#define SCTP_DEBUG_PEEL1 0x40000000 +#if defined(__Userspace__) +#define SCTP_DEBUG_USR 0x80000000 +#else +#define SCTP_DEBUG_XXXXX 0x80000000 /* unused */ +#endif +#define SCTP_DEBUG_ALL 0x7ff3ffff +#define SCTP_DEBUG_NOISY 0x00040000 + +/* What sender needs to see to avoid SWS or we consider peers rwnd 0 */ +#define SCTP_SWS_SENDER_DEF 1420 + +/* + * SWS is scaled to the sb_hiwat of the socket. A value of 2 is hiwat/4, 1 + * would be hiwat/2 etc. + */ +/* What receiver needs to see in sockbuf or we tell peer its 1 */ +#define SCTP_SWS_RECEIVER_DEF 3000 + +#define SCTP_INITIAL_CWND 4380 + +#define SCTP_DEFAULT_MTU 1500 /* emergency default MTU */ +/* amount peer is obligated to have in rwnd or I will abort */ +#define SCTP_MIN_RWND 1500 + +#define SCTP_CHUNK_BUFFER_SIZE 512 +#define SCTP_PARAM_BUFFER_SIZE 512 + +/* small chunk store for looking at chunk_list in auth */ +#define SCTP_SMALL_CHUNK_STORE 260 + +#define SCTP_HOW_MANY_SECRETS 2 /* how many secrets I keep */ + +#define SCTP_NUMBER_OF_SECRETS 8 /* or 8 * 4 = 32 octets */ +#define SCTP_SECRET_SIZE 32 /* number of octets in a 256 bits */ + +/* + * SCTP upper layer notifications + */ +#define SCTP_NOTIFY_ASSOC_UP 1 +#define SCTP_NOTIFY_ASSOC_DOWN 2 +#define SCTP_NOTIFY_INTERFACE_DOWN 3 +#define SCTP_NOTIFY_INTERFACE_UP 4 +#define SCTP_NOTIFY_SENT_DG_FAIL 5 +#define SCTP_NOTIFY_UNSENT_DG_FAIL 6 +#define SCTP_NOTIFY_SPECIAL_SP_FAIL 7 +#define SCTP_NOTIFY_ASSOC_LOC_ABORTED 8 +#define SCTP_NOTIFY_ASSOC_REM_ABORTED 9 +#define SCTP_NOTIFY_ASSOC_RESTART 10 +#define SCTP_NOTIFY_PEER_SHUTDOWN 11 +#define SCTP_NOTIFY_ASCONF_ADD_IP 12 +#define SCTP_NOTIFY_ASCONF_DELETE_IP 13 +#define SCTP_NOTIFY_ASCONF_SET_PRIMARY 14 +#define SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION 15 +#define SCTP_NOTIFY_INTERFACE_CONFIRMED 16 +#define SCTP_NOTIFY_STR_RESET_RECV 17 +#define SCTP_NOTIFY_STR_RESET_SEND 18 +#define SCTP_NOTIFY_STR_RESET_FAILED_OUT 19 +#define SCTP_NOTIFY_STR_RESET_FAILED_IN 20 +#define SCTP_NOTIFY_STR_RESET_DENIED_OUT 21 +#define SCTP_NOTIFY_STR_RESET_DENIED_IN 22 +#define SCTP_NOTIFY_AUTH_NEW_KEY 23 +#define SCTP_NOTIFY_AUTH_FREE_KEY 24 +#define SCTP_NOTIFY_NO_PEER_AUTH 25 +#define SCTP_NOTIFY_SENDER_DRY 26 +#define SCTP_NOTIFY_REMOTE_ERROR 27 +#define SCTP_NOTIFY_ASSOC_TIMEDOUT 28 + +/* This is the value for messages that are NOT completely + * copied down where we will start to split the message. + * So, with our default, we split only if the piece we + * want to take will fill up a full MTU (assuming + * a 1500 byte MTU). + */ +#define SCTP_DEFAULT_SPLIT_POINT_MIN 2904 + +/* Maximum length of diagnostic information in error causes */ +#if defined(__Userspace__) +#define SCTP_DIAG_INFO_LEN 256 +#else +#define SCTP_DIAG_INFO_LEN 128 +#endif + +/* ABORT CODES and other tell-tale location + * codes are generated by adding the below + * to the instance id. + */ + +/* File defines */ +#define SCTP_FROM_SCTP_INPUT 0x10000000 +#define SCTP_FROM_SCTP_PCB 0x20000000 +#define SCTP_FROM_SCTP_INDATA 0x30000000 +#define SCTP_FROM_SCTP_TIMER 0x40000000 +#define SCTP_FROM_SCTP_USRREQ 0x50000000 +#define SCTP_FROM_SCTPUTIL 0x60000000 +#define SCTP_FROM_SCTP6_USRREQ 0x70000000 +#define SCTP_FROM_SCTP_ASCONF 0x80000000 +#define SCTP_FROM_SCTP_OUTPUT 0x90000000 +#define SCTP_FROM_SCTP_PEELOFF 0xa0000000 +#define SCTP_FROM_SCTP_SYSCTL 0xb0000000 +#define SCTP_FROM_SCTP_CC_FUNCTIONS 0xc0000000 + +/* Location ID's */ +#define SCTP_LOC_1 0x00000001 +#define SCTP_LOC_2 0x00000002 +#define SCTP_LOC_3 0x00000003 +#define SCTP_LOC_4 0x00000004 +#define SCTP_LOC_5 0x00000005 +#define SCTP_LOC_6 0x00000006 +#define SCTP_LOC_7 0x00000007 +#define SCTP_LOC_8 0x00000008 +#define SCTP_LOC_9 0x00000009 +#define SCTP_LOC_10 0x0000000a +#define SCTP_LOC_11 0x0000000b +#define SCTP_LOC_12 0x0000000c +#define SCTP_LOC_13 0x0000000d +#define SCTP_LOC_14 0x0000000e +#define SCTP_LOC_15 0x0000000f +#define SCTP_LOC_16 0x00000010 +#define SCTP_LOC_17 0x00000011 +#define SCTP_LOC_18 0x00000012 +#define SCTP_LOC_19 0x00000013 +#define SCTP_LOC_20 0x00000014 +#define SCTP_LOC_21 0x00000015 +#define SCTP_LOC_22 0x00000016 +#define SCTP_LOC_23 0x00000017 +#define SCTP_LOC_24 0x00000018 +#define SCTP_LOC_25 0x00000019 +#define SCTP_LOC_26 0x0000001a +#define SCTP_LOC_27 0x0000001b +#define SCTP_LOC_28 0x0000001c +#define SCTP_LOC_29 0x0000001d +#define SCTP_LOC_30 0x0000001e +#define SCTP_LOC_31 0x0000001f +#define SCTP_LOC_32 0x00000020 +#define SCTP_LOC_33 0x00000021 +#define SCTP_LOC_34 0x00000022 +#define SCTP_LOC_35 0x00000023 +#define SCTP_LOC_36 0x00000024 +#define SCTP_LOC_37 0x00000025 + +/* Free assoc codes */ +#define SCTP_NORMAL_PROC 0 +#define SCTP_PCBFREE_NOFORCE 1 +#define SCTP_PCBFREE_FORCE 2 + +/* From codes for adding addresses */ +#define SCTP_ADDR_IS_CONFIRMED 8 +#define SCTP_ADDR_DYNAMIC_ADDED 6 +#define SCTP_IN_COOKIE_PROC 100 +#define SCTP_ALLOC_ASOC 1 +#define SCTP_LOAD_ADDR_2 2 +#define SCTP_LOAD_ADDR_3 3 +#define SCTP_LOAD_ADDR_4 4 +#define SCTP_LOAD_ADDR_5 5 + +#define SCTP_DONOT_SETSCOPE 0 +#define SCTP_DO_SETSCOPE 1 + +/* This value determines the default for when + * we try to add more on the send queue., if + * there is room. This prevents us from cycling + * into the copy_resume routine to often if + * we have not got enough space to add a decent + * enough size message. Note that if we have enough + * space to complete the message copy we will always + * add to the message, no matter what the size. Its + * only when we reach the point that we have some left + * to add, there is only room for part of it that we + * will use this threshold. Its also a sysctl. + */ +#define SCTP_DEFAULT_ADD_MORE 1452 + +#ifndef SCTP_PCBHASHSIZE +/* default number of association hash buckets in each endpoint */ +#define SCTP_PCBHASHSIZE 256 +#endif +#ifndef SCTP_TCBHASHSIZE +#define SCTP_TCBHASHSIZE 1024 +#endif + +#ifndef SCTP_CHUNKQUEUE_SCALE +#define SCTP_CHUNKQUEUE_SCALE 10 +#endif + +#if defined(__FreeBSD__) && !defined(__Userspace__) +/* clock variance is 1 ms */ +#define SCTP_CLOCK_GRANULARITY 1 +#else +/* clock variance is 10 ms */ +#define SCTP_CLOCK_GRANULARITY 10 +#endif +#define IP_HDR_SIZE 40 /* we use the size of a IP6 header here this + * detracts a small amount for ipv4 but it + * simplifies the ipv6 addition */ + +/* Argument magic number for sctp_inpcb_free() */ + +/* third argument */ +#define SCTP_CALLED_DIRECTLY_NOCMPSET 0 +#define SCTP_CALLED_AFTER_CMPSET_OFCLOSE 1 +#define SCTP_CALLED_FROM_INPKILL_TIMER 2 +/* second argument */ +#define SCTP_FREE_SHOULD_USE_ABORT 1 +#define SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE 0 + +#ifndef IPPROTO_SCTP +#define IPPROTO_SCTP 132 /* the Official IANA number :-) */ +#endif /* !IPPROTO_SCTP */ + +#define SCTP_MAX_DATA_BUNDLING 256 + +/* modular comparison */ +/* See RFC 1982 for details. */ +#define SCTP_UINT16_GT(a, b) (((a < b) && ((uint16_t)(b - a) > (1U<<15))) || \ + ((a > b) && ((uint16_t)(a - b) < (1U<<15)))) +#define SCTP_UINT16_GE(a, b) (SCTP_UINT16_GT(a, b) || (a == b)) +#define SCTP_UINT32_GT(a, b) (((a < b) && ((uint32_t)(b - a) > (1U<<31))) || \ + ((a > b) && ((uint32_t)(a - b) < (1U<<31)))) +#define SCTP_UINT32_GE(a, b) (SCTP_UINT32_GT(a, b) || (a == b)) + +#define SCTP_SSN_GT(a, b) SCTP_UINT16_GT(a, b) +#define SCTP_SSN_GE(a, b) SCTP_UINT16_GE(a, b) +#define SCTP_TSN_GT(a, b) SCTP_UINT32_GT(a, b) +#define SCTP_TSN_GE(a, b) SCTP_UINT32_GE(a, b) +#define SCTP_MID_GT(i, a, b) (((i) == 1) ? SCTP_UINT32_GT(a, b) : SCTP_UINT16_GT((uint16_t)a, (uint16_t)b)) +#define SCTP_MID_GE(i, a, b) (((i) == 1) ? SCTP_UINT32_GE(a, b) : SCTP_UINT16_GE((uint16_t)a, (uint16_t)b)) +#define SCTP_MID_EQ(i, a, b) (((i) == 1) ? a == b : (uint16_t)a == (uint16_t)b) + +/* Mapping array manipulation routines */ +#define SCTP_IS_TSN_PRESENT(arry, gap) ((arry[(gap >> 3)] >> (gap & 0x07)) & 0x01) +#define SCTP_SET_TSN_PRESENT(arry, gap) (arry[(gap >> 3)] |= (0x01 << ((gap & 0x07)))) +#define SCTP_UNSET_TSN_PRESENT(arry, gap) (arry[(gap >> 3)] &= ((~(0x01 << ((gap & 0x07)))) & 0xff)) +#define SCTP_CALC_TSN_TO_GAP(gap, tsn, mapping_tsn) do { \ + if (tsn >= mapping_tsn) { \ + gap = tsn - mapping_tsn; \ + } else { \ + gap = (MAX_TSN - mapping_tsn) + tsn + 1; \ + } \ + } while (0) + +#define SCTP_RETRAN_DONE -1 +#define SCTP_RETRAN_EXIT -2 + +/* + * This value defines the number of vtag block time wait entry's per list + * element. Each entry will take 2 4 byte ints (and of course the overhead + * of the next pointer as well). Using 15 as an example will yield * ((8 * + * 15) + 8) or 128 bytes of overhead for each timewait block that gets + * initialized. Increasing it to 31 would yield 256 bytes per block. + */ +#define SCTP_NUMBER_IN_VTAG_BLOCK 15 +/* + * If we use the STACK option, we have an array of this size head pointers. + * This array is mod'd the with the size to find which bucket and then all + * entries must be searched to see if the tag is in timed wait. If so we + * reject it. + */ +#define SCTP_STACK_VTAG_HASH_SIZE 32 + +/* + * Number of seconds of time wait for a vtag. + */ +#define SCTP_TIME_WAIT 60 + +/* How many micro seconds is the cutoff from + * local lan type rtt's + */ + /* + * We allow 900us for the rtt. + */ +#define SCTP_LOCAL_LAN_RTT 900 +#define SCTP_LAN_UNKNOWN 0 +#define SCTP_LAN_LOCAL 1 +#define SCTP_LAN_INTERNET 2 + +#define SCTP_SEND_BUFFER_SPLITTING 0x00000001 +#define SCTP_RECV_BUFFER_SPLITTING 0x00000002 + +/* The system retains a cache of free chunks such to + * cut down on calls the memory allocation system. There + * is a per association limit of free items and a overall + * system limit. If either one gets hit then the resource + * stops being cached. + */ + +#define SCTP_DEF_ASOC_RESC_LIMIT 10 +#define SCTP_DEF_SYSTEM_RESC_LIMIT 1000 + +/*- + * defines for socket lock states. + * Used by __APPLE__ + */ +#define SCTP_SO_LOCKED 1 +#define SCTP_SO_NOT_LOCKED 0 + +/*- + * For address locks, do we hold the lock? + */ +#define SCTP_ADDR_LOCKED 1 +#define SCTP_ADDR_NOT_LOCKED 0 + +#define IN4_ISPRIVATE_ADDRESS(a) \ + ((((uint8_t *)&(a)->s_addr)[0] == 10) || \ + ((((uint8_t *)&(a)->s_addr)[0] == 172) && \ + (((uint8_t *)&(a)->s_addr)[1] >= 16) && \ + (((uint8_t *)&(a)->s_addr)[1] <= 32)) || \ + ((((uint8_t *)&(a)->s_addr)[0] == 192) && \ + (((uint8_t *)&(a)->s_addr)[1] == 168))) + +#define IN4_ISLOOPBACK_ADDRESS(a) \ + (((uint8_t *)&(a)->s_addr)[0] == 127) + +#define IN4_ISLINKLOCAL_ADDRESS(a) \ + ((((uint8_t *)&(a)->s_addr)[0] == 169) && \ + (((uint8_t *)&(a)->s_addr)[1] == 254)) + +/* Maximum size of optval for IPPROTO_SCTP level socket options. */ +#define SCTP_SOCKET_OPTION_LIMIT (64 * 1024) + +#if defined(__Userspace__) +#if defined(_WIN32) +#define SCTP_GETTIME_TIMEVAL(x) getwintimeofday(x) +#define SCTP_GETPTIME_TIMEVAL(x) getwintimeofday(x) /* this doesn't seem to ever be used.. */ +#else +#define SCTP_GETTIME_TIMEVAL(x) gettimeofday(x, NULL) +#define SCTP_GETPTIME_TIMEVAL(x) gettimeofday(x, NULL) +#endif +#endif +#if defined(_KERNEL) +#define SCTP_GETTIME_TIMEVAL(x) (getmicrouptime(x)) +#define SCTP_GETPTIME_TIMEVAL(x) (microuptime(x)) +#endif + +#if defined(_KERNEL) || defined(__Userspace__) +#define sctp_sowwakeup(inp, so) \ +do { \ + if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { \ + sctp_pcb_add_flags(inp, SCTP_PCB_FLAGS_WAKEOUTPUT); \ + } else { \ + sowwakeup(so); \ + } \ +} while (0) + +#if defined(__FreeBSD__) || defined(_WIN32) || defined(__Userspace__) +#define sctp_sowwakeup_locked(inp, so) \ +do { \ + if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { \ + sctp_pcb_add_flags(inp, SCTP_PCB_FLAGS_WAKEOUTPUT); \ + SOCKBUF_UNLOCK(&((so)->so_snd)); \ + } else { \ + sowwakeup_locked(so); \ + } \ +} while (0) +#else +#define sctp_sowwakeup_locked(inp, so) \ +do { \ + if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { \ + sctp_pcb_add_flags(inp, SCTP_PCB_FLAGS_WAKEOUTPUT); \ + SOCKBUF_UNLOCK(&((so)->so_snd)); \ + } else { \ + sowwakeup(so); \ + } \ +} while (0) +#endif + +#define sctp_sorwakeup(inp, so) \ +do { \ + if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { \ + sctp_pcb_add_flags(inp, SCTP_PCB_FLAGS_WAKEINPUT); \ + } else { \ + sorwakeup(so); \ + } \ +} while (0) + +#if defined(__FreeBSD__) || defined(_WIN32) || defined(__Userspace__) +#define sctp_sorwakeup_locked(inp, so) \ +do { \ + if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { \ + sctp_pcb_add_flags(inp, SCTP_PCB_FLAGS_WAKEINPUT); \ + SOCKBUF_UNLOCK(&((so)->so_rcv)); \ + } else { \ + sorwakeup_locked(so); \ + } \ +} while (0) +#else + +#define sctp_sorwakeup_locked(inp, so) \ +do { \ + if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { \ + sctp_pcb_add_flags(inp, SCTP_PCB_FLAGS_WAKEINPUT); \ + SOCKBUF_UNLOCK(&((so)->so_rcv)); \ + } else { \ + sorwakeup(so); \ + } \ +} while (0) +#endif + +#endif /* _KERNEL || __Userspace__*/ +#endif diff --git a/netwerk/sctp/src/netinet/sctp_crc32.c b/netwerk/sctp/src/netinet/sctp_crc32.c new file mode 100644 index 0000000000..51b4e49d2e --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_crc32.c @@ -0,0 +1,821 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); + +#include "opt_sctp.h" + +#include <sys/param.h> +#include <sys/systm.h> +#include <sys/gsb_crc32.h> +#include <sys/mbuf.h> + +#include <netinet/sctp.h> +#include <netinet/sctp_crc32.h> +#if defined(SCTP) || defined(SCTP_SUPPORT) +#include <netinet/sctp_os.h> +#include <netinet/sctp_pcb.h> +#endif +#else +#include <netinet/sctp_os.h> +#include <netinet/sctp.h> +#include <netinet/sctp_crc32.h> +#include <netinet/sctp_pcb.h> +#endif + +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) +/** + * + * Routine Description: + * + * Computes the CRC32c checksum for the specified buffer using the slicing by 8 + * algorithm over 64 bit quantities. + * + * Arguments: + * + * p_running_crc - pointer to the initial or final remainder value + * used in CRC computations. It should be set to + * non-NULL if the mode argument is equal to CONT or END + * p_buf - the packet buffer where crc computations are being performed + * length - the length of p_buf in bytes + * init_bytes - the number of initial bytes that need to be procesed before + * aligning p_buf to multiples of 4 bytes + * mode - can be any of the following: BEGIN, CONT, END, BODY, ALIGN + * + * Return value: + * + * The computed CRC32c value + */ + + +/* + * Copyright (c) 2004-2006 Intel Corporation - All Rights Reserved + * + * + * This software program is licensed subject to the BSD License, available at + * http://www.opensource.org/licenses/bsd-license.html. + * + * Abstract: + * + * Tables for software CRC generation + */ + +/* + * The following CRC lookup table was generated automagically using the + * following model parameters: + * + * Generator Polynomial = ................. 0x1EDC6F41 + * Generator Polynomial Length = .......... 32 bits + * Reflected Bits = ....................... TRUE + * Table Generation Offset = .............. 32 bits + * Number of Slices = ..................... 8 slices + * Slice Lengths = ........................ 8 8 8 8 8 8 8 8 + * Directory Name = ....................... .\ + * File Name = ............................ 8x256_tables.c + */ + +static const uint32_t sctp_crc_tableil8_o32[256] = +{ + 0x00000000, 0xF26B8303, 0xE13B70F7, 0x1350F3F4, 0xC79A971F, 0x35F1141C, 0x26A1E7E8, 0xD4CA64EB, + 0x8AD958CF, 0x78B2DBCC, 0x6BE22838, 0x9989AB3B, 0x4D43CFD0, 0xBF284CD3, 0xAC78BF27, 0x5E133C24, + 0x105EC76F, 0xE235446C, 0xF165B798, 0x030E349B, 0xD7C45070, 0x25AFD373, 0x36FF2087, 0xC494A384, + 0x9A879FA0, 0x68EC1CA3, 0x7BBCEF57, 0x89D76C54, 0x5D1D08BF, 0xAF768BBC, 0xBC267848, 0x4E4DFB4B, + 0x20BD8EDE, 0xD2D60DDD, 0xC186FE29, 0x33ED7D2A, 0xE72719C1, 0x154C9AC2, 0x061C6936, 0xF477EA35, + 0xAA64D611, 0x580F5512, 0x4B5FA6E6, 0xB93425E5, 0x6DFE410E, 0x9F95C20D, 0x8CC531F9, 0x7EAEB2FA, + 0x30E349B1, 0xC288CAB2, 0xD1D83946, 0x23B3BA45, 0xF779DEAE, 0x05125DAD, 0x1642AE59, 0xE4292D5A, + 0xBA3A117E, 0x4851927D, 0x5B016189, 0xA96AE28A, 0x7DA08661, 0x8FCB0562, 0x9C9BF696, 0x6EF07595, + 0x417B1DBC, 0xB3109EBF, 0xA0406D4B, 0x522BEE48, 0x86E18AA3, 0x748A09A0, 0x67DAFA54, 0x95B17957, + 0xCBA24573, 0x39C9C670, 0x2A993584, 0xD8F2B687, 0x0C38D26C, 0xFE53516F, 0xED03A29B, 0x1F682198, + 0x5125DAD3, 0xA34E59D0, 0xB01EAA24, 0x42752927, 0x96BF4DCC, 0x64D4CECF, 0x77843D3B, 0x85EFBE38, + 0xDBFC821C, 0x2997011F, 0x3AC7F2EB, 0xC8AC71E8, 0x1C661503, 0xEE0D9600, 0xFD5D65F4, 0x0F36E6F7, + 0x61C69362, 0x93AD1061, 0x80FDE395, 0x72966096, 0xA65C047D, 0x5437877E, 0x4767748A, 0xB50CF789, + 0xEB1FCBAD, 0x197448AE, 0x0A24BB5A, 0xF84F3859, 0x2C855CB2, 0xDEEEDFB1, 0xCDBE2C45, 0x3FD5AF46, + 0x7198540D, 0x83F3D70E, 0x90A324FA, 0x62C8A7F9, 0xB602C312, 0x44694011, 0x5739B3E5, 0xA55230E6, + 0xFB410CC2, 0x092A8FC1, 0x1A7A7C35, 0xE811FF36, 0x3CDB9BDD, 0xCEB018DE, 0xDDE0EB2A, 0x2F8B6829, + 0x82F63B78, 0x709DB87B, 0x63CD4B8F, 0x91A6C88C, 0x456CAC67, 0xB7072F64, 0xA457DC90, 0x563C5F93, + 0x082F63B7, 0xFA44E0B4, 0xE9141340, 0x1B7F9043, 0xCFB5F4A8, 0x3DDE77AB, 0x2E8E845F, 0xDCE5075C, + 0x92A8FC17, 0x60C37F14, 0x73938CE0, 0x81F80FE3, 0x55326B08, 0xA759E80B, 0xB4091BFF, 0x466298FC, + 0x1871A4D8, 0xEA1A27DB, 0xF94AD42F, 0x0B21572C, 0xDFEB33C7, 0x2D80B0C4, 0x3ED04330, 0xCCBBC033, + 0xA24BB5A6, 0x502036A5, 0x4370C551, 0xB11B4652, 0x65D122B9, 0x97BAA1BA, 0x84EA524E, 0x7681D14D, + 0x2892ED69, 0xDAF96E6A, 0xC9A99D9E, 0x3BC21E9D, 0xEF087A76, 0x1D63F975, 0x0E330A81, 0xFC588982, + 0xB21572C9, 0x407EF1CA, 0x532E023E, 0xA145813D, 0x758FE5D6, 0x87E466D5, 0x94B49521, 0x66DF1622, + 0x38CC2A06, 0xCAA7A905, 0xD9F75AF1, 0x2B9CD9F2, 0xFF56BD19, 0x0D3D3E1A, 0x1E6DCDEE, 0xEC064EED, + 0xC38D26C4, 0x31E6A5C7, 0x22B65633, 0xD0DDD530, 0x0417B1DB, 0xF67C32D8, 0xE52CC12C, 0x1747422F, + 0x49547E0B, 0xBB3FFD08, 0xA86F0EFC, 0x5A048DFF, 0x8ECEE914, 0x7CA56A17, 0x6FF599E3, 0x9D9E1AE0, + 0xD3D3E1AB, 0x21B862A8, 0x32E8915C, 0xC083125F, 0x144976B4, 0xE622F5B7, 0xF5720643, 0x07198540, + 0x590AB964, 0xAB613A67, 0xB831C993, 0x4A5A4A90, 0x9E902E7B, 0x6CFBAD78, 0x7FAB5E8C, 0x8DC0DD8F, + 0xE330A81A, 0x115B2B19, 0x020BD8ED, 0xF0605BEE, 0x24AA3F05, 0xD6C1BC06, 0xC5914FF2, 0x37FACCF1, + 0x69E9F0D5, 0x9B8273D6, 0x88D28022, 0x7AB90321, 0xAE7367CA, 0x5C18E4C9, 0x4F48173D, 0xBD23943E, + 0xF36E6F75, 0x0105EC76, 0x12551F82, 0xE03E9C81, 0x34F4F86A, 0xC69F7B69, 0xD5CF889D, 0x27A40B9E, + 0x79B737BA, 0x8BDCB4B9, 0x988C474D, 0x6AE7C44E, 0xBE2DA0A5, 0x4C4623A6, 0x5F16D052, 0xAD7D5351 +}; + +/* + * end of the CRC lookup table crc_tableil8_o32 + */ + + + +/* + * The following CRC lookup table was generated automagically using the + * following model parameters: + * + * Generator Polynomial = ................. 0x1EDC6F41 + * Generator Polynomial Length = .......... 32 bits + * Reflected Bits = ....................... TRUE + * Table Generation Offset = .............. 32 bits + * Number of Slices = ..................... 8 slices + * Slice Lengths = ........................ 8 8 8 8 8 8 8 8 + * Directory Name = ....................... .\ + * File Name = ............................ 8x256_tables.c + */ + +static const uint32_t sctp_crc_tableil8_o40[256] = +{ + 0x00000000, 0x13A29877, 0x274530EE, 0x34E7A899, 0x4E8A61DC, 0x5D28F9AB, 0x69CF5132, 0x7A6DC945, + 0x9D14C3B8, 0x8EB65BCF, 0xBA51F356, 0xA9F36B21, 0xD39EA264, 0xC03C3A13, 0xF4DB928A, 0xE7790AFD, + 0x3FC5F181, 0x2C6769F6, 0x1880C16F, 0x0B225918, 0x714F905D, 0x62ED082A, 0x560AA0B3, 0x45A838C4, + 0xA2D13239, 0xB173AA4E, 0x859402D7, 0x96369AA0, 0xEC5B53E5, 0xFFF9CB92, 0xCB1E630B, 0xD8BCFB7C, + 0x7F8BE302, 0x6C297B75, 0x58CED3EC, 0x4B6C4B9B, 0x310182DE, 0x22A31AA9, 0x1644B230, 0x05E62A47, + 0xE29F20BA, 0xF13DB8CD, 0xC5DA1054, 0xD6788823, 0xAC154166, 0xBFB7D911, 0x8B507188, 0x98F2E9FF, + 0x404E1283, 0x53EC8AF4, 0x670B226D, 0x74A9BA1A, 0x0EC4735F, 0x1D66EB28, 0x298143B1, 0x3A23DBC6, + 0xDD5AD13B, 0xCEF8494C, 0xFA1FE1D5, 0xE9BD79A2, 0x93D0B0E7, 0x80722890, 0xB4958009, 0xA737187E, + 0xFF17C604, 0xECB55E73, 0xD852F6EA, 0xCBF06E9D, 0xB19DA7D8, 0xA23F3FAF, 0x96D89736, 0x857A0F41, + 0x620305BC, 0x71A19DCB, 0x45463552, 0x56E4AD25, 0x2C896460, 0x3F2BFC17, 0x0BCC548E, 0x186ECCF9, + 0xC0D23785, 0xD370AFF2, 0xE797076B, 0xF4359F1C, 0x8E585659, 0x9DFACE2E, 0xA91D66B7, 0xBABFFEC0, + 0x5DC6F43D, 0x4E646C4A, 0x7A83C4D3, 0x69215CA4, 0x134C95E1, 0x00EE0D96, 0x3409A50F, 0x27AB3D78, + 0x809C2506, 0x933EBD71, 0xA7D915E8, 0xB47B8D9F, 0xCE1644DA, 0xDDB4DCAD, 0xE9537434, 0xFAF1EC43, + 0x1D88E6BE, 0x0E2A7EC9, 0x3ACDD650, 0x296F4E27, 0x53028762, 0x40A01F15, 0x7447B78C, 0x67E52FFB, + 0xBF59D487, 0xACFB4CF0, 0x981CE469, 0x8BBE7C1E, 0xF1D3B55B, 0xE2712D2C, 0xD69685B5, 0xC5341DC2, + 0x224D173F, 0x31EF8F48, 0x050827D1, 0x16AABFA6, 0x6CC776E3, 0x7F65EE94, 0x4B82460D, 0x5820DE7A, + 0xFBC3FAF9, 0xE861628E, 0xDC86CA17, 0xCF245260, 0xB5499B25, 0xA6EB0352, 0x920CABCB, 0x81AE33BC, + 0x66D73941, 0x7575A136, 0x419209AF, 0x523091D8, 0x285D589D, 0x3BFFC0EA, 0x0F186873, 0x1CBAF004, + 0xC4060B78, 0xD7A4930F, 0xE3433B96, 0xF0E1A3E1, 0x8A8C6AA4, 0x992EF2D3, 0xADC95A4A, 0xBE6BC23D, + 0x5912C8C0, 0x4AB050B7, 0x7E57F82E, 0x6DF56059, 0x1798A91C, 0x043A316B, 0x30DD99F2, 0x237F0185, + 0x844819FB, 0x97EA818C, 0xA30D2915, 0xB0AFB162, 0xCAC27827, 0xD960E050, 0xED8748C9, 0xFE25D0BE, + 0x195CDA43, 0x0AFE4234, 0x3E19EAAD, 0x2DBB72DA, 0x57D6BB9F, 0x447423E8, 0x70938B71, 0x63311306, + 0xBB8DE87A, 0xA82F700D, 0x9CC8D894, 0x8F6A40E3, 0xF50789A6, 0xE6A511D1, 0xD242B948, 0xC1E0213F, + 0x26992BC2, 0x353BB3B5, 0x01DC1B2C, 0x127E835B, 0x68134A1E, 0x7BB1D269, 0x4F567AF0, 0x5CF4E287, + 0x04D43CFD, 0x1776A48A, 0x23910C13, 0x30339464, 0x4A5E5D21, 0x59FCC556, 0x6D1B6DCF, 0x7EB9F5B8, + 0x99C0FF45, 0x8A626732, 0xBE85CFAB, 0xAD2757DC, 0xD74A9E99, 0xC4E806EE, 0xF00FAE77, 0xE3AD3600, + 0x3B11CD7C, 0x28B3550B, 0x1C54FD92, 0x0FF665E5, 0x759BACA0, 0x663934D7, 0x52DE9C4E, 0x417C0439, + 0xA6050EC4, 0xB5A796B3, 0x81403E2A, 0x92E2A65D, 0xE88F6F18, 0xFB2DF76F, 0xCFCA5FF6, 0xDC68C781, + 0x7B5FDFFF, 0x68FD4788, 0x5C1AEF11, 0x4FB87766, 0x35D5BE23, 0x26772654, 0x12908ECD, 0x013216BA, + 0xE64B1C47, 0xF5E98430, 0xC10E2CA9, 0xD2ACB4DE, 0xA8C17D9B, 0xBB63E5EC, 0x8F844D75, 0x9C26D502, + 0x449A2E7E, 0x5738B609, 0x63DF1E90, 0x707D86E7, 0x0A104FA2, 0x19B2D7D5, 0x2D557F4C, 0x3EF7E73B, + 0xD98EEDC6, 0xCA2C75B1, 0xFECBDD28, 0xED69455F, 0x97048C1A, 0x84A6146D, 0xB041BCF4, 0xA3E32483 +}; + +/* + * end of the CRC lookup table crc_tableil8_o40 + */ + + + +/* + * The following CRC lookup table was generated automagically using the + * following model parameters: + * + * Generator Polynomial = ................. 0x1EDC6F41 + * Generator Polynomial Length = .......... 32 bits + * Reflected Bits = ....................... TRUE + * Table Generation Offset = .............. 32 bits + * Number of Slices = ..................... 8 slices + * Slice Lengths = ........................ 8 8 8 8 8 8 8 8 + * Directory Name = ....................... .\ + * File Name = ............................ 8x256_tables.c + */ + +static const uint32_t sctp_crc_tableil8_o48[256] = +{ + 0x00000000, 0xA541927E, 0x4F6F520D, 0xEA2EC073, 0x9EDEA41A, 0x3B9F3664, 0xD1B1F617, 0x74F06469, + 0x38513EC5, 0x9D10ACBB, 0x773E6CC8, 0xD27FFEB6, 0xA68F9ADF, 0x03CE08A1, 0xE9E0C8D2, 0x4CA15AAC, + 0x70A27D8A, 0xD5E3EFF4, 0x3FCD2F87, 0x9A8CBDF9, 0xEE7CD990, 0x4B3D4BEE, 0xA1138B9D, 0x045219E3, + 0x48F3434F, 0xEDB2D131, 0x079C1142, 0xA2DD833C, 0xD62DE755, 0x736C752B, 0x9942B558, 0x3C032726, + 0xE144FB14, 0x4405696A, 0xAE2BA919, 0x0B6A3B67, 0x7F9A5F0E, 0xDADBCD70, 0x30F50D03, 0x95B49F7D, + 0xD915C5D1, 0x7C5457AF, 0x967A97DC, 0x333B05A2, 0x47CB61CB, 0xE28AF3B5, 0x08A433C6, 0xADE5A1B8, + 0x91E6869E, 0x34A714E0, 0xDE89D493, 0x7BC846ED, 0x0F382284, 0xAA79B0FA, 0x40577089, 0xE516E2F7, + 0xA9B7B85B, 0x0CF62A25, 0xE6D8EA56, 0x43997828, 0x37691C41, 0x92288E3F, 0x78064E4C, 0xDD47DC32, + 0xC76580D9, 0x622412A7, 0x880AD2D4, 0x2D4B40AA, 0x59BB24C3, 0xFCFAB6BD, 0x16D476CE, 0xB395E4B0, + 0xFF34BE1C, 0x5A752C62, 0xB05BEC11, 0x151A7E6F, 0x61EA1A06, 0xC4AB8878, 0x2E85480B, 0x8BC4DA75, + 0xB7C7FD53, 0x12866F2D, 0xF8A8AF5E, 0x5DE93D20, 0x29195949, 0x8C58CB37, 0x66760B44, 0xC337993A, + 0x8F96C396, 0x2AD751E8, 0xC0F9919B, 0x65B803E5, 0x1148678C, 0xB409F5F2, 0x5E273581, 0xFB66A7FF, + 0x26217BCD, 0x8360E9B3, 0x694E29C0, 0xCC0FBBBE, 0xB8FFDFD7, 0x1DBE4DA9, 0xF7908DDA, 0x52D11FA4, + 0x1E704508, 0xBB31D776, 0x511F1705, 0xF45E857B, 0x80AEE112, 0x25EF736C, 0xCFC1B31F, 0x6A802161, + 0x56830647, 0xF3C29439, 0x19EC544A, 0xBCADC634, 0xC85DA25D, 0x6D1C3023, 0x8732F050, 0x2273622E, + 0x6ED23882, 0xCB93AAFC, 0x21BD6A8F, 0x84FCF8F1, 0xF00C9C98, 0x554D0EE6, 0xBF63CE95, 0x1A225CEB, + 0x8B277743, 0x2E66E53D, 0xC448254E, 0x6109B730, 0x15F9D359, 0xB0B84127, 0x5A968154, 0xFFD7132A, + 0xB3764986, 0x1637DBF8, 0xFC191B8B, 0x595889F5, 0x2DA8ED9C, 0x88E97FE2, 0x62C7BF91, 0xC7862DEF, + 0xFB850AC9, 0x5EC498B7, 0xB4EA58C4, 0x11ABCABA, 0x655BAED3, 0xC01A3CAD, 0x2A34FCDE, 0x8F756EA0, + 0xC3D4340C, 0x6695A672, 0x8CBB6601, 0x29FAF47F, 0x5D0A9016, 0xF84B0268, 0x1265C21B, 0xB7245065, + 0x6A638C57, 0xCF221E29, 0x250CDE5A, 0x804D4C24, 0xF4BD284D, 0x51FCBA33, 0xBBD27A40, 0x1E93E83E, + 0x5232B292, 0xF77320EC, 0x1D5DE09F, 0xB81C72E1, 0xCCEC1688, 0x69AD84F6, 0x83834485, 0x26C2D6FB, + 0x1AC1F1DD, 0xBF8063A3, 0x55AEA3D0, 0xF0EF31AE, 0x841F55C7, 0x215EC7B9, 0xCB7007CA, 0x6E3195B4, + 0x2290CF18, 0x87D15D66, 0x6DFF9D15, 0xC8BE0F6B, 0xBC4E6B02, 0x190FF97C, 0xF321390F, 0x5660AB71, + 0x4C42F79A, 0xE90365E4, 0x032DA597, 0xA66C37E9, 0xD29C5380, 0x77DDC1FE, 0x9DF3018D, 0x38B293F3, + 0x7413C95F, 0xD1525B21, 0x3B7C9B52, 0x9E3D092C, 0xEACD6D45, 0x4F8CFF3B, 0xA5A23F48, 0x00E3AD36, + 0x3CE08A10, 0x99A1186E, 0x738FD81D, 0xD6CE4A63, 0xA23E2E0A, 0x077FBC74, 0xED517C07, 0x4810EE79, + 0x04B1B4D5, 0xA1F026AB, 0x4BDEE6D8, 0xEE9F74A6, 0x9A6F10CF, 0x3F2E82B1, 0xD50042C2, 0x7041D0BC, + 0xAD060C8E, 0x08479EF0, 0xE2695E83, 0x4728CCFD, 0x33D8A894, 0x96993AEA, 0x7CB7FA99, 0xD9F668E7, + 0x9557324B, 0x3016A035, 0xDA386046, 0x7F79F238, 0x0B899651, 0xAEC8042F, 0x44E6C45C, 0xE1A75622, + 0xDDA47104, 0x78E5E37A, 0x92CB2309, 0x378AB177, 0x437AD51E, 0xE63B4760, 0x0C158713, 0xA954156D, + 0xE5F54FC1, 0x40B4DDBF, 0xAA9A1DCC, 0x0FDB8FB2, 0x7B2BEBDB, 0xDE6A79A5, 0x3444B9D6, 0x91052BA8 +}; + +/* + * end of the CRC lookup table crc_tableil8_o48 + */ + + + +/* + * The following CRC lookup table was generated automagically using the + * following model parameters: + * + * Generator Polynomial = ................. 0x1EDC6F41 + * Generator Polynomial Length = .......... 32 bits + * Reflected Bits = ....................... TRUE + * Table Generation Offset = .............. 32 bits + * Number of Slices = ..................... 8 slices + * Slice Lengths = ........................ 8 8 8 8 8 8 8 8 + * Directory Name = ....................... .\ + * File Name = ............................ 8x256_tables.c + */ + +static const uint32_t sctp_crc_tableil8_o56[256] = +{ + 0x00000000, 0xDD45AAB8, 0xBF672381, 0x62228939, 0x7B2231F3, 0xA6679B4B, 0xC4451272, 0x1900B8CA, + 0xF64463E6, 0x2B01C95E, 0x49234067, 0x9466EADF, 0x8D665215, 0x5023F8AD, 0x32017194, 0xEF44DB2C, + 0xE964B13D, 0x34211B85, 0x560392BC, 0x8B463804, 0x924680CE, 0x4F032A76, 0x2D21A34F, 0xF06409F7, + 0x1F20D2DB, 0xC2657863, 0xA047F15A, 0x7D025BE2, 0x6402E328, 0xB9474990, 0xDB65C0A9, 0x06206A11, + 0xD725148B, 0x0A60BE33, 0x6842370A, 0xB5079DB2, 0xAC072578, 0x71428FC0, 0x136006F9, 0xCE25AC41, + 0x2161776D, 0xFC24DDD5, 0x9E0654EC, 0x4343FE54, 0x5A43469E, 0x8706EC26, 0xE524651F, 0x3861CFA7, + 0x3E41A5B6, 0xE3040F0E, 0x81268637, 0x5C632C8F, 0x45639445, 0x98263EFD, 0xFA04B7C4, 0x27411D7C, + 0xC805C650, 0x15406CE8, 0x7762E5D1, 0xAA274F69, 0xB327F7A3, 0x6E625D1B, 0x0C40D422, 0xD1057E9A, + 0xABA65FE7, 0x76E3F55F, 0x14C17C66, 0xC984D6DE, 0xD0846E14, 0x0DC1C4AC, 0x6FE34D95, 0xB2A6E72D, + 0x5DE23C01, 0x80A796B9, 0xE2851F80, 0x3FC0B538, 0x26C00DF2, 0xFB85A74A, 0x99A72E73, 0x44E284CB, + 0x42C2EEDA, 0x9F874462, 0xFDA5CD5B, 0x20E067E3, 0x39E0DF29, 0xE4A57591, 0x8687FCA8, 0x5BC25610, + 0xB4868D3C, 0x69C32784, 0x0BE1AEBD, 0xD6A40405, 0xCFA4BCCF, 0x12E11677, 0x70C39F4E, 0xAD8635F6, + 0x7C834B6C, 0xA1C6E1D4, 0xC3E468ED, 0x1EA1C255, 0x07A17A9F, 0xDAE4D027, 0xB8C6591E, 0x6583F3A6, + 0x8AC7288A, 0x57828232, 0x35A00B0B, 0xE8E5A1B3, 0xF1E51979, 0x2CA0B3C1, 0x4E823AF8, 0x93C79040, + 0x95E7FA51, 0x48A250E9, 0x2A80D9D0, 0xF7C57368, 0xEEC5CBA2, 0x3380611A, 0x51A2E823, 0x8CE7429B, + 0x63A399B7, 0xBEE6330F, 0xDCC4BA36, 0x0181108E, 0x1881A844, 0xC5C402FC, 0xA7E68BC5, 0x7AA3217D, + 0x52A0C93F, 0x8FE56387, 0xEDC7EABE, 0x30824006, 0x2982F8CC, 0xF4C75274, 0x96E5DB4D, 0x4BA071F5, + 0xA4E4AAD9, 0x79A10061, 0x1B838958, 0xC6C623E0, 0xDFC69B2A, 0x02833192, 0x60A1B8AB, 0xBDE41213, + 0xBBC47802, 0x6681D2BA, 0x04A35B83, 0xD9E6F13B, 0xC0E649F1, 0x1DA3E349, 0x7F816A70, 0xA2C4C0C8, + 0x4D801BE4, 0x90C5B15C, 0xF2E73865, 0x2FA292DD, 0x36A22A17, 0xEBE780AF, 0x89C50996, 0x5480A32E, + 0x8585DDB4, 0x58C0770C, 0x3AE2FE35, 0xE7A7548D, 0xFEA7EC47, 0x23E246FF, 0x41C0CFC6, 0x9C85657E, + 0x73C1BE52, 0xAE8414EA, 0xCCA69DD3, 0x11E3376B, 0x08E38FA1, 0xD5A62519, 0xB784AC20, 0x6AC10698, + 0x6CE16C89, 0xB1A4C631, 0xD3864F08, 0x0EC3E5B0, 0x17C35D7A, 0xCA86F7C2, 0xA8A47EFB, 0x75E1D443, + 0x9AA50F6F, 0x47E0A5D7, 0x25C22CEE, 0xF8878656, 0xE1873E9C, 0x3CC29424, 0x5EE01D1D, 0x83A5B7A5, + 0xF90696D8, 0x24433C60, 0x4661B559, 0x9B241FE1, 0x8224A72B, 0x5F610D93, 0x3D4384AA, 0xE0062E12, + 0x0F42F53E, 0xD2075F86, 0xB025D6BF, 0x6D607C07, 0x7460C4CD, 0xA9256E75, 0xCB07E74C, 0x16424DF4, + 0x106227E5, 0xCD278D5D, 0xAF050464, 0x7240AEDC, 0x6B401616, 0xB605BCAE, 0xD4273597, 0x09629F2F, + 0xE6264403, 0x3B63EEBB, 0x59416782, 0x8404CD3A, 0x9D0475F0, 0x4041DF48, 0x22635671, 0xFF26FCC9, + 0x2E238253, 0xF36628EB, 0x9144A1D2, 0x4C010B6A, 0x5501B3A0, 0x88441918, 0xEA669021, 0x37233A99, + 0xD867E1B5, 0x05224B0D, 0x6700C234, 0xBA45688C, 0xA345D046, 0x7E007AFE, 0x1C22F3C7, 0xC167597F, + 0xC747336E, 0x1A0299D6, 0x782010EF, 0xA565BA57, 0xBC65029D, 0x6120A825, 0x0302211C, 0xDE478BA4, + 0x31035088, 0xEC46FA30, 0x8E647309, 0x5321D9B1, 0x4A21617B, 0x9764CBC3, 0xF54642FA, 0x2803E842 +}; + +/* + * end of the CRC lookup table crc_tableil8_o56 + */ + + + +/* + * The following CRC lookup table was generated automagically using the + * following model parameters: + * + * Generator Polynomial = ................. 0x1EDC6F41 + * Generator Polynomial Length = .......... 32 bits + * Reflected Bits = ....................... TRUE + * Table Generation Offset = .............. 32 bits + * Number of Slices = ..................... 8 slices + * Slice Lengths = ........................ 8 8 8 8 8 8 8 8 + * Directory Name = ....................... .\ + * File Name = ............................ 8x256_tables.c + */ + +static const uint32_t sctp_crc_tableil8_o64[256] = +{ + 0x00000000, 0x38116FAC, 0x7022DF58, 0x4833B0F4, 0xE045BEB0, 0xD854D11C, 0x906761E8, 0xA8760E44, + 0xC5670B91, 0xFD76643D, 0xB545D4C9, 0x8D54BB65, 0x2522B521, 0x1D33DA8D, 0x55006A79, 0x6D1105D5, + 0x8F2261D3, 0xB7330E7F, 0xFF00BE8B, 0xC711D127, 0x6F67DF63, 0x5776B0CF, 0x1F45003B, 0x27546F97, + 0x4A456A42, 0x725405EE, 0x3A67B51A, 0x0276DAB6, 0xAA00D4F2, 0x9211BB5E, 0xDA220BAA, 0xE2336406, + 0x1BA8B557, 0x23B9DAFB, 0x6B8A6A0F, 0x539B05A3, 0xFBED0BE7, 0xC3FC644B, 0x8BCFD4BF, 0xB3DEBB13, + 0xDECFBEC6, 0xE6DED16A, 0xAEED619E, 0x96FC0E32, 0x3E8A0076, 0x069B6FDA, 0x4EA8DF2E, 0x76B9B082, + 0x948AD484, 0xAC9BBB28, 0xE4A80BDC, 0xDCB96470, 0x74CF6A34, 0x4CDE0598, 0x04EDB56C, 0x3CFCDAC0, + 0x51EDDF15, 0x69FCB0B9, 0x21CF004D, 0x19DE6FE1, 0xB1A861A5, 0x89B90E09, 0xC18ABEFD, 0xF99BD151, + 0x37516AAE, 0x0F400502, 0x4773B5F6, 0x7F62DA5A, 0xD714D41E, 0xEF05BBB2, 0xA7360B46, 0x9F2764EA, + 0xF236613F, 0xCA270E93, 0x8214BE67, 0xBA05D1CB, 0x1273DF8F, 0x2A62B023, 0x625100D7, 0x5A406F7B, + 0xB8730B7D, 0x806264D1, 0xC851D425, 0xF040BB89, 0x5836B5CD, 0x6027DA61, 0x28146A95, 0x10050539, + 0x7D1400EC, 0x45056F40, 0x0D36DFB4, 0x3527B018, 0x9D51BE5C, 0xA540D1F0, 0xED736104, 0xD5620EA8, + 0x2CF9DFF9, 0x14E8B055, 0x5CDB00A1, 0x64CA6F0D, 0xCCBC6149, 0xF4AD0EE5, 0xBC9EBE11, 0x848FD1BD, + 0xE99ED468, 0xD18FBBC4, 0x99BC0B30, 0xA1AD649C, 0x09DB6AD8, 0x31CA0574, 0x79F9B580, 0x41E8DA2C, + 0xA3DBBE2A, 0x9BCAD186, 0xD3F96172, 0xEBE80EDE, 0x439E009A, 0x7B8F6F36, 0x33BCDFC2, 0x0BADB06E, + 0x66BCB5BB, 0x5EADDA17, 0x169E6AE3, 0x2E8F054F, 0x86F90B0B, 0xBEE864A7, 0xF6DBD453, 0xCECABBFF, + 0x6EA2D55C, 0x56B3BAF0, 0x1E800A04, 0x269165A8, 0x8EE76BEC, 0xB6F60440, 0xFEC5B4B4, 0xC6D4DB18, + 0xABC5DECD, 0x93D4B161, 0xDBE70195, 0xE3F66E39, 0x4B80607D, 0x73910FD1, 0x3BA2BF25, 0x03B3D089, + 0xE180B48F, 0xD991DB23, 0x91A26BD7, 0xA9B3047B, 0x01C50A3F, 0x39D46593, 0x71E7D567, 0x49F6BACB, + 0x24E7BF1E, 0x1CF6D0B2, 0x54C56046, 0x6CD40FEA, 0xC4A201AE, 0xFCB36E02, 0xB480DEF6, 0x8C91B15A, + 0x750A600B, 0x4D1B0FA7, 0x0528BF53, 0x3D39D0FF, 0x954FDEBB, 0xAD5EB117, 0xE56D01E3, 0xDD7C6E4F, + 0xB06D6B9A, 0x887C0436, 0xC04FB4C2, 0xF85EDB6E, 0x5028D52A, 0x6839BA86, 0x200A0A72, 0x181B65DE, + 0xFA2801D8, 0xC2396E74, 0x8A0ADE80, 0xB21BB12C, 0x1A6DBF68, 0x227CD0C4, 0x6A4F6030, 0x525E0F9C, + 0x3F4F0A49, 0x075E65E5, 0x4F6DD511, 0x777CBABD, 0xDF0AB4F9, 0xE71BDB55, 0xAF286BA1, 0x9739040D, + 0x59F3BFF2, 0x61E2D05E, 0x29D160AA, 0x11C00F06, 0xB9B60142, 0x81A76EEE, 0xC994DE1A, 0xF185B1B6, + 0x9C94B463, 0xA485DBCF, 0xECB66B3B, 0xD4A70497, 0x7CD10AD3, 0x44C0657F, 0x0CF3D58B, 0x34E2BA27, + 0xD6D1DE21, 0xEEC0B18D, 0xA6F30179, 0x9EE26ED5, 0x36946091, 0x0E850F3D, 0x46B6BFC9, 0x7EA7D065, + 0x13B6D5B0, 0x2BA7BA1C, 0x63940AE8, 0x5B856544, 0xF3F36B00, 0xCBE204AC, 0x83D1B458, 0xBBC0DBF4, + 0x425B0AA5, 0x7A4A6509, 0x3279D5FD, 0x0A68BA51, 0xA21EB415, 0x9A0FDBB9, 0xD23C6B4D, 0xEA2D04E1, + 0x873C0134, 0xBF2D6E98, 0xF71EDE6C, 0xCF0FB1C0, 0x6779BF84, 0x5F68D028, 0x175B60DC, 0x2F4A0F70, + 0xCD796B76, 0xF56804DA, 0xBD5BB42E, 0x854ADB82, 0x2D3CD5C6, 0x152DBA6A, 0x5D1E0A9E, 0x650F6532, + 0x081E60E7, 0x300F0F4B, 0x783CBFBF, 0x402DD013, 0xE85BDE57, 0xD04AB1FB, 0x9879010F, 0xA0686EA3 +}; + +/* + * end of the CRC lookup table crc_tableil8_o64 + */ + + + +/* + * The following CRC lookup table was generated automagically using the + * following model parameters: + * + * Generator Polynomial = ................. 0x1EDC6F41 + * Generator Polynomial Length = .......... 32 bits + * Reflected Bits = ....................... TRUE + * Table Generation Offset = .............. 32 bits + * Number of Slices = ..................... 8 slices + * Slice Lengths = ........................ 8 8 8 8 8 8 8 8 + * Directory Name = ....................... .\ + * File Name = ............................ 8x256_tables.c + */ + +static const uint32_t sctp_crc_tableil8_o72[256] = +{ + 0x00000000, 0xEF306B19, 0xDB8CA0C3, 0x34BCCBDA, 0xB2F53777, 0x5DC55C6E, 0x697997B4, 0x8649FCAD, + 0x6006181F, 0x8F367306, 0xBB8AB8DC, 0x54BAD3C5, 0xD2F32F68, 0x3DC34471, 0x097F8FAB, 0xE64FE4B2, + 0xC00C303E, 0x2F3C5B27, 0x1B8090FD, 0xF4B0FBE4, 0x72F90749, 0x9DC96C50, 0xA975A78A, 0x4645CC93, + 0xA00A2821, 0x4F3A4338, 0x7B8688E2, 0x94B6E3FB, 0x12FF1F56, 0xFDCF744F, 0xC973BF95, 0x2643D48C, + 0x85F4168D, 0x6AC47D94, 0x5E78B64E, 0xB148DD57, 0x370121FA, 0xD8314AE3, 0xEC8D8139, 0x03BDEA20, + 0xE5F20E92, 0x0AC2658B, 0x3E7EAE51, 0xD14EC548, 0x570739E5, 0xB83752FC, 0x8C8B9926, 0x63BBF23F, + 0x45F826B3, 0xAAC84DAA, 0x9E748670, 0x7144ED69, 0xF70D11C4, 0x183D7ADD, 0x2C81B107, 0xC3B1DA1E, + 0x25FE3EAC, 0xCACE55B5, 0xFE729E6F, 0x1142F576, 0x970B09DB, 0x783B62C2, 0x4C87A918, 0xA3B7C201, + 0x0E045BEB, 0xE13430F2, 0xD588FB28, 0x3AB89031, 0xBCF16C9C, 0x53C10785, 0x677DCC5F, 0x884DA746, + 0x6E0243F4, 0x813228ED, 0xB58EE337, 0x5ABE882E, 0xDCF77483, 0x33C71F9A, 0x077BD440, 0xE84BBF59, + 0xCE086BD5, 0x213800CC, 0x1584CB16, 0xFAB4A00F, 0x7CFD5CA2, 0x93CD37BB, 0xA771FC61, 0x48419778, + 0xAE0E73CA, 0x413E18D3, 0x7582D309, 0x9AB2B810, 0x1CFB44BD, 0xF3CB2FA4, 0xC777E47E, 0x28478F67, + 0x8BF04D66, 0x64C0267F, 0x507CEDA5, 0xBF4C86BC, 0x39057A11, 0xD6351108, 0xE289DAD2, 0x0DB9B1CB, + 0xEBF65579, 0x04C63E60, 0x307AF5BA, 0xDF4A9EA3, 0x5903620E, 0xB6330917, 0x828FC2CD, 0x6DBFA9D4, + 0x4BFC7D58, 0xA4CC1641, 0x9070DD9B, 0x7F40B682, 0xF9094A2F, 0x16392136, 0x2285EAEC, 0xCDB581F5, + 0x2BFA6547, 0xC4CA0E5E, 0xF076C584, 0x1F46AE9D, 0x990F5230, 0x763F3929, 0x4283F2F3, 0xADB399EA, + 0x1C08B7D6, 0xF338DCCF, 0xC7841715, 0x28B47C0C, 0xAEFD80A1, 0x41CDEBB8, 0x75712062, 0x9A414B7B, + 0x7C0EAFC9, 0x933EC4D0, 0xA7820F0A, 0x48B26413, 0xCEFB98BE, 0x21CBF3A7, 0x1577387D, 0xFA475364, + 0xDC0487E8, 0x3334ECF1, 0x0788272B, 0xE8B84C32, 0x6EF1B09F, 0x81C1DB86, 0xB57D105C, 0x5A4D7B45, + 0xBC029FF7, 0x5332F4EE, 0x678E3F34, 0x88BE542D, 0x0EF7A880, 0xE1C7C399, 0xD57B0843, 0x3A4B635A, + 0x99FCA15B, 0x76CCCA42, 0x42700198, 0xAD406A81, 0x2B09962C, 0xC439FD35, 0xF08536EF, 0x1FB55DF6, + 0xF9FAB944, 0x16CAD25D, 0x22761987, 0xCD46729E, 0x4B0F8E33, 0xA43FE52A, 0x90832EF0, 0x7FB345E9, + 0x59F09165, 0xB6C0FA7C, 0x827C31A6, 0x6D4C5ABF, 0xEB05A612, 0x0435CD0B, 0x308906D1, 0xDFB96DC8, + 0x39F6897A, 0xD6C6E263, 0xE27A29B9, 0x0D4A42A0, 0x8B03BE0D, 0x6433D514, 0x508F1ECE, 0xBFBF75D7, + 0x120CEC3D, 0xFD3C8724, 0xC9804CFE, 0x26B027E7, 0xA0F9DB4A, 0x4FC9B053, 0x7B757B89, 0x94451090, + 0x720AF422, 0x9D3A9F3B, 0xA98654E1, 0x46B63FF8, 0xC0FFC355, 0x2FCFA84C, 0x1B736396, 0xF443088F, + 0xD200DC03, 0x3D30B71A, 0x098C7CC0, 0xE6BC17D9, 0x60F5EB74, 0x8FC5806D, 0xBB794BB7, 0x544920AE, + 0xB206C41C, 0x5D36AF05, 0x698A64DF, 0x86BA0FC6, 0x00F3F36B, 0xEFC39872, 0xDB7F53A8, 0x344F38B1, + 0x97F8FAB0, 0x78C891A9, 0x4C745A73, 0xA344316A, 0x250DCDC7, 0xCA3DA6DE, 0xFE816D04, 0x11B1061D, + 0xF7FEE2AF, 0x18CE89B6, 0x2C72426C, 0xC3422975, 0x450BD5D8, 0xAA3BBEC1, 0x9E87751B, 0x71B71E02, + 0x57F4CA8E, 0xB8C4A197, 0x8C786A4D, 0x63480154, 0xE501FDF9, 0x0A3196E0, 0x3E8D5D3A, 0xD1BD3623, + 0x37F2D291, 0xD8C2B988, 0xEC7E7252, 0x034E194B, 0x8507E5E6, 0x6A378EFF, 0x5E8B4525, 0xB1BB2E3C +}; + +/* + * end of the CRC lookup table crc_tableil8_o72 + */ + + + +/* + * The following CRC lookup table was generated automagically using the + * following model parameters: + * + * Generator Polynomial = ................. 0x1EDC6F41 + * Generator Polynomial Length = .......... 32 bits + * Reflected Bits = ....................... TRUE + * Table Generation Offset = .............. 32 bits + * Number of Slices = ..................... 8 slices + * Slice Lengths = ........................ 8 8 8 8 8 8 8 8 + * Directory Name = ....................... .\ + * File Name = ............................ 8x256_tables.c + */ + +static const uint32_t sctp_crc_tableil8_o80[256] = +{ + 0x00000000, 0x68032CC8, 0xD0065990, 0xB8057558, 0xA5E0C5D1, 0xCDE3E919, 0x75E69C41, 0x1DE5B089, + 0x4E2DFD53, 0x262ED19B, 0x9E2BA4C3, 0xF628880B, 0xEBCD3882, 0x83CE144A, 0x3BCB6112, 0x53C84DDA, + 0x9C5BFAA6, 0xF458D66E, 0x4C5DA336, 0x245E8FFE, 0x39BB3F77, 0x51B813BF, 0xE9BD66E7, 0x81BE4A2F, + 0xD27607F5, 0xBA752B3D, 0x02705E65, 0x6A7372AD, 0x7796C224, 0x1F95EEEC, 0xA7909BB4, 0xCF93B77C, + 0x3D5B83BD, 0x5558AF75, 0xED5DDA2D, 0x855EF6E5, 0x98BB466C, 0xF0B86AA4, 0x48BD1FFC, 0x20BE3334, + 0x73767EEE, 0x1B755226, 0xA370277E, 0xCB730BB6, 0xD696BB3F, 0xBE9597F7, 0x0690E2AF, 0x6E93CE67, + 0xA100791B, 0xC90355D3, 0x7106208B, 0x19050C43, 0x04E0BCCA, 0x6CE39002, 0xD4E6E55A, 0xBCE5C992, + 0xEF2D8448, 0x872EA880, 0x3F2BDDD8, 0x5728F110, 0x4ACD4199, 0x22CE6D51, 0x9ACB1809, 0xF2C834C1, + 0x7AB7077A, 0x12B42BB2, 0xAAB15EEA, 0xC2B27222, 0xDF57C2AB, 0xB754EE63, 0x0F519B3B, 0x6752B7F3, + 0x349AFA29, 0x5C99D6E1, 0xE49CA3B9, 0x8C9F8F71, 0x917A3FF8, 0xF9791330, 0x417C6668, 0x297F4AA0, + 0xE6ECFDDC, 0x8EEFD114, 0x36EAA44C, 0x5EE98884, 0x430C380D, 0x2B0F14C5, 0x930A619D, 0xFB094D55, + 0xA8C1008F, 0xC0C22C47, 0x78C7591F, 0x10C475D7, 0x0D21C55E, 0x6522E996, 0xDD279CCE, 0xB524B006, + 0x47EC84C7, 0x2FEFA80F, 0x97EADD57, 0xFFE9F19F, 0xE20C4116, 0x8A0F6DDE, 0x320A1886, 0x5A09344E, + 0x09C17994, 0x61C2555C, 0xD9C72004, 0xB1C40CCC, 0xAC21BC45, 0xC422908D, 0x7C27E5D5, 0x1424C91D, + 0xDBB77E61, 0xB3B452A9, 0x0BB127F1, 0x63B20B39, 0x7E57BBB0, 0x16549778, 0xAE51E220, 0xC652CEE8, + 0x959A8332, 0xFD99AFFA, 0x459CDAA2, 0x2D9FF66A, 0x307A46E3, 0x58796A2B, 0xE07C1F73, 0x887F33BB, + 0xF56E0EF4, 0x9D6D223C, 0x25685764, 0x4D6B7BAC, 0x508ECB25, 0x388DE7ED, 0x808892B5, 0xE88BBE7D, + 0xBB43F3A7, 0xD340DF6F, 0x6B45AA37, 0x034686FF, 0x1EA33676, 0x76A01ABE, 0xCEA56FE6, 0xA6A6432E, + 0x6935F452, 0x0136D89A, 0xB933ADC2, 0xD130810A, 0xCCD53183, 0xA4D61D4B, 0x1CD36813, 0x74D044DB, + 0x27180901, 0x4F1B25C9, 0xF71E5091, 0x9F1D7C59, 0x82F8CCD0, 0xEAFBE018, 0x52FE9540, 0x3AFDB988, + 0xC8358D49, 0xA036A181, 0x1833D4D9, 0x7030F811, 0x6DD54898, 0x05D66450, 0xBDD31108, 0xD5D03DC0, + 0x8618701A, 0xEE1B5CD2, 0x561E298A, 0x3E1D0542, 0x23F8B5CB, 0x4BFB9903, 0xF3FEEC5B, 0x9BFDC093, + 0x546E77EF, 0x3C6D5B27, 0x84682E7F, 0xEC6B02B7, 0xF18EB23E, 0x998D9EF6, 0x2188EBAE, 0x498BC766, + 0x1A438ABC, 0x7240A674, 0xCA45D32C, 0xA246FFE4, 0xBFA34F6D, 0xD7A063A5, 0x6FA516FD, 0x07A63A35, + 0x8FD9098E, 0xE7DA2546, 0x5FDF501E, 0x37DC7CD6, 0x2A39CC5F, 0x423AE097, 0xFA3F95CF, 0x923CB907, + 0xC1F4F4DD, 0xA9F7D815, 0x11F2AD4D, 0x79F18185, 0x6414310C, 0x0C171DC4, 0xB412689C, 0xDC114454, + 0x1382F328, 0x7B81DFE0, 0xC384AAB8, 0xAB878670, 0xB66236F9, 0xDE611A31, 0x66646F69, 0x0E6743A1, + 0x5DAF0E7B, 0x35AC22B3, 0x8DA957EB, 0xE5AA7B23, 0xF84FCBAA, 0x904CE762, 0x2849923A, 0x404ABEF2, + 0xB2828A33, 0xDA81A6FB, 0x6284D3A3, 0x0A87FF6B, 0x17624FE2, 0x7F61632A, 0xC7641672, 0xAF673ABA, + 0xFCAF7760, 0x94AC5BA8, 0x2CA92EF0, 0x44AA0238, 0x594FB2B1, 0x314C9E79, 0x8949EB21, 0xE14AC7E9, + 0x2ED97095, 0x46DA5C5D, 0xFEDF2905, 0x96DC05CD, 0x8B39B544, 0xE33A998C, 0x5B3FECD4, 0x333CC01C, + 0x60F48DC6, 0x08F7A10E, 0xB0F2D456, 0xD8F1F89E, 0xC5144817, 0xAD1764DF, 0x15121187, 0x7D113D4F +}; + +/* + * end of the CRC lookup table crc_tableil8_o80 + */ + + + +/* + * The following CRC lookup table was generated automagically using the + * following model parameters: + * + * Generator Polynomial = ................. 0x1EDC6F41 + * Generator Polynomial Length = .......... 32 bits + * Reflected Bits = ....................... TRUE + * Table Generation Offset = .............. 32 bits + * Number of Slices = ..................... 8 slices + * Slice Lengths = ........................ 8 8 8 8 8 8 8 8 + * Directory Name = ....................... .\ + * File Name = ............................ 8x256_tables.c + */ + +static const uint32_t sctp_crc_tableil8_o88[256] = +{ + 0x00000000, 0x493C7D27, 0x9278FA4E, 0xDB448769, 0x211D826D, 0x6821FF4A, 0xB3657823, 0xFA590504, + 0x423B04DA, 0x0B0779FD, 0xD043FE94, 0x997F83B3, 0x632686B7, 0x2A1AFB90, 0xF15E7CF9, 0xB86201DE, + 0x847609B4, 0xCD4A7493, 0x160EF3FA, 0x5F328EDD, 0xA56B8BD9, 0xEC57F6FE, 0x37137197, 0x7E2F0CB0, + 0xC64D0D6E, 0x8F717049, 0x5435F720, 0x1D098A07, 0xE7508F03, 0xAE6CF224, 0x7528754D, 0x3C14086A, + 0x0D006599, 0x443C18BE, 0x9F789FD7, 0xD644E2F0, 0x2C1DE7F4, 0x65219AD3, 0xBE651DBA, 0xF759609D, + 0x4F3B6143, 0x06071C64, 0xDD439B0D, 0x947FE62A, 0x6E26E32E, 0x271A9E09, 0xFC5E1960, 0xB5626447, + 0x89766C2D, 0xC04A110A, 0x1B0E9663, 0x5232EB44, 0xA86BEE40, 0xE1579367, 0x3A13140E, 0x732F6929, + 0xCB4D68F7, 0x827115D0, 0x593592B9, 0x1009EF9E, 0xEA50EA9A, 0xA36C97BD, 0x782810D4, 0x31146DF3, + 0x1A00CB32, 0x533CB615, 0x8878317C, 0xC1444C5B, 0x3B1D495F, 0x72213478, 0xA965B311, 0xE059CE36, + 0x583BCFE8, 0x1107B2CF, 0xCA4335A6, 0x837F4881, 0x79264D85, 0x301A30A2, 0xEB5EB7CB, 0xA262CAEC, + 0x9E76C286, 0xD74ABFA1, 0x0C0E38C8, 0x453245EF, 0xBF6B40EB, 0xF6573DCC, 0x2D13BAA5, 0x642FC782, + 0xDC4DC65C, 0x9571BB7B, 0x4E353C12, 0x07094135, 0xFD504431, 0xB46C3916, 0x6F28BE7F, 0x2614C358, + 0x1700AEAB, 0x5E3CD38C, 0x857854E5, 0xCC4429C2, 0x361D2CC6, 0x7F2151E1, 0xA465D688, 0xED59ABAF, + 0x553BAA71, 0x1C07D756, 0xC743503F, 0x8E7F2D18, 0x7426281C, 0x3D1A553B, 0xE65ED252, 0xAF62AF75, + 0x9376A71F, 0xDA4ADA38, 0x010E5D51, 0x48322076, 0xB26B2572, 0xFB575855, 0x2013DF3C, 0x692FA21B, + 0xD14DA3C5, 0x9871DEE2, 0x4335598B, 0x0A0924AC, 0xF05021A8, 0xB96C5C8F, 0x6228DBE6, 0x2B14A6C1, + 0x34019664, 0x7D3DEB43, 0xA6796C2A, 0xEF45110D, 0x151C1409, 0x5C20692E, 0x8764EE47, 0xCE589360, + 0x763A92BE, 0x3F06EF99, 0xE44268F0, 0xAD7E15D7, 0x572710D3, 0x1E1B6DF4, 0xC55FEA9D, 0x8C6397BA, + 0xB0779FD0, 0xF94BE2F7, 0x220F659E, 0x6B3318B9, 0x916A1DBD, 0xD856609A, 0x0312E7F3, 0x4A2E9AD4, + 0xF24C9B0A, 0xBB70E62D, 0x60346144, 0x29081C63, 0xD3511967, 0x9A6D6440, 0x4129E329, 0x08159E0E, + 0x3901F3FD, 0x703D8EDA, 0xAB7909B3, 0xE2457494, 0x181C7190, 0x51200CB7, 0x8A648BDE, 0xC358F6F9, + 0x7B3AF727, 0x32068A00, 0xE9420D69, 0xA07E704E, 0x5A27754A, 0x131B086D, 0xC85F8F04, 0x8163F223, + 0xBD77FA49, 0xF44B876E, 0x2F0F0007, 0x66337D20, 0x9C6A7824, 0xD5560503, 0x0E12826A, 0x472EFF4D, + 0xFF4CFE93, 0xB67083B4, 0x6D3404DD, 0x240879FA, 0xDE517CFE, 0x976D01D9, 0x4C2986B0, 0x0515FB97, + 0x2E015D56, 0x673D2071, 0xBC79A718, 0xF545DA3F, 0x0F1CDF3B, 0x4620A21C, 0x9D642575, 0xD4585852, + 0x6C3A598C, 0x250624AB, 0xFE42A3C2, 0xB77EDEE5, 0x4D27DBE1, 0x041BA6C6, 0xDF5F21AF, 0x96635C88, + 0xAA7754E2, 0xE34B29C5, 0x380FAEAC, 0x7133D38B, 0x8B6AD68F, 0xC256ABA8, 0x19122CC1, 0x502E51E6, + 0xE84C5038, 0xA1702D1F, 0x7A34AA76, 0x3308D751, 0xC951D255, 0x806DAF72, 0x5B29281B, 0x1215553C, + 0x230138CF, 0x6A3D45E8, 0xB179C281, 0xF845BFA6, 0x021CBAA2, 0x4B20C785, 0x906440EC, 0xD9583DCB, + 0x613A3C15, 0x28064132, 0xF342C65B, 0xBA7EBB7C, 0x4027BE78, 0x091BC35F, 0xD25F4436, 0x9B633911, + 0xA777317B, 0xEE4B4C5C, 0x350FCB35, 0x7C33B612, 0x866AB316, 0xCF56CE31, 0x14124958, 0x5D2E347F, + 0xE54C35A1, 0xAC704886, 0x7734CFEF, 0x3E08B2C8, 0xC451B7CC, 0x8D6DCAEB, 0x56294D82, 0x1F1530A5 +}; + +/* + * end of the CRC lookup table crc_tableil8_o88 + */ + + +static uint32_t +sctp_crc32c_sb8_64_bit(uint32_t crc, + const unsigned char *p_buf, + uint32_t length, + uint32_t init_bytes) +{ + uint32_t li; + uint32_t term1, term2; + uint32_t running_length; + uint32_t end_bytes; + + running_length = ((length - init_bytes) / 8) * 8; + end_bytes = length - init_bytes - running_length; + + for (li = 0; li < init_bytes; li++) + crc = sctp_crc_tableil8_o32[(crc ^ *p_buf++) & 0x000000FF] ^ + (crc >> 8); + for (li = 0; li < running_length / 8; li++) { +#if BYTE_ORDER == BIG_ENDIAN + crc ^= *p_buf++; + crc ^= (*p_buf++) << 8; + crc ^= (*p_buf++) << 16; + crc ^= (*p_buf++) << 24; +#else + crc ^= *(const uint32_t *) p_buf; + p_buf += 4; +#endif + term1 = sctp_crc_tableil8_o88[crc & 0x000000FF] ^ + sctp_crc_tableil8_o80[(crc >> 8) & 0x000000FF]; + term2 = crc >> 16; + crc = term1 ^ + sctp_crc_tableil8_o72[term2 & 0x000000FF] ^ + sctp_crc_tableil8_o64[(term2 >> 8) & 0x000000FF]; + +#if BYTE_ORDER == BIG_ENDIAN + crc ^= sctp_crc_tableil8_o56[*p_buf++]; + crc ^= sctp_crc_tableil8_o48[*p_buf++]; + crc ^= sctp_crc_tableil8_o40[*p_buf++]; + crc ^= sctp_crc_tableil8_o32[*p_buf++]; +#else + term1 = sctp_crc_tableil8_o56[(*(const uint32_t *) p_buf) & 0x000000FF] ^ + sctp_crc_tableil8_o48[((*(const uint32_t *) p_buf) >> 8) & 0x000000FF]; + + term2 = (*(const uint32_t *) p_buf) >> 16; + crc = crc ^ + term1 ^ + sctp_crc_tableil8_o40[term2 & 0x000000FF] ^ + sctp_crc_tableil8_o32[(term2 >> 8) & 0x000000FF]; + p_buf += 4; +#endif + } + for (li = 0; li < end_bytes; li++) + crc = sctp_crc_tableil8_o32[(crc ^ *p_buf++) & 0x000000FF] ^ + (crc >> 8); + return (crc); +} + + +/** + * + * Routine Description: + * + * warms the tables + * + * Arguments: + * + * none + * + * Return value: + * + * none + */ +static uint32_t +multitable_crc32c(uint32_t crc32c, + const unsigned char *buffer, + unsigned int length) +{ + uint32_t to_even_word; + + if (length == 0) { + return (crc32c); + } + to_even_word = (4 - (((uintptr_t) buffer) & 0x3)); + return (sctp_crc32c_sb8_64_bit(crc32c, buffer, length, to_even_word)); +} + +static const uint32_t sctp_crc_c[256] = { + 0x00000000, 0xF26B8303, 0xE13B70F7, 0x1350F3F4, + 0xC79A971F, 0x35F1141C, 0x26A1E7E8, 0xD4CA64EB, + 0x8AD958CF, 0x78B2DBCC, 0x6BE22838, 0x9989AB3B, + 0x4D43CFD0, 0xBF284CD3, 0xAC78BF27, 0x5E133C24, + 0x105EC76F, 0xE235446C, 0xF165B798, 0x030E349B, + 0xD7C45070, 0x25AFD373, 0x36FF2087, 0xC494A384, + 0x9A879FA0, 0x68EC1CA3, 0x7BBCEF57, 0x89D76C54, + 0x5D1D08BF, 0xAF768BBC, 0xBC267848, 0x4E4DFB4B, + 0x20BD8EDE, 0xD2D60DDD, 0xC186FE29, 0x33ED7D2A, + 0xE72719C1, 0x154C9AC2, 0x061C6936, 0xF477EA35, + 0xAA64D611, 0x580F5512, 0x4B5FA6E6, 0xB93425E5, + 0x6DFE410E, 0x9F95C20D, 0x8CC531F9, 0x7EAEB2FA, + 0x30E349B1, 0xC288CAB2, 0xD1D83946, 0x23B3BA45, + 0xF779DEAE, 0x05125DAD, 0x1642AE59, 0xE4292D5A, + 0xBA3A117E, 0x4851927D, 0x5B016189, 0xA96AE28A, + 0x7DA08661, 0x8FCB0562, 0x9C9BF696, 0x6EF07595, + 0x417B1DBC, 0xB3109EBF, 0xA0406D4B, 0x522BEE48, + 0x86E18AA3, 0x748A09A0, 0x67DAFA54, 0x95B17957, + 0xCBA24573, 0x39C9C670, 0x2A993584, 0xD8F2B687, + 0x0C38D26C, 0xFE53516F, 0xED03A29B, 0x1F682198, + 0x5125DAD3, 0xA34E59D0, 0xB01EAA24, 0x42752927, + 0x96BF4DCC, 0x64D4CECF, 0x77843D3B, 0x85EFBE38, + 0xDBFC821C, 0x2997011F, 0x3AC7F2EB, 0xC8AC71E8, + 0x1C661503, 0xEE0D9600, 0xFD5D65F4, 0x0F36E6F7, + 0x61C69362, 0x93AD1061, 0x80FDE395, 0x72966096, + 0xA65C047D, 0x5437877E, 0x4767748A, 0xB50CF789, + 0xEB1FCBAD, 0x197448AE, 0x0A24BB5A, 0xF84F3859, + 0x2C855CB2, 0xDEEEDFB1, 0xCDBE2C45, 0x3FD5AF46, + 0x7198540D, 0x83F3D70E, 0x90A324FA, 0x62C8A7F9, + 0xB602C312, 0x44694011, 0x5739B3E5, 0xA55230E6, + 0xFB410CC2, 0x092A8FC1, 0x1A7A7C35, 0xE811FF36, + 0x3CDB9BDD, 0xCEB018DE, 0xDDE0EB2A, 0x2F8B6829, + 0x82F63B78, 0x709DB87B, 0x63CD4B8F, 0x91A6C88C, + 0x456CAC67, 0xB7072F64, 0xA457DC90, 0x563C5F93, + 0x082F63B7, 0xFA44E0B4, 0xE9141340, 0x1B7F9043, + 0xCFB5F4A8, 0x3DDE77AB, 0x2E8E845F, 0xDCE5075C, + 0x92A8FC17, 0x60C37F14, 0x73938CE0, 0x81F80FE3, + 0x55326B08, 0xA759E80B, 0xB4091BFF, 0x466298FC, + 0x1871A4D8, 0xEA1A27DB, 0xF94AD42F, 0x0B21572C, + 0xDFEB33C7, 0x2D80B0C4, 0x3ED04330, 0xCCBBC033, + 0xA24BB5A6, 0x502036A5, 0x4370C551, 0xB11B4652, + 0x65D122B9, 0x97BAA1BA, 0x84EA524E, 0x7681D14D, + 0x2892ED69, 0xDAF96E6A, 0xC9A99D9E, 0x3BC21E9D, + 0xEF087A76, 0x1D63F975, 0x0E330A81, 0xFC588982, + 0xB21572C9, 0x407EF1CA, 0x532E023E, 0xA145813D, + 0x758FE5D6, 0x87E466D5, 0x94B49521, 0x66DF1622, + 0x38CC2A06, 0xCAA7A905, 0xD9F75AF1, 0x2B9CD9F2, + 0xFF56BD19, 0x0D3D3E1A, 0x1E6DCDEE, 0xEC064EED, + 0xC38D26C4, 0x31E6A5C7, 0x22B65633, 0xD0DDD530, + 0x0417B1DB, 0xF67C32D8, 0xE52CC12C, 0x1747422F, + 0x49547E0B, 0xBB3FFD08, 0xA86F0EFC, 0x5A048DFF, + 0x8ECEE914, 0x7CA56A17, 0x6FF599E3, 0x9D9E1AE0, + 0xD3D3E1AB, 0x21B862A8, 0x32E8915C, 0xC083125F, + 0x144976B4, 0xE622F5B7, 0xF5720643, 0x07198540, + 0x590AB964, 0xAB613A67, 0xB831C993, 0x4A5A4A90, + 0x9E902E7B, 0x6CFBAD78, 0x7FAB5E8C, 0x8DC0DD8F, + 0xE330A81A, 0x115B2B19, 0x020BD8ED, 0xF0605BEE, + 0x24AA3F05, 0xD6C1BC06, 0xC5914FF2, 0x37FACCF1, + 0x69E9F0D5, 0x9B8273D6, 0x88D28022, 0x7AB90321, + 0xAE7367CA, 0x5C18E4C9, 0x4F48173D, 0xBD23943E, + 0xF36E6F75, 0x0105EC76, 0x12551F82, 0xE03E9C81, + 0x34F4F86A, 0xC69F7B69, 0xD5CF889D, 0x27A40B9E, + 0x79B737BA, 0x8BDCB4B9, 0x988C474D, 0x6AE7C44E, + 0xBE2DA0A5, 0x4C4623A6, 0x5F16D052, 0xAD7D5351, +}; + + +#define SCTP_CRC32C(c,d) (c=(c>>8)^sctp_crc_c[(c^(d))&0xFF]) + +static uint32_t +singletable_crc32c(uint32_t crc32c, + const unsigned char *buffer, + unsigned int length) +{ + unsigned int i; + + for (i = 0; i < length; i++) { + SCTP_CRC32C(crc32c, buffer[i]); + } + return (crc32c); +} + +#if defined(__Userspace__) +uint32_t +#else +static uint32_t +#endif +calculate_crc32c(uint32_t crc32c, + const unsigned char *buffer, + unsigned int length) +{ + if (length < 4) { + return (singletable_crc32c(crc32c, buffer, length)); + } else { + return (multitable_crc32c(crc32c, buffer, length)); + } +} + +#endif +#if defined(__Userspace__) +uint32_t +#else +static uint32_t +#endif +sctp_finalize_crc32c(uint32_t crc32c) +{ +#if BYTE_ORDER == BIG_ENDIAN + uint32_t byte0, byte1, byte2, byte3; +#endif + +#if BYTE_ORDER == BIG_ENDIAN + /* + * For BIG-ENDIAN platforms, the result is in LITTLE-ENDIAN byte order. + * For LITTLE-ENDIAN platforms, the result is in in BIG-ENDIAN byte + * order. So for BIG-ENDIAN platforms the bytes must be swapped to + * return the result always in network byte order (aka BIG-ENDIAN). + */ + byte0 = crc32c & 0x000000ff; + byte1 = (crc32c >> 8) & 0x000000ff; + byte2 = (crc32c >> 16) & 0x000000ff; + byte3 = (crc32c >> 24) & 0x000000ff; + crc32c = ((byte0 << 24) | (byte1 << 16) | (byte2 << 8) | byte3); +#endif + return (~crc32c); +} + +static int +sctp_calculate_cksum_cb(void *arg, void *data, u_int len) +{ + uint32_t *basep; + + basep = arg; + *basep = calculate_crc32c(*basep, data, len); + return (0); +} + +/* + * Compute the SCTP checksum in network byte order for a given mbuf chain m + * which contains an SCTP packet starting at offset. + * Since this function is also called by ipfw, don't assume that + * it is compiled on a kernel with SCTP support. + */ +uint32_t +sctp_calculate_cksum(struct mbuf *m, int32_t offset) +{ + uint32_t base; + int len; + + M_ASSERTPKTHDR(m); + KASSERT(offset < m->m_pkthdr.len, + ("%s: invalid offset %u into mbuf %p", __func__, offset, m)); + + base = 0xffffffff; + len = m->m_pkthdr.len - offset; + (void)m_apply(m, offset, len, sctp_calculate_cksum_cb, &base); + return (sctp_finalize_crc32c(base)); +} + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SCTP) || defined(SCTP_SUPPORT) + +VNET_DEFINE(struct sctp_base_info, system_base_info); + +/* + * Compute and insert the SCTP checksum in network byte order for a given + * mbuf chain m which contains an SCTP packet starting at offset. + */ +void +sctp_delayed_cksum(struct mbuf *m, uint32_t offset) +{ + uint32_t checksum; + + checksum = sctp_calculate_cksum(m, offset); + SCTP_STAT_DECR(sctps_sendhwcrc); + SCTP_STAT_INCR(sctps_sendswcrc); + offset += offsetof(struct sctphdr, checksum); + + if (offset + sizeof(uint32_t) > (uint32_t)(m->m_pkthdr.len)) { +#ifdef INVARIANTS + panic("sctp_delayed_cksum(): m->m_pkthdr.len: %d, offset: %u.", + m->m_pkthdr.len, offset); +#else + SCTP_PRINTF("sctp_delayed_cksum(): m->m_pkthdr.len: %d, offset: %u.\n", + m->m_pkthdr.len, offset); +#endif + return; + } + m_copyback(m, (int)offset, (int)sizeof(uint32_t), (caddr_t)&checksum); +} +#endif +#endif + diff --git a/netwerk/sctp/src/netinet/sctp_crc32.h b/netwerk/sctp/src/netinet/sctp_crc32.h new file mode 100644 index 0000000000..18be779d62 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_crc32.h @@ -0,0 +1,56 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_CRC32_H_ +#define _NETINET_SCTP_CRC32_H_ + +#if defined(_KERNEL) +uint32_t sctp_calculate_cksum(struct mbuf *, int32_t); +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SCTP) || defined(SCTP_SUPPORT) +void sctp_delayed_cksum(struct mbuf *, uint32_t offset); +#endif +#endif +#endif /* _KERNEL */ +#if defined(__Userspace__) +uint32_t calculate_crc32c(uint32_t, const unsigned char *, unsigned int); +uint32_t sctp_finalize_crc32c(uint32_t); +uint32_t sctp_calculate_cksum(struct mbuf *, int32_t); +#endif +#endif /* __crc32c_h__ */ diff --git a/netwerk/sctp/src/netinet/sctp_header.h b/netwerk/sctp/src/netinet/sctp_header.h new file mode 100644 index 0000000000..64a27eccf6 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_header.h @@ -0,0 +1,596 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_HEADER_H_ +#define _NETINET_SCTP_HEADER_H_ + +#if defined(_WIN32) && !defined(__Userspace__) +#include <packon.h> +#endif +#if !defined(_WIN32) +#include <sys/time.h> +#endif +#include <netinet/sctp.h> +#include <netinet/sctp_constants.h> + +#if !defined(_WIN32) +#define SCTP_PACKED __attribute__((packed)) +#else +#pragma pack (push, 1) +#define SCTP_PACKED +#endif + +/* + * Parameter structures + */ +struct sctp_ipv4addr_param { + struct sctp_paramhdr ph;/* type=SCTP_IPV4_PARAM_TYPE, len=8 */ + uint32_t addr; /* IPV4 address */ +} SCTP_PACKED; + +#define SCTP_V6_ADDR_BYTES 16 + +struct sctp_ipv6addr_param { + struct sctp_paramhdr ph;/* type=SCTP_IPV6_PARAM_TYPE, len=20 */ + uint8_t addr[SCTP_V6_ADDR_BYTES]; /* IPV6 address */ +} SCTP_PACKED; + +/* Cookie Preservative */ +struct sctp_cookie_perserve_param { + struct sctp_paramhdr ph;/* type=SCTP_COOKIE_PRESERVE, len=8 */ + uint32_t time; /* time in ms to extend cookie */ +} SCTP_PACKED; + +#define SCTP_ARRAY_MIN_LEN 1 +/* Host Name Address */ +struct sctp_host_name_param { + struct sctp_paramhdr ph;/* type=SCTP_HOSTNAME_ADDRESS */ + char name[SCTP_ARRAY_MIN_LEN]; /* host name */ +} SCTP_PACKED; + +/* + * This is the maximum padded size of a s-a-p + * so paramheadr + 3 address types (6 bytes) + 2 byte pad = 12 + */ +#define SCTP_MAX_ADDR_PARAMS_SIZE 12 +/* supported address type */ +struct sctp_supported_addr_param { + struct sctp_paramhdr ph;/* type=SCTP_SUPPORTED_ADDRTYPE */ + uint16_t addr_type[2]; /* array of supported address types */ +} SCTP_PACKED; + +/* heartbeat info parameter */ +struct sctp_heartbeat_info_param { + struct sctp_paramhdr ph; + uint32_t time_value_1; + uint32_t time_value_2; + uint32_t random_value1; + uint32_t random_value2; + uint8_t addr_family; + uint8_t addr_len; + /* make sure that this structure is 4 byte aligned */ + uint8_t padding[2]; + char address[SCTP_ADDRMAX]; +} SCTP_PACKED; + +/* draft-ietf-tsvwg-prsctp */ +/* PR-SCTP supported parameter */ +struct sctp_prsctp_supported_param { + struct sctp_paramhdr ph; +} SCTP_PACKED; + +/* draft-ietf-tsvwg-addip-sctp */ +struct sctp_asconf_paramhdr { /* an ASCONF "parameter" */ + struct sctp_paramhdr ph;/* a SCTP parameter header */ + uint32_t correlation_id;/* correlation id for this param */ +} SCTP_PACKED; + +struct sctp_asconf_addr_param { /* an ASCONF address parameter */ + struct sctp_asconf_paramhdr aph; /* asconf "parameter" */ + struct sctp_ipv6addr_param addrp; /* max storage size */ +} SCTP_PACKED; + +struct sctp_asconf_tag_param { /* an ASCONF NAT-Vtag parameter */ + struct sctp_asconf_paramhdr aph; /* asconf "parameter" */ + uint32_t local_vtag; + uint32_t remote_vtag; +} SCTP_PACKED; + +struct sctp_asconf_addrv4_param { /* an ASCONF address (v4) parameter */ + struct sctp_asconf_paramhdr aph; /* asconf "parameter" */ + struct sctp_ipv4addr_param addrp; /* max storage size */ +} SCTP_PACKED; + +#define SCTP_MAX_SUPPORTED_EXT 256 + +struct sctp_supported_chunk_types_param { + struct sctp_paramhdr ph;/* type = 0x8008 len = x */ + uint8_t chunk_types[]; +} SCTP_PACKED; + +/* + * Structures for DATA chunks + */ +struct sctp_data { + uint32_t tsn; + uint16_t sid; + uint16_t ssn; + uint32_t ppid; + /* user data follows */ +} SCTP_PACKED; + +struct sctp_data_chunk { + struct sctp_chunkhdr ch; + struct sctp_data dp; +} SCTP_PACKED; + +struct sctp_idata { + uint32_t tsn; + uint16_t sid; + uint16_t reserved; /* Where does the SSN go? */ + uint32_t mid; + union { + uint32_t ppid; + uint32_t fsn; /* Fragment Sequence Number */ + } ppid_fsn; + /* user data follows */ +} SCTP_PACKED; + +struct sctp_idata_chunk { + struct sctp_chunkhdr ch; + struct sctp_idata dp; +} SCTP_PACKED; + +/* + * Structures for the control chunks + */ + +/* Initiate (INIT)/Initiate Ack (INIT ACK) */ +struct sctp_init { + uint32_t initiate_tag; /* initiate tag */ + uint32_t a_rwnd; /* a_rwnd */ + uint16_t num_outbound_streams; /* OS */ + uint16_t num_inbound_streams; /* MIS */ + uint32_t initial_tsn; /* I-TSN */ + /* optional param's follow */ +} SCTP_PACKED; +#define SCTP_IDENTIFICATION_SIZE 16 +#define SCTP_ADDRESS_SIZE 4 +#if defined(__Userspace__) +#define SCTP_RESERVE_SPACE 4 +#else +#define SCTP_RESERVE_SPACE 5 +#endif +/* state cookie header */ +struct sctp_state_cookie { /* this is our definition... */ + uint8_t identification[SCTP_IDENTIFICATION_SIZE];/* id of who we are */ + struct timeval time_entered; /* the time I built cookie */ + uint32_t cookie_life; /* life I will award this cookie */ + uint32_t tie_tag_my_vtag; /* my tag in old association */ + + uint32_t tie_tag_peer_vtag; /* peers tag in old association */ + uint32_t peers_vtag; /* peers tag in INIT (for quick ref) */ + + uint32_t my_vtag; /* my tag in INIT-ACK (for quick ref) */ + uint32_t address[SCTP_ADDRESS_SIZE]; /* 4 ints/128 bits */ + uint32_t addr_type; /* address type */ + uint32_t laddress[SCTP_ADDRESS_SIZE]; /* my local from address */ + uint32_t laddr_type; /* my local from address type */ + uint32_t scope_id; /* v6 scope id for link-locals */ + + uint16_t peerport; /* port address of the peer in the INIT */ + uint16_t myport; /* my port address used in the INIT */ + uint8_t ipv4_addr_legal;/* Are V4 addr legal? */ + uint8_t ipv6_addr_legal;/* Are V6 addr legal? */ +#if defined(__Userspace__) + uint8_t conn_addr_legal; +#endif + uint8_t local_scope; /* IPv6 local scope flag */ + uint8_t site_scope; /* IPv6 site scope flag */ + + uint8_t ipv4_scope; /* IPv4 private addr scope */ + uint8_t loopback_scope; /* loopback scope information */ + uint8_t zero_checksum; /* copy of the inp value */ + uint8_t reserved[SCTP_RESERVE_SPACE]; /* Align to 64 bits */ + /* + * at the end is tacked on the INIT chunk and the INIT-ACK chunk + * (minus the cookie). + */ +} SCTP_PACKED; + +/* state cookie parameter */ +struct sctp_state_cookie_param { + struct sctp_paramhdr ph; + struct sctp_state_cookie cookie; +} SCTP_PACKED; + +struct sctp_init_chunk { + struct sctp_chunkhdr ch; + struct sctp_init init; +} SCTP_PACKED; + +struct sctp_init_msg { + struct sctphdr sh; + struct sctp_init_chunk msg; +} SCTP_PACKED; + +/* ... used for both INIT and INIT ACK */ +#define sctp_init_ack sctp_init +#define sctp_init_ack_chunk sctp_init_chunk +#define sctp_init_ack_msg sctp_init_msg + +/* Selective Ack (SACK) */ +struct sctp_gap_ack_block { + uint16_t start; /* Gap Ack block start */ + uint16_t end; /* Gap Ack block end */ +} SCTP_PACKED; + +struct sctp_sack { + uint32_t cum_tsn_ack; /* cumulative TSN Ack */ + uint32_t a_rwnd; /* updated a_rwnd of sender */ + uint16_t num_gap_ack_blks; /* number of Gap Ack blocks */ + uint16_t num_dup_tsns; /* number of duplicate TSNs */ + /* struct sctp_gap_ack_block's follow */ + /* uint32_t duplicate_tsn's follow */ +} SCTP_PACKED; + +struct sctp_sack_chunk { + struct sctp_chunkhdr ch; + struct sctp_sack sack; +} SCTP_PACKED; + +struct sctp_nr_sack { + uint32_t cum_tsn_ack; /* cumulative TSN Ack */ + uint32_t a_rwnd; /* updated a_rwnd of sender */ + uint16_t num_gap_ack_blks; /* number of Gap Ack blocks */ + uint16_t num_nr_gap_ack_blks; /* number of NR Gap Ack blocks */ + uint16_t num_dup_tsns; /* number of duplicate TSNs */ + uint16_t reserved; /* not currently used*/ + /* struct sctp_gap_ack_block's follow */ + /* uint32_t duplicate_tsn's follow */ +} SCTP_PACKED; + +struct sctp_nr_sack_chunk { + struct sctp_chunkhdr ch; + struct sctp_nr_sack nr_sack; +} SCTP_PACKED; + +/* Heartbeat Request (HEARTBEAT) */ +struct sctp_heartbeat { + struct sctp_heartbeat_info_param hb_info; +} SCTP_PACKED; + +struct sctp_heartbeat_chunk { + struct sctp_chunkhdr ch; + struct sctp_heartbeat heartbeat; +} SCTP_PACKED; + +/* ... used for Heartbeat Ack (HEARTBEAT ACK) */ +#define sctp_heartbeat_ack sctp_heartbeat +#define sctp_heartbeat_ack_chunk sctp_heartbeat_chunk + +/* Abort Asssociation (ABORT) */ +struct sctp_abort_chunk { + struct sctp_chunkhdr ch; + /* optional error cause may follow */ +} SCTP_PACKED; + +struct sctp_abort_msg { + struct sctphdr sh; + struct sctp_abort_chunk msg; +} SCTP_PACKED; + +/* Shutdown Association (SHUTDOWN) */ +struct sctp_shutdown_chunk { + struct sctp_chunkhdr ch; + uint32_t cumulative_tsn_ack; +} SCTP_PACKED; + +/* Shutdown Acknowledgment (SHUTDOWN ACK) */ +struct sctp_shutdown_ack_chunk { + struct sctp_chunkhdr ch; +} SCTP_PACKED; + +/* Operation Error (ERROR) */ +struct sctp_error_chunk { + struct sctp_chunkhdr ch; + /* optional error causes follow */ +} SCTP_PACKED; + +/* Cookie Echo (COOKIE ECHO) */ +struct sctp_cookie_echo_chunk { + struct sctp_chunkhdr ch; + struct sctp_state_cookie cookie; +} SCTP_PACKED; + +/* Cookie Acknowledgment (COOKIE ACK) */ +struct sctp_cookie_ack_chunk { + struct sctp_chunkhdr ch; +} SCTP_PACKED; + +/* Explicit Congestion Notification Echo (ECNE) */ +struct old_sctp_ecne_chunk { + struct sctp_chunkhdr ch; + uint32_t tsn; +} SCTP_PACKED; + +struct sctp_ecne_chunk { + struct sctp_chunkhdr ch; + uint32_t tsn; + uint32_t num_pkts_since_cwr; +} SCTP_PACKED; + +/* Congestion Window Reduced (CWR) */ +struct sctp_cwr_chunk { + struct sctp_chunkhdr ch; + uint32_t tsn; +} SCTP_PACKED; + +/* Shutdown Complete (SHUTDOWN COMPLETE) */ +struct sctp_shutdown_complete_chunk { + struct sctp_chunkhdr ch; +} SCTP_PACKED; + +struct sctp_adaptation_layer_indication { + struct sctp_paramhdr ph; + uint32_t indication; +} SCTP_PACKED; + +/* + * draft-ietf-tsvwg-addip-sctp + */ +/* Address/Stream Configuration Change (ASCONF) */ +struct sctp_asconf_chunk { + struct sctp_chunkhdr ch; + uint32_t serial_number; + /* lookup address parameter (mandatory) */ + /* asconf parameters follow */ +} SCTP_PACKED; + +/* Address/Stream Configuration Acknowledge (ASCONF ACK) */ +struct sctp_asconf_ack_chunk { + struct sctp_chunkhdr ch; + uint32_t serial_number; + /* asconf parameters follow */ +} SCTP_PACKED; + +/* draft-ietf-tsvwg-prsctp */ +/* Forward Cumulative TSN (FORWARD TSN) */ +struct sctp_forward_tsn_chunk { + struct sctp_chunkhdr ch; + uint32_t new_cumulative_tsn; + /* stream/sequence pairs (sctp_strseq) follow */ +} SCTP_PACKED; + +struct sctp_strseq { + uint16_t sid; + uint16_t ssn; +} SCTP_PACKED; + +struct sctp_strseq_mid { + uint16_t sid; + uint16_t flags; + uint32_t mid; +}; + +struct sctp_forward_tsn_msg { + struct sctphdr sh; + struct sctp_forward_tsn_chunk msg; +} SCTP_PACKED; + +/* should be a multiple of 4 - 1 aka 3/7/11 etc. */ + +#define SCTP_NUM_DB_TO_VERIFY 31 + +struct sctp_chunk_desc { + uint8_t chunk_type; + uint8_t data_bytes[SCTP_NUM_DB_TO_VERIFY]; + uint32_t tsn_ifany; +} SCTP_PACKED; + +struct sctp_pktdrop_chunk { + struct sctp_chunkhdr ch; + uint32_t bottle_bw; + uint32_t current_onq; + uint16_t trunc_len; + uint16_t reserved; + uint8_t data[]; +} SCTP_PACKED; + +/**********STREAM RESET STUFF ******************/ + +struct sctp_stream_reset_request { + struct sctp_paramhdr ph; + uint32_t request_seq; +} SCTP_PACKED; + +struct sctp_stream_reset_out_request { + struct sctp_paramhdr ph; + uint32_t request_seq; /* monotonically increasing seq no */ + uint32_t response_seq; /* if a response, the resp seq no */ + uint32_t send_reset_at_tsn; /* last TSN I assigned outbound */ + uint16_t list_of_streams[]; /* if not all list of streams */ +} SCTP_PACKED; + +struct sctp_stream_reset_in_request { + struct sctp_paramhdr ph; + uint32_t request_seq; + uint16_t list_of_streams[]; /* if not all list of streams */ +} SCTP_PACKED; + +struct sctp_stream_reset_tsn_request { + struct sctp_paramhdr ph; + uint32_t request_seq; +} SCTP_PACKED; + +struct sctp_stream_reset_response { + struct sctp_paramhdr ph; + uint32_t response_seq; /* if a response, the resp seq no */ + uint32_t result; +} SCTP_PACKED; + +struct sctp_stream_reset_response_tsn { + struct sctp_paramhdr ph; + uint32_t response_seq; /* if a response, the resp seq no */ + uint32_t result; + uint32_t senders_next_tsn; + uint32_t receivers_next_tsn; +} SCTP_PACKED; + +struct sctp_stream_reset_add_strm { + struct sctp_paramhdr ph; + uint32_t request_seq; + uint16_t number_of_streams; + uint16_t reserved; +} SCTP_PACKED; + +#define SCTP_STREAM_RESET_RESULT_NOTHING_TO_DO 0x00000000 /* XXX: unused */ +#define SCTP_STREAM_RESET_RESULT_PERFORMED 0x00000001 +#define SCTP_STREAM_RESET_RESULT_DENIED 0x00000002 +#define SCTP_STREAM_RESET_RESULT_ERR__WRONG_SSN 0x00000003 /* XXX: unused */ +#define SCTP_STREAM_RESET_RESULT_ERR_IN_PROGRESS 0x00000004 +#define SCTP_STREAM_RESET_RESULT_ERR_BAD_SEQNO 0x00000005 +#define SCTP_STREAM_RESET_RESULT_IN_PROGRESS 0x00000006 /* XXX: unused */ + +/* + * convience structures, note that if you are making a request for specific + * streams then the request will need to be an overlay structure. + */ + +struct sctp_stream_reset_tsn_req { + struct sctp_chunkhdr ch; + struct sctp_stream_reset_tsn_request sr_req; +} SCTP_PACKED; + +struct sctp_stream_reset_resp { + struct sctp_chunkhdr ch; + struct sctp_stream_reset_response sr_resp; +} SCTP_PACKED; + +/* respone only valid with a TSN request */ +struct sctp_stream_reset_resp_tsn { + struct sctp_chunkhdr ch; + struct sctp_stream_reset_response_tsn sr_resp; +} SCTP_PACKED; + +/****************************************************/ + +/* + * Authenticated chunks support draft-ietf-tsvwg-sctp-auth + */ + +/* Should we make the max be 32? */ +#define SCTP_RANDOM_MAX_SIZE 256 +struct sctp_auth_random { + struct sctp_paramhdr ph;/* type = 0x8002 */ + uint8_t random_data[]; +} SCTP_PACKED; + +struct sctp_auth_chunk_list { + struct sctp_paramhdr ph;/* type = 0x8003 */ + uint8_t chunk_types[]; +} SCTP_PACKED; + +struct sctp_auth_hmac_algo { + struct sctp_paramhdr ph;/* type = 0x8004 */ + uint16_t hmac_ids[]; +} SCTP_PACKED; + +struct sctp_auth_chunk { + struct sctp_chunkhdr ch; + uint16_t shared_key_id; + uint16_t hmac_id; + uint8_t hmac[]; +} SCTP_PACKED; + +/* + * we pre-reserve enough room for a ECNE or CWR AND a SACK with no missing + * pieces. If ENCE is missing we could have a couple of blocks. This way we + * optimize so we MOST likely can bundle a SACK/ECN with the smallest size + * data chunk I will split into. We could increase throughput slightly by + * taking out these two but the 24-sack/8-CWR i.e. 32 bytes I pre-reserve I + * feel is worth it for now. + */ +#ifndef SCTP_MAX_OVERHEAD +#ifdef INET6 +#define SCTP_MAX_OVERHEAD (sizeof(struct sctp_data_chunk) + \ + sizeof(struct sctphdr) + \ + sizeof(struct sctp_ecne_chunk) + \ + sizeof(struct sctp_sack_chunk) + \ + sizeof(struct ip6_hdr)) + +#define SCTP_MED_OVERHEAD (sizeof(struct sctp_data_chunk) + \ + sizeof(struct sctphdr) + \ + sizeof(struct ip6_hdr)) + +#define SCTP_MIN_OVERHEAD (sizeof(struct ip6_hdr) + \ + sizeof(struct sctphdr)) + +#else +#define SCTP_MAX_OVERHEAD (sizeof(struct sctp_data_chunk) + \ + sizeof(struct sctphdr) + \ + sizeof(struct sctp_ecne_chunk) + \ + sizeof(struct sctp_sack_chunk) + \ + sizeof(struct ip)) + +#define SCTP_MED_OVERHEAD (sizeof(struct sctp_data_chunk) + \ + sizeof(struct sctphdr) + \ + sizeof(struct ip)) + +#define SCTP_MIN_OVERHEAD (sizeof(struct ip) + \ + sizeof(struct sctphdr)) + +#endif /* INET6 */ +#endif /* !SCTP_MAX_OVERHEAD */ + +#define SCTP_MED_V4_OVERHEAD (sizeof(struct sctp_data_chunk) + \ + sizeof(struct sctphdr) + \ + sizeof(struct ip)) + +#define SCTP_MIN_V4_OVERHEAD (sizeof(struct ip) + \ + sizeof(struct sctphdr)) + +#if defined(_WIN32) && !defined(__Userspace__) +#include <packoff.h> +#endif +#if defined(_WIN32) && defined(__Userspace__) +#pragma pack(pop) +#endif +#undef SCTP_PACKED +#endif /* !__sctp_header_h__ */ diff --git a/netwerk/sctp/src/netinet/sctp_indata.c b/netwerk/sctp/src/netinet/sctp_indata.c new file mode 100644 index 0000000000..cc8aed3410 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_indata.c @@ -0,0 +1,5798 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/proc.h> +#endif +#include <netinet/sctp_var.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_header.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_uio.h> +#include <netinet/sctp_auth.h> +#include <netinet/sctp_timer.h> +#include <netinet/sctp_asconf.h> +#include <netinet/sctp_indata.h> +#include <netinet/sctp_bsd_addr.h> +#include <netinet/sctp_input.h> +#include <netinet/sctp_crc32.h> +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <netinet/sctp_lock_bsd.h> +#endif +/* + * NOTES: On the outbound side of things I need to check the sack timer to + * see if I should generate a sack into the chunk queue (if I have data to + * send that is and will be sending it .. for bundling. + * + * The callback in sctp_usrreq.c will get called when the socket is read from. + * This will cause sctp_service_queues() to get called on the top entry in + * the list. + */ +static uint32_t +sctp_add_chk_to_control(struct sctp_queued_to_read *control, + struct sctp_stream_in *strm, + struct sctp_tcb *stcb, + struct sctp_association *asoc, + struct sctp_tmit_chunk *chk, int hold_rlock); + +void +sctp_set_rwnd(struct sctp_tcb *stcb, struct sctp_association *asoc) +{ + asoc->my_rwnd = sctp_calc_rwnd(stcb, asoc); +} + +/* Calculate what the rwnd would be */ +uint32_t +sctp_calc_rwnd(struct sctp_tcb *stcb, struct sctp_association *asoc) +{ + uint32_t calc = 0; + + /* + * This is really set wrong with respect to a 1-2-m socket. Since + * the sb_cc is the count that everyone as put up. When we re-write + * sctp_soreceive then we will fix this so that ONLY this + * associations data is taken into account. + */ + if (stcb->sctp_socket == NULL) { + return (calc); + } + + KASSERT(asoc->cnt_on_reasm_queue > 0 || asoc->size_on_reasm_queue == 0, + ("size_on_reasm_queue is %u", asoc->size_on_reasm_queue)); + KASSERT(asoc->cnt_on_all_streams > 0 || asoc->size_on_all_streams == 0, + ("size_on_all_streams is %u", asoc->size_on_all_streams)); + if (stcb->asoc.sb_cc == 0 && + asoc->cnt_on_reasm_queue == 0 && + asoc->cnt_on_all_streams == 0) { + /* Full rwnd granted */ + calc = max(SCTP_SB_LIMIT_RCV(stcb->sctp_socket), SCTP_MINIMAL_RWND); + return (calc); + } + /* get actual space */ + calc = (uint32_t) sctp_sbspace(&stcb->asoc, &stcb->sctp_socket->so_rcv); + /* + * take out what has NOT been put on socket queue and we yet hold + * for putting up. + */ + calc = sctp_sbspace_sub(calc, (uint32_t)(asoc->size_on_reasm_queue + + asoc->cnt_on_reasm_queue * MSIZE)); + calc = sctp_sbspace_sub(calc, (uint32_t)(asoc->size_on_all_streams + + asoc->cnt_on_all_streams * MSIZE)); + if (calc == 0) { + /* out of space */ + return (calc); + } + + /* what is the overhead of all these rwnd's */ + calc = sctp_sbspace_sub(calc, stcb->asoc.my_rwnd_control_len); + /* If the window gets too small due to ctrl-stuff, reduce it + * to 1, even it is 0. SWS engaged + */ + if (calc < stcb->asoc.my_rwnd_control_len) { + calc = 1; + } + return (calc); +} + +/* + * Build out our readq entry based on the incoming packet. + */ +struct sctp_queued_to_read * +sctp_build_readq_entry(struct sctp_tcb *stcb, + struct sctp_nets *net, + uint32_t tsn, uint32_t ppid, + uint32_t context, uint16_t sid, + uint32_t mid, uint8_t flags, + struct mbuf *dm) +{ + struct sctp_queued_to_read *read_queue_e = NULL; + + sctp_alloc_a_readq(stcb, read_queue_e); + if (read_queue_e == NULL) { + goto failed_build; + } + memset(read_queue_e, 0, sizeof(struct sctp_queued_to_read)); + read_queue_e->sinfo_stream = sid; + read_queue_e->sinfo_flags = (flags << 8); + read_queue_e->sinfo_ppid = ppid; + read_queue_e->sinfo_context = context; + read_queue_e->sinfo_tsn = tsn; + read_queue_e->sinfo_cumtsn = tsn; + read_queue_e->sinfo_assoc_id = sctp_get_associd(stcb); + read_queue_e->mid = mid; + read_queue_e->top_fsn = read_queue_e->fsn_included = 0xffffffff; + TAILQ_INIT(&read_queue_e->reasm); + read_queue_e->whoFrom = net; + atomic_add_int(&net->ref_count, 1); + read_queue_e->data = dm; + read_queue_e->stcb = stcb; + read_queue_e->port_from = stcb->rport; + if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + read_queue_e->do_not_ref_stcb = 1; + } +failed_build: + return (read_queue_e); +} + +struct mbuf * +sctp_build_ctl_nchunk(struct sctp_inpcb *inp, struct sctp_sndrcvinfo *sinfo) +{ + struct sctp_extrcvinfo *seinfo; + struct sctp_sndrcvinfo *outinfo; + struct sctp_rcvinfo *rcvinfo; + struct sctp_nxtinfo *nxtinfo; +#if defined(_WIN32) + WSACMSGHDR *cmh; +#else + struct cmsghdr *cmh; +#endif + struct mbuf *ret; + int len; + int use_extended; + int provide_nxt; + + if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT) && + sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVRCVINFO) && + sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVNXTINFO)) { + /* user does not want any ancillary data */ + return (NULL); + } + + len = 0; + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVRCVINFO)) { + len += CMSG_SPACE(sizeof(struct sctp_rcvinfo)); + } + seinfo = (struct sctp_extrcvinfo *)sinfo; + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVNXTINFO) && + (seinfo->serinfo_next_flags & SCTP_NEXT_MSG_AVAIL)) { + provide_nxt = 1; + len += CMSG_SPACE(sizeof(struct sctp_nxtinfo)); + } else { + provide_nxt = 0; + } + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) { + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO)) { + use_extended = 1; + len += CMSG_SPACE(sizeof(struct sctp_extrcvinfo)); + } else { + use_extended = 0; + len += CMSG_SPACE(sizeof(struct sctp_sndrcvinfo)); + } + } else { + use_extended = 0; + } + + ret = sctp_get_mbuf_for_msg(len, 0, M_NOWAIT, 1, MT_DATA); + if (ret == NULL) { + /* No space */ + return (ret); + } + SCTP_BUF_LEN(ret) = 0; + + /* We need a CMSG header followed by the struct */ +#if defined(_WIN32) + cmh = mtod(ret, WSACMSGHDR *); +#else + cmh = mtod(ret, struct cmsghdr *); +#endif + /* + * Make sure that there is no un-initialized padding between + * the cmsg header and cmsg data and after the cmsg data. + */ + memset(cmh, 0, len); + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVRCVINFO)) { + cmh->cmsg_level = IPPROTO_SCTP; + cmh->cmsg_len = CMSG_LEN(sizeof(struct sctp_rcvinfo)); + cmh->cmsg_type = SCTP_RCVINFO; + rcvinfo = (struct sctp_rcvinfo *)CMSG_DATA(cmh); + rcvinfo->rcv_sid = sinfo->sinfo_stream; + rcvinfo->rcv_ssn = sinfo->sinfo_ssn; + rcvinfo->rcv_flags = sinfo->sinfo_flags; + rcvinfo->rcv_ppid = sinfo->sinfo_ppid; + rcvinfo->rcv_tsn = sinfo->sinfo_tsn; + rcvinfo->rcv_cumtsn = sinfo->sinfo_cumtsn; + rcvinfo->rcv_context = sinfo->sinfo_context; + rcvinfo->rcv_assoc_id = sinfo->sinfo_assoc_id; +#if defined(_WIN32) + cmh = (WSACMSGHDR *)((caddr_t)cmh + CMSG_SPACE(sizeof(struct sctp_rcvinfo))); +#else + cmh = (struct cmsghdr *)((caddr_t)cmh + CMSG_SPACE(sizeof(struct sctp_rcvinfo))); +#endif + SCTP_BUF_LEN(ret) += CMSG_SPACE(sizeof(struct sctp_rcvinfo)); + } + if (provide_nxt) { + cmh->cmsg_level = IPPROTO_SCTP; + cmh->cmsg_len = CMSG_LEN(sizeof(struct sctp_nxtinfo)); + cmh->cmsg_type = SCTP_NXTINFO; + nxtinfo = (struct sctp_nxtinfo *)CMSG_DATA(cmh); + nxtinfo->nxt_sid = seinfo->serinfo_next_stream; + nxtinfo->nxt_flags = 0; + if (seinfo->serinfo_next_flags & SCTP_NEXT_MSG_IS_UNORDERED) { + nxtinfo->nxt_flags |= SCTP_UNORDERED; + } + if (seinfo->serinfo_next_flags & SCTP_NEXT_MSG_IS_NOTIFICATION) { + nxtinfo->nxt_flags |= SCTP_NOTIFICATION; + } + if (seinfo->serinfo_next_flags & SCTP_NEXT_MSG_ISCOMPLETE) { + nxtinfo->nxt_flags |= SCTP_COMPLETE; + } + nxtinfo->nxt_ppid = seinfo->serinfo_next_ppid; + nxtinfo->nxt_length = seinfo->serinfo_next_length; + nxtinfo->nxt_assoc_id = seinfo->serinfo_next_aid; +#if defined(_WIN32) + cmh = (WSACMSGHDR *)((caddr_t)cmh + CMSG_SPACE(sizeof(struct sctp_nxtinfo))); +#else + cmh = (struct cmsghdr *)((caddr_t)cmh + CMSG_SPACE(sizeof(struct sctp_nxtinfo))); +#endif + SCTP_BUF_LEN(ret) += CMSG_SPACE(sizeof(struct sctp_nxtinfo)); + } + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) { + cmh->cmsg_level = IPPROTO_SCTP; + outinfo = (struct sctp_sndrcvinfo *)CMSG_DATA(cmh); + if (use_extended) { + cmh->cmsg_len = CMSG_LEN(sizeof(struct sctp_extrcvinfo)); + cmh->cmsg_type = SCTP_EXTRCV; + memcpy(outinfo, sinfo, sizeof(struct sctp_extrcvinfo)); + SCTP_BUF_LEN(ret) += CMSG_SPACE(sizeof(struct sctp_extrcvinfo)); + } else { + cmh->cmsg_len = CMSG_LEN(sizeof(struct sctp_sndrcvinfo)); + cmh->cmsg_type = SCTP_SNDRCV; + *outinfo = *sinfo; + SCTP_BUF_LEN(ret) += CMSG_SPACE(sizeof(struct sctp_sndrcvinfo)); + } + } + return (ret); +} + +static void +sctp_mark_non_revokable(struct sctp_association *asoc, uint32_t tsn) +{ + uint32_t gap, i; + int in_r, in_nr; + + if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) { + return; + } + if (SCTP_TSN_GE(asoc->cumulative_tsn, tsn)) { + /* + * This tsn is behind the cum ack and thus we don't + * need to worry about it being moved from one to the other. + */ + return; + } + SCTP_CALC_TSN_TO_GAP(gap, tsn, asoc->mapping_array_base_tsn); + in_r = SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap); + in_nr = SCTP_IS_TSN_PRESENT(asoc->nr_mapping_array, gap); + KASSERT(in_r || in_nr, ("%s: Things are really messed up now", __func__)); + if (!in_nr) { + SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap); + if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) { + asoc->highest_tsn_inside_nr_map = tsn; + } + } + if (in_r) { + SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); + if (tsn == asoc->highest_tsn_inside_map) { + /* We must back down to see what the new highest is. */ + for (i = tsn - 1; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) { + SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn); + if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) { + asoc->highest_tsn_inside_map = i; + break; + } + } + if (!SCTP_TSN_GE(i, asoc->mapping_array_base_tsn)) { + asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1; + } + } + } +} + +static int +sctp_place_control_in_stream(struct sctp_stream_in *strm, + struct sctp_association *asoc, + struct sctp_queued_to_read *control) +{ + struct sctp_queued_to_read *at; + struct sctp_readhead *q; + uint8_t flags, unordered; + + flags = (control->sinfo_flags >> 8); + unordered = flags & SCTP_DATA_UNORDERED; + if (unordered) { + q = &strm->uno_inqueue; + if (asoc->idata_supported == 0) { + if (!TAILQ_EMPTY(q)) { + /* Only one stream can be here in old style -- abort */ + return (-1); + } + TAILQ_INSERT_TAIL(q, control, next_instrm); + control->on_strm_q = SCTP_ON_UNORDERED; + return (0); + } + } else { + q = &strm->inqueue; + } + if ((flags & SCTP_DATA_NOT_FRAG) == SCTP_DATA_NOT_FRAG) { + control->end_added = 1; + control->first_frag_seen = 1; + control->last_frag_seen = 1; + } + if (TAILQ_EMPTY(q)) { + /* Empty queue */ + TAILQ_INSERT_HEAD(q, control, next_instrm); + if (unordered) { + control->on_strm_q = SCTP_ON_UNORDERED; + } else { + control->on_strm_q = SCTP_ON_ORDERED; + } + return (0); + } else { + TAILQ_FOREACH(at, q, next_instrm) { + if (SCTP_MID_GT(asoc->idata_supported, at->mid, control->mid)) { + /* + * one in queue is bigger than the + * new one, insert before this one + */ + TAILQ_INSERT_BEFORE(at, control, next_instrm); + if (unordered) { + control->on_strm_q = SCTP_ON_UNORDERED; + } else { + control->on_strm_q = SCTP_ON_ORDERED; + } + break; + } else if (SCTP_MID_EQ(asoc->idata_supported, at->mid, control->mid)) { + /* + * Gak, He sent me a duplicate msg + * id number?? return -1 to abort. + */ + return (-1); + } else { + if (TAILQ_NEXT(at, next_instrm) == NULL) { + /* + * We are at the end, insert + * it after this one + */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { + sctp_log_strm_del(control, at, + SCTP_STR_LOG_FROM_INSERT_TL); + } + TAILQ_INSERT_AFTER(q, at, control, next_instrm); + if (unordered) { + control->on_strm_q = SCTP_ON_UNORDERED; + } else { + control->on_strm_q = SCTP_ON_ORDERED; + } + break; + } + } + } + } + return (0); +} + +static void +sctp_abort_in_reasm(struct sctp_tcb *stcb, + struct sctp_queued_to_read *control, + struct sctp_tmit_chunk *chk, + int *abort_flag, int opspot) +{ + char msg[SCTP_DIAG_INFO_LEN]; + struct mbuf *oper; + + if (stcb->asoc.idata_supported) { + SCTP_SNPRINTF(msg, sizeof(msg), + "Reass %x,CF:%x,TSN=%8.8x,SID=%4.4x,FSN=%8.8x,MID:%8.8x", + opspot, + control->fsn_included, + chk->rec.data.tsn, + chk->rec.data.sid, + chk->rec.data.fsn, chk->rec.data.mid); + } else { + SCTP_SNPRINTF(msg, sizeof(msg), + "Reass %x,CI:%x,TSN=%8.8x,SID=%4.4x,FSN=%4.4x,SSN:%4.4x", + opspot, + control->fsn_included, + chk->rec.data.tsn, + chk->rec.data.sid, + chk->rec.data.fsn, + (uint16_t)chk->rec.data.mid); + } + oper = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + sctp_m_freem(chk->data); + chk->data = NULL; + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_1; + sctp_abort_an_association(stcb->sctp_ep, stcb, oper, false, SCTP_SO_NOT_LOCKED); + *abort_flag = 1; +} + +static void +sctp_clean_up_control(struct sctp_tcb *stcb, struct sctp_queued_to_read *control) +{ + /* + * The control could not be placed and must be cleaned. + */ + struct sctp_tmit_chunk *chk, *nchk; + TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) { + TAILQ_REMOVE(&control->reasm, chk, sctp_next); + if (chk->data) + sctp_m_freem(chk->data); + chk->data = NULL; + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + } + sctp_free_remote_addr(control->whoFrom); + if (control->data) { + sctp_m_freem(control->data); + control->data = NULL; + } + sctp_free_a_readq(stcb, control); +} + +/* + * Queue the chunk either right into the socket buffer if it is the next one + * to go OR put it in the correct place in the delivery queue. If we do + * append to the so_buf, keep doing so until we are out of order as + * long as the control's entered are non-fragmented. + */ +static void +sctp_queue_data_to_stream(struct sctp_tcb *stcb, + struct sctp_association *asoc, + struct sctp_queued_to_read *control, int *abort_flag, int *need_reasm) +{ + /* + * FIX-ME maybe? What happens when the ssn wraps? If we are getting + * all the data in one stream this could happen quite rapidly. One + * could use the TSN to keep track of things, but this scheme breaks + * down in the other type of stream usage that could occur. Send a + * single msg to stream 0, send 4Billion messages to stream 1, now + * send a message to stream 0. You have a situation where the TSN + * has wrapped but not in the stream. Is this worth worrying about + * or should we just change our queue sort at the bottom to be by + * TSN. + * + * Could it also be legal for a peer to send ssn 1 with TSN 2 and ssn 2 + * with TSN 1? If the peer is doing some sort of funky TSN/SSN + * assignment this could happen... and I don't see how this would be + * a violation. So for now I am undecided an will leave the sort by + * SSN alone. Maybe a hybrid approach is the answer + * + */ + struct sctp_queued_to_read *at; + int queue_needed; + uint32_t nxt_todel; + struct mbuf *op_err; + struct sctp_stream_in *strm; + char msg[SCTP_DIAG_INFO_LEN]; + + strm = &asoc->strmin[control->sinfo_stream]; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { + sctp_log_strm_del(control, NULL, SCTP_STR_LOG_FROM_INTO_STRD); + } + if (SCTP_MID_GT((asoc->idata_supported), strm->last_mid_delivered, control->mid)) { + /* The incoming sseq is behind where we last delivered? */ + SCTPDBG(SCTP_DEBUG_INDATA1, "Duplicate S-SEQ: %u delivered: %u from peer, Abort association\n", + strm->last_mid_delivered, control->mid); + /* + * throw it in the stream so it gets cleaned up in + * association destruction + */ + TAILQ_INSERT_HEAD(&strm->inqueue, control, next_instrm); + if (asoc->idata_supported) { + SCTP_SNPRINTF(msg, sizeof(msg), "Delivered MID=%8.8x, got TSN=%8.8x, SID=%4.4x, MID=%8.8x", + strm->last_mid_delivered, control->sinfo_tsn, + control->sinfo_stream, control->mid); + } else { + SCTP_SNPRINTF(msg, sizeof(msg), "Delivered SSN=%4.4x, got TSN=%8.8x, SID=%4.4x, SSN=%4.4x", + (uint16_t)strm->last_mid_delivered, + control->sinfo_tsn, + control->sinfo_stream, + (uint16_t)control->mid); + } + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_2; + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + *abort_flag = 1; + return; + } + queue_needed = 1; + asoc->size_on_all_streams += control->length; + sctp_ucount_incr(asoc->cnt_on_all_streams); + nxt_todel = strm->last_mid_delivered + 1; + if (SCTP_MID_EQ(asoc->idata_supported, nxt_todel, control->mid)) { +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; + + so = SCTP_INP_SO(stcb->sctp_ep); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + SCTP_SOCKET_UNLOCK(so, 1); + return; + } +#endif + /* can be delivered right away? */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { + sctp_log_strm_del(control, NULL, SCTP_STR_LOG_FROM_IMMED_DEL); + } + /* EY it wont be queued if it could be delivered directly */ + queue_needed = 0; + if (asoc->size_on_all_streams >= control->length) { + asoc->size_on_all_streams -= control->length; + } else { +#ifdef INVARIANTS + panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); +#else + asoc->size_on_all_streams = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_all_streams); + strm->last_mid_delivered++; + sctp_mark_non_revokable(asoc, control->sinfo_tsn); + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, + SCTP_READ_LOCK_NOT_HELD, SCTP_SO_LOCKED); + TAILQ_FOREACH_SAFE(control, &strm->inqueue, next_instrm, at) { + /* all delivered */ + nxt_todel = strm->last_mid_delivered + 1; + if (SCTP_MID_EQ(asoc->idata_supported, nxt_todel, control->mid) && + (((control->sinfo_flags >> 8) & SCTP_DATA_NOT_FRAG) == SCTP_DATA_NOT_FRAG)) { + if (control->on_strm_q == SCTP_ON_ORDERED) { + TAILQ_REMOVE(&strm->inqueue, control, next_instrm); + if (asoc->size_on_all_streams >= control->length) { + asoc->size_on_all_streams -= control->length; + } else { +#ifdef INVARIANTS + panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); +#else + asoc->size_on_all_streams = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_all_streams); +#ifdef INVARIANTS + } else { + panic("Huh control: %p is on_strm_q: %d", + control, control->on_strm_q); +#endif + } + control->on_strm_q = 0; + strm->last_mid_delivered++; + /* + * We ignore the return of deliver_data here + * since we always can hold the chunk on the + * d-queue. And we have a finite number that + * can be delivered from the strq. + */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { + sctp_log_strm_del(control, NULL, + SCTP_STR_LOG_FROM_IMMED_DEL); + } + sctp_mark_non_revokable(asoc, control->sinfo_tsn); + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, + SCTP_READ_LOCK_NOT_HELD, + SCTP_SO_LOCKED); + continue; + } else if (SCTP_MID_EQ(asoc->idata_supported, nxt_todel, control->mid)) { + *need_reasm = 1; + } + break; + } +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } + if (queue_needed) { + /* + * Ok, we did not deliver this guy, find the correct place + * to put it on the queue. + */ + if (sctp_place_control_in_stream(strm, asoc, control)) { + SCTP_SNPRINTF(msg, sizeof(msg), + "Queue to str MID: %u duplicate", control->mid); + sctp_clean_up_control(stcb, control); + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_3; + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + *abort_flag = 1; + } + } +} + +static void +sctp_setup_tail_pointer(struct sctp_queued_to_read *control) +{ + struct mbuf *m, *prev = NULL; + struct sctp_tcb *stcb; + + stcb = control->stcb; + control->held_length = 0; + control->length = 0; + m = control->data; + while (m) { + if (SCTP_BUF_LEN(m) == 0) { + /* Skip mbufs with NO length */ + if (prev == NULL) { + /* First one */ + control->data = sctp_m_free(m); + m = control->data; + } else { + SCTP_BUF_NEXT(prev) = sctp_m_free(m); + m = SCTP_BUF_NEXT(prev); + } + if (m == NULL) { + control->tail_mbuf = prev; + } + continue; + } + prev = m; + atomic_add_int(&control->length, SCTP_BUF_LEN(m)); + if (control->on_read_q) { + /* + * On read queue so we must increment the + * SB stuff, we assume caller has done any locks of SB. + */ + sctp_sballoc(stcb, &stcb->sctp_socket->so_rcv, m); + } + m = SCTP_BUF_NEXT(m); + } + if (prev) { + control->tail_mbuf = prev; + } +} + +static void +sctp_add_to_tail_pointer(struct sctp_queued_to_read *control, struct mbuf *m, uint32_t *added) +{ + struct mbuf *prev=NULL; + struct sctp_tcb *stcb; + + stcb = control->stcb; + if (stcb == NULL) { +#ifdef INVARIANTS + panic("Control broken"); +#else + return; +#endif + } + if (control->tail_mbuf == NULL) { + /* TSNH */ + sctp_m_freem(control->data); + control->data = m; + sctp_setup_tail_pointer(control); + return; + } + control->tail_mbuf->m_next = m; + while (m) { + if (SCTP_BUF_LEN(m) == 0) { + /* Skip mbufs with NO length */ + if (prev == NULL) { + /* First one */ + control->tail_mbuf->m_next = sctp_m_free(m); + m = control->tail_mbuf->m_next; + } else { + SCTP_BUF_NEXT(prev) = sctp_m_free(m); + m = SCTP_BUF_NEXT(prev); + } + if (m == NULL) { + control->tail_mbuf = prev; + } + continue; + } + prev = m; + if (control->on_read_q) { + /* + * On read queue so we must increment the + * SB stuff, we assume caller has done any locks of SB. + */ + sctp_sballoc(stcb, &stcb->sctp_socket->so_rcv, m); + } + *added += SCTP_BUF_LEN(m); + atomic_add_int(&control->length, SCTP_BUF_LEN(m)); + m = SCTP_BUF_NEXT(m); + } + if (prev) { + control->tail_mbuf = prev; + } +} + +static void +sctp_build_readq_entry_from_ctl(struct sctp_queued_to_read *nc, struct sctp_queued_to_read *control) +{ + memset(nc, 0, sizeof(struct sctp_queued_to_read)); + nc->sinfo_stream = control->sinfo_stream; + nc->mid = control->mid; + TAILQ_INIT(&nc->reasm); + nc->top_fsn = control->top_fsn; + nc->mid = control->mid; + nc->sinfo_flags = control->sinfo_flags; + nc->sinfo_ppid = control->sinfo_ppid; + nc->sinfo_context = control->sinfo_context; + nc->fsn_included = 0xffffffff; + nc->sinfo_tsn = control->sinfo_tsn; + nc->sinfo_cumtsn = control->sinfo_cumtsn; + nc->sinfo_assoc_id = control->sinfo_assoc_id; + nc->whoFrom = control->whoFrom; + atomic_add_int(&nc->whoFrom->ref_count, 1); + nc->stcb = control->stcb; + nc->port_from = control->port_from; + nc->do_not_ref_stcb = control->do_not_ref_stcb; +} + +static void +sctp_reset_a_control(struct sctp_queued_to_read *control, + struct sctp_inpcb *inp, uint32_t tsn) +{ + control->fsn_included = tsn; + if (control->on_read_q) { + /* + * We have to purge it from there, + * hopefully this will work :-) + */ + TAILQ_REMOVE(&inp->read_queue, control, next); + control->on_read_q = 0; + } +} + +static int +sctp_handle_old_unordered_data(struct sctp_tcb *stcb, + struct sctp_association *asoc, + struct sctp_stream_in *strm, + struct sctp_queued_to_read *control, + uint32_t pd_point, + int inp_read_lock_held) +{ + /* Special handling for the old un-ordered data chunk. + * All the chunks/TSN's go to mid 0. So + * we have to do the old style watching to see + * if we have it all. If you return one, no other + * control entries on the un-ordered queue will + * be looked at. In theory there should be no others + * entries in reality, unless the guy is sending both + * unordered NDATA and unordered DATA... + */ + struct sctp_tmit_chunk *chk, *lchk, *tchk; + uint32_t fsn; + struct sctp_queued_to_read *nc; + int cnt_added; + + if (control->first_frag_seen == 0) { + /* Nothing we can do, we have not seen the first piece yet */ + return (1); + } + /* Collapse any we can */ + cnt_added = 0; +restart: + fsn = control->fsn_included + 1; + /* Now what can we add? */ + TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, lchk) { + if (chk->rec.data.fsn == fsn) { + /* Ok lets add it */ + sctp_alloc_a_readq(stcb, nc); + if (nc == NULL) { + break; + } + memset(nc, 0, sizeof(struct sctp_queued_to_read)); + TAILQ_REMOVE(&control->reasm, chk, sctp_next); + sctp_add_chk_to_control(control, strm, stcb, asoc, chk, inp_read_lock_held); + fsn++; + cnt_added++; + chk = NULL; + if (control->end_added) { + /* We are done */ + if (!TAILQ_EMPTY(&control->reasm)) { + /* + * Ok we have to move anything left on + * the control queue to a new control. + */ + sctp_build_readq_entry_from_ctl(nc, control); + tchk = TAILQ_FIRST(&control->reasm); + if (tchk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) { + TAILQ_REMOVE(&control->reasm, tchk, sctp_next); + if (asoc->size_on_reasm_queue >= tchk->send_size) { + asoc->size_on_reasm_queue -= tchk->send_size; + } else { +#ifdef INVARIANTS + panic("size_on_reasm_queue = %u smaller than chunk length %u", asoc->size_on_reasm_queue, tchk->send_size); +#else + asoc->size_on_reasm_queue = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_reasm_queue); + nc->first_frag_seen = 1; + nc->fsn_included = tchk->rec.data.fsn; + nc->data = tchk->data; + nc->sinfo_ppid = tchk->rec.data.ppid; + nc->sinfo_tsn = tchk->rec.data.tsn; + sctp_mark_non_revokable(asoc, tchk->rec.data.tsn); + tchk->data = NULL; + sctp_free_a_chunk(stcb, tchk, SCTP_SO_NOT_LOCKED); + sctp_setup_tail_pointer(nc); + tchk = TAILQ_FIRST(&control->reasm); + } + /* Spin the rest onto the queue */ + while (tchk) { + TAILQ_REMOVE(&control->reasm, tchk, sctp_next); + TAILQ_INSERT_TAIL(&nc->reasm, tchk, sctp_next); + tchk = TAILQ_FIRST(&control->reasm); + } + /* Now lets add it to the queue after removing control */ + TAILQ_INSERT_TAIL(&strm->uno_inqueue, nc, next_instrm); + nc->on_strm_q = SCTP_ON_UNORDERED; + if (control->on_strm_q) { + TAILQ_REMOVE(&strm->uno_inqueue, control, next_instrm); + control->on_strm_q = 0; + } + } + if (control->pdapi_started) { + strm->pd_api_started = 0; + control->pdapi_started = 0; + } + if (control->on_strm_q) { + TAILQ_REMOVE(&strm->uno_inqueue, control, next_instrm); + control->on_strm_q = 0; + SCTP_STAT_INCR_COUNTER64(sctps_reasmusrmsgs); + } + if (control->on_read_q == 0) { + sctp_add_to_readq(stcb->sctp_ep, stcb, control, + &stcb->sctp_socket->so_rcv, control->end_added, + inp_read_lock_held, SCTP_SO_NOT_LOCKED); +#if defined(__Userspace__) + } else { + sctp_invoke_recv_callback(stcb->sctp_ep, stcb, control, inp_read_lock_held); +#endif + } + sctp_wakeup_the_read_socket(stcb->sctp_ep, stcb, SCTP_SO_NOT_LOCKED); + if ((nc->first_frag_seen) && !TAILQ_EMPTY(&nc->reasm)) { + /* Switch to the new guy and continue */ + control = nc; + goto restart; + } else { + if (nc->on_strm_q == 0) { + sctp_free_a_readq(stcb, nc); + } + } + return (1); + } else { + sctp_free_a_readq(stcb, nc); + } + } else { + /* Can't add more */ + break; + } + } + if (cnt_added && strm->pd_api_started) { +#if defined(__Userspace__) + sctp_invoke_recv_callback(stcb->sctp_ep, stcb, control, inp_read_lock_held); +#endif + sctp_wakeup_the_read_socket(stcb->sctp_ep, stcb, SCTP_SO_NOT_LOCKED); + } + if ((control->length > pd_point) && (strm->pd_api_started == 0)) { + strm->pd_api_started = 1; + control->pdapi_started = 1; + sctp_add_to_readq(stcb->sctp_ep, stcb, control, + &stcb->sctp_socket->so_rcv, control->end_added, + inp_read_lock_held, SCTP_SO_NOT_LOCKED); + sctp_wakeup_the_read_socket(stcb->sctp_ep, stcb, SCTP_SO_NOT_LOCKED); + return (0); + } else { + return (1); + } +} + +static void +sctp_inject_old_unordered_data(struct sctp_tcb *stcb, + struct sctp_association *asoc, + struct sctp_queued_to_read *control, + struct sctp_tmit_chunk *chk, + int *abort_flag) +{ + struct sctp_tmit_chunk *at; + int inserted; + /* + * Here we need to place the chunk into the control structure + * sorted in the correct order. + */ + if (chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) { + /* Its the very first one. */ + SCTPDBG(SCTP_DEBUG_XXX, + "chunk is a first fsn: %u becomes fsn_included\n", + chk->rec.data.fsn); + at = TAILQ_FIRST(&control->reasm); + if (at && SCTP_TSN_GT(chk->rec.data.fsn, at->rec.data.fsn)) { + /* + * The first chunk in the reassembly is + * a smaller TSN than this one, even though + * this has a first, it must be from a subsequent + * msg. + */ + goto place_chunk; + } + if (control->first_frag_seen) { + /* + * In old un-ordered we can reassembly on + * one control multiple messages. As long + * as the next FIRST is greater then the old + * first (TSN i.e. FSN wise) + */ + struct mbuf *tdata; + uint32_t tmp; + + if (SCTP_TSN_GT(chk->rec.data.fsn, control->fsn_included)) { + /* Easy way the start of a new guy beyond the lowest */ + goto place_chunk; + } + if ((chk->rec.data.fsn == control->fsn_included) || + (control->pdapi_started)) { + /* + * Ok this should not happen, if it does + * we started the pd-api on the higher TSN (since + * the equals part is a TSN failure it must be that). + * + * We are completely hosed in that case since I have + * no way to recover. This really will only happen + * if we can get more TSN's higher before the pd-api-point. + */ + sctp_abort_in_reasm(stcb, control, chk, + abort_flag, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_4); + + return; + } + /* + * Ok we have two firsts and the one we just got + * is smaller than the one we previously placed.. yuck! + * We must swap them out. + */ + /* swap the mbufs */ + tdata = control->data; + control->data = chk->data; + chk->data = tdata; + /* Save the lengths */ + chk->send_size = control->length; + /* Recompute length of control and tail pointer */ + sctp_setup_tail_pointer(control); + /* Fix the FSN included */ + tmp = control->fsn_included; + control->fsn_included = chk->rec.data.fsn; + chk->rec.data.fsn = tmp; + /* Fix the TSN included */ + tmp = control->sinfo_tsn; + control->sinfo_tsn = chk->rec.data.tsn; + chk->rec.data.tsn = tmp; + /* Fix the PPID included */ + tmp = control->sinfo_ppid; + control->sinfo_ppid = chk->rec.data.ppid; + chk->rec.data.ppid = tmp; + /* Fix tail pointer */ + goto place_chunk; + } + control->first_frag_seen = 1; + control->fsn_included = chk->rec.data.fsn; + control->top_fsn = chk->rec.data.fsn; + control->sinfo_tsn = chk->rec.data.tsn; + control->sinfo_ppid = chk->rec.data.ppid; + control->data = chk->data; + sctp_mark_non_revokable(asoc, chk->rec.data.tsn); + chk->data = NULL; + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + sctp_setup_tail_pointer(control); + return; + } +place_chunk: + inserted = 0; + TAILQ_FOREACH(at, &control->reasm, sctp_next) { + if (SCTP_TSN_GT(at->rec.data.fsn, chk->rec.data.fsn)) { + /* + * This one in queue is bigger than the new one, insert + * the new one before at. + */ + asoc->size_on_reasm_queue += chk->send_size; + sctp_ucount_incr(asoc->cnt_on_reasm_queue); + inserted = 1; + TAILQ_INSERT_BEFORE(at, chk, sctp_next); + break; + } else if (at->rec.data.fsn == chk->rec.data.fsn) { + /* + * They sent a duplicate fsn number. This + * really should not happen since the FSN is + * a TSN and it should have been dropped earlier. + */ + sctp_abort_in_reasm(stcb, control, chk, + abort_flag, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_5); + return; + } + } + if (inserted == 0) { + /* Its at the end */ + asoc->size_on_reasm_queue += chk->send_size; + sctp_ucount_incr(asoc->cnt_on_reasm_queue); + control->top_fsn = chk->rec.data.fsn; + TAILQ_INSERT_TAIL(&control->reasm, chk, sctp_next); + } +} + +static int +sctp_deliver_reasm_check(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_in *strm, int inp_read_lock_held) +{ + /* + * Given a stream, strm, see if any of + * the SSN's on it that are fragmented + * are ready to deliver. If so go ahead + * and place them on the read queue. In + * so placing if we have hit the end, then + * we need to remove them from the stream's queue. + */ + struct sctp_queued_to_read *control, *nctl = NULL; + uint32_t next_to_del; + uint32_t pd_point; + int ret = 0; + + if (stcb->sctp_socket) { + pd_point = min(SCTP_SB_LIMIT_RCV(stcb->sctp_socket) >> SCTP_PARTIAL_DELIVERY_SHIFT, + stcb->sctp_ep->partial_delivery_point); + } else { + pd_point = stcb->sctp_ep->partial_delivery_point; + } + control = TAILQ_FIRST(&strm->uno_inqueue); + + if ((control != NULL) && + (asoc->idata_supported == 0)) { + /* Special handling needed for "old" data format */ + if (sctp_handle_old_unordered_data(stcb, asoc, strm, control, pd_point, inp_read_lock_held)) { + goto done_un; + } + } + if (strm->pd_api_started) { + /* Can't add more */ + return (0); + } + while (control) { + SCTPDBG(SCTP_DEBUG_XXX, "Looking at control: %p e(%d) ssn: %u top_fsn: %u inc_fsn: %u -uo\n", + control, control->end_added, control->mid, control->top_fsn, control->fsn_included); + nctl = TAILQ_NEXT(control, next_instrm); + if (control->end_added) { + /* We just put the last bit on */ + if (control->on_strm_q) { +#ifdef INVARIANTS + if (control->on_strm_q != SCTP_ON_UNORDERED) { + panic("Huh control: %p on_q: %d -- not unordered?", + control, control->on_strm_q); + } +#endif + SCTP_STAT_INCR_COUNTER64(sctps_reasmusrmsgs); + TAILQ_REMOVE(&strm->uno_inqueue, control, next_instrm); + if (asoc->size_on_all_streams >= control->length) { + asoc->size_on_all_streams -= control->length; + } else { +#ifdef INVARIANTS + panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); +#else + asoc->size_on_all_streams = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_all_streams); + control->on_strm_q = 0; + } + if (control->on_read_q == 0) { + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, control->end_added, + inp_read_lock_held, SCTP_SO_NOT_LOCKED); + } + } else { + /* Can we do a PD-API for this un-ordered guy? */ + if ((control->length >= pd_point) && (strm->pd_api_started == 0)) { + strm->pd_api_started = 1; + control->pdapi_started = 1; + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, control->end_added, + inp_read_lock_held, SCTP_SO_NOT_LOCKED); + + break; + } + } + control = nctl; + } +done_un: + control = TAILQ_FIRST(&strm->inqueue); + if (strm->pd_api_started) { + /* Can't add more */ + return (0); + } + if (control == NULL) { + return (ret); + } + if (SCTP_MID_EQ(asoc->idata_supported, strm->last_mid_delivered, control->mid)) { + /* Ok the guy at the top was being partially delivered + * completed, so we remove it. Note + * the pd_api flag was taken off when the + * chunk was merged on in sctp_queue_data_for_reasm below. + */ + nctl = TAILQ_NEXT(control, next_instrm); + SCTPDBG(SCTP_DEBUG_XXX, + "Looking at control: %p e(%d) ssn: %u top_fsn: %u inc_fsn: %u (lastdel: %u)- o\n", + control, control->end_added, control->mid, + control->top_fsn, control->fsn_included, + strm->last_mid_delivered); + if (control->end_added) { + if (control->on_strm_q) { +#ifdef INVARIANTS + if (control->on_strm_q != SCTP_ON_ORDERED) { + panic("Huh control: %p on_q: %d -- not ordered?", + control, control->on_strm_q); + } +#endif + SCTP_STAT_INCR_COUNTER64(sctps_reasmusrmsgs); + TAILQ_REMOVE(&strm->inqueue, control, next_instrm); + if (asoc->size_on_all_streams >= control->length) { + asoc->size_on_all_streams -= control->length; + } else { +#ifdef INVARIANTS + panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); +#else + asoc->size_on_all_streams = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_all_streams); + control->on_strm_q = 0; + } + if (strm->pd_api_started && control->pdapi_started) { + control->pdapi_started = 0; + strm->pd_api_started = 0; + } + if (control->on_read_q == 0) { + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, control->end_added, + inp_read_lock_held, SCTP_SO_NOT_LOCKED); + } + control = nctl; + } + } + if (strm->pd_api_started) { + /* Can't add more must have gotten an un-ordered above being partially delivered. */ + return (0); + } +deliver_more: + next_to_del = strm->last_mid_delivered + 1; + if (control) { + SCTPDBG(SCTP_DEBUG_XXX, + "Looking at control: %p e(%d) ssn: %u top_fsn: %u inc_fsn: %u (nxtdel: %u)- o\n", + control, control->end_added, control->mid, control->top_fsn, control->fsn_included, + next_to_del); + nctl = TAILQ_NEXT(control, next_instrm); + if (SCTP_MID_EQ(asoc->idata_supported, control->mid, next_to_del) && + (control->first_frag_seen)) { + int done; + + /* Ok we can deliver it onto the stream. */ + if (control->end_added) { + /* We are done with it afterwards */ + if (control->on_strm_q) { +#ifdef INVARIANTS + if (control->on_strm_q != SCTP_ON_ORDERED) { + panic("Huh control: %p on_q: %d -- not ordered?", + control, control->on_strm_q); + } +#endif + SCTP_STAT_INCR_COUNTER64(sctps_reasmusrmsgs); + TAILQ_REMOVE(&strm->inqueue, control, next_instrm); + if (asoc->size_on_all_streams >= control->length) { + asoc->size_on_all_streams -= control->length; + } else { +#ifdef INVARIANTS + panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); +#else + asoc->size_on_all_streams = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_all_streams); + control->on_strm_q = 0; + } + ret++; + } + if (((control->sinfo_flags >> 8) & SCTP_DATA_NOT_FRAG) == SCTP_DATA_NOT_FRAG) { + /* A singleton now slipping through - mark it non-revokable too */ + sctp_mark_non_revokable(asoc, control->sinfo_tsn); + } else if (control->end_added == 0) { + /* Check if we can defer adding until its all there */ + if ((control->length < pd_point) || (strm->pd_api_started)) { + /* Don't need it or cannot add more (one being delivered that way) */ + goto out; + } + } + done = (control->end_added) && (control->last_frag_seen); + if (control->on_read_q == 0) { + if (!done) { + if (asoc->size_on_all_streams >= control->length) { + asoc->size_on_all_streams -= control->length; + } else { +#ifdef INVARIANTS + panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); +#else + asoc->size_on_all_streams = 0; +#endif + } + strm->pd_api_started = 1; + control->pdapi_started = 1; + } + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, control->end_added, + inp_read_lock_held, SCTP_SO_NOT_LOCKED); + } + strm->last_mid_delivered = next_to_del; + if (done) { + control = nctl; + goto deliver_more; + } + } + } +out: + return (ret); +} + +uint32_t +sctp_add_chk_to_control(struct sctp_queued_to_read *control, + struct sctp_stream_in *strm, + struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_tmit_chunk *chk, int hold_rlock) +{ + /* + * Given a control and a chunk, merge the + * data from the chk onto the control and free + * up the chunk resources. + */ + uint32_t added=0; + int i_locked = 0; + + if (control->on_read_q && (hold_rlock == 0)) { + /* + * Its being pd-api'd so we must + * do some locks. + */ + SCTP_INP_READ_LOCK(stcb->sctp_ep); + i_locked = 1; + } + if (control->data == NULL) { + control->data = chk->data; + sctp_setup_tail_pointer(control); + } else { + sctp_add_to_tail_pointer(control, chk->data, &added); + } + control->fsn_included = chk->rec.data.fsn; + asoc->size_on_reasm_queue -= chk->send_size; + sctp_ucount_decr(asoc->cnt_on_reasm_queue); + sctp_mark_non_revokable(asoc, chk->rec.data.tsn); + chk->data = NULL; + if (chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) { + control->first_frag_seen = 1; + control->sinfo_tsn = chk->rec.data.tsn; + control->sinfo_ppid = chk->rec.data.ppid; + } + if (chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) { + /* Its complete */ + if ((control->on_strm_q) && (control->on_read_q)) { + if (control->pdapi_started) { + control->pdapi_started = 0; + strm->pd_api_started = 0; + } + if (control->on_strm_q == SCTP_ON_UNORDERED) { + /* Unordered */ + TAILQ_REMOVE(&strm->uno_inqueue, control, next_instrm); + control->on_strm_q = 0; + } else if (control->on_strm_q == SCTP_ON_ORDERED) { + /* Ordered */ + TAILQ_REMOVE(&strm->inqueue, control, next_instrm); + /* + * Don't need to decrement size_on_all_streams, + * since control is on the read queue. + */ + sctp_ucount_decr(asoc->cnt_on_all_streams); + control->on_strm_q = 0; +#ifdef INVARIANTS + } else if (control->on_strm_q) { + panic("Unknown state on ctrl: %p on_strm_q: %d", control, + control->on_strm_q); +#endif + } + } + control->end_added = 1; + control->last_frag_seen = 1; + } + if (i_locked) { + SCTP_INP_READ_UNLOCK(stcb->sctp_ep); + } + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + return (added); +} + +/* + * Dump onto the re-assembly queue, in its proper place. After dumping on the + * queue, see if anthing can be delivered. If so pull it off (or as much as + * we can. If we run out of space then we must dump what we can and set the + * appropriate flag to say we queued what we could. + */ +static void +sctp_queue_data_for_reasm(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_queued_to_read *control, + struct sctp_tmit_chunk *chk, + int created_control, + int *abort_flag, uint32_t tsn) +{ + uint32_t next_fsn; + struct sctp_tmit_chunk *at, *nat; + struct sctp_stream_in *strm; + int do_wakeup, unordered; + uint32_t lenadded; + + strm = &asoc->strmin[control->sinfo_stream]; + /* + * For old un-ordered data chunks. + */ + if ((control->sinfo_flags >> 8) & SCTP_DATA_UNORDERED) { + unordered = 1; + } else { + unordered = 0; + } + /* Must be added to the stream-in queue */ + if (created_control) { + if ((unordered == 0) || (asoc->idata_supported)) { + sctp_ucount_incr(asoc->cnt_on_all_streams); + } + if (sctp_place_control_in_stream(strm, asoc, control)) { + /* Duplicate SSN? */ + sctp_abort_in_reasm(stcb, control, chk, + abort_flag, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_6); + sctp_clean_up_control(stcb, control); + return; + } + if ((tsn == (asoc->cumulative_tsn + 1) && (asoc->idata_supported == 0))) { + /* Ok we created this control and now + * lets validate that its legal i.e. there + * is a B bit set, if not and we have + * up to the cum-ack then its invalid. + */ + if ((chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) == 0) { + sctp_abort_in_reasm(stcb, control, chk, + abort_flag, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_7); + return; + } + } + } + if ((asoc->idata_supported == 0) && (unordered == 1)) { + sctp_inject_old_unordered_data(stcb, asoc, control, chk, abort_flag); + return; + } + /* + * Ok we must queue the chunk into the reasembly portion: + * o if its the first it goes to the control mbuf. + * o if its not first but the next in sequence it goes to the control, + * and each succeeding one in order also goes. + * o if its not in order we place it on the list in its place. + */ + if (chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) { + /* Its the very first one. */ + SCTPDBG(SCTP_DEBUG_XXX, + "chunk is a first fsn: %u becomes fsn_included\n", + chk->rec.data.fsn); + if (control->first_frag_seen) { + /* + * Error on senders part, they either + * sent us two data chunks with FIRST, + * or they sent two un-ordered chunks that + * were fragmented at the same time in the same stream. + */ + sctp_abort_in_reasm(stcb, control, chk, + abort_flag, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_8); + return; + } + control->first_frag_seen = 1; + control->sinfo_ppid = chk->rec.data.ppid; + control->sinfo_tsn = chk->rec.data.tsn; + control->fsn_included = chk->rec.data.fsn; + control->data = chk->data; + sctp_mark_non_revokable(asoc, chk->rec.data.tsn); + chk->data = NULL; + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + sctp_setup_tail_pointer(control); + asoc->size_on_all_streams += control->length; + } else { + /* Place the chunk in our list */ + int inserted=0; + if (control->last_frag_seen == 0) { + /* Still willing to raise highest FSN seen */ + if (SCTP_TSN_GT(chk->rec.data.fsn, control->top_fsn)) { + SCTPDBG(SCTP_DEBUG_XXX, + "We have a new top_fsn: %u\n", + chk->rec.data.fsn); + control->top_fsn = chk->rec.data.fsn; + } + if (chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) { + SCTPDBG(SCTP_DEBUG_XXX, + "The last fsn is now in place fsn: %u\n", + chk->rec.data.fsn); + control->last_frag_seen = 1; + if (SCTP_TSN_GT(control->top_fsn, chk->rec.data.fsn)) { + SCTPDBG(SCTP_DEBUG_XXX, + "New fsn: %u is not at top_fsn: %u -- abort\n", + chk->rec.data.fsn, + control->top_fsn); + sctp_abort_in_reasm(stcb, control, chk, + abort_flag, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_9); + return; + } + } + if (asoc->idata_supported || control->first_frag_seen) { + /* + * For IDATA we always check since we know that + * the first fragment is 0. For old DATA we have + * to receive the first before we know the first FSN + * (which is the TSN). + */ + if (SCTP_TSN_GE(control->fsn_included, chk->rec.data.fsn)) { + /* We have already delivered up to this so its a dup */ + sctp_abort_in_reasm(stcb, control, chk, + abort_flag, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_10); + return; + } + } + } else { + if (chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) { + /* Second last? huh? */ + SCTPDBG(SCTP_DEBUG_XXX, + "Duplicate last fsn: %u (top: %u) -- abort\n", + chk->rec.data.fsn, control->top_fsn); + sctp_abort_in_reasm(stcb, control, + chk, abort_flag, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_11); + return; + } + if (asoc->idata_supported || control->first_frag_seen) { + /* + * For IDATA we always check since we know that + * the first fragment is 0. For old DATA we have + * to receive the first before we know the first FSN + * (which is the TSN). + */ + + if (SCTP_TSN_GE(control->fsn_included, chk->rec.data.fsn)) { + /* We have already delivered up to this so its a dup */ + SCTPDBG(SCTP_DEBUG_XXX, + "New fsn: %u is already seen in included_fsn: %u -- abort\n", + chk->rec.data.fsn, control->fsn_included); + sctp_abort_in_reasm(stcb, control, chk, + abort_flag, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_12); + return; + } + } + /* validate not beyond top FSN if we have seen last one */ + if (SCTP_TSN_GT(chk->rec.data.fsn, control->top_fsn)) { + SCTPDBG(SCTP_DEBUG_XXX, + "New fsn: %u is beyond or at top_fsn: %u -- abort\n", + chk->rec.data.fsn, + control->top_fsn); + sctp_abort_in_reasm(stcb, control, chk, + abort_flag, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_13); + return; + } + } + /* + * If we reach here, we need to place the + * new chunk in the reassembly for this + * control. + */ + SCTPDBG(SCTP_DEBUG_XXX, + "chunk is a not first fsn: %u needs to be inserted\n", + chk->rec.data.fsn); + TAILQ_FOREACH(at, &control->reasm, sctp_next) { + if (SCTP_TSN_GT(at->rec.data.fsn, chk->rec.data.fsn)) { + if (chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) { + /* Last not at the end? huh? */ + SCTPDBG(SCTP_DEBUG_XXX, + "Last fragment not last in list: -- abort\n"); + sctp_abort_in_reasm(stcb, control, + chk, abort_flag, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_14); + return; + } + /* + * This one in queue is bigger than the new one, insert + * the new one before at. + */ + SCTPDBG(SCTP_DEBUG_XXX, + "Insert it before fsn: %u\n", + at->rec.data.fsn); + asoc->size_on_reasm_queue += chk->send_size; + sctp_ucount_incr(asoc->cnt_on_reasm_queue); + TAILQ_INSERT_BEFORE(at, chk, sctp_next); + inserted = 1; + break; + } else if (at->rec.data.fsn == chk->rec.data.fsn) { + /* Gak, He sent me a duplicate str seq number */ + /* + * foo bar, I guess I will just free this new guy, + * should we abort too? FIX ME MAYBE? Or it COULD be + * that the SSN's have wrapped. Maybe I should + * compare to TSN somehow... sigh for now just blow + * away the chunk! + */ + SCTPDBG(SCTP_DEBUG_XXX, + "Duplicate to fsn: %u -- abort\n", + at->rec.data.fsn); + sctp_abort_in_reasm(stcb, control, + chk, abort_flag, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_15); + return; + } + } + if (inserted == 0) { + /* Goes on the end */ + SCTPDBG(SCTP_DEBUG_XXX, "Inserting at tail of list fsn: %u\n", + chk->rec.data.fsn); + asoc->size_on_reasm_queue += chk->send_size; + sctp_ucount_incr(asoc->cnt_on_reasm_queue); + TAILQ_INSERT_TAIL(&control->reasm, chk, sctp_next); + } + } + /* + * Ok lets see if we can suck any up into the control + * structure that are in seq if it makes sense. + */ + do_wakeup = 0; + /* + * If the first fragment has not been + * seen there is no sense in looking. + */ + if (control->first_frag_seen) { + next_fsn = control->fsn_included + 1; + TAILQ_FOREACH_SAFE(at, &control->reasm, sctp_next, nat) { + if (at->rec.data.fsn == next_fsn) { + /* We can add this one now to the control */ + SCTPDBG(SCTP_DEBUG_XXX, + "Adding more to control: %p at: %p fsn: %u next_fsn: %u included: %u\n", + control, at, + at->rec.data.fsn, + next_fsn, control->fsn_included); + TAILQ_REMOVE(&control->reasm, at, sctp_next); + lenadded = sctp_add_chk_to_control(control, strm, stcb, asoc, at, SCTP_READ_LOCK_NOT_HELD); + if (control->on_read_q) { + do_wakeup = 1; + } else { + /* + * We only add to the size-on-all-streams + * if its not on the read q. The read q + * flag will cause a sballoc so its accounted + * for there. + */ + asoc->size_on_all_streams += lenadded; + } + next_fsn++; + if (control->end_added && control->pdapi_started) { + if (strm->pd_api_started) { + strm->pd_api_started = 0; + control->pdapi_started = 0; + } + if (control->on_read_q == 0) { + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, control->end_added, + SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); + } + break; + } + } else { + break; + } + } + } + if (do_wakeup) { +#if defined(__Userspace__) + sctp_invoke_recv_callback(stcb->sctp_ep, stcb, control, SCTP_READ_LOCK_NOT_HELD); +#endif + /* Need to wakeup the reader */ + sctp_wakeup_the_read_socket(stcb->sctp_ep, stcb, SCTP_SO_NOT_LOCKED); + } +} + +static struct sctp_queued_to_read * +sctp_find_reasm_entry(struct sctp_stream_in *strm, uint32_t mid, int ordered, int idata_supported) +{ + struct sctp_queued_to_read *control; + + if (ordered) { + TAILQ_FOREACH(control, &strm->inqueue, next_instrm) { + if (SCTP_MID_EQ(idata_supported, control->mid, mid)) { + break; + } + } + } else { + if (idata_supported) { + TAILQ_FOREACH(control, &strm->uno_inqueue, next_instrm) { + if (SCTP_MID_EQ(idata_supported, control->mid, mid)) { + break; + } + } + } else { + control = TAILQ_FIRST(&strm->uno_inqueue); + } + } + return (control); +} + +static int +sctp_process_a_data_chunk(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct mbuf **m, int offset, int chk_length, + struct sctp_nets *net, uint32_t *high_tsn, int *abort_flag, + int *break_flag, int last_chunk, uint8_t chk_type) +{ + struct sctp_tmit_chunk *chk = NULL; /* make gcc happy */ + struct sctp_stream_in *strm; + uint32_t tsn, fsn, gap, mid; + struct mbuf *dmbuf; + int the_len; + int need_reasm_check = 0; + uint16_t sid; + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + struct sctp_queued_to_read *control, *ncontrol; + uint32_t ppid; + uint8_t chk_flags; + struct sctp_stream_reset_list *liste; + int ordered; + size_t clen; + int created_control = 0; + + if (chk_type == SCTP_IDATA) { + struct sctp_idata_chunk *chunk, chunk_buf; + + chunk = (struct sctp_idata_chunk *)sctp_m_getptr(*m, offset, + sizeof(struct sctp_idata_chunk), (uint8_t *)&chunk_buf); + chk_flags = chunk->ch.chunk_flags; + clen = sizeof(struct sctp_idata_chunk); + tsn = ntohl(chunk->dp.tsn); + sid = ntohs(chunk->dp.sid); + mid = ntohl(chunk->dp.mid); + if (chk_flags & SCTP_DATA_FIRST_FRAG) { + fsn = 0; + ppid = chunk->dp.ppid_fsn.ppid; + } else { + fsn = ntohl(chunk->dp.ppid_fsn.fsn); + ppid = 0xffffffff; /* Use as an invalid value. */ + } + } else { + struct sctp_data_chunk *chunk, chunk_buf; + + chunk = (struct sctp_data_chunk *)sctp_m_getptr(*m, offset, + sizeof(struct sctp_data_chunk), (uint8_t *)&chunk_buf); + chk_flags = chunk->ch.chunk_flags; + clen = sizeof(struct sctp_data_chunk); + tsn = ntohl(chunk->dp.tsn); + sid = ntohs(chunk->dp.sid); + mid = (uint32_t)(ntohs(chunk->dp.ssn)); + fsn = tsn; + ppid = chunk->dp.ppid; + } + if ((size_t)chk_length == clen) { + /* + * Need to send an abort since we had a + * empty data chunk. + */ + op_err = sctp_generate_no_user_data_cause(tsn); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_16; + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + *abort_flag = 1; + return (0); + } + if ((chk_flags & SCTP_DATA_SACK_IMMEDIATELY) == SCTP_DATA_SACK_IMMEDIATELY) { + asoc->send_sack = 1; + } + ordered = ((chk_flags & SCTP_DATA_UNORDERED) == 0); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { + sctp_log_map(tsn, asoc->cumulative_tsn, asoc->highest_tsn_inside_map, SCTP_MAP_TSN_ENTERS); + } + if (stcb == NULL) { + return (0); + } + SCTP_LTRACE_CHK(stcb->sctp_ep, stcb, chk_type, tsn); + if (SCTP_TSN_GE(asoc->cumulative_tsn, tsn)) { + /* It is a duplicate */ + SCTP_STAT_INCR(sctps_recvdupdata); + if (asoc->numduptsns < SCTP_MAX_DUP_TSNS) { + /* Record a dup for the next outbound sack */ + asoc->dup_tsns[asoc->numduptsns] = tsn; + asoc->numduptsns++; + } + asoc->send_sack = 1; + return (0); + } + /* Calculate the number of TSN's between the base and this TSN */ + SCTP_CALC_TSN_TO_GAP(gap, tsn, asoc->mapping_array_base_tsn); + if (gap >= (SCTP_MAPPING_ARRAY << 3)) { + /* Can't hold the bit in the mapping at max array, toss it */ + return (0); + } + if (gap >= (uint32_t) (asoc->mapping_array_size << 3)) { + SCTP_TCB_LOCK_ASSERT(stcb); + if (sctp_expand_mapping_array(asoc, gap)) { + /* Can't expand, drop it */ + return (0); + } + } + if (SCTP_TSN_GT(tsn, *high_tsn)) { + *high_tsn = tsn; + } + /* See if we have received this one already */ + if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap) || + SCTP_IS_TSN_PRESENT(asoc->nr_mapping_array, gap)) { + SCTP_STAT_INCR(sctps_recvdupdata); + if (asoc->numduptsns < SCTP_MAX_DUP_TSNS) { + /* Record a dup for the next outbound sack */ + asoc->dup_tsns[asoc->numduptsns] = tsn; + asoc->numduptsns++; + } + asoc->send_sack = 1; + return (0); + } + /* + * Check to see about the GONE flag, duplicates would cause a sack + * to be sent up above + */ + if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET))) { + /* + * wait a minute, this guy is gone, there is no longer a + * receiver. Send peer an ABORT! + */ + op_err = sctp_generate_cause(SCTP_CAUSE_OUT_OF_RESC, ""); + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + *abort_flag = 1; + return (0); + } + /* + * Now before going further we see if there is room. If NOT then we + * MAY let one through only IF this TSN is the one we are waiting + * for on a partial delivery API. + */ + + /* Is the stream valid? */ + if (sid >= asoc->streamincnt) { + struct sctp_error_invalid_stream *cause; + + op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_error_invalid_stream), + 0, M_NOWAIT, 1, MT_DATA); + if (op_err != NULL) { + /* add some space up front so prepend will work well */ + SCTP_BUF_RESV_UF(op_err, sizeof(struct sctp_chunkhdr)); + cause = mtod(op_err, struct sctp_error_invalid_stream *); + /* + * Error causes are just param's and this one has + * two back to back phdr, one with the error type + * and size, the other with the streamid and a rsvd + */ + SCTP_BUF_LEN(op_err) = sizeof(struct sctp_error_invalid_stream); + cause->cause.code = htons(SCTP_CAUSE_INVALID_STREAM); + cause->cause.length = htons(sizeof(struct sctp_error_invalid_stream)); + cause->stream_id = htons(sid); + cause->reserved = htons(0); + sctp_queue_op_err(stcb, op_err); + } + SCTP_STAT_INCR(sctps_badsid); + SCTP_TCB_LOCK_ASSERT(stcb); + SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap); + if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) { + asoc->highest_tsn_inside_nr_map = tsn; + } + if (tsn == (asoc->cumulative_tsn + 1)) { + /* Update cum-ack */ + asoc->cumulative_tsn = tsn; + } + return (0); + } + /* + * If its a fragmented message, lets see if we can + * find the control on the reassembly queues. + */ + if ((chk_type == SCTP_IDATA) && + ((chk_flags & SCTP_DATA_FIRST_FRAG) == 0) && + (fsn == 0)) { + /* + * The first *must* be fsn 0, and other + * (middle/end) pieces can *not* be fsn 0. + * XXX: This can happen in case of a wrap around. + * Ignore is for now. + */ + SCTP_SNPRINTF(msg, sizeof(msg), "FSN zero for MID=%8.8x, but flags=%2.2x", mid, chk_flags); + goto err_out; + } + control = sctp_find_reasm_entry(&asoc->strmin[sid], mid, ordered, asoc->idata_supported); + SCTPDBG(SCTP_DEBUG_XXX, "chunk_flags:0x%x look for control on queues %p\n", + chk_flags, control); + if ((chk_flags & SCTP_DATA_NOT_FRAG) != SCTP_DATA_NOT_FRAG) { + /* See if we can find the re-assembly entity */ + if (control != NULL) { + /* We found something, does it belong? */ + if (ordered && (mid != control->mid)) { + SCTP_SNPRINTF(msg, sizeof(msg), "Reassembly problem (MID=%8.8x)", mid); + err_out: + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_17; + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + *abort_flag = 1; + return (0); + } + if (ordered && ((control->sinfo_flags >> 8) & SCTP_DATA_UNORDERED)) { + /* We can't have a switched order with an unordered chunk */ + SCTP_SNPRINTF(msg, sizeof(msg), + "All fragments of a user message must be ordered or unordered (TSN=%8.8x)", + tsn); + goto err_out; + } + if (!ordered && (((control->sinfo_flags >> 8) & SCTP_DATA_UNORDERED) == 0)) { + /* We can't have a switched unordered with a ordered chunk */ + SCTP_SNPRINTF(msg, sizeof(msg), + "All fragments of a user message must be ordered or unordered (TSN=%8.8x)", + tsn); + goto err_out; + } + } + } else { + /* Its a complete segment. Lets validate we + * don't have a re-assembly going on with + * the same Stream/Seq (for ordered) or in + * the same Stream for unordered. + */ + if (control != NULL) { + if (ordered || asoc->idata_supported) { + SCTPDBG(SCTP_DEBUG_XXX, "chunk_flags: 0x%x dup detected on MID: %u\n", + chk_flags, mid); + SCTP_SNPRINTF(msg, sizeof(msg), "Duplicate MID=%8.8x detected.", mid); + goto err_out; + } else { + if ((tsn == control->fsn_included + 1) && + (control->end_added == 0)) { + SCTP_SNPRINTF(msg, sizeof(msg), + "Illegal message sequence, missing end for MID: %8.8x", + control->fsn_included); + goto err_out; + } else { + control = NULL; + } + } + } + } + /* now do the tests */ + if (((asoc->cnt_on_all_streams + + asoc->cnt_on_reasm_queue + + asoc->cnt_msg_on_sb) >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue)) || + (((int)asoc->my_rwnd) <= 0)) { + /* + * When we have NO room in the rwnd we check to make sure + * the reader is doing its job... + */ + if (SCTP_SBAVAIL(&stcb->sctp_socket->so_rcv) > 0) { + /* some to read, wake-up */ +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; + + so = SCTP_INP_SO(stcb->sctp_ep); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { + /* assoc was freed while we were unlocked */ + SCTP_SOCKET_UNLOCK(so, 1); + return (0); + } +#endif + sctp_sorwakeup(stcb->sctp_ep, stcb->sctp_socket); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } + /* now is it in the mapping array of what we have accepted? */ + if (chk_type == SCTP_DATA) { + if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_map) && + SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) { + /* Nope not in the valid range dump it */ + dump_packet: + sctp_set_rwnd(stcb, asoc); + if ((asoc->cnt_on_all_streams + + asoc->cnt_on_reasm_queue + + asoc->cnt_msg_on_sb) >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue)) { + SCTP_STAT_INCR(sctps_datadropchklmt); + } else { + SCTP_STAT_INCR(sctps_datadroprwnd); + } + *break_flag = 1; + return (0); + } + } else { + if (control == NULL) { + goto dump_packet; + } + if (SCTP_TSN_GT(fsn, control->top_fsn)) { + goto dump_packet; + } + } + } +#ifdef SCTP_ASOCLOG_OF_TSNS + SCTP_TCB_LOCK_ASSERT(stcb); + if (asoc->tsn_in_at >= SCTP_TSN_LOG_SIZE) { + asoc->tsn_in_at = 0; + asoc->tsn_in_wrapped = 1; + } + asoc->in_tsnlog[asoc->tsn_in_at].tsn = tsn; + asoc->in_tsnlog[asoc->tsn_in_at].strm = sid; + asoc->in_tsnlog[asoc->tsn_in_at].seq = mid; + asoc->in_tsnlog[asoc->tsn_in_at].sz = chk_length; + asoc->in_tsnlog[asoc->tsn_in_at].flgs = chunk_flags; + asoc->in_tsnlog[asoc->tsn_in_at].stcb = (void *)stcb; + asoc->in_tsnlog[asoc->tsn_in_at].in_pos = asoc->tsn_in_at; + asoc->in_tsnlog[asoc->tsn_in_at].in_out = 1; + asoc->tsn_in_at++; +#endif + /* + * Before we continue lets validate that we are not being fooled by + * an evil attacker. We can only have Nk chunks based on our TSN + * spread allowed by the mapping array N * 8 bits, so there is no + * way our stream sequence numbers could have wrapped. We of course + * only validate the FIRST fragment so the bit must be set. + */ + if ((chk_flags & SCTP_DATA_FIRST_FRAG) && + (TAILQ_EMPTY(&asoc->resetHead)) && + (chk_flags & SCTP_DATA_UNORDERED) == 0 && + SCTP_MID_GE(asoc->idata_supported, asoc->strmin[sid].last_mid_delivered, mid)) { + /* The incoming sseq is behind where we last delivered? */ + SCTPDBG(SCTP_DEBUG_INDATA1, "EVIL/Broken-Dup S-SEQ: %u delivered: %u from peer, Abort!\n", + mid, asoc->strmin[sid].last_mid_delivered); + + if (asoc->idata_supported) { + SCTP_SNPRINTF(msg, sizeof(msg), "Delivered MID=%8.8x, got TSN=%8.8x, SID=%4.4x, MID=%8.8x", + asoc->strmin[sid].last_mid_delivered, + tsn, + sid, + mid); + } else { + SCTP_SNPRINTF(msg, sizeof(msg), "Delivered SSN=%4.4x, got TSN=%8.8x, SID=%4.4x, SSN=%4.4x", + (uint16_t)asoc->strmin[sid].last_mid_delivered, + tsn, + sid, + (uint16_t)mid); + } + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_18; + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + *abort_flag = 1; + return (0); + } + if (chk_type == SCTP_IDATA) { + the_len = (chk_length - sizeof(struct sctp_idata_chunk)); + } else { + the_len = (chk_length - sizeof(struct sctp_data_chunk)); + } + if (last_chunk == 0) { + if (chk_type == SCTP_IDATA) { + dmbuf = SCTP_M_COPYM(*m, + (offset + sizeof(struct sctp_idata_chunk)), + the_len, M_NOWAIT); + } else { + dmbuf = SCTP_M_COPYM(*m, + (offset + sizeof(struct sctp_data_chunk)), + the_len, M_NOWAIT); + } +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(dmbuf, SCTP_MBUF_ICOPY); + } +#endif + } else { + /* We can steal the last chunk */ + int l_len; + dmbuf = *m; + /* lop off the top part */ + if (chk_type == SCTP_IDATA) { + m_adj(dmbuf, (offset + sizeof(struct sctp_idata_chunk))); + } else { + m_adj(dmbuf, (offset + sizeof(struct sctp_data_chunk))); + } + if (SCTP_BUF_NEXT(dmbuf) == NULL) { + l_len = SCTP_BUF_LEN(dmbuf); + } else { + /* need to count up the size hopefully + * does not hit this to often :-0 + */ + struct mbuf *lat; + + l_len = 0; + for (lat = dmbuf; lat; lat = SCTP_BUF_NEXT(lat)) { + l_len += SCTP_BUF_LEN(lat); + } + } + if (l_len > the_len) { + /* Trim the end round bytes off too */ + m_adj(dmbuf, -(l_len - the_len)); + } + } + if (dmbuf == NULL) { + SCTP_STAT_INCR(sctps_nomem); + return (0); + } + /* + * Now no matter what, we need a control, get one + * if we don't have one (we may have gotten it + * above when we found the message was fragmented + */ + if (control == NULL) { + sctp_alloc_a_readq(stcb, control); + sctp_build_readq_entry_mac(control, stcb, asoc->context, net, tsn, + ppid, + sid, + chk_flags, + NULL, fsn, mid); + if (control == NULL) { + SCTP_STAT_INCR(sctps_nomem); + return (0); + } + if ((chk_flags & SCTP_DATA_NOT_FRAG) == SCTP_DATA_NOT_FRAG) { + struct mbuf *mm; + + control->data = dmbuf; + control->tail_mbuf = NULL; + for (mm = control->data; mm; mm = mm->m_next) { + control->length += SCTP_BUF_LEN(mm); + if (SCTP_BUF_NEXT(mm) == NULL) { + control->tail_mbuf = mm; + } + } + control->end_added = 1; + control->last_frag_seen = 1; + control->first_frag_seen = 1; + control->fsn_included = fsn; + control->top_fsn = fsn; + } + created_control = 1; + } + SCTPDBG(SCTP_DEBUG_XXX, "chunk_flags: 0x%x ordered: %d MID: %u control: %p\n", + chk_flags, ordered, mid, control); + if ((chk_flags & SCTP_DATA_NOT_FRAG) == SCTP_DATA_NOT_FRAG && + TAILQ_EMPTY(&asoc->resetHead) && + ((ordered == 0) || + (SCTP_MID_EQ(asoc->idata_supported, asoc->strmin[sid].last_mid_delivered + 1, mid) && + TAILQ_EMPTY(&asoc->strmin[sid].inqueue)))) { + /* Candidate for express delivery */ + /* + * Its not fragmented, No PD-API is up, Nothing in the + * delivery queue, Its un-ordered OR ordered and the next to + * deliver AND nothing else is stuck on the stream queue, + * And there is room for it in the socket buffer. Lets just + * stuff it up the buffer.... + */ + SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap); + if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) { + asoc->highest_tsn_inside_nr_map = tsn; + } + SCTPDBG(SCTP_DEBUG_XXX, "Injecting control: %p to be read (MID: %u)\n", + control, mid); + + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, &stcb->sctp_socket->so_rcv, + 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); + + if ((chk_flags & SCTP_DATA_UNORDERED) == 0) { + /* for ordered, bump what we delivered */ + asoc->strmin[sid].last_mid_delivered++; + } + SCTP_STAT_INCR(sctps_recvexpress); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { + sctp_log_strm_del_alt(stcb, tsn, mid, sid, + SCTP_STR_LOG_FROM_EXPRS_DEL); + } + control = NULL; + goto finish_express_del; + } + + /* Now will we need a chunk too? */ + if ((chk_flags & SCTP_DATA_NOT_FRAG) != SCTP_DATA_NOT_FRAG) { + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + /* No memory so we drop the chunk */ + SCTP_STAT_INCR(sctps_nomem); + if (last_chunk == 0) { + /* we copied it, free the copy */ + sctp_m_freem(dmbuf); + } + return (0); + } + chk->rec.data.tsn = tsn; + chk->no_fr_allowed = 0; + chk->rec.data.fsn = fsn; + chk->rec.data.mid = mid; + chk->rec.data.sid = sid; + chk->rec.data.ppid = ppid; + chk->rec.data.context = stcb->asoc.context; + chk->rec.data.doing_fast_retransmit = 0; + chk->rec.data.rcv_flags = chk_flags; + chk->asoc = asoc; + chk->send_size = the_len; + chk->whoTo = net; + SCTPDBG(SCTP_DEBUG_XXX, "Building ck: %p for control: %p to be read (MID: %u)\n", + chk, + control, mid); + atomic_add_int(&net->ref_count, 1); + chk->data = dmbuf; + } + /* Set the appropriate TSN mark */ + if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) { + SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, gap); + if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_nr_map)) { + asoc->highest_tsn_inside_nr_map = tsn; + } + } else { + SCTP_SET_TSN_PRESENT(asoc->mapping_array, gap); + if (SCTP_TSN_GT(tsn, asoc->highest_tsn_inside_map)) { + asoc->highest_tsn_inside_map = tsn; + } + } + /* Now is it complete (i.e. not fragmented)? */ + if ((chk_flags & SCTP_DATA_NOT_FRAG) == SCTP_DATA_NOT_FRAG) { + /* + * Special check for when streams are resetting. We + * could be more smart about this and check the + * actual stream to see if it is not being reset.. + * that way we would not create a HOLB when amongst + * streams being reset and those not being reset. + * + */ + if (((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) && + SCTP_TSN_GT(tsn, liste->tsn)) { + /* + * yep its past where we need to reset... go + * ahead and queue it. + */ + if (TAILQ_EMPTY(&asoc->pending_reply_queue)) { + /* first one on */ + TAILQ_INSERT_TAIL(&asoc->pending_reply_queue, control, next); + } else { + struct sctp_queued_to_read *lcontrol, *nlcontrol; + unsigned char inserted = 0; + TAILQ_FOREACH_SAFE(lcontrol, &asoc->pending_reply_queue, next, nlcontrol) { + if (SCTP_TSN_GT(control->sinfo_tsn, lcontrol->sinfo_tsn)) { + continue; + } else { + /* found it */ + TAILQ_INSERT_BEFORE(lcontrol, control, next); + inserted = 1; + break; + } + } + if (inserted == 0) { + /* + * must be put at end, use + * prevP (all setup from + * loop) to setup nextP. + */ + TAILQ_INSERT_TAIL(&asoc->pending_reply_queue, control, next); + } + } + goto finish_express_del; + } + if (chk_flags & SCTP_DATA_UNORDERED) { + /* queue directly into socket buffer */ + SCTPDBG(SCTP_DEBUG_XXX, "Unordered data to be read control: %p MID: %u\n", + control, mid); + sctp_mark_non_revokable(asoc, control->sinfo_tsn); + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, + SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); + + } else { + SCTPDBG(SCTP_DEBUG_XXX, "Queue control: %p for reordering MID: %u\n", control, + mid); + sctp_queue_data_to_stream(stcb, asoc, control, abort_flag, &need_reasm_check); + if (*abort_flag) { + if (last_chunk) { + *m = NULL; + } + return (0); + } + } + goto finish_express_del; + } + /* If we reach here its a reassembly */ + need_reasm_check = 1; + SCTPDBG(SCTP_DEBUG_XXX, + "Queue data to stream for reasm control: %p MID: %u\n", + control, mid); + sctp_queue_data_for_reasm(stcb, asoc, control, chk, created_control, abort_flag, tsn); + if (*abort_flag) { + /* + * the assoc is now gone and chk was put onto the + * reasm queue, which has all been freed. + */ + if (last_chunk) { + *m = NULL; + } + return (0); + } +finish_express_del: + /* Here we tidy up things */ + if (tsn == (asoc->cumulative_tsn + 1)) { + /* Update cum-ack */ + asoc->cumulative_tsn = tsn; + } + if (last_chunk) { + *m = NULL; + } + if (ordered) { + SCTP_STAT_INCR_COUNTER64(sctps_inorderchunks); + } else { + SCTP_STAT_INCR_COUNTER64(sctps_inunorderchunks); + } + SCTP_STAT_INCR(sctps_recvdata); + /* Set it present please */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_STR_LOGGING_ENABLE) { + sctp_log_strm_del_alt(stcb, tsn, mid, sid, SCTP_STR_LOG_FROM_MARK_TSN); + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { + sctp_log_map(asoc->mapping_array_base_tsn, asoc->cumulative_tsn, + asoc->highest_tsn_inside_map, SCTP_MAP_PREPARE_SLIDE); + } + if (need_reasm_check) { + (void)sctp_deliver_reasm_check(stcb, asoc, &asoc->strmin[sid], SCTP_READ_LOCK_NOT_HELD); + need_reasm_check = 0; + } + /* check the special flag for stream resets */ + if (((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) && + SCTP_TSN_GE(asoc->cumulative_tsn, liste->tsn)) { + /* + * we have finished working through the backlogged TSN's now + * time to reset streams. 1: call reset function. 2: free + * pending_reply space 3: distribute any chunks in + * pending_reply_queue. + */ + sctp_reset_in_stream(stcb, liste->number_entries, liste->list_of_streams); + TAILQ_REMOVE(&asoc->resetHead, liste, next_resp); + sctp_send_deferred_reset_response(stcb, liste, SCTP_STREAM_RESET_RESULT_PERFORMED); + SCTP_FREE(liste, SCTP_M_STRESET); + /*sa_ignore FREED_MEMORY*/ + liste = TAILQ_FIRST(&asoc->resetHead); + if (TAILQ_EMPTY(&asoc->resetHead)) { + /* All can be removed */ + TAILQ_FOREACH_SAFE(control, &asoc->pending_reply_queue, next, ncontrol) { + TAILQ_REMOVE(&asoc->pending_reply_queue, control, next); + strm = &asoc->strmin[control->sinfo_stream]; + sctp_queue_data_to_stream(stcb, asoc, control, abort_flag, &need_reasm_check); + if (*abort_flag) { + return (0); + } + if (need_reasm_check) { + (void)sctp_deliver_reasm_check(stcb, asoc, strm, SCTP_READ_LOCK_NOT_HELD); + need_reasm_check = 0; + } + } + } else { + TAILQ_FOREACH_SAFE(control, &asoc->pending_reply_queue, next, ncontrol) { + if (SCTP_TSN_GT(control->sinfo_tsn, liste->tsn)) { + break; + } + /* + * if control->sinfo_tsn is <= liste->tsn we can + * process it which is the NOT of + * control->sinfo_tsn > liste->tsn + */ + TAILQ_REMOVE(&asoc->pending_reply_queue, control, next); + strm = &asoc->strmin[control->sinfo_stream]; + sctp_queue_data_to_stream(stcb, asoc, control, abort_flag, &need_reasm_check); + if (*abort_flag) { + return (0); + } + if (need_reasm_check) { + (void)sctp_deliver_reasm_check(stcb, asoc, strm, SCTP_READ_LOCK_NOT_HELD); + need_reasm_check = 0; + } + } + } + } + return (1); +} + +static const int8_t sctp_map_lookup_tab[256] = { + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 4, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 5, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 4, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 6, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 4, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 5, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 4, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 7, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 4, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 5, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 4, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 6, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 4, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 5, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 4, + 0, 1, 0, 2, 0, 1, 0, 3, + 0, 1, 0, 2, 0, 1, 0, 8 +}; + +void +sctp_slide_mapping_arrays(struct sctp_tcb *stcb) +{ + /* + * Now we also need to check the mapping array in a couple of ways. + * 1) Did we move the cum-ack point? + * + * When you first glance at this you might think + * that all entries that make up the position + * of the cum-ack would be in the nr-mapping array + * only.. i.e. things up to the cum-ack are always + * deliverable. Thats true with one exception, when + * its a fragmented message we may not deliver the data + * until some threshold (or all of it) is in place. So + * we must OR the nr_mapping_array and mapping_array to + * get a true picture of the cum-ack. + */ + struct sctp_association *asoc; + int at; + uint8_t val; + int slide_from, slide_end, lgap, distance; + uint32_t old_cumack, old_base, old_highest, highest_tsn; + + asoc = &stcb->asoc; + + old_cumack = asoc->cumulative_tsn; + old_base = asoc->mapping_array_base_tsn; + old_highest = asoc->highest_tsn_inside_map; + /* + * We could probably improve this a small bit by calculating the + * offset of the current cum-ack as the starting point. + */ + at = 0; + for (slide_from = 0; slide_from < stcb->asoc.mapping_array_size; slide_from++) { + val = asoc->nr_mapping_array[slide_from] | asoc->mapping_array[slide_from]; + if (val == 0xff) { + at += 8; + } else { + /* there is a 0 bit */ + at += sctp_map_lookup_tab[val]; + break; + } + } + asoc->cumulative_tsn = asoc->mapping_array_base_tsn + (at-1); + + if (SCTP_TSN_GT(asoc->cumulative_tsn, asoc->highest_tsn_inside_map) && + SCTP_TSN_GT(asoc->cumulative_tsn, asoc->highest_tsn_inside_nr_map)) { +#ifdef INVARIANTS + panic("huh, cumack 0x%x greater than high-tsn 0x%x in map", + asoc->cumulative_tsn, asoc->highest_tsn_inside_map); +#else + SCTP_PRINTF("huh, cumack 0x%x greater than high-tsn 0x%x in map - should panic?\n", + asoc->cumulative_tsn, asoc->highest_tsn_inside_map); + sctp_print_mapping_array(asoc); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { + sctp_log_map(0, 6, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); + } + asoc->highest_tsn_inside_map = asoc->cumulative_tsn; + asoc->highest_tsn_inside_nr_map = asoc->cumulative_tsn; +#endif + } + if (SCTP_TSN_GT(asoc->highest_tsn_inside_nr_map, asoc->highest_tsn_inside_map)) { + highest_tsn = asoc->highest_tsn_inside_nr_map; + } else { + highest_tsn = asoc->highest_tsn_inside_map; + } + if ((asoc->cumulative_tsn == highest_tsn) && (at >= 8)) { + /* The complete array was completed by a single FR */ + /* highest becomes the cum-ack */ + int clr; +#ifdef INVARIANTS + unsigned int i; +#endif + + /* clear the array */ + clr = ((at+7) >> 3); + if (clr > asoc->mapping_array_size) { + clr = asoc->mapping_array_size; + } + memset(asoc->mapping_array, 0, clr); + memset(asoc->nr_mapping_array, 0, clr); +#ifdef INVARIANTS + for (i = 0; i < asoc->mapping_array_size; i++) { + if ((asoc->mapping_array[i]) || (asoc->nr_mapping_array[i])) { + SCTP_PRINTF("Error Mapping array's not clean at clear\n"); + sctp_print_mapping_array(asoc); + } + } +#endif + asoc->mapping_array_base_tsn = asoc->cumulative_tsn + 1; + asoc->highest_tsn_inside_nr_map = asoc->highest_tsn_inside_map = asoc->cumulative_tsn; + } else if (at >= 8) { + /* we can slide the mapping array down */ + /* slide_from holds where we hit the first NON 0xff byte */ + + /* + * now calculate the ceiling of the move using our highest + * TSN value + */ + SCTP_CALC_TSN_TO_GAP(lgap, highest_tsn, asoc->mapping_array_base_tsn); + slide_end = (lgap >> 3); + if (slide_end < slide_from) { + sctp_print_mapping_array(asoc); +#ifdef INVARIANTS + panic("impossible slide"); +#else + SCTP_PRINTF("impossible slide lgap: %x slide_end: %x slide_from: %x? at: %d\n", + lgap, slide_end, slide_from, at); + return; +#endif + } + if (slide_end > asoc->mapping_array_size) { +#ifdef INVARIANTS + panic("would overrun buffer"); +#else + SCTP_PRINTF("Gak, would have overrun map end: %d slide_end: %d\n", + asoc->mapping_array_size, slide_end); + slide_end = asoc->mapping_array_size; +#endif + } + distance = (slide_end - slide_from) + 1; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { + sctp_log_map(old_base, old_cumack, old_highest, + SCTP_MAP_PREPARE_SLIDE); + sctp_log_map((uint32_t) slide_from, (uint32_t) slide_end, + (uint32_t) lgap, SCTP_MAP_SLIDE_FROM); + } + if (distance + slide_from > asoc->mapping_array_size || + distance < 0) { + /* + * Here we do NOT slide forward the array so that + * hopefully when more data comes in to fill it up + * we will be able to slide it forward. Really I + * don't think this should happen :-0 + */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { + sctp_log_map((uint32_t) distance, (uint32_t) slide_from, + (uint32_t) asoc->mapping_array_size, + SCTP_MAP_SLIDE_NONE); + } + } else { + int ii; + + for (ii = 0; ii < distance; ii++) { + asoc->mapping_array[ii] = asoc->mapping_array[slide_from + ii]; + asoc->nr_mapping_array[ii] = asoc->nr_mapping_array[slide_from + ii]; + } + for (ii = distance; ii < asoc->mapping_array_size; ii++) { + asoc->mapping_array[ii] = 0; + asoc->nr_mapping_array[ii] = 0; + } + if (asoc->highest_tsn_inside_map + 1 == asoc->mapping_array_base_tsn) { + asoc->highest_tsn_inside_map += (slide_from << 3); + } + if (asoc->highest_tsn_inside_nr_map + 1 == asoc->mapping_array_base_tsn) { + asoc->highest_tsn_inside_nr_map += (slide_from << 3); + } + asoc->mapping_array_base_tsn += (slide_from << 3); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { + sctp_log_map(asoc->mapping_array_base_tsn, + asoc->cumulative_tsn, asoc->highest_tsn_inside_map, + SCTP_MAP_SLIDE_RESULT); + } + } + } +} + +void +sctp_sack_check(struct sctp_tcb *stcb, int was_a_gap) +{ + struct sctp_association *asoc; + uint32_t highest_tsn; + int is_a_gap; + + sctp_slide_mapping_arrays(stcb); + asoc = &stcb->asoc; + if (SCTP_TSN_GT(asoc->highest_tsn_inside_nr_map, asoc->highest_tsn_inside_map)) { + highest_tsn = asoc->highest_tsn_inside_nr_map; + } else { + highest_tsn = asoc->highest_tsn_inside_map; + } + /* Is there a gap now? */ + is_a_gap = SCTP_TSN_GT(highest_tsn, stcb->asoc.cumulative_tsn); + + /* + * Now we need to see if we need to queue a sack or just start the + * timer (if allowed). + */ + if (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_SENT) { + /* + * Ok special case, in SHUTDOWN-SENT case. here we + * maker sure SACK timer is off and instead send a + * SHUTDOWN and a SACK + */ + if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) { + sctp_timer_stop(SCTP_TIMER_TYPE_RECV, + stcb->sctp_ep, stcb, NULL, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_19); + } + sctp_send_shutdown(stcb, + ((stcb->asoc.alternate) ? stcb->asoc.alternate : stcb->asoc.primary_destination)); + if (is_a_gap) { + sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED); + } + } else { + /* + * CMT DAC algorithm: increase number of packets + * received since last ack + */ + stcb->asoc.cmt_dac_pkts_rcvd++; + + if ((stcb->asoc.send_sack == 1) || /* We need to send a SACK */ + ((was_a_gap) && (is_a_gap == 0)) || /* was a gap, but no + * longer is one */ + (stcb->asoc.numduptsns) || /* we have dup's */ + (is_a_gap) || /* is still a gap */ + (stcb->asoc.delayed_ack == 0) || /* Delayed sack disabled */ + (stcb->asoc.data_pkts_seen >= stcb->asoc.sack_freq)) { /* hit limit of pkts */ + if ((stcb->asoc.sctp_cmt_on_off > 0) && + (SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) && + (stcb->asoc.send_sack == 0) && + (stcb->asoc.numduptsns == 0) && + (stcb->asoc.delayed_ack) && + (!SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer))) { + /* + * CMT DAC algorithm: With CMT, + * delay acks even in the face of + * reordering. Therefore, if acks + * that do not have to be sent + * because of the above reasons, + * will be delayed. That is, acks + * that would have been sent due to + * gap reports will be delayed with + * DAC. Start the delayed ack timer. + */ + sctp_timer_start(SCTP_TIMER_TYPE_RECV, + stcb->sctp_ep, stcb, NULL); + } else { + /* + * Ok we must build a SACK since the + * timer is pending, we got our + * first packet OR there are gaps or + * duplicates. + */ + sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_20); + sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED); + } + } else { + if (!SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) { + sctp_timer_start(SCTP_TIMER_TYPE_RECV, + stcb->sctp_ep, stcb, NULL); + } + } + } +} + +int +sctp_process_data(struct mbuf **mm, int iphlen, int *offset, int length, + struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct sctp_nets *net, uint32_t *high_tsn) +{ + struct sctp_chunkhdr *ch, chunk_buf; + struct sctp_association *asoc; + int num_chunks = 0; /* number of control chunks processed */ + int stop_proc = 0; + int break_flag, last_chunk; + int abort_flag = 0, was_a_gap; + struct mbuf *m; + uint32_t highest_tsn; + uint16_t chk_length; + + /* set the rwnd */ + sctp_set_rwnd(stcb, &stcb->asoc); + + m = *mm; + SCTP_TCB_LOCK_ASSERT(stcb); + asoc = &stcb->asoc; + if (SCTP_TSN_GT(asoc->highest_tsn_inside_nr_map, asoc->highest_tsn_inside_map)) { + highest_tsn = asoc->highest_tsn_inside_nr_map; + } else { + highest_tsn = asoc->highest_tsn_inside_map; + } + was_a_gap = SCTP_TSN_GT(highest_tsn, stcb->asoc.cumulative_tsn); + /* + * setup where we got the last DATA packet from for any SACK that + * may need to go out. Don't bump the net. This is done ONLY when a + * chunk is assigned. + */ + asoc->last_data_chunk_from = net; + + /*- + * Now before we proceed we must figure out if this is a wasted + * cluster... i.e. it is a small packet sent in and yet the driver + * underneath allocated a full cluster for it. If so we must copy it + * to a smaller mbuf and free up the cluster mbuf. This will help + * with cluster starvation. + */ + if (SCTP_BUF_LEN(m) < (long)MLEN && SCTP_BUF_NEXT(m) == NULL) { + /* we only handle mbufs that are singletons.. not chains */ + m = sctp_get_mbuf_for_msg(SCTP_BUF_LEN(m), 0, M_NOWAIT, 1, MT_DATA); + if (m) { + /* ok lets see if we can copy the data up */ + caddr_t *from, *to; + /* get the pointers and copy */ + to = mtod(m, caddr_t *); + from = mtod((*mm), caddr_t *); + memcpy(to, from, SCTP_BUF_LEN((*mm))); + /* copy the length and free up the old */ + SCTP_BUF_LEN(m) = SCTP_BUF_LEN((*mm)); + sctp_m_freem(*mm); + /* success, back copy */ + *mm = m; + } else { + /* We are in trouble in the mbuf world .. yikes */ + m = *mm; + } + } + /* get pointer to the first chunk header */ + ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, *offset, + sizeof(struct sctp_chunkhdr), + (uint8_t *)&chunk_buf); + if (ch == NULL) { + return (1); + } + /* + * process all DATA chunks... + */ + *high_tsn = asoc->cumulative_tsn; + break_flag = 0; + asoc->data_pkts_seen++; + while (stop_proc == 0) { + /* validate chunk length */ + chk_length = ntohs(ch->chunk_length); + if (length - *offset < chk_length) { + /* all done, mutulated chunk */ + stop_proc = 1; + continue; + } + if ((asoc->idata_supported == 1) && + (ch->chunk_type == SCTP_DATA)) { + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + SCTP_SNPRINTF(msg, sizeof(msg), "%s", "DATA chunk received when I-DATA was negotiated"); + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_21; + sctp_abort_an_association(inp, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + return (2); + } + if ((asoc->idata_supported == 0) && + (ch->chunk_type == SCTP_IDATA)) { + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + SCTP_SNPRINTF(msg, sizeof(msg), "%s", "I-DATA chunk received when DATA was negotiated"); + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_22; + sctp_abort_an_association(inp, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + return (2); + } + if ((ch->chunk_type == SCTP_DATA) || + (ch->chunk_type == SCTP_IDATA)) { + uint16_t clen; + + if (ch->chunk_type == SCTP_DATA) { + clen = sizeof(struct sctp_data_chunk); + } else { + clen = sizeof(struct sctp_idata_chunk); + } + if (chk_length < clen) { + /* + * Need to send an abort since we had a + * invalid data chunk. + */ + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + SCTP_SNPRINTF(msg, sizeof(msg), "%s chunk of length %u", + ch->chunk_type == SCTP_DATA ? "DATA" : "I-DATA", + chk_length); + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_23; + sctp_abort_an_association(inp, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + return (2); + } +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xB1, 0); +#endif + if (SCTP_SIZE32(chk_length) == (length - *offset)) { + last_chunk = 1; + } else { + last_chunk = 0; + } + if (sctp_process_a_data_chunk(stcb, asoc, mm, *offset, + chk_length, net, high_tsn, &abort_flag, &break_flag, + last_chunk, ch->chunk_type)) { + num_chunks++; + } + if (abort_flag) + return (2); + + if (break_flag) { + /* + * Set because of out of rwnd space and no + * drop rep space left. + */ + stop_proc = 1; + continue; + } + } else { + /* not a data chunk in the data region */ + switch (ch->chunk_type) { + case SCTP_INITIATION: + case SCTP_INITIATION_ACK: + case SCTP_SELECTIVE_ACK: + case SCTP_NR_SELECTIVE_ACK: + case SCTP_HEARTBEAT_REQUEST: + case SCTP_HEARTBEAT_ACK: + case SCTP_ABORT_ASSOCIATION: + case SCTP_SHUTDOWN: + case SCTP_SHUTDOWN_ACK: + case SCTP_OPERATION_ERROR: + case SCTP_COOKIE_ECHO: + case SCTP_COOKIE_ACK: + case SCTP_ECN_ECHO: + case SCTP_ECN_CWR: + case SCTP_SHUTDOWN_COMPLETE: + case SCTP_AUTHENTICATION: + case SCTP_ASCONF_ACK: + case SCTP_PACKET_DROPPED: + case SCTP_STREAM_RESET: + case SCTP_FORWARD_CUM_TSN: + case SCTP_ASCONF: + { + /* + * Now, what do we do with KNOWN chunks that + * are NOT in the right place? + * + * For now, I do nothing but ignore them. We + * may later want to add sysctl stuff to + * switch out and do either an ABORT() or + * possibly process them. + */ + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + SCTP_SNPRINTF(msg, sizeof(msg), "DATA chunk followed by chunk of type %2.2x", + ch->chunk_type); + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + sctp_abort_an_association(inp, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + return (2); + } + default: + /* + * Unknown chunk type: use bit rules after + * checking length + */ + if (chk_length < sizeof(struct sctp_chunkhdr)) { + /* + * Need to send an abort since we had a + * invalid chunk. + */ + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + SCTP_SNPRINTF(msg, sizeof(msg), "Chunk of length %u", chk_length); + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_24; + sctp_abort_an_association(inp, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + return (2); + } + if (ch->chunk_type & 0x40) { + /* Add a error report to the queue */ + struct mbuf *op_err; + struct sctp_gen_error_cause *cause; + + op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_gen_error_cause), + 0, M_NOWAIT, 1, MT_DATA); + if (op_err != NULL) { + cause = mtod(op_err, struct sctp_gen_error_cause *); + cause->code = htons(SCTP_CAUSE_UNRECOG_CHUNK); + cause->length = htons((uint16_t)(chk_length + sizeof(struct sctp_gen_error_cause))); + SCTP_BUF_LEN(op_err) = sizeof(struct sctp_gen_error_cause); + SCTP_BUF_NEXT(op_err) = SCTP_M_COPYM(m, *offset, chk_length, M_NOWAIT); + if (SCTP_BUF_NEXT(op_err) != NULL) { + sctp_queue_op_err(stcb, op_err); + } else { + sctp_m_freem(op_err); + } + } + } + if ((ch->chunk_type & 0x80) == 0) { + /* discard the rest of this packet */ + stop_proc = 1; + } /* else skip this bad chunk and + * continue... */ + break; + } /* switch of chunk type */ + } + *offset += SCTP_SIZE32(chk_length); + if ((*offset >= length) || stop_proc) { + /* no more data left in the mbuf chain */ + stop_proc = 1; + continue; + } + ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, *offset, + sizeof(struct sctp_chunkhdr), + (uint8_t *)&chunk_buf); + if (ch == NULL) { + *offset = length; + stop_proc = 1; + continue; + } + } + if (break_flag) { + /* + * we need to report rwnd overrun drops. + */ + sctp_send_packet_dropped(stcb, net, *mm, length, iphlen, 0); + } + if (num_chunks) { + /* + * Did we get data, if so update the time for auto-close and + * give peer credit for being alive. + */ + SCTP_STAT_INCR(sctps_recvpktwithdata); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { + sctp_misc_ints(SCTP_THRESHOLD_CLEAR, + stcb->asoc.overall_error_count, + 0, + SCTP_FROM_SCTP_INDATA, + __LINE__); + } + stcb->asoc.overall_error_count = 0; + (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_last_rcvd); + } + /* now service all of the reassm queue if needed */ + if (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_SENT) { + /* Assure that we ack right away */ + stcb->asoc.send_sack = 1; + } + /* Start a sack timer or QUEUE a SACK for sending */ + sctp_sack_check(stcb, was_a_gap); + return (0); +} + +static int +sctp_process_segment_range(struct sctp_tcb *stcb, struct sctp_tmit_chunk **p_tp1, uint32_t last_tsn, + uint16_t frag_strt, uint16_t frag_end, int nr_sacking, + int *num_frs, + uint32_t *biggest_newly_acked_tsn, + uint32_t *this_sack_lowest_newack, + int *rto_ok) +{ + struct sctp_tmit_chunk *tp1; + unsigned int theTSN; + int j, wake_him = 0, circled = 0; + + /* Recover the tp1 we last saw */ + tp1 = *p_tp1; + if (tp1 == NULL) { + tp1 = TAILQ_FIRST(&stcb->asoc.sent_queue); + } + for (j = frag_strt; j <= frag_end; j++) { + theTSN = j + last_tsn; + while (tp1) { + if (tp1->rec.data.doing_fast_retransmit) + (*num_frs) += 1; + + /*- + * CMT: CUCv2 algorithm. For each TSN being + * processed from the sent queue, track the + * next expected pseudo-cumack, or + * rtx_pseudo_cumack, if required. Separate + * cumack trackers for first transmissions, + * and retransmissions. + */ + if ((tp1->sent < SCTP_DATAGRAM_RESEND) && + (tp1->whoTo->find_pseudo_cumack == 1) && + (tp1->snd_count == 1)) { + tp1->whoTo->pseudo_cumack = tp1->rec.data.tsn; + tp1->whoTo->find_pseudo_cumack = 0; + } + if ((tp1->sent < SCTP_DATAGRAM_RESEND) && + (tp1->whoTo->find_rtx_pseudo_cumack == 1) && + (tp1->snd_count > 1)) { + tp1->whoTo->rtx_pseudo_cumack = tp1->rec.data.tsn; + tp1->whoTo->find_rtx_pseudo_cumack = 0; + } + if (tp1->rec.data.tsn == theTSN) { + if (tp1->sent != SCTP_DATAGRAM_UNSENT) { + /*- + * must be held until + * cum-ack passes + */ + if (tp1->sent < SCTP_DATAGRAM_RESEND) { + /*- + * If it is less than RESEND, it is + * now no-longer in flight. + * Higher values may already be set + * via previous Gap Ack Blocks... + * i.e. ACKED or RESEND. + */ + if (SCTP_TSN_GT(tp1->rec.data.tsn, + *biggest_newly_acked_tsn)) { + *biggest_newly_acked_tsn = tp1->rec.data.tsn; + } + /*- + * CMT: SFR algo (and HTNA) - set + * saw_newack to 1 for dest being + * newly acked. update + * this_sack_highest_newack if + * appropriate. + */ + if (tp1->rec.data.chunk_was_revoked == 0) + tp1->whoTo->saw_newack = 1; + + if (SCTP_TSN_GT(tp1->rec.data.tsn, + tp1->whoTo->this_sack_highest_newack)) { + tp1->whoTo->this_sack_highest_newack = + tp1->rec.data.tsn; + } + /*- + * CMT DAC algo: also update + * this_sack_lowest_newack + */ + if (*this_sack_lowest_newack == 0) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { + sctp_log_sack(*this_sack_lowest_newack, + last_tsn, + tp1->rec.data.tsn, + 0, + 0, + SCTP_LOG_TSN_ACKED); + } + *this_sack_lowest_newack = tp1->rec.data.tsn; + } + /*- + * CMT: CUCv2 algorithm. If (rtx-)pseudo-cumack for corresp + * dest is being acked, then we have a new (rtx-)pseudo-cumack. Set + * new_(rtx_)pseudo_cumack to TRUE so that the cwnd for this dest can be + * updated. Also trigger search for the next expected (rtx-)pseudo-cumack. + * Separate pseudo_cumack trackers for first transmissions and + * retransmissions. + */ + if (tp1->rec.data.tsn == tp1->whoTo->pseudo_cumack) { + if (tp1->rec.data.chunk_was_revoked == 0) { + tp1->whoTo->new_pseudo_cumack = 1; + } + tp1->whoTo->find_pseudo_cumack = 1; + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, tp1->whoTo, tp1->rec.data.tsn, SCTP_CWND_LOG_FROM_SACK); + } + if (tp1->rec.data.tsn == tp1->whoTo->rtx_pseudo_cumack) { + if (tp1->rec.data.chunk_was_revoked == 0) { + tp1->whoTo->new_pseudo_cumack = 1; + } + tp1->whoTo->find_rtx_pseudo_cumack = 1; + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { + sctp_log_sack(*biggest_newly_acked_tsn, + last_tsn, + tp1->rec.data.tsn, + frag_strt, + frag_end, + SCTP_LOG_TSN_ACKED); + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_GAP, + tp1->whoTo->flight_size, + tp1->book_size, + (uint32_t)(uintptr_t)tp1->whoTo, + tp1->rec.data.tsn); + } + sctp_flight_size_decrease(tp1); + if (stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) { + (*stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged)(tp1->whoTo, + tp1); + } + sctp_total_flight_decrease(stcb, tp1); + + tp1->whoTo->net_ack += tp1->send_size; + if (tp1->snd_count < 2) { + /*- + * True non-retransmitted chunk + */ + tp1->whoTo->net_ack2 += tp1->send_size; + + /*- + * update RTO too ? + */ + if (tp1->do_rtt) { + if (*rto_ok && + sctp_calculate_rto(stcb, + &stcb->asoc, + tp1->whoTo, + &tp1->sent_rcv_time, + SCTP_RTT_FROM_DATA)) { + *rto_ok = 0; + } + if (tp1->whoTo->rto_needed == 0) { + tp1->whoTo->rto_needed = 1; + } + tp1->do_rtt = 0; + } + } + } + if (tp1->sent <= SCTP_DATAGRAM_RESEND) { + if (SCTP_TSN_GT(tp1->rec.data.tsn, + stcb->asoc.this_sack_highest_gap)) { + stcb->asoc.this_sack_highest_gap = + tp1->rec.data.tsn; + } + if (tp1->sent == SCTP_DATAGRAM_RESEND) { + sctp_ucount_decr(stcb->asoc.sent_queue_retran_cnt); +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xB2, + (stcb->asoc.sent_queue_retran_cnt & 0x000000ff)); +#endif + } + } + /*- + * All chunks NOT UNSENT fall through here and are marked + * (leave PR-SCTP ones that are to skip alone though) + */ + if ((tp1->sent != SCTP_FORWARD_TSN_SKIP) && + (tp1->sent != SCTP_DATAGRAM_NR_ACKED)) { + tp1->sent = SCTP_DATAGRAM_MARKED; + } + if (tp1->rec.data.chunk_was_revoked) { + /* deflate the cwnd */ + tp1->whoTo->cwnd -= tp1->book_size; + tp1->rec.data.chunk_was_revoked = 0; + } + /* NR Sack code here */ + if (nr_sacking && + (tp1->sent != SCTP_DATAGRAM_NR_ACKED)) { + if (stcb->asoc.strmout[tp1->rec.data.sid].chunks_on_queues > 0) { + stcb->asoc.strmout[tp1->rec.data.sid].chunks_on_queues--; +#ifdef INVARIANTS + } else { + panic("No chunks on the queues for sid %u.", tp1->rec.data.sid); +#endif + } + if ((stcb->asoc.strmout[tp1->rec.data.sid].chunks_on_queues == 0) && + (stcb->asoc.strmout[tp1->rec.data.sid].state == SCTP_STREAM_RESET_PENDING) && + TAILQ_EMPTY(&stcb->asoc.strmout[tp1->rec.data.sid].outqueue)) { + stcb->asoc.trigger_reset = 1; + } + tp1->sent = SCTP_DATAGRAM_NR_ACKED; + if (tp1->data) { + /* sa_ignore NO_NULL_CHK */ + sctp_free_bufspace(stcb, &stcb->asoc, tp1, 1); + sctp_m_freem(tp1->data); + tp1->data = NULL; + } + wake_him++; + } + } + break; + } /* if (tp1->tsn == theTSN) */ + if (SCTP_TSN_GT(tp1->rec.data.tsn, theTSN)) { + break; + } + tp1 = TAILQ_NEXT(tp1, sctp_next); + if ((tp1 == NULL) && (circled == 0)) { + circled++; + tp1 = TAILQ_FIRST(&stcb->asoc.sent_queue); + } + } /* end while (tp1) */ + if (tp1 == NULL) { + circled = 0; + tp1 = TAILQ_FIRST(&stcb->asoc.sent_queue); + } + /* In case the fragments were not in order we must reset */ + } /* end for (j = fragStart */ + *p_tp1 = tp1; + return (wake_him); /* Return value only used for nr-sack */ +} + +static int +sctp_handle_segments(struct mbuf *m, int *offset, struct sctp_tcb *stcb, struct sctp_association *asoc, + uint32_t last_tsn, uint32_t *biggest_tsn_acked, + uint32_t *biggest_newly_acked_tsn, uint32_t *this_sack_lowest_newack, + int num_seg, int num_nr_seg, int *rto_ok) +{ + struct sctp_gap_ack_block *frag, block; + struct sctp_tmit_chunk *tp1; + int i; + int num_frs = 0; + int chunk_freed; + int non_revocable; + uint16_t frag_strt, frag_end, prev_frag_end; + + tp1 = TAILQ_FIRST(&asoc->sent_queue); + prev_frag_end = 0; + chunk_freed = 0; + + for (i = 0; i < (num_seg + num_nr_seg); i++) { + if (i == num_seg) { + prev_frag_end = 0; + tp1 = TAILQ_FIRST(&asoc->sent_queue); + } + frag = (struct sctp_gap_ack_block *)sctp_m_getptr(m, *offset, + sizeof(struct sctp_gap_ack_block), (uint8_t *) &block); + *offset += sizeof(block); + if (frag == NULL) { + return (chunk_freed); + } + frag_strt = ntohs(frag->start); + frag_end = ntohs(frag->end); + + if (frag_strt > frag_end) { + /* This gap report is malformed, skip it. */ + continue; + } + if (frag_strt <= prev_frag_end) { + /* This gap report is not in order, so restart. */ + tp1 = TAILQ_FIRST(&asoc->sent_queue); + } + if (SCTP_TSN_GT((last_tsn + frag_end), *biggest_tsn_acked)) { + *biggest_tsn_acked = last_tsn + frag_end; + } + if (i < num_seg) { + non_revocable = 0; + } else { + non_revocable = 1; + } + if (sctp_process_segment_range(stcb, &tp1, last_tsn, frag_strt, frag_end, + non_revocable, &num_frs, biggest_newly_acked_tsn, + this_sack_lowest_newack, rto_ok)) { + chunk_freed = 1; + } + prev_frag_end = frag_end; + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + if (num_frs) + sctp_log_fr(*biggest_tsn_acked, + *biggest_newly_acked_tsn, + last_tsn, SCTP_FR_LOG_BIGGEST_TSNS); + } + return (chunk_freed); +} + +static void +sctp_check_for_revoked(struct sctp_tcb *stcb, + struct sctp_association *asoc, uint32_t cumack, + uint32_t biggest_tsn_acked) +{ + struct sctp_tmit_chunk *tp1; + + TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { + if (SCTP_TSN_GT(tp1->rec.data.tsn, cumack)) { + /* + * ok this guy is either ACK or MARKED. If it is + * ACKED it has been previously acked but not this + * time i.e. revoked. If it is MARKED it was ACK'ed + * again. + */ + if (SCTP_TSN_GT(tp1->rec.data.tsn, biggest_tsn_acked)) { + break; + } + if (tp1->sent == SCTP_DATAGRAM_ACKED) { + /* it has been revoked */ + tp1->sent = SCTP_DATAGRAM_SENT; + tp1->rec.data.chunk_was_revoked = 1; + /* We must add this stuff back in to + * assure timers and such get started. + */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_UP_REVOKE, + tp1->whoTo->flight_size, + tp1->book_size, + (uint32_t)(uintptr_t)tp1->whoTo, + tp1->rec.data.tsn); + } + sctp_flight_size_increase(tp1); + sctp_total_flight_increase(stcb, tp1); + /* We inflate the cwnd to compensate for our + * artificial inflation of the flight_size. + */ + tp1->whoTo->cwnd += tp1->book_size; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { + sctp_log_sack(asoc->last_acked_seq, + cumack, + tp1->rec.data.tsn, + 0, + 0, + SCTP_LOG_TSN_REVOKED); + } + } else if (tp1->sent == SCTP_DATAGRAM_MARKED) { + /* it has been re-acked in this SACK */ + tp1->sent = SCTP_DATAGRAM_ACKED; + } + } + if (tp1->sent == SCTP_DATAGRAM_UNSENT) + break; + } +} + +static void +sctp_strike_gap_ack_chunks(struct sctp_tcb *stcb, struct sctp_association *asoc, + uint32_t biggest_tsn_acked, uint32_t biggest_tsn_newly_acked, uint32_t this_sack_lowest_newack, int accum_moved) +{ + struct sctp_tmit_chunk *tp1; + int strike_flag = 0; + struct timeval now; + uint32_t sending_seq; + struct sctp_nets *net; + int num_dests_sacked = 0; + + /* + * select the sending_seq, this is either the next thing ready to be + * sent but not transmitted, OR, the next seq we assign. + */ + tp1 = TAILQ_FIRST(&stcb->asoc.send_queue); + if (tp1 == NULL) { + sending_seq = asoc->sending_seq; + } else { + sending_seq = tp1->rec.data.tsn; + } + + /* CMT DAC algo: finding out if SACK is a mixed SACK */ + if ((asoc->sctp_cmt_on_off > 0) && + SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) { + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if (net->saw_newack) + num_dests_sacked++; + } + } + if (stcb->asoc.prsctp_supported) { + (void)SCTP_GETTIME_TIMEVAL(&now); + } + TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { + strike_flag = 0; + if (tp1->no_fr_allowed) { + /* this one had a timeout or something */ + continue; + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + if (tp1->sent < SCTP_DATAGRAM_RESEND) + sctp_log_fr(biggest_tsn_newly_acked, + tp1->rec.data.tsn, + tp1->sent, + SCTP_FR_LOG_CHECK_STRIKE); + } + if (SCTP_TSN_GT(tp1->rec.data.tsn, biggest_tsn_acked) || + tp1->sent == SCTP_DATAGRAM_UNSENT) { + /* done */ + break; + } + if (stcb->asoc.prsctp_supported) { + if ((PR_SCTP_TTL_ENABLED(tp1->flags)) && tp1->sent < SCTP_DATAGRAM_ACKED) { + /* Is it expired? */ +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + if (timercmp(&now, &tp1->rec.data.timetodrop, >)) { +#else + if (timevalcmp(&now, &tp1->rec.data.timetodrop, >)) { +#endif + /* Yes so drop it */ + if (tp1->data != NULL) { + (void)sctp_release_pr_sctp_chunk(stcb, tp1, 1, + SCTP_SO_NOT_LOCKED); + } + continue; + } + } + } + if (SCTP_TSN_GT(tp1->rec.data.tsn, asoc->this_sack_highest_gap) && + !(accum_moved && asoc->fast_retran_loss_recovery)) { + /* we are beyond the tsn in the sack */ + break; + } + if (tp1->sent >= SCTP_DATAGRAM_RESEND) { + /* either a RESEND, ACKED, or MARKED */ + /* skip */ + if (tp1->sent == SCTP_FORWARD_TSN_SKIP) { + /* Continue strikin FWD-TSN chunks */ + tp1->rec.data.fwd_tsn_cnt++; + } + continue; + } + /* + * CMT : SFR algo (covers part of DAC and HTNA as well) + */ + if (tp1->whoTo && tp1->whoTo->saw_newack == 0) { + /* + * No new acks were received for data sent to this + * dest. Therefore, according to the SFR algo for + * CMT, no data sent to this dest can be marked for + * FR using this SACK. + */ + continue; + } else if (tp1->whoTo && + SCTP_TSN_GT(tp1->rec.data.tsn, + tp1->whoTo->this_sack_highest_newack) && + !(accum_moved && asoc->fast_retran_loss_recovery)) { + /* + * CMT: New acks were received for data sent to + * this dest. But no new acks were seen for data + * sent after tp1. Therefore, according to the SFR + * algo for CMT, tp1 cannot be marked for FR using + * this SACK. This step covers part of the DAC algo + * and the HTNA algo as well. + */ + continue; + } + /* + * Here we check to see if we were have already done a FR + * and if so we see if the biggest TSN we saw in the sack is + * smaller than the recovery point. If so we don't strike + * the tsn... otherwise we CAN strike the TSN. + */ + /* + * @@@ JRI: Check for CMT + * if (accum_moved && asoc->fast_retran_loss_recovery && (sctp_cmt_on_off == 0)) { + */ + if (accum_moved && asoc->fast_retran_loss_recovery) { + /* + * Strike the TSN if in fast-recovery and cum-ack + * moved. + */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(biggest_tsn_newly_acked, + tp1->rec.data.tsn, + tp1->sent, + SCTP_FR_LOG_STRIKE_CHUNK); + } + if (tp1->sent < SCTP_DATAGRAM_RESEND) { + tp1->sent++; + } + if ((asoc->sctp_cmt_on_off > 0) && + SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) { + /* + * CMT DAC algorithm: If SACK flag is set to + * 0, then lowest_newack test will not pass + * because it would have been set to the + * cumack earlier. If not already to be + * rtx'd, If not a mixed sack and if tp1 is + * not between two sacked TSNs, then mark by + * one more. + * NOTE that we are marking by one additional time since the SACK DAC flag indicates that + * two packets have been received after this missing TSN. + */ + if ((tp1->sent < SCTP_DATAGRAM_RESEND) && (num_dests_sacked == 1) && + SCTP_TSN_GT(this_sack_lowest_newack, tp1->rec.data.tsn)) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(16 + num_dests_sacked, + tp1->rec.data.tsn, + tp1->sent, + SCTP_FR_LOG_STRIKE_CHUNK); + } + tp1->sent++; + } + } + } else if ((tp1->rec.data.doing_fast_retransmit) && + (asoc->sctp_cmt_on_off == 0)) { + /* + * For those that have done a FR we must take + * special consideration if we strike. I.e the + * biggest_newly_acked must be higher than the + * sending_seq at the time we did the FR. + */ + if ( +#ifdef SCTP_FR_TO_ALTERNATE + /* + * If FR's go to new networks, then we must only do + * this for singly homed asoc's. However if the FR's + * go to the same network (Armando's work) then its + * ok to FR multiple times. + */ + (asoc->numnets < 2) +#else + (1) +#endif + ) { + if (SCTP_TSN_GE(biggest_tsn_newly_acked, + tp1->rec.data.fast_retran_tsn)) { + /* + * Strike the TSN, since this ack is + * beyond where things were when we + * did a FR. + */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(biggest_tsn_newly_acked, + tp1->rec.data.tsn, + tp1->sent, + SCTP_FR_LOG_STRIKE_CHUNK); + } + if (tp1->sent < SCTP_DATAGRAM_RESEND) { + tp1->sent++; + } + strike_flag = 1; + if ((asoc->sctp_cmt_on_off > 0) && + SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) { + /* + * CMT DAC algorithm: If + * SACK flag is set to 0, + * then lowest_newack test + * will not pass because it + * would have been set to + * the cumack earlier. If + * not already to be rtx'd, + * If not a mixed sack and + * if tp1 is not between two + * sacked TSNs, then mark by + * one more. + * NOTE that we are marking by one additional time since the SACK DAC flag indicates that + * two packets have been received after this missing TSN. + */ + if ((tp1->sent < SCTP_DATAGRAM_RESEND) && + (num_dests_sacked == 1) && + SCTP_TSN_GT(this_sack_lowest_newack, + tp1->rec.data.tsn)) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(32 + num_dests_sacked, + tp1->rec.data.tsn, + tp1->sent, + SCTP_FR_LOG_STRIKE_CHUNK); + } + if (tp1->sent < SCTP_DATAGRAM_RESEND) { + tp1->sent++; + } + } + } + } + } + /* + * JRI: TODO: remove code for HTNA algo. CMT's + * SFR algo covers HTNA. + */ + } else if (SCTP_TSN_GT(tp1->rec.data.tsn, + biggest_tsn_newly_acked)) { + /* + * We don't strike these: This is the HTNA + * algorithm i.e. we don't strike If our TSN is + * larger than the Highest TSN Newly Acked. + */ + ; + } else { + /* Strike the TSN */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(biggest_tsn_newly_acked, + tp1->rec.data.tsn, + tp1->sent, + SCTP_FR_LOG_STRIKE_CHUNK); + } + if (tp1->sent < SCTP_DATAGRAM_RESEND) { + tp1->sent++; + } + if ((asoc->sctp_cmt_on_off > 0) && + SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) { + /* + * CMT DAC algorithm: If SACK flag is set to + * 0, then lowest_newack test will not pass + * because it would have been set to the + * cumack earlier. If not already to be + * rtx'd, If not a mixed sack and if tp1 is + * not between two sacked TSNs, then mark by + * one more. + * NOTE that we are marking by one additional time since the SACK DAC flag indicates that + * two packets have been received after this missing TSN. + */ + if ((tp1->sent < SCTP_DATAGRAM_RESEND) && (num_dests_sacked == 1) && + SCTP_TSN_GT(this_sack_lowest_newack, tp1->rec.data.tsn)) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(48 + num_dests_sacked, + tp1->rec.data.tsn, + tp1->sent, + SCTP_FR_LOG_STRIKE_CHUNK); + } + tp1->sent++; + } + } + } + if (tp1->sent == SCTP_DATAGRAM_RESEND) { + struct sctp_nets *alt; + + /* fix counts and things */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND, + (tp1->whoTo ? (tp1->whoTo->flight_size) : 0), + tp1->book_size, + (uint32_t)(uintptr_t)tp1->whoTo, + tp1->rec.data.tsn); + } + if (tp1->whoTo) { + tp1->whoTo->net_ack++; + sctp_flight_size_decrease(tp1); + if (stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) { + (*stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged)(tp1->whoTo, + tp1); + } + } + + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { + sctp_log_rwnd(SCTP_INCREASE_PEER_RWND, + asoc->peers_rwnd, tp1->send_size, SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)); + } + /* add back to the rwnd */ + asoc->peers_rwnd += (tp1->send_size + SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)); + + /* remove from the total flight */ + sctp_total_flight_decrease(stcb, tp1); + + if ((stcb->asoc.prsctp_supported) && + (PR_SCTP_RTX_ENABLED(tp1->flags))) { + /* Has it been retransmitted tv_sec times? - we store the retran count there. */ + if (tp1->snd_count > tp1->rec.data.timetodrop.tv_sec) { + /* Yes, so drop it */ + if (tp1->data != NULL) { + (void)sctp_release_pr_sctp_chunk(stcb, tp1, 1, + SCTP_SO_NOT_LOCKED); + } + /* Make sure to flag we had a FR */ + if (tp1->whoTo != NULL) { + tp1->whoTo->net_ack++; + } + continue; + } + } + /* SCTP_PRINTF("OK, we are now ready to FR this guy\n"); */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(tp1->rec.data.tsn, tp1->snd_count, + 0, SCTP_FR_MARKED); + } + if (strike_flag) { + /* This is a subsequent FR */ + SCTP_STAT_INCR(sctps_sendmultfastretrans); + } + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + if (asoc->sctp_cmt_on_off > 0) { + /* + * CMT: Using RTX_SSTHRESH policy for CMT. + * If CMT is being used, then pick dest with + * largest ssthresh for any retransmission. + */ + tp1->no_fr_allowed = 1; + alt = tp1->whoTo; + /*sa_ignore NO_NULL_CHK*/ + if (asoc->sctp_cmt_pf > 0) { + /* JRS 5/18/07 - If CMT PF is on, use the PF version of find_alt_net() */ + alt = sctp_find_alternate_net(stcb, alt, 2); + } else { + /* JRS 5/18/07 - If only CMT is on, use the CMT version of find_alt_net() */ + /*sa_ignore NO_NULL_CHK*/ + alt = sctp_find_alternate_net(stcb, alt, 1); + } + if (alt == NULL) { + alt = tp1->whoTo; + } + /* + * CUCv2: If a different dest is picked for + * the retransmission, then new + * (rtx-)pseudo_cumack needs to be tracked + * for orig dest. Let CUCv2 track new (rtx-) + * pseudo-cumack always. + */ + if (tp1->whoTo) { + tp1->whoTo->find_pseudo_cumack = 1; + tp1->whoTo->find_rtx_pseudo_cumack = 1; + } + } else {/* CMT is OFF */ +#ifdef SCTP_FR_TO_ALTERNATE + /* Can we find an alternate? */ + alt = sctp_find_alternate_net(stcb, tp1->whoTo, 0); +#else + /* + * default behavior is to NOT retransmit + * FR's to an alternate. Armando Caro's + * paper details why. + */ + alt = tp1->whoTo; +#endif + } + + tp1->rec.data.doing_fast_retransmit = 1; + /* mark the sending seq for possible subsequent FR's */ + /* + * SCTP_PRINTF("Marking TSN for FR new value %x\n", + * (uint32_t)tpi->rec.data.tsn); + */ + if (TAILQ_EMPTY(&asoc->send_queue)) { + /* + * If the queue of send is empty then its + * the next sequence number that will be + * assigned so we subtract one from this to + * get the one we last sent. + */ + tp1->rec.data.fast_retran_tsn = sending_seq; + } else { + /* + * If there are chunks on the send queue + * (unsent data that has made it from the + * stream queues but not out the door, we + * take the first one (which will have the + * lowest TSN) and subtract one to get the + * one we last sent. + */ + struct sctp_tmit_chunk *ttt; + + ttt = TAILQ_FIRST(&asoc->send_queue); + tp1->rec.data.fast_retran_tsn = + ttt->rec.data.tsn; + } + + if (tp1->do_rtt) { + /* + * this guy had a RTO calculation pending on + * it, cancel it + */ + if ((tp1->whoTo != NULL) && + (tp1->whoTo->rto_needed == 0)) { + tp1->whoTo->rto_needed = 1; + } + tp1->do_rtt = 0; + } + if (alt != tp1->whoTo) { + /* yes, there is an alternate. */ + sctp_free_remote_addr(tp1->whoTo); + /*sa_ignore FREED_MEMORY*/ + tp1->whoTo = alt; + atomic_add_int(&alt->ref_count, 1); + } + } + } +} + +struct sctp_tmit_chunk * +sctp_try_advance_peer_ack_point(struct sctp_tcb *stcb, + struct sctp_association *asoc) +{ + struct sctp_tmit_chunk *tp1, *tp2, *a_adv = NULL; + struct timeval now; + int now_filled = 0; + + if (asoc->prsctp_supported == 0) { + return (NULL); + } + TAILQ_FOREACH_SAFE(tp1, &asoc->sent_queue, sctp_next, tp2) { + if (tp1->sent != SCTP_FORWARD_TSN_SKIP && + tp1->sent != SCTP_DATAGRAM_RESEND && + tp1->sent != SCTP_DATAGRAM_NR_ACKED) { + /* no chance to advance, out of here */ + break; + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_TRY_ADVANCE) { + if ((tp1->sent == SCTP_FORWARD_TSN_SKIP) || + (tp1->sent == SCTP_DATAGRAM_NR_ACKED)) { + sctp_misc_ints(SCTP_FWD_TSN_CHECK, + asoc->advanced_peer_ack_point, + tp1->rec.data.tsn, 0, 0); + } + } + if (!PR_SCTP_ENABLED(tp1->flags)) { + /* + * We can't fwd-tsn past any that are reliable aka + * retransmitted until the asoc fails. + */ + break; + } + if (!now_filled) { + (void)SCTP_GETTIME_TIMEVAL(&now); + now_filled = 1; + } + /* + * now we got a chunk which is marked for another + * retransmission to a PR-stream but has run out its chances + * already maybe OR has been marked to skip now. Can we skip + * it if its a resend? + */ + if (tp1->sent == SCTP_DATAGRAM_RESEND && + (PR_SCTP_TTL_ENABLED(tp1->flags))) { + /* + * Now is this one marked for resend and its time is + * now up? + */ +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + if (timercmp(&now, &tp1->rec.data.timetodrop, >)) { +#else + if (timevalcmp(&now, &tp1->rec.data.timetodrop, >)) { +#endif + /* Yes so drop it */ + if (tp1->data) { + (void)sctp_release_pr_sctp_chunk(stcb, tp1, + 1, SCTP_SO_NOT_LOCKED); + } + } else { + /* + * No, we are done when hit one for resend + * whos time as not expired. + */ + break; + } + } + /* + * Ok now if this chunk is marked to drop it we can clean up + * the chunk, advance our peer ack point and we can check + * the next chunk. + */ + if ((tp1->sent == SCTP_FORWARD_TSN_SKIP) || + (tp1->sent == SCTP_DATAGRAM_NR_ACKED)) { + /* advance PeerAckPoint goes forward */ + if (SCTP_TSN_GT(tp1->rec.data.tsn, asoc->advanced_peer_ack_point)) { + asoc->advanced_peer_ack_point = tp1->rec.data.tsn; + a_adv = tp1; + } else if (tp1->rec.data.tsn == asoc->advanced_peer_ack_point) { + /* No update but we do save the chk */ + a_adv = tp1; + } + } else { + /* + * If it is still in RESEND we can advance no + * further + */ + break; + } + } + return (a_adv); +} + +static int +sctp_fs_audit(struct sctp_association *asoc) +{ + struct sctp_tmit_chunk *chk; + int inflight = 0, resend = 0, inbetween = 0, acked = 0, above = 0; + int ret; +#ifndef INVARIANTS + int entry_flight, entry_cnt; +#endif + + ret = 0; +#ifndef INVARIANTS + entry_flight = asoc->total_flight; + entry_cnt = asoc->total_flight_count; +#endif + if (asoc->pr_sctp_cnt >= asoc->sent_queue_cnt) + return (0); + + TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { + if (chk->sent < SCTP_DATAGRAM_RESEND) { + SCTP_PRINTF("Chk TSN: %u size: %d inflight cnt: %d\n", + chk->rec.data.tsn, + chk->send_size, + chk->snd_count); + inflight++; + } else if (chk->sent == SCTP_DATAGRAM_RESEND) { + resend++; + } else if (chk->sent < SCTP_DATAGRAM_ACKED) { + inbetween++; + } else if (chk->sent > SCTP_DATAGRAM_ACKED) { + above++; + } else { + acked++; + } + } + + if ((inflight > 0) || (inbetween > 0)) { +#ifdef INVARIANTS + panic("Flight size-express incorrect F: %d I: %d R: %d Ab: %d ACK: %d", + inflight, inbetween, resend, above, acked); +#else + SCTP_PRINTF("asoc->total_flight: %d cnt: %d\n", + entry_flight, entry_cnt); + SCTP_PRINTF("Flight size-express incorrect F: %d I: %d R: %d Ab: %d ACK: %d\n", + inflight, inbetween, resend, above, acked); + ret = 1; +#endif + } + return (ret); +} + +static void +sctp_window_probe_recovery(struct sctp_tcb *stcb, + struct sctp_association *asoc, + struct sctp_tmit_chunk *tp1) +{ + tp1->window_probe = 0; + if ((tp1->sent >= SCTP_DATAGRAM_ACKED) || (tp1->data == NULL)) { + /* TSN's skipped we do NOT move back. */ + sctp_misc_ints(SCTP_FLIGHT_LOG_DWN_WP_FWD, + tp1->whoTo ? tp1->whoTo->flight_size : 0, + tp1->book_size, + (uint32_t)(uintptr_t)tp1->whoTo, + tp1->rec.data.tsn); + return; + } + /* First setup this by shrinking flight */ + if (stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) { + (*stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged)(tp1->whoTo, + tp1); + } + sctp_flight_size_decrease(tp1); + sctp_total_flight_decrease(stcb, tp1); + /* Now mark for resend */ + tp1->sent = SCTP_DATAGRAM_RESEND; + sctp_ucount_incr(asoc->sent_queue_retran_cnt); + + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_WP, + tp1->whoTo->flight_size, + tp1->book_size, + (uint32_t)(uintptr_t)tp1->whoTo, + tp1->rec.data.tsn); + } +} + +void +sctp_express_handle_sack(struct sctp_tcb *stcb, uint32_t cumack, + uint32_t rwnd, int *abort_now, int ecne_seen) +{ + struct sctp_nets *net; + struct sctp_association *asoc; + struct sctp_tmit_chunk *tp1, *tp2; + uint32_t old_rwnd; + int win_probe_recovery = 0; + int win_probe_recovered = 0; + int j, done_once = 0; + int rto_ok = 1; + uint32_t send_s; + + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_SACK_ARRIVALS_ENABLE) { + sctp_misc_ints(SCTP_SACK_LOG_EXPRESS, cumack, + rwnd, stcb->asoc.last_acked_seq, stcb->asoc.peers_rwnd); + } + SCTP_TCB_LOCK_ASSERT(stcb); +#ifdef SCTP_ASOCLOG_OF_TSNS + stcb->asoc.cumack_log[stcb->asoc.cumack_log_at] = cumack; + stcb->asoc.cumack_log_at++; + if (stcb->asoc.cumack_log_at > SCTP_TSN_LOG_SIZE) { + stcb->asoc.cumack_log_at = 0; + } +#endif + asoc = &stcb->asoc; + old_rwnd = asoc->peers_rwnd; + if (SCTP_TSN_GT(asoc->last_acked_seq, cumack)) { + /* old ack */ + return; + } else if (asoc->last_acked_seq == cumack) { + /* Window update sack */ + asoc->peers_rwnd = sctp_sbspace_sub(rwnd, + (uint32_t) (asoc->total_flight + (asoc->total_flight_count * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)))); + if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { + /* SWS sender side engages */ + asoc->peers_rwnd = 0; + } + if (asoc->peers_rwnd > old_rwnd) { + goto again; + } + return; + } + + /* First setup for CC stuff */ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if (SCTP_TSN_GT(cumack, net->cwr_window_tsn)) { + /* Drag along the window_tsn for cwr's */ + net->cwr_window_tsn = cumack; + } + net->prev_cwnd = net->cwnd; + net->net_ack = 0; + net->net_ack2 = 0; + + /* + * CMT: Reset CUC and Fast recovery algo variables before + * SACK processing + */ + net->new_pseudo_cumack = 0; + net->will_exit_fast_recovery = 0; + if (stcb->asoc.cc_functions.sctp_cwnd_prepare_net_for_sack) { + (*stcb->asoc.cc_functions.sctp_cwnd_prepare_net_for_sack)(stcb, net); + } + } + if (!TAILQ_EMPTY(&asoc->sent_queue)) { + tp1 = TAILQ_LAST(&asoc->sent_queue, + sctpchunk_listhead); + send_s = tp1->rec.data.tsn + 1; + } else { + send_s = asoc->sending_seq; + } + if (SCTP_TSN_GE(cumack, send_s)) { + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + *abort_now = 1; + /* XXX */ + SCTP_SNPRINTF(msg, sizeof(msg), + "Cum ack %8.8x greater or equal than TSN %8.8x", + cumack, send_s); + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_25; + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + return; + } + asoc->this_sack_highest_gap = cumack; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { + sctp_misc_ints(SCTP_THRESHOLD_CLEAR, + stcb->asoc.overall_error_count, + 0, + SCTP_FROM_SCTP_INDATA, + __LINE__); + } + stcb->asoc.overall_error_count = 0; + if (SCTP_TSN_GT(cumack, asoc->last_acked_seq)) { + /* process the new consecutive TSN first */ + TAILQ_FOREACH_SAFE(tp1, &asoc->sent_queue, sctp_next, tp2) { + if (SCTP_TSN_GE(cumack, tp1->rec.data.tsn)) { + if (tp1->sent == SCTP_DATAGRAM_UNSENT) { + SCTP_PRINTF("Warning, an unsent is now acked?\n"); + } + if (tp1->sent < SCTP_DATAGRAM_ACKED) { + /* + * If it is less than ACKED, it is + * now no-longer in flight. Higher + * values may occur during marking + */ + if (tp1->sent < SCTP_DATAGRAM_RESEND) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_CA, + tp1->whoTo->flight_size, + tp1->book_size, + (uint32_t)(uintptr_t)tp1->whoTo, + tp1->rec.data.tsn); + } + sctp_flight_size_decrease(tp1); + if (stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) { + (*stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged)(tp1->whoTo, + tp1); + } + /* sa_ignore NO_NULL_CHK */ + sctp_total_flight_decrease(stcb, tp1); + } + tp1->whoTo->net_ack += tp1->send_size; + if (tp1->snd_count < 2) { + /* + * True non-retransmitted + * chunk + */ + tp1->whoTo->net_ack2 += + tp1->send_size; + + /* update RTO too? */ + if (tp1->do_rtt) { + if (rto_ok && + sctp_calculate_rto(stcb, + &stcb->asoc, + tp1->whoTo, + &tp1->sent_rcv_time, + SCTP_RTT_FROM_DATA)) { + rto_ok = 0; + } + if (tp1->whoTo->rto_needed == 0) { + tp1->whoTo->rto_needed = 1; + } + tp1->do_rtt = 0; + } + } + /* + * CMT: CUCv2 algorithm. From the + * cumack'd TSNs, for each TSN being + * acked for the first time, set the + * following variables for the + * corresp destination. + * new_pseudo_cumack will trigger a + * cwnd update. + * find_(rtx_)pseudo_cumack will + * trigger search for the next + * expected (rtx-)pseudo-cumack. + */ + tp1->whoTo->new_pseudo_cumack = 1; + tp1->whoTo->find_pseudo_cumack = 1; + tp1->whoTo->find_rtx_pseudo_cumack = 1; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + /* sa_ignore NO_NULL_CHK */ + sctp_log_cwnd(stcb, tp1->whoTo, tp1->rec.data.tsn, SCTP_CWND_LOG_FROM_SACK); + } + } + if (tp1->sent == SCTP_DATAGRAM_RESEND) { + sctp_ucount_decr(asoc->sent_queue_retran_cnt); + } + if (tp1->rec.data.chunk_was_revoked) { + /* deflate the cwnd */ + tp1->whoTo->cwnd -= tp1->book_size; + tp1->rec.data.chunk_was_revoked = 0; + } + if (tp1->sent != SCTP_DATAGRAM_NR_ACKED) { + if (asoc->strmout[tp1->rec.data.sid].chunks_on_queues > 0) { + asoc->strmout[tp1->rec.data.sid].chunks_on_queues--; +#ifdef INVARIANTS + } else { + panic("No chunks on the queues for sid %u.", tp1->rec.data.sid); +#endif + } + } + if ((asoc->strmout[tp1->rec.data.sid].chunks_on_queues == 0) && + (asoc->strmout[tp1->rec.data.sid].state == SCTP_STREAM_RESET_PENDING) && + TAILQ_EMPTY(&asoc->strmout[tp1->rec.data.sid].outqueue)) { + asoc->trigger_reset = 1; + } + TAILQ_REMOVE(&asoc->sent_queue, tp1, sctp_next); + if (tp1->data) { + /* sa_ignore NO_NULL_CHK */ + sctp_free_bufspace(stcb, asoc, tp1, 1); + sctp_m_freem(tp1->data); + tp1->data = NULL; + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { + sctp_log_sack(asoc->last_acked_seq, + cumack, + tp1->rec.data.tsn, + 0, + 0, + SCTP_LOG_FREE_SENT); + } + asoc->sent_queue_cnt--; + sctp_free_a_chunk(stcb, tp1, SCTP_SO_NOT_LOCKED); + } else { + break; + } + } + } +#if defined(__Userspace__) + if (stcb->sctp_ep->recv_callback) { + if (stcb->sctp_socket) { + uint32_t inqueue_bytes, sb_free_now; + struct sctp_inpcb *inp; + + inp = stcb->sctp_ep; + inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); + sb_free_now = SCTP_SB_LIMIT_SND(stcb->sctp_socket) - (inqueue_bytes + stcb->asoc.sb_send_resv); + + /* check if the amount free in the send socket buffer crossed the threshold */ + if (inp->send_callback && + (((inp->send_sb_threshold > 0) && + (sb_free_now >= inp->send_sb_threshold) && + (stcb->asoc.chunks_on_out_queue <= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue))) || + (inp->send_sb_threshold == 0))) { + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + inp->send_callback(stcb->sctp_socket, sb_free_now, inp->ulp_info); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } + } + } else if (stcb->sctp_socket) { +#else + /* sa_ignore NO_NULL_CHK */ + if (stcb->sctp_socket) { +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; + +#endif + SOCKBUF_LOCK(&stcb->sctp_socket->so_snd); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_WAKE_LOGGING_ENABLE) { + /* sa_ignore NO_NULL_CHK */ + sctp_wakeup_log(stcb, 1, SCTP_WAKESND_FROM_SACK); + } +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(stcb->sctp_ep); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { + /* assoc was freed while we were unlocked */ + SCTP_SOCKET_UNLOCK(so, 1); + return; + } +#endif + sctp_sowwakeup_locked(stcb->sctp_ep, stcb->sctp_socket); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_WAKE_LOGGING_ENABLE) { + sctp_wakeup_log(stcb, 1, SCTP_NOWAKE_FROM_SACK); + } + } + + /* JRS - Use the congestion control given in the CC module */ + if ((asoc->last_acked_seq != cumack) && (ecne_seen == 0)) { + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if (net->net_ack2 > 0) { + /* + * Karn's rule applies to clearing error count, this + * is optional. + */ + net->error_count = 0; + if ((net->dest_state & SCTP_ADDR_REACHABLE) == 0) { + /* addr came good */ + net->dest_state |= SCTP_ADDR_REACHABLE; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, + 0, (void *)net, SCTP_SO_NOT_LOCKED); + } + if (net == stcb->asoc.primary_destination) { + if (stcb->asoc.alternate) { + /* release the alternate, primary is good */ + sctp_free_remote_addr(stcb->asoc.alternate); + stcb->asoc.alternate = NULL; + } + } + if (net->dest_state & SCTP_ADDR_PF) { + net->dest_state &= ~SCTP_ADDR_PF; + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, + stcb->sctp_ep, stcb, net, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_26); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); + asoc->cc_functions.sctp_cwnd_update_exit_pf(stcb, net); + /* Done with this net */ + net->net_ack = 0; + } + /* restore any doubled timers */ + net->RTO = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv; + if (net->RTO < stcb->asoc.minrto) { + net->RTO = stcb->asoc.minrto; + } + if (net->RTO > stcb->asoc.maxrto) { + net->RTO = stcb->asoc.maxrto; + } + } + } + asoc->cc_functions.sctp_cwnd_update_after_sack(stcb, asoc, 1, 0, 0); + } + asoc->last_acked_seq = cumack; + + if (TAILQ_EMPTY(&asoc->sent_queue)) { + /* nothing left in-flight */ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + net->flight_size = 0; + net->partial_bytes_acked = 0; + } + asoc->total_flight = 0; + asoc->total_flight_count = 0; + } + + /* RWND update */ + asoc->peers_rwnd = sctp_sbspace_sub(rwnd, + (uint32_t) (asoc->total_flight + (asoc->total_flight_count * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)))); + if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { + /* SWS sender side engages */ + asoc->peers_rwnd = 0; + } + if (asoc->peers_rwnd > old_rwnd) { + win_probe_recovery = 1; + } + /* Now assure a timer where data is queued at */ +again: + j = 0; + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if (win_probe_recovery && (net->window_probe)) { + win_probe_recovered = 1; + /* + * Find first chunk that was used with window probe + * and clear the sent + */ + /* sa_ignore FREED_MEMORY */ + TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { + if (tp1->window_probe) { + /* move back to data send queue */ + sctp_window_probe_recovery(stcb, asoc, tp1); + break; + } + } + } + if (net->flight_size) { + j++; + sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net); + if (net->window_probe) { + net->window_probe = 0; + } + } else { + if (net->window_probe) { + /* In window probes we must assure a timer is still running there */ + net->window_probe = 0; + if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { + sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, net); + } + } else if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, + stcb, net, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_27); + } + } + } + if ((j == 0) && + (!TAILQ_EMPTY(&asoc->sent_queue)) && + (asoc->sent_queue_retran_cnt == 0) && + (win_probe_recovered == 0) && + (done_once == 0)) { + /* huh, this should not happen unless all packets + * are PR-SCTP and marked to skip of course. + */ + if (sctp_fs_audit(asoc)) { + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + net->flight_size = 0; + } + asoc->total_flight = 0; + asoc->total_flight_count = 0; + asoc->sent_queue_retran_cnt = 0; + TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { + if (tp1->sent < SCTP_DATAGRAM_RESEND) { + sctp_flight_size_increase(tp1); + sctp_total_flight_increase(stcb, tp1); + } else if (tp1->sent == SCTP_DATAGRAM_RESEND) { + sctp_ucount_incr(asoc->sent_queue_retran_cnt); + } + } + } + done_once = 1; + goto again; + } + /**********************************/ + /* Now what about shutdown issues */ + /**********************************/ + if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue)) { + /* nothing left on sendqueue.. consider done */ + /* clean up */ + if ((asoc->stream_queue_cnt == 1) && + ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) && + ((*asoc->ss_functions.sctp_ss_is_user_msgs_incomplete)(stcb, asoc))) { + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_PARTIAL_MSG_LEFT); + } + if (((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) && + (asoc->stream_queue_cnt == 1) && + (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { + struct mbuf *op_err; + + *abort_now = 1; + /* XXX */ + op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_28; + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + return; + } + if ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) && + (asoc->stream_queue_cnt == 0)) { + struct sctp_nets *netp; + + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_SENT); + sctp_stop_timers_for_shutdown(stcb); + if (asoc->alternate) { + netp = asoc->alternate; + } else { + netp = asoc->primary_destination; + } + sctp_send_shutdown(stcb, netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, + stcb->sctp_ep, stcb, netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, + stcb->sctp_ep, stcb, NULL); + } else if ((SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED) && + (asoc->stream_queue_cnt == 0)) { + struct sctp_nets *netp; + + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_ACK_SENT); + sctp_stop_timers_for_shutdown(stcb); + if (asoc->alternate) { + netp = asoc->alternate; + } else { + netp = asoc->primary_destination; + } + sctp_send_shutdown_ack(stcb, netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, + stcb->sctp_ep, stcb, netp); + } + } + /*********************************************/ + /* Here we perform PR-SCTP procedures */ + /* (section 4.2) */ + /*********************************************/ + /* C1. update advancedPeerAckPoint */ + if (SCTP_TSN_GT(cumack, asoc->advanced_peer_ack_point)) { + asoc->advanced_peer_ack_point = cumack; + } + /* PR-Sctp issues need to be addressed too */ + if ((asoc->prsctp_supported) && (asoc->pr_sctp_cnt > 0)) { + struct sctp_tmit_chunk *lchk; + uint32_t old_adv_peer_ack_point; + + old_adv_peer_ack_point = asoc->advanced_peer_ack_point; + lchk = sctp_try_advance_peer_ack_point(stcb, asoc); + /* C3. See if we need to send a Fwd-TSN */ + if (SCTP_TSN_GT(asoc->advanced_peer_ack_point, cumack)) { + /* + * ISSUE with ECN, see FWD-TSN processing. + */ + if (SCTP_TSN_GT(asoc->advanced_peer_ack_point, old_adv_peer_ack_point)) { + send_forward_tsn(stcb, asoc); + } else if (lchk) { + /* try to FR fwd-tsn's that get lost too */ + if (lchk->rec.data.fwd_tsn_cnt >= 3) { + send_forward_tsn(stcb, asoc); + } + } + } + for (; lchk != NULL; lchk = TAILQ_NEXT(lchk, sctp_next)) { + if (lchk->whoTo != NULL) { + break; + } + } + if (lchk != NULL) { + /* Assure a timer is up */ + sctp_timer_start(SCTP_TIMER_TYPE_SEND, + stcb->sctp_ep, stcb, lchk->whoTo); + } + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_RWND_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_SACK_RWND_UPDATE, + rwnd, + stcb->asoc.peers_rwnd, + stcb->asoc.total_flight, + stcb->asoc.total_output_queue_size); + } +} + +void +sctp_handle_sack(struct mbuf *m, int offset_seg, int offset_dup, + struct sctp_tcb *stcb, + uint16_t num_seg, uint16_t num_nr_seg, uint16_t num_dup, + int *abort_now, uint8_t flags, + uint32_t cum_ack, uint32_t rwnd, int ecne_seen) +{ + struct sctp_association *asoc; + struct sctp_tmit_chunk *tp1, *tp2; + uint32_t last_tsn, biggest_tsn_acked, biggest_tsn_newly_acked, this_sack_lowest_newack; + uint16_t wake_him = 0; + uint32_t send_s = 0; + long j; + int accum_moved = 0; + int will_exit_fast_recovery = 0; + uint32_t a_rwnd, old_rwnd; + int win_probe_recovery = 0; + int win_probe_recovered = 0; + struct sctp_nets *net = NULL; + int done_once; + int rto_ok = 1; + uint8_t reneged_all = 0; + uint8_t cmt_dac_flag; + /* + * we take any chance we can to service our queues since we cannot + * get awoken when the socket is read from :< + */ + /* + * Now perform the actual SACK handling: 1) Verify that it is not an + * old sack, if so discard. 2) If there is nothing left in the send + * queue (cum-ack is equal to last acked) then you have a duplicate + * too, update any rwnd change and verify no timers are running. + * then return. 3) Process any new consecutive data i.e. cum-ack + * moved process these first and note that it moved. 4) Process any + * sack blocks. 5) Drop any acked from the queue. 6) Check for any + * revoked blocks and mark. 7) Update the cwnd. 8) Nothing left, + * sync up flightsizes and things, stop all timers and also check + * for shutdown_pending state. If so then go ahead and send off the + * shutdown. If in shutdown recv, send off the shutdown-ack and + * start that timer, Ret. 9) Strike any non-acked things and do FR + * procedure if needed being sure to set the FR flag. 10) Do pr-sctp + * procedures. 11) Apply any FR penalties. 12) Assure we will SACK + * if in shutdown_recv state. + */ + SCTP_TCB_LOCK_ASSERT(stcb); + /* CMT DAC algo */ + this_sack_lowest_newack = 0; + SCTP_STAT_INCR(sctps_slowpath_sack); + last_tsn = cum_ack; + cmt_dac_flag = flags & SCTP_SACK_CMT_DAC; +#ifdef SCTP_ASOCLOG_OF_TSNS + stcb->asoc.cumack_log[stcb->asoc.cumack_log_at] = cum_ack; + stcb->asoc.cumack_log_at++; + if (stcb->asoc.cumack_log_at > SCTP_TSN_LOG_SIZE) { + stcb->asoc.cumack_log_at = 0; + } +#endif + a_rwnd = rwnd; + + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_SACK_ARRIVALS_ENABLE) { + sctp_misc_ints(SCTP_SACK_LOG_NORMAL, cum_ack, + rwnd, stcb->asoc.last_acked_seq, stcb->asoc.peers_rwnd); + } + + old_rwnd = stcb->asoc.peers_rwnd; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { + sctp_misc_ints(SCTP_THRESHOLD_CLEAR, + stcb->asoc.overall_error_count, + 0, + SCTP_FROM_SCTP_INDATA, + __LINE__); + } + stcb->asoc.overall_error_count = 0; + asoc = &stcb->asoc; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { + sctp_log_sack(asoc->last_acked_seq, + cum_ack, + 0, + num_seg, + num_dup, + SCTP_LOG_NEW_SACK); + } + if ((num_dup) && (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE)) { + uint16_t i; + uint32_t *dupdata, dblock; + + for (i = 0; i < num_dup; i++) { + dupdata = (uint32_t *)sctp_m_getptr(m, offset_dup + i * sizeof(uint32_t), + sizeof(uint32_t), (uint8_t *)&dblock); + if (dupdata == NULL) { + break; + } + sctp_log_fr(*dupdata, 0, 0, SCTP_FR_DUPED); + } + } + /* reality check */ + if (!TAILQ_EMPTY(&asoc->sent_queue)) { + tp1 = TAILQ_LAST(&asoc->sent_queue, + sctpchunk_listhead); + send_s = tp1->rec.data.tsn + 1; + } else { + tp1 = NULL; + send_s = asoc->sending_seq; + } + if (SCTP_TSN_GE(cum_ack, send_s)) { + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + /* + * no way, we have not even sent this TSN out yet. + * Peer is hopelessly messed up with us. + */ + SCTP_PRINTF("NEW cum_ack:%x send_s:%x is smaller or equal\n", + cum_ack, send_s); + if (tp1) { + SCTP_PRINTF("Got send_s from tsn:%x + 1 of tp1: %p\n", + tp1->rec.data.tsn, (void *)tp1); + } + hopeless_peer: + *abort_now = 1; + /* XXX */ + SCTP_SNPRINTF(msg, sizeof(msg), + "Cum ack %8.8x greater or equal than TSN %8.8x", + cum_ack, send_s); + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_29; + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + return; + } + /**********************/ + /* 1) check the range */ + /**********************/ + if (SCTP_TSN_GT(asoc->last_acked_seq, last_tsn)) { + /* acking something behind */ + return; + } + + /* update the Rwnd of the peer */ + if (TAILQ_EMPTY(&asoc->sent_queue) && + TAILQ_EMPTY(&asoc->send_queue) && + (asoc->stream_queue_cnt == 0)) { + /* nothing left on send/sent and strmq */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { + sctp_log_rwnd_set(SCTP_SET_PEER_RWND_VIA_SACK, + asoc->peers_rwnd, 0, 0, a_rwnd); + } + asoc->peers_rwnd = a_rwnd; + if (asoc->sent_queue_retran_cnt) { + asoc->sent_queue_retran_cnt = 0; + } + if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { + /* SWS sender side engages */ + asoc->peers_rwnd = 0; + } + /* stop any timers */ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, + stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_30); + net->partial_bytes_acked = 0; + net->flight_size = 0; + } + asoc->total_flight = 0; + asoc->total_flight_count = 0; + return; + } + /* + * We init netAckSz and netAckSz2 to 0. These are used to track 2 + * things. The total byte count acked is tracked in netAckSz AND + * netAck2 is used to track the total bytes acked that are un- + * ambiguous and were never retransmitted. We track these on a per + * destination address basis. + */ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if (SCTP_TSN_GT(cum_ack, net->cwr_window_tsn)) { + /* Drag along the window_tsn for cwr's */ + net->cwr_window_tsn = cum_ack; + } + net->prev_cwnd = net->cwnd; + net->net_ack = 0; + net->net_ack2 = 0; + + /* + * CMT: Reset CUC and Fast recovery algo variables before + * SACK processing + */ + net->new_pseudo_cumack = 0; + net->will_exit_fast_recovery = 0; + if (stcb->asoc.cc_functions.sctp_cwnd_prepare_net_for_sack) { + (*stcb->asoc.cc_functions.sctp_cwnd_prepare_net_for_sack)(stcb, net); + } + + /* + * CMT: SFR algo (and HTNA) - this_sack_highest_newack has + * to be greater than the cumack. Also reset saw_newack to 0 + * for all dests. + */ + net->saw_newack = 0; + net->this_sack_highest_newack = last_tsn; + } + /* process the new consecutive TSN first */ + TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { + if (SCTP_TSN_GE(last_tsn, tp1->rec.data.tsn)) { + if (tp1->sent != SCTP_DATAGRAM_UNSENT) { + accum_moved = 1; + if (tp1->sent < SCTP_DATAGRAM_ACKED) { + /* + * If it is less than ACKED, it is + * now no-longer in flight. Higher + * values may occur during marking + */ + if ((tp1->whoTo->dest_state & + SCTP_ADDR_UNCONFIRMED) && + (tp1->snd_count < 2)) { + /* + * If there was no retran + * and the address is + * un-confirmed and we sent + * there and are now + * sacked.. its confirmed, + * mark it so. + */ + tp1->whoTo->dest_state &= + ~SCTP_ADDR_UNCONFIRMED; + } + if (tp1->sent < SCTP_DATAGRAM_RESEND) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_CA, + tp1->whoTo->flight_size, + tp1->book_size, + (uint32_t)(uintptr_t)tp1->whoTo, + tp1->rec.data.tsn); + } + sctp_flight_size_decrease(tp1); + sctp_total_flight_decrease(stcb, tp1); + if (stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged) { + (*stcb->asoc.cc_functions.sctp_cwnd_update_tsn_acknowledged)(tp1->whoTo, + tp1); + } + } + tp1->whoTo->net_ack += tp1->send_size; + + /* CMT SFR and DAC algos */ + this_sack_lowest_newack = tp1->rec.data.tsn; + tp1->whoTo->saw_newack = 1; + + if (tp1->snd_count < 2) { + /* + * True non-retransmitted + * chunk + */ + tp1->whoTo->net_ack2 += + tp1->send_size; + + /* update RTO too? */ + if (tp1->do_rtt) { + if (rto_ok && + sctp_calculate_rto(stcb, + &stcb->asoc, + tp1->whoTo, + &tp1->sent_rcv_time, + SCTP_RTT_FROM_DATA)) { + rto_ok = 0; + } + if (tp1->whoTo->rto_needed == 0) { + tp1->whoTo->rto_needed = 1; + } + tp1->do_rtt = 0; + } + } + /* + * CMT: CUCv2 algorithm. From the + * cumack'd TSNs, for each TSN being + * acked for the first time, set the + * following variables for the + * corresp destination. + * new_pseudo_cumack will trigger a + * cwnd update. + * find_(rtx_)pseudo_cumack will + * trigger search for the next + * expected (rtx-)pseudo-cumack. + */ + tp1->whoTo->new_pseudo_cumack = 1; + tp1->whoTo->find_pseudo_cumack = 1; + tp1->whoTo->find_rtx_pseudo_cumack = 1; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { + sctp_log_sack(asoc->last_acked_seq, + cum_ack, + tp1->rec.data.tsn, + 0, + 0, + SCTP_LOG_TSN_ACKED); + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, tp1->whoTo, tp1->rec.data.tsn, SCTP_CWND_LOG_FROM_SACK); + } + } + if (tp1->sent == SCTP_DATAGRAM_RESEND) { + sctp_ucount_decr(asoc->sent_queue_retran_cnt); +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xB3, + (asoc->sent_queue_retran_cnt & 0x000000ff)); +#endif + } + if (tp1->rec.data.chunk_was_revoked) { + /* deflate the cwnd */ + tp1->whoTo->cwnd -= tp1->book_size; + tp1->rec.data.chunk_was_revoked = 0; + } + if (tp1->sent != SCTP_DATAGRAM_NR_ACKED) { + tp1->sent = SCTP_DATAGRAM_ACKED; + } + } + } else { + break; + } + } + biggest_tsn_newly_acked = biggest_tsn_acked = last_tsn; + /* always set this up to cum-ack */ + asoc->this_sack_highest_gap = last_tsn; + + if ((num_seg > 0) || (num_nr_seg > 0)) { + /* + * thisSackHighestGap will increase while handling NEW + * segments this_sack_highest_newack will increase while + * handling NEWLY ACKED chunks. this_sack_lowest_newack is + * used for CMT DAC algo. saw_newack will also change. + */ + if (sctp_handle_segments(m, &offset_seg, stcb, asoc, last_tsn, &biggest_tsn_acked, + &biggest_tsn_newly_acked, &this_sack_lowest_newack, + num_seg, num_nr_seg, &rto_ok)) { + wake_him++; + } + /* + * validate the biggest_tsn_acked in the gap acks if + * strict adherence is wanted. + */ + if (SCTP_TSN_GE(biggest_tsn_acked, send_s)) { + /* + * peer is either confused or we are under + * attack. We must abort. + */ + SCTP_PRINTF("Hopeless peer! biggest_tsn_acked:%x largest seq:%x\n", + biggest_tsn_acked, send_s); + goto hopeless_peer; + } + } + /*******************************************/ + /* cancel ALL T3-send timer if accum moved */ + /*******************************************/ + if (asoc->sctp_cmt_on_off > 0) { + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if (net->new_pseudo_cumack) + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, + stcb, net, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_31); + } + } else { + if (accum_moved) { + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, + stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_32); + } + } + } + /********************************************/ + /* drop the acked chunks from the sentqueue */ + /********************************************/ + asoc->last_acked_seq = cum_ack; + + TAILQ_FOREACH_SAFE(tp1, &asoc->sent_queue, sctp_next, tp2) { + if (SCTP_TSN_GT(tp1->rec.data.tsn, cum_ack)) { + break; + } + if (tp1->sent != SCTP_DATAGRAM_NR_ACKED) { + if (asoc->strmout[tp1->rec.data.sid].chunks_on_queues > 0) { + asoc->strmout[tp1->rec.data.sid].chunks_on_queues--; +#ifdef INVARIANTS + } else { + panic("No chunks on the queues for sid %u.", tp1->rec.data.sid); +#endif + } + } + if ((asoc->strmout[tp1->rec.data.sid].chunks_on_queues == 0) && + (asoc->strmout[tp1->rec.data.sid].state == SCTP_STREAM_RESET_PENDING) && + TAILQ_EMPTY(&asoc->strmout[tp1->rec.data.sid].outqueue)) { + asoc->trigger_reset = 1; + } + TAILQ_REMOVE(&asoc->sent_queue, tp1, sctp_next); + if (PR_SCTP_ENABLED(tp1->flags)) { + if (asoc->pr_sctp_cnt != 0) + asoc->pr_sctp_cnt--; + } + asoc->sent_queue_cnt--; + if (tp1->data) { + /* sa_ignore NO_NULL_CHK */ + sctp_free_bufspace(stcb, asoc, tp1, 1); + sctp_m_freem(tp1->data); + tp1->data = NULL; + if (asoc->prsctp_supported && PR_SCTP_BUF_ENABLED(tp1->flags)) { + asoc->sent_queue_cnt_removeable--; + } + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_LOGGING_ENABLE) { + sctp_log_sack(asoc->last_acked_seq, + cum_ack, + tp1->rec.data.tsn, + 0, + 0, + SCTP_LOG_FREE_SENT); + } + sctp_free_a_chunk(stcb, tp1, SCTP_SO_NOT_LOCKED); + wake_him++; + } + if (TAILQ_EMPTY(&asoc->sent_queue) && (asoc->total_flight > 0)) { +#ifdef INVARIANTS + panic("Warning flight size is positive and should be 0"); +#else + SCTP_PRINTF("Warning flight size incorrect should be 0 is %d\n", + asoc->total_flight); +#endif + asoc->total_flight = 0; + } + +#if defined(__Userspace__) + if (stcb->sctp_ep->recv_callback) { + if (stcb->sctp_socket) { + uint32_t inqueue_bytes, sb_free_now; + struct sctp_inpcb *inp; + + inp = stcb->sctp_ep; + inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); + sb_free_now = SCTP_SB_LIMIT_SND(stcb->sctp_socket) - (inqueue_bytes + stcb->asoc.sb_send_resv); + + /* check if the amount free in the send socket buffer crossed the threshold */ + if (inp->send_callback && + (((inp->send_sb_threshold > 0) && (sb_free_now >= inp->send_sb_threshold)) || + (inp->send_sb_threshold == 0))) { + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + inp->send_callback(stcb->sctp_socket, sb_free_now, inp->ulp_info); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } + } + } else if ((wake_him) && (stcb->sctp_socket)) { +#else + /* sa_ignore NO_NULL_CHK */ + if ((wake_him) && (stcb->sctp_socket)) { +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; + +#endif + SOCKBUF_LOCK(&stcb->sctp_socket->so_snd); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_WAKE_LOGGING_ENABLE) { + sctp_wakeup_log(stcb, wake_him, SCTP_WAKESND_FROM_SACK); + } +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(stcb->sctp_ep); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { + /* assoc was freed while we were unlocked */ + SCTP_SOCKET_UNLOCK(so, 1); + return; + } +#endif + sctp_sowwakeup_locked(stcb->sctp_ep, stcb->sctp_socket); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_WAKE_LOGGING_ENABLE) { + sctp_wakeup_log(stcb, wake_him, SCTP_NOWAKE_FROM_SACK); + } + } + + if (asoc->fast_retran_loss_recovery && accum_moved) { + if (SCTP_TSN_GE(asoc->last_acked_seq, asoc->fast_recovery_tsn)) { + /* Setup so we will exit RFC2582 fast recovery */ + will_exit_fast_recovery = 1; + } + } + /* + * Check for revoked fragments: + * + * if Previous sack - Had no frags then we can't have any revoked if + * Previous sack - Had frag's then - If we now have frags aka + * num_seg > 0 call sctp_check_for_revoked() to tell if peer revoked + * some of them. else - The peer revoked all ACKED fragments, since + * we had some before and now we have NONE. + */ + + if (num_seg) { + sctp_check_for_revoked(stcb, asoc, cum_ack, biggest_tsn_acked); + asoc->saw_sack_with_frags = 1; + } else if (asoc->saw_sack_with_frags) { + int cnt_revoked = 0; + + /* Peer revoked all dg's marked or acked */ + TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { + if (tp1->sent == SCTP_DATAGRAM_ACKED) { + tp1->sent = SCTP_DATAGRAM_SENT; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_UP_REVOKE, + tp1->whoTo->flight_size, + tp1->book_size, + (uint32_t)(uintptr_t)tp1->whoTo, + tp1->rec.data.tsn); + } + sctp_flight_size_increase(tp1); + sctp_total_flight_increase(stcb, tp1); + tp1->rec.data.chunk_was_revoked = 1; + /* + * To ensure that this increase in + * flightsize, which is artificial, + * does not throttle the sender, we + * also increase the cwnd + * artificially. + */ + tp1->whoTo->cwnd += tp1->book_size; + cnt_revoked++; + } + } + if (cnt_revoked) { + reneged_all = 1; + } + asoc->saw_sack_with_frags = 0; + } + if (num_nr_seg > 0) + asoc->saw_sack_with_nr_frags = 1; + else + asoc->saw_sack_with_nr_frags = 0; + + /* JRS - Use the congestion control given in the CC module */ + if (ecne_seen == 0) { + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if (net->net_ack2 > 0) { + /* + * Karn's rule applies to clearing error count, this + * is optional. + */ + net->error_count = 0; + if ((net->dest_state & SCTP_ADDR_REACHABLE) == 0) { + /* addr came good */ + net->dest_state |= SCTP_ADDR_REACHABLE; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, + 0, (void *)net, SCTP_SO_NOT_LOCKED); + } + + if (net == stcb->asoc.primary_destination) { + if (stcb->asoc.alternate) { + /* release the alternate, primary is good */ + sctp_free_remote_addr(stcb->asoc.alternate); + stcb->asoc.alternate = NULL; + } + } + + if (net->dest_state & SCTP_ADDR_PF) { + net->dest_state &= ~SCTP_ADDR_PF; + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, + stcb->sctp_ep, stcb, net, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_33); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); + asoc->cc_functions.sctp_cwnd_update_exit_pf(stcb, net); + /* Done with this net */ + net->net_ack = 0; + } + /* restore any doubled timers */ + net->RTO = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv; + if (net->RTO < stcb->asoc.minrto) { + net->RTO = stcb->asoc.minrto; + } + if (net->RTO > stcb->asoc.maxrto) { + net->RTO = stcb->asoc.maxrto; + } + } + } + asoc->cc_functions.sctp_cwnd_update_after_sack(stcb, asoc, accum_moved, reneged_all, will_exit_fast_recovery); + } + + if (TAILQ_EMPTY(&asoc->sent_queue)) { + /* nothing left in-flight */ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + /* stop all timers */ + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, + stcb, net, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_34); + net->flight_size = 0; + net->partial_bytes_acked = 0; + } + asoc->total_flight = 0; + asoc->total_flight_count = 0; + } + + /**********************************/ + /* Now what about shutdown issues */ + /**********************************/ + if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue)) { + /* nothing left on sendqueue.. consider done */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { + sctp_log_rwnd_set(SCTP_SET_PEER_RWND_VIA_SACK, + asoc->peers_rwnd, 0, 0, a_rwnd); + } + asoc->peers_rwnd = a_rwnd; + if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { + /* SWS sender side engages */ + asoc->peers_rwnd = 0; + } + /* clean up */ + if ((asoc->stream_queue_cnt == 1) && + ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) && + ((*asoc->ss_functions.sctp_ss_is_user_msgs_incomplete)(stcb, asoc))) { + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_PARTIAL_MSG_LEFT); + } + if (((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) && + (asoc->stream_queue_cnt == 1) && + (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { + struct mbuf *op_err; + + *abort_now = 1; + /* XXX */ + op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_35; + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + return; + } + if ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) && + (asoc->stream_queue_cnt == 0)) { + struct sctp_nets *netp; + + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_SENT); + sctp_stop_timers_for_shutdown(stcb); + if (asoc->alternate) { + netp = asoc->alternate; + } else { + netp = asoc->primary_destination; + } + sctp_send_shutdown(stcb, netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, + stcb->sctp_ep, stcb, netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, + stcb->sctp_ep, stcb, NULL); + return; + } else if ((SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED) && + (asoc->stream_queue_cnt == 0)) { + struct sctp_nets *netp; + + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_ACK_SENT); + sctp_stop_timers_for_shutdown(stcb); + if (asoc->alternate) { + netp = asoc->alternate; + } else { + netp = asoc->primary_destination; + } + sctp_send_shutdown_ack(stcb, netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, + stcb->sctp_ep, stcb, netp); + return; + } + } + /* + * Now here we are going to recycle net_ack for a different use... + * HEADS UP. + */ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + net->net_ack = 0; + } + + /* + * CMT DAC algorithm: If SACK DAC flag was 0, then no extra marking + * to be done. Setting this_sack_lowest_newack to the cum_ack will + * automatically ensure that. + */ + if ((asoc->sctp_cmt_on_off > 0) && + SCTP_BASE_SYSCTL(sctp_cmt_use_dac) && + (cmt_dac_flag == 0)) { + this_sack_lowest_newack = cum_ack; + } + if ((num_seg > 0) || (num_nr_seg > 0)) { + sctp_strike_gap_ack_chunks(stcb, asoc, biggest_tsn_acked, + biggest_tsn_newly_acked, this_sack_lowest_newack, accum_moved); + } + /* JRS - Use the congestion control given in the CC module */ + asoc->cc_functions.sctp_cwnd_update_after_fr(stcb, asoc); + + /* Now are we exiting loss recovery ? */ + if (will_exit_fast_recovery) { + /* Ok, we must exit fast recovery */ + asoc->fast_retran_loss_recovery = 0; + } + if ((asoc->sat_t3_loss_recovery) && + SCTP_TSN_GE(asoc->last_acked_seq, asoc->sat_t3_recovery_tsn)) { + /* end satellite t3 loss recovery */ + asoc->sat_t3_loss_recovery = 0; + } + /* + * CMT Fast recovery + */ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if (net->will_exit_fast_recovery) { + /* Ok, we must exit fast recovery */ + net->fast_retran_loss_recovery = 0; + } + } + + /* Adjust and set the new rwnd value */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { + sctp_log_rwnd_set(SCTP_SET_PEER_RWND_VIA_SACK, + asoc->peers_rwnd, asoc->total_flight, (asoc->total_flight_count * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)), a_rwnd); + } + asoc->peers_rwnd = sctp_sbspace_sub(a_rwnd, + (uint32_t) (asoc->total_flight + (asoc->total_flight_count * SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)))); + if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { + /* SWS sender side engages */ + asoc->peers_rwnd = 0; + } + if (asoc->peers_rwnd > old_rwnd) { + win_probe_recovery = 1; + } + + /* + * Now we must setup so we have a timer up for anyone with + * outstanding data. + */ + done_once = 0; +again: + j = 0; + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if (win_probe_recovery && (net->window_probe)) { + win_probe_recovered = 1; + /*- + * Find first chunk that was used with + * window probe and clear the event. Put + * it back into the send queue as if has + * not been sent. + */ + TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { + if (tp1->window_probe) { + sctp_window_probe_recovery(stcb, asoc, tp1); + break; + } + } + } + if (net->flight_size) { + j++; + if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { + sctp_timer_start(SCTP_TIMER_TYPE_SEND, + stcb->sctp_ep, stcb, net); + } + if (net->window_probe) { + net->window_probe = 0; + } + } else { + if (net->window_probe) { + /* In window probes we must assure a timer is still running there */ + if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { + sctp_timer_start(SCTP_TIMER_TYPE_SEND, + stcb->sctp_ep, stcb, net); + } + } else if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, + stcb, net, + SCTP_FROM_SCTP_INDATA + SCTP_LOC_36); + } + } + } + if ((j == 0) && + (!TAILQ_EMPTY(&asoc->sent_queue)) && + (asoc->sent_queue_retran_cnt == 0) && + (win_probe_recovered == 0) && + (done_once == 0)) { + /* huh, this should not happen unless all packets + * are PR-SCTP and marked to skip of course. + */ + if (sctp_fs_audit(asoc)) { + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + net->flight_size = 0; + } + asoc->total_flight = 0; + asoc->total_flight_count = 0; + asoc->sent_queue_retran_cnt = 0; + TAILQ_FOREACH(tp1, &asoc->sent_queue, sctp_next) { + if (tp1->sent < SCTP_DATAGRAM_RESEND) { + sctp_flight_size_increase(tp1); + sctp_total_flight_increase(stcb, tp1); + } else if (tp1->sent == SCTP_DATAGRAM_RESEND) { + sctp_ucount_incr(asoc->sent_queue_retran_cnt); + } + } + } + done_once = 1; + goto again; + } + /*********************************************/ + /* Here we perform PR-SCTP procedures */ + /* (section 4.2) */ + /*********************************************/ + /* C1. update advancedPeerAckPoint */ + if (SCTP_TSN_GT(cum_ack, asoc->advanced_peer_ack_point)) { + asoc->advanced_peer_ack_point = cum_ack; + } + /* C2. try to further move advancedPeerAckPoint ahead */ + if ((asoc->prsctp_supported) && (asoc->pr_sctp_cnt > 0)) { + struct sctp_tmit_chunk *lchk; + uint32_t old_adv_peer_ack_point; + + old_adv_peer_ack_point = asoc->advanced_peer_ack_point; + lchk = sctp_try_advance_peer_ack_point(stcb, asoc); + /* C3. See if we need to send a Fwd-TSN */ + if (SCTP_TSN_GT(asoc->advanced_peer_ack_point, cum_ack)) { + /* + * ISSUE with ECN, see FWD-TSN processing. + */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_TRY_ADVANCE) { + sctp_misc_ints(SCTP_FWD_TSN_CHECK, + 0xee, cum_ack, asoc->advanced_peer_ack_point, + old_adv_peer_ack_point); + } + if (SCTP_TSN_GT(asoc->advanced_peer_ack_point, old_adv_peer_ack_point)) { + send_forward_tsn(stcb, asoc); + } else if (lchk) { + /* try to FR fwd-tsn's that get lost too */ + if (lchk->rec.data.fwd_tsn_cnt >= 3) { + send_forward_tsn(stcb, asoc); + } + } + } + for (; lchk != NULL; lchk = TAILQ_NEXT(lchk, sctp_next)) { + if (lchk->whoTo != NULL) { + break; + } + } + if (lchk != NULL) { + /* Assure a timer is up */ + sctp_timer_start(SCTP_TIMER_TYPE_SEND, + stcb->sctp_ep, stcb, lchk->whoTo); + } + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SACK_RWND_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_SACK_RWND_UPDATE, + a_rwnd, + stcb->asoc.peers_rwnd, + stcb->asoc.total_flight, + stcb->asoc.total_output_queue_size); + } +} + +void +sctp_update_acked(struct sctp_tcb *stcb, struct sctp_shutdown_chunk *cp, int *abort_flag) +{ + /* Copy cum-ack */ + uint32_t cum_ack, a_rwnd; + + cum_ack = ntohl(cp->cumulative_tsn_ack); + /* Arrange so a_rwnd does NOT change */ + a_rwnd = stcb->asoc.peers_rwnd + stcb->asoc.total_flight; + + /* Now call the express sack handling */ + sctp_express_handle_sack(stcb, cum_ack, a_rwnd, abort_flag, 0); +} + +static void +sctp_kick_prsctp_reorder_queue(struct sctp_tcb *stcb, + struct sctp_stream_in *strmin) +{ + struct sctp_queued_to_read *control, *ncontrol; + struct sctp_association *asoc; + uint32_t mid; + int need_reasm_check = 0; + + asoc = &stcb->asoc; + mid = strmin->last_mid_delivered; + /* + * First deliver anything prior to and including the stream no that + * came in. + */ + TAILQ_FOREACH_SAFE(control, &strmin->inqueue, next_instrm, ncontrol) { + if (SCTP_MID_GE(asoc->idata_supported, mid, control->mid)) { + /* this is deliverable now */ + if (((control->sinfo_flags >> 8) & SCTP_DATA_NOT_FRAG) == SCTP_DATA_NOT_FRAG) { + if (control->on_strm_q) { + if (control->on_strm_q == SCTP_ON_ORDERED) { + TAILQ_REMOVE(&strmin->inqueue, control, next_instrm); + } else if (control->on_strm_q == SCTP_ON_UNORDERED) { + TAILQ_REMOVE(&strmin->uno_inqueue, control, next_instrm); +#ifdef INVARIANTS + } else { + panic("strmin: %p ctl: %p unknown %d", + strmin, control, control->on_strm_q); +#endif + } + control->on_strm_q = 0; + } + /* subtract pending on streams */ + if (asoc->size_on_all_streams >= control->length) { + asoc->size_on_all_streams -= control->length; + } else { +#ifdef INVARIANTS + panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); +#else + asoc->size_on_all_streams = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_all_streams); + /* deliver it to at least the delivery-q */ + if (stcb->sctp_socket) { + sctp_mark_non_revokable(asoc, control->sinfo_tsn); + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, + 1, SCTP_READ_LOCK_HELD, + SCTP_SO_NOT_LOCKED); + } + } else { + /* Its a fragmented message */ + if (control->first_frag_seen) { + /* Make it so this is next to deliver, we restore later */ + strmin->last_mid_delivered = control->mid - 1; + need_reasm_check = 1; + break; + } + } + } else { + /* no more delivery now. */ + break; + } + } + if (need_reasm_check) { + int ret; + ret = sctp_deliver_reasm_check(stcb, &stcb->asoc, strmin, SCTP_READ_LOCK_HELD); + if (SCTP_MID_GT(asoc->idata_supported, mid, strmin->last_mid_delivered)) { + /* Restore the next to deliver unless we are ahead */ + strmin->last_mid_delivered = mid; + } + if (ret == 0) { + /* Left the front Partial one on */ + return; + } + need_reasm_check = 0; + } + /* + * now we must deliver things in queue the normal way if any are + * now ready. + */ + mid = strmin->last_mid_delivered + 1; + TAILQ_FOREACH_SAFE(control, &strmin->inqueue, next_instrm, ncontrol) { + if (SCTP_MID_EQ(asoc->idata_supported, mid, control->mid)) { + if (((control->sinfo_flags >> 8) & SCTP_DATA_NOT_FRAG) == SCTP_DATA_NOT_FRAG) { + /* this is deliverable now */ + if (control->on_strm_q) { + if (control->on_strm_q == SCTP_ON_ORDERED) { + TAILQ_REMOVE(&strmin->inqueue, control, next_instrm); + } else if (control->on_strm_q == SCTP_ON_UNORDERED) { + TAILQ_REMOVE(&strmin->uno_inqueue, control, next_instrm); +#ifdef INVARIANTS + } else { + panic("strmin: %p ctl: %p unknown %d", + strmin, control, control->on_strm_q); +#endif + } + control->on_strm_q = 0; + } + /* subtract pending on streams */ + if (asoc->size_on_all_streams >= control->length) { + asoc->size_on_all_streams -= control->length; + } else { +#ifdef INVARIANTS + panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); +#else + asoc->size_on_all_streams = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_all_streams); + /* deliver it to at least the delivery-q */ + strmin->last_mid_delivered = control->mid; + if (stcb->sctp_socket) { + sctp_mark_non_revokable(asoc, control->sinfo_tsn); + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, + SCTP_READ_LOCK_HELD, SCTP_SO_NOT_LOCKED); + } + mid = strmin->last_mid_delivered + 1; + } else { + /* Its a fragmented message */ + if (control->first_frag_seen) { + /* Make it so this is next to deliver */ + strmin->last_mid_delivered = control->mid - 1; + need_reasm_check = 1; + break; + } + } + } else { + break; + } + } + if (need_reasm_check) { + (void)sctp_deliver_reasm_check(stcb, &stcb->asoc, strmin, SCTP_READ_LOCK_HELD); + } +} + +static void +sctp_flush_reassm_for_str_seq(struct sctp_tcb *stcb, + struct sctp_association *asoc, struct sctp_stream_in *strm, + struct sctp_queued_to_read *control, int ordered, uint32_t cumtsn) +{ + struct sctp_tmit_chunk *chk, *nchk; + + /* + * For now large messages held on the stream reasm that are + * complete will be tossed too. We could in theory do more + * work to spin through and stop after dumping one msg aka + * seeing the start of a new msg at the head, and call the + * delivery function... to see if it can be delivered... But + * for now we just dump everything on the queue. + */ + if (!asoc->idata_supported && !ordered && + control->first_frag_seen && + SCTP_TSN_GT(control->fsn_included, cumtsn)) { + return; + } + TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) { + /* Purge hanging chunks */ + if (!asoc->idata_supported && !ordered) { + if (SCTP_TSN_GT(chk->rec.data.tsn, cumtsn)) { + break; + } + } + TAILQ_REMOVE(&control->reasm, chk, sctp_next); + if (asoc->size_on_reasm_queue >= chk->send_size) { + asoc->size_on_reasm_queue -= chk->send_size; + } else { +#ifdef INVARIANTS + panic("size_on_reasm_queue = %u smaller than chunk length %u", asoc->size_on_reasm_queue, chk->send_size); +#else + asoc->size_on_reasm_queue = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_reasm_queue); + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + } + if (!TAILQ_EMPTY(&control->reasm)) { + /* This has to be old data, unordered */ + if (control->data) { + sctp_m_freem(control->data); + control->data = NULL; + } + sctp_reset_a_control(control, stcb->sctp_ep, cumtsn); + chk = TAILQ_FIRST(&control->reasm); + if (chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) { + TAILQ_REMOVE(&control->reasm, chk, sctp_next); + sctp_add_chk_to_control(control, strm, stcb, asoc, + chk, SCTP_READ_LOCK_HELD); + } + sctp_deliver_reasm_check(stcb, asoc, strm, SCTP_READ_LOCK_HELD); + return; + } + if (control->on_strm_q == SCTP_ON_ORDERED) { + TAILQ_REMOVE(&strm->inqueue, control, next_instrm); + if (asoc->size_on_all_streams >= control->length) { + asoc->size_on_all_streams -= control->length; + } else { +#ifdef INVARIANTS + panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); +#else + asoc->size_on_all_streams = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_all_streams); + control->on_strm_q = 0; + } else if (control->on_strm_q == SCTP_ON_UNORDERED) { + TAILQ_REMOVE(&strm->uno_inqueue, control, next_instrm); + control->on_strm_q = 0; +#ifdef INVARIANTS + } else if (control->on_strm_q) { + panic("strm: %p ctl: %p unknown %d", + strm, control, control->on_strm_q); +#endif + } + control->on_strm_q = 0; + if (control->on_read_q == 0) { + sctp_free_remote_addr(control->whoFrom); + if (control->data) { + sctp_m_freem(control->data); + control->data = NULL; + } + sctp_free_a_readq(stcb, control); + } +} + +void +sctp_handle_forward_tsn(struct sctp_tcb *stcb, + struct sctp_forward_tsn_chunk *fwd, + int *abort_flag, struct mbuf *m , int offset) +{ + /* The pr-sctp fwd tsn */ + /* + * here we will perform all the data receiver side steps for + * processing FwdTSN, as required in by pr-sctp draft: + * + * Assume we get FwdTSN(x): + * + * 1) update local cumTSN to x + * 2) try to further advance cumTSN to x + others we have + * 3) examine and update re-ordering queue on pr-in-streams + * 4) clean up re-assembly queue + * 5) Send a sack to report where we are. + */ + struct sctp_association *asoc; + uint32_t new_cum_tsn, gap; + unsigned int i, fwd_sz, m_size; + uint32_t str_seq; + struct sctp_stream_in *strm; + struct sctp_queued_to_read *control, *ncontrol, *sv; + + asoc = &stcb->asoc; + if ((fwd_sz = ntohs(fwd->ch.chunk_length)) < sizeof(struct sctp_forward_tsn_chunk)) { + SCTPDBG(SCTP_DEBUG_INDATA1, + "Bad size too small/big fwd-tsn\n"); + return; + } + m_size = (stcb->asoc.mapping_array_size << 3); + /*************************************************************/ + /* 1. Here we update local cumTSN and shift the bitmap array */ + /*************************************************************/ + new_cum_tsn = ntohl(fwd->new_cumulative_tsn); + + if (SCTP_TSN_GE(asoc->cumulative_tsn, new_cum_tsn)) { + /* Already got there ... */ + return; + } + /* + * now we know the new TSN is more advanced, let's find the actual + * gap + */ + SCTP_CALC_TSN_TO_GAP(gap, new_cum_tsn, asoc->mapping_array_base_tsn); + asoc->cumulative_tsn = new_cum_tsn; + if (gap >= m_size) { + if ((long)gap > sctp_sbspace(&stcb->asoc, &stcb->sctp_socket->so_rcv)) { + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + /* + * out of range (of single byte chunks in the rwnd I + * give out). This must be an attacker. + */ + *abort_flag = 1; + SCTP_SNPRINTF(msg, sizeof(msg), + "New cum ack %8.8x too high, highest TSN %8.8x", + new_cum_tsn, asoc->highest_tsn_inside_map); + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_INDATA + SCTP_LOC_37; + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + return; + } + SCTP_STAT_INCR(sctps_fwdtsn_map_over); + + memset(stcb->asoc.mapping_array, 0, stcb->asoc.mapping_array_size); + asoc->mapping_array_base_tsn = new_cum_tsn + 1; + asoc->highest_tsn_inside_map = new_cum_tsn; + + memset(stcb->asoc.nr_mapping_array, 0, stcb->asoc.mapping_array_size); + asoc->highest_tsn_inside_nr_map = new_cum_tsn; + + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { + sctp_log_map(0, 3, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); + } + } else { + SCTP_TCB_LOCK_ASSERT(stcb); + for (i = 0; i <= gap; i++) { + if (!SCTP_IS_TSN_PRESENT(asoc->mapping_array, i) && + !SCTP_IS_TSN_PRESENT(asoc->nr_mapping_array, i)) { + SCTP_SET_TSN_PRESENT(asoc->nr_mapping_array, i); + if (SCTP_TSN_GT(asoc->mapping_array_base_tsn + i, asoc->highest_tsn_inside_nr_map)) { + asoc->highest_tsn_inside_nr_map = asoc->mapping_array_base_tsn + i; + } + } + } + } + /*************************************************************/ + /* 2. Clear up re-assembly queue */ + /*************************************************************/ + + /* This is now done as part of clearing up the stream/seq */ + if (asoc->idata_supported == 0) { + uint16_t sid; + + /* Flush all the un-ordered data based on cum-tsn */ + SCTP_INP_READ_LOCK(stcb->sctp_ep); + for (sid = 0; sid < asoc->streamincnt; sid++) { + strm = &asoc->strmin[sid]; + if (!TAILQ_EMPTY(&strm->uno_inqueue)) { + sctp_flush_reassm_for_str_seq(stcb, asoc, strm, TAILQ_FIRST(&strm->uno_inqueue), 0, new_cum_tsn); + } + } + SCTP_INP_READ_UNLOCK(stcb->sctp_ep); + } + /*******************************************************/ + /* 3. Update the PR-stream re-ordering queues and fix */ + /* delivery issues as needed. */ + /*******************************************************/ + fwd_sz -= sizeof(*fwd); + if (m && fwd_sz) { + /* New method. */ + unsigned int num_str; + uint32_t mid; + uint16_t sid; + uint16_t ordered, flags; + struct sctp_strseq *stseq, strseqbuf; + struct sctp_strseq_mid *stseq_m, strseqbuf_m; + offset += sizeof(*fwd); + + SCTP_INP_READ_LOCK(stcb->sctp_ep); + if (asoc->idata_supported) { + num_str = fwd_sz / sizeof(struct sctp_strseq_mid); + } else { + num_str = fwd_sz / sizeof(struct sctp_strseq); + } + for (i = 0; i < num_str; i++) { + if (asoc->idata_supported) { + stseq_m = (struct sctp_strseq_mid *)sctp_m_getptr(m, offset, + sizeof(struct sctp_strseq_mid), + (uint8_t *)&strseqbuf_m); + offset += sizeof(struct sctp_strseq_mid); + if (stseq_m == NULL) { + break; + } + sid = ntohs(stseq_m->sid); + mid = ntohl(stseq_m->mid); + flags = ntohs(stseq_m->flags); + if (flags & PR_SCTP_UNORDERED_FLAG) { + ordered = 0; + } else { + ordered = 1; + } + } else { + stseq = (struct sctp_strseq *)sctp_m_getptr(m, offset, + sizeof(struct sctp_strseq), + (uint8_t *)&strseqbuf); + offset += sizeof(struct sctp_strseq); + if (stseq == NULL) { + break; + } + sid = ntohs(stseq->sid); + mid = (uint32_t)ntohs(stseq->ssn); + ordered = 1; + } + /* Convert */ + + /* now process */ + + /* + * Ok we now look for the stream/seq on the read queue + * where its not all delivered. If we find it we transmute the + * read entry into a PDI_ABORTED. + */ + if (sid >= asoc->streamincnt) { + /* screwed up streams, stop! */ + break; + } + if ((asoc->str_of_pdapi == sid) && + (asoc->ssn_of_pdapi == mid)) { + /* If this is the one we were partially delivering + * now then we no longer are. Note this will change + * with the reassembly re-write. + */ + asoc->fragmented_delivery_inprogress = 0; + } + strm = &asoc->strmin[sid]; + if (ordered) { + TAILQ_FOREACH_SAFE(control, &strm->inqueue, next_instrm, ncontrol) { + if (SCTP_MID_GE(asoc->idata_supported, mid, control->mid)) { + sctp_flush_reassm_for_str_seq(stcb, asoc, strm, control, ordered, new_cum_tsn); + } + } + } else { + if (asoc->idata_supported) { + TAILQ_FOREACH_SAFE(control, &strm->uno_inqueue, next_instrm, ncontrol) { + if (SCTP_MID_GE(asoc->idata_supported, mid, control->mid)) { + sctp_flush_reassm_for_str_seq(stcb, asoc, strm, control, ordered, new_cum_tsn); + } + } + } else { + if (!TAILQ_EMPTY(&strm->uno_inqueue)) { + sctp_flush_reassm_for_str_seq(stcb, asoc, strm, TAILQ_FIRST(&strm->uno_inqueue), ordered, new_cum_tsn); + } + } + } + TAILQ_FOREACH(control, &stcb->sctp_ep->read_queue, next) { + if ((control->sinfo_stream == sid) && + (SCTP_MID_EQ(asoc->idata_supported, control->mid, mid))) { + str_seq = (sid << 16) | (0x0000ffff & mid); + control->pdapi_aborted = 1; + sv = stcb->asoc.control_pdapi; + control->end_added = 1; + if (control->on_strm_q == SCTP_ON_ORDERED) { + TAILQ_REMOVE(&strm->inqueue, control, next_instrm); + if (asoc->size_on_all_streams >= control->length) { + asoc->size_on_all_streams -= control->length; + } else { +#ifdef INVARIANTS + panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); +#else + asoc->size_on_all_streams = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_all_streams); + } else if (control->on_strm_q == SCTP_ON_UNORDERED) { + TAILQ_REMOVE(&strm->uno_inqueue, control, next_instrm); +#ifdef INVARIANTS + } else if (control->on_strm_q) { + panic("strm: %p ctl: %p unknown %d", + strm, control, control->on_strm_q); +#endif + } + control->on_strm_q = 0; + stcb->asoc.control_pdapi = control; + sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION, + stcb, + SCTP_PARTIAL_DELIVERY_ABORTED, + (void *)&str_seq, + SCTP_SO_NOT_LOCKED); + stcb->asoc.control_pdapi = sv; + break; + } else if ((control->sinfo_stream == sid) && + SCTP_MID_GT(asoc->idata_supported, control->mid, mid)) { + /* We are past our victim SSN */ + break; + } + } + if (SCTP_MID_GT(asoc->idata_supported, mid, strm->last_mid_delivered)) { + /* Update the sequence number */ + strm->last_mid_delivered = mid; + } + /* now kick the stream the new way */ + /*sa_ignore NO_NULL_CHK*/ + sctp_kick_prsctp_reorder_queue(stcb, strm); + } + SCTP_INP_READ_UNLOCK(stcb->sctp_ep); + } + /* + * Now slide thing forward. + */ + sctp_slide_mapping_arrays(stcb); +} diff --git a/netwerk/sctp/src/netinet/sctp_indata.h b/netwerk/sctp/src/netinet/sctp_indata.h new file mode 100644 index 0000000000..c9b71054d3 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_indata.h @@ -0,0 +1,121 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_INDATA_H_ +#define _NETINET_SCTP_INDATA_H_ + +#if defined(_KERNEL) || defined(__Userspace__) + +struct sctp_queued_to_read * +sctp_build_readq_entry(struct sctp_tcb *stcb, + struct sctp_nets *net, + uint32_t tsn, uint32_t ppid, + uint32_t context, uint16_t sid, + uint32_t mid, uint8_t flags, + struct mbuf *dm); + +#define sctp_build_readq_entry_mac(_ctl, in_it, context, net, tsn, ppid, sid, flags, dm, tfsn, mid) do { \ + if (_ctl) { \ + atomic_add_int(&((net)->ref_count), 1); \ + memset(_ctl, 0, sizeof(struct sctp_queued_to_read)); \ + (_ctl)->sinfo_stream = sid; \ + TAILQ_INIT(&_ctl->reasm); \ + (_ctl)->top_fsn = tfsn; \ + (_ctl)->mid = mid; \ + (_ctl)->sinfo_flags = (flags << 8); \ + (_ctl)->sinfo_ppid = ppid; \ + (_ctl)->sinfo_context = context; \ + (_ctl)->fsn_included = 0xffffffff; \ + (_ctl)->sinfo_tsn = tsn; \ + (_ctl)->sinfo_cumtsn = tsn; \ + (_ctl)->sinfo_assoc_id = sctp_get_associd((in_it)); \ + (_ctl)->whoFrom = net; \ + (_ctl)->data = dm; \ + (_ctl)->stcb = (in_it); \ + (_ctl)->port_from = (in_it)->rport; \ + if ((in_it)->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { \ + (_ctl)->do_not_ref_stcb = 1; \ + }\ + } \ +} while (0) + +struct mbuf * +sctp_build_ctl_nchunk(struct sctp_inpcb *inp, + struct sctp_sndrcvinfo *sinfo); + +void sctp_set_rwnd(struct sctp_tcb *, struct sctp_association *); + +uint32_t +sctp_calc_rwnd(struct sctp_tcb *stcb, struct sctp_association *asoc); + +void +sctp_express_handle_sack(struct sctp_tcb *stcb, uint32_t cumack, + uint32_t rwnd, int *abort_now, int ecne_seen); + +void +sctp_handle_sack(struct mbuf *m, int offset_seg, int offset_dup, + struct sctp_tcb *stcb, + uint16_t num_seg, uint16_t num_nr_seg, uint16_t num_dup, + int *abort_now, uint8_t flags, + uint32_t cum_ack, uint32_t rwnd, int ecne_seen); + +/* draft-ietf-tsvwg-usctp */ +void +sctp_handle_forward_tsn(struct sctp_tcb *, + struct sctp_forward_tsn_chunk *, int *, struct mbuf *, int); + +struct sctp_tmit_chunk * +sctp_try_advance_peer_ack_point(struct sctp_tcb *, struct sctp_association *); + +void sctp_service_queues(struct sctp_tcb *, struct sctp_association *); + +void +sctp_update_acked(struct sctp_tcb *, struct sctp_shutdown_chunk *, int *); + +int +sctp_process_data(struct mbuf **, int, int *, int, + struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *, uint32_t *); + +void sctp_slide_mapping_arrays(struct sctp_tcb *stcb); + +void sctp_sack_check(struct sctp_tcb *, int); + +#endif +#endif diff --git a/netwerk/sctp/src/netinet/sctp_input.c b/netwerk/sctp/src/netinet/sctp_input.c new file mode 100644 index 0000000000..651d078ab7 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_input.c @@ -0,0 +1,6465 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#include <netinet/sctp_var.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctp_header.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_input.h> +#include <netinet/sctp_auth.h> +#include <netinet/sctp_indata.h> +#include <netinet/sctp_asconf.h> +#include <netinet/sctp_bsd_addr.h> +#include <netinet/sctp_timer.h> +#include <netinet/sctp_crc32.h> +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <netinet/sctp_kdtrace.h> +#endif +#if defined(INET) || defined(INET6) +#if !defined(_WIN32) +#include <netinet/udp.h> +#endif +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/smp.h> +#endif + +static void +sctp_stop_all_cookie_timers(struct sctp_tcb *stcb) +{ + struct sctp_nets *net; + + /* This now not only stops all cookie timers + * it also stops any INIT timers as well. This + * will make sure that the timers are stopped in + * all collision cases. + */ + SCTP_TCB_LOCK_ASSERT(stcb); + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if (net->rxt_timer.type == SCTP_TIMER_TYPE_COOKIE) { + sctp_timer_stop(SCTP_TIMER_TYPE_COOKIE, + stcb->sctp_ep, + stcb, + net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_1); + } else if (net->rxt_timer.type == SCTP_TIMER_TYPE_INIT) { + sctp_timer_stop(SCTP_TIMER_TYPE_INIT, + stcb->sctp_ep, + stcb, + net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_2); + } + } +} + +/* INIT handler */ +static void +sctp_handle_init(struct mbuf *m, int iphlen, int offset, + struct sockaddr *src, struct sockaddr *dst, struct sctphdr *sh, + struct sctp_init_chunk *cp, struct sctp_inpcb *inp, + struct sctp_tcb *stcb, struct sctp_nets *net, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, +#endif + uint32_t vrf_id, uint16_t port) +{ + struct sctp_init *init; + struct mbuf *op_err; + + SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_init: handling INIT tcb:%p\n", + (void *)stcb); + if (stcb == NULL) { + SCTP_INP_RLOCK(inp); + } + /* Validate parameters */ + init = &cp->init; + if (ntohl(init->initiate_tag) == 0) { + goto outnow; + } + if ((ntohl(init->a_rwnd) < SCTP_MIN_RWND) || + (ntohs(init->num_inbound_streams) == 0) || + (ntohs(init->num_outbound_streams) == 0)) { + /* protocol error... send abort */ + op_err = sctp_generate_cause(SCTP_CAUSE_INVALID_PARAM, ""); + sctp_send_abort(m, iphlen, src, dst, sh, init->initiate_tag, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, port); + goto outnow; + } + if (sctp_validate_init_auth_params(m, offset + sizeof(*cp), + offset + ntohs(cp->ch.chunk_length))) { + /* auth parameter(s) error... send abort */ + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + "Problem with AUTH parameters"); + sctp_send_abort(m, iphlen, src, dst, sh, init->initiate_tag, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, port); + goto outnow; + } + /* We are only accepting if we have a listening socket.*/ + if ((stcb == NULL) && + ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (!SCTP_IS_LISTENING(inp)))) { + /* + * FIX ME ?? What about TCP model and we have a + * match/restart case? Actually no fix is needed. + * the lookup will always find the existing assoc so stcb + * would not be NULL. It may be questionable to do this + * since we COULD just send back the INIT-ACK and hope that + * the app did accept()'s by the time the COOKIE was sent. But + * there is a price to pay for COOKIE generation and I don't + * want to pay it on the chance that the app will actually do + * some accepts(). The App just looses and should NOT be in + * this state :-) + */ + if (SCTP_BASE_SYSCTL(sctp_blackhole) == 0) { + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + "No listener"); + sctp_send_abort(m, iphlen, src, dst, sh, 0, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, port); + } + goto outnow; + } + if ((stcb != NULL) && + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_ACK_SENT)) { + SCTPDBG(SCTP_DEBUG_INPUT3, "sctp_handle_init: sending SHUTDOWN-ACK\n"); + sctp_send_shutdown_ack(stcb, NULL); + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CONTROL_PROC, SCTP_SO_NOT_LOCKED); + } else { + SCTPDBG(SCTP_DEBUG_INPUT3, "sctp_handle_init: sending INIT-ACK\n"); + sctp_send_initiate_ack(inp, stcb, net, m, iphlen, offset, + src, dst, sh, cp, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, port); + } + outnow: + if (stcb == NULL) { + SCTP_INP_RUNLOCK(inp); + } +} + +/* + * process peer "INIT/INIT-ACK" chunk returns value < 0 on error + */ + +int +sctp_is_there_unsent_data(struct sctp_tcb *stcb, int so_locked) +{ + int unsent_data; + unsigned int i; + struct sctp_stream_queue_pending *sp; + struct sctp_association *asoc; + + SCTP_TCB_LOCK_ASSERT(stcb); + + /* This function returns if any stream has true unsent data on it. + * Note that as it looks through it will clean up any places that + * have old data that has been sent but left at top of stream queue. + */ + asoc = &stcb->asoc; + unsent_data = 0; + if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, asoc)) { + /* Check to see if some data queued */ + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + /*sa_ignore FREED_MEMORY*/ + sp = TAILQ_FIRST(&stcb->asoc.strmout[i].outqueue); + if (sp == NULL) { + continue; + } + if ((sp->msg_is_complete) && + (sp->length == 0) && + (sp->sender_all_done)) { + /* We are doing differed cleanup. Last + * time through when we took all the data + * the sender_all_done was not set. + */ + if (sp->put_last_out == 0) { + SCTP_PRINTF("Gak, put out entire msg with NO end!-1\n"); + SCTP_PRINTF("sender_done:%d len:%d msg_comp:%d put_last_out:%d\n", + sp->sender_all_done, + sp->length, + sp->msg_is_complete, + sp->put_last_out); + } + atomic_subtract_int(&stcb->asoc.stream_queue_cnt, 1); + TAILQ_REMOVE(&stcb->asoc.strmout[i].outqueue, sp, next); + stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, &asoc->strmout[i], sp); + if (sp->net) { + sctp_free_remote_addr(sp->net); + sp->net = NULL; + } + if (sp->data) { + sctp_m_freem(sp->data); + sp->data = NULL; + } + sctp_free_a_strmoq(stcb, sp, so_locked); + if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) { + unsent_data++; + } + } else { + unsent_data++; + } + if (unsent_data > 0) { + break; + } + } + } + return (unsent_data); +} + +static int +sctp_process_init(struct sctp_init_chunk *cp, struct sctp_tcb *stcb) +{ + struct sctp_init *init; + struct sctp_association *asoc; + struct sctp_nets *lnet; + unsigned int i; + + SCTP_TCB_LOCK_ASSERT(stcb); + + init = &cp->init; + asoc = &stcb->asoc; + /* save off parameters */ + asoc->peer_vtag = ntohl(init->initiate_tag); + asoc->peers_rwnd = ntohl(init->a_rwnd); + /* init tsn's */ + asoc->highest_tsn_inside_map = asoc->asconf_seq_in = ntohl(init->initial_tsn) - 1; + + if (!TAILQ_EMPTY(&asoc->nets)) { + /* update any ssthresh's that may have a default */ + TAILQ_FOREACH(lnet, &asoc->nets, sctp_next) { + lnet->ssthresh = asoc->peers_rwnd; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_CWND_MONITOR_ENABLE|SCTP_CWND_LOGGING_ENABLE)) { + sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_INITIALIZATION); + } + } + } + if (asoc->pre_open_streams > ntohs(init->num_inbound_streams)) { + unsigned int newcnt; + struct sctp_stream_out *outs; + struct sctp_stream_queue_pending *sp, *nsp; + struct sctp_tmit_chunk *chk, *nchk; + + /* abandon the upper streams */ + newcnt = ntohs(init->num_inbound_streams); + TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) { + if (chk->rec.data.sid >= newcnt) { + TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next); + asoc->send_queue_cnt--; + if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) { + asoc->strmout[chk->rec.data.sid].chunks_on_queues--; +#ifdef INVARIANTS + } else { + panic("No chunks on the queues for sid %u.", chk->rec.data.sid); +#endif + } + if (chk->data != NULL) { + sctp_free_bufspace(stcb, asoc, chk, 1); + sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb, + 0, chk, SCTP_SO_NOT_LOCKED); + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + } + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + /*sa_ignore FREED_MEMORY*/ + } + } + if (asoc->strmout) { + for (i = newcnt; i < asoc->pre_open_streams; i++) { + outs = &asoc->strmout[i]; + TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) { + atomic_subtract_int(&stcb->asoc.stream_queue_cnt, 1); + TAILQ_REMOVE(&outs->outqueue, sp, next); + stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, outs, sp); + sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, + stcb, 0, sp, SCTP_SO_NOT_LOCKED); + if (sp->data) { + sctp_m_freem(sp->data); + sp->data = NULL; + } + if (sp->net) { + sctp_free_remote_addr(sp->net); + sp->net = NULL; + } + /* Free the chunk */ + sctp_free_a_strmoq(stcb, sp, SCTP_SO_NOT_LOCKED); + /*sa_ignore FREED_MEMORY*/ + } + outs->state = SCTP_STREAM_CLOSED; + } + } + /* cut back the count */ + asoc->pre_open_streams = newcnt; + } + asoc->streamoutcnt = asoc->pre_open_streams; + if (asoc->strmout) { + for (i = 0; i < asoc->streamoutcnt; i++) { + asoc->strmout[i].state = SCTP_STREAM_OPEN; + } + } + /* EY - nr_sack: initialize highest tsn in nr_mapping_array */ + asoc->highest_tsn_inside_nr_map = asoc->highest_tsn_inside_map; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { + sctp_log_map(0, 5, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); + } + /* This is the next one we expect */ + asoc->str_reset_seq_in = asoc->asconf_seq_in + 1; + + asoc->mapping_array_base_tsn = ntohl(init->initial_tsn); + asoc->tsn_last_delivered = asoc->cumulative_tsn = asoc->asconf_seq_in; + + asoc->advanced_peer_ack_point = asoc->last_acked_seq; + /* open the requested streams */ + + if (asoc->strmin != NULL) { + /* Free the old ones */ + for (i = 0; i < asoc->streamincnt; i++) { + sctp_clean_up_stream(stcb, &asoc->strmin[i].inqueue); + sctp_clean_up_stream(stcb, &asoc->strmin[i].uno_inqueue); + } + SCTP_FREE(asoc->strmin, SCTP_M_STRMI); + } + if (asoc->max_inbound_streams > ntohs(init->num_outbound_streams)) { + asoc->streamincnt = ntohs(init->num_outbound_streams); + } else { + asoc->streamincnt = asoc->max_inbound_streams; + } + SCTP_MALLOC(asoc->strmin, struct sctp_stream_in *, asoc->streamincnt * + sizeof(struct sctp_stream_in), SCTP_M_STRMI); + if (asoc->strmin == NULL) { + /* we didn't get memory for the streams! */ + SCTPDBG(SCTP_DEBUG_INPUT2, "process_init: couldn't get memory for the streams!\n"); + return (-1); + } + for (i = 0; i < asoc->streamincnt; i++) { + asoc->strmin[i].sid = i; + asoc->strmin[i].last_mid_delivered = 0xffffffff; + TAILQ_INIT(&asoc->strmin[i].inqueue); + TAILQ_INIT(&asoc->strmin[i].uno_inqueue); + asoc->strmin[i].pd_api_started = 0; + asoc->strmin[i].delivery_started = 0; + } + /* + * load_address_from_init will put the addresses into the + * association when the COOKIE is processed or the INIT-ACK is + * processed. Both types of COOKIE's existing and new call this + * routine. It will remove addresses that are no longer in the + * association (for the restarting case where addresses are + * removed). Up front when the INIT arrives we will discard it if it + * is a restart and new addresses have been added. + */ + /* sa_ignore MEMLEAK */ + return (0); +} + +/* + * INIT-ACK message processing/consumption returns value < 0 on error + */ +static int +sctp_process_init_ack(struct mbuf *m, int iphlen, int offset, + struct sockaddr *src, struct sockaddr *dst, struct sctphdr *sh, + struct sctp_init_ack_chunk *cp, struct sctp_tcb *stcb, + struct sctp_nets *net, int *abort_no_unlock, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, +#endif + uint32_t vrf_id) +{ + struct sctp_association *asoc; + struct mbuf *op_err; + int retval, abort_flag, cookie_found; + int initack_limit; + int nat_friendly = 0; + + /* First verify that we have no illegal param's */ + abort_flag = 0; + cookie_found = 0; + + op_err = sctp_arethere_unrecognized_parameters(m, + (offset + sizeof(struct sctp_init_chunk)), + &abort_flag, (struct sctp_chunkhdr *)cp, + &nat_friendly, &cookie_found); + if (abort_flag) { + /* Send an abort and notify peer */ + sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, net->port); + *abort_no_unlock = 1; + return (-1); + } + if (!cookie_found) { + uint16_t len; + + /* Only report the missing cookie parameter */ + if (op_err != NULL) { + sctp_m_freem(op_err); + } + len = (uint16_t)(sizeof(struct sctp_error_missing_param) + sizeof(uint16_t)); + /* We abort with an error of missing mandatory param */ + op_err = sctp_get_mbuf_for_msg(len, 0, M_NOWAIT, 1, MT_DATA); + if (op_err != NULL) { + struct sctp_error_missing_param *cause; + + SCTP_BUF_LEN(op_err) = len; + cause = mtod(op_err, struct sctp_error_missing_param *); + /* Subtract the reserved param */ + cause->cause.code = htons(SCTP_CAUSE_MISSING_PARAM); + cause->cause.length = htons(len); + cause->num_missing_params = htonl(1); + cause->type[0] = htons(SCTP_STATE_COOKIE); + } + sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, net->port); + *abort_no_unlock = 1; + return (-3); + } + asoc = &stcb->asoc; + asoc->peer_supports_nat = (uint8_t)nat_friendly; + /* process the peer's parameters in the INIT-ACK */ + if (sctp_process_init((struct sctp_init_chunk *)cp, stcb) < 0) { + if (op_err != NULL) { + sctp_m_freem(op_err); + } + op_err = sctp_generate_cause(SCTP_CAUSE_OUT_OF_RESC, ""); + SCTPDBG(SCTP_DEBUG_INPUT1, "sctp_process_init() failed\n"); + sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, net->port); + *abort_no_unlock = 1; + return (-1); + } + initack_limit = offset + ntohs(cp->ch.chunk_length); + /* load all addresses */ + if ((retval = sctp_load_addresses_from_init(stcb, m, + offset + sizeof(struct sctp_init_chunk), + initack_limit, src, dst, NULL, stcb->asoc.port)) < 0) { + if (op_err != NULL) { + sctp_m_freem(op_err); + } + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + "Problem with address parameters"); + SCTPDBG(SCTP_DEBUG_INPUT1, + "Load addresses from INIT causes an abort %d\n", + retval); + sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, net->port); + *abort_no_unlock = 1; + return (-1); + } + /* if the peer doesn't support asconf, flush the asconf queue */ + if (asoc->asconf_supported == 0) { + struct sctp_asconf_addr *param, *nparam; + + TAILQ_FOREACH_SAFE(param, &asoc->asconf_queue, next, nparam) { + TAILQ_REMOVE(&asoc->asconf_queue, param, next); + SCTP_FREE(param, SCTP_M_ASC_ADDR); + } + } + + stcb->asoc.peer_hmac_id = sctp_negotiate_hmacid(stcb->asoc.peer_hmacs, + stcb->asoc.local_hmacs); + if (op_err) { + sctp_queue_op_err(stcb, op_err); + /* queuing will steal away the mbuf chain to the out queue */ + op_err = NULL; + } + /* extract the cookie and queue it to "echo" it back... */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { + sctp_misc_ints(SCTP_THRESHOLD_CLEAR, + stcb->asoc.overall_error_count, + 0, + SCTP_FROM_SCTP_INPUT, + __LINE__); + } + + /* + * Cancel the INIT timer, We do this first before queueing the + * cookie. We always cancel at the primary to assume that we are + * canceling the timer started by the INIT which always goes to the + * primary. + */ + sctp_timer_stop(SCTP_TIMER_TYPE_INIT, stcb->sctp_ep, stcb, + asoc->primary_destination, SCTP_FROM_SCTP_INPUT + SCTP_LOC_3); + + /* calculate the RTO */ + if (asoc->overall_error_count == 0) { + sctp_calculate_rto(stcb, asoc, net, &asoc->time_entered, + SCTP_RTT_FROM_NON_DATA); + } + stcb->asoc.overall_error_count = 0; + net->error_count = 0; +#if defined(__Userspace__) + if (stcb->sctp_ep->recv_callback) { + if (stcb->sctp_socket) { + uint32_t inqueue_bytes, sb_free_now; + struct sctp_inpcb *inp; + + inp = stcb->sctp_ep; + inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); + sb_free_now = SCTP_SB_LIMIT_SND(stcb->sctp_socket) - (inqueue_bytes + stcb->asoc.sb_send_resv); + + /* check if the amount free in the send socket buffer crossed the threshold */ + if (inp->send_callback && + (((inp->send_sb_threshold > 0) && + (sb_free_now >= inp->send_sb_threshold) && + (stcb->asoc.chunks_on_out_queue <= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue))) || + (inp->send_sb_threshold == 0))) { + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + inp->send_callback(stcb->sctp_socket, sb_free_now, inp->ulp_info); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } + } + } +#endif + retval = sctp_send_cookie_echo(m, offset, initack_limit, stcb, net); + return (retval); +} + +static void +sctp_handle_heartbeat_ack(struct sctp_heartbeat_chunk *cp, + struct sctp_tcb *stcb, struct sctp_nets *net) +{ + union sctp_sockstore store; + struct sctp_nets *r_net, *f_net; + struct timeval tv; + int req_prim = 0; + uint16_t old_error_counter; + + if (ntohs(cp->ch.chunk_length) != sizeof(struct sctp_heartbeat_chunk)) { + /* Invalid length */ + return; + } + + memset(&store, 0, sizeof(store)); + switch (cp->heartbeat.hb_info.addr_family) { +#ifdef INET + case AF_INET: + if (cp->heartbeat.hb_info.addr_len == sizeof(struct sockaddr_in)) { + store.sin.sin_family = cp->heartbeat.hb_info.addr_family; +#ifdef HAVE_SIN_LEN + store.sin.sin_len = cp->heartbeat.hb_info.addr_len; +#endif + store.sin.sin_port = stcb->rport; + memcpy(&store.sin.sin_addr, cp->heartbeat.hb_info.address, + sizeof(store.sin.sin_addr)); + } else { + return; + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (cp->heartbeat.hb_info.addr_len == sizeof(struct sockaddr_in6)) { + store.sin6.sin6_family = cp->heartbeat.hb_info.addr_family; +#ifdef HAVE_SIN6_LEN + store.sin6.sin6_len = cp->heartbeat.hb_info.addr_len; +#endif + store.sin6.sin6_port = stcb->rport; + memcpy(&store.sin6.sin6_addr, cp->heartbeat.hb_info.address, sizeof(struct in6_addr)); + } else { + return; + } + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + if (cp->heartbeat.hb_info.addr_len == sizeof(struct sockaddr_conn)) { + store.sconn.sconn_family = cp->heartbeat.hb_info.addr_family; +#ifdef HAVE_SCONN_LEN + store.sconn.sconn_len = cp->heartbeat.hb_info.addr_len; +#endif + store.sconn.sconn_port = stcb->rport; + memcpy(&store.sconn.sconn_addr, cp->heartbeat.hb_info.address, sizeof(void *)); + } else { + return; + } + break; +#endif + default: + return; + } + r_net = sctp_findnet(stcb, &store.sa); + if (r_net == NULL) { + SCTPDBG(SCTP_DEBUG_INPUT1, "Huh? I can't find the address I sent it to, discard\n"); + return; + } + if ((r_net && (r_net->dest_state & SCTP_ADDR_UNCONFIRMED)) && + (r_net->heartbeat_random1 == cp->heartbeat.hb_info.random_value1) && + (r_net->heartbeat_random2 == cp->heartbeat.hb_info.random_value2)) { + /* + * If the its a HB and it's random value is correct when can + * confirm the destination. + */ + r_net->dest_state &= ~SCTP_ADDR_UNCONFIRMED; + if (r_net->dest_state & SCTP_ADDR_REQ_PRIMARY) { + stcb->asoc.primary_destination = r_net; + r_net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY; + f_net = TAILQ_FIRST(&stcb->asoc.nets); + if (f_net != r_net) { + /* first one on the list is NOT the primary + * sctp_cmpaddr() is much more efficient if + * the primary is the first on the list, make it + * so. + */ + TAILQ_REMOVE(&stcb->asoc.nets, r_net, sctp_next); + TAILQ_INSERT_HEAD(&stcb->asoc.nets, r_net, sctp_next); + } + req_prim = 1; + } + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_CONFIRMED, + stcb, 0, (void *)r_net, SCTP_SO_NOT_LOCKED); + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, + r_net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_4); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, r_net); + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { + sctp_misc_ints(SCTP_THRESHOLD_CLEAR, + stcb->asoc.overall_error_count, + 0, + SCTP_FROM_SCTP_INPUT, + __LINE__); + } + stcb->asoc.overall_error_count = 0; + old_error_counter = r_net->error_count; + r_net->error_count = 0; + r_net->hb_responded = 1; + tv.tv_sec = cp->heartbeat.hb_info.time_value_1; + tv.tv_usec = cp->heartbeat.hb_info.time_value_2; + /* Now lets do a RTO with this */ + sctp_calculate_rto(stcb, &stcb->asoc, r_net, &tv, + SCTP_RTT_FROM_NON_DATA); + if ((r_net->dest_state & SCTP_ADDR_REACHABLE) == 0) { + r_net->dest_state |= SCTP_ADDR_REACHABLE; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, + 0, (void *)r_net, SCTP_SO_NOT_LOCKED); + } + if (r_net->dest_state & SCTP_ADDR_PF) { + r_net->dest_state &= ~SCTP_ADDR_PF; + stcb->asoc.cc_functions.sctp_cwnd_update_exit_pf(stcb, net); + } + if (old_error_counter > 0) { + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, + stcb, r_net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_5); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, r_net); + } + if (r_net == stcb->asoc.primary_destination) { + if (stcb->asoc.alternate) { + /* release the alternate, primary is good */ + sctp_free_remote_addr(stcb->asoc.alternate); + stcb->asoc.alternate = NULL; + } + } + /* Mobility adaptation */ + if (req_prim) { + if ((sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_BASE) || + sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_FASTHANDOFF)) && + sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_PRIM_DELETED)) { + sctp_timer_stop(SCTP_TIMER_TYPE_PRIM_DELETED, + stcb->sctp_ep, stcb, NULL, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_6); + if (sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_FASTHANDOFF)) { + sctp_assoc_immediate_retrans(stcb, + stcb->asoc.primary_destination); + } + if (sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_BASE)) { + sctp_move_chunks_from_net(stcb, + stcb->asoc.deleted_primary); + } + sctp_delete_prim_timer(stcb->sctp_ep, stcb); + } + } +} + +static int +sctp_handle_nat_colliding_state(struct sctp_tcb *stcb) +{ + /* + * Return 0 means we want you to proceed with the abort + * non-zero means no abort processing. + */ + uint32_t new_vtag; + struct sctpasochead *head; + + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) { + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_INFO_WLOCK(); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } else { + return (0); + } + new_vtag = sctp_select_a_tag(stcb->sctp_ep, stcb->sctp_ep->sctp_lport, stcb->rport, 1); + if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) { + /* generate a new vtag and send init */ + LIST_REMOVE(stcb, sctp_asocs); + stcb->asoc.my_vtag = new_vtag; + head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))]; + /* put it in the bucket in the vtag hash of assoc's for the system */ + LIST_INSERT_HEAD(head, stcb, sctp_asocs); + SCTP_INP_INFO_WUNLOCK(); + sctp_send_initiate(stcb->sctp_ep, stcb, SCTP_SO_NOT_LOCKED); + return (1); + } else { + /* treat like a case where the cookie expired i.e.: + * - dump current cookie. + * - generate a new vtag. + * - resend init. + */ + /* generate a new vtag and send init */ + LIST_REMOVE(stcb, sctp_asocs); + SCTP_SET_STATE(stcb, SCTP_STATE_COOKIE_WAIT); + sctp_stop_all_cookie_timers(stcb); + sctp_toss_old_cookies(stcb, &stcb->asoc); + stcb->asoc.my_vtag = new_vtag; + head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))]; + /* put it in the bucket in the vtag hash of assoc's for the system */ + LIST_INSERT_HEAD(head, stcb, sctp_asocs); + SCTP_INP_INFO_WUNLOCK(); + sctp_send_initiate(stcb->sctp_ep, stcb, SCTP_SO_NOT_LOCKED); + return (1); + } + return (0); +} + +static int +sctp_handle_nat_missing_state(struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + /* return 0 means we want you to proceed with the abort + * non-zero means no abort processing + */ + if (stcb->asoc.auth_supported == 0) { + SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_nat_missing_state: Peer does not support AUTH, cannot send an asconf\n"); + return (0); + } + sctp_asconf_send_nat_state_update(stcb, net); + return (1); +} + +/* Returns 1 if the stcb was aborted, 0 otherwise */ +static int +sctp_handle_abort(struct sctp_abort_chunk *abort, + struct sctp_tcb *stcb, struct sctp_nets *net) +{ +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif + uint16_t len; + uint16_t error; + + SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_abort: handling ABORT\n"); + if (stcb == NULL) + return (0); + + len = ntohs(abort->ch.chunk_length); + if (len >= sizeof(struct sctp_chunkhdr) + sizeof(struct sctp_error_cause)) { + /* Need to check the cause codes for our + * two magic nat aborts which don't kill the assoc + * necessarily. + */ + struct sctp_error_cause *cause; + + cause = (struct sctp_error_cause *)(abort + 1); + error = ntohs(cause->code); + if (error == SCTP_CAUSE_NAT_COLLIDING_STATE) { + SCTPDBG(SCTP_DEBUG_INPUT2, "Received Colliding state, ABORT flags:%x\n", + abort->ch.chunk_flags); + if (sctp_handle_nat_colliding_state(stcb)) { + return (0); + } + } else if (error == SCTP_CAUSE_NAT_MISSING_STATE) { + SCTPDBG(SCTP_DEBUG_INPUT2, "Received missing state, ABORT flags:%x\n", + abort->ch.chunk_flags); + if (sctp_handle_nat_missing_state(stcb, net)) { + return (0); + } + } + } else { + error = 0; + } + /* stop any receive timers */ + sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_7); + /* notify user of the abort and clean up... */ + sctp_abort_notification(stcb, true, false, error, abort, SCTP_SO_NOT_LOCKED); + /* free the tcb */ + SCTP_STAT_INCR_COUNTER32(sctps_aborted); + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } +#ifdef SCTP_ASOCLOG_OF_TSNS + sctp_print_out_track_log(stcb); +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(stcb->sctp_ep); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(stcb->sctp_ep, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_8); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_handle_abort: finished\n"); + return (1); +} + +static void +sctp_start_net_timers(struct sctp_tcb *stcb) +{ + uint32_t cnt_hb_sent; + struct sctp_nets *net; + + cnt_hb_sent = 0; + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + /* For each network start: + * 1) A pmtu timer. + * 2) A HB timer + * 3) If the dest in unconfirmed send + * a hb as well if under max_hb_burst have + * been sent. + */ + sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, stcb->sctp_ep, stcb, net); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); + if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) && + (cnt_hb_sent < SCTP_BASE_SYSCTL(sctp_hb_maxburst))) { + sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED); + cnt_hb_sent++; + } + } + if (cnt_hb_sent) { + sctp_chunk_output(stcb->sctp_ep, stcb, + SCTP_OUTPUT_FROM_COOKIE_ACK, + SCTP_SO_NOT_LOCKED); + } +} + +static void +sctp_handle_shutdown(struct sctp_shutdown_chunk *cp, + struct sctp_tcb *stcb, struct sctp_nets *net, int *abort_flag) +{ + struct sctp_association *asoc; + int some_on_streamwheel; + int old_state; +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif + + SCTPDBG(SCTP_DEBUG_INPUT2, + "sctp_handle_shutdown: handling SHUTDOWN\n"); + if (stcb == NULL) + return; + asoc = &stcb->asoc; + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) { + return; + } + if (ntohs(cp->ch.chunk_length) != sizeof(struct sctp_shutdown_chunk)) { + /* Shutdown NOT the expected size */ + return; + } + old_state = SCTP_GET_STATE(stcb); + sctp_update_acked(stcb, cp, abort_flag); + if (*abort_flag) { + return; + } + if (asoc->control_pdapi) { + /* With a normal shutdown + * we assume the end of last record. + */ + SCTP_INP_READ_LOCK(stcb->sctp_ep); + if (asoc->control_pdapi->on_strm_q) { + struct sctp_stream_in *strm; + + strm = &asoc->strmin[asoc->control_pdapi->sinfo_stream]; + if (asoc->control_pdapi->on_strm_q == SCTP_ON_UNORDERED) { + /* Unordered */ + TAILQ_REMOVE(&strm->uno_inqueue, asoc->control_pdapi, next_instrm); + asoc->control_pdapi->on_strm_q = 0; + } else if (asoc->control_pdapi->on_strm_q == SCTP_ON_ORDERED) { + /* Ordered */ + TAILQ_REMOVE(&strm->inqueue, asoc->control_pdapi, next_instrm); + asoc->control_pdapi->on_strm_q = 0; +#ifdef INVARIANTS + } else { + panic("Unknown state on ctrl:%p on_strm_q:%d", + asoc->control_pdapi, + asoc->control_pdapi->on_strm_q); +#endif + } + } + asoc->control_pdapi->end_added = 1; + asoc->control_pdapi->pdapi_aborted = 1; + asoc->control_pdapi = NULL; + SCTP_INP_READ_UNLOCK(stcb->sctp_ep); +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(stcb->sctp_ep); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { + /* assoc was freed while we were unlocked */ + SCTP_SOCKET_UNLOCK(so, 1); + return; + } +#endif + if (stcb->sctp_socket) { + sctp_sorwakeup(stcb->sctp_ep, stcb->sctp_socket); + } +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } + /* goto SHUTDOWN_RECEIVED state to block new requests */ + if (stcb->sctp_socket) { + if ((SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_RECEIVED) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_ACK_SENT) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT)) { + SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_RECEIVED); + /* notify upper layer that peer has initiated a shutdown */ + sctp_ulp_notify(SCTP_NOTIFY_PEER_SHUTDOWN, stcb, 0, NULL, SCTP_SO_NOT_LOCKED); + + /* reset time */ + (void)SCTP_GETTIME_TIMEVAL(&asoc->time_entered); + } + } + if (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_SENT) { + /* + * stop the shutdown timer, since we WILL move to + * SHUTDOWN-ACK-SENT. + */ + sctp_timer_stop(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, + net, SCTP_FROM_SCTP_INPUT + SCTP_LOC_9); + } + /* Now is there unsent data on a stream somewhere? */ + some_on_streamwheel = sctp_is_there_unsent_data(stcb, SCTP_SO_NOT_LOCKED); + + if (!TAILQ_EMPTY(&asoc->send_queue) || + !TAILQ_EMPTY(&asoc->sent_queue) || + some_on_streamwheel) { + /* By returning we will push more data out */ + return; + } else { + /* no outstanding data to send, so move on... */ + /* send SHUTDOWN-ACK */ + /* move to SHUTDOWN-ACK-SENT state */ + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + if (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_ACK_SENT) { + SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_ACK_SENT); + sctp_stop_timers_for_shutdown(stcb); + sctp_send_shutdown_ack(stcb, net); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, + stcb->sctp_ep, stcb, net); + } else if (old_state == SCTP_STATE_SHUTDOWN_ACK_SENT) { + sctp_send_shutdown_ack(stcb, net); + } + } +} + +static void +sctp_handle_shutdown_ack(struct sctp_shutdown_ack_chunk *cp SCTP_UNUSED, + struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + struct sctp_association *asoc; +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; + + so = SCTP_INP_SO(stcb->sctp_ep); +#endif + SCTPDBG(SCTP_DEBUG_INPUT2, + "sctp_handle_shutdown_ack: handling SHUTDOWN ACK\n"); + if (stcb == NULL) + return; + + asoc = &stcb->asoc; + /* process according to association state */ + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) { + /* unexpected SHUTDOWN-ACK... do OOTB handling... */ + sctp_send_shutdown_complete(stcb, net, 1); + SCTP_TCB_UNLOCK(stcb); + return; + } + if ((SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { + /* unexpected SHUTDOWN-ACK... so ignore... */ + SCTP_TCB_UNLOCK(stcb); + return; + } + if (asoc->control_pdapi) { + /* With a normal shutdown + * we assume the end of last record. + */ + SCTP_INP_READ_LOCK(stcb->sctp_ep); + asoc->control_pdapi->end_added = 1; + asoc->control_pdapi->pdapi_aborted = 1; + asoc->control_pdapi = NULL; + SCTP_INP_READ_UNLOCK(stcb->sctp_ep); +#if defined(__APPLE__) && !defined(__Userspace__) + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { + /* assoc was freed while we were unlocked */ + SCTP_SOCKET_UNLOCK(so, 1); + return; + } +#endif + sctp_sorwakeup(stcb->sctp_ep, stcb->sctp_socket); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } +#ifdef INVARIANTS + if (!TAILQ_EMPTY(&asoc->send_queue) || + !TAILQ_EMPTY(&asoc->sent_queue) || + sctp_is_there_unsent_data(stcb, SCTP_SO_NOT_LOCKED)) { + panic("Queues are not empty when handling SHUTDOWN-ACK"); + } +#endif + /* stop the timer */ + sctp_timer_stop(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, net, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_10); + /* send SHUTDOWN-COMPLETE */ + sctp_send_shutdown_complete(stcb, net, 0); + /* notify upper layer protocol */ + if (stcb->sctp_socket) { + if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { + SCTP_SB_CLEAR(stcb->sctp_socket->so_snd); + } + sctp_ulp_notify(SCTP_NOTIFY_ASSOC_DOWN, stcb, 0, NULL, SCTP_SO_NOT_LOCKED); + } + SCTP_STAT_INCR_COUNTER32(sctps_shutdown); + /* free the TCB but first save off the ep */ +#if defined(__APPLE__) && !defined(__Userspace__) + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(stcb->sctp_ep, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_11); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif +} + +static void +sctp_process_unrecog_chunk(struct sctp_tcb *stcb, uint8_t chunk_type) +{ + switch (chunk_type) { + case SCTP_ASCONF_ACK: + case SCTP_ASCONF: + sctp_asconf_cleanup(stcb); + break; + case SCTP_IFORWARD_CUM_TSN: + case SCTP_FORWARD_CUM_TSN: + stcb->asoc.prsctp_supported = 0; + break; + default: + SCTPDBG(SCTP_DEBUG_INPUT2, + "Peer does not support chunk type %d (0x%x).\n", + chunk_type, chunk_type); + break; + } +} + +/* + * Skip past the param header and then we will find the param that caused the + * problem. There are a number of param's in a ASCONF OR the prsctp param + * these will turn of specific features. + * XXX: Is this the right thing to do? + */ +static void +sctp_process_unrecog_param(struct sctp_tcb *stcb, uint16_t parameter_type) +{ + switch (parameter_type) { + /* pr-sctp draft */ + case SCTP_PRSCTP_SUPPORTED: + stcb->asoc.prsctp_supported = 0; + break; + case SCTP_SUPPORTED_CHUNK_EXT: + break; + /* draft-ietf-tsvwg-addip-sctp */ + case SCTP_HAS_NAT_SUPPORT: + stcb->asoc.peer_supports_nat = 0; + break; + case SCTP_ADD_IP_ADDRESS: + case SCTP_DEL_IP_ADDRESS: + case SCTP_SET_PRIM_ADDR: + stcb->asoc.asconf_supported = 0; + break; + case SCTP_SUCCESS_REPORT: + case SCTP_ERROR_CAUSE_IND: + SCTPDBG(SCTP_DEBUG_INPUT2, "Huh, the peer does not support success? or error cause?\n"); + SCTPDBG(SCTP_DEBUG_INPUT2, + "Turning off ASCONF to this strange peer\n"); + stcb->asoc.asconf_supported = 0; + break; + default: + SCTPDBG(SCTP_DEBUG_INPUT2, + "Peer does not support param type %d (0x%x)??\n", + parameter_type, parameter_type); + break; + } +} + +static int +sctp_handle_error(struct sctp_chunkhdr *ch, + struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t limit) +{ + struct sctp_error_cause *cause; + struct sctp_association *asoc; + uint32_t remaining_length, adjust; + uint16_t code, cause_code, cause_length; +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif + + /* parse through all of the errors and process */ + asoc = &stcb->asoc; + cause = (struct sctp_error_cause *)((caddr_t)ch + + sizeof(struct sctp_chunkhdr)); + remaining_length = ntohs(ch->chunk_length); + if (remaining_length > limit) { + remaining_length = limit; + } + if (remaining_length >= sizeof(struct sctp_chunkhdr)) { + remaining_length -= sizeof(struct sctp_chunkhdr); + } else { + remaining_length = 0; + } + code = 0; + while (remaining_length >= sizeof(struct sctp_error_cause)) { + /* Process an Error Cause */ + cause_code = ntohs(cause->code); + cause_length = ntohs(cause->length); + if ((cause_length > remaining_length) || (cause_length == 0)) { + /* Invalid cause length, possibly due to truncation. */ + SCTPDBG(SCTP_DEBUG_INPUT1, "Bogus length in cause - bytes left: %u cause length: %u\n", + remaining_length, cause_length); + return (0); + } + if (code == 0) { + /* report the first error cause */ + code = cause_code; + } + switch (cause_code) { + case SCTP_CAUSE_INVALID_STREAM: + case SCTP_CAUSE_MISSING_PARAM: + case SCTP_CAUSE_INVALID_PARAM: + case SCTP_CAUSE_NO_USER_DATA: + SCTPDBG(SCTP_DEBUG_INPUT1, "Software error we got a %u back? We have a bug :/ (or do they?)\n", + cause_code); + break; + case SCTP_CAUSE_NAT_COLLIDING_STATE: + SCTPDBG(SCTP_DEBUG_INPUT2, "Received Colliding state, ERROR flags: %x\n", + ch->chunk_flags); + if (sctp_handle_nat_colliding_state(stcb)) { + return (0); + } + break; + case SCTP_CAUSE_NAT_MISSING_STATE: + SCTPDBG(SCTP_DEBUG_INPUT2, "Received missing state, ERROR flags: %x\n", + ch->chunk_flags); + if (sctp_handle_nat_missing_state(stcb, net)) { + return (0); + } + break; + case SCTP_CAUSE_STALE_COOKIE: + /* + * We only act if we have echoed a cookie and are + * waiting. + */ + if ((cause_length >= sizeof(struct sctp_error_stale_cookie)) && + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) { + struct sctp_error_stale_cookie *stale_cookie; + + stale_cookie = (struct sctp_error_stale_cookie *)cause; + /* stable_time is in usec, convert to msec. */ + asoc->cookie_preserve_req = ntohl(stale_cookie->stale_time) / 1000; + /* Double it to be more robust on RTX. */ + asoc->cookie_preserve_req *= 2; + asoc->stale_cookie_count++; + if (asoc->stale_cookie_count > + asoc->max_init_times) { + sctp_abort_notification(stcb, false, true, 0, NULL, SCTP_SO_NOT_LOCKED); + /* now free the asoc */ +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(stcb->sctp_ep); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(stcb->sctp_ep, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_12); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + return (-1); + } + /* blast back to INIT state */ + sctp_toss_old_cookies(stcb, &stcb->asoc); + SCTP_SET_STATE(stcb, SCTP_STATE_COOKIE_WAIT); + sctp_stop_all_cookie_timers(stcb); + sctp_send_initiate(stcb->sctp_ep, stcb, SCTP_SO_NOT_LOCKED); + } + break; + case SCTP_CAUSE_UNRESOLVABLE_ADDR: + /* + * Nothing we can do here, we don't do hostname + * addresses so if the peer does not like my IPv6 + * (or IPv4 for that matter) it does not matter. If + * they don't support that type of address, they can + * NOT possibly get that packet type... i.e. with no + * IPv6 you can't receive a IPv6 packet. so we can + * safely ignore this one. If we ever added support + * for HOSTNAME Addresses, then we would need to do + * something here. + */ + break; + case SCTP_CAUSE_UNRECOG_CHUNK: + if (cause_length >= sizeof(struct sctp_error_unrecognized_chunk)) { + struct sctp_error_unrecognized_chunk *unrec_chunk; + + unrec_chunk = (struct sctp_error_unrecognized_chunk *)cause; + sctp_process_unrecog_chunk(stcb, unrec_chunk->ch.chunk_type); + } + break; + case SCTP_CAUSE_UNRECOG_PARAM: + /* XXX: We only consider the first parameter */ + if (cause_length >= sizeof(struct sctp_error_cause) + sizeof(struct sctp_paramhdr)) { + struct sctp_paramhdr *unrec_parameter; + + unrec_parameter = (struct sctp_paramhdr *)(cause + 1); + sctp_process_unrecog_param(stcb, ntohs(unrec_parameter->param_type)); + } + break; + case SCTP_CAUSE_COOKIE_IN_SHUTDOWN: + /* + * We ignore this since the timer will drive out a + * new cookie anyway and there timer will drive us + * to send a SHUTDOWN_COMPLETE. We can't send one + * here since we don't have their tag. + */ + break; + case SCTP_CAUSE_DELETING_LAST_ADDR: + case SCTP_CAUSE_RESOURCE_SHORTAGE: + case SCTP_CAUSE_DELETING_SRC_ADDR: + /* + * We should NOT get these here, but in a + * ASCONF-ACK. + */ + SCTPDBG(SCTP_DEBUG_INPUT2, "Peer sends ASCONF errors in a error cause with code %u.\n", + cause_code); + break; + case SCTP_CAUSE_OUT_OF_RESC: + /* + * And what, pray tell do we do with the fact that + * the peer is out of resources? Not really sure we + * could do anything but abort. I suspect this + * should have came WITH an abort instead of in a + * OP-ERROR. + */ + break; + default: + SCTPDBG(SCTP_DEBUG_INPUT1, "sctp_handle_error: unknown code 0x%x\n", + cause_code); + break; + } + adjust = SCTP_SIZE32(cause_length); + if (remaining_length >= adjust) { + remaining_length -= adjust; + } else { + remaining_length = 0; + } + cause = (struct sctp_error_cause *)((caddr_t)cause + adjust); + } + sctp_ulp_notify(SCTP_NOTIFY_REMOTE_ERROR, stcb, code, ch, SCTP_SO_NOT_LOCKED); + return (0); +} + +static int +sctp_handle_init_ack(struct mbuf *m, int iphlen, int offset, + struct sockaddr *src, struct sockaddr *dst, struct sctphdr *sh, + struct sctp_init_ack_chunk *cp, struct sctp_tcb *stcb, + struct sctp_nets *net, int *abort_no_unlock, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, +#endif + uint32_t vrf_id) +{ + struct sctp_init_ack *init_ack; + struct mbuf *op_err; + + SCTPDBG(SCTP_DEBUG_INPUT2, + "sctp_handle_init_ack: handling INIT-ACK\n"); + + if (stcb == NULL) { + SCTPDBG(SCTP_DEBUG_INPUT2, + "sctp_handle_init_ack: TCB is null\n"); + return (-1); + } + /* Only process the INIT-ACK chunk in COOKIE WAIT state.*/ + if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) { + init_ack = &cp->init; + /* Validate parameters. */ + if ((ntohl(init_ack->initiate_tag) == 0) || + (ntohl(init_ack->a_rwnd) < SCTP_MIN_RWND) || + (ntohs(init_ack->num_inbound_streams) == 0) || + (ntohs(init_ack->num_outbound_streams) == 0)) { + /* One of the mandatory parameters is illegal. */ + op_err = sctp_generate_cause(SCTP_CAUSE_INVALID_PARAM, ""); + sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, + src, dst, sh, op_err, + #if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, + #endif + vrf_id, net->port); + *abort_no_unlock = 1; + return (-1); + } + if (stcb->asoc.primary_destination->dest_state & + SCTP_ADDR_UNCONFIRMED) { + /* + * The primary is where we sent the INIT, we can + * always consider it confirmed when the INIT-ACK is + * returned. Do this before we load addresses + * though. + */ + stcb->asoc.primary_destination->dest_state &= + ~SCTP_ADDR_UNCONFIRMED; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_CONFIRMED, + stcb, 0, (void *)stcb->asoc.primary_destination, SCTP_SO_NOT_LOCKED); + } + if (sctp_process_init_ack(m, iphlen, offset, src, dst, sh, cp, stcb, + net, abort_no_unlock, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id) < 0) { + /* error in parsing parameters */ + return (-1); + } + /* Update our state. */ + SCTPDBG(SCTP_DEBUG_INPUT2, "moving to COOKIE-ECHOED state\n"); + SCTP_SET_STATE(stcb, SCTP_STATE_COOKIE_ECHOED); + + /* Reset the RTO calculation. */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { + sctp_misc_ints(SCTP_THRESHOLD_CLEAR, + stcb->asoc.overall_error_count, + 0, + SCTP_FROM_SCTP_INPUT, + __LINE__); + } + stcb->asoc.overall_error_count = 0; + (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); + /* + * Collapse the init timer back in case of a exponential + * backoff. + */ + sctp_timer_start(SCTP_TIMER_TYPE_COOKIE, stcb->sctp_ep, + stcb, net); + /* + * The output routine at the end of the inbound data processing + * will cause the cookie to be sent. + */ + SCTPDBG(SCTP_DEBUG_INPUT1, "Leaving handle-init-ack end\n"); + return (0); + } else { + return (-1); + } +} + +static struct sctp_tcb * +sctp_process_cookie_new(struct mbuf *m, int iphlen, int offset, + struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, struct sctp_state_cookie *cookie, int cookie_len, + struct sctp_inpcb *inp, struct sctp_nets **netp, + struct sockaddr *init_src, int *notification, + int auth_skipped, uint32_t auth_offset, uint32_t auth_len, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, +#endif + uint32_t vrf_id, uint16_t port); + +/* + * handle a state cookie for an existing association m: input packet mbuf + * chain-- assumes a pullup on IP/SCTP/COOKIE-ECHO chunk note: this is a + * "split" mbuf and the cookie signature does not exist offset: offset into + * mbuf to the cookie-echo chunk + */ +static struct sctp_tcb * +sctp_process_cookie_existing(struct mbuf *m, int iphlen, int offset, + struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, struct sctp_state_cookie *cookie, int cookie_len, + struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets **netp, + struct sockaddr *init_src, int *notification, + int auth_skipped, uint32_t auth_offset, uint32_t auth_len, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, +#endif + uint32_t vrf_id, uint16_t port) +{ + struct sctp_association *asoc; + struct sctp_init_chunk *init_cp, init_buf; + struct sctp_init_ack_chunk *initack_cp, initack_buf; + struct sctp_asconf_addr *aparam, *naparam; + struct sctp_asconf_ack *aack, *naack; + struct sctp_tmit_chunk *chk, *nchk; + struct sctp_stream_reset_list *strrst, *nstrrst; + struct sctp_queued_to_read *sq, *nsq; + struct sctp_nets *net; + struct mbuf *op_err; + int init_offset, initack_offset, i; + int retval; + int spec_flag = 0; + uint32_t how_indx; +#if defined(SCTP_DETAILED_STR_STATS) + int j; +#endif + + net = *netp; + /* I know that the TCB is non-NULL from the caller */ + asoc = &stcb->asoc; + for (how_indx = 0; how_indx < sizeof(asoc->cookie_how); how_indx++) { + if (asoc->cookie_how[how_indx] == 0) + break; + } + if (how_indx < sizeof(asoc->cookie_how)) { + asoc->cookie_how[how_indx] = 1; + } + if (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_ACK_SENT) { + /* SHUTDOWN came in after sending INIT-ACK */ + sctp_send_shutdown_ack(stcb, stcb->asoc.primary_destination); + op_err = sctp_generate_cause(SCTP_CAUSE_COOKIE_IN_SHUTDOWN, ""); + sctp_send_operr_to(src, dst, sh, cookie->peers_vtag, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, net->port); + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 2; + SCTP_TCB_UNLOCK(stcb); + return (NULL); + } + /* + * find and validate the INIT chunk in the cookie (peer's info) the + * INIT should start after the cookie-echo header struct (chunk + * header, state cookie header struct) + */ + init_offset = offset += sizeof(struct sctp_cookie_echo_chunk); + + init_cp = (struct sctp_init_chunk *) + sctp_m_getptr(m, init_offset, sizeof(struct sctp_init_chunk), + (uint8_t *) & init_buf); + if (init_cp == NULL) { + /* could not pull a INIT chunk in cookie */ + SCTP_TCB_UNLOCK(stcb); + return (NULL); + } + if (init_cp->ch.chunk_type != SCTP_INITIATION) { + SCTP_TCB_UNLOCK(stcb); + return (NULL); + } + /* + * find and validate the INIT-ACK chunk in the cookie (my info) the + * INIT-ACK follows the INIT chunk + */ + initack_offset = init_offset + SCTP_SIZE32(ntohs(init_cp->ch.chunk_length)); + initack_cp = (struct sctp_init_ack_chunk *) + sctp_m_getptr(m, initack_offset, sizeof(struct sctp_init_ack_chunk), + (uint8_t *) & initack_buf); + if (initack_cp == NULL) { + /* could not pull INIT-ACK chunk in cookie */ + SCTP_TCB_UNLOCK(stcb); + return (NULL); + } + if (initack_cp->ch.chunk_type != SCTP_INITIATION_ACK) { + SCTP_TCB_UNLOCK(stcb); + return (NULL); + } + if ((ntohl(initack_cp->init.initiate_tag) == asoc->my_vtag) && + (ntohl(init_cp->init.initiate_tag) == asoc->peer_vtag)) { + /* + * case D in Section 5.2.4 Table 2: MMAA process accordingly + * to get into the OPEN state + */ + if (ntohl(initack_cp->init.initial_tsn) != asoc->init_seq_number) { + /*- + * Opps, this means that we somehow generated two vtag's + * the same. I.e. we did: + * Us Peer + * <---INIT(tag=a)------ + * ----INIT-ACK(tag=t)--> + * ----INIT(tag=t)------> *1 + * <---INIT-ACK(tag=a)--- + * <----CE(tag=t)------------- *2 + * + * At point *1 we should be generating a different + * tag t'. Which means we would throw away the CE and send + * ours instead. Basically this is case C (throw away side). + */ + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 17; + SCTP_TCB_UNLOCK(stcb); + return (NULL); + } + switch (SCTP_GET_STATE(stcb)) { + case SCTP_STATE_COOKIE_WAIT: + case SCTP_STATE_COOKIE_ECHOED: + /* + * INIT was sent but got a COOKIE_ECHO with the + * correct tags... just accept it...but we must + * process the init so that we can make sure we + * have the right seq no's. + */ + /* First we must process the INIT !! */ + if (sctp_process_init(init_cp, stcb) < 0) { + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 3; + op_err = sctp_generate_cause(SCTP_CAUSE_OUT_OF_RESC, ""); + SCTPDBG(SCTP_DEBUG_INPUT1, "sctp_process_init() failed\n"); + sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, net->port); + return (NULL); + } + /* we have already processed the INIT so no problem */ + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, + stcb, net, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_13); + sctp_timer_stop(SCTP_TIMER_TYPE_INIT, inp, + stcb, net, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_14); + /* update current state */ + if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED) + SCTP_STAT_INCR_COUNTER32(sctps_activeestab); + else + SCTP_STAT_INCR_COUNTER32(sctps_collisionestab); + + SCTP_SET_STATE(stcb, SCTP_STATE_OPEN); + if (asoc->state & SCTP_STATE_SHUTDOWN_PENDING) { + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, + stcb->sctp_ep, stcb, NULL); + } + SCTP_STAT_INCR_GAUGE32(sctps_currestab); + sctp_stop_all_cookie_timers(stcb); + if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) && + (!SCTP_IS_LISTENING(inp))) { +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif + /* + * Here is where collision would go if we + * did a connect() and instead got a + * init/init-ack/cookie done before the + * init-ack came back.. + */ + sctp_pcb_add_flags(stcb->sctp_ep, SCTP_PCB_FLAGS_CONNECTED); +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(stcb->sctp_ep); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_UNLOCK(so, 1); + return (NULL); + } +#endif + soisconnected(stcb->sctp_socket); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } + /* notify upper layer */ + *notification = SCTP_NOTIFY_ASSOC_UP; + net->hb_responded = 1; + if (stcb->asoc.sctp_autoclose_ticks && + (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE))) { + sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, + inp, stcb, NULL); + } + break; + default: + /* + * we're in the OPEN state (or beyond), so + * peer must have simply lost the COOKIE-ACK + */ + break; + } /* end switch */ + sctp_stop_all_cookie_timers(stcb); + if ((retval = sctp_load_addresses_from_init(stcb, m, + init_offset + sizeof(struct sctp_init_chunk), + initack_offset, src, dst, init_src, stcb->asoc.port)) < 0) { + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 4; + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + "Problem with address parameters"); + SCTPDBG(SCTP_DEBUG_INPUT1, + "Load addresses from INIT causes an abort %d\n", + retval); + sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, net->port); + return (NULL); + } + /* respond with a COOKIE-ACK */ + sctp_toss_old_cookies(stcb, asoc); + sctp_send_cookie_ack(stcb); + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 5; + return (stcb); + } + + if (ntohl(initack_cp->init.initiate_tag) != asoc->my_vtag && + ntohl(init_cp->init.initiate_tag) == asoc->peer_vtag && + cookie->tie_tag_my_vtag == 0 && + cookie->tie_tag_peer_vtag == 0) { + /* + * case C in Section 5.2.4 Table 2: XMOO silently discard + */ + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 6; + SCTP_TCB_UNLOCK(stcb); + return (NULL); + } + /* If nat support, and the below and stcb is established, + * send back a ABORT(colliding state) if we are established. + */ + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) && + (asoc->peer_supports_nat) && + ((ntohl(initack_cp->init.initiate_tag) == asoc->my_vtag) && + ((ntohl(init_cp->init.initiate_tag) != asoc->peer_vtag) || + (asoc->peer_vtag == 0)))) { + /* Special case - Peer's support nat. We may have + * two init's that we gave out the same tag on since + * one was not established.. i.e. we get INIT from host-1 + * behind the nat and we respond tag-a, we get a INIT from + * host-2 behind the nat and we get tag-a again. Then we + * bring up host-1 (or 2's) assoc, Then comes the cookie + * from hsot-2 (or 1). Now we have colliding state. We must + * send an abort here with colliding state indication. + */ + op_err = sctp_generate_cause(SCTP_CAUSE_NAT_COLLIDING_STATE, ""); + sctp_send_abort(m, iphlen, src, dst, sh, 0, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, port); + SCTP_TCB_UNLOCK(stcb); + return (NULL); + } + if ((ntohl(initack_cp->init.initiate_tag) == asoc->my_vtag) && + ((ntohl(init_cp->init.initiate_tag) != asoc->peer_vtag) || + (asoc->peer_vtag == 0))) { + /* + * case B in Section 5.2.4 Table 2: MXAA or MOAA my info + * should be ok, re-accept peer info + */ + if (ntohl(initack_cp->init.initial_tsn) != asoc->init_seq_number) { + /* Extension of case C. + * If we hit this, then the random number + * generator returned the same vtag when we + * first sent our INIT-ACK and when we later sent + * our INIT. The side with the seq numbers that are + * different will be the one that normally would + * have hit case C. This in effect "extends" our vtags + * in this collision case to be 64 bits. The same collision + * could occur aka you get both vtag and seq number the + * same twice in a row.. but is much less likely. If it + * did happen then we would proceed through and bring + * up the assoc.. we may end up with the wrong stream + * setup however.. which would be bad.. but there is + * no way to tell.. until we send on a stream that does + * not exist :-) + */ + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 7; + + SCTP_TCB_UNLOCK(stcb); + return (NULL); + } + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 8; + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_15); + sctp_stop_all_cookie_timers(stcb); + /* + * since we did not send a HB make sure we don't double + * things + */ + net->hb_responded = 1; + if (stcb->asoc.sctp_autoclose_ticks && + sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) { + sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb, + NULL); + } + asoc->my_rwnd = ntohl(initack_cp->init.a_rwnd); + if (asoc->pre_open_streams < asoc->streamoutcnt) { + asoc->pre_open_streams = asoc->streamoutcnt; + } + + if (ntohl(init_cp->init.initiate_tag) != asoc->peer_vtag) { + /* Ok the peer probably discarded our + * data (if we echoed a cookie+data). So anything + * on the sent_queue should be marked for + * retransmit, we may not get something to + * kick us so it COULD still take a timeout + * to move these.. but it can't hurt to mark them. + */ + + TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { + if (chk->sent < SCTP_DATAGRAM_RESEND) { + chk->sent = SCTP_DATAGRAM_RESEND; + sctp_flight_size_decrease(chk); + sctp_total_flight_decrease(stcb, chk); + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + spec_flag++; + } + } + } + /* process the INIT info (peer's info) */ + if (sctp_process_init(init_cp, stcb) < 0) { + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 9; + op_err = sctp_generate_cause(SCTP_CAUSE_OUT_OF_RESC, ""); + SCTPDBG(SCTP_DEBUG_INPUT1, "sctp_process_init() failed\n"); + sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, net->port); + return (NULL); + } + if ((retval = sctp_load_addresses_from_init(stcb, m, + init_offset + sizeof(struct sctp_init_chunk), + initack_offset, src, dst, init_src, stcb->asoc.port)) < 0) { + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 10; + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + "Problem with address parameters"); + SCTPDBG(SCTP_DEBUG_INPUT1, + "Load addresses from INIT causes an abort %d\n", + retval); + sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, net->port); + return (NULL); + } + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) { + *notification = SCTP_NOTIFY_ASSOC_UP; + + if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) && + (!SCTP_IS_LISTENING(inp))) { +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif + sctp_pcb_add_flags(stcb->sctp_ep, SCTP_PCB_FLAGS_CONNECTED); +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(stcb->sctp_ep); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_UNLOCK(so, 1); + return (NULL); + } +#endif + soisconnected(stcb->sctp_socket); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } + if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED) + SCTP_STAT_INCR_COUNTER32(sctps_activeestab); + else + SCTP_STAT_INCR_COUNTER32(sctps_collisionestab); + SCTP_STAT_INCR_GAUGE32(sctps_currestab); + } else if (SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) { + SCTP_STAT_INCR_COUNTER32(sctps_restartestab); + } else { + SCTP_STAT_INCR_COUNTER32(sctps_collisionestab); + } + SCTP_SET_STATE(stcb, SCTP_STATE_OPEN); + if (asoc->state & SCTP_STATE_SHUTDOWN_PENDING) { + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, + stcb->sctp_ep, stcb, NULL); + } + sctp_stop_all_cookie_timers(stcb); + sctp_toss_old_cookies(stcb, asoc); + sctp_send_cookie_ack(stcb); + if (spec_flag) { + /* only if we have retrans set do we do this. What + * this call does is get only the COOKIE-ACK out + * and then when we return the normal call to + * sctp_chunk_output will get the retrans out + * behind this. + */ + sctp_chunk_output(inp,stcb, SCTP_OUTPUT_FROM_COOKIE_ACK, SCTP_SO_NOT_LOCKED); + } + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 11; + + return (stcb); + } + if ((ntohl(initack_cp->init.initiate_tag) != asoc->my_vtag && + ntohl(init_cp->init.initiate_tag) != asoc->peer_vtag) && + cookie->tie_tag_my_vtag == asoc->my_vtag_nonce && + cookie->tie_tag_peer_vtag == asoc->peer_vtag_nonce && + cookie->tie_tag_peer_vtag != 0) { + struct sctpasochead *head; +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif + + if (asoc->peer_supports_nat) { + struct sctp_tcb *local_stcb; + + /* This is a gross gross hack. + * Just call the cookie_new code since we + * are allowing a duplicate association. + * I hope this works... + */ + local_stcb = sctp_process_cookie_new(m, iphlen, offset, src, dst, + sh, cookie, cookie_len, + inp, netp, init_src,notification, + auth_skipped, auth_offset, auth_len, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, port); + if (local_stcb == NULL) { + SCTP_TCB_UNLOCK(stcb); + } + return (local_stcb); + } + /* + * case A in Section 5.2.4 Table 2: XXMM (peer restarted) + */ + /* temp code */ + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 12; + sctp_stop_association_timers(stcb, false); + /* notify upper layer */ + *notification = SCTP_NOTIFY_ASSOC_RESTART; + atomic_add_int(&stcb->asoc.refcnt, 1); + if ((SCTP_GET_STATE(stcb) != SCTP_STATE_OPEN) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_RECEIVED) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT)) { + SCTP_STAT_INCR_GAUGE32(sctps_currestab); + } + if (SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) { + SCTP_STAT_INCR_GAUGE32(sctps_restartestab); + } else if (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) { + SCTP_STAT_INCR_GAUGE32(sctps_collisionestab); + } + if (asoc->state & SCTP_STATE_SHUTDOWN_PENDING) { + SCTP_SET_STATE(stcb, SCTP_STATE_OPEN); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, + stcb->sctp_ep, stcb, NULL); + + } else if (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) { + /* move to OPEN state, if not in SHUTDOWN_SENT */ + SCTP_SET_STATE(stcb, SCTP_STATE_OPEN); + } + if (asoc->pre_open_streams < asoc->streamoutcnt) { + asoc->pre_open_streams = asoc->streamoutcnt; + } + asoc->init_seq_number = ntohl(initack_cp->init.initial_tsn); + asoc->sending_seq = asoc->asconf_seq_out = asoc->str_reset_seq_out = asoc->init_seq_number; + asoc->asconf_seq_out_acked = asoc->asconf_seq_out - 1; + asoc->asconf_seq_in = asoc->last_acked_seq = asoc->init_seq_number - 1; + asoc->str_reset_seq_in = asoc->init_seq_number; + asoc->advanced_peer_ack_point = asoc->last_acked_seq; + asoc->send_sack = 1; + asoc->data_pkts_seen = 0; + asoc->last_data_chunk_from = NULL; + asoc->last_control_chunk_from = NULL; + asoc->last_net_cmt_send_started = NULL; + if (asoc->mapping_array) { + memset(asoc->mapping_array, 0, + asoc->mapping_array_size); + } + if (asoc->nr_mapping_array) { + memset(asoc->nr_mapping_array, 0, + asoc->mapping_array_size); + } + SCTP_TCB_UNLOCK(stcb); +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(stcb->sctp_ep); + SCTP_SOCKET_LOCK(so, 1); +#endif + SCTP_INP_INFO_WLOCK(); + SCTP_INP_WLOCK(stcb->sctp_ep); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + /* send up all the data */ + sctp_report_all_outbound(stcb, 0, SCTP_SO_LOCKED); + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + stcb->asoc.strmout[i].chunks_on_queues = 0; +#if defined(SCTP_DETAILED_STR_STATS) + for (j = 0; j < SCTP_PR_SCTP_MAX + 1; j++) { + asoc->strmout[i].abandoned_sent[j] = 0; + asoc->strmout[i].abandoned_unsent[j] = 0; + } +#else + asoc->strmout[i].abandoned_sent[0] = 0; + asoc->strmout[i].abandoned_unsent[0] = 0; +#endif + stcb->asoc.strmout[i].next_mid_ordered = 0; + stcb->asoc.strmout[i].next_mid_unordered = 0; + stcb->asoc.strmout[i].sid = i; + stcb->asoc.strmout[i].last_msg_incomplete = 0; + } + TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) { + TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp); + SCTP_FREE(strrst, SCTP_M_STRESET); + } + TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) { + TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next); + if (sq->data) { + sctp_m_freem(sq->data); + sq->data = NULL; + } + sctp_free_remote_addr(sq->whoFrom); + sq->whoFrom = NULL; + sq->stcb = NULL; + sctp_free_a_readq(stcb, sq); + } + TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) { + TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + if (chk->holds_key_ref) + sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); + sctp_free_remote_addr(chk->whoTo); + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); + SCTP_DECR_CHK_COUNT(); + } + asoc->ctrl_queue_cnt = 0; + asoc->str_reset = NULL; + asoc->stream_reset_outstanding = 0; + TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) { + TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next); + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + if (chk->holds_key_ref) + sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); + sctp_free_remote_addr(chk->whoTo); + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); + SCTP_DECR_CHK_COUNT(); + } + TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) { + TAILQ_REMOVE(&asoc->asconf_queue, aparam, next); + SCTP_FREE(aparam,SCTP_M_ASC_ADDR); + } + TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) { + TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next); + if (aack->data != NULL) { + sctp_m_freem(aack->data); + } + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack); + } + asoc->zero_checksum = cookie->zero_checksum; + + /* process the INIT-ACK info (my info) */ + asoc->my_vtag = ntohl(initack_cp->init.initiate_tag); + asoc->my_rwnd = ntohl(initack_cp->init.a_rwnd); + + /* pull from vtag hash */ + LIST_REMOVE(stcb, sctp_asocs); + /* re-insert to new vtag position */ + head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, + SCTP_BASE_INFO(hashasocmark))]; + /* + * put it in the bucket in the vtag hash of assoc's for the + * system + */ + LIST_INSERT_HEAD(head, stcb, sctp_asocs); + + SCTP_INP_WUNLOCK(stcb->sctp_ep); + SCTP_INP_INFO_WUNLOCK(); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + asoc->total_flight = 0; + asoc->total_flight_count = 0; + /* process the INIT info (peer's info) */ + if (sctp_process_init(init_cp, stcb) < 0) { + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 13; + op_err = sctp_generate_cause(SCTP_CAUSE_OUT_OF_RESC, ""); + SCTPDBG(SCTP_DEBUG_INPUT1, "sctp_process_init() failed\n"); + sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, net->port); + return (NULL); + } + /* + * since we did not send a HB make sure we don't double + * things + */ + net->hb_responded = 1; + + if ((retval = sctp_load_addresses_from_init(stcb, m, + init_offset + sizeof(struct sctp_init_chunk), + initack_offset, src, dst, init_src, stcb->asoc.port)) < 0) { + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 14; + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + "Problem with address parameters"); + SCTPDBG(SCTP_DEBUG_INPUT1, + "Load addresses from INIT causes an abort %d\n", + retval); + sctp_abort_association(stcb->sctp_ep, stcb, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, net->port); + return (NULL); + } + /* respond with a COOKIE-ACK */ + sctp_send_cookie_ack(stcb); + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 15; + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE) && + (asoc->sctp_autoclose_ticks > 0)) { + sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb, NULL); + } + return (stcb); + } + if (how_indx < sizeof(asoc->cookie_how)) + asoc->cookie_how[how_indx] = 16; + /* all other cases... */ + SCTP_TCB_UNLOCK(stcb); + return (NULL); +} + +/* + * handle a state cookie for a new association m: input packet mbuf chain-- + * assumes a pullup on IP/SCTP/COOKIE-ECHO chunk note: this is a "split" mbuf + * and the cookie signature does not exist offset: offset into mbuf to the + * cookie-echo chunk length: length of the cookie chunk to: where the init + * was from returns a new TCB + */ +static struct sctp_tcb * +sctp_process_cookie_new(struct mbuf *m, int iphlen, int offset, + struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, struct sctp_state_cookie *cookie, int cookie_len, + struct sctp_inpcb *inp, struct sctp_nets **netp, + struct sockaddr *init_src, int *notification, + int auth_skipped, uint32_t auth_offset, uint32_t auth_len, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, +#endif + uint32_t vrf_id, uint16_t port) +{ + struct sctp_tcb *stcb; + struct sctp_init_chunk *init_cp, init_buf; + struct sctp_init_ack_chunk *initack_cp, initack_buf; + union sctp_sockstore store; + struct sctp_association *asoc; + int init_offset, initack_offset, initack_limit; + int error = 0; + uint8_t auth_chunk_buf[SCTP_CHUNK_BUFFER_SIZE]; +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; + + so = SCTP_INP_SO(inp); +#endif + + /* + * find and validate the INIT chunk in the cookie (peer's info) the + * INIT should start after the cookie-echo header struct (chunk + * header, state cookie header struct) + */ + init_offset = offset + sizeof(struct sctp_cookie_echo_chunk); + init_cp = (struct sctp_init_chunk *) + sctp_m_getptr(m, init_offset, sizeof(struct sctp_init_chunk), + (uint8_t *) & init_buf); + if (init_cp == NULL) { + /* could not pull a INIT chunk in cookie */ + SCTPDBG(SCTP_DEBUG_INPUT1, + "process_cookie_new: could not pull INIT chunk hdr\n"); + return (NULL); + } + if (init_cp->ch.chunk_type != SCTP_INITIATION) { + SCTPDBG(SCTP_DEBUG_INPUT1, "HUH? process_cookie_new: could not find INIT chunk!\n"); + return (NULL); + } + initack_offset = init_offset + SCTP_SIZE32(ntohs(init_cp->ch.chunk_length)); + /* + * find and validate the INIT-ACK chunk in the cookie (my info) the + * INIT-ACK follows the INIT chunk + */ + initack_cp = (struct sctp_init_ack_chunk *) + sctp_m_getptr(m, initack_offset, sizeof(struct sctp_init_ack_chunk), + (uint8_t *) & initack_buf); + if (initack_cp == NULL) { + /* could not pull INIT-ACK chunk in cookie */ + SCTPDBG(SCTP_DEBUG_INPUT1, "process_cookie_new: could not pull INIT-ACK chunk hdr\n"); + return (NULL); + } + if (initack_cp->ch.chunk_type != SCTP_INITIATION_ACK) { + return (NULL); + } + /* + * NOTE: We can't use the INIT_ACK's chk_length to determine the + * "initack_limit" value. This is because the chk_length field + * includes the length of the cookie, but the cookie is omitted when + * the INIT and INIT_ACK are tacked onto the cookie... + */ + initack_limit = offset + cookie_len; + + /* + * now that we know the INIT/INIT-ACK are in place, create a new TCB + * and populate + */ + + /* + * Here we do a trick, we set in NULL for the proc/thread argument. We + * do this since in effect we only use the p argument when + * the socket is unbound and we must do an implicit bind. + * Since we are getting a cookie, we cannot be unbound. + */ + stcb = sctp_aloc_assoc(inp, init_src, &error, + ntohl(initack_cp->init.initiate_tag), + ntohl(initack_cp->init.initial_tsn), vrf_id, + ntohs(initack_cp->init.num_outbound_streams), + port, +#if defined(__FreeBSD__) && !defined(__Userspace__) + (struct thread *)NULL, +#elif defined(_WIN32) && !defined(__Userspace__) + (PKTHREAD)NULL, +#else + (struct proc *)NULL, +#endif + SCTP_DONT_INITIALIZE_AUTH_PARAMS); + if (stcb == NULL) { + struct mbuf *op_err; + + /* memory problem? */ + SCTPDBG(SCTP_DEBUG_INPUT1, + "process_cookie_new: no room for another TCB!\n"); + op_err = sctp_generate_cause(SCTP_CAUSE_OUT_OF_RESC, ""); + sctp_abort_association(inp, (struct sctp_tcb *)NULL, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, port); + return (NULL); + } + asoc = &stcb->asoc; + /* get scope variables out of cookie */ + asoc->scope.ipv4_local_scope = cookie->ipv4_scope; + asoc->scope.site_scope = cookie->site_scope; + asoc->scope.local_scope = cookie->local_scope; + asoc->scope.loopback_scope = cookie->loopback_scope; + +#if defined(__Userspace__) + if ((asoc->scope.ipv4_addr_legal != cookie->ipv4_addr_legal) || + (asoc->scope.ipv6_addr_legal != cookie->ipv6_addr_legal) || + (asoc->scope.conn_addr_legal != cookie->conn_addr_legal)) { +#else + if ((asoc->scope.ipv4_addr_legal != cookie->ipv4_addr_legal) || + (asoc->scope.ipv6_addr_legal != cookie->ipv6_addr_legal)) { +#endif + struct mbuf *op_err; + + /* + * Houston we have a problem. The EP changed while the + * cookie was in flight. Only recourse is to abort the + * association. + */ + op_err = sctp_generate_cause(SCTP_CAUSE_OUT_OF_RESC, ""); + sctp_abort_association(inp, (struct sctp_tcb *)NULL, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, port); +#if defined(__APPLE__) && !defined(__Userspace__) + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_18); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + return (NULL); + } + asoc->zero_checksum = cookie->zero_checksum; + /* process the INIT-ACK info (my info) */ + asoc->my_rwnd = ntohl(initack_cp->init.a_rwnd); + + /* process the INIT info (peer's info) */ + if (sctp_process_init(init_cp, stcb) < 0) { +#if defined(__APPLE__) && !defined(__Userspace__) + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_19); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + return (NULL); + } + /* load all addresses */ + if (sctp_load_addresses_from_init(stcb, m, + init_offset + sizeof(struct sctp_init_chunk), + initack_offset, src, dst, init_src, port) < 0) { +#if defined(__APPLE__) && !defined(__Userspace__) + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_20); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + return (NULL); + } + /* + * verify any preceding AUTH chunk that was skipped + */ + /* pull the local authentication parameters from the cookie/init-ack */ + sctp_auth_get_cookie_params(stcb, m, + initack_offset + sizeof(struct sctp_init_ack_chunk), + initack_limit - (initack_offset + sizeof(struct sctp_init_ack_chunk))); + if (auth_skipped) { + struct sctp_auth_chunk *auth; + + if (auth_len <= SCTP_CHUNK_BUFFER_SIZE) { + auth = (struct sctp_auth_chunk *)sctp_m_getptr(m, auth_offset, auth_len, auth_chunk_buf); + } else { + auth = NULL; + } + if ((auth == NULL) || sctp_handle_auth(stcb, auth, m, auth_offset)) { + /* auth HMAC failed, dump the assoc and packet */ + SCTPDBG(SCTP_DEBUG_AUTH1, + "COOKIE-ECHO: AUTH failed\n"); +#if defined(__APPLE__) && !defined(__Userspace__) + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_21); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + return (NULL); + } else { + /* remaining chunks checked... good to go */ + stcb->asoc.authenticated = 1; + } + } + + /* + * if we're doing ASCONFs, check to see if we have any new local + * addresses that need to get added to the peer (eg. addresses + * changed while cookie echo in flight). This needs to be done + * after we go to the OPEN state to do the correct asconf + * processing. else, make sure we have the correct addresses in our + * lists + */ + + /* warning, we re-use sin, sin6, sa_store here! */ + /* pull in local_address (our "from" address) */ + switch (cookie->laddr_type) { +#ifdef INET + case SCTP_IPV4_ADDRESS: + /* source addr is IPv4 */ + memset(&store.sin, 0, sizeof(struct sockaddr_in)); + store.sin.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + store.sin.sin_len = sizeof(struct sockaddr_in); +#endif + store.sin.sin_addr.s_addr = cookie->laddress[0]; + break; +#endif +#ifdef INET6 + case SCTP_IPV6_ADDRESS: + /* source addr is IPv6 */ + memset(&store.sin6, 0, sizeof(struct sockaddr_in6)); + store.sin6.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + store.sin6.sin6_len = sizeof(struct sockaddr_in6); +#endif + store.sin6.sin6_scope_id = cookie->scope_id; + memcpy(&store.sin6.sin6_addr, cookie->laddress, sizeof(struct in6_addr)); + break; +#endif +#if defined(__Userspace__) + case SCTP_CONN_ADDRESS: + /* source addr is conn */ + memset(&store.sconn, 0, sizeof(struct sockaddr_conn)); + store.sconn.sconn_family = AF_CONN; +#ifdef HAVE_SCONN_LEN + store.sconn.sconn_len = sizeof(struct sockaddr_conn); +#endif + memcpy(&store.sconn.sconn_addr, cookie->laddress, sizeof(void *)); + break; +#endif + default: +#if defined(__APPLE__) && !defined(__Userspace__) + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_22); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + return (NULL); + } + + /* update current state */ + SCTPDBG(SCTP_DEBUG_INPUT2, "moving to OPEN state\n"); + SCTP_SET_STATE(stcb, SCTP_STATE_OPEN); + if (asoc->state & SCTP_STATE_SHUTDOWN_PENDING) { + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, + stcb->sctp_ep, stcb, NULL); + } + sctp_stop_all_cookie_timers(stcb); + SCTP_STAT_INCR_COUNTER32(sctps_passiveestab); + SCTP_STAT_INCR_GAUGE32(sctps_currestab); + + /* set up to notify upper layer */ + *notification = SCTP_NOTIFY_ASSOC_UP; + if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) && + (!SCTP_IS_LISTENING(inp))) { + /* + * This is an endpoint that called connect() how it got a + * cookie that is NEW is a bit of a mystery. It must be that + * the INIT was sent, but before it got there.. a complete + * INIT/INIT-ACK/COOKIE arrived. But of course then it + * should have went to the other code.. not here.. oh well.. + * a bit of protection is worth having.. + * + * XXXMJ unlocked + */ + sctp_pcb_add_flags(stcb->sctp_ep, SCTP_PCB_FLAGS_CONNECTED); +#if defined(__APPLE__) && !defined(__Userspace__) + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { + SCTP_SOCKET_UNLOCK(so, 1); + return (NULL); + } +#endif + soisconnected(stcb->sctp_socket); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } else if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && + (SCTP_IS_LISTENING(inp))) { + /* + * We don't want to do anything with this one. Since it is + * the listening guy. The timer will get started for + * accepted connections in the caller. + */ + ; + } + if (stcb->asoc.sctp_autoclose_ticks && + sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) { + sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb, NULL); + } + (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); + *netp = sctp_findnet(stcb, init_src); + if (*netp != NULL) { + /* + * Since we did not send a HB, make sure we don't double + * things. + */ + (*netp)->hb_responded = 1; + } + /* respond with a COOKIE-ACK */ + sctp_send_cookie_ack(stcb); + + /* + * check the address lists for any ASCONFs that need to be sent + * AFTER the cookie-ack is sent + */ + sctp_check_address_list(stcb, m, + initack_offset + sizeof(struct sctp_init_ack_chunk), + initack_limit - (initack_offset + sizeof(struct sctp_init_ack_chunk)), + &store.sa, cookie->local_scope, cookie->site_scope, + cookie->ipv4_scope, cookie->loopback_scope); + + return (stcb); +} + +/* + * CODE LIKE THIS NEEDS TO RUN IF the peer supports the NAT extension, i.e + * we NEED to make sure we are not already using the vtag. If so we + * need to send back an ABORT-TRY-AGAIN-WITH-NEW-TAG No middle box bit! + head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag, + SCTP_BASE_INFO(hashasocmark))]; + LIST_FOREACH(stcb, head, sctp_asocs) { + if ((stcb->asoc.my_vtag == tag) && (stcb->rport == rport) && (inp == stcb->sctp_ep)) { + -- SEND ABORT - TRY AGAIN -- + } + } +*/ + +/* + * handles a COOKIE-ECHO message stcb: modified to either a new or left as + * existing (non-NULL) TCB + */ +static struct mbuf * +sctp_handle_cookie_echo(struct mbuf *m, int iphlen, int offset, + struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, struct sctp_cookie_echo_chunk *cp, + struct sctp_inpcb **inp_p, struct sctp_tcb **stcb, struct sctp_nets **netp, + int auth_skipped, uint32_t auth_offset, uint32_t auth_len, + struct sctp_tcb **locked_tcb, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, +#endif + uint32_t vrf_id, uint16_t port) +{ + struct sctp_state_cookie *cookie; + struct sctp_tcb *l_stcb = *stcb; + struct sctp_inpcb *l_inp; + struct sockaddr *to; + struct sctp_pcb *ep; + struct mbuf *m_sig; + uint8_t calc_sig[SCTP_SIGNATURE_SIZE], tmp_sig[SCTP_SIGNATURE_SIZE]; + uint8_t *sig; +#if defined(__Userspace__) && defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) + uint8_t cookie_ok = 1; +#else + uint8_t cookie_ok = 0; +#endif + unsigned int sig_offset, cookie_offset; + unsigned int cookie_len; + struct timeval now; + struct timeval time_entered, time_expires; + int notification = 0; + struct sctp_nets *netl; + int had_a_existing_tcb = 0; + int send_int_conf = 0; +#ifdef INET + struct sockaddr_in sin; +#endif +#ifdef INET6 + struct sockaddr_in6 sin6; +#endif +#if defined(__Userspace__) + struct sockaddr_conn sconn; +#endif + + SCTPDBG(SCTP_DEBUG_INPUT2, + "sctp_handle_cookie: handling COOKIE-ECHO\n"); + + if (inp_p == NULL) { + return (NULL); + } + cookie = &cp->cookie; + cookie_offset = offset + sizeof(struct sctp_chunkhdr); + cookie_len = ntohs(cp->ch.chunk_length); + + if (cookie_len < sizeof(struct sctp_cookie_echo_chunk) + + sizeof(struct sctp_init_chunk) + + sizeof(struct sctp_init_ack_chunk) + SCTP_SIGNATURE_SIZE) { + /* cookie too small */ + return (NULL); + } + if ((cookie->peerport != sh->src_port) || + (cookie->myport != sh->dest_port) || + (cookie->my_vtag != sh->v_tag)) { + /* + * invalid ports or bad tag. Note that we always leave the + * v_tag in the header in network order and when we stored + * it in the my_vtag slot we also left it in network order. + * This maintains the match even though it may be in the + * opposite byte order of the machine :-> + */ + return (NULL); + } +#if defined(__Userspace__) + /* + * Recover the AF_CONN addresses within the cookie. + * This needs to be done in the buffer provided for later processing + * of the cookie and in the mbuf chain for HMAC validation. + */ + if ((cookie->addr_type == SCTP_CONN_ADDRESS) && (src->sa_family == AF_CONN)) { + struct sockaddr_conn *sconnp = (struct sockaddr_conn *)src; + + memcpy(cookie->address, &sconnp->sconn_addr , sizeof(void *)); + m_copyback(m, cookie_offset + offsetof(struct sctp_state_cookie, address), + (int)sizeof(void *), (caddr_t)&sconnp->sconn_addr); + } + if ((cookie->laddr_type == SCTP_CONN_ADDRESS) && (dst->sa_family == AF_CONN)) { + struct sockaddr_conn *sconnp = (struct sockaddr_conn *)dst; + + memcpy(cookie->laddress, &sconnp->sconn_addr , sizeof(void *)); + m_copyback(m, cookie_offset + offsetof(struct sctp_state_cookie, laddress), + (int)sizeof(void *), (caddr_t)&sconnp->sconn_addr); + } +#endif + /* + * split off the signature into its own mbuf (since it should not be + * calculated in the sctp_hmac_m() call). + */ + sig_offset = offset + cookie_len - SCTP_SIGNATURE_SIZE; + m_sig = m_split(m, sig_offset, M_NOWAIT); + if (m_sig == NULL) { + /* out of memory or ?? */ + return (NULL); + } +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(m_sig, SCTP_MBUF_SPLIT); + } +#endif + + /* + * compute the signature/digest for the cookie + */ + if (l_stcb != NULL) { + atomic_add_int(&l_stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(l_stcb); + } + l_inp = *inp_p; + SCTP_INP_RLOCK(l_inp); + if (l_stcb != NULL) { + SCTP_TCB_LOCK(l_stcb); + atomic_subtract_int(&l_stcb->asoc.refcnt, 1); + } + if (l_inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { + SCTP_INP_RUNLOCK(l_inp); + sctp_m_freem(m_sig); + return (NULL); + } + ep = &(*inp_p)->sctp_ep; + /* which cookie is it? */ + if ((cookie->time_entered.tv_sec < (long)ep->time_of_secret_change) && + (ep->current_secret_number != ep->last_secret_number)) { + /* it's the old cookie */ + (void)sctp_hmac_m(SCTP_HMAC, + (uint8_t *)ep->secret_key[(int)ep->last_secret_number], + SCTP_SECRET_SIZE, m, cookie_offset, calc_sig, 0); + } else { + /* it's the current cookie */ + (void)sctp_hmac_m(SCTP_HMAC, + (uint8_t *)ep->secret_key[(int)ep->current_secret_number], + SCTP_SECRET_SIZE, m, cookie_offset, calc_sig, 0); + } + /* get the signature */ + SCTP_INP_RUNLOCK(l_inp); + sig = (uint8_t *) sctp_m_getptr(m_sig, 0, SCTP_SIGNATURE_SIZE, (uint8_t *) & tmp_sig); + if (sig == NULL) { + /* couldn't find signature */ + sctp_m_freem(m_sig); + return (NULL); + } + /* compare the received digest with the computed digest */ + if (timingsafe_bcmp(calc_sig, sig, SCTP_SIGNATURE_SIZE) != 0) { + /* try the old cookie? */ + if ((cookie->time_entered.tv_sec == (long)ep->time_of_secret_change) && + (ep->current_secret_number != ep->last_secret_number)) { + /* compute digest with old */ + (void)sctp_hmac_m(SCTP_HMAC, + (uint8_t *)ep->secret_key[(int)ep->last_secret_number], + SCTP_SECRET_SIZE, m, cookie_offset, calc_sig, 0); + /* compare */ + if (timingsafe_bcmp(calc_sig, sig, SCTP_SIGNATURE_SIZE) == 0) + cookie_ok = 1; + } + } else { + cookie_ok = 1; + } + + /* + * Now before we continue we must reconstruct our mbuf so that + * normal processing of any other chunks will work. + */ + { + struct mbuf *m_at; + + m_at = m; + while (SCTP_BUF_NEXT(m_at) != NULL) { + m_at = SCTP_BUF_NEXT(m_at); + } + SCTP_BUF_NEXT(m_at) = m_sig; + } + + if (cookie_ok == 0) { + SCTPDBG(SCTP_DEBUG_INPUT2, "handle_cookie_echo: cookie signature validation failed!\n"); + SCTPDBG(SCTP_DEBUG_INPUT2, + "offset = %u, cookie_offset = %u, sig_offset = %u\n", + (uint32_t) offset, cookie_offset, sig_offset); + return (NULL); + } + + if (sctp_ticks_to_msecs(cookie->cookie_life) > SCTP_MAX_COOKIE_LIFE) { + SCTPDBG(SCTP_DEBUG_INPUT2, "handle_cookie_echo: Invalid cookie lifetime\n"); + return (NULL); + } + time_entered.tv_sec = cookie->time_entered.tv_sec; + time_entered.tv_usec = cookie->time_entered.tv_usec; + if ((time_entered.tv_sec < 0) || + (time_entered.tv_usec < 0) || + (time_entered.tv_usec >= 1000000)) { + /* Invalid time stamp. Cookie must have been modified. */ + SCTPDBG(SCTP_DEBUG_INPUT2, "handle_cookie_echo: Invalid time stamp\n"); + return (NULL); + } + (void)SCTP_GETTIME_TIMEVAL(&now); +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + if (timercmp(&now, &time_entered, <)) { +#else + if (timevalcmp(&now, &time_entered, <)) { +#endif + SCTPDBG(SCTP_DEBUG_INPUT2, "handle_cookie_echo: cookie generated in the future!\n"); + return (NULL); + } + /* + * Check the cookie timestamps to be sure it's not stale. + * cookie_life is in ticks, so we convert to seconds. + */ + time_expires.tv_sec = time_entered.tv_sec + sctp_ticks_to_secs(cookie->cookie_life); + time_expires.tv_usec = time_entered.tv_usec; +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + if (timercmp(&now, &time_expires, >)) +#else + if (timevalcmp(&now, &time_expires, >)) +#endif + { + /* cookie is stale! */ + struct mbuf *op_err; + struct sctp_error_stale_cookie *cause; + struct timeval diff; + uint32_t staleness; + + op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_error_stale_cookie), + 0, M_NOWAIT, 1, MT_DATA); + if (op_err == NULL) { + /* FOOBAR */ + return (NULL); + } + /* Set the len */ + SCTP_BUF_LEN(op_err) = sizeof(struct sctp_error_stale_cookie); + cause = mtod(op_err, struct sctp_error_stale_cookie *); + cause->cause.code = htons(SCTP_CAUSE_STALE_COOKIE); + cause->cause.length = htons(sizeof(struct sctp_error_stale_cookie)); +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + timersub(&now, &time_expires, &diff); +#else + diff = now; + timevalsub(&diff, &time_expires); +#endif + if ((uint32_t)diff.tv_sec > UINT32_MAX / 1000000) { + staleness = UINT32_MAX; + } else { + staleness = (uint32_t)diff.tv_sec * 1000000; + } + if (UINT32_MAX - staleness >= (uint32_t)diff.tv_usec) { + staleness += (uint32_t)diff.tv_usec; + } else { + staleness = UINT32_MAX; + } + cause->stale_time = htonl(staleness); + sctp_send_operr_to(src, dst, sh, cookie->peers_vtag, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, l_inp->fibnum, +#endif + vrf_id, port); + return (NULL); + } + /* + * Now we must see with the lookup address if we have an existing + * asoc. This will only happen if we were in the COOKIE-WAIT state + * and a INIT collided with us and somewhere the peer sent the + * cookie on another address besides the single address our assoc + * had for him. In this case we will have one of the tie-tags set at + * least AND the address field in the cookie can be used to look it + * up. + */ + to = NULL; + switch (cookie->addr_type) { +#ifdef INET6 + case SCTP_IPV6_ADDRESS: + memset(&sin6, 0, sizeof(sin6)); + sin6.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6.sin6_len = sizeof(sin6); +#endif + sin6.sin6_port = sh->src_port; + sin6.sin6_scope_id = cookie->scope_id; + memcpy(&sin6.sin6_addr.s6_addr, cookie->address, + sizeof(sin6.sin6_addr.s6_addr)); + to = (struct sockaddr *)&sin6; + break; +#endif +#ifdef INET + case SCTP_IPV4_ADDRESS: + memset(&sin, 0, sizeof(sin)); + sin.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + sin.sin_len = sizeof(sin); +#endif + sin.sin_port = sh->src_port; + sin.sin_addr.s_addr = cookie->address[0]; + to = (struct sockaddr *)&sin; + break; +#endif +#if defined(__Userspace__) + case SCTP_CONN_ADDRESS: + memset(&sconn, 0, sizeof(struct sockaddr_conn)); + sconn.sconn_family = AF_CONN; +#ifdef HAVE_SCONN_LEN + sconn.sconn_len = sizeof(struct sockaddr_conn); +#endif + sconn.sconn_port = sh->src_port; + memcpy(&sconn.sconn_addr, cookie->address, sizeof(void *)); + to = (struct sockaddr *)&sconn; + break; +#endif + default: + /* This should not happen */ + return (NULL); + } + if (*stcb == NULL) { + /* Yep, lets check */ + *stcb = sctp_findassociation_ep_addr(inp_p, to, netp, dst, NULL); + if (*stcb == NULL) { + /* + * We should have only got back the same inp. If we + * got back a different ep we have a problem. The + * original findep got back l_inp and now + */ + if (l_inp != *inp_p) { + SCTP_PRINTF("Bad problem find_ep got a diff inp then special_locate?\n"); + } + } else { + if (*locked_tcb == NULL) { + /* In this case we found the assoc only + * after we locked the create lock. This means + * we are in a colliding case and we must make + * sure that we unlock the tcb if its one of the + * cases where we throw away the incoming packets. + */ + *locked_tcb = *stcb; + + /* We must also increment the inp ref count + * since the ref_count flags was set when we + * did not find the TCB, now we found it which + * reduces the refcount.. we must raise it back + * out to balance it all :-) + */ + SCTP_INP_INCR_REF((*stcb)->sctp_ep); + if ((*stcb)->sctp_ep != l_inp) { + SCTP_PRINTF("Huh? ep:%p diff then l_inp:%p?\n", + (void *)(*stcb)->sctp_ep, (void *)l_inp); + } + } + } + } + + cookie_len -= SCTP_SIGNATURE_SIZE; + if (*stcb == NULL) { + /* this is the "normal" case... get a new TCB */ + *stcb = sctp_process_cookie_new(m, iphlen, offset, src, dst, sh, + cookie, cookie_len, *inp_p, + netp, to, ¬ification, + auth_skipped, auth_offset, auth_len, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, port); + } else { + /* this is abnormal... cookie-echo on existing TCB */ + had_a_existing_tcb = 1; + *stcb = sctp_process_cookie_existing(m, iphlen, offset, + src, dst, sh, + cookie, cookie_len, *inp_p, *stcb, netp, to, + ¬ification, auth_skipped, auth_offset, auth_len, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, port); + if (*stcb == NULL) { + *locked_tcb = NULL; + } + } + + if (*stcb == NULL) { + /* still no TCB... must be bad cookie-echo */ + return (NULL); + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (*netp != NULL) { + (*netp)->flowtype = mflowtype; + (*netp)->flowid = mflowid; + } +#endif + /* + * Ok, we built an association so confirm the address we sent the + * INIT-ACK to. + */ + netl = sctp_findnet(*stcb, to); + /* + * This code should in theory NOT run but + */ + if (netl == NULL) { + /* TSNH! Huh, why do I need to add this address here? */ + if (sctp_add_remote_addr(*stcb, to, NULL, port, + SCTP_DONOT_SETSCOPE, SCTP_IN_COOKIE_PROC)) { + return (NULL); + } + netl = sctp_findnet(*stcb, to); + } + if (netl) { + if (netl->dest_state & SCTP_ADDR_UNCONFIRMED) { + netl->dest_state &= ~SCTP_ADDR_UNCONFIRMED; + (void)sctp_set_primary_addr((*stcb), (struct sockaddr *)NULL, + netl); + send_int_conf = 1; + } + } + sctp_start_net_timers(*stcb); + if ((*inp_p)->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) { + if (!had_a_existing_tcb || + (((*inp_p)->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0)) { + /* + * If we have a NEW cookie or the connect never + * reached the connected state during collision we + * must do the TCP accept thing. + */ + struct socket *so, *oso; + struct sctp_inpcb *inp; + + if (notification == SCTP_NOTIFY_ASSOC_RESTART) { + /* + * For a restart we will keep the same + * socket, no need to do anything. I THINK!! + */ + sctp_ulp_notify(notification, *stcb, 0, NULL, SCTP_SO_NOT_LOCKED); + if (send_int_conf) { + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_CONFIRMED, + (*stcb), 0, (void *)netl, SCTP_SO_NOT_LOCKED); + } + return (m); + } + oso = (*inp_p)->sctp_socket; + atomic_add_int(&(*stcb)->asoc.refcnt, 1); + SCTP_TCB_UNLOCK((*stcb)); +#if defined(__FreeBSD__) && !defined(__Userspace__) + CURVNET_SET(oso->so_vnet); +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_LOCK(oso, 1); +#endif + so = sonewconn(oso, 0 +#if defined(__APPLE__) && !defined(__Userspace__) + ,NULL +#endif + ); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(oso, 1); +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + CURVNET_RESTORE(); +#endif + SCTP_TCB_LOCK((*stcb)); + atomic_subtract_int(&(*stcb)->asoc.refcnt, 1); + + if (so == NULL) { + struct mbuf *op_err; +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *pcb_so; +#endif + /* Too many sockets */ + SCTPDBG(SCTP_DEBUG_INPUT1, "process_cookie_new: no room for another socket!\n"); + op_err = sctp_generate_cause(SCTP_CAUSE_OUT_OF_RESC, ""); + sctp_abort_association(*inp_p, NULL, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, port); +#if defined(__APPLE__) && !defined(__Userspace__) + pcb_so = SCTP_INP_SO(*inp_p); + atomic_add_int(&(*stcb)->asoc.refcnt, 1); + SCTP_TCB_UNLOCK((*stcb)); + SCTP_SOCKET_LOCK(pcb_so, 1); + SCTP_TCB_LOCK((*stcb)); + atomic_subtract_int(&(*stcb)->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(*inp_p, *stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_23); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(pcb_so, 1); +#endif + return (NULL); + } + inp = (struct sctp_inpcb *)so->so_pcb; + SCTP_INP_INCR_REF(inp); + /* + * We add the unbound flag here so that + * if we get an soabort() before we get the + * move_pcb done, we will properly cleanup. + */ + inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE | + SCTP_PCB_FLAGS_CONNECTED | + SCTP_PCB_FLAGS_IN_TCPPOOL | + SCTP_PCB_FLAGS_UNBOUND | + (SCTP_PCB_COPY_FLAGS & (*inp_p)->sctp_flags) | + SCTP_PCB_FLAGS_DONT_WAKE); + inp->sctp_features = (*inp_p)->sctp_features; + inp->sctp_mobility_features = (*inp_p)->sctp_mobility_features; + inp->sctp_socket = so; + inp->sctp_frag_point = (*inp_p)->sctp_frag_point; + inp->max_cwnd = (*inp_p)->max_cwnd; + inp->sctp_cmt_on_off = (*inp_p)->sctp_cmt_on_off; + inp->ecn_supported = (*inp_p)->ecn_supported; + inp->prsctp_supported = (*inp_p)->prsctp_supported; + inp->auth_supported = (*inp_p)->auth_supported; + inp->asconf_supported = (*inp_p)->asconf_supported; + inp->reconfig_supported = (*inp_p)->reconfig_supported; + inp->nrsack_supported = (*inp_p)->nrsack_supported; + inp->pktdrop_supported = (*inp_p)->pktdrop_supported; + inp->partial_delivery_point = (*inp_p)->partial_delivery_point; + inp->sctp_context = (*inp_p)->sctp_context; + inp->local_strreset_support = (*inp_p)->local_strreset_support; + inp->fibnum = (*inp_p)->fibnum; +#if defined(__Userspace__) + inp->ulp_info = (*inp_p)->ulp_info; + inp->recv_callback = (*inp_p)->recv_callback; + inp->send_callback = (*inp_p)->send_callback; + inp->send_sb_threshold = (*inp_p)->send_sb_threshold; +#endif + /* + * copy in the authentication parameters from the + * original endpoint + */ + if (inp->sctp_ep.local_hmacs) + sctp_free_hmaclist(inp->sctp_ep.local_hmacs); + inp->sctp_ep.local_hmacs = + sctp_copy_hmaclist((*inp_p)->sctp_ep.local_hmacs); + if (inp->sctp_ep.local_auth_chunks) + sctp_free_chunklist(inp->sctp_ep.local_auth_chunks); + inp->sctp_ep.local_auth_chunks = + sctp_copy_chunklist((*inp_p)->sctp_ep.local_auth_chunks); + + /* + * Now we must move it from one hash table to + * another and get the tcb in the right place. + */ + + /* This is where the one-2-one socket is put into + * the accept state waiting for the accept! + */ + if (*stcb) { + SCTP_ADD_SUBSTATE(*stcb, SCTP_STATE_IN_ACCEPT_QUEUE); + } + sctp_move_pcb_and_assoc(*inp_p, inp, *stcb); + + atomic_add_int(&(*stcb)->asoc.refcnt, 1); + SCTP_TCB_UNLOCK((*stcb)); + +#if defined(__FreeBSD__) && !defined(__Userspace__) + sctp_pull_off_control_to_new_inp((*inp_p), inp, *stcb, + 0); +#else + sctp_pull_off_control_to_new_inp((*inp_p), inp, *stcb, M_NOWAIT); +#endif + SCTP_TCB_LOCK((*stcb)); + atomic_subtract_int(&(*stcb)->asoc.refcnt, 1); + + /* now we must check to see if we were aborted while + * the move was going on and the lock/unlock happened. + */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + /* yep it was, we leave the + * assoc attached to the socket since + * the sctp_inpcb_free() call will send + * an abort for us. + */ + SCTP_INP_DECR_REF(inp); + return (NULL); + } + SCTP_INP_DECR_REF(inp); + /* Switch over to the new guy */ + *inp_p = inp; + sctp_ulp_notify(notification, *stcb, 0, NULL, SCTP_SO_NOT_LOCKED); + if (send_int_conf) { + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_CONFIRMED, + (*stcb), 0, (void *)netl, SCTP_SO_NOT_LOCKED); + } + + /* Pull it from the incomplete queue and wake the guy */ +#if defined(__APPLE__) && !defined(__Userspace__) + atomic_add_int(&(*stcb)->asoc.refcnt, 1); + SCTP_TCB_UNLOCK((*stcb)); + SCTP_SOCKET_LOCK(so, 1); +#endif + soisconnected(so); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_TCB_LOCK((*stcb)); + atomic_subtract_int(&(*stcb)->asoc.refcnt, 1); + SCTP_SOCKET_UNLOCK(so, 1); +#endif + return (m); + } + } + if (notification) { + sctp_ulp_notify(notification, *stcb, 0, NULL, SCTP_SO_NOT_LOCKED); + } + if (send_int_conf) { + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_CONFIRMED, + (*stcb), 0, (void *)netl, SCTP_SO_NOT_LOCKED); + } + return (m); +} + +static void +sctp_handle_cookie_ack(struct sctp_cookie_ack_chunk *cp SCTP_UNUSED, + struct sctp_tcb *stcb, struct sctp_nets *net) +{ + /* cp must not be used, others call this without a c-ack :-) */ + struct sctp_association *asoc; + struct sctp_tmit_chunk *chk; + + SCTPDBG(SCTP_DEBUG_INPUT2, + "sctp_handle_cookie_ack: handling COOKIE-ACK\n"); + if ((stcb == NULL) || (net == NULL)) { + return; + } + + asoc = &stcb->asoc; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { + sctp_misc_ints(SCTP_THRESHOLD_CLEAR, + asoc->overall_error_count, + 0, + SCTP_FROM_SCTP_INPUT, + __LINE__); + } + sctp_stop_all_cookie_timers(stcb); + /* process according to association state */ + if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED) { + /* state change only needed when I am in right state */ + SCTPDBG(SCTP_DEBUG_INPUT2, "moving to OPEN state\n"); + SCTP_SET_STATE(stcb, SCTP_STATE_OPEN); + sctp_start_net_timers(stcb); + if (asoc->state & SCTP_STATE_SHUTDOWN_PENDING) { + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, + stcb->sctp_ep, stcb, NULL); + } + /* update RTO */ + SCTP_STAT_INCR_COUNTER32(sctps_activeestab); + SCTP_STAT_INCR_GAUGE32(sctps_currestab); + if (asoc->overall_error_count == 0) { + sctp_calculate_rto(stcb, asoc, net, &asoc->time_entered, + SCTP_RTT_FROM_NON_DATA); + } + /* + * Since we did not send a HB make sure we don't double + * things. + */ + asoc->overall_error_count = 0; + net->hb_responded = 1; + (void)SCTP_GETTIME_TIMEVAL(&asoc->time_entered); + sctp_ulp_notify(SCTP_NOTIFY_ASSOC_UP, stcb, 0, NULL, SCTP_SO_NOT_LOCKED); + if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; + +#endif + sctp_pcb_add_flags(stcb->sctp_ep, SCTP_PCB_FLAGS_CONNECTED); +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(stcb->sctp_ep); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + if ((stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) == 0) { + soisconnected(stcb->sctp_socket); + } +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } + + if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { + /* We don't need to do the asconf thing, + * nor hb or autoclose if the socket is closed. + */ + goto closed_socket; + } + + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, + stcb, net); + + if (stcb->asoc.sctp_autoclose_ticks && + sctp_is_feature_on(stcb->sctp_ep, SCTP_PCB_FLAGS_AUTOCLOSE)) { + sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, + stcb->sctp_ep, stcb, NULL); + } + /* + * send ASCONF if parameters are pending and ASCONFs are + * allowed (eg. addresses changed when init/cookie echo were + * in flight) + */ + if ((sctp_is_feature_on(stcb->sctp_ep, SCTP_PCB_FLAGS_DO_ASCONF)) && + (stcb->asoc.asconf_supported == 1) && + (!TAILQ_EMPTY(&stcb->asoc.asconf_queue))) { +#ifdef SCTP_TIMER_BASED_ASCONF + sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, + stcb->sctp_ep, stcb, + stcb->asoc.primary_destination); +#else + sctp_send_asconf(stcb, stcb->asoc.primary_destination, + SCTP_ADDR_NOT_LOCKED); +#endif + } + } +closed_socket: + /* Toss the cookie if I can */ + sctp_toss_old_cookies(stcb, asoc); + /* Restart the timer if we have pending data */ + TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { + if (chk->whoTo != NULL) { + break; + } + } + if (chk != NULL) { + sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, chk->whoTo); + } +} + +static void +sctp_handle_ecn_echo(struct sctp_ecne_chunk *cp, + struct sctp_tcb *stcb) +{ + struct sctp_nets *net; + struct sctp_tmit_chunk *lchk; + struct sctp_ecne_chunk bkup; + uint8_t override_bit; + uint32_t tsn, window_data_tsn; + int len; + unsigned int pkt_cnt; + + len = ntohs(cp->ch.chunk_length); + if (len == sizeof(struct old_sctp_ecne_chunk)) { + /* Its the old format */ + memcpy(&bkup, cp, sizeof(struct old_sctp_ecne_chunk)); + bkup.num_pkts_since_cwr = htonl(1); + cp = &bkup; + } + SCTP_STAT_INCR(sctps_recvecne); + tsn = ntohl(cp->tsn); + pkt_cnt = ntohl(cp->num_pkts_since_cwr); + lchk = TAILQ_LAST(&stcb->asoc.send_queue, sctpchunk_listhead); + if (lchk == NULL) { + window_data_tsn = stcb->asoc.sending_seq - 1; + } else { + window_data_tsn = lchk->rec.data.tsn; + } + + /* Find where it was sent to if possible. */ + net = NULL; + TAILQ_FOREACH(lchk, &stcb->asoc.sent_queue, sctp_next) { + if (lchk->rec.data.tsn == tsn) { + net = lchk->whoTo; + net->ecn_prev_cwnd = lchk->rec.data.cwnd_at_send; + break; + } + if (SCTP_TSN_GT(lchk->rec.data.tsn, tsn)) { + break; + } + } + if (net == NULL) { + /* + * What to do. A previous send of a + * CWR was possibly lost. See how old it is, we + * may have it marked on the actual net. + */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if (tsn == net->last_cwr_tsn) { + /* Found him, send it off */ + break; + } + } + if (net == NULL) { + /* + * If we reach here, we need to send a special + * CWR that says hey, we did this a long time + * ago and you lost the response. + */ + net = TAILQ_FIRST(&stcb->asoc.nets); + if (net == NULL) { + /* TSNH */ + return; + } + override_bit = SCTP_CWR_REDUCE_OVERRIDE; + } else { + override_bit = 0; + } + } else { + override_bit = 0; + } + if (SCTP_TSN_GT(tsn, net->cwr_window_tsn) && + ((override_bit&SCTP_CWR_REDUCE_OVERRIDE) == 0)) { + /* JRS - Use the congestion control given in the pluggable CC module */ + stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo(stcb, net, 0, pkt_cnt); + /* + * We reduce once every RTT. So we will only lower cwnd at + * the next sending seq i.e. the window_data_tsn + */ + net->cwr_window_tsn = window_data_tsn; + net->ecn_ce_pkt_cnt += pkt_cnt; + net->lost_cnt = pkt_cnt; + net->last_cwr_tsn = tsn; + } else { + override_bit |= SCTP_CWR_IN_SAME_WINDOW; + if (SCTP_TSN_GT(tsn, net->last_cwr_tsn) && + ((override_bit&SCTP_CWR_REDUCE_OVERRIDE) == 0)) { + /* + * Another loss in the same window update how + * many marks/packets lost we have had. + */ + int cnt = 1; + if (pkt_cnt > net->lost_cnt) { + /* Should be the case */ + cnt = (pkt_cnt - net->lost_cnt); + net->ecn_ce_pkt_cnt += cnt; + } + net->lost_cnt = pkt_cnt; + net->last_cwr_tsn = tsn; + /* + * Most CC functions will ignore this call, since we are in-window + * yet of the initial CE the peer saw. + */ + stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo(stcb, net, 1, cnt); + } + } + /* + * We always send a CWR this way if our previous one was lost our + * peer will get an update, or if it is not time again to reduce we + * still get the cwr to the peer. Note we set the override when we + * could not find the TSN on the chunk or the destination network. + */ + sctp_send_cwr(stcb, net, net->last_cwr_tsn, override_bit); +} + +static void +sctp_handle_ecn_cwr(struct sctp_cwr_chunk *cp, struct sctp_tcb *stcb, struct sctp_nets *net) +{ + /* + * Here we get a CWR from the peer. We must look in the outqueue and + * make sure that we have a covered ECNE in the control chunk part. + * If so remove it. + */ + struct sctp_tmit_chunk *chk, *nchk; + struct sctp_ecne_chunk *ecne; + int override; + uint32_t cwr_tsn; + + cwr_tsn = ntohl(cp->tsn); + override = cp->ch.chunk_flags & SCTP_CWR_REDUCE_OVERRIDE; + TAILQ_FOREACH_SAFE(chk, &stcb->asoc.control_send_queue, sctp_next, nchk) { + if (chk->rec.chunk_id.id != SCTP_ECN_ECHO) { + continue; + } + if ((override == 0) && (chk->whoTo != net)) { + /* Must be from the right src unless override is set */ + continue; + } + ecne = mtod(chk->data, struct sctp_ecne_chunk *); + if (SCTP_TSN_GE(cwr_tsn, ntohl(ecne->tsn))) { + /* this covers this ECNE, we can remove it */ + stcb->asoc.ecn_echo_cnt_onq--; + TAILQ_REMOVE(&stcb->asoc.control_send_queue, chk, + sctp_next); + stcb->asoc.ctrl_queue_cnt--; + sctp_m_freem(chk->data); + chk->data = NULL; + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + if (override == 0) { + break; + } + } + } +} + +static void +sctp_handle_shutdown_complete(struct sctp_shutdown_complete_chunk *cp SCTP_UNUSED, + struct sctp_tcb *stcb, struct sctp_nets *net) +{ +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif + + SCTPDBG(SCTP_DEBUG_INPUT2, + "sctp_handle_shutdown_complete: handling SHUTDOWN-COMPLETE\n"); + if (stcb == NULL) + return; + + /* process according to association state */ + if (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_ACK_SENT) { + /* unexpected SHUTDOWN-COMPLETE... so ignore... */ + SCTPDBG(SCTP_DEBUG_INPUT2, + "sctp_handle_shutdown_complete: not in SCTP_STATE_SHUTDOWN_ACK_SENT --- ignore\n"); + SCTP_TCB_UNLOCK(stcb); + return; + } + /* notify upper layer protocol */ + if (stcb->sctp_socket) { + sctp_ulp_notify(SCTP_NOTIFY_ASSOC_DOWN, stcb, 0, NULL, SCTP_SO_NOT_LOCKED); + } +#ifdef INVARIANTS + if (!TAILQ_EMPTY(&stcb->asoc.send_queue) || + !TAILQ_EMPTY(&stcb->asoc.sent_queue) || + sctp_is_there_unsent_data(stcb, SCTP_SO_NOT_LOCKED)) { + panic("Queues are not empty when handling SHUTDOWN-COMPLETE"); + } +#endif + /* stop the timer */ + sctp_timer_stop(SCTP_TIMER_TYPE_SHUTDOWNACK, stcb->sctp_ep, stcb, net, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_24); + SCTP_STAT_INCR_COUNTER32(sctps_shutdown); + /* free the TCB */ + SCTPDBG(SCTP_DEBUG_INPUT2, + "sctp_handle_shutdown_complete: calls free-asoc\n"); +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(stcb->sctp_ep); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(stcb->sctp_ep, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_25); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + return; +} + +static int +process_chunk_drop(struct sctp_tcb *stcb, struct sctp_chunk_desc *desc, + struct sctp_nets *net, uint8_t flg) +{ + switch (desc->chunk_type) { + case SCTP_DATA: + case SCTP_IDATA: + /* find the tsn to resend (possibly) */ + { + uint32_t tsn; + struct sctp_tmit_chunk *tp1; + + tsn = ntohl(desc->tsn_ifany); + TAILQ_FOREACH(tp1, &stcb->asoc.sent_queue, sctp_next) { + if (tp1->rec.data.tsn == tsn) { + /* found it */ + break; + } + if (SCTP_TSN_GT(tp1->rec.data.tsn, tsn)) { + /* not found */ + tp1 = NULL; + break; + } + } + if (tp1 == NULL) { + /* + * Do it the other way , aka without paying + * attention to queue seq order. + */ + SCTP_STAT_INCR(sctps_pdrpdnfnd); + TAILQ_FOREACH(tp1, &stcb->asoc.sent_queue, sctp_next) { + if (tp1->rec.data.tsn == tsn) { + /* found it */ + break; + } + } + } + if (tp1 == NULL) { + SCTP_STAT_INCR(sctps_pdrptsnnf); + } + if ((tp1) && (tp1->sent < SCTP_DATAGRAM_ACKED)) { + if (((flg & SCTP_BADCRC) == 0) && + ((flg & SCTP_FROM_MIDDLE_BOX) == 0)) { + return (0); + } + if ((stcb->asoc.peers_rwnd == 0) && + ((flg & SCTP_FROM_MIDDLE_BOX) == 0)) { + SCTP_STAT_INCR(sctps_pdrpdiwnp); + return (0); + } + if (stcb->asoc.peers_rwnd == 0 && + (flg & SCTP_FROM_MIDDLE_BOX)) { + SCTP_STAT_INCR(sctps_pdrpdizrw); + return (0); + } + if ((uint32_t)SCTP_BUF_LEN(tp1->data) < + SCTP_DATA_CHUNK_OVERHEAD(stcb) + SCTP_NUM_DB_TO_VERIFY) { + /* Payload not matching. */ + SCTP_STAT_INCR(sctps_pdrpbadd); + return (-1); + } + if (memcmp(mtod(tp1->data, caddr_t) + SCTP_DATA_CHUNK_OVERHEAD(stcb), + desc->data_bytes, SCTP_NUM_DB_TO_VERIFY) != 0) { + /* Payload not matching. */ + SCTP_STAT_INCR(sctps_pdrpbadd); + return (-1); + } + if (tp1->do_rtt) { + /* + * this guy had a RTO calculation + * pending on it, cancel it + */ + if (tp1->whoTo->rto_needed == 0) { + tp1->whoTo->rto_needed = 1; + } + tp1->do_rtt = 0; + } + SCTP_STAT_INCR(sctps_pdrpmark); + if (tp1->sent != SCTP_DATAGRAM_RESEND) + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + /* + * mark it as if we were doing a FR, since + * we will be getting gap ack reports behind + * the info from the router. + */ + tp1->rec.data.doing_fast_retransmit = 1; + /* + * mark the tsn with what sequences can + * cause a new FR. + */ + if (TAILQ_EMPTY(&stcb->asoc.send_queue)) { + tp1->rec.data.fast_retran_tsn = stcb->asoc.sending_seq; + } else { + tp1->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.tsn; + } + + /* restart the timer */ + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, + stcb, tp1->whoTo, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_26); + sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, + stcb, tp1->whoTo); + + /* fix counts and things */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_PDRP, + tp1->whoTo->flight_size, + tp1->book_size, + (uint32_t)(uintptr_t)stcb, + tp1->rec.data.tsn); + } + if (tp1->sent < SCTP_DATAGRAM_RESEND) { + sctp_flight_size_decrease(tp1); + sctp_total_flight_decrease(stcb, tp1); + } + tp1->sent = SCTP_DATAGRAM_RESEND; + } { + /* audit code */ + unsigned int audit; + + audit = 0; + TAILQ_FOREACH(tp1, &stcb->asoc.sent_queue, sctp_next) { + if (tp1->sent == SCTP_DATAGRAM_RESEND) + audit++; + } + TAILQ_FOREACH(tp1, &stcb->asoc.control_send_queue, + sctp_next) { + if (tp1->sent == SCTP_DATAGRAM_RESEND) + audit++; + } + if (audit != stcb->asoc.sent_queue_retran_cnt) { + SCTP_PRINTF("**Local Audit finds cnt:%d asoc cnt:%d\n", + audit, stcb->asoc.sent_queue_retran_cnt); +#ifndef SCTP_AUDITING_ENABLED + stcb->asoc.sent_queue_retran_cnt = audit; +#endif + } + } + } + break; + case SCTP_ASCONF: + { + struct sctp_tmit_chunk *asconf; + + TAILQ_FOREACH(asconf, &stcb->asoc.control_send_queue, + sctp_next) { + if (asconf->rec.chunk_id.id == SCTP_ASCONF) { + break; + } + } + if (asconf) { + if (asconf->sent != SCTP_DATAGRAM_RESEND) + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + asconf->sent = SCTP_DATAGRAM_RESEND; + asconf->snd_count--; + } + } + break; + case SCTP_INITIATION: + /* resend the INIT */ + stcb->asoc.dropped_special_cnt++; + if (stcb->asoc.dropped_special_cnt < SCTP_RETRY_DROPPED_THRESH) { + /* + * If we can get it in, in a few attempts we do + * this, otherwise we let the timer fire. + */ + sctp_timer_stop(SCTP_TIMER_TYPE_INIT, stcb->sctp_ep, + stcb, net, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_27); + sctp_send_initiate(stcb->sctp_ep, stcb, SCTP_SO_NOT_LOCKED); + } + break; + case SCTP_SELECTIVE_ACK: + case SCTP_NR_SELECTIVE_ACK: + /* resend the sack */ + sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED); + break; + case SCTP_HEARTBEAT_REQUEST: + /* resend a demand HB */ + if ((stcb->asoc.overall_error_count + 3) < stcb->asoc.max_send_times) { + /* Only retransmit if we KNOW we wont destroy the tcb */ + sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED); + } + break; + case SCTP_SHUTDOWN: + sctp_send_shutdown(stcb, net); + break; + case SCTP_SHUTDOWN_ACK: + sctp_send_shutdown_ack(stcb, net); + break; + case SCTP_COOKIE_ECHO: + { + struct sctp_tmit_chunk *cookie; + + cookie = NULL; + TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, + sctp_next) { + if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) { + break; + } + } + if (cookie) { + if (cookie->sent != SCTP_DATAGRAM_RESEND) + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + cookie->sent = SCTP_DATAGRAM_RESEND; + sctp_stop_all_cookie_timers(stcb); + } + } + break; + case SCTP_COOKIE_ACK: + sctp_send_cookie_ack(stcb); + break; + case SCTP_ASCONF_ACK: + /* resend last asconf ack */ + sctp_send_asconf_ack(stcb); + break; + case SCTP_IFORWARD_CUM_TSN: + case SCTP_FORWARD_CUM_TSN: + send_forward_tsn(stcb, &stcb->asoc); + break; + /* can't do anything with these */ + case SCTP_PACKET_DROPPED: + case SCTP_INITIATION_ACK: /* this should not happen */ + case SCTP_HEARTBEAT_ACK: + case SCTP_ABORT_ASSOCIATION: + case SCTP_OPERATION_ERROR: + case SCTP_SHUTDOWN_COMPLETE: + case SCTP_ECN_ECHO: + case SCTP_ECN_CWR: + default: + break; + } + return (0); +} + +void +sctp_reset_in_stream(struct sctp_tcb *stcb, uint32_t number_entries, uint16_t *list) +{ + uint32_t i; + uint16_t temp; + + /* + * We set things to 0xffffffff since this is the last delivered sequence + * and we will be sending in 0 after the reset. + */ + + if (number_entries) { + for (i = 0; i < number_entries; i++) { + temp = ntohs(list[i]); + if (temp >= stcb->asoc.streamincnt) { + continue; + } + stcb->asoc.strmin[temp].last_mid_delivered = 0xffffffff; + } + } else { + list = NULL; + for (i = 0; i < stcb->asoc.streamincnt; i++) { + stcb->asoc.strmin[i].last_mid_delivered = 0xffffffff; + } + } + sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_RECV, stcb, number_entries, (void *)list, SCTP_SO_NOT_LOCKED); +} + +static void +sctp_reset_out_streams(struct sctp_tcb *stcb, uint32_t number_entries, uint16_t *list) +{ + uint32_t i; + uint16_t temp; + + if (number_entries > 0) { + for (i = 0; i < number_entries; i++) { + temp = ntohs(list[i]); + if (temp >= stcb->asoc.streamoutcnt) { + /* no such stream */ + continue; + } + stcb->asoc.strmout[temp].next_mid_ordered = 0; + stcb->asoc.strmout[temp].next_mid_unordered = 0; + } + } else { + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + stcb->asoc.strmout[i].next_mid_ordered = 0; + stcb->asoc.strmout[i].next_mid_unordered = 0; + } + } + sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_SEND, stcb, number_entries, (void *)list, SCTP_SO_NOT_LOCKED); +} + +static void +sctp_reset_clear_pending(struct sctp_tcb *stcb, uint32_t number_entries, uint16_t *list) +{ + uint32_t i; + uint16_t temp; + + if (number_entries > 0) { + for (i = 0; i < number_entries; i++) { + temp = ntohs(list[i]); + if (temp >= stcb->asoc.streamoutcnt) { + /* no such stream */ + continue; + } + stcb->asoc.strmout[temp].state = SCTP_STREAM_OPEN; + } + } else { + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + stcb->asoc.strmout[i].state = SCTP_STREAM_OPEN; + } + } +} + +struct sctp_stream_reset_request * +sctp_find_stream_reset(struct sctp_tcb *stcb, uint32_t seq, struct sctp_tmit_chunk **bchk) +{ + struct sctp_association *asoc; + struct sctp_chunkhdr *ch; + struct sctp_stream_reset_request *r; + struct sctp_tmit_chunk *chk; + int len, clen; + + asoc = &stcb->asoc; + chk = asoc->str_reset; + if (TAILQ_EMPTY(&asoc->control_send_queue) || + (chk == NULL)) { + asoc->stream_reset_outstanding = 0; + return (NULL); + } + if (chk->data == NULL) { + return (NULL); + } + if (bchk != NULL) { + /* he wants a copy of the chk pointer */ + *bchk = chk; + } + clen = chk->send_size; + ch = mtod(chk->data, struct sctp_chunkhdr *); + r = (struct sctp_stream_reset_request *)(ch + 1); + if (ntohl(r->request_seq) == seq) { + /* found it */ + return (r); + } + len = SCTP_SIZE32(ntohs(r->ph.param_length)); + if (clen > (len + (int)sizeof(struct sctp_chunkhdr))) { + /* move to the next one, there can only be a max of two */ + r = (struct sctp_stream_reset_request *)((caddr_t)r + len); + if (ntohl(r->request_seq) == seq) { + return (r); + } + } + /* that seq is not here */ + return (NULL); +} + +static void +sctp_clean_up_stream_reset(struct sctp_tcb *stcb) +{ + struct sctp_association *asoc; + struct sctp_tmit_chunk *chk; + + asoc = &stcb->asoc; + chk = asoc->str_reset; + if (chk == NULL) { + return; + } + asoc->str_reset = NULL; + sctp_timer_stop(SCTP_TIMER_TYPE_STRRESET, stcb->sctp_ep, stcb, + NULL, SCTP_FROM_SCTP_INPUT + SCTP_LOC_28); + TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); + asoc->ctrl_queue_cnt--; + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); +} + +static int +sctp_handle_stream_reset_response(struct sctp_tcb *stcb, + uint32_t seq, uint32_t action, + struct sctp_stream_reset_response *respin) +{ + uint16_t type; + int lparam_len; + struct sctp_association *asoc = &stcb->asoc; + struct sctp_tmit_chunk *chk; + struct sctp_stream_reset_request *req_param; + struct sctp_stream_reset_out_request *req_out_param; + struct sctp_stream_reset_in_request *req_in_param; + uint32_t number_entries; + + if (asoc->stream_reset_outstanding == 0) { + /* duplicate */ + return (0); + } + if (seq == stcb->asoc.str_reset_seq_out) { + req_param = sctp_find_stream_reset(stcb, seq, &chk); + if (req_param != NULL) { + stcb->asoc.str_reset_seq_out++; + type = ntohs(req_param->ph.param_type); + lparam_len = ntohs(req_param->ph.param_length); + if (type == SCTP_STR_RESET_OUT_REQUEST) { + int no_clear = 0; + + req_out_param = (struct sctp_stream_reset_out_request *)req_param; + number_entries = (lparam_len - sizeof(struct sctp_stream_reset_out_request)) / sizeof(uint16_t); + asoc->stream_reset_out_is_outstanding = 0; + if (asoc->stream_reset_outstanding) + asoc->stream_reset_outstanding--; + if (action == SCTP_STREAM_RESET_RESULT_PERFORMED) { + /* do it */ + sctp_reset_out_streams(stcb, number_entries, req_out_param->list_of_streams); + } else if (action == SCTP_STREAM_RESET_RESULT_DENIED) { + sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_DENIED_OUT, stcb, number_entries, req_out_param->list_of_streams, SCTP_SO_NOT_LOCKED); + } else if (action == SCTP_STREAM_RESET_RESULT_IN_PROGRESS) { + /* Set it up so we don't stop retransmitting */ + asoc->stream_reset_outstanding++; + stcb->asoc.str_reset_seq_out--; + asoc->stream_reset_out_is_outstanding = 1; + no_clear = 1; + } else { + sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_FAILED_OUT, stcb, number_entries, req_out_param->list_of_streams, SCTP_SO_NOT_LOCKED); + } + if (no_clear == 0) { + sctp_reset_clear_pending(stcb, number_entries, req_out_param->list_of_streams); + } + } else if (type == SCTP_STR_RESET_IN_REQUEST) { + req_in_param = (struct sctp_stream_reset_in_request *)req_param; + number_entries = (lparam_len - sizeof(struct sctp_stream_reset_in_request)) / sizeof(uint16_t); + if (asoc->stream_reset_outstanding) + asoc->stream_reset_outstanding--; + if (action == SCTP_STREAM_RESET_RESULT_DENIED) { + sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_DENIED_IN, stcb, + number_entries, req_in_param->list_of_streams, SCTP_SO_NOT_LOCKED); + } else if (action != SCTP_STREAM_RESET_RESULT_PERFORMED) { + sctp_ulp_notify(SCTP_NOTIFY_STR_RESET_FAILED_IN, stcb, + number_entries, req_in_param->list_of_streams, SCTP_SO_NOT_LOCKED); + } + } else if (type == SCTP_STR_RESET_ADD_OUT_STREAMS) { + /* Ok we now may have more streams */ + int num_stream; + + num_stream = stcb->asoc.strm_pending_add_size; + if (num_stream > (stcb->asoc.strm_realoutsize - stcb->asoc.streamoutcnt)) { + /* TSNH */ + num_stream = stcb->asoc.strm_realoutsize - stcb->asoc.streamoutcnt; + } + stcb->asoc.strm_pending_add_size = 0; + if (asoc->stream_reset_outstanding) + asoc->stream_reset_outstanding--; + if (action == SCTP_STREAM_RESET_RESULT_PERFORMED) { + /* Put the new streams into effect */ + int i; + for (i = asoc->streamoutcnt; i < (asoc->streamoutcnt + num_stream); i++) { + asoc->strmout[i].state = SCTP_STREAM_OPEN; + } + asoc->streamoutcnt += num_stream; + sctp_notify_stream_reset_add(stcb, stcb->asoc.streamincnt, stcb->asoc.streamoutcnt, 0); + } else if (action == SCTP_STREAM_RESET_RESULT_DENIED) { + sctp_notify_stream_reset_add(stcb, stcb->asoc.streamincnt, stcb->asoc.streamoutcnt, + SCTP_STREAM_CHANGE_DENIED); + } else { + sctp_notify_stream_reset_add(stcb, stcb->asoc.streamincnt, stcb->asoc.streamoutcnt, + SCTP_STREAM_CHANGE_FAILED); + } + } else if (type == SCTP_STR_RESET_ADD_IN_STREAMS) { + if (asoc->stream_reset_outstanding) + asoc->stream_reset_outstanding--; + if (action == SCTP_STREAM_RESET_RESULT_DENIED) { + sctp_notify_stream_reset_add(stcb, stcb->asoc.streamincnt, stcb->asoc.streamoutcnt, + SCTP_STREAM_CHANGE_DENIED); + } else if (action != SCTP_STREAM_RESET_RESULT_PERFORMED) { + sctp_notify_stream_reset_add(stcb, stcb->asoc.streamincnt, stcb->asoc.streamoutcnt, + SCTP_STREAM_CHANGE_FAILED); + } + } else if (type == SCTP_STR_RESET_TSN_REQUEST) { + /** + * a) Adopt the new in tsn. + * b) reset the map + * c) Adopt the new out-tsn + */ + struct sctp_stream_reset_response_tsn *resp; + struct sctp_forward_tsn_chunk fwdtsn; + int abort_flag = 0; + if (respin == NULL) { + /* huh ? */ + return (0); + } + if (ntohs(respin->ph.param_length) < sizeof(struct sctp_stream_reset_response_tsn)) { + return (0); + } + if (action == SCTP_STREAM_RESET_RESULT_PERFORMED) { + resp = (struct sctp_stream_reset_response_tsn *)respin; + asoc->stream_reset_outstanding--; + fwdtsn.ch.chunk_length = htons(sizeof(struct sctp_forward_tsn_chunk)); + fwdtsn.ch.chunk_type = SCTP_FORWARD_CUM_TSN; + fwdtsn.new_cumulative_tsn = htonl(ntohl(resp->senders_next_tsn) - 1); + sctp_handle_forward_tsn(stcb, &fwdtsn, &abort_flag, NULL, 0); + if (abort_flag) { + return (1); + } + stcb->asoc.highest_tsn_inside_map = (ntohl(resp->senders_next_tsn) - 1); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { + sctp_log_map(0, 7, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); + } + + stcb->asoc.tsn_last_delivered = stcb->asoc.cumulative_tsn = stcb->asoc.highest_tsn_inside_map; + stcb->asoc.mapping_array_base_tsn = ntohl(resp->senders_next_tsn); + memset(stcb->asoc.mapping_array, 0, stcb->asoc.mapping_array_size); + + stcb->asoc.highest_tsn_inside_nr_map = stcb->asoc.highest_tsn_inside_map; + memset(stcb->asoc.nr_mapping_array, 0, stcb->asoc.mapping_array_size); + + stcb->asoc.sending_seq = ntohl(resp->receivers_next_tsn); + stcb->asoc.last_acked_seq = stcb->asoc.cumulative_tsn; + + sctp_reset_out_streams(stcb, 0, (uint16_t *) NULL); + sctp_reset_in_stream(stcb, 0, (uint16_t *) NULL); + sctp_notify_stream_reset_tsn(stcb, stcb->asoc.sending_seq, (stcb->asoc.mapping_array_base_tsn + 1), 0); + } else if (action == SCTP_STREAM_RESET_RESULT_DENIED) { + sctp_notify_stream_reset_tsn(stcb, stcb->asoc.sending_seq, (stcb->asoc.mapping_array_base_tsn + 1), + SCTP_ASSOC_RESET_DENIED); + } else { + sctp_notify_stream_reset_tsn(stcb, stcb->asoc.sending_seq, (stcb->asoc.mapping_array_base_tsn + 1), + SCTP_ASSOC_RESET_FAILED); + } + } + /* get rid of the request and get the request flags */ + if (asoc->stream_reset_outstanding == 0) { + sctp_clean_up_stream_reset(stcb); + } + } + } + if (asoc->stream_reset_outstanding == 0) { + sctp_send_stream_reset_out_if_possible(stcb, SCTP_SO_NOT_LOCKED); + } + return (0); +} + +static void +sctp_handle_str_reset_request_in(struct sctp_tcb *stcb, + struct sctp_tmit_chunk *chk, + struct sctp_stream_reset_in_request *req, int trunc) +{ + uint32_t seq; + int len, i; + int number_entries; + uint16_t temp; + + /* + * peer wants me to send a str-reset to him for my outgoing seq's if + * seq_in is right. + */ + struct sctp_association *asoc = &stcb->asoc; + + seq = ntohl(req->request_seq); + if (asoc->str_reset_seq_in == seq) { + asoc->last_reset_action[1] = asoc->last_reset_action[0]; + if ((asoc->local_strreset_support & SCTP_ENABLE_RESET_STREAM_REQ) == 0) { + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_DENIED; + } else if (trunc) { + /* Can't do it, since they exceeded our buffer size */ + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_DENIED; + } else if (stcb->asoc.stream_reset_out_is_outstanding == 0) { + len = ntohs(req->ph.param_length); + number_entries = ((len - sizeof(struct sctp_stream_reset_in_request)) / sizeof(uint16_t)); + if (number_entries) { + for (i = 0; i < number_entries; i++) { + temp = ntohs(req->list_of_streams[i]); + if (temp >= stcb->asoc.streamoutcnt) { + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_DENIED; + goto bad_boy; + } + req->list_of_streams[i] = temp; + } + for (i = 0; i < number_entries; i++) { + if (stcb->asoc.strmout[req->list_of_streams[i]].state == SCTP_STREAM_OPEN) { + stcb->asoc.strmout[req->list_of_streams[i]].state = SCTP_STREAM_RESET_PENDING; + } + } + } else { + /* Its all */ + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + if (stcb->asoc.strmout[i].state == SCTP_STREAM_OPEN) + stcb->asoc.strmout[i].state = SCTP_STREAM_RESET_PENDING; + } + } + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_PERFORMED; + } else { + /* Can't do it, since we have sent one out */ + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_ERR_IN_PROGRESS; + } + bad_boy: + sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); + asoc->str_reset_seq_in++; + } else if (asoc->str_reset_seq_in - 1 == seq) { + sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); + } else if (asoc->str_reset_seq_in - 2 == seq) { + sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[1]); + } else { + sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_RESULT_ERR_BAD_SEQNO); + } + sctp_send_stream_reset_out_if_possible(stcb, SCTP_SO_NOT_LOCKED); +} + +static int +sctp_handle_str_reset_request_tsn(struct sctp_tcb *stcb, + struct sctp_tmit_chunk *chk, + struct sctp_stream_reset_tsn_request *req) +{ + /* reset all in and out and update the tsn */ + /* + * A) reset my str-seq's on in and out. B) Select a receive next, + * and set cum-ack to it. Also process this selected number as a + * fwd-tsn as well. C) set in the response my next sending seq. + */ + struct sctp_forward_tsn_chunk fwdtsn; + struct sctp_association *asoc = &stcb->asoc; + int abort_flag = 0; + uint32_t seq; + + seq = ntohl(req->request_seq); + if (asoc->str_reset_seq_in == seq) { + asoc->last_reset_action[1] = stcb->asoc.last_reset_action[0]; + if ((asoc->local_strreset_support & SCTP_ENABLE_CHANGE_ASSOC_REQ) == 0) { + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_DENIED; + } else { + fwdtsn.ch.chunk_length = htons(sizeof(struct sctp_forward_tsn_chunk)); + fwdtsn.ch.chunk_type = SCTP_FORWARD_CUM_TSN; + fwdtsn.ch.chunk_flags = 0; + fwdtsn.new_cumulative_tsn = htonl(stcb->asoc.highest_tsn_inside_map + 1); + sctp_handle_forward_tsn(stcb, &fwdtsn, &abort_flag, NULL, 0); + if (abort_flag) { + return (1); + } + asoc->highest_tsn_inside_map += SCTP_STREAM_RESET_TSN_DELTA; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { + sctp_log_map(0, 10, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); + } + asoc->tsn_last_delivered = asoc->cumulative_tsn = asoc->highest_tsn_inside_map; + asoc->mapping_array_base_tsn = asoc->highest_tsn_inside_map + 1; + memset(asoc->mapping_array, 0, asoc->mapping_array_size); + asoc->highest_tsn_inside_nr_map = asoc->highest_tsn_inside_map; + memset(asoc->nr_mapping_array, 0, asoc->mapping_array_size); + atomic_add_int(&asoc->sending_seq, 1); + /* save off historical data for retrans */ + asoc->last_sending_seq[1] = asoc->last_sending_seq[0]; + asoc->last_sending_seq[0] = asoc->sending_seq; + asoc->last_base_tsnsent[1] = asoc->last_base_tsnsent[0]; + asoc->last_base_tsnsent[0] = asoc->mapping_array_base_tsn; + sctp_reset_out_streams(stcb, 0, (uint16_t *) NULL); + sctp_reset_in_stream(stcb, 0, (uint16_t *) NULL); + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_PERFORMED; + sctp_notify_stream_reset_tsn(stcb, asoc->sending_seq, (asoc->mapping_array_base_tsn + 1), 0); + } + sctp_add_stream_reset_result_tsn(chk, seq, asoc->last_reset_action[0], + asoc->last_sending_seq[0], asoc->last_base_tsnsent[0]); + asoc->str_reset_seq_in++; + } else if (asoc->str_reset_seq_in - 1 == seq) { + sctp_add_stream_reset_result_tsn(chk, seq, asoc->last_reset_action[0], + asoc->last_sending_seq[0], asoc->last_base_tsnsent[0]); + } else if (asoc->str_reset_seq_in - 2 == seq) { + sctp_add_stream_reset_result_tsn(chk, seq, asoc->last_reset_action[1], + asoc->last_sending_seq[1], asoc->last_base_tsnsent[1]); + } else { + sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_RESULT_ERR_BAD_SEQNO); + } + return (0); +} + +static void +sctp_handle_str_reset_request_out(struct sctp_tcb *stcb, + struct sctp_tmit_chunk *chk, + struct sctp_stream_reset_out_request *req, int trunc) +{ + uint32_t seq, tsn; + int number_entries, len; + struct sctp_association *asoc = &stcb->asoc; + + seq = ntohl(req->request_seq); + + /* now if its not a duplicate we process it */ + if (asoc->str_reset_seq_in == seq) { + len = ntohs(req->ph.param_length); + number_entries = ((len - sizeof(struct sctp_stream_reset_out_request)) / sizeof(uint16_t)); + /* + * the sender is resetting, handle the list issue.. we must + * a) verify if we can do the reset, if so no problem b) If + * we can't do the reset we must copy the request. c) queue + * it, and setup the data in processor to trigger it off + * when needed and dequeue all the queued data. + */ + tsn = ntohl(req->send_reset_at_tsn); + + /* move the reset action back one */ + asoc->last_reset_action[1] = asoc->last_reset_action[0]; + if ((asoc->local_strreset_support & SCTP_ENABLE_RESET_STREAM_REQ) == 0) { + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_DENIED; + } else if (trunc) { + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_DENIED; + } else if (SCTP_TSN_GE(asoc->cumulative_tsn, tsn)) { + /* we can do it now */ + sctp_reset_in_stream(stcb, number_entries, req->list_of_streams); + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_PERFORMED; + } else { + /* + * we must queue it up and thus wait for the TSN's + * to arrive that are at or before tsn + */ + struct sctp_stream_reset_list *liste; + int siz; + + siz = sizeof(struct sctp_stream_reset_list) + (number_entries * sizeof(uint16_t)); + SCTP_MALLOC(liste, struct sctp_stream_reset_list *, + siz, SCTP_M_STRESET); + if (liste == NULL) { + /* gak out of memory */ + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_DENIED; + sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); + return; + } + liste->seq = seq; + liste->tsn = tsn; + liste->number_entries = number_entries; + memcpy(&liste->list_of_streams, req->list_of_streams, number_entries * sizeof(uint16_t)); + TAILQ_INSERT_TAIL(&asoc->resetHead, liste, next_resp); + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_IN_PROGRESS; + } + sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); + asoc->str_reset_seq_in++; + } else if ((asoc->str_reset_seq_in - 1) == seq) { + /* + * one seq back, just echo back last action since my + * response was lost. + */ + sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); + } else if ((asoc->str_reset_seq_in - 2) == seq) { + /* + * two seq back, just echo back last action since my + * response was lost. + */ + sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[1]); + } else { + sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_RESULT_ERR_BAD_SEQNO); + } +} + +static void +sctp_handle_str_reset_add_strm(struct sctp_tcb *stcb, struct sctp_tmit_chunk *chk, + struct sctp_stream_reset_add_strm *str_add) +{ + /* + * Peer is requesting to add more streams. + * If its within our max-streams we will + * allow it. + */ + uint32_t num_stream, i; + uint32_t seq; + struct sctp_association *asoc = &stcb->asoc; + struct sctp_queued_to_read *ctl, *nctl; + + /* Get the number. */ + seq = ntohl(str_add->request_seq); + num_stream = ntohs(str_add->number_of_streams); + /* Now what would be the new total? */ + if (asoc->str_reset_seq_in == seq) { + num_stream += stcb->asoc.streamincnt; + stcb->asoc.last_reset_action[1] = stcb->asoc.last_reset_action[0]; + if ((asoc->local_strreset_support & SCTP_ENABLE_CHANGE_ASSOC_REQ) == 0) { + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_DENIED; + } else if ((num_stream > stcb->asoc.max_inbound_streams) || + (num_stream > 0xffff)) { + /* We must reject it they ask for to many */ + denied: + stcb->asoc.last_reset_action[0] = SCTP_STREAM_RESET_RESULT_DENIED; + } else { + /* Ok, we can do that :-) */ + struct sctp_stream_in *oldstrm; + + /* save off the old */ + oldstrm = stcb->asoc.strmin; + SCTP_MALLOC(stcb->asoc.strmin, struct sctp_stream_in *, + (num_stream * sizeof(struct sctp_stream_in)), + SCTP_M_STRMI); + if (stcb->asoc.strmin == NULL) { + stcb->asoc.strmin = oldstrm; + goto denied; + } + /* copy off the old data */ + for (i = 0; i < stcb->asoc.streamincnt; i++) { + TAILQ_INIT(&stcb->asoc.strmin[i].inqueue); + TAILQ_INIT(&stcb->asoc.strmin[i].uno_inqueue); + stcb->asoc.strmin[i].sid = i; + stcb->asoc.strmin[i].last_mid_delivered = oldstrm[i].last_mid_delivered; + stcb->asoc.strmin[i].delivery_started = oldstrm[i].delivery_started; + stcb->asoc.strmin[i].pd_api_started = oldstrm[i].pd_api_started; + /* now anything on those queues? */ + TAILQ_FOREACH_SAFE(ctl, &oldstrm[i].inqueue, next_instrm, nctl) { + TAILQ_REMOVE(&oldstrm[i].inqueue, ctl, next_instrm); + TAILQ_INSERT_TAIL(&stcb->asoc.strmin[i].inqueue, ctl, next_instrm); + } + TAILQ_FOREACH_SAFE(ctl, &oldstrm[i].uno_inqueue, next_instrm, nctl) { + TAILQ_REMOVE(&oldstrm[i].uno_inqueue, ctl, next_instrm); + TAILQ_INSERT_TAIL(&stcb->asoc.strmin[i].uno_inqueue, ctl, next_instrm); + } + } + /* Init the new streams */ + for (i = stcb->asoc.streamincnt; i < num_stream; i++) { + TAILQ_INIT(&stcb->asoc.strmin[i].inqueue); + TAILQ_INIT(&stcb->asoc.strmin[i].uno_inqueue); + stcb->asoc.strmin[i].sid = i; + stcb->asoc.strmin[i].last_mid_delivered = 0xffffffff; + stcb->asoc.strmin[i].pd_api_started = 0; + stcb->asoc.strmin[i].delivery_started = 0; + } + SCTP_FREE(oldstrm, SCTP_M_STRMI); + /* update the size */ + stcb->asoc.streamincnt = num_stream; + stcb->asoc.last_reset_action[0] = SCTP_STREAM_RESET_RESULT_PERFORMED; + sctp_notify_stream_reset_add(stcb, stcb->asoc.streamincnt, stcb->asoc.streamoutcnt, 0); + } + sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); + asoc->str_reset_seq_in++; + } else if ((asoc->str_reset_seq_in - 1) == seq) { + /* + * one seq back, just echo back last action since my + * response was lost. + */ + sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); + } else if ((asoc->str_reset_seq_in - 2) == seq) { + /* + * two seq back, just echo back last action since my + * response was lost. + */ + sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[1]); + } else { + sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_RESULT_ERR_BAD_SEQNO); + } +} + +static void +sctp_handle_str_reset_add_out_strm(struct sctp_tcb *stcb, struct sctp_tmit_chunk *chk, + struct sctp_stream_reset_add_strm *str_add) +{ + /* + * Peer is requesting to add more streams. + * If its within our max-streams we will + * allow it. + */ + uint16_t num_stream; + uint32_t seq; + struct sctp_association *asoc = &stcb->asoc; + + /* Get the number. */ + seq = ntohl(str_add->request_seq); + num_stream = ntohs(str_add->number_of_streams); + /* Now what would be the new total? */ + if (asoc->str_reset_seq_in == seq) { + stcb->asoc.last_reset_action[1] = stcb->asoc.last_reset_action[0]; + if ((asoc->local_strreset_support & SCTP_ENABLE_CHANGE_ASSOC_REQ) == 0) { + asoc->last_reset_action[0] = SCTP_STREAM_RESET_RESULT_DENIED; + } else if (stcb->asoc.stream_reset_outstanding) { + /* We must reject it we have something pending */ + stcb->asoc.last_reset_action[0] = SCTP_STREAM_RESET_RESULT_ERR_IN_PROGRESS; + } else { + /* Ok, we can do that :-) */ + int mychk; + mychk = stcb->asoc.streamoutcnt; + mychk += num_stream; + if (mychk < 0x10000) { + stcb->asoc.last_reset_action[0] = SCTP_STREAM_RESET_RESULT_PERFORMED; + if (sctp_send_str_reset_req(stcb, 0, NULL, 0, 0, 1, num_stream, 0, 1)) { + stcb->asoc.last_reset_action[0] = SCTP_STREAM_RESET_RESULT_DENIED; + } + } else { + stcb->asoc.last_reset_action[0] = SCTP_STREAM_RESET_RESULT_DENIED; + } + } + sctp_add_stream_reset_result(chk, seq, stcb->asoc.last_reset_action[0]); + asoc->str_reset_seq_in++; + } else if ((asoc->str_reset_seq_in - 1) == seq) { + /* + * one seq back, just echo back last action since my + * response was lost. + */ + sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[0]); + } else if ((asoc->str_reset_seq_in - 2) == seq) { + /* + * two seq back, just echo back last action since my + * response was lost. + */ + sctp_add_stream_reset_result(chk, seq, asoc->last_reset_action[1]); + } else { + sctp_add_stream_reset_result(chk, seq, SCTP_STREAM_RESET_RESULT_ERR_BAD_SEQNO); + } +} + +#ifdef __GNUC__ +__attribute__ ((noinline)) +#endif +static int +sctp_handle_stream_reset(struct sctp_tcb *stcb, struct mbuf *m, int offset, + struct sctp_chunkhdr *ch_req) +{ + uint16_t remaining_length, param_len, ptype; + struct sctp_paramhdr pstore; + uint8_t cstore[SCTP_CHUNK_BUFFER_SIZE]; + uint32_t seq = 0; + int num_req = 0; + int trunc = 0; + struct sctp_tmit_chunk *chk; + struct sctp_chunkhdr *ch; + struct sctp_paramhdr *ph; + int ret_code = 0; + int num_param = 0; + + /* now it may be a reset or a reset-response */ + remaining_length = ntohs(ch_req->chunk_length) - sizeof(struct sctp_chunkhdr); + + /* setup for adding the response */ + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + return (ret_code); + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_STREAM_RESET; + chk->rec.chunk_id.can_take_data = 0; + chk->flags = 0; + chk->asoc = &stcb->asoc; + chk->no_fr_allowed = 0; + chk->book_size = chk->send_size = sizeof(struct sctp_chunkhdr); + chk->book_size_scale = 0; + chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_NOWAIT, 1, MT_DATA); + if (chk->data == NULL) { + strres_nochunk: + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + return (ret_code); + } + SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); + + /* setup chunk parameters */ + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + chk->whoTo = NULL; + + ch = mtod(chk->data, struct sctp_chunkhdr *); + ch->chunk_type = SCTP_STREAM_RESET; + ch->chunk_flags = 0; + ch->chunk_length = htons(chk->send_size); + SCTP_BUF_LEN(chk->data) = SCTP_SIZE32(chk->send_size); + offset += sizeof(struct sctp_chunkhdr); + while (remaining_length >= sizeof(struct sctp_paramhdr)) { + ph = (struct sctp_paramhdr *)sctp_m_getptr(m, offset, sizeof(pstore), (uint8_t *)&pstore); + if (ph == NULL) { + /* TSNH */ + break; + } + param_len = ntohs(ph->param_length); + if ((param_len > remaining_length) || + (param_len < (sizeof(struct sctp_paramhdr) + sizeof(uint32_t)))) { + /* bad parameter length */ + break; + } + ph = (struct sctp_paramhdr *)sctp_m_getptr(m, offset, min(param_len, sizeof(cstore)), + (uint8_t *)&cstore); + if (ph == NULL) { + /* TSNH */ + break; + } + ptype = ntohs(ph->param_type); + num_param++; + if (param_len > sizeof(cstore)) { + trunc = 1; + } else { + trunc = 0; + } + if (num_param > SCTP_MAX_RESET_PARAMS) { + /* hit the max of parameters already sorry.. */ + break; + } + if (ptype == SCTP_STR_RESET_OUT_REQUEST) { + struct sctp_stream_reset_out_request *req_out; + + if (param_len < sizeof(struct sctp_stream_reset_out_request)) { + break; + } + req_out = (struct sctp_stream_reset_out_request *)ph; + num_req++; + if (stcb->asoc.stream_reset_outstanding) { + seq = ntohl(req_out->response_seq); + if (seq == stcb->asoc.str_reset_seq_out) { + /* implicit ack */ + (void)sctp_handle_stream_reset_response(stcb, seq, SCTP_STREAM_RESET_RESULT_PERFORMED, NULL); + } + } + sctp_handle_str_reset_request_out(stcb, chk, req_out, trunc); + } else if (ptype == SCTP_STR_RESET_ADD_OUT_STREAMS) { + struct sctp_stream_reset_add_strm *str_add; + + if (param_len < sizeof(struct sctp_stream_reset_add_strm)) { + break; + } + str_add = (struct sctp_stream_reset_add_strm *)ph; + num_req++; + sctp_handle_str_reset_add_strm(stcb, chk, str_add); + } else if (ptype == SCTP_STR_RESET_ADD_IN_STREAMS) { + struct sctp_stream_reset_add_strm *str_add; + + if (param_len < sizeof(struct sctp_stream_reset_add_strm)) { + break; + } + str_add = (struct sctp_stream_reset_add_strm *)ph; + num_req++; + sctp_handle_str_reset_add_out_strm(stcb, chk, str_add); + } else if (ptype == SCTP_STR_RESET_IN_REQUEST) { + struct sctp_stream_reset_in_request *req_in; + + num_req++; + req_in = (struct sctp_stream_reset_in_request *)ph; + sctp_handle_str_reset_request_in(stcb, chk, req_in, trunc); + } else if (ptype == SCTP_STR_RESET_TSN_REQUEST) { + struct sctp_stream_reset_tsn_request *req_tsn; + + num_req++; + req_tsn = (struct sctp_stream_reset_tsn_request *)ph; + if (sctp_handle_str_reset_request_tsn(stcb, chk, req_tsn)) { + ret_code = 1; + goto strres_nochunk; + } + /* no more */ + break; + } else if (ptype == SCTP_STR_RESET_RESPONSE) { + struct sctp_stream_reset_response *resp; + uint32_t result; + + if (param_len < sizeof(struct sctp_stream_reset_response)) { + break; + } + resp = (struct sctp_stream_reset_response *)ph; + seq = ntohl(resp->response_seq); + result = ntohl(resp->result); + if (sctp_handle_stream_reset_response(stcb, seq, result, resp)) { + ret_code = 1; + goto strres_nochunk; + } + } else { + break; + } + offset += SCTP_SIZE32(param_len); + if (remaining_length >= SCTP_SIZE32(param_len)) { + remaining_length -= SCTP_SIZE32(param_len); + } else { + remaining_length = 0; + } + } + if (num_req == 0) { + /* we have no response free the stuff */ + goto strres_nochunk; + } + /* ok we have a chunk to link in */ + TAILQ_INSERT_TAIL(&stcb->asoc.control_send_queue, + chk, + sctp_next); + stcb->asoc.ctrl_queue_cnt++; + return (ret_code); +} + +/* + * Handle a router or endpoints report of a packet loss, there are two ways + * to handle this, either we get the whole packet and must disect it + * ourselves (possibly with truncation and or corruption) or it is a summary + * from a middle box that did the disecting for us. + */ +static void +sctp_handle_packet_dropped(struct sctp_pktdrop_chunk *cp, + struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t limit) +{ + struct sctp_chunk_desc desc; + struct sctp_chunkhdr *chk_hdr; + struct sctp_data_chunk *data_chunk; + struct sctp_idata_chunk *idata_chunk; + uint32_t bottle_bw, on_queue; + uint32_t offset, chk_len; + uint16_t pktdrp_len; + uint8_t pktdrp_flags; + + KASSERT(sizeof(struct sctp_pktdrop_chunk) <= limit, + ("PKTDROP chunk too small")); + pktdrp_flags = cp->ch.chunk_flags; + pktdrp_len = ntohs(cp->ch.chunk_length); + KASSERT(limit <= pktdrp_len, ("Inconsistent limit")); + if (pktdrp_flags & SCTP_PACKET_TRUNCATED) { + if (ntohs(cp->trunc_len) <= pktdrp_len - sizeof(struct sctp_pktdrop_chunk)) { + /* The peer plays games with us. */ + return; + } + } + limit -= sizeof(struct sctp_pktdrop_chunk); + offset = 0; + if (offset == limit) { + if (pktdrp_flags & SCTP_FROM_MIDDLE_BOX) { + SCTP_STAT_INCR(sctps_pdrpbwrpt); + } + } else if (offset + sizeof(struct sctphdr) > limit) { + /* Only a partial SCTP common header. */ + SCTP_STAT_INCR(sctps_pdrpcrupt); + offset = limit; + } else { + /* XXX: Check embedded SCTP common header. */ + offset += sizeof(struct sctphdr); + } + /* Now parse through the chunks themselves. */ + while (offset < limit) { + if (offset + sizeof(struct sctp_chunkhdr) > limit) { + SCTP_STAT_INCR(sctps_pdrpcrupt); + break; + } + chk_hdr = (struct sctp_chunkhdr *)(cp->data + offset); + desc.chunk_type = chk_hdr->chunk_type; + /* get amount we need to move */ + chk_len = (uint32_t)ntohs(chk_hdr->chunk_length); + if (chk_len < sizeof(struct sctp_chunkhdr)) { + /* Someone is lying... */ + break; + } + if (desc.chunk_type == SCTP_DATA) { + if (stcb->asoc.idata_supported) { + /* Some is playing games with us. */ + break; + } + if (chk_len <= sizeof(struct sctp_data_chunk)) { + /* Some is playing games with us. */ + break; + } + if (chk_len < sizeof(struct sctp_data_chunk) + SCTP_NUM_DB_TO_VERIFY) { + /* Not enough data bytes available in the chunk. */ + SCTP_STAT_INCR(sctps_pdrpnedat); + goto next_chunk; + } + if (offset + sizeof(struct sctp_data_chunk) + SCTP_NUM_DB_TO_VERIFY > limit) { + /* Not enough data in buffer. */ + break; + } + data_chunk = (struct sctp_data_chunk *)(cp->data + offset); + memcpy(desc.data_bytes, data_chunk + 1, SCTP_NUM_DB_TO_VERIFY); + desc.tsn_ifany = data_chunk->dp.tsn; + if (pktdrp_flags & SCTP_FROM_MIDDLE_BOX) { + SCTP_STAT_INCR(sctps_pdrpmbda); + } + } else if (desc.chunk_type == SCTP_IDATA) { + if (!stcb->asoc.idata_supported) { + /* Some is playing games with us. */ + break; + } + if (chk_len <= sizeof(struct sctp_idata_chunk)) { + /* Some is playing games with us. */ + break; + } + if (chk_len < sizeof(struct sctp_idata_chunk) + SCTP_NUM_DB_TO_VERIFY) { + /* Not enough data bytes available in the chunk. */ + SCTP_STAT_INCR(sctps_pdrpnedat); + goto next_chunk; + } + if (offset + sizeof(struct sctp_idata_chunk) + SCTP_NUM_DB_TO_VERIFY > limit) { + /* Not enough data in buffer. */ + break; + } + idata_chunk = (struct sctp_idata_chunk *)(cp->data + offset); + memcpy(desc.data_bytes, idata_chunk + 1, SCTP_NUM_DB_TO_VERIFY); + desc.tsn_ifany = idata_chunk->dp.tsn; + if (pktdrp_flags & SCTP_FROM_MIDDLE_BOX) { + SCTP_STAT_INCR(sctps_pdrpmbda); + } + } else { + if (pktdrp_flags & SCTP_FROM_MIDDLE_BOX) { + SCTP_STAT_INCR(sctps_pdrpmbct); + } + } + if (process_chunk_drop(stcb, &desc, net, pktdrp_flags)) { + SCTP_STAT_INCR(sctps_pdrppdbrk); + break; + } +next_chunk: + offset += SCTP_SIZE32(chk_len); + } + /* Now update any rwnd --- possibly */ + if ((pktdrp_flags & SCTP_FROM_MIDDLE_BOX) == 0) { + /* From a peer, we get a rwnd report */ + uint32_t a_rwnd; + + SCTP_STAT_INCR(sctps_pdrpfehos); + + bottle_bw = ntohl(cp->bottle_bw); + on_queue = ntohl(cp->current_onq); + if (bottle_bw && on_queue) { + /* a rwnd report is in here */ + if (bottle_bw > on_queue) + a_rwnd = bottle_bw - on_queue; + else + a_rwnd = 0; + + if (a_rwnd == 0) + stcb->asoc.peers_rwnd = 0; + else { + if (a_rwnd > stcb->asoc.total_flight) { + stcb->asoc.peers_rwnd = + a_rwnd - stcb->asoc.total_flight; + } else { + stcb->asoc.peers_rwnd = 0; + } + if (stcb->asoc.peers_rwnd < + stcb->sctp_ep->sctp_ep.sctp_sws_sender) { + /* SWS sender side engages */ + stcb->asoc.peers_rwnd = 0; + } + } + } + } else { + SCTP_STAT_INCR(sctps_pdrpfmbox); + } + + /* now middle boxes in sat networks get a cwnd bump */ + if ((pktdrp_flags & SCTP_FROM_MIDDLE_BOX) && + (stcb->asoc.sat_t3_loss_recovery == 0) && + (stcb->asoc.sat_network)) { + /* + * This is debatable but for sat networks it makes sense + * Note if a T3 timer has went off, we will prohibit any + * changes to cwnd until we exit the t3 loss recovery. + */ + stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped(stcb, + net, cp, &bottle_bw, &on_queue); + } +} + +/* + * handles all control chunks in a packet inputs: - m: mbuf chain, assumed to + * still contain IP/SCTP header - stcb: is the tcb found for this packet - + * offset: offset into the mbuf chain to first chunkhdr - length: is the + * length of the complete packet outputs: - length: modified to remaining + * length after control processing - netp: modified to new sctp_nets after + * cookie-echo processing - return NULL to discard the packet (ie. no asoc, + * bad packet,...) otherwise return the tcb for this packet + */ +#ifdef __GNUC__ +__attribute__ ((noinline)) +#endif +static struct sctp_tcb * +sctp_process_control(struct mbuf *m, int iphlen, int *offset, int length, + struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, struct sctp_chunkhdr *ch, struct sctp_inpcb *inp, + struct sctp_tcb *stcb, struct sctp_nets **netp, int *fwd_tsn_seen, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, uint16_t fibnum, +#endif + uint32_t vrf_id, uint16_t port) +{ + struct sctp_association *asoc; + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + uint32_t vtag_in; + int num_chunks = 0; /* number of control chunks processed */ + uint32_t chk_length, contiguous; + int ret; + int abort_no_unlock = 0; + int ecne_seen = 0; + int abort_flag; + /* + * How big should this be, and should it be alloc'd? Lets try the + * d-mtu-ceiling for now (2k) and that should hopefully work ... + * until we get into jumbo grams and such.. + */ + uint8_t chunk_buf[SCTP_CHUNK_BUFFER_SIZE]; + int got_auth = 0; + uint32_t auth_offset = 0, auth_len = 0; + int auth_skipped = 0; + int asconf_cnt = 0; +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif + + SCTPDBG(SCTP_DEBUG_INPUT1, "sctp_process_control: iphlen=%u, offset=%u, length=%u stcb:%p\n", + iphlen, *offset, length, (void *)stcb); + + if (stcb) { + SCTP_TCB_LOCK_ASSERT(stcb); + } + /* validate chunk header length... */ + if (ntohs(ch->chunk_length) < sizeof(*ch)) { + SCTPDBG(SCTP_DEBUG_INPUT1, "Invalid header length %d\n", + ntohs(ch->chunk_length)); + *offset = length; + return (stcb); + } + /* + * validate the verification tag + */ + vtag_in = ntohl(sh->v_tag); + + if (ch->chunk_type == SCTP_INITIATION) { + SCTPDBG(SCTP_DEBUG_INPUT1, "Its an INIT of len:%d vtag:%x\n", + ntohs(ch->chunk_length), vtag_in); + if (vtag_in != 0) { + /* protocol error- silently discard... */ + SCTP_STAT_INCR(sctps_badvtag); + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + return (NULL); + } + } else if (ch->chunk_type != SCTP_COOKIE_ECHO) { + /* + * If there is no stcb, skip the AUTH chunk and process + * later after a stcb is found (to validate the lookup was + * valid. + */ + if ((ch->chunk_type == SCTP_AUTHENTICATION) && + (stcb == NULL) && + (inp->auth_supported == 1)) { + /* save this chunk for later processing */ + auth_skipped = 1; + auth_offset = *offset; + auth_len = ntohs(ch->chunk_length); + + /* (temporarily) move past this chunk */ + *offset += SCTP_SIZE32(auth_len); + if (*offset >= length) { + /* no more data left in the mbuf chain */ + *offset = length; + return (NULL); + } + ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, *offset, + sizeof(struct sctp_chunkhdr), chunk_buf); + } + if (ch == NULL) { + /* Help */ + *offset = length; + return (stcb); + } + if (ch->chunk_type == SCTP_COOKIE_ECHO) { + goto process_control_chunks; + } + /* + * first check if it's an ASCONF with an unknown src addr we + * need to look inside to find the association + */ + if (ch->chunk_type == SCTP_ASCONF && stcb == NULL) { + struct sctp_chunkhdr *asconf_ch = ch; + uint32_t asconf_offset = 0, asconf_len = 0; + + /* inp's refcount may be reduced */ + SCTP_INP_INCR_REF(inp); + + asconf_offset = *offset; + do { + asconf_len = ntohs(asconf_ch->chunk_length); + if (asconf_len < sizeof(struct sctp_asconf_paramhdr)) + break; + stcb = sctp_findassociation_ep_asconf(m, + *offset, + dst, + sh, &inp, netp, vrf_id); + if (stcb != NULL) + break; + asconf_offset += SCTP_SIZE32(asconf_len); + asconf_ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, asconf_offset, + sizeof(struct sctp_chunkhdr), chunk_buf); + } while (asconf_ch != NULL && asconf_ch->chunk_type == SCTP_ASCONF); + if (stcb == NULL) { + /* + * reduce inp's refcount if not reduced in + * sctp_findassociation_ep_asconf(). + */ + SCTP_INP_DECR_REF(inp); + } + + /* now go back and verify any auth chunk to be sure */ + if (auth_skipped && (stcb != NULL)) { + struct sctp_auth_chunk *auth; + + if (auth_len <= SCTP_CHUNK_BUFFER_SIZE) { + auth = (struct sctp_auth_chunk *)sctp_m_getptr(m, auth_offset, auth_len, chunk_buf); + got_auth = 1; + auth_skipped = 0; + } else { + auth = NULL; + } + if ((auth == NULL) || sctp_handle_auth(stcb, auth, m, + auth_offset)) { + /* auth HMAC failed so dump it */ + *offset = length; + return (stcb); + } else { + /* remaining chunks are HMAC checked */ + stcb->asoc.authenticated = 1; + } + } + } + if (stcb == NULL) { + SCTP_SNPRINTF(msg, sizeof(msg), "OOTB, %s:%d at %s", __FILE__, __LINE__, __func__); + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + msg); + /* no association, so it's out of the blue... */ + sctp_handle_ootb(m, iphlen, *offset, src, dst, sh, inp, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, port); + *offset = length; + return (NULL); + } + asoc = &stcb->asoc; + /* ABORT and SHUTDOWN can use either v_tag... */ + if ((ch->chunk_type == SCTP_ABORT_ASSOCIATION) || + (ch->chunk_type == SCTP_SHUTDOWN_COMPLETE) || + (ch->chunk_type == SCTP_PACKET_DROPPED)) { + /* Take the T-bit always into account. */ + if ((((ch->chunk_flags & SCTP_HAD_NO_TCB) == 0) && + (vtag_in == asoc->my_vtag)) || + (((ch->chunk_flags & SCTP_HAD_NO_TCB) == SCTP_HAD_NO_TCB) && + (asoc->peer_vtag != htonl(0)) && + (vtag_in == asoc->peer_vtag))) { + /* this is valid */ + } else { + /* drop this packet... */ + SCTP_STAT_INCR(sctps_badvtag); + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + return (NULL); + } + } else { + /* for all other chunks, vtag must match */ + if (vtag_in != asoc->my_vtag) { + /* invalid vtag... */ + SCTPDBG(SCTP_DEBUG_INPUT3, + "invalid vtag: %xh, expect %xh\n", + vtag_in, asoc->my_vtag); + SCTP_STAT_INCR(sctps_badvtag); + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + *offset = length; + return (NULL); + } + } + } /* end if !SCTP_COOKIE_ECHO */ + /* + * process all control chunks... + */ + if (((ch->chunk_type == SCTP_SELECTIVE_ACK) || + (ch->chunk_type == SCTP_NR_SELECTIVE_ACK) || + (ch->chunk_type == SCTP_HEARTBEAT_REQUEST)) && + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) { + /* implied cookie-ack.. we must have lost the ack */ + sctp_handle_cookie_ack((struct sctp_cookie_ack_chunk *)ch, stcb, + *netp); + } + + process_control_chunks: + while (IS_SCTP_CONTROL(ch)) { + /* validate chunk length */ + chk_length = ntohs(ch->chunk_length); + SCTPDBG(SCTP_DEBUG_INPUT2, "sctp_process_control: processing a chunk type=%u, len=%u\n", + ch->chunk_type, chk_length); + SCTP_LTRACE_CHK(inp, stcb, ch->chunk_type, chk_length); + if (chk_length < sizeof(*ch) || + (*offset + (int)chk_length) > length) { + *offset = length; + return (stcb); + } + SCTP_STAT_INCR_COUNTER64(sctps_incontrolchunks); + /* + * INIT and INIT-ACK only gets the init ack "header" portion + * only because we don't have to process the peer's COOKIE. + * All others get a complete chunk. + */ + switch (ch->chunk_type) { + case SCTP_INITIATION: + contiguous = sizeof(struct sctp_init_chunk); + break; + case SCTP_INITIATION_ACK: + contiguous = sizeof(struct sctp_init_ack_chunk); + break; + default: + contiguous = min(chk_length, sizeof(chunk_buf)); + break; + } + ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, *offset, + contiguous, + chunk_buf); + if (ch == NULL) { + *offset = length; + return (stcb); + } + + num_chunks++; + /* Save off the last place we got a control from */ + if (stcb != NULL) { + if (((netp != NULL) && (*netp != NULL)) || (ch->chunk_type == SCTP_ASCONF)) { + /* + * allow last_control to be NULL if + * ASCONF... ASCONF processing will find the + * right net later + */ + if ((netp != NULL) && (*netp != NULL)) + stcb->asoc.last_control_chunk_from = *netp; + } + } +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xB0, ch->chunk_type); +#endif + + /* check to see if this chunk required auth, but isn't */ + if ((stcb != NULL) && + sctp_auth_is_required_chunk(ch->chunk_type, stcb->asoc.local_auth_chunks) && + !stcb->asoc.authenticated) { + /* "silently" ignore */ + SCTP_STAT_INCR(sctps_recvauthmissing); + goto next_chunk; + } + switch (ch->chunk_type) { + case SCTP_INITIATION: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_INIT\n"); + /* The INIT chunk must be the only chunk. */ + if ((num_chunks > 1) || + (length - *offset > (int)SCTP_SIZE32(chk_length))) { + /* + * RFC 4960bis requires stopping the + * processing of the packet. + */ + *offset = length; + return (stcb); + } + /* Honor our resource limit. */ + if (chk_length > SCTP_LARGEST_INIT_ACCEPTED) { + op_err = sctp_generate_cause(SCTP_CAUSE_OUT_OF_RESC, ""); + sctp_send_abort(m, iphlen, src, dst, sh, 0, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, port); + *offset = length; + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + return (NULL); + } + sctp_handle_init(m, iphlen, *offset, src, dst, sh, + (struct sctp_init_chunk *)ch, inp, + stcb, *netp, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, port); + *offset = length; + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + return (NULL); + break; + case SCTP_PAD_CHUNK: + break; + case SCTP_INITIATION_ACK: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_INIT_ACK\n"); + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + /* We are not interested anymore */ + if ((stcb != NULL) && (stcb->asoc.total_output_queue_size)) { + ; + } else { + *offset = length; + if (stcb != NULL) { +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(inp); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_29); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } + return (NULL); + } + } + /* The INIT-ACK chunk must be the only chunk. */ + if ((num_chunks > 1) || + (length - *offset > (int)SCTP_SIZE32(chk_length))) { + *offset = length; + return (stcb); + } + if ((netp != NULL) && (*netp != NULL)) { + ret = sctp_handle_init_ack(m, iphlen, *offset, + src, dst, sh, + (struct sctp_init_ack_chunk *)ch, + stcb, *netp, + &abort_no_unlock, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id); + } else { + ret = -1; + } + *offset = length; + if (abort_no_unlock) { + return (NULL); + } + /* + * Special case, I must call the output routine to + * get the cookie echoed + */ + if ((stcb != NULL) && (ret == 0)) { + sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_CONTROL_PROC, SCTP_SO_NOT_LOCKED); + } + return (stcb); + break; + case SCTP_SELECTIVE_ACK: + case SCTP_NR_SELECTIVE_ACK: + { + int abort_now = 0; + uint32_t a_rwnd, cum_ack; + uint16_t num_seg, num_nr_seg, num_dup; + uint8_t flags; + int offset_seg, offset_dup; + + SCTPDBG(SCTP_DEBUG_INPUT3, "%s\n", + ch->chunk_type == SCTP_SELECTIVE_ACK ? "SCTP_SACK" : "SCTP_NR_SACK"); + SCTP_STAT_INCR(sctps_recvsacks); + if (stcb == NULL) { + SCTPDBG(SCTP_DEBUG_INDATA1, "No stcb when processing %s chunk\n", + (ch->chunk_type == SCTP_SELECTIVE_ACK) ? "SCTP_SACK" : "SCTP_NR_SACK"); + break; + } + if (ch->chunk_type == SCTP_SELECTIVE_ACK) { + if (chk_length < sizeof(struct sctp_sack_chunk)) { + SCTPDBG(SCTP_DEBUG_INDATA1, "Bad size on SACK chunk, too small\n"); + break; + } + } else { + if (stcb->asoc.nrsack_supported == 0) { + goto unknown_chunk; + } + if (chk_length < sizeof(struct sctp_nr_sack_chunk)) { + SCTPDBG(SCTP_DEBUG_INDATA1, "Bad size on NR_SACK chunk, too small\n"); + break; + } + } + if (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_ACK_SENT) { + /*- + * If we have sent a shutdown-ack, we will pay no + * attention to a sack sent in to us since + * we don't care anymore. + */ + break; + } + flags = ch->chunk_flags; + if (ch->chunk_type == SCTP_SELECTIVE_ACK) { + struct sctp_sack_chunk *sack; + + sack = (struct sctp_sack_chunk *)ch; + cum_ack = ntohl(sack->sack.cum_tsn_ack); + num_seg = ntohs(sack->sack.num_gap_ack_blks); + num_nr_seg = 0; + num_dup = ntohs(sack->sack.num_dup_tsns); + a_rwnd = ntohl(sack->sack.a_rwnd); + if (sizeof(struct sctp_sack_chunk) + + num_seg * sizeof(struct sctp_gap_ack_block) + + num_dup * sizeof(uint32_t) != chk_length) { + SCTPDBG(SCTP_DEBUG_INDATA1, "Bad size of SACK chunk\n"); + break; + } + offset_seg = *offset + sizeof(struct sctp_sack_chunk); + offset_dup = offset_seg + num_seg * sizeof(struct sctp_gap_ack_block); + } else { + struct sctp_nr_sack_chunk *nr_sack; + + nr_sack = (struct sctp_nr_sack_chunk *)ch; + cum_ack = ntohl(nr_sack->nr_sack.cum_tsn_ack); + num_seg = ntohs(nr_sack->nr_sack.num_gap_ack_blks); + num_nr_seg = ntohs(nr_sack->nr_sack.num_nr_gap_ack_blks); + num_dup = ntohs(nr_sack->nr_sack.num_dup_tsns); + a_rwnd = ntohl(nr_sack->nr_sack.a_rwnd); + if (sizeof(struct sctp_nr_sack_chunk) + + (num_seg + num_nr_seg) * sizeof(struct sctp_gap_ack_block) + + num_dup * sizeof(uint32_t) != chk_length) { + SCTPDBG(SCTP_DEBUG_INDATA1, "Bad size of NR_SACK chunk\n"); + break; + } + offset_seg = *offset + sizeof(struct sctp_nr_sack_chunk); + offset_dup = offset_seg + (num_seg + num_nr_seg) * sizeof(struct sctp_gap_ack_block); + } + SCTPDBG(SCTP_DEBUG_INPUT3, "%s process cum_ack:%x num_seg:%d a_rwnd:%d\n", + (ch->chunk_type == SCTP_SELECTIVE_ACK) ? "SCTP_SACK" : "SCTP_NR_SACK", + cum_ack, num_seg, a_rwnd); + stcb->asoc.seen_a_sack_this_pkt = 1; + if ((stcb->asoc.pr_sctp_cnt == 0) && + (num_seg == 0) && (num_nr_seg == 0) && + SCTP_TSN_GE(cum_ack, stcb->asoc.last_acked_seq) && + (stcb->asoc.saw_sack_with_frags == 0) && + (stcb->asoc.saw_sack_with_nr_frags == 0) && + (!TAILQ_EMPTY(&stcb->asoc.sent_queue))) { + /* + * We have a SIMPLE sack having no + * prior segments and data on sent + * queue to be acked. Use the + * faster path sack processing. We + * also allow window update sacks + * with no missing segments to go + * this way too. + */ + sctp_express_handle_sack(stcb, cum_ack, a_rwnd, + &abort_now, ecne_seen); + } else { + if ((netp != NULL) && (*netp != NULL)) { + sctp_handle_sack(m, offset_seg, offset_dup, stcb, + num_seg, num_nr_seg, num_dup, &abort_now, flags, + cum_ack, a_rwnd, ecne_seen); + } + } + if (abort_now) { + /* ABORT signal from sack processing */ + *offset = length; + return (NULL); + } + if (TAILQ_EMPTY(&stcb->asoc.send_queue) && + TAILQ_EMPTY(&stcb->asoc.sent_queue) && + (stcb->asoc.stream_queue_cnt == 0)) { + sctp_ulp_notify(SCTP_NOTIFY_SENDER_DRY, stcb, 0, NULL, SCTP_SO_NOT_LOCKED); + } + break; + } + case SCTP_HEARTBEAT_REQUEST: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_HEARTBEAT\n"); + if ((stcb != NULL) && (netp != NULL) && (*netp != NULL)) { + SCTP_STAT_INCR(sctps_recvheartbeat); + sctp_send_heartbeat_ack(stcb, m, *offset, + chk_length, *netp); + } + break; + case SCTP_HEARTBEAT_ACK: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_HEARTBEAT_ACK\n"); + if ((stcb == NULL) || (chk_length != sizeof(struct sctp_heartbeat_chunk))) { + /* Its not ours */ + break; + } + SCTP_STAT_INCR(sctps_recvheartbeatack); + if ((netp != NULL) && (*netp != NULL)) { + sctp_handle_heartbeat_ack((struct sctp_heartbeat_chunk *)ch, + stcb, *netp); + } + break; + case SCTP_ABORT_ASSOCIATION: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_ABORT, stcb %p\n", + (void *)stcb); + *offset = length; + if ((stcb != NULL) && (netp != NULL) && (*netp != NULL)) { + if (sctp_handle_abort((struct sctp_abort_chunk *)ch, stcb, *netp)) { + return (NULL); + } else { + return (stcb); + } + } else { + return (NULL); + } + break; + case SCTP_SHUTDOWN: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_SHUTDOWN, stcb %p\n", + (void *)stcb); + if ((stcb == NULL) || (chk_length != sizeof(struct sctp_shutdown_chunk))) { + break; + } + if ((netp != NULL) && (*netp != NULL)) { + abort_flag = 0; + sctp_handle_shutdown((struct sctp_shutdown_chunk *)ch, + stcb, *netp, &abort_flag); + if (abort_flag) { + *offset = length; + return (NULL); + } + } + break; + case SCTP_SHUTDOWN_ACK: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_SHUTDOWN_ACK, stcb %p\n", (void *)stcb); + if ((chk_length == sizeof(struct sctp_shutdown_ack_chunk)) && + (stcb != NULL) && (netp != NULL) && (*netp != NULL)) { + sctp_handle_shutdown_ack((struct sctp_shutdown_ack_chunk *)ch, stcb, *netp); + *offset = length; + return (NULL); + } + break; + case SCTP_OPERATION_ERROR: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_OP_ERR\n"); + if ((stcb != NULL) && (netp != NULL) && (*netp != NULL) && + sctp_handle_error(ch, stcb, *netp, contiguous) < 0) { + *offset = length; + return (NULL); + } + break; + case SCTP_COOKIE_ECHO: + SCTPDBG(SCTP_DEBUG_INPUT3, + "SCTP_COOKIE_ECHO, stcb %p\n", (void *)stcb); + if ((stcb != NULL) && (stcb->asoc.total_output_queue_size > 0)) { + ; + } else { + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + /* We are not interested anymore */ + abend: + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + *offset = length; + return (NULL); + } + } + /*- + * First are we accepting? We do this again here + * since it is possible that a previous endpoint WAS + * listening responded to a INIT-ACK and then + * closed. We opened and bound.. and are now no + * longer listening. + * + * XXXGL: notes on checking listen queue length. + * 1) SCTP_IS_LISTENING() doesn't necessarily mean + * SOLISTENING(), because a listening "UDP type" + * socket isn't listening in terms of the socket + * layer. It is a normal data flow socket, that + * can fork off new connections. Thus, we should + * look into sol_qlen only in case we are !UDP. + * 2) Checking sol_qlen in general requires locking + * the socket, and this code lacks that. + */ + if ((stcb == NULL) && + (!SCTP_IS_LISTENING(inp) || + (((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) == 0) && +#if defined(__FreeBSD__) && !defined(__Userspace__) + inp->sctp_socket->sol_qlen >= inp->sctp_socket->sol_qlimit))) { +#else + inp->sctp_socket->so_qlen >= inp->sctp_socket->so_qlimit))) { +#endif + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && + (SCTP_BASE_SYSCTL(sctp_abort_if_one_2_one_hits_limit))) { + op_err = sctp_generate_cause(SCTP_CAUSE_OUT_OF_RESC, ""); + sctp_abort_association(inp, stcb, m, iphlen, + src, dst, sh, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + vrf_id, port); + } + *offset = length; + return (NULL); + } else { + struct mbuf *ret_buf; + struct sctp_inpcb *linp; + struct sctp_tmit_chunk *chk; + + if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | + SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { + goto abend; + } + + if (stcb) { + linp = NULL; + } else { + linp = inp; + } + + if (linp != NULL) { + SCTP_ASOC_CREATE_LOCK(linp); + } + + if (netp != NULL) { + struct sctp_tcb *locked_stcb; + + locked_stcb = stcb; + ret_buf = + sctp_handle_cookie_echo(m, iphlen, + *offset, + src, dst, + sh, + (struct sctp_cookie_echo_chunk *)ch, + &inp, &stcb, netp, + auth_skipped, + auth_offset, + auth_len, + &locked_stcb, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, + mflowid, +#endif + vrf_id, + port); + if ((locked_stcb != NULL) && (locked_stcb != stcb)) { + SCTP_TCB_UNLOCK(locked_stcb); + } + if (stcb != NULL) { + SCTP_TCB_LOCK_ASSERT(stcb); + } + } else { + ret_buf = NULL; + } + if (linp != NULL) { + SCTP_ASOC_CREATE_UNLOCK(linp); + } + if (ret_buf == NULL) { + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + SCTPDBG(SCTP_DEBUG_INPUT3, + "GAK, null buffer\n"); + *offset = length; + return (NULL); + } + /* if AUTH skipped, see if it verified... */ + if (auth_skipped) { + got_auth = 1; + auth_skipped = 0; + } + /* Restart the timer if we have pending data */ + TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { + if (chk->whoTo != NULL) { + break; + } + } + if (chk != NULL) { + sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, chk->whoTo); + } + } + break; + case SCTP_COOKIE_ACK: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_COOKIE_ACK, stcb %p\n", (void *)stcb); + if ((stcb == NULL) || chk_length != sizeof(struct sctp_cookie_ack_chunk)) { + break; + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + /* We are not interested anymore */ + if ((stcb) && (stcb->asoc.total_output_queue_size)) { + ; + } else if (stcb) { +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(inp); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_30); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + *offset = length; + return (NULL); + } + } + if ((netp != NULL) && (*netp != NULL)) { + sctp_handle_cookie_ack((struct sctp_cookie_ack_chunk *)ch, stcb, *netp); + } + break; + case SCTP_ECN_ECHO: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_ECN_ECHO\n"); + if (stcb == NULL) { + break; + } + if (stcb->asoc.ecn_supported == 0) { + goto unknown_chunk; + } + if ((chk_length != sizeof(struct sctp_ecne_chunk)) && + (chk_length != sizeof(struct old_sctp_ecne_chunk))) { + break; + } + sctp_handle_ecn_echo((struct sctp_ecne_chunk *)ch, stcb); + ecne_seen = 1; + break; + case SCTP_ECN_CWR: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_ECN_CWR\n"); + if (stcb == NULL) { + break; + } + if (stcb->asoc.ecn_supported == 0) { + goto unknown_chunk; + } + if (chk_length != sizeof(struct sctp_cwr_chunk)) { + break; + } + sctp_handle_ecn_cwr((struct sctp_cwr_chunk *)ch, stcb, *netp); + break; + case SCTP_SHUTDOWN_COMPLETE: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_SHUTDOWN_COMPLETE, stcb %p\n", (void *)stcb); + /* must be first and only chunk */ + if ((num_chunks > 1) || + (length - *offset > (int)SCTP_SIZE32(chk_length))) { + *offset = length; + return (stcb); + } + if ((chk_length == sizeof(struct sctp_shutdown_complete_chunk)) && + (stcb != NULL) && (netp != NULL) && (*netp != NULL)) { + sctp_handle_shutdown_complete((struct sctp_shutdown_complete_chunk *)ch, + stcb, *netp); + *offset = length; + return (NULL); + } + break; + case SCTP_ASCONF: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_ASCONF\n"); + if (stcb != NULL) { + if (stcb->asoc.asconf_supported == 0) { + goto unknown_chunk; + } + sctp_handle_asconf(m, *offset, src, + (struct sctp_asconf_chunk *)ch, stcb, asconf_cnt == 0); + asconf_cnt++; + } + break; + case SCTP_ASCONF_ACK: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_ASCONF_ACK\n"); + if (stcb == NULL) { + break; + } + if (stcb->asoc.asconf_supported == 0) { + goto unknown_chunk; + } + if (chk_length < sizeof(struct sctp_asconf_ack_chunk)) { + break; + } + if ((netp != NULL) && (*netp != NULL)) { + /* He's alive so give him credit */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { + sctp_misc_ints(SCTP_THRESHOLD_CLEAR, + stcb->asoc.overall_error_count, + 0, + SCTP_FROM_SCTP_INPUT, + __LINE__); + } + stcb->asoc.overall_error_count = 0; + sctp_handle_asconf_ack(m, *offset, + (struct sctp_asconf_ack_chunk *)ch, stcb, *netp, &abort_no_unlock); + if (abort_no_unlock) + return (NULL); + } + break; + case SCTP_FORWARD_CUM_TSN: + case SCTP_IFORWARD_CUM_TSN: + SCTPDBG(SCTP_DEBUG_INPUT3, "%s\n", + ch->chunk_type == SCTP_FORWARD_CUM_TSN ? "FORWARD_TSN" : "I_FORWARD_TSN"); + if (stcb == NULL) { + break; + } + if (stcb->asoc.prsctp_supported == 0) { + goto unknown_chunk; + } + if (chk_length < sizeof(struct sctp_forward_tsn_chunk)) { + break; + } + if (((stcb->asoc.idata_supported == 1) && (ch->chunk_type == SCTP_FORWARD_CUM_TSN)) || + ((stcb->asoc.idata_supported == 0) && (ch->chunk_type == SCTP_IFORWARD_CUM_TSN))) { + if (ch->chunk_type == SCTP_FORWARD_CUM_TSN) { + SCTP_SNPRINTF(msg, sizeof(msg), "%s", "FORWARD-TSN chunk received when I-FORWARD-TSN was negotiated"); + } else { + SCTP_SNPRINTF(msg, sizeof(msg), "%s", "I-FORWARD-TSN chunk received when FORWARD-TSN was negotiated"); + } + op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, msg); + sctp_abort_an_association(inp, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + *offset = length; + return (NULL); + } + *fwd_tsn_seen = 1; + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + /* We are not interested anymore */ +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(inp); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_INPUT + SCTP_LOC_31); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + *offset = length; + return (NULL); + } + /* + * For sending a SACK this looks like DATA + * chunks. + */ + stcb->asoc.last_data_chunk_from = stcb->asoc.last_control_chunk_from; + abort_flag = 0; + sctp_handle_forward_tsn(stcb, + (struct sctp_forward_tsn_chunk *)ch, &abort_flag, m, *offset); + if (abort_flag) { + *offset = length; + return (NULL); + } + break; + case SCTP_STREAM_RESET: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_STREAM_RESET\n"); + if (stcb == NULL) { + break; + } + if (stcb->asoc.reconfig_supported == 0) { + goto unknown_chunk; + } + if (chk_length < sizeof(struct sctp_stream_reset_tsn_req)) { + break; + } + if (sctp_handle_stream_reset(stcb, m, *offset, ch)) { + /* stop processing */ + *offset = length; + return (NULL); + } + break; + case SCTP_PACKET_DROPPED: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_PACKET_DROPPED\n"); + if (stcb == NULL) { + break; + } + if (stcb->asoc.pktdrop_supported == 0) { + goto unknown_chunk; + } + if (chk_length < sizeof(struct sctp_pktdrop_chunk)) { + break; + } + if ((netp != NULL) && (*netp != NULL)) { + sctp_handle_packet_dropped((struct sctp_pktdrop_chunk *)ch, + stcb, *netp, + min(chk_length, contiguous)); + } + break; + case SCTP_AUTHENTICATION: + SCTPDBG(SCTP_DEBUG_INPUT3, "SCTP_AUTHENTICATION\n"); + if (stcb == NULL) { + /* save the first AUTH for later processing */ + if (auth_skipped == 0) { + auth_offset = *offset; + auth_len = chk_length; + auth_skipped = 1; + } + /* skip this chunk (temporarily) */ + break; + } + if (stcb->asoc.auth_supported == 0) { + goto unknown_chunk; + } + if ((chk_length < (sizeof(struct sctp_auth_chunk))) || + (chk_length > (sizeof(struct sctp_auth_chunk) + + SCTP_AUTH_DIGEST_LEN_MAX))) { + /* Its not ours */ + *offset = length; + return (stcb); + } + if (got_auth == 1) { + /* skip this chunk... it's already auth'd */ + break; + } + got_auth = 1; + if (sctp_handle_auth(stcb, (struct sctp_auth_chunk *)ch, m, *offset)) { + /* auth HMAC failed so dump the packet */ + *offset = length; + return (stcb); + } else { + /* remaining chunks are HMAC checked */ + stcb->asoc.authenticated = 1; + } + break; + + default: + unknown_chunk: + /* it's an unknown chunk! */ + if ((ch->chunk_type & 0x40) && + (stcb != NULL) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_EMPTY) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_INUSE) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_COOKIE_WAIT)) { + struct sctp_gen_error_cause *cause; + int len; + + op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_gen_error_cause), + 0, M_NOWAIT, 1, MT_DATA); + if (op_err != NULL) { + len = min(SCTP_SIZE32(chk_length), (uint32_t)(length - *offset)); + cause = mtod(op_err, struct sctp_gen_error_cause *); + cause->code = htons(SCTP_CAUSE_UNRECOG_CHUNK); + cause->length = htons((uint16_t)(len + sizeof(struct sctp_gen_error_cause))); + SCTP_BUF_LEN(op_err) = sizeof(struct sctp_gen_error_cause); + SCTP_BUF_NEXT(op_err) = SCTP_M_COPYM(m, *offset, len, M_NOWAIT); + if (SCTP_BUF_NEXT(op_err) != NULL) { +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(SCTP_BUF_NEXT(op_err), SCTP_MBUF_ICOPY); + } +#endif + sctp_queue_op_err(stcb, op_err); + } else { + sctp_m_freem(op_err); + } + } + } + if ((ch->chunk_type & 0x80) == 0) { + /* discard this packet */ + *offset = length; + return (stcb); + } /* else skip this bad chunk and continue... */ + break; + } /* switch (ch->chunk_type) */ + + next_chunk: + /* get the next chunk */ + *offset += SCTP_SIZE32(chk_length); + if (*offset >= length) { + /* no more data left in the mbuf chain */ + break; + } + ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, *offset, + sizeof(struct sctp_chunkhdr), chunk_buf); + if (ch == NULL) { + *offset = length; + return (stcb); + } + } /* while */ + + if ((asconf_cnt > 0) && (stcb != NULL)) { + sctp_send_asconf_ack(stcb); + } + return (stcb); +} + +/* + * common input chunk processing (v4 and v6) + */ +void +sctp_common_input_processing(struct mbuf **mm, int iphlen, int offset, int length, + struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, struct sctp_chunkhdr *ch, + uint8_t compute_crc, + uint8_t ecn_bits, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, uint16_t fibnum, +#endif + uint32_t vrf_id, uint16_t port) +{ + char msg[SCTP_DIAG_INFO_LEN]; + struct mbuf *m = *mm, *op_err; + struct sctp_inpcb *inp = NULL, *inp_decr = NULL; + struct sctp_tcb *stcb = NULL; + struct sctp_nets *net = NULL; +#if defined(__Userspace__) + struct socket *upcall_socket = NULL; +#endif + uint32_t high_tsn; + uint32_t cksum_in_hdr; + int un_sent; + int cnt_ctrl_ready = 0; + int fwd_tsn_seen = 0, data_processed = 0; + bool cksum_validated, stcb_looked_up; + + SCTP_STAT_INCR(sctps_recvdatagrams); +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xE0, 1); + sctp_auditing(0, inp, stcb, net); +#endif + + stcb_looked_up = false; + if (compute_crc != 0) { + cksum_validated = false; + cksum_in_hdr = sh->checksum; + if (cksum_in_hdr != htonl(0)) { + uint32_t cksum_calculated; + +validate_cksum: + sh->checksum = 0; + cksum_calculated = sctp_calculate_cksum(m, iphlen); + sh->checksum = cksum_in_hdr; + if (cksum_calculated != cksum_in_hdr) { + if (stcb_looked_up) { + /* + * The packet has a zero checksum, which + * is not the correct CRC, no stcb has + * been found or an stcb has been found + * but an incorrect zero checksum is not + * acceptable. + */ + KASSERT(cksum_in_hdr == htonl(0), + ("cksum in header not zero: %x", + ntohl(cksum_in_hdr))); + if ((inp == NULL) && + (SCTP_BASE_SYSCTL(sctp_ootb_with_zero_cksum) == 1)) { + /* + * This is an OOTB packet, + * depending on the sysctl + * variable, pretend that the + * checksum is acceptable, + * to allow an appropriate + * response (ABORT, for examlpe) + * to be sent. + */ + KASSERT(stcb == NULL, + ("stcb is %p", stcb)); + SCTP_STAT_INCR(sctps_recvzerocrc); + goto cksum_validated; + } + } else { + stcb = sctp_findassociation_addr(m, offset, src, dst, + sh, ch, &inp, &net, vrf_id); + } + SCTPDBG(SCTP_DEBUG_INPUT1, "Bad cksum in SCTP packet:%x calculated:%x m:%p mlen:%d iphlen:%d\n", + ntohl(cksum_in_hdr), ntohl(cksum_calculated), (void *)m, length, iphlen); +#if defined(INET) || defined(INET6) + if ((ch->chunk_type != SCTP_INITIATION) && + (net != NULL) && (net->port != port)) { + if (net->port == 0) { + /* UDP encapsulation turned on. */ + net->mtu -= sizeof(struct udphdr); + if (stcb->asoc.smallest_mtu > net->mtu) { + sctp_pathmtu_adjustment(stcb, net->mtu, true); + } + } else if (port == 0) { + /* UDP encapsulation turned off. */ + net->mtu += sizeof(struct udphdr); + /* XXX Update smallest_mtu */ + } + net->port = port; + } +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (net != NULL) { + net->flowtype = mflowtype; + net->flowid = mflowid; + } + SCTP_PROBE5(receive, NULL, stcb, m, stcb, sh); +#endif + if ((inp != NULL) && (stcb != NULL)) { + if (stcb->asoc.pktdrop_supported) { + sctp_send_packet_dropped(stcb, net, m, length, iphlen, 1); + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_INPUT_ERROR, SCTP_SO_NOT_LOCKED); + } + } else if ((inp != NULL) && (stcb == NULL)) { + inp_decr = inp; + } + SCTP_STAT_INCR(sctps_badsum); + SCTP_STAT_INCR_COUNTER32(sctps_checksumerrors); + goto out; + } else { + cksum_validated = true; + } + } + KASSERT(cksum_validated || cksum_in_hdr == htonl(0), + ("cksum 0x%08x not zero and not validated", ntohl(cksum_in_hdr))); + if (!cksum_validated) { + stcb = sctp_findassociation_addr(m, offset, src, dst, + sh, ch, &inp, &net, vrf_id); + stcb_looked_up = true; + if ((stcb == NULL) || (stcb->asoc.zero_checksum == 0)) { + goto validate_cksum; + } + SCTP_STAT_INCR(sctps_recvzerocrc); + } + } +cksum_validated: + /* Destination port of 0 is illegal, based on RFC4960. */ + if (sh->dest_port == htons(0)) { + SCTP_STAT_INCR(sctps_hdrops); + if ((stcb == NULL) && (inp != NULL)) { + inp_decr = inp; + } + goto out; + } + if (!stcb_looked_up) { + stcb = sctp_findassociation_addr(m, offset, src, dst, + sh, ch, &inp, &net, vrf_id); + } +#if defined(INET) || defined(INET6) + if ((ch->chunk_type != SCTP_INITIATION) && + (net != NULL) && (net->port != port)) { + if (net->port == 0) { + /* UDP encapsulation turned on. */ + net->mtu -= sizeof(struct udphdr); + if (stcb->asoc.smallest_mtu > net->mtu) { + sctp_pathmtu_adjustment(stcb, net->mtu, true); + } + } else if (port == 0) { + /* UDP encapsulation turned off. */ + net->mtu += sizeof(struct udphdr); + /* XXX Update smallest_mtu */ + } + net->port = port; + } +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (net != NULL) { + net->flowtype = mflowtype; + net->flowid = mflowid; + } +#endif + if (inp == NULL) { + SCTP_STAT_INCR(sctps_noport); +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_PROBE5(receive, NULL, stcb, m, stcb, sh); + if (badport_bandlim(BANDLIM_SCTP_OOTB) < 0) { + goto out; + } +#endif + if (ch->chunk_type == SCTP_SHUTDOWN_ACK) { + SCTP_STAT_INCR_COUNTER64(sctps_incontrolchunks); + sctp_send_shutdown_complete2(src, dst, sh, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, fibnum, +#endif + vrf_id, port); + goto out; + } + if (ch->chunk_type == SCTP_SHUTDOWN_COMPLETE) { + SCTP_STAT_INCR_COUNTER64(sctps_incontrolchunks); + goto out; + } + if (ch->chunk_type != SCTP_ABORT_ASSOCIATION) { + if ((SCTP_BASE_SYSCTL(sctp_blackhole) == 0) || + ((SCTP_BASE_SYSCTL(sctp_blackhole) == 1) && + (ch->chunk_type != SCTP_INIT))) { + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + "Out of the blue"); + sctp_send_abort(m, iphlen, src, dst, + sh, 0, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, fibnum, +#endif + vrf_id, port); + } + } + goto out; + } else if (stcb == NULL) { + inp_decr = inp; + } + SCTPDBG(SCTP_DEBUG_INPUT1, "Ok, Common input processing called, m:%p iphlen:%d offset:%d length:%d stcb:%p\n", + (void *)m, iphlen, offset, length, (void *)stcb); + if (stcb) { + /* always clear this before beginning a packet */ + stcb->asoc.authenticated = 0; + stcb->asoc.seen_a_sack_this_pkt = 0; + SCTPDBG(SCTP_DEBUG_INPUT1, "stcb:%p state:%x\n", + (void *)stcb, stcb->asoc.state); + + if ((stcb->asoc.state & SCTP_STATE_WAS_ABORTED) || + (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED)) { + /*- + * If we hit here, we had a ref count + * up when the assoc was aborted and the + * timer is clearing out the assoc, we should + * NOT respond to any packet.. its OOTB. + */ + SCTP_TCB_UNLOCK(stcb); + stcb = NULL; +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_PROBE5(receive, NULL, stcb, m, stcb, sh); +#endif + SCTP_SNPRINTF(msg, sizeof(msg), "OOTB, %s:%d at %s", __FILE__, __LINE__, __func__); + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + msg); + sctp_handle_ootb(m, iphlen, offset, src, dst, sh, inp, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, port); + goto out; + } + } +#if defined(__Userspace__) + if ((stcb != NULL) && + ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (stcb->sctp_socket != NULL)) { + if (stcb->sctp_socket->so_head != NULL) { + upcall_socket = stcb->sctp_socket->so_head; + } else { + upcall_socket = stcb->sctp_socket; + } + SOCK_LOCK(upcall_socket); + soref(upcall_socket); + SOCK_UNLOCK(upcall_socket); + } +#endif + if (IS_SCTP_CONTROL(ch)) { + /* process the control portion of the SCTP packet */ + /* sa_ignore NO_NULL_CHK */ + stcb = sctp_process_control(m, iphlen, &offset, length, + src, dst, sh, ch, + inp, stcb, &net, &fwd_tsn_seen, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, fibnum, +#endif + vrf_id, port); + if (stcb) { + /* This covers us if the cookie-echo was there + * and it changes our INP. + */ + inp = stcb->sctp_ep; +#if defined(INET) || defined(INET6) + if ((ch->chunk_type != SCTP_INITIATION) && + (net != NULL) && (net->port != port)) { + if (net->port == 0) { + /* UDP encapsulation turned on. */ + net->mtu -= sizeof(struct udphdr); + if (stcb->asoc.smallest_mtu > net->mtu) { + sctp_pathmtu_adjustment(stcb, net->mtu, true); + } + } else if (port == 0) { + /* UDP encapsulation turned off. */ + net->mtu += sizeof(struct udphdr); + /* XXX Update smallest_mtu */ + } + net->port = port; + } +#endif + } + } else { + /* + * no control chunks, so pre-process DATA chunks (these + * checks are taken care of by control processing) + */ + + /* + * if DATA only packet, and auth is required, then punt... + * can't have authenticated without any AUTH (control) + * chunks + */ + if ((stcb != NULL) && + sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.local_auth_chunks)) { + /* "silently" ignore */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_PROBE5(receive, NULL, stcb, m, stcb, sh); +#endif + SCTP_STAT_INCR(sctps_recvauthmissing); + goto out; + } + if (stcb == NULL) { + /* out of the blue DATA chunk */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_PROBE5(receive, NULL, NULL, m, NULL, sh); +#endif + SCTP_SNPRINTF(msg, sizeof(msg), "OOTB, %s:%d at %s", __FILE__, __LINE__, __func__); + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + msg); + sctp_handle_ootb(m, iphlen, offset, src, dst, sh, inp, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, fibnum, +#endif + vrf_id, port); + goto out; + } + if (stcb->asoc.my_vtag != ntohl(sh->v_tag)) { + /* v_tag mismatch! */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_PROBE5(receive, NULL, stcb, m, stcb, sh); +#endif + SCTP_STAT_INCR(sctps_badvtag); + goto out; + } + } + +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_PROBE5(receive, NULL, stcb, m, stcb, sh); +#endif + if (stcb == NULL) { + /* + * no valid TCB for this packet, or we found it's a bad + * packet while processing control, or we're done with this + * packet (done or skip rest of data), so we drop it... + */ + goto out; + } +#if defined(__Userspace__) + if ((upcall_socket == NULL) && + ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (stcb->sctp_socket != NULL)) { + if (stcb->sctp_socket->so_head != NULL) { + upcall_socket = stcb->sctp_socket->so_head; + } else { + upcall_socket = stcb->sctp_socket; + } + SOCK_LOCK(upcall_socket); + soref(upcall_socket); + SOCK_UNLOCK(upcall_socket); + } +#endif + + /* + * DATA chunk processing + */ + /* plow through the data chunks while length > offset */ + + /* + * Rest should be DATA only. Check authentication state if AUTH for + * DATA is required. + */ + if ((length > offset) && + (stcb != NULL) && + sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.local_auth_chunks) && + !stcb->asoc.authenticated) { + /* "silently" ignore */ + SCTP_STAT_INCR(sctps_recvauthmissing); + SCTPDBG(SCTP_DEBUG_AUTH1, + "Data chunk requires AUTH, skipped\n"); + goto trigger_send; + } + if (length > offset) { + int retval; + + /* + * First check to make sure our state is correct. We would + * not get here unless we really did have a tag, so we don't + * abort if this happens, just dump the chunk silently. + */ + switch (SCTP_GET_STATE(stcb)) { + case SCTP_STATE_COOKIE_ECHOED: + /* + * we consider data with valid tags in this state + * shows us the cookie-ack was lost. Imply it was + * there. + */ + sctp_handle_cookie_ack((struct sctp_cookie_ack_chunk *)ch, stcb, net); + break; + case SCTP_STATE_COOKIE_WAIT: + /* + * We consider OOTB any data sent during asoc setup. + */ + SCTP_SNPRINTF(msg, sizeof(msg), "OOTB, %s:%d at %s", __FILE__, __LINE__, __func__); + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + msg); + sctp_handle_ootb(m, iphlen, offset, src, dst, sh, inp, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, port); + goto out; + /*sa_ignore NOTREACHED*/ + break; + case SCTP_STATE_EMPTY: /* should not happen */ + case SCTP_STATE_INUSE: /* should not happen */ + case SCTP_STATE_SHUTDOWN_RECEIVED: /* This is a peer error */ + case SCTP_STATE_SHUTDOWN_ACK_SENT: + default: + goto out; + /*sa_ignore NOTREACHED*/ + break; + case SCTP_STATE_OPEN: + case SCTP_STATE_SHUTDOWN_SENT: + break; + } + /* plow through the data chunks while length > offset */ + retval = sctp_process_data(mm, iphlen, &offset, length, + inp, stcb, net, &high_tsn); + if (retval == 2) { + /* + * The association aborted, NO UNLOCK needed since + * the association is destroyed. + */ + stcb = NULL; + goto out; + } + if (retval == 0) { + data_processed = 1; + } + /* + * Anything important needs to have been m_copy'ed in + * process_data + */ + } + + /* take care of ecn */ + if ((data_processed == 1) && + (stcb->asoc.ecn_supported == 1) && + ((ecn_bits & SCTP_CE_BITS) == SCTP_CE_BITS)) { + /* Yep, we need to add a ECNE */ + sctp_send_ecn_echo(stcb, net, high_tsn); + } + + if ((data_processed == 0) && (fwd_tsn_seen)) { + int was_a_gap; + uint32_t highest_tsn; + + if (SCTP_TSN_GT(stcb->asoc.highest_tsn_inside_nr_map, stcb->asoc.highest_tsn_inside_map)) { + highest_tsn = stcb->asoc.highest_tsn_inside_nr_map; + } else { + highest_tsn = stcb->asoc.highest_tsn_inside_map; + } + was_a_gap = SCTP_TSN_GT(highest_tsn, stcb->asoc.cumulative_tsn); + stcb->asoc.send_sack = 1; + sctp_sack_check(stcb, was_a_gap); + } else if (fwd_tsn_seen) { + stcb->asoc.send_sack = 1; + } + /* trigger send of any chunks in queue... */ +trigger_send: +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xE0, 2); + sctp_auditing(1, inp, stcb, net); +#endif + SCTPDBG(SCTP_DEBUG_INPUT1, + "Check for chunk output prw:%d tqe:%d tf=%d\n", + stcb->asoc.peers_rwnd, + TAILQ_EMPTY(&stcb->asoc.control_send_queue), + stcb->asoc.total_flight); + un_sent = (stcb->asoc.total_output_queue_size - stcb->asoc.total_flight); + if (!TAILQ_EMPTY(&stcb->asoc.control_send_queue)) { + cnt_ctrl_ready = stcb->asoc.ctrl_queue_cnt - stcb->asoc.ecn_echo_cnt_onq; + } + if (!TAILQ_EMPTY(&stcb->asoc.asconf_send_queue) || + cnt_ctrl_ready || + stcb->asoc.trigger_reset || + ((un_sent > 0) && + (stcb->asoc.peers_rwnd > 0 || stcb->asoc.total_flight == 0))) { + SCTPDBG(SCTP_DEBUG_INPUT3, "Calling chunk OUTPUT\n"); + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CONTROL_PROC, SCTP_SO_NOT_LOCKED); + SCTPDBG(SCTP_DEBUG_INPUT3, "chunk OUTPUT returns\n"); + } +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xE0, 3); + sctp_auditing(2, inp, stcb, net); +#endif + out: + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } +#if defined(__Userspace__) + if (upcall_socket != NULL) { + if (upcall_socket->so_upcall != NULL) { + if (soreadable(upcall_socket) || + sowriteable(upcall_socket) || + upcall_socket->so_error) { + (*upcall_socket->so_upcall)(upcall_socket, upcall_socket->so_upcallarg, M_NOWAIT); + } + } + ACCEPT_LOCK(); + SOCK_LOCK(upcall_socket); + sorele(upcall_socket); + } +#endif + if (inp_decr != NULL) { + /* reduce ref-count */ + SCTP_INP_WLOCK(inp_decr); + SCTP_INP_DECR_REF(inp_decr); + SCTP_INP_WUNLOCK(inp_decr); + } + return; +} + +#ifdef INET +#if !defined(__Userspace__) +void +sctp_input_with_port(struct mbuf *i_pak, int off, uint16_t port) +{ + struct mbuf *m; + int iphlen; + uint32_t vrf_id = 0; + uint8_t ecn_bits; + struct sockaddr_in src, dst; + struct ip *ip; + struct sctphdr *sh; + struct sctp_chunkhdr *ch; + int length, offset; + uint8_t compute_crc; +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint32_t mflowid; + uint8_t mflowtype; + uint16_t fibnum; +#endif +#if defined(__Userspace__) + uint16_t port = 0; +#endif + + iphlen = off; + if (SCTP_GET_PKT_VRFID(i_pak, vrf_id)) { + SCTP_RELEASE_PKT(i_pak); + return; + } + m = SCTP_HEADER_TO_CHAIN(i_pak); +#ifdef SCTP_MBUF_LOGGING + /* Log in any input mbufs */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(m, SCTP_MBUF_INPUT); + } +#endif +#ifdef SCTP_PACKET_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) { + sctp_packet_log(m); + } +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTPDBG(SCTP_DEBUG_CRCOFFLOAD, + "sctp_input(): Packet of length %d received on %s with csum_flags 0x%b.\n", + m->m_pkthdr.len, + if_name(m->m_pkthdr.rcvif), + (int)m->m_pkthdr.csum_flags, CSUM_BITS); +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + SCTPDBG(SCTP_DEBUG_CRCOFFLOAD, + "sctp_input(): Packet of length %d received on %s%d with csum_flags 0x%x.\n", + m->m_pkthdr.len, + m->m_pkthdr.rcvif->if_name, + m->m_pkthdr.rcvif->if_unit, + m->m_pkthdr.csum_flags); +#endif +#if defined(_WIN32) && !defined(__Userspace__) + SCTPDBG(SCTP_DEBUG_CRCOFFLOAD, + "sctp_input(): Packet of length %d received on %s with csum_flags 0x%x.\n", + m->m_pkthdr.len, + m->m_pkthdr.rcvif->if_xname, + m->m_pkthdr.csum_flags); +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowid = m->m_pkthdr.flowid; + mflowtype = M_HASHTYPE_GET(m); + fibnum = M_GETFIB(m); +#endif + SCTP_STAT_INCR(sctps_recvpackets); + SCTP_STAT_INCR_COUNTER64(sctps_inpackets); + /* Get IP, SCTP, and first chunk header together in the first mbuf. */ + offset = iphlen + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); + if (SCTP_BUF_LEN(m) < offset) { + if ((m = m_pullup(m, offset)) == NULL) { + SCTP_STAT_INCR(sctps_hdrops); + return; + } + } + ip = mtod(m, struct ip *); + sh = (struct sctphdr *)((caddr_t)ip + iphlen); + ch = (struct sctp_chunkhdr *)((caddr_t)sh + sizeof(struct sctphdr)); + offset -= sizeof(struct sctp_chunkhdr); + memset(&src, 0, sizeof(struct sockaddr_in)); + src.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + src.sin_len = sizeof(struct sockaddr_in); +#endif + src.sin_port = sh->src_port; + src.sin_addr = ip->ip_src; + memset(&dst, 0, sizeof(struct sockaddr_in)); + dst.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + dst.sin_len = sizeof(struct sockaddr_in); +#endif + dst.sin_port = sh->dest_port; + dst.sin_addr = ip->ip_dst; +#if defined(_WIN32) && !defined(__Userspace__) + NTOHS(ip->ip_len); +#endif +#if defined(__linux__) || (defined(_WIN32) && defined(__Userspace__)) + ip->ip_len = ntohs(ip->ip_len); +#endif +#if defined(__Userspace__) +#if defined(__linux__) || defined(_WIN32) + length = ip->ip_len; +#else + length = ip->ip_len + iphlen; +#endif +#elif defined(__FreeBSD__) + length = ntohs(ip->ip_len); +#elif defined(__APPLE__) + length = ip->ip_len + iphlen; +#else + length = ip->ip_len; +#endif + /* Validate mbuf chain length with IP payload length. */ + if (SCTP_HEADER_LEN(m) != length) { + SCTPDBG(SCTP_DEBUG_INPUT1, + "sctp_input() length:%d reported length:%d\n", length, SCTP_HEADER_LEN(m)); + SCTP_STAT_INCR(sctps_hdrops); + goto out; + } + /* SCTP does not allow broadcasts or multicasts */ + if (IN_MULTICAST(ntohl(dst.sin_addr.s_addr))) { + goto out; + } + if (SCTP_IS_IT_BROADCAST(dst.sin_addr, m)) { + goto out; + } + ecn_bits = ip->ip_tos; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (m->m_pkthdr.csum_flags & CSUM_SCTP_VALID) { + SCTP_STAT_INCR(sctps_recvhwcrc); + compute_crc = 0; + } else { +#else + if (SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) && + ((src.sin_addr.s_addr == dst.sin_addr.s_addr) || + (SCTP_IS_IT_LOOPBACK(m)))) { + SCTP_STAT_INCR(sctps_recvhwcrc); + compute_crc = 0; + } else { +#endif + SCTP_STAT_INCR(sctps_recvswcrc); + compute_crc = 1; + } + sctp_common_input_processing(&m, iphlen, offset, length, + (struct sockaddr *)&src, + (struct sockaddr *)&dst, + sh, ch, + compute_crc, + ecn_bits, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, fibnum, +#endif + vrf_id, port); + out: + if (m) { + sctp_m_freem(m); + } + return; +} + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SCTP_MCORE_INPUT) && defined(SMP) +extern int *sctp_cpuarry; +#endif +#endif + +#if defined(__FreeBSD__) && !defined(__Userspace__) +int +sctp_input(struct mbuf **mp, int *offp, int proto SCTP_UNUSED) +{ + struct mbuf *m; + int off; + + m = *mp; + off = *offp; +#else +void +sctp_input(struct mbuf *m, int off) +{ +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SCTP_MCORE_INPUT) && defined(SMP) + if (mp_ncpus > 1) { + struct ip *ip; + struct sctphdr *sh; + int offset; + int cpu_to_use; + uint32_t flowid, tag; + + if (M_HASHTYPE_GET(m) != M_HASHTYPE_NONE) { + flowid = m->m_pkthdr.flowid; + } else { + /* No flow id built by lower layers + * fix it so we create one. + */ + offset = off + sizeof(struct sctphdr); + if (SCTP_BUF_LEN(m) < offset) { + if ((m = m_pullup(m, offset)) == NULL) { + SCTP_STAT_INCR(sctps_hdrops); + return (IPPROTO_DONE); + } + } + ip = mtod(m, struct ip *); + sh = (struct sctphdr *)((caddr_t)ip + off); + tag = htonl(sh->v_tag); + flowid = tag ^ ntohs(sh->dest_port) ^ ntohs(sh->src_port); + m->m_pkthdr.flowid = flowid; + M_HASHTYPE_SET(m, M_HASHTYPE_OPAQUE_HASH); + } + cpu_to_use = sctp_cpuarry[flowid % mp_ncpus]; + sctp_queue_to_mcore(m, off, cpu_to_use); + return (IPPROTO_DONE); + } +#endif +#endif + sctp_input_with_port(m, off, 0); +#if defined(__FreeBSD__) && !defined(__Userspace__) + return (IPPROTO_DONE); +#endif +} +#endif +#endif diff --git a/netwerk/sctp/src/netinet/sctp_input.h b/netwerk/sctp/src/netinet/sctp_input.h new file mode 100644 index 0000000000..d2cd215761 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_input.h @@ -0,0 +1,66 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_INPUT_H_ +#define _NETINET_SCTP_INPUT_H_ + +#if defined(_KERNEL) || defined(__Userspace__) +void +sctp_common_input_processing(struct mbuf **, int, int, int, + struct sockaddr *, struct sockaddr *, + struct sctphdr *, struct sctp_chunkhdr *, + uint8_t, + uint8_t, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t, uint32_t, uint16_t, +#endif + uint32_t, uint16_t); + +struct sctp_stream_reset_request * +sctp_find_stream_reset(struct sctp_tcb *stcb, uint32_t seq, + struct sctp_tmit_chunk **bchk); + +void +sctp_reset_in_stream(struct sctp_tcb *stcb, uint32_t number_entries, + uint16_t *list); + +int sctp_is_there_unsent_data(struct sctp_tcb *stcb, int so_locked); + +#endif +#endif diff --git a/netwerk/sctp/src/netinet/sctp_lock_userspace.h b/netwerk/sctp/src/netinet/sctp_lock_userspace.h new file mode 100644 index 0000000000..f3b9b3d961 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_lock_userspace.h @@ -0,0 +1,243 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * Copyright (c) 2008-2012, by Brad Penoff. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_LOCK_EMPTY_H_ +#define _NETINET_SCTP_LOCK_EMPTY_H_ + +/* + * Empty Lock declarations for all other platforms. Pre-process away to + * nothing. + */ + +/* __Userspace__ putting lock macros in same order as sctp_lock_bsd.h ...*/ + +#define SCTP_IPI_COUNT_INIT() + +#define SCTP_STATLOG_INIT_LOCK() +#define SCTP_STATLOG_LOCK() +#define SCTP_STATLOG_UNLOCK() +#define SCTP_STATLOG_DESTROY() + +#define SCTP_INP_INFO_LOCK_DESTROY() + +#define SCTP_INP_INFO_LOCK_INIT() +#define SCTP_INP_INFO_RLOCK() +#define SCTP_INP_INFO_WLOCK() +#define SCTP_INP_INFO_TRYLOCK() 1 +#define SCTP_INP_INFO_RUNLOCK() +#define SCTP_INP_INFO_WUNLOCK() +#define SCTP_INP_INFO_LOCK_ASSERT() +#define SCTP_INP_INFO_RLOCK_ASSERT() +#define SCTP_INP_INFO_WLOCK_ASSERT() + +#define SCTP_WQ_ADDR_INIT() +#define SCTP_WQ_ADDR_DESTROY() +#define SCTP_WQ_ADDR_LOCK() +#define SCTP_WQ_ADDR_UNLOCK() +#define SCTP_WQ_ADDR_LOCK_ASSERT() + +#define SCTP_IPI_ADDR_INIT() +#define SCTP_IPI_ADDR_DESTROY() +#define SCTP_IPI_ADDR_RLOCK() +#define SCTP_IPI_ADDR_WLOCK() +#define SCTP_IPI_ADDR_RUNLOCK() +#define SCTP_IPI_ADDR_WUNLOCK() +#define SCTP_IPI_ADDR_LOCK_ASSERT() +#define SCTP_IPI_ADDR_WLOCK_ASSERT() + +#define SCTP_IPI_ITERATOR_WQ_INIT() +#define SCTP_IPI_ITERATOR_WQ_DESTROY() +#define SCTP_IPI_ITERATOR_WQ_LOCK() +#define SCTP_IPI_ITERATOR_WQ_UNLOCK() + +#define SCTP_IP_PKTLOG_INIT() +#define SCTP_IP_PKTLOG_LOCK() +#define SCTP_IP_PKTLOG_UNLOCK() +#define SCTP_IP_PKTLOG_DESTROY() + +#define SCTP_INP_READ_INIT(_inp) +#define SCTP_INP_READ_DESTROY(_inp) +#define SCTP_INP_READ_LOCK(_inp) +#define SCTP_INP_READ_UNLOCK(_inp) + +#define SCTP_INP_LOCK_INIT(_inp) +#define SCTP_ASOC_CREATE_LOCK_INIT(_inp) +#define SCTP_INP_LOCK_DESTROY(_inp) +#define SCTP_ASOC_CREATE_LOCK_DESTROY(_inp) + +#define SCTP_INP_RLOCK(_inp) +#define SCTP_INP_WLOCK(_inp) +#define SCTP_INP_RLOCK_ASSERT(_inp) +#define SCTP_INP_WLOCK_ASSERT(_inp) + +#define SCTP_INP_LOCK_CONTENDED(_inp) (0) /* Don't know if this is possible */ + +#define SCTP_INP_READ_CONTENDED(_inp) (0) /* Don't know if this is possible */ + +#define SCTP_ASOC_CREATE_LOCK_CONTENDED(_inp) (0) /* Don't know if this is possible */ + + +#define SCTP_INP_INCR_REF(_inp) +#define SCTP_INP_DECR_REF(_inp) + +#define SCTP_ASOC_CREATE_LOCK(_inp) + +#define SCTP_INP_RUNLOCK(_inp) +#define SCTP_INP_WUNLOCK(_inp) +#define SCTP_ASOC_CREATE_UNLOCK(_inp) + + +#define SCTP_TCB_LOCK_INIT(_tcb) +#define SCTP_TCB_LOCK_DESTROY(_tcb) +#define SCTP_TCB_LOCK(_tcb) +#define SCTP_TCB_TRYLOCK(_tcb) 1 +#define SCTP_TCB_UNLOCK(_tcb) +#define SCTP_TCB_UNLOCK_IFOWNED(_tcb) +#define SCTP_TCB_LOCK_ASSERT(_tcb) + + + +#define SCTP_ITERATOR_LOCK_INIT() +#define SCTP_ITERATOR_LOCK() +#define SCTP_ITERATOR_UNLOCK() +#define SCTP_ITERATOR_LOCK_DESTROY() + + + +#define SCTP_INCR_EP_COUNT() \ + do { \ + sctppcbinfo.ipi_count_ep++; \ + } while (0) + +#define SCTP_DECR_EP_COUNT() \ + do { \ + sctppcbinfo.ipi_count_ep--; \ + } while (0) + +#define SCTP_INCR_ASOC_COUNT() \ + do { \ + sctppcbinfo.ipi_count_asoc++; \ + } while (0) + +#define SCTP_DECR_ASOC_COUNT() \ + do { \ + sctppcbinfo.ipi_count_asoc--; \ + } while (0) + +#define SCTP_INCR_LADDR_COUNT() \ + do { \ + sctppcbinfo.ipi_count_laddr++; \ + } while (0) + +#define SCTP_DECR_LADDR_COUNT() \ + do { \ + sctppcbinfo.ipi_count_laddr--; \ + } while (0) + +#define SCTP_INCR_RADDR_COUNT() \ + do { \ + sctppcbinfo.ipi_count_raddr++; \ + } while (0) + +#define SCTP_DECR_RADDR_COUNT() \ + do { \ + sctppcbinfo.ipi_count_raddr--; \ + } while (0) + +#define SCTP_INCR_CHK_COUNT() \ + do { \ + sctppcbinfo.ipi_count_chunk++; \ + } while (0) + +#define SCTP_DECR_CHK_COUNT() \ + do { \ + sctppcbinfo.ipi_count_chunk--; \ + } while (0) + +#define SCTP_INCR_READQ_COUNT() \ + do { \ + sctppcbinfo.ipi_count_readq++; \ + } while (0) + +#define SCTP_DECR_READQ_COUNT() \ + do { \ + sctppcbinfo.ipi_count_readq--; \ + } while (0) + +#define SCTP_INCR_STRMOQ_COUNT() \ + do { \ + sctppcbinfo.ipi_count_strmoq++; \ + } while (0) + +#define SCTP_DECR_STRMOQ_COUNT() \ + do { \ + sctppcbinfo.ipi_count_strmoq--; \ + } while (0) + + +/* these were in sctp_lock_empty.h but aren't in sctp_lock_bsd.h ... */ +#if 0 +#define SCTP_IPI_ADDR_LOCK() +#define SCTP_IPI_ADDR_UNLOCK() +#endif + + +/* These were in sctp_lock_empty.h because they were commented out within + * within user_include/user_socketvar.h . If they are NOT commented out + * in user_socketvar.h (because that seems the more natural place for them + * to live), then change this "if" to 0. Keep the "if" as 1 if these ARE + * indeed commented out in user_socketvar.h . + * + * This modularity is kept so this file can easily be chosen as an alternative + * to SCTP_PROCESS_LEVEL_LOCKS. If one defines SCTP_PROCESS_LEVEL_LOCKS in + * user_include/opt_sctp.h, then the file sctp_process_lock.h (which we didn't + * implement) is used, and that declares these locks already (so using + * SCTP_PROCESS_LEVEL_LOCKS *requires* that these defintions be commented out + * in user_socketvar.h). + */ +#if 1 +#define SOCK_LOCK(_so) +#define SOCK_UNLOCK(_so) +#define SOCKBUF_LOCK(_so_buf) +#define SOCKBUF_UNLOCK(_so_buf) +#define SOCKBUF_LOCK_ASSERT(_so_buf) +#endif + +#endif diff --git a/netwerk/sctp/src/netinet/sctp_os.h b/netwerk/sctp/src/netinet/sctp_os.h new file mode 100644 index 0000000000..dae612bc57 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_os.h @@ -0,0 +1,89 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2006-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_OS_H_ +#define _NETINET_SCTP_OS_H_ + +/* + * General kernel memory allocation: + * SCTP_MALLOC(element, type, size, name) + * SCTP_FREE(element) + * Kernel memory allocation for "soname"- memory must be zeroed. + * SCTP_MALLOC_SONAME(name, type, size) + * SCTP_FREE_SONAME(name) + */ + +/* + * Zone(pool) allocation routines: MUST be defined for each OS. + * zone = zone/pool pointer. + * name = string name of the zone/pool. + * size = size of each zone/pool element. + * number = number of elements in zone/pool. + * type = structure type to allocate + * + * sctp_zone_t + * SCTP_ZONE_INIT(zone, name, size, number) + * SCTP_ZONE_GET(zone, type) + * SCTP_ZONE_FREE(zone, element) + * SCTP_ZONE_DESTROY(zone) + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <netinet/sctp_os_bsd.h> +#else +#define MODULE_GLOBAL(_B) (_B) +#endif +#if defined(__Userspace__) +#include <netinet/sctp_os_userspace.h> +#endif +#if defined(__APPLE__) && !defined(__Userspace__) +#include <netinet/sctp_os_macosx.h> +#endif +#if defined(_WIN32) && !defined(__Userspace__) +#include <netinet/sctp_os_windows.h> +#endif + +/* All os's must implement this address gatherer. If + * no VRF's exist, then vrf 0 is the only one and all + * addresses and ifn's live here. + */ +#define SCTP_DEFAULT_VRF 0 +void sctp_init_vrf_list(int vrfid); + +#endif diff --git a/netwerk/sctp/src/netinet/sctp_os_userspace.h b/netwerk/sctp/src/netinet/sctp_os_userspace.h new file mode 100644 index 0000000000..05f655aa94 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_os_userspace.h @@ -0,0 +1,1147 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2006-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2011, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2011, by Michael Tuexen. All rights reserved. + * Copyright (c) 2008-2011, by Brad Penoff. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#ifndef __sctp_os_userspace_h__ +#define __sctp_os_userspace_h__ +/* + * Userspace includes + * All the opt_xxx.h files are placed in the kernel build directory. + * We will place them in userspace stack build directory. + */ + +#include <errno.h> + +#if defined(_WIN32) +#include <winsock2.h> +#include <ws2tcpip.h> +#include <iphlpapi.h> +#include <mswsock.h> +#include <windows.h> +#include "user_environment.h" +typedef CRITICAL_SECTION userland_mutex_t; +#if WINVER < 0x0600 +typedef CRITICAL_SECTION userland_rwlock_t; +enum { + C_SIGNAL = 0, + C_BROADCAST = 1, + C_MAX_EVENTS = 2 +}; +typedef struct +{ + u_int waiters_count; + CRITICAL_SECTION waiters_count_lock; + HANDLE events_[C_MAX_EVENTS]; +} userland_cond_t; +void InitializeXPConditionVariable(userland_cond_t *); +void DeleteXPConditionVariable(userland_cond_t *); +int SleepXPConditionVariable(userland_cond_t *, userland_mutex_t *); +void WakeAllXPConditionVariable(userland_cond_t *); +#define InitializeConditionVariable(cond) InitializeXPConditionVariable(cond) +#define DeleteConditionVariable(cond) DeleteXPConditionVariable(cond) +#define SleepConditionVariableCS(cond, mtx, time) SleepXPConditionVariable(cond, mtx) +#define WakeAllConditionVariable(cond) WakeAllXPConditionVariable(cond) +#else +typedef SRWLOCK userland_rwlock_t; +#define DeleteConditionVariable(cond) +typedef CONDITION_VARIABLE userland_cond_t; +#endif +typedef HANDLE userland_thread_t; +#define ADDRESS_FAMILY unsigned __int8 +#define IPVERSION 4 +#define MAXTTL 255 +/* VS2010 comes with stdint.h */ +#if !defined(_MSC_VER) || (_MSC_VER >= 1600) +#include <stdint.h> +#else +typedef unsigned __int64 uint64_t; +typedef unsigned __int32 uint32_t; +typedef __int32 int32_t; +typedef unsigned __int16 uint16_t; +typedef __int16 int16_t; +typedef unsigned __int8 uint8_t; +typedef __int8 int8_t; +#endif +#ifndef _SIZE_T_DEFINED +#typedef __int32 size_t; +#endif +typedef unsigned __int32 u_int; +typedef unsigned char u_char; +typedef unsigned __int16 u_short; +typedef unsigned __int8 sa_family_t; +#ifndef _SSIZE_T_DEFINED +typedef __int64 ssize_t; +#endif +#if !defined(__MINGW32__) +#define __func__ __FUNCTION__ +#endif +#ifndef EWOULDBLOCK +#define EWOULDBLOCK WSAEWOULDBLOCK +#endif +#ifndef EINPROGRESS +#define EINPROGRESS WSAEINPROGRESS +#endif +#ifndef EALREADY +#define EALREADY WSAEALREADY +#endif +#ifndef ENOTSOCK +#define ENOTSOCK WSAENOTSOCK +#endif +#ifndef EDESTADDRREQ +#define EDESTADDRREQ WSAEDESTADDRREQ +#endif +#ifndef EMSGSIZE +#define EMSGSIZE WSAEMSGSIZE +#endif +#ifndef EPROTOTYPE +#define EPROTOTYPE WSAEPROTOTYPE +#endif +#ifndef ENOPROTOOPT +#define ENOPROTOOPT WSAENOPROTOOPT +#endif +#ifndef EPROTONOSUPPORT +#define EPROTONOSUPPORT WSAEPROTONOSUPPORT +#endif +#ifndef ESOCKTNOSUPPORT +#define ESOCKTNOSUPPORT WSAESOCKTNOSUPPORT +#endif +#ifndef EOPNOTSUPP +#define EOPNOTSUPP WSAEOPNOTSUPP +#endif +#ifndef ENOTSUP +#define ENOTSUP WSAEOPNOTSUPP +#endif +#ifndef EPFNOSUPPORT +#define EPFNOSUPPORT WSAEPFNOSUPPORT +#endif +#ifndef EAFNOSUPPORT +#define EAFNOSUPPORT WSAEAFNOSUPPORT +#endif +#ifndef EADDRINUSE +#define EADDRINUSE WSAEADDRINUSE +#endif +#ifndef EADDRNOTAVAIL +#define EADDRNOTAVAIL WSAEADDRNOTAVAIL +#endif +#ifndef ENETDOWN +#define ENETDOWN WSAENETDOWN +#endif +#ifndef ENETUNREACH +#define ENETUNREACH WSAENETUNREACH +#endif +#ifndef ENETRESET +#define ENETRESET WSAENETRESET +#endif +#ifndef ECONNABORTED +#define ECONNABORTED WSAECONNABORTED +#endif +#ifndef ECONNRESET +#define ECONNRESET WSAECONNRESET +#endif +#ifndef ENOBUFS +#define ENOBUFS WSAENOBUFS +#endif +#ifndef EISCONN +#define EISCONN WSAEISCONN +#endif +#ifndef ENOTCONN +#define ENOTCONN WSAENOTCONN +#endif +#ifndef ESHUTDOWN +#define ESHUTDOWN WSAESHUTDOWN +#endif +#ifndef ETOOMANYREFS +#define ETOOMANYREFS WSAETOOMANYREFS +#endif +#ifndef ETIMEDOUT +#define ETIMEDOUT WSAETIMEDOUT +#endif +#ifndef ECONNREFUSED +#define ECONNREFUSED WSAECONNREFUSED +#endif +#ifndef ELOOP +#define ELOOP WSAELOOP +#endif +#ifndef EHOSTDOWN +#define EHOSTDOWN WSAEHOSTDOWN +#endif +#ifndef EHOSTUNREACH +#define EHOSTUNREACH WSAEHOSTUNREACH +#endif +#ifndef EPROCLIM +#define EPROCLIM WSAEPROCLIM +#endif +#ifndef EUSERS +#define EUSERS WSAEUSERS +#endif +#ifndef EDQUOT +#define EDQUOT WSAEDQUOT +#endif +#ifndef ESTALE +#define ESTALE WSAESTALE +#endif +#ifndef EREMOTE +#define EREMOTE WSAEREMOTE +#endif + +typedef char* caddr_t; + +#define bzero(buf, len) memset(buf, 0, len) +#define bcopy(srcKey, dstKey, len) memcpy(dstKey, srcKey, len) + +#if defined(_MSC_VER) && (_MSC_VER < 1900) && !defined(__MINGW32__) +#define SCTP_SNPRINTF(data, size, format, ...) \ + if (_snprintf_s(data, size, _TRUNCATE, format, __VA_ARGS__) < 0) { \ + data[0] = '\0'; \ + } +#else +#define SCTP_SNPRINTF(data, ...) \ + if (snprintf(data, __VA_ARGS__) < 0 ) { \ + data[0] = '\0'; \ + } +#endif + +#define inline __inline +#define __inline__ __inline +#define MSG_EOR 0x8 /* data completes record */ +#define MSG_DONTWAIT 0x80 /* this message should be nonblocking */ + +#ifdef CMSG_DATA +#undef CMSG_DATA +#endif +/* + * The following definitions should apply iff WINVER < 0x0600 + * but that check doesn't work in all cases. So be more pedantic... + */ +#define CMSG_DATA(x) WSA_CMSG_DATA(x) +#define CMSG_ALIGN(x) WSA_CMSGDATA_ALIGN(x) +#ifndef CMSG_FIRSTHDR +#define CMSG_FIRSTHDR(x) WSA_CMSG_FIRSTHDR(x) +#endif +#ifndef CMSG_NXTHDR +#define CMSG_NXTHDR(x, y) WSA_CMSG_NXTHDR(x, y) +#endif +#ifndef CMSG_SPACE +#define CMSG_SPACE(x) WSA_CMSG_SPACE(x) +#endif +#ifndef CMSG_LEN +#define CMSG_LEN(x) WSA_CMSG_LEN(x) +#endif + +/**** from sctp_os_windows.h ***************/ +#define SCTP_IFN_IS_IFT_LOOP(ifn) ((ifn)->ifn_type == IFT_LOOP) +#define SCTP_ROUTE_IS_REAL_LOOP(ro) ((ro)->ro_rt && (ro)->ro_rt->rt_ifa && (ro)->ro_rt->rt_ifa->ifa_ifp && (ro)->ro_rt->rt_ifa->ifa_ifp->if_type == IFT_LOOP) + +/* + * Access to IFN's to help with src-addr-selection + */ +/* This could return VOID if the index works but for BSD we provide both. */ +#define SCTP_GET_IFN_VOID_FROM_ROUTE(ro) \ + ((ro)->ro_rt != NULL ? (ro)->ro_rt->rt_ifp : NULL) +#define SCTP_ROUTE_HAS_VALID_IFN(ro) \ + ((ro)->ro_rt && (ro)->ro_rt->rt_ifp) +/******************************************/ + +#define SCTP_GET_IF_INDEX_FROM_ROUTE(ro) 1 /* compiles... TODO use routing socket to determine */ + +#define BIG_ENDIAN 1 +#define LITTLE_ENDIAN 0 +#ifdef WORDS_BIGENDIAN +#define BYTE_ORDER BIG_ENDIAN +#else +#define BYTE_ORDER LITTLE_ENDIAN +#endif + +#else /* !defined(Userspace_os_Windows) */ +#include <sys/socket.h> + +#if defined(__EMSCRIPTEN__) && !defined(__EMSCRIPTEN_PTHREADS__) +#error "Unsupported build configuration." +#endif + +#include <pthread.h> + +typedef pthread_mutex_t userland_mutex_t; +typedef pthread_rwlock_t userland_rwlock_t; +typedef pthread_cond_t userland_cond_t; +typedef pthread_t userland_thread_t; +#endif + +#if defined(_WIN32) || defined(__native_client__) + +#define IFNAMSIZ 64 + +#define random() rand() +#define srandom(s) srand(s) + +#define timeradd(tvp, uvp, vvp) \ + do { \ + (vvp)->tv_sec = (tvp)->tv_sec + (uvp)->tv_sec; \ + (vvp)->tv_usec = (tvp)->tv_usec + (uvp)->tv_usec; \ + if ((vvp)->tv_usec >= 1000000) { \ + (vvp)->tv_sec++; \ + (vvp)->tv_usec -= 1000000; \ + } \ + } while (0) + +#define timersub(tvp, uvp, vvp) \ + do { \ + (vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec; \ + (vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec; \ + if ((vvp)->tv_usec < 0) { \ + (vvp)->tv_sec--; \ + (vvp)->tv_usec += 1000000; \ + } \ + } while (0) + +/*#include <packon.h> +#pragma pack(push, 1)*/ +struct ip { + u_char ip_hl:4, ip_v:4; + u_char ip_tos; + u_short ip_len; + u_short ip_id; + u_short ip_off; +#define IP_RP 0x8000 +#define IP_DF 0x4000 +#define IP_MF 0x2000 +#define IP_OFFMASK 0x1fff + u_char ip_ttl; + u_char ip_p; + u_short ip_sum; + struct in_addr ip_src, ip_dst; +}; + +struct ifaddrs { + struct ifaddrs *ifa_next; + char *ifa_name; + unsigned int ifa_flags; + struct sockaddr *ifa_addr; + struct sockaddr *ifa_netmask; + struct sockaddr *ifa_dstaddr; + void *ifa_data; +}; + +struct udphdr { + uint16_t uh_sport; + uint16_t uh_dport; + uint16_t uh_ulen; + uint16_t uh_sum; +}; + +struct iovec { + size_t len; + char *buf; +}; + +#define iov_base buf +#define iov_len len + +struct ifa_msghdr { + uint16_t ifam_msglen; + unsigned char ifam_version; + unsigned char ifam_type; + uint32_t ifam_addrs; + uint32_t ifam_flags; + uint16_t ifam_index; + uint32_t ifam_metric; +}; + +struct ifdevmtu { + int ifdm_current; + int ifdm_min; + int ifdm_max; +}; + +struct ifkpi { + unsigned int ifk_module_id; + unsigned int ifk_type; + union { + void *ifk_ptr; + int ifk_value; + } ifk_data; +}; +#endif + +#if defined(_WIN32) +int Win_getifaddrs(struct ifaddrs**); +#define getifaddrs(interfaces) (int)Win_getifaddrs(interfaces) +int win_if_nametoindex(const char *); +#define if_nametoindex(x) win_if_nametoindex(x) +#endif + +#define mtx_lock(arg1) +#define mtx_unlock(arg1) +#define mtx_assert(arg1,arg2) +#define MA_OWNED 7 /* sys/mutex.h typically on FreeBSD */ +#if !defined(__FreeBSD__) +struct mtx {int dummy;}; +#if !defined(__NetBSD__) +struct selinfo {int dummy;}; +#endif +struct sx {int dummy;}; +#endif + +#include <stdio.h> +#include <string.h> +#include <stdbool.h> +/* #include <sys/param.h> in FreeBSD defines MSIZE */ +/* #include <sys/ktr.h> */ +/* #include <sys/systm.h> */ +#if defined(HAVE_SYS_QUEUE_H) +#include <sys/queue.h> +#else +#include <user_queue.h> +#endif +#include <user_malloc.h> +/* #include <sys/kernel.h> */ +/* #include <sys/sysctl.h> */ +/* #include <sys/protosw.h> */ +/* on FreeBSD, this results in a redefintion of SOCK(BUF)_(UN)LOCK and + * uknown type of struct mtx for sb_mtx in struct sockbuf */ +#include "user_socketvar.h" /* MALLOC_DECLARE's M_PCB. Replacement for sys/socketvar.h */ +/* #include <sys/jail.h> */ +/* #include <sys/sysctl.h> */ +#include <user_environment.h> +#include <user_atomic.h> +#include <user_mbuf.h> +/* #include <sys/uio.h> */ +/* #include <sys/lock.h> */ +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/rwlock.h> +#endif +/* #include <sys/kthread.h> */ +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/priv.h> +#endif +/* #include <sys/random.h> */ +#include <limits.h> +/* #include <machine/cpu.h> */ + +#if defined(__APPLE__) +/* was a 0 byte file. needed for structs if_data(64) and net_event_data */ +#include <net/if_var.h> +#endif +#if defined(__FreeBSD__) +#include <net/if_types.h> +/* #include <net/if_var.h> was a 0 byte file. causes struct mtx redefinition */ +#endif +/* OOTB only - dummy route used at the moment. should we port route to + * userspace as well? */ +/* on FreeBSD, this results in a redefintion of struct route */ +/* #include <net/route.h> */ +#if !defined(_WIN32) && !defined(__native_client__) +#include <net/if.h> +#include <netinet/in.h> +#include <netinet/in_systm.h> +#include <netinet/ip.h> +#endif +#if defined(HAVE_NETINET_IP_ICMP_H) +#include <netinet/ip_icmp.h> +#else +#include <user_ip_icmp.h> +#endif +/* #include <netinet/in_pcb.h> ported to userspace */ +#include <user_inpcb.h> + +/* for getifaddrs */ +#include <sys/types.h> +#if !defined(_WIN32) +#if defined(INET) || defined(INET6) +#include <ifaddrs.h> +#endif + +/* for ioctl */ +#include <sys/ioctl.h> + +/* for close, etc. */ +#include <unistd.h> +/* for gettimeofday */ +#include <sys/time.h> +#endif + +/* lots of errno's used and needed in userspace */ + +/* for offsetof */ +#include <stddef.h> + +#if defined(SCTP_PROCESS_LEVEL_LOCKS) && !defined(_WIN32) +/* for pthread_mutex_lock, pthread_mutex_unlock, etc. */ +#include <pthread.h> +#endif + +#ifdef IPSEC +#include <netipsec/ipsec.h> +#include <netipsec/key.h> +#endif /* IPSEC */ + +#ifdef INET6 +#if defined(__FreeBSD__) +#include <sys/domain.h> +#endif +#ifdef IPSEC +#include <netipsec/ipsec6.h> +#endif +#if !defined(_WIN32) +#include <netinet/ip6.h> +#endif +#if defined(__APPLE__) || defined(__FreeBSD__) || defined(__linux__) || defined(__NetBSD__) || defined(__OpenBSD__) || defined(_WIN32) || defined(__EMSCRIPTEN__) +#include "user_ip6_var.h" +#else +#include <netinet6/ip6_var.h> +#endif +#if defined(__FreeBSD__) +#include <netinet6/in6_pcb.h> +#include <netinet6/ip6protosw.h> +/* #include <netinet6/nd6.h> was a 0 byte file */ +#include <netinet6/scope6_var.h> +#endif +#endif /* INET6 */ + +#if defined(HAVE_SCTP_PEELOFF_SOCKOPT) +#include <sys/file.h> +#include <sys/filedesc.h> +#endif + +#include "netinet/sctp_sha1.h" + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <netinet/ip_options.h> +#endif + +#define SCTP_PRINTF(...) \ + if (SCTP_BASE_VAR(debug_printf)) { \ + SCTP_BASE_VAR(debug_printf)(__VA_ARGS__); \ + } + +/* Declare all the malloc names for all the various mallocs */ +MALLOC_DECLARE(SCTP_M_MAP); +MALLOC_DECLARE(SCTP_M_STRMI); +MALLOC_DECLARE(SCTP_M_STRMO); +MALLOC_DECLARE(SCTP_M_ASC_ADDR); +MALLOC_DECLARE(SCTP_M_ASC_IT); +MALLOC_DECLARE(SCTP_M_AUTH_CL); +MALLOC_DECLARE(SCTP_M_AUTH_KY); +MALLOC_DECLARE(SCTP_M_AUTH_HL); +MALLOC_DECLARE(SCTP_M_AUTH_IF); +MALLOC_DECLARE(SCTP_M_STRESET); +MALLOC_DECLARE(SCTP_M_CMSG); +MALLOC_DECLARE(SCTP_M_COPYAL); +MALLOC_DECLARE(SCTP_M_VRF); +MALLOC_DECLARE(SCTP_M_IFA); +MALLOC_DECLARE(SCTP_M_IFN); +MALLOC_DECLARE(SCTP_M_TIMW); +MALLOC_DECLARE(SCTP_M_MVRF); +MALLOC_DECLARE(SCTP_M_ITER); +MALLOC_DECLARE(SCTP_M_SOCKOPT); + +#if defined(SCTP_LOCAL_TRACE_BUF) + +#define SCTP_GET_CYCLECOUNT get_cyclecount() +#define SCTP_CTR6 sctp_log_trace + +#else +#define SCTP_CTR6 CTR6 +#endif + +/* Empty ktr statement for _Userspace__ (similar to what is done for mac) */ +#define CTR6(m, d, p1, p2, p3, p4, p5, p6) + + + +#define SCTP_BASE_INFO(__m) system_base_info.sctppcbinfo.__m +#define SCTP_BASE_STATS system_base_info.sctpstat +#define SCTP_BASE_STAT(__m) system_base_info.sctpstat.__m +#define SCTP_BASE_SYSCTL(__m) system_base_info.sctpsysctl.__m +#define SCTP_BASE_VAR(__m) system_base_info.__m + +/* + * + */ +#if !defined(__APPLE__) +#define USER_ADDR_NULL (NULL) /* FIX ME: temp */ +#endif + +#include <netinet/sctp_constants.h> +#if defined(SCTP_DEBUG) +#define SCTPDBG(level, ...) \ +{ \ + do { \ + if (SCTP_BASE_SYSCTL(sctp_debug_on) & level) { \ + SCTP_PRINTF(__VA_ARGS__); \ + } \ + } while (0); \ +} +#define SCTPDBG_ADDR(level, addr) \ +{ \ + do { \ + if (SCTP_BASE_SYSCTL(sctp_debug_on) & level ) { \ + sctp_print_address(addr); \ + } \ + } while (0); \ +} +#else +#define SCTPDBG(level, ...) +#define SCTPDBG_ADDR(level, addr) +#endif + +#ifdef SCTP_LTRACE_CHUNKS +#define SCTP_LTRACE_CHK(a, b, c, d) if(sctp_logging_level & SCTP_LTRACE_CHUNK_ENABLE) CTR6(KTR_SUBSYS, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_CHUNK_PROC, 0, a, b, c, d) +#else +#define SCTP_LTRACE_CHK(a, b, c, d) +#endif + +#ifdef SCTP_LTRACE_ERRORS +#define SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, file, err) \ + if (sctp_logging_level & SCTP_LTRACE_ERROR_ENABLE) \ + SCTP_PRINTF("mbuf:%p inp:%p stcb:%p net:%p file:%x line:%d error:%d\n", \ + (void *)m, (void *)inp, (void *)stcb, (void *)net, file, __LINE__, err); +#define SCTP_LTRACE_ERR_RET(inp, stcb, net, file, err) \ + if (sctp_logging_level & SCTP_LTRACE_ERROR_ENABLE) \ + SCTP_PRINTF("inp:%p stcb:%p net:%p file:%x line:%d error:%d\n", \ + (void *)inp, (void *)stcb, (void *)net, file, __LINE__, err); +#else +#define SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, file, err) +#define SCTP_LTRACE_ERR_RET(inp, stcb, net, file, err) +#endif + + +/* + * Local address and interface list handling + */ +#define SCTP_MAX_VRF_ID 0 +#define SCTP_SIZE_OF_VRF_HASH 3 +#define SCTP_IFNAMSIZ IFNAMSIZ +#define SCTP_DEFAULT_VRFID 0 +#define SCTP_VRF_ADDR_HASH_SIZE 16 +#define SCTP_VRF_IFN_HASH_SIZE 3 +#define SCTP_INIT_VRF_TABLEID(vrf) + +#if !defined(_WIN32) +#define SCTP_IFN_IS_IFT_LOOP(ifn) (strncmp((ifn)->ifn_name, "lo", 2) == 0) +/* BSD definition */ +/* #define SCTP_ROUTE_IS_REAL_LOOP(ro) ((ro)->ro_rt && (ro)->ro_rt->rt_ifa && (ro)->ro_rt->rt_ifa->ifa_ifp && (ro)->ro_rt->rt_ifa->ifa_ifp->if_type == IFT_LOOP) */ +/* only used in IPv6 scenario, which isn't supported yet */ +#define SCTP_ROUTE_IS_REAL_LOOP(ro) 0 + +/* + * Access to IFN's to help with src-addr-selection + */ +/* This could return VOID if the index works but for BSD we provide both. */ +#define SCTP_GET_IFN_VOID_FROM_ROUTE(ro) (void *)ro->ro_rt->rt_ifp +#define SCTP_GET_IF_INDEX_FROM_ROUTE(ro) 1 /* compiles... TODO use routing socket to determine */ +#define SCTP_ROUTE_HAS_VALID_IFN(ro) ((ro)->ro_rt && (ro)->ro_rt->rt_ifp) +#endif + +/* + * general memory allocation + */ +#define SCTP_MALLOC(var, type, size, name) \ + do { \ + MALLOC(var, type, size, name, M_NOWAIT); \ + } while (0) + +#define SCTP_FREE(var, type) FREE(var, type) + +#define SCTP_MALLOC_SONAME(var, type, size) \ + do { \ + MALLOC(var, type, size, M_SONAME, (M_WAITOK | M_ZERO)); \ + } while (0) + +#define SCTP_FREE_SONAME(var) FREE(var, M_SONAME) + +#define SCTP_PROCESS_STRUCT struct proc * + +/* + * zone allocation functions + */ + + +#if defined(SCTP_SIMPLE_ALLOCATOR) +/*typedef size_t sctp_zone_t;*/ +#define SCTP_ZONE_INIT(zone, name, size, number) { \ + zone = size; \ +} + +/* __Userspace__ SCTP_ZONE_GET: allocate element from the zone */ +#define SCTP_ZONE_GET(zone, type) \ + (type *)malloc(zone); + + +/* __Userspace__ SCTP_ZONE_FREE: free element from the zone */ +#define SCTP_ZONE_FREE(zone, element) { \ + free(element); \ +} + +#define SCTP_ZONE_DESTROY(zone) +#else +/*__Userspace__ + Compiling & linking notes: Needs libumem, which has been placed in ./user_lib + All userspace header files are in ./user_include. Makefile will need the + following. + CFLAGS = -I./ -Wall + LDFLAGS = -L./user_lib -R./user_lib -lumem +*/ +#include "user_include/umem.h" + +/* __Userspace__ SCTP_ZONE_INIT: initialize the zone */ +/* + __Userspace__ + No equivalent function to uma_zone_set_max added yet. (See SCTP_ZONE_INIT in sctp_os_bsd.h + for reference). It may not be required as mentioned in + http://nixdoc.net/man-pages/FreeBSD/uma_zalloc.9.html that + max limits may not enforced on systems with more than one CPU. +*/ +#define SCTP_ZONE_INIT(zone, name, size, number) { \ + zone = umem_cache_create(name, size, 0, NULL, NULL, NULL, NULL, NULL, 0); \ + } + +/* __Userspace__ SCTP_ZONE_GET: allocate element from the zone */ +#define SCTP_ZONE_GET(zone, type) \ + (type *)umem_cache_alloc(zone, UMEM_DEFAULT); + + +/* __Userspace__ SCTP_ZONE_FREE: free element from the zone */ +#define SCTP_ZONE_FREE(zone, element) \ + umem_cache_free(zone, element); + + +/* __Userspace__ SCTP_ZONE_DESTROY: destroy the zone */ +#define SCTP_ZONE_DESTROY(zone) \ + umem_cache_destroy(zone); +#endif + +/* + * __Userspace__ Defining sctp_hashinit_flags() and sctp_hashdestroy() for userland. + */ +void *sctp_hashinit_flags(int elements, struct malloc_type *type, + u_long *hashmask, int flags); +void +sctp_hashdestroy(void *vhashtbl, struct malloc_type *type, u_long hashmask); + +void +sctp_hashfreedestroy(void *vhashtbl, struct malloc_type *type, u_long hashmask); + + +#define HASH_NOWAIT 0x00000001 +#define HASH_WAITOK 0x00000002 + +/* M_PCB is MALLOC_DECLARE'd in sys/socketvar.h */ +#define SCTP_HASH_INIT(size, hashmark) sctp_hashinit_flags(size, M_PCB, hashmark, HASH_NOWAIT) + +#define SCTP_HASH_FREE(table, hashmark) sctp_hashdestroy(table, M_PCB, hashmark) + +#define SCTP_HASH_FREE_DESTROY(table, hashmark) sctp_hashfreedestroy(table, M_PCB, hashmark) +#define SCTP_M_COPYM m_copym + +/* + * timers + */ +/* __Userspace__ + * user_sctp_callout.h has typedef struct sctp_callout sctp_os_timer_t; + * which is used in the timer related functions such as + * SCTP_OS_TIMER_INIT etc. +*/ +#include <netinet/sctp_callout.h> + +/* __Userspace__ Creating a receive thread */ +#include <user_recv_thread.h> + +/*__Userspace__ defining KTR_SUBSYS 1 as done in sctp_os_macosx.h */ +#define KTR_SUBSYS 1 + +/* The packed define for 64 bit platforms */ +#if !defined(_WIN32) +#define SCTP_PACKED __attribute__((packed)) +#define SCTP_UNUSED __attribute__((unused)) +#else +#define SCTP_PACKED +#define SCTP_UNUSED +#endif + +/* + * Functions + */ +/* Mbuf manipulation and access macros */ +#define SCTP_BUF_LEN(m) (m->m_len) +#define SCTP_BUF_NEXT(m) (m->m_next) +#define SCTP_BUF_NEXT_PKT(m) (m->m_nextpkt) +#define SCTP_BUF_RESV_UF(m, size) m->m_data += size +#define SCTP_BUF_AT(m, size) m->m_data + size +#define SCTP_BUF_IS_EXTENDED(m) (m->m_flags & M_EXT) +#define SCTP_BUF_EXTEND_SIZE(m) (m->m_ext.ext_size) +#define SCTP_BUF_TYPE(m) (m->m_type) +#define SCTP_BUF_RECVIF(m) (m->m_pkthdr.rcvif) +#define SCTP_BUF_PREPEND M_PREPEND + +#define SCTP_ALIGN_TO_END(m, len) if(m->m_flags & M_PKTHDR) { \ + MH_ALIGN(m, len); \ + } else if ((m->m_flags & M_EXT) == 0) { \ + M_ALIGN(m, len); \ + } + +#if !defined(_WIN32) +#define SCTP_SNPRINTF(data, ...) \ + if (snprintf(data, __VA_ARGS__) < 0) { \ + data[0] = '\0'; \ + } +#endif + +/* We make it so if you have up to 4 threads + * writting based on the default size of + * the packet log 65 k, that would be + * 4 16k packets before we would hit + * a problem. + */ +#define SCTP_PKTLOG_WRITERS_NEED_LOCK 3 + + +/* + * routes, output, etc. + */ + +typedef struct sctp_route sctp_route_t; +typedef struct sctp_rtentry sctp_rtentry_t; + +static inline void sctp_userspace_rtalloc(sctp_route_t *ro) +{ + if (ro->ro_rt != NULL) { + ro->ro_rt->rt_refcnt++; + return; + } + + ro->ro_rt = (sctp_rtentry_t *) malloc(sizeof(sctp_rtentry_t)); + if (ro->ro_rt == NULL) + return; + + /* initialize */ + memset(ro->ro_rt, 0, sizeof(sctp_rtentry_t)); + ro->ro_rt->rt_refcnt = 1; + + /* set MTU */ + /* TODO set this based on the ro->ro_dst, looking up MTU with routing socket */ +#if 0 + if (userspace_rawroute == -1) { + userspace_rawroute = socket(AF_ROUTE, SOCK_RAW, 0); + if (userspace_rawroute == -1) + return; + } +#endif + ro->ro_rt->rt_rmx.rmx_mtu = 1500; /* FIXME temporary solution */ + + /* TODO enable the ability to obtain interface index of route for + * SCTP_GET_IF_INDEX_FROM_ROUTE macro. + */ +} +#define SCTP_RTALLOC(ro, vrf_id, fibnum) sctp_userspace_rtalloc((sctp_route_t *)ro) + +/* dummy rtfree needed once user_route.h is included */ +static inline void sctp_userspace_rtfree(sctp_rtentry_t *rt) +{ + if(rt == NULL) { + return; + } + if(--rt->rt_refcnt > 0) { + return; + } + free(rt); +} +#define rtfree(arg1) sctp_userspace_rtfree(arg1) + + +/*************************/ +/* MTU */ +/*************************/ +int sctp_userspace_get_mtu_from_ifn(uint32_t if_index); + +#define SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index) sctp_userspace_get_mtu_from_ifn(ifn_index) + +#define SCTP_GATHER_MTU_FROM_ROUTE(sctp_ifa, sa, rt) ((rt != NULL) ? rt->rt_rmx.rmx_mtu : 0) + +#define SCTP_SET_MTU_OF_ROUTE(sa, rt, mtu) do { \ + if (rt != NULL) \ + rt->rt_rmx.rmx_mtu = mtu; \ + } while(0) + + +/*************************/ +/* These are for logging */ +/*************************/ +/* return the base ext data pointer */ +#define SCTP_BUF_EXTEND_BASE(m) (m->m_ext.ext_buf) + /* return the refcnt of the data pointer */ +#define SCTP_BUF_EXTEND_REFCNT(m) (*m->m_ext.ref_cnt) +/* return any buffer related flags, this is + * used beyond logging for apple only. + */ +#define SCTP_BUF_GET_FLAGS(m) (m->m_flags) + +/* For BSD this just accesses the M_PKTHDR length + * so it operates on an mbuf with hdr flag. Other + * O/S's may have seperate packet header and mbuf + * chain pointers.. thus the macro. + */ +#define SCTP_HEADER_TO_CHAIN(m) (m) +#define SCTP_DETACH_HEADER_FROM_CHAIN(m) +#define SCTP_HEADER_LEN(m) ((m)->m_pkthdr.len) +#define SCTP_GET_HEADER_FOR_OUTPUT(o_pak) 0 +#define SCTP_RELEASE_HEADER(m) +#define SCTP_RELEASE_PKT(m) sctp_m_freem(m) + +#define SCTP_GET_PKT_VRFID(m, vrf_id) ((vrf_id = SCTP_DEFAULT_VRFID) != SCTP_DEFAULT_VRFID) + + + +/* Attach the chain of data into the sendable packet. */ +#define SCTP_ATTACH_CHAIN(pak, m, packet_length) do { \ + pak = m; \ + pak->m_pkthdr.len = packet_length; \ + } while(0) + +/* Other m_pkthdr type things */ +/* FIXME need real definitions */ +#define SCTP_IS_IT_BROADCAST(dst, m) 0 +/* OOTB only #define SCTP_IS_IT_BROADCAST(dst, m) ((m->m_flags & M_PKTHDR) ? in_broadcast(dst, m->m_pkthdr.rcvif) : 0) BSD def */ +#define SCTP_IS_IT_LOOPBACK(m) 0 +/* OOTB ONLY #define SCTP_IS_IT_LOOPBACK(m) ((m->m_flags & M_PKTHDR) && ((m->m_pkthdr.rcvif == NULL) || (m->m_pkthdr.rcvif->if_type == IFT_LOOP))) BSD def */ + + +/* This converts any input packet header + * into the chain of data holders, for BSD + * its a NOP. + */ + +/* get the v6 hop limit */ +#define SCTP_GET_HLIM(inp, ro) 128 +#define IPv6_HOP_LIMIT 128 + +/* is the endpoint v6only? */ +#define SCTP_IPV6_V6ONLY(sctp_inpcb) ((sctp_inpcb)->ip_inp.inp.inp_flags & IN6P_IPV6_V6ONLY) +/* is the socket non-blocking? */ +#define SCTP_SO_IS_NBIO(so) ((so)->so_state & SS_NBIO) +#define SCTP_SET_SO_NBIO(so) ((so)->so_state |= SS_NBIO) +#define SCTP_CLEAR_SO_NBIO(so) ((so)->so_state &= ~SS_NBIO) +/* get the socket type */ +#define SCTP_SO_TYPE(so) ((so)->so_type) + +/* reserve sb space for a socket */ +#define SCTP_SORESERVE(so, send, recv) soreserve(so, send, recv) + +/* wakeup a socket */ +#define SCTP_SOWAKEUP(so) wakeup(&(so)->so_timeo, so) +/* number of bytes ready to read */ +#define SCTP_SBAVAIL(sb) (sb)->sb_cc +/* clear the socket buffer state */ +#define SCTP_SB_CLEAR(sb) \ + (sb).sb_cc = 0; \ + (sb).sb_mb = NULL; \ + (sb).sb_mbcnt = 0; + +#define SCTP_SB_LIMIT_RCV(so) so->so_rcv.sb_hiwat +#define SCTP_SB_LIMIT_SND(so) so->so_snd.sb_hiwat + +#define SCTP_READ_RANDOM(buf, len) read_random(buf, len) + +#define SCTP_SHA1_CTX struct sctp_sha1_context +#define SCTP_SHA1_INIT sctp_sha1_init +#define SCTP_SHA1_UPDATE sctp_sha1_update +#define SCTP_SHA1_FINAL(x,y) sctp_sha1_final((unsigned char *)x, y) + +/* start OOTB only stuff */ +/* TODO IFT_LOOP is in net/if_types.h on Linux */ +#define IFT_LOOP 0x18 + +/* sctp_pcb.h */ + +#if defined(_WIN32) +#define SHUT_RD 1 +#define SHUT_WR 2 +#define SHUT_RDWR 3 +#endif +#define PRU_FLUSH_RD SHUT_RD +#define PRU_FLUSH_WR SHUT_WR +#define PRU_FLUSH_RDWR SHUT_RDWR + +/* netinet/ip_var.h defintions are behind an if defined for _KERNEL on FreeBSD */ +#define IP_RAWOUTPUT 0x2 + + +/* end OOTB only stuff */ + +#define AF_CONN 123 +struct sockaddr_conn { +#ifdef HAVE_SCONN_LEN + uint8_t sconn_len; + uint8_t sconn_family; +#else + uint16_t sconn_family; +#endif + uint16_t sconn_port; + void *sconn_addr; +}; + +typedef void *(*start_routine_t)(void *); + +extern int +sctp_userspace_thread_create(userland_thread_t *thread, start_routine_t start_routine); + +void +sctp_userspace_set_threadname(const char *name); + +/* + * SCTP protocol specific mbuf flags. + */ +#define M_NOTIFICATION M_PROTO5 /* SCTP notification */ + +/* + * IP output routines + */ + +/* Defining SCTP_IP_ID macro. + In netinet/ip_output.c, we have u_short ip_id; + In netinet/ip_var.h, we have extern u_short ip_id; (enclosed within _KERNEL_) + See static __inline uint16_t ip_newid(void) in netinet/ip_var.h + */ +#define SCTP_IP_ID(inp) (ip_id) + +/* need sctphdr to get port in SCTP_IP_OUTPUT. sctphdr defined in sctp.h */ +#include <netinet/sctp.h> +extern void sctp_userspace_ip_output(int *result, struct mbuf *o_pak, + sctp_route_t *ro, void *stcb, + uint32_t vrf_id); + +#define SCTP_IP_OUTPUT(result, o_pak, ro, inp, vrf_id) sctp_userspace_ip_output(&result, o_pak, ro, inp, vrf_id); + +#if defined(INET6) +extern void sctp_userspace_ip6_output(int *result, struct mbuf *o_pak, + struct route_in6 *ro, void *stcb, + uint32_t vrf_id); +#define SCTP_IP6_OUTPUT(result, o_pak, ro, ifp, inp, vrf_id) sctp_userspace_ip6_output(&result, o_pak, ro, inp, vrf_id); +#endif + + + +#if 0 +#define SCTP_IP6_OUTPUT(result, o_pak, ro, ifp, stcb, vrf_id) \ +{ \ + if (stcb && stcb->sctp_ep) \ + result = ip6_output(o_pak, \ + ((struct inpcb *)(stcb->sctp_ep))->in6p_outputopts, \ + (ro), 0, 0, ifp, NULL); \ + else \ + result = ip6_output(o_pak, NULL, (ro), 0, 0, ifp, NULL); \ +} +#endif + +struct mbuf * +sctp_get_mbuf_for_msg(unsigned int space_needed, int want_header, int how, int allonebuf, int type); + + +/* with the current included files, this is defined in Linux but + * in FreeBSD, it is behind a _KERNEL in sys/socket.h ... + */ +#if defined(__DragonFly__) || defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__native_client__) +/* stolen from /usr/include/sys/socket.h */ +#define CMSG_ALIGN(n) _ALIGN(n) +#elif defined(__NetBSD__) +#define CMSG_ALIGN(n) (((n) + __ALIGNBYTES) & ~__ALIGNBYTES) +#elif defined(__APPLE__) +#if !defined(__DARWIN_ALIGNBYTES) +#define __DARWIN_ALIGNBYTES (sizeof(__darwin_size_t) - 1) +#endif + +#if !defined(__DARWIN_ALIGN) +#define __DARWIN_ALIGN(p) ((__darwin_size_t)((char *)(uintptr_t)(p) + __DARWIN_ALIGNBYTES) &~ __DARWIN_ALIGNBYTES) +#endif + +#if !defined(__DARWIN_ALIGNBYTES32) +#define __DARWIN_ALIGNBYTES32 (sizeof(__uint32_t) - 1) +#endif + +#if !defined(__DARWIN_ALIGN32) +#define __DARWIN_ALIGN32(p) ((__darwin_size_t)((char *)(uintptr_t)(p) + __DARWIN_ALIGNBYTES32) &~ __DARWIN_ALIGNBYTES32) +#endif +#define CMSG_ALIGN(n) __DARWIN_ALIGN32(n) +#endif +#define I_AM_HERE \ + do { \ + SCTP_PRINTF("%s:%d at %s\n", __FILE__, __LINE__ , __func__); \ + } while (0) + +#ifndef timevalsub +#define timevalsub(tp1, tp2) \ + do { \ + (tp1)->tv_sec -= (tp2)->tv_sec; \ + (tp1)->tv_usec -= (tp2)->tv_usec; \ + if ((tp1)->tv_usec < 0) { \ + (tp1)->tv_sec--; \ + (tp1)->tv_usec += 1000000; \ + } \ + } while (0) +#endif + +#if defined(__linux__) +#if !defined(TAILQ_FOREACH_SAFE) +#define TAILQ_FOREACH_SAFE(var, head, field, tvar) \ + for ((var) = ((head)->tqh_first); \ + (var) && ((tvar) = TAILQ_NEXT((var), field), 1); \ + (var) = (tvar)) +#endif +#if !defined(LIST_FOREACH_SAFE) +#define LIST_FOREACH_SAFE(var, head, field, tvar) \ + for ((var) = ((head)->lh_first); \ + (var) && ((tvar) = LIST_NEXT((var), field), 1); \ + (var) = (tvar)) +#endif +#endif +#if defined(__DragonFly__) +#define TAILQ_FOREACH_SAFE TAILQ_FOREACH_MUTABLE +#define LIST_FOREACH_SAFE LIST_FOREACH_MUTABLE +#endif + +#if defined(__native_client__) +#define timercmp(tvp, uvp, cmp) \ + (((tvp)->tv_sec == (uvp)->tv_sec) ? \ + ((tvp)->tv_usec cmp (uvp)->tv_usec) : \ + ((tvp)->tv_sec cmp (uvp)->tv_sec)) +#endif + +#define SCTP_IS_LISTENING(inp) ((inp->sctp_flags & SCTP_PCB_FLAGS_ACCEPTING) != 0) + +#if defined(__APPLE__) || defined(__DragonFly__) || defined(__linux__) || defined(__native_client__) || defined(__NetBSD__) || defined(_WIN32) || defined(__Fuchsia__) || defined(__EMSCRIPTEN__) +int +timingsafe_bcmp(const void *, const void *, size_t); +#endif + +#endif diff --git a/netwerk/sctp/src/netinet/sctp_output.c b/netwerk/sctp/src/netinet/sctp_output.c new file mode 100644 index 0000000000..aa0aae260f --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_output.c @@ -0,0 +1,15037 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/proc.h> +#endif +#include <netinet/sctp_var.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_header.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_uio.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_auth.h> +#include <netinet/sctp_timer.h> +#include <netinet/sctp_asconf.h> +#include <netinet/sctp_indata.h> +#include <netinet/sctp_bsd_addr.h> +#include <netinet/sctp_input.h> +#include <netinet/sctp_crc32.h> +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <netinet/sctp_kdtrace.h> +#endif +#if defined(__linux__) +#define __FAVOR_BSD /* (on Ubuntu at least) enables UDP header field names like BSD in RFC 768 */ +#endif +#if defined(INET) || defined(INET6) +#if !defined(_WIN32) +#include <netinet/udp.h> +#endif +#endif +#if !defined(__Userspace__) +#if defined(__APPLE__) +#include <netinet/in.h> +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <netinet/udp_var.h> +#include <machine/in_cksum.h> +#endif +#endif +#if defined(__Userspace__) && defined(INET6) +#include <netinet6/sctp6_var.h> +#endif +#if defined(__APPLE__) && !defined(__Userspace__) +#if !(defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD)) +#define SCTP_MAX_LINKHDR 16 +#endif +#endif + +#define SCTP_MAX_GAPS_INARRAY 4 +struct sack_track { + uint8_t right_edge; /* mergable on the right edge */ + uint8_t left_edge; /* mergable on the left edge */ + uint8_t num_entries; + uint8_t spare; + struct sctp_gap_ack_block gaps[SCTP_MAX_GAPS_INARRAY]; +}; + +const struct sack_track sack_array[256] = { + {0, 0, 0, 0, /* 0x00 */ + {{0, 0}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 1, 0, /* 0x01 */ + {{0, 0}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x02 */ + {{1, 1}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 1, 0, /* 0x03 */ + {{0, 1}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x04 */ + {{2, 2}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x05 */ + {{0, 0}, + {2, 2}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x06 */ + {{1, 2}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 1, 0, /* 0x07 */ + {{0, 2}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x08 */ + {{3, 3}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x09 */ + {{0, 0}, + {3, 3}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x0a */ + {{1, 1}, + {3, 3}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x0b */ + {{0, 1}, + {3, 3}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x0c */ + {{2, 3}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x0d */ + {{0, 0}, + {2, 3}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x0e */ + {{1, 3}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 1, 0, /* 0x0f */ + {{0, 3}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x10 */ + {{4, 4}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x11 */ + {{0, 0}, + {4, 4}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x12 */ + {{1, 1}, + {4, 4}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x13 */ + {{0, 1}, + {4, 4}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x14 */ + {{2, 2}, + {4, 4}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x15 */ + {{0, 0}, + {2, 2}, + {4, 4}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x16 */ + {{1, 2}, + {4, 4}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x17 */ + {{0, 2}, + {4, 4}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x18 */ + {{3, 4}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x19 */ + {{0, 0}, + {3, 4}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x1a */ + {{1, 1}, + {3, 4}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x1b */ + {{0, 1}, + {3, 4}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x1c */ + {{2, 4}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x1d */ + {{0, 0}, + {2, 4}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x1e */ + {{1, 4}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 1, 0, /* 0x1f */ + {{0, 4}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x20 */ + {{5, 5}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x21 */ + {{0, 0}, + {5, 5}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x22 */ + {{1, 1}, + {5, 5}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x23 */ + {{0, 1}, + {5, 5}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x24 */ + {{2, 2}, + {5, 5}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x25 */ + {{0, 0}, + {2, 2}, + {5, 5}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x26 */ + {{1, 2}, + {5, 5}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x27 */ + {{0, 2}, + {5, 5}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x28 */ + {{3, 3}, + {5, 5}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x29 */ + {{0, 0}, + {3, 3}, + {5, 5}, + {0, 0} + } + }, + {0, 0, 3, 0, /* 0x2a */ + {{1, 1}, + {3, 3}, + {5, 5}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x2b */ + {{0, 1}, + {3, 3}, + {5, 5}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x2c */ + {{2, 3}, + {5, 5}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x2d */ + {{0, 0}, + {2, 3}, + {5, 5}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x2e */ + {{1, 3}, + {5, 5}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x2f */ + {{0, 3}, + {5, 5}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x30 */ + {{4, 5}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x31 */ + {{0, 0}, + {4, 5}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x32 */ + {{1, 1}, + {4, 5}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x33 */ + {{0, 1}, + {4, 5}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x34 */ + {{2, 2}, + {4, 5}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x35 */ + {{0, 0}, + {2, 2}, + {4, 5}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x36 */ + {{1, 2}, + {4, 5}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x37 */ + {{0, 2}, + {4, 5}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x38 */ + {{3, 5}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x39 */ + {{0, 0}, + {3, 5}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x3a */ + {{1, 1}, + {3, 5}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x3b */ + {{0, 1}, + {3, 5}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x3c */ + {{2, 5}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x3d */ + {{0, 0}, + {2, 5}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x3e */ + {{1, 5}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 1, 0, /* 0x3f */ + {{0, 5}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x40 */ + {{6, 6}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x41 */ + {{0, 0}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x42 */ + {{1, 1}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x43 */ + {{0, 1}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x44 */ + {{2, 2}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x45 */ + {{0, 0}, + {2, 2}, + {6, 6}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x46 */ + {{1, 2}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x47 */ + {{0, 2}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x48 */ + {{3, 3}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x49 */ + {{0, 0}, + {3, 3}, + {6, 6}, + {0, 0} + } + }, + {0, 0, 3, 0, /* 0x4a */ + {{1, 1}, + {3, 3}, + {6, 6}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x4b */ + {{0, 1}, + {3, 3}, + {6, 6}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x4c */ + {{2, 3}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x4d */ + {{0, 0}, + {2, 3}, + {6, 6}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x4e */ + {{1, 3}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x4f */ + {{0, 3}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x50 */ + {{4, 4}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x51 */ + {{0, 0}, + {4, 4}, + {6, 6}, + {0, 0} + } + }, + {0, 0, 3, 0, /* 0x52 */ + {{1, 1}, + {4, 4}, + {6, 6}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x53 */ + {{0, 1}, + {4, 4}, + {6, 6}, + {0, 0} + } + }, + {0, 0, 3, 0, /* 0x54 */ + {{2, 2}, + {4, 4}, + {6, 6}, + {0, 0} + } + }, + {1, 0, 4, 0, /* 0x55 */ + {{0, 0}, + {2, 2}, + {4, 4}, + {6, 6} + } + }, + {0, 0, 3, 0, /* 0x56 */ + {{1, 2}, + {4, 4}, + {6, 6}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x57 */ + {{0, 2}, + {4, 4}, + {6, 6}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x58 */ + {{3, 4}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x59 */ + {{0, 0}, + {3, 4}, + {6, 6}, + {0, 0} + } + }, + {0, 0, 3, 0, /* 0x5a */ + {{1, 1}, + {3, 4}, + {6, 6}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x5b */ + {{0, 1}, + {3, 4}, + {6, 6}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x5c */ + {{2, 4}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x5d */ + {{0, 0}, + {2, 4}, + {6, 6}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x5e */ + {{1, 4}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x5f */ + {{0, 4}, + {6, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x60 */ + {{5, 6}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x61 */ + {{0, 0}, + {5, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x62 */ + {{1, 1}, + {5, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x63 */ + {{0, 1}, + {5, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x64 */ + {{2, 2}, + {5, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x65 */ + {{0, 0}, + {2, 2}, + {5, 6}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x66 */ + {{1, 2}, + {5, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x67 */ + {{0, 2}, + {5, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x68 */ + {{3, 3}, + {5, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x69 */ + {{0, 0}, + {3, 3}, + {5, 6}, + {0, 0} + } + }, + {0, 0, 3, 0, /* 0x6a */ + {{1, 1}, + {3, 3}, + {5, 6}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x6b */ + {{0, 1}, + {3, 3}, + {5, 6}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x6c */ + {{2, 3}, + {5, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x6d */ + {{0, 0}, + {2, 3}, + {5, 6}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x6e */ + {{1, 3}, + {5, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x6f */ + {{0, 3}, + {5, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x70 */ + {{4, 6}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x71 */ + {{0, 0}, + {4, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x72 */ + {{1, 1}, + {4, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x73 */ + {{0, 1}, + {4, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x74 */ + {{2, 2}, + {4, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 3, 0, /* 0x75 */ + {{0, 0}, + {2, 2}, + {4, 6}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x76 */ + {{1, 2}, + {4, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x77 */ + {{0, 2}, + {4, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x78 */ + {{3, 6}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x79 */ + {{0, 0}, + {3, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 2, 0, /* 0x7a */ + {{1, 1}, + {3, 6}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x7b */ + {{0, 1}, + {3, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x7c */ + {{2, 6}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 2, 0, /* 0x7d */ + {{0, 0}, + {2, 6}, + {0, 0}, + {0, 0} + } + }, + {0, 0, 1, 0, /* 0x7e */ + {{1, 6}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 0, 1, 0, /* 0x7f */ + {{0, 6}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 1, 0, /* 0x80 */ + {{7, 7}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0x81 */ + {{0, 0}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0x82 */ + {{1, 1}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0x83 */ + {{0, 1}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0x84 */ + {{2, 2}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0x85 */ + {{0, 0}, + {2, 2}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0x86 */ + {{1, 2}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0x87 */ + {{0, 2}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0x88 */ + {{3, 3}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0x89 */ + {{0, 0}, + {3, 3}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0x8a */ + {{1, 1}, + {3, 3}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0x8b */ + {{0, 1}, + {3, 3}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0x8c */ + {{2, 3}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0x8d */ + {{0, 0}, + {2, 3}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0x8e */ + {{1, 3}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0x8f */ + {{0, 3}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0x90 */ + {{4, 4}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0x91 */ + {{0, 0}, + {4, 4}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0x92 */ + {{1, 1}, + {4, 4}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0x93 */ + {{0, 1}, + {4, 4}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0x94 */ + {{2, 2}, + {4, 4}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 4, 0, /* 0x95 */ + {{0, 0}, + {2, 2}, + {4, 4}, + {7, 7} + } + }, + {0, 1, 3, 0, /* 0x96 */ + {{1, 2}, + {4, 4}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0x97 */ + {{0, 2}, + {4, 4}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0x98 */ + {{3, 4}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0x99 */ + {{0, 0}, + {3, 4}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0x9a */ + {{1, 1}, + {3, 4}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0x9b */ + {{0, 1}, + {3, 4}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0x9c */ + {{2, 4}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0x9d */ + {{0, 0}, + {2, 4}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0x9e */ + {{1, 4}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0x9f */ + {{0, 4}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xa0 */ + {{5, 5}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xa1 */ + {{0, 0}, + {5, 5}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0xa2 */ + {{1, 1}, + {5, 5}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xa3 */ + {{0, 1}, + {5, 5}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0xa4 */ + {{2, 2}, + {5, 5}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 4, 0, /* 0xa5 */ + {{0, 0}, + {2, 2}, + {5, 5}, + {7, 7} + } + }, + {0, 1, 3, 0, /* 0xa6 */ + {{1, 2}, + {5, 5}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xa7 */ + {{0, 2}, + {5, 5}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0xa8 */ + {{3, 3}, + {5, 5}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 4, 0, /* 0xa9 */ + {{0, 0}, + {3, 3}, + {5, 5}, + {7, 7} + } + }, + {0, 1, 4, 0, /* 0xaa */ + {{1, 1}, + {3, 3}, + {5, 5}, + {7, 7} + } + }, + {1, 1, 4, 0, /* 0xab */ + {{0, 1}, + {3, 3}, + {5, 5}, + {7, 7} + } + }, + {0, 1, 3, 0, /* 0xac */ + {{2, 3}, + {5, 5}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 4, 0, /* 0xad */ + {{0, 0}, + {2, 3}, + {5, 5}, + {7, 7} + } + }, + {0, 1, 3, 0, /* 0xae */ + {{1, 3}, + {5, 5}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xaf */ + {{0, 3}, + {5, 5}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xb0 */ + {{4, 5}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xb1 */ + {{0, 0}, + {4, 5}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0xb2 */ + {{1, 1}, + {4, 5}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xb3 */ + {{0, 1}, + {4, 5}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0xb4 */ + {{2, 2}, + {4, 5}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 4, 0, /* 0xb5 */ + {{0, 0}, + {2, 2}, + {4, 5}, + {7, 7} + } + }, + {0, 1, 3, 0, /* 0xb6 */ + {{1, 2}, + {4, 5}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xb7 */ + {{0, 2}, + {4, 5}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xb8 */ + {{3, 5}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xb9 */ + {{0, 0}, + {3, 5}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0xba */ + {{1, 1}, + {3, 5}, + {7, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xbb */ + {{0, 1}, + {3, 5}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xbc */ + {{2, 5}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xbd */ + {{0, 0}, + {2, 5}, + {7, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xbe */ + {{1, 5}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xbf */ + {{0, 5}, + {7, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 1, 0, /* 0xc0 */ + {{6, 7}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xc1 */ + {{0, 0}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xc2 */ + {{1, 1}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xc3 */ + {{0, 1}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xc4 */ + {{2, 2}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xc5 */ + {{0, 0}, + {2, 2}, + {6, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xc6 */ + {{1, 2}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xc7 */ + {{0, 2}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xc8 */ + {{3, 3}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xc9 */ + {{0, 0}, + {3, 3}, + {6, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0xca */ + {{1, 1}, + {3, 3}, + {6, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xcb */ + {{0, 1}, + {3, 3}, + {6, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xcc */ + {{2, 3}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xcd */ + {{0, 0}, + {2, 3}, + {6, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xce */ + {{1, 3}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xcf */ + {{0, 3}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xd0 */ + {{4, 4}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xd1 */ + {{0, 0}, + {4, 4}, + {6, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0xd2 */ + {{1, 1}, + {4, 4}, + {6, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xd3 */ + {{0, 1}, + {4, 4}, + {6, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0xd4 */ + {{2, 2}, + {4, 4}, + {6, 7}, + {0, 0} + } + }, + {1, 1, 4, 0, /* 0xd5 */ + {{0, 0}, + {2, 2}, + {4, 4}, + {6, 7} + } + }, + {0, 1, 3, 0, /* 0xd6 */ + {{1, 2}, + {4, 4}, + {6, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xd7 */ + {{0, 2}, + {4, 4}, + {6, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xd8 */ + {{3, 4}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xd9 */ + {{0, 0}, + {3, 4}, + {6, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0xda */ + {{1, 1}, + {3, 4}, + {6, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xdb */ + {{0, 1}, + {3, 4}, + {6, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xdc */ + {{2, 4}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xdd */ + {{0, 0}, + {2, 4}, + {6, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xde */ + {{1, 4}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xdf */ + {{0, 4}, + {6, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 1, 0, /* 0xe0 */ + {{5, 7}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xe1 */ + {{0, 0}, + {5, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xe2 */ + {{1, 1}, + {5, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xe3 */ + {{0, 1}, + {5, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xe4 */ + {{2, 2}, + {5, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xe5 */ + {{0, 0}, + {2, 2}, + {5, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xe6 */ + {{1, 2}, + {5, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xe7 */ + {{0, 2}, + {5, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xe8 */ + {{3, 3}, + {5, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xe9 */ + {{0, 0}, + {3, 3}, + {5, 7}, + {0, 0} + } + }, + {0, 1, 3, 0, /* 0xea */ + {{1, 1}, + {3, 3}, + {5, 7}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xeb */ + {{0, 1}, + {3, 3}, + {5, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xec */ + {{2, 3}, + {5, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xed */ + {{0, 0}, + {2, 3}, + {5, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xee */ + {{1, 3}, + {5, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xef */ + {{0, 3}, + {5, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 1, 0, /* 0xf0 */ + {{4, 7}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xf1 */ + {{0, 0}, + {4, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xf2 */ + {{1, 1}, + {4, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xf3 */ + {{0, 1}, + {4, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xf4 */ + {{2, 2}, + {4, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 3, 0, /* 0xf5 */ + {{0, 0}, + {2, 2}, + {4, 7}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xf6 */ + {{1, 2}, + {4, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xf7 */ + {{0, 2}, + {4, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 1, 0, /* 0xf8 */ + {{3, 7}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xf9 */ + {{0, 0}, + {3, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 2, 0, /* 0xfa */ + {{1, 1}, + {3, 7}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xfb */ + {{0, 1}, + {3, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 1, 0, /* 0xfc */ + {{2, 7}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 2, 0, /* 0xfd */ + {{0, 0}, + {2, 7}, + {0, 0}, + {0, 0} + } + }, + {0, 1, 1, 0, /* 0xfe */ + {{1, 7}, + {0, 0}, + {0, 0}, + {0, 0} + } + }, + {1, 1, 1, 0, /* 0xff */ + {{0, 7}, + {0, 0}, + {0, 0}, + {0, 0} + } + } +}; + +int +sctp_is_address_in_scope(struct sctp_ifa *ifa, + struct sctp_scoping *scope, + int do_update) +{ + if ((scope->loopback_scope == 0) && + (ifa->ifn_p) && SCTP_IFN_IS_IFT_LOOP(ifa->ifn_p)) { + /* + * skip loopback if not in scope * + */ + return (0); + } + switch (ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: + if (scope->ipv4_addr_legal) { + struct sockaddr_in *sin; + + sin = &ifa->address.sin; + if (sin->sin_addr.s_addr == 0) { + /* not in scope , unspecified */ + return (0); + } + if ((scope->ipv4_local_scope == 0) && + (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { + /* private address not in scope */ + return (0); + } + } else { + return (0); + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (scope->ipv6_addr_legal) { + struct sockaddr_in6 *sin6; + + /* Must update the flags, bummer, which + * means any IFA locks must now be applied HERE <-> + */ + if (do_update) { + sctp_gather_internal_ifa_flags(ifa); + } + if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { + return (0); + } + /* ok to use deprecated addresses? */ + sin6 = &ifa->address.sin6; + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + /* skip unspecified addresses */ + return (0); + } + if ( /* (local_scope == 0) && */ + (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr))) { + return (0); + } + if ((scope->site_scope == 0) && + (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { + return (0); + } + } else { + return (0); + } + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + if (!scope->conn_addr_legal) { + return (0); + } + break; +#endif + default: + return (0); + } + return (1); +} + +static struct mbuf * +sctp_add_addr_to_mbuf(struct mbuf *m, struct sctp_ifa *ifa, uint16_t *len) +{ +#if defined(INET) || defined(INET6) + struct sctp_paramhdr *paramh; + struct mbuf *mret; + uint16_t plen; +#endif + + switch (ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: + plen = (uint16_t)sizeof(struct sctp_ipv4addr_param); + break; +#endif +#ifdef INET6 + case AF_INET6: + plen = (uint16_t)sizeof(struct sctp_ipv6addr_param); + break; +#endif + default: + return (m); + } +#if defined(INET) || defined(INET6) + if (M_TRAILINGSPACE(m) >= plen) { + /* easy side we just drop it on the end */ + paramh = (struct sctp_paramhdr *)(SCTP_BUF_AT(m, SCTP_BUF_LEN(m))); + mret = m; + } else { + /* Need more space */ + mret = m; + while (SCTP_BUF_NEXT(mret) != NULL) { + mret = SCTP_BUF_NEXT(mret); + } + SCTP_BUF_NEXT(mret) = sctp_get_mbuf_for_msg(plen, 0, M_NOWAIT, 1, MT_DATA); + if (SCTP_BUF_NEXT(mret) == NULL) { + /* We are hosed, can't add more addresses */ + return (m); + } + mret = SCTP_BUF_NEXT(mret); + paramh = mtod(mret, struct sctp_paramhdr *); + } + /* now add the parameter */ + switch (ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: + { + struct sctp_ipv4addr_param *ipv4p; + struct sockaddr_in *sin; + + sin = &ifa->address.sin; + ipv4p = (struct sctp_ipv4addr_param *)paramh; + paramh->param_type = htons(SCTP_IPV4_ADDRESS); + paramh->param_length = htons(plen); + ipv4p->addr = sin->sin_addr.s_addr; + SCTP_BUF_LEN(mret) += plen; + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + struct sctp_ipv6addr_param *ipv6p; + struct sockaddr_in6 *sin6; + + sin6 = &ifa->address.sin6; + ipv6p = (struct sctp_ipv6addr_param *)paramh; + paramh->param_type = htons(SCTP_IPV6_ADDRESS); + paramh->param_length = htons(plen); + memcpy(ipv6p->addr, &sin6->sin6_addr, + sizeof(ipv6p->addr)); +#if defined(SCTP_EMBEDDED_V6_SCOPE) + /* clear embedded scope in the address */ + in6_clearscope((struct in6_addr *)ipv6p->addr); +#endif + SCTP_BUF_LEN(mret) += plen; + break; + } +#endif + default: + return (m); + } + if (len != NULL) { + *len += plen; + } + return (mret); +#endif +} + +struct mbuf * +sctp_add_addresses_to_i_ia(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct sctp_scoping *scope, + struct mbuf *m_at, int cnt_inits_to, + uint16_t *padding_len, uint16_t *chunk_len) +{ + struct sctp_vrf *vrf = NULL; + int cnt, limit_out = 0, total_count; + uint32_t vrf_id; + + vrf_id = inp->def_vrf_id; + SCTP_IPI_ADDR_RLOCK(); + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) { + SCTP_IPI_ADDR_RUNLOCK(); + return (m_at); + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + struct sctp_ifa *sctp_ifap; + struct sctp_ifn *sctp_ifnp; + + cnt = cnt_inits_to; + if (vrf->total_ifa_count > SCTP_COUNT_LIMIT) { + limit_out = 1; + cnt = SCTP_ADDRESS_LIMIT; + goto skip_count; + } + LIST_FOREACH(sctp_ifnp, &vrf->ifnlist, next_ifn) { + if ((scope->loopback_scope == 0) && + SCTP_IFN_IS_IFT_LOOP(sctp_ifnp)) { + /* + * Skip loopback devices if loopback_scope + * not set + */ + continue; + } + LIST_FOREACH(sctp_ifap, &sctp_ifnp->ifalist, next_ifa) { +#if defined(__FreeBSD__) && !defined(__Userspace__) +#ifdef INET + if ((sctp_ifap->address.sa.sa_family == AF_INET) && + (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sctp_ifap->address.sin.sin_addr) != 0)) { + continue; + } +#endif +#ifdef INET6 + if ((sctp_ifap->address.sa.sa_family == AF_INET6) && + (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sctp_ifap->address.sin6.sin6_addr) != 0)) { + continue; + } +#endif +#endif + if (sctp_is_addr_restricted(stcb, sctp_ifap)) { + continue; + } +#if defined(__Userspace__) + if (sctp_ifap->address.sa.sa_family == AF_CONN) { + continue; + } +#endif + if (sctp_is_address_in_scope(sctp_ifap, scope, 1) == 0) { + continue; + } + cnt++; + if (cnt > SCTP_ADDRESS_LIMIT) { + break; + } + } + if (cnt > SCTP_ADDRESS_LIMIT) { + break; + } + } + skip_count: + if (cnt > 1) { + total_count = 0; + LIST_FOREACH(sctp_ifnp, &vrf->ifnlist, next_ifn) { + cnt = 0; + if ((scope->loopback_scope == 0) && + SCTP_IFN_IS_IFT_LOOP(sctp_ifnp)) { + /* + * Skip loopback devices if + * loopback_scope not set + */ + continue; + } + LIST_FOREACH(sctp_ifap, &sctp_ifnp->ifalist, next_ifa) { +#if defined(__FreeBSD__) && !defined(__Userspace__) +#ifdef INET + if ((sctp_ifap->address.sa.sa_family == AF_INET) && + (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sctp_ifap->address.sin.sin_addr) != 0)) { + continue; + } +#endif +#ifdef INET6 + if ((sctp_ifap->address.sa.sa_family == AF_INET6) && + (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sctp_ifap->address.sin6.sin6_addr) != 0)) { + continue; + } +#endif +#endif + if (sctp_is_addr_restricted(stcb, sctp_ifap)) { + continue; + } +#if defined(__Userspace__) + if (sctp_ifap->address.sa.sa_family == AF_CONN) { + continue; + } +#endif + if (sctp_is_address_in_scope(sctp_ifap, + scope, 0) == 0) { + continue; + } + if ((chunk_len != NULL) && + (padding_len != NULL) && + (*padding_len > 0)) { + memset(mtod(m_at, caddr_t) + *chunk_len, 0, *padding_len); + SCTP_BUF_LEN(m_at) += *padding_len; + *chunk_len += *padding_len; + *padding_len = 0; + } + m_at = sctp_add_addr_to_mbuf(m_at, sctp_ifap, chunk_len); + if (limit_out) { + cnt++; + total_count++; + if (cnt >= 2) { + /* two from each address */ + break; + } + if (total_count > SCTP_ADDRESS_LIMIT) { + /* No more addresses */ + break; + } + } + } + } + } + } else { + struct sctp_laddr *laddr; + + cnt = cnt_inits_to; + /* First, how many ? */ + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa == NULL) { + continue; + } + if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) + /* Address being deleted by the system, dont + * list. + */ + continue; + if (laddr->action == SCTP_DEL_IP_ADDRESS) { + /* Address being deleted on this ep + * don't list. + */ + continue; + } +#if defined(__Userspace__) + if (laddr->ifa->address.sa.sa_family == AF_CONN) { + continue; + } +#endif + if (sctp_is_address_in_scope(laddr->ifa, + scope, 1) == 0) { + continue; + } + cnt++; + } + /* + * To get through a NAT we only list addresses if we have + * more than one. That way if you just bind a single address + * we let the source of the init dictate our address. + */ + if (cnt > 1) { + cnt = cnt_inits_to; + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa == NULL) { + continue; + } + if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { + continue; + } +#if defined(__Userspace__) + if (laddr->ifa->address.sa.sa_family == AF_CONN) { + continue; + } +#endif + if (sctp_is_address_in_scope(laddr->ifa, + scope, 0) == 0) { + continue; + } + if ((chunk_len != NULL) && + (padding_len != NULL) && + (*padding_len > 0)) { + memset(mtod(m_at, caddr_t) + *chunk_len, 0, *padding_len); + SCTP_BUF_LEN(m_at) += *padding_len; + *chunk_len += *padding_len; + *padding_len = 0; + } + m_at = sctp_add_addr_to_mbuf(m_at, laddr->ifa, chunk_len); + cnt++; + if (cnt >= SCTP_ADDRESS_LIMIT) { + break; + } + } + } + } + SCTP_IPI_ADDR_RUNLOCK(); + return (m_at); +} + +static struct sctp_ifa * +sctp_is_ifa_addr_preferred(struct sctp_ifa *ifa, + uint8_t dest_is_loop, + uint8_t dest_is_priv, + sa_family_t fam) +{ + uint8_t dest_is_global = 0; + /* dest_is_priv is true if destination is a private address */ + /* dest_is_loop is true if destination is a loopback addresses */ + + /** + * Here we determine if its a preferred address. A preferred address + * means it is the same scope or higher scope then the destination. + * L = loopback, P = private, G = global + * ----------------------------------------- + * src | dest | result + * ---------------------------------------- + * L | L | yes + * ----------------------------------------- + * P | L | yes-v4 no-v6 + * ----------------------------------------- + * G | L | yes-v4 no-v6 + * ----------------------------------------- + * L | P | no + * ----------------------------------------- + * P | P | yes + * ----------------------------------------- + * G | P | no + * ----------------------------------------- + * L | G | no + * ----------------------------------------- + * P | G | no + * ----------------------------------------- + * G | G | yes + * ----------------------------------------- + */ + + if (ifa->address.sa.sa_family != fam) { + /* forget mis-matched family */ + return (NULL); + } + if ((dest_is_priv == 0) && (dest_is_loop == 0)) { + dest_is_global = 1; + } + SCTPDBG(SCTP_DEBUG_OUTPUT2, "Is destination preferred:"); + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &ifa->address.sa); + /* Ok the address may be ok */ +#ifdef INET6 + if (fam == AF_INET6) { + /* ok to use deprecated addresses? no lets not! */ + if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { + SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:1\n"); + return (NULL); + } + if (ifa->src_is_priv && !ifa->src_is_loop) { + if (dest_is_loop) { + SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:2\n"); + return (NULL); + } + } + if (ifa->src_is_glob) { + if (dest_is_loop) { + SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:3\n"); + return (NULL); + } + } + } +#endif + /* Now that we know what is what, implement or table + * this could in theory be done slicker (it used to be), but this + * is straightforward and easier to validate :-) + */ + SCTPDBG(SCTP_DEBUG_OUTPUT3, "src_loop:%d src_priv:%d src_glob:%d\n", + ifa->src_is_loop, ifa->src_is_priv, ifa->src_is_glob); + SCTPDBG(SCTP_DEBUG_OUTPUT3, "dest_loop:%d dest_priv:%d dest_glob:%d\n", + dest_is_loop, dest_is_priv, dest_is_global); + + if ((ifa->src_is_loop) && (dest_is_priv)) { + SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:4\n"); + return (NULL); + } + if ((ifa->src_is_glob) && (dest_is_priv)) { + SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:5\n"); + return (NULL); + } + if ((ifa->src_is_loop) && (dest_is_global)) { + SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:6\n"); + return (NULL); + } + if ((ifa->src_is_priv) && (dest_is_global)) { + SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:7\n"); + return (NULL); + } + SCTPDBG(SCTP_DEBUG_OUTPUT3, "YES\n"); + /* its a preferred address */ + return (ifa); +} + +static struct sctp_ifa * +sctp_is_ifa_addr_acceptable(struct sctp_ifa *ifa, + uint8_t dest_is_loop, + uint8_t dest_is_priv, + sa_family_t fam) +{ + uint8_t dest_is_global = 0; + + /** + * Here we determine if its a acceptable address. A acceptable + * address means it is the same scope or higher scope but we can + * allow for NAT which means its ok to have a global dest and a + * private src. + * + * L = loopback, P = private, G = global + * ----------------------------------------- + * src | dest | result + * ----------------------------------------- + * L | L | yes + * ----------------------------------------- + * P | L | yes-v4 no-v6 + * ----------------------------------------- + * G | L | yes + * ----------------------------------------- + * L | P | no + * ----------------------------------------- + * P | P | yes + * ----------------------------------------- + * G | P | yes - May not work + * ----------------------------------------- + * L | G | no + * ----------------------------------------- + * P | G | yes - May not work + * ----------------------------------------- + * G | G | yes + * ----------------------------------------- + */ + + if (ifa->address.sa.sa_family != fam) { + /* forget non matching family */ + SCTPDBG(SCTP_DEBUG_OUTPUT3, "ifa_fam:%d fam:%d\n", + ifa->address.sa.sa_family, fam); + return (NULL); + } + /* Ok the address may be ok */ + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT3, &ifa->address.sa); + SCTPDBG(SCTP_DEBUG_OUTPUT3, "dst_is_loop:%d dest_is_priv:%d\n", + dest_is_loop, dest_is_priv); + if ((dest_is_loop == 0) && (dest_is_priv == 0)) { + dest_is_global = 1; + } +#ifdef INET6 + if (fam == AF_INET6) { + /* ok to use deprecated addresses? */ + if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { + return (NULL); + } + if (ifa->src_is_priv) { + /* Special case, linklocal to loop */ + if (dest_is_loop) + return (NULL); + } + } +#endif + /* + * Now that we know what is what, implement our table. + * This could in theory be done slicker (it used to be), but this + * is straightforward and easier to validate :-) + */ + SCTPDBG(SCTP_DEBUG_OUTPUT3, "ifa->src_is_loop:%d dest_is_priv:%d\n", + ifa->src_is_loop, + dest_is_priv); + if ((ifa->src_is_loop == 1) && (dest_is_priv)) { + return (NULL); + } + SCTPDBG(SCTP_DEBUG_OUTPUT3, "ifa->src_is_loop:%d dest_is_glob:%d\n", + ifa->src_is_loop, + dest_is_global); + if ((ifa->src_is_loop == 1) && (dest_is_global)) { + return (NULL); + } + SCTPDBG(SCTP_DEBUG_OUTPUT3, "address is acceptable\n"); + /* its an acceptable address */ + return (ifa); +} + +int +sctp_is_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) +{ + struct sctp_laddr *laddr; + + if (stcb == NULL) { + /* There are no restrictions, no TCB :-) */ + return (0); + } + LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) { + if (laddr->ifa == NULL) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "%s: NULL ifa\n", + __func__); + continue; + } + if (laddr->ifa == ifa) { + /* Yes it is on the list */ + return (1); + } + } + return (0); +} + +int +sctp_is_addr_in_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa) +{ + struct sctp_laddr *laddr; + + if (ifa == NULL) + return (0); + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa == NULL) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "%s: NULL ifa\n", + __func__); + continue; + } + if ((laddr->ifa == ifa) && laddr->action == 0) + /* same pointer */ + return (1); + } + return (0); +} + +static struct sctp_ifa * +sctp_choose_boundspecific_inp(struct sctp_inpcb *inp, + sctp_route_t *ro, + uint32_t vrf_id, + int non_asoc_addr_ok, + uint8_t dest_is_priv, + uint8_t dest_is_loop, + sa_family_t fam) +{ + struct sctp_laddr *laddr, *starting_point; + void *ifn; + int resettotop = 0; + struct sctp_ifn *sctp_ifn; + struct sctp_ifa *sctp_ifa, *sifa; + struct sctp_vrf *vrf; + uint32_t ifn_index; + + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) + return (NULL); + + ifn = SCTP_GET_IFN_VOID_FROM_ROUTE(ro); + ifn_index = SCTP_GET_IF_INDEX_FROM_ROUTE(ro); + sctp_ifn = sctp_find_ifn(ifn, ifn_index); + /* + * first question, is the ifn we will emit on in our list, if so, we + * want such an address. Note that we first looked for a + * preferred address. + */ + if (sctp_ifn) { + /* is a preferred one on the interface we route out? */ + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { +#if defined(__FreeBSD__) && !defined(__Userspace__) +#ifdef INET + if ((sctp_ifa->address.sa.sa_family == AF_INET) && + (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sctp_ifa->address.sin.sin_addr) != 0)) { + continue; + } +#endif +#ifdef INET6 + if ((sctp_ifa->address.sa.sa_family == AF_INET6) && + (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sctp_ifa->address.sin6.sin6_addr) != 0)) { + continue; + } +#endif +#endif + if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && + (non_asoc_addr_ok == 0)) + continue; + sifa = sctp_is_ifa_addr_preferred(sctp_ifa, + dest_is_loop, + dest_is_priv, fam); + if (sifa == NULL) + continue; + if (sctp_is_addr_in_ep(inp, sifa)) { + atomic_add_int(&sifa->refcount, 1); + return (sifa); + } + } + } + /* + * ok, now we now need to find one on the list of the addresses. + * We can't get one on the emitting interface so let's find first + * a preferred one. If not that an acceptable one otherwise... + * we return NULL. + */ + starting_point = inp->next_addr_touse; + once_again: + if (inp->next_addr_touse == NULL) { + inp->next_addr_touse = LIST_FIRST(&inp->sctp_addr_list); + resettotop = 1; + } + for (laddr = inp->next_addr_touse; laddr; + laddr = LIST_NEXT(laddr, sctp_nxt_addr)) { + if (laddr->ifa == NULL) { + /* address has been removed */ + continue; + } + if (laddr->action == SCTP_DEL_IP_ADDRESS) { + /* address is being deleted */ + continue; + } + sifa = sctp_is_ifa_addr_preferred(laddr->ifa, dest_is_loop, + dest_is_priv, fam); + if (sifa == NULL) + continue; + atomic_add_int(&sifa->refcount, 1); + return (sifa); + } + if (resettotop == 0) { + inp->next_addr_touse = NULL; + goto once_again; + } + + inp->next_addr_touse = starting_point; + resettotop = 0; + once_again_too: + if (inp->next_addr_touse == NULL) { + inp->next_addr_touse = LIST_FIRST(&inp->sctp_addr_list); + resettotop = 1; + } + + /* ok, what about an acceptable address in the inp */ + for (laddr = inp->next_addr_touse; laddr; + laddr = LIST_NEXT(laddr, sctp_nxt_addr)) { + if (laddr->ifa == NULL) { + /* address has been removed */ + continue; + } + if (laddr->action == SCTP_DEL_IP_ADDRESS) { + /* address is being deleted */ + continue; + } + sifa = sctp_is_ifa_addr_acceptable(laddr->ifa, dest_is_loop, + dest_is_priv, fam); + if (sifa == NULL) + continue; + atomic_add_int(&sifa->refcount, 1); + return (sifa); + } + if (resettotop == 0) { + inp->next_addr_touse = NULL; + goto once_again_too; + } + + /* + * no address bound can be a source for the destination we are in + * trouble + */ + return (NULL); +} + +static struct sctp_ifa * +sctp_choose_boundspecific_stcb(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + sctp_route_t *ro, + uint32_t vrf_id, + uint8_t dest_is_priv, + uint8_t dest_is_loop, + int non_asoc_addr_ok, + sa_family_t fam) +{ + struct sctp_laddr *laddr, *starting_point; + void *ifn; + struct sctp_ifn *sctp_ifn; + struct sctp_ifa *sctp_ifa, *sifa; + uint8_t start_at_beginning = 0; + struct sctp_vrf *vrf; + uint32_t ifn_index; + + /* + * first question, is the ifn we will emit on in our list, if so, we + * want that one. + */ + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) + return (NULL); + + ifn = SCTP_GET_IFN_VOID_FROM_ROUTE(ro); + ifn_index = SCTP_GET_IF_INDEX_FROM_ROUTE(ro); + sctp_ifn = sctp_find_ifn(ifn, ifn_index); + + /* + * first question, is the ifn we will emit on in our list? If so, + * we want that one. First we look for a preferred. Second, we go + * for an acceptable. + */ + if (sctp_ifn) { + /* first try for a preferred address on the ep */ + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { +#if defined(__FreeBSD__) && !defined(__Userspace__) +#ifdef INET + if ((sctp_ifa->address.sa.sa_family == AF_INET) && + (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sctp_ifa->address.sin.sin_addr) != 0)) { + continue; + } +#endif +#ifdef INET6 + if ((sctp_ifa->address.sa.sa_family == AF_INET6) && + (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sctp_ifa->address.sin6.sin6_addr) != 0)) { + continue; + } +#endif +#endif + if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) + continue; + if (sctp_is_addr_in_ep(inp, sctp_ifa)) { + sifa = sctp_is_ifa_addr_preferred(sctp_ifa, dest_is_loop, dest_is_priv, fam); + if (sifa == NULL) + continue; + if (((non_asoc_addr_ok == 0) && + (sctp_is_addr_restricted(stcb, sifa))) || + (non_asoc_addr_ok && + (sctp_is_addr_restricted(stcb, sifa)) && + (!sctp_is_addr_pending(stcb, sifa)))) { + /* on the no-no list */ + continue; + } + atomic_add_int(&sifa->refcount, 1); + return (sifa); + } + } + /* next try for an acceptable address on the ep */ + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { +#if defined(__FreeBSD__) && !defined(__Userspace__) +#ifdef INET + if ((sctp_ifa->address.sa.sa_family == AF_INET) && + (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sctp_ifa->address.sin.sin_addr) != 0)) { + continue; + } +#endif +#ifdef INET6 + if ((sctp_ifa->address.sa.sa_family == AF_INET6) && + (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sctp_ifa->address.sin6.sin6_addr) != 0)) { + continue; + } +#endif +#endif + if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) + continue; + if (sctp_is_addr_in_ep(inp, sctp_ifa)) { + sifa= sctp_is_ifa_addr_acceptable(sctp_ifa, dest_is_loop, dest_is_priv,fam); + if (sifa == NULL) + continue; + if (((non_asoc_addr_ok == 0) && + (sctp_is_addr_restricted(stcb, sifa))) || + (non_asoc_addr_ok && + (sctp_is_addr_restricted(stcb, sifa)) && + (!sctp_is_addr_pending(stcb, sifa)))) { + /* on the no-no list */ + continue; + } + atomic_add_int(&sifa->refcount, 1); + return (sifa); + } + } + } + /* + * if we can't find one like that then we must look at all + * addresses bound to pick one at first preferable then + * secondly acceptable. + */ + starting_point = stcb->asoc.last_used_address; + sctp_from_the_top: + if (stcb->asoc.last_used_address == NULL) { + start_at_beginning = 1; + stcb->asoc.last_used_address = LIST_FIRST(&inp->sctp_addr_list); + } + /* search beginning with the last used address */ + for (laddr = stcb->asoc.last_used_address; laddr; + laddr = LIST_NEXT(laddr, sctp_nxt_addr)) { + if (laddr->ifa == NULL) { + /* address has been removed */ + continue; + } + if (laddr->action == SCTP_DEL_IP_ADDRESS) { + /* address is being deleted */ + continue; + } + sifa = sctp_is_ifa_addr_preferred(laddr->ifa, dest_is_loop, dest_is_priv, fam); + if (sifa == NULL) + continue; + if (((non_asoc_addr_ok == 0) && + (sctp_is_addr_restricted(stcb, sifa))) || + (non_asoc_addr_ok && + (sctp_is_addr_restricted(stcb, sifa)) && + (!sctp_is_addr_pending(stcb, sifa)))) { + /* on the no-no list */ + continue; + } + stcb->asoc.last_used_address = laddr; + atomic_add_int(&sifa->refcount, 1); + return (sifa); + } + if (start_at_beginning == 0) { + stcb->asoc.last_used_address = NULL; + goto sctp_from_the_top; + } + /* now try for any higher scope than the destination */ + stcb->asoc.last_used_address = starting_point; + start_at_beginning = 0; + sctp_from_the_top2: + if (stcb->asoc.last_used_address == NULL) { + start_at_beginning = 1; + stcb->asoc.last_used_address = LIST_FIRST(&inp->sctp_addr_list); + } + /* search beginning with the last used address */ + for (laddr = stcb->asoc.last_used_address; laddr; + laddr = LIST_NEXT(laddr, sctp_nxt_addr)) { + if (laddr->ifa == NULL) { + /* address has been removed */ + continue; + } + if (laddr->action == SCTP_DEL_IP_ADDRESS) { + /* address is being deleted */ + continue; + } + sifa = sctp_is_ifa_addr_acceptable(laddr->ifa, dest_is_loop, + dest_is_priv, fam); + if (sifa == NULL) + continue; + if (((non_asoc_addr_ok == 0) && + (sctp_is_addr_restricted(stcb, sifa))) || + (non_asoc_addr_ok && + (sctp_is_addr_restricted(stcb, sifa)) && + (!sctp_is_addr_pending(stcb, sifa)))) { + /* on the no-no list */ + continue; + } + stcb->asoc.last_used_address = laddr; + atomic_add_int(&sifa->refcount, 1); + return (sifa); + } + if (start_at_beginning == 0) { + stcb->asoc.last_used_address = NULL; + goto sctp_from_the_top2; + } + return (NULL); +} + +static struct sctp_ifa * +sctp_select_nth_preferred_addr_from_ifn_boundall(struct sctp_ifn *ifn, +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct sctp_inpcb *inp, +#else + struct sctp_inpcb *inp SCTP_UNUSED, +#endif + struct sctp_tcb *stcb, + int non_asoc_addr_ok, + uint8_t dest_is_loop, + uint8_t dest_is_priv, + int addr_wanted, + sa_family_t fam, + sctp_route_t *ro) +{ + struct sctp_ifa *ifa, *sifa; + int num_eligible_addr = 0; +#ifdef INET6 +#ifdef SCTP_EMBEDDED_V6_SCOPE + struct sockaddr_in6 sin6, lsa6; + + if (fam == AF_INET6) { + memcpy(&sin6, &ro->ro_dst, sizeof(struct sockaddr_in6)); +#ifdef SCTP_KAME + (void)sa6_recoverscope(&sin6); +#else + (void)in6_recoverscope(&sin6, &sin6.sin6_addr, NULL); +#endif /* SCTP_KAME */ + } +#endif /* SCTP_EMBEDDED_V6_SCOPE */ +#endif /* INET6 */ + LIST_FOREACH(ifa, &ifn->ifalist, next_ifa) { +#if defined(__FreeBSD__) && !defined(__Userspace__) +#ifdef INET + if ((ifa->address.sa.sa_family == AF_INET) && + (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &ifa->address.sin.sin_addr) != 0)) { + continue; + } +#endif +#ifdef INET6 + if ((ifa->address.sa.sa_family == AF_INET6) && + (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &ifa->address.sin6.sin6_addr) != 0)) { + continue; + } +#endif +#endif + if ((ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && + (non_asoc_addr_ok == 0)) + continue; + sifa = sctp_is_ifa_addr_preferred(ifa, dest_is_loop, + dest_is_priv, fam); + if (sifa == NULL) + continue; +#ifdef INET6 + if (fam == AF_INET6 && + dest_is_loop && + sifa->src_is_loop && sifa->src_is_priv) { + /* don't allow fe80::1 to be a src on loop ::1, we don't list it + * to the peer so we will get an abort. + */ + continue; + } +#ifdef SCTP_EMBEDDED_V6_SCOPE + if (fam == AF_INET6 && + IN6_IS_ADDR_LINKLOCAL(&sifa->address.sin6.sin6_addr) && + IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) { + /* link-local <-> link-local must belong to the same scope. */ + memcpy(&lsa6, &sifa->address.sin6, sizeof(struct sockaddr_in6)); +#ifdef SCTP_KAME + (void)sa6_recoverscope(&lsa6); +#else + (void)in6_recoverscope(&lsa6, &lsa6.sin6_addr, NULL); +#endif /* SCTP_KAME */ + if (sin6.sin6_scope_id != lsa6.sin6_scope_id) { + continue; + } + } +#endif /* SCTP_EMBEDDED_V6_SCOPE */ +#endif /* INET6 */ + +#if defined(__FreeBSD__) || defined(__APPLE__) || defined(__Userspace__) + /* Check if the IPv6 address matches to next-hop. + In the mobile case, old IPv6 address may be not deleted + from the interface. Then, the interface has previous and + new addresses. We should use one corresponding to the + next-hop. (by micchie) + */ +#ifdef INET6 + if (stcb && fam == AF_INET6 && + sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_BASE)) { + if (sctp_v6src_match_nexthop(&sifa->address.sin6, ro) == 0) { + continue; + } + } +#endif +#ifdef INET + /* Avoid topologically incorrect IPv4 address */ + if (stcb && fam == AF_INET && + sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_BASE)) { + if (sctp_v4src_match_nexthop(sifa, ro) == 0) { + continue; + } + } +#endif +#endif + if (stcb) { + if (sctp_is_address_in_scope(ifa, &stcb->asoc.scope, 0) == 0) { + continue; + } + if (((non_asoc_addr_ok == 0) && + (sctp_is_addr_restricted(stcb, sifa))) || + (non_asoc_addr_ok && + (sctp_is_addr_restricted(stcb, sifa)) && + (!sctp_is_addr_pending(stcb, sifa)))) { + /* + * It is restricted for some reason.. + * probably not yet added. + */ + continue; + } + } + if (num_eligible_addr >= addr_wanted) { + return (sifa); + } + num_eligible_addr++; + } + return (NULL); +} + +static int +sctp_count_num_preferred_boundall(struct sctp_ifn *ifn, +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct sctp_inpcb *inp, +#else + struct sctp_inpcb *inp SCTP_UNUSED, +#endif + struct sctp_tcb *stcb, + int non_asoc_addr_ok, + uint8_t dest_is_loop, + uint8_t dest_is_priv, + sa_family_t fam) +{ + struct sctp_ifa *ifa, *sifa; + int num_eligible_addr = 0; + + LIST_FOREACH(ifa, &ifn->ifalist, next_ifa) { +#if defined(__FreeBSD__) && !defined(__Userspace__) +#ifdef INET + if ((ifa->address.sa.sa_family == AF_INET) && + (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &ifa->address.sin.sin_addr) != 0)) { + continue; + } +#endif +#ifdef INET6 + if ((ifa->address.sa.sa_family == AF_INET6) && + (stcb != NULL) && + (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &ifa->address.sin6.sin6_addr) != 0)) { + continue; + } +#endif +#endif + if ((ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && + (non_asoc_addr_ok == 0)) { + continue; + } + sifa = sctp_is_ifa_addr_preferred(ifa, dest_is_loop, + dest_is_priv, fam); + if (sifa == NULL) { + continue; + } + if (stcb) { + if (sctp_is_address_in_scope(ifa, &stcb->asoc.scope, 0) == 0) { + continue; + } + if (((non_asoc_addr_ok == 0) && + (sctp_is_addr_restricted(stcb, sifa))) || + (non_asoc_addr_ok && + (sctp_is_addr_restricted(stcb, sifa)) && + (!sctp_is_addr_pending(stcb, sifa)))) { + /* + * It is restricted for some reason.. + * probably not yet added. + */ + continue; + } + } + num_eligible_addr++; + } + return (num_eligible_addr); +} + +static struct sctp_ifa * +sctp_choose_boundall(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_nets *net, + sctp_route_t *ro, + uint32_t vrf_id, + uint8_t dest_is_priv, + uint8_t dest_is_loop, + int non_asoc_addr_ok, + sa_family_t fam) +{ + int cur_addr_num = 0, num_preferred = 0; + void *ifn; + struct sctp_ifn *sctp_ifn, *looked_at = NULL, *emit_ifn; + struct sctp_ifa *sctp_ifa, *sifa; + uint32_t ifn_index; + struct sctp_vrf *vrf; +#ifdef INET + int retried = 0; +#endif + + /*- + * For boundall we can use any address in the association. + * If non_asoc_addr_ok is set we can use any address (at least in + * theory). So we look for preferred addresses first. If we find one, + * we use it. Otherwise we next try to get an address on the + * interface, which we should be able to do (unless non_asoc_addr_ok + * is false and we are routed out that way). In these cases where we + * can't use the address of the interface we go through all the + * ifn's looking for an address we can use and fill that in. Punting + * means we send back address 0, which will probably cause problems + * actually since then IP will fill in the address of the route ifn, + * which means we probably already rejected it.. i.e. here comes an + * abort :-<. + */ + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) + return (NULL); + + ifn = SCTP_GET_IFN_VOID_FROM_ROUTE(ro); + ifn_index = SCTP_GET_IF_INDEX_FROM_ROUTE(ro); + SCTPDBG(SCTP_DEBUG_OUTPUT2,"ifn from route:%p ifn_index:%d\n", ifn, ifn_index); + emit_ifn = looked_at = sctp_ifn = sctp_find_ifn(ifn, ifn_index); + if (sctp_ifn == NULL) { + /* ?? We don't have this guy ?? */ + SCTPDBG(SCTP_DEBUG_OUTPUT2,"No ifn emit interface?\n"); + goto bound_all_plan_b; + } + SCTPDBG(SCTP_DEBUG_OUTPUT2,"ifn_index:%d name:%s is emit interface\n", + ifn_index, sctp_ifn->ifn_name); + + if (net) { + cur_addr_num = net->indx_of_eligible_next_to_use; + } + num_preferred = sctp_count_num_preferred_boundall(sctp_ifn, + inp, stcb, + non_asoc_addr_ok, + dest_is_loop, + dest_is_priv, fam); + SCTPDBG(SCTP_DEBUG_OUTPUT2, "Found %d preferred source addresses for intf:%s\n", + num_preferred, sctp_ifn->ifn_name); + if (num_preferred == 0) { + /* + * no eligible addresses, we must use some other interface + * address if we can find one. + */ + goto bound_all_plan_b; + } + /* + * Ok we have num_eligible_addr set with how many we can use, this + * may vary from call to call due to addresses being deprecated + * etc.. + */ + if (cur_addr_num >= num_preferred) { + cur_addr_num = 0; + } + /* + * select the nth address from the list (where cur_addr_num is the + * nth) and 0 is the first one, 1 is the second one etc... + */ + SCTPDBG(SCTP_DEBUG_OUTPUT2, "cur_addr_num:%d\n", cur_addr_num); + + sctp_ifa = sctp_select_nth_preferred_addr_from_ifn_boundall(sctp_ifn, inp, stcb, non_asoc_addr_ok, dest_is_loop, + dest_is_priv, cur_addr_num, fam, ro); + + /* if sctp_ifa is NULL something changed??, fall to plan b. */ + if (sctp_ifa) { + atomic_add_int(&sctp_ifa->refcount, 1); + if (net) { + /* save off where the next one we will want */ + net->indx_of_eligible_next_to_use = cur_addr_num + 1; + } + return (sctp_ifa); + } + /* + * plan_b: Look at all interfaces and find a preferred address. If + * no preferred fall through to plan_c. + */ + bound_all_plan_b: + SCTPDBG(SCTP_DEBUG_OUTPUT2, "Trying Plan B\n"); + LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { + SCTPDBG(SCTP_DEBUG_OUTPUT2, "Examine interface %s\n", + sctp_ifn->ifn_name); + if (dest_is_loop == 0 && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { + /* wrong base scope */ + SCTPDBG(SCTP_DEBUG_OUTPUT2, "skip\n"); + continue; + } + if ((sctp_ifn == looked_at) && looked_at) { + /* already looked at this guy */ + SCTPDBG(SCTP_DEBUG_OUTPUT2, "already seen\n"); + continue; + } + num_preferred = sctp_count_num_preferred_boundall(sctp_ifn, inp, stcb, non_asoc_addr_ok, + dest_is_loop, dest_is_priv, fam); + SCTPDBG(SCTP_DEBUG_OUTPUT2, + "Found ifn:%p %d preferred source addresses\n", + ifn, num_preferred); + if (num_preferred == 0) { + /* None on this interface. */ + SCTPDBG(SCTP_DEBUG_OUTPUT2, "No preferred -- skipping to next\n"); + continue; + } + SCTPDBG(SCTP_DEBUG_OUTPUT2, + "num preferred:%d on interface:%p cur_addr_num:%d\n", + num_preferred, (void *)sctp_ifn, cur_addr_num); + + /* + * Ok we have num_eligible_addr set with how many we can + * use, this may vary from call to call due to addresses + * being deprecated etc.. + */ + if (cur_addr_num >= num_preferred) { + cur_addr_num = 0; + } + sifa = sctp_select_nth_preferred_addr_from_ifn_boundall(sctp_ifn, inp, stcb, non_asoc_addr_ok, dest_is_loop, + dest_is_priv, cur_addr_num, fam, ro); + if (sifa == NULL) + continue; + if (net) { + net->indx_of_eligible_next_to_use = cur_addr_num + 1; + SCTPDBG(SCTP_DEBUG_OUTPUT2, "we selected %d\n", + cur_addr_num); + SCTPDBG(SCTP_DEBUG_OUTPUT2, "Source:"); + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &sifa->address.sa); + SCTPDBG(SCTP_DEBUG_OUTPUT2, "Dest:"); + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &net->ro._l_addr.sa); + } + atomic_add_int(&sifa->refcount, 1); + return (sifa); + } +#ifdef INET +again_with_private_addresses_allowed: +#endif + /* plan_c: do we have an acceptable address on the emit interface */ + sifa = NULL; + SCTPDBG(SCTP_DEBUG_OUTPUT2,"Trying Plan C: find acceptable on interface\n"); + if (emit_ifn == NULL) { + SCTPDBG(SCTP_DEBUG_OUTPUT2,"Jump to Plan D - no emit_ifn\n"); + goto plan_d; + } + LIST_FOREACH(sctp_ifa, &emit_ifn->ifalist, next_ifa) { + SCTPDBG(SCTP_DEBUG_OUTPUT2, "ifa:%p\n", (void *)sctp_ifa); +#if defined(__FreeBSD__) && !defined(__Userspace__) +#ifdef INET + if ((sctp_ifa->address.sa.sa_family == AF_INET) && + (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sctp_ifa->address.sin.sin_addr) != 0)) { + SCTPDBG(SCTP_DEBUG_OUTPUT2,"Jailed\n"); + continue; + } +#endif +#ifdef INET6 + if ((sctp_ifa->address.sa.sa_family == AF_INET6) && + (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sctp_ifa->address.sin6.sin6_addr) != 0)) { + SCTPDBG(SCTP_DEBUG_OUTPUT2,"Jailed\n"); + continue; + } +#endif +#endif + if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && + (non_asoc_addr_ok == 0)) { + SCTPDBG(SCTP_DEBUG_OUTPUT2,"Defer\n"); + continue; + } + sifa = sctp_is_ifa_addr_acceptable(sctp_ifa, dest_is_loop, + dest_is_priv, fam); + if (sifa == NULL) { + SCTPDBG(SCTP_DEBUG_OUTPUT2, "IFA not acceptable\n"); + continue; + } + if (stcb) { + if (sctp_is_address_in_scope(sifa, &stcb->asoc.scope, 0) == 0) { + SCTPDBG(SCTP_DEBUG_OUTPUT2, "NOT in scope\n"); + sifa = NULL; + continue; + } + if (((non_asoc_addr_ok == 0) && + (sctp_is_addr_restricted(stcb, sifa))) || + (non_asoc_addr_ok && + (sctp_is_addr_restricted(stcb, sifa)) && + (!sctp_is_addr_pending(stcb, sifa)))) { + /* + * It is restricted for some + * reason.. probably not yet added. + */ + SCTPDBG(SCTP_DEBUG_OUTPUT2, "Its restricted\n"); + sifa = NULL; + continue; + } + } + atomic_add_int(&sifa->refcount, 1); + goto out; + } + plan_d: + /* + * plan_d: We are in trouble. No preferred address on the emit + * interface. And not even a preferred address on all interfaces. + * Go out and see if we can find an acceptable address somewhere + * amongst all interfaces. + */ + SCTPDBG(SCTP_DEBUG_OUTPUT2, "Trying Plan D looked_at is %p\n", (void *)looked_at); + LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { + if (dest_is_loop == 0 && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { + /* wrong base scope */ + continue; + } + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { +#if defined(__FreeBSD__) && !defined(__Userspace__) +#ifdef INET + if ((sctp_ifa->address.sa.sa_family == AF_INET) && + (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sctp_ifa->address.sin.sin_addr) != 0)) { + continue; + } +#endif +#ifdef INET6 + if ((sctp_ifa->address.sa.sa_family == AF_INET6) && + (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sctp_ifa->address.sin6.sin6_addr) != 0)) { + continue; + } +#endif +#endif + if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && + (non_asoc_addr_ok == 0)) + continue; + sifa = sctp_is_ifa_addr_acceptable(sctp_ifa, + dest_is_loop, + dest_is_priv, fam); + if (sifa == NULL) + continue; + if (stcb) { + if (sctp_is_address_in_scope(sifa, &stcb->asoc.scope, 0) == 0) { + sifa = NULL; + continue; + } + if (((non_asoc_addr_ok == 0) && + (sctp_is_addr_restricted(stcb, sifa))) || + (non_asoc_addr_ok && + (sctp_is_addr_restricted(stcb, sifa)) && + (!sctp_is_addr_pending(stcb, sifa)))) { + /* + * It is restricted for some + * reason.. probably not yet added. + */ + sifa = NULL; + continue; + } + } + goto out; + } + } +#ifdef INET + if (stcb) { + if ((retried == 0) && (stcb->asoc.scope.ipv4_local_scope == 0)) { + stcb->asoc.scope.ipv4_local_scope = 1; + retried = 1; + goto again_with_private_addresses_allowed; + } else if (retried == 1) { + stcb->asoc.scope.ipv4_local_scope = 0; + } + } +#endif +out: +#ifdef INET + if (sifa) { + if (retried == 1) { + LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { + if (dest_is_loop == 0 && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { + /* wrong base scope */ + continue; + } + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { + struct sctp_ifa *tmp_sifa; + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#ifdef INET + if ((sctp_ifa->address.sa.sa_family == AF_INET) && + (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sctp_ifa->address.sin.sin_addr) != 0)) { + continue; + } +#endif +#ifdef INET6 + if ((sctp_ifa->address.sa.sa_family == AF_INET6) && + (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sctp_ifa->address.sin6.sin6_addr) != 0)) { + continue; + } +#endif +#endif + if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && + (non_asoc_addr_ok == 0)) + continue; + tmp_sifa = sctp_is_ifa_addr_acceptable(sctp_ifa, + dest_is_loop, + dest_is_priv, fam); + if (tmp_sifa == NULL) { + continue; + } + if (tmp_sifa == sifa) { + continue; + } + if (stcb) { + if (sctp_is_address_in_scope(tmp_sifa, + &stcb->asoc.scope, 0) == 0) { + continue; + } + if (((non_asoc_addr_ok == 0) && + (sctp_is_addr_restricted(stcb, tmp_sifa))) || + (non_asoc_addr_ok && + (sctp_is_addr_restricted(stcb, tmp_sifa)) && + (!sctp_is_addr_pending(stcb, tmp_sifa)))) { + /* + * It is restricted for some + * reason.. probably not yet added. + */ + continue; + } + } + if ((tmp_sifa->address.sin.sin_family == AF_INET) && + (IN4_ISPRIVATE_ADDRESS(&(tmp_sifa->address.sin.sin_addr)))) { + sctp_add_local_addr_restricted(stcb, tmp_sifa); + } + } + } + } + atomic_add_int(&sifa->refcount, 1); + } +#endif + return (sifa); +} + +/* tcb may be NULL */ +struct sctp_ifa * +sctp_source_address_selection(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + sctp_route_t *ro, + struct sctp_nets *net, + int non_asoc_addr_ok, uint32_t vrf_id) +{ + struct sctp_ifa *answer; + uint8_t dest_is_priv, dest_is_loop; + sa_family_t fam; +#ifdef INET + struct sockaddr_in *to = (struct sockaddr_in *)&ro->ro_dst; +#endif +#ifdef INET6 + struct sockaddr_in6 *to6 = (struct sockaddr_in6 *)&ro->ro_dst; +#endif + + /** + * Rules: + * - Find the route if needed, cache if I can. + * - Look at interface address in route, Is it in the bound list. If so we + * have the best source. + * - If not we must rotate amongst the addresses. + * + * Caveats and issues + * + * Do we need to pay attention to scope. We can have a private address + * or a global address we are sourcing or sending to. So if we draw + * it out + * zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz + * For V4 + * ------------------------------------------ + * source * dest * result + * ----------------------------------------- + * <a> Private * Global * NAT + * ----------------------------------------- + * <b> Private * Private * No problem + * ----------------------------------------- + * <c> Global * Private * Huh, How will this work? + * ----------------------------------------- + * <d> Global * Global * No Problem + *------------------------------------------ + * zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz + * For V6 + *------------------------------------------ + * source * dest * result + * ----------------------------------------- + * <a> Linklocal * Global * + * ----------------------------------------- + * <b> Linklocal * Linklocal * No problem + * ----------------------------------------- + * <c> Global * Linklocal * Huh, How will this work? + * ----------------------------------------- + * <d> Global * Global * No Problem + *------------------------------------------ + * zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz + * + * And then we add to that what happens if there are multiple addresses + * assigned to an interface. Remember the ifa on a ifn is a linked + * list of addresses. So one interface can have more than one IP + * address. What happens if we have both a private and a global + * address? Do we then use context of destination to sort out which + * one is best? And what about NAT's sending P->G may get you a NAT + * translation, or should you select the G thats on the interface in + * preference. + * + * Decisions: + * + * - count the number of addresses on the interface. + * - if it is one, no problem except case <c>. + * For <a> we will assume a NAT out there. + * - if there are more than one, then we need to worry about scope P + * or G. We should prefer G -> G and P -> P if possible. + * Then as a secondary fall back to mixed types G->P being a last + * ditch one. + * - The above all works for bound all, but bound specific we need to + * use the same concept but instead only consider the bound + * addresses. If the bound set is NOT assigned to the interface then + * we must use rotation amongst the bound addresses.. + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (ro->ro_nh == NULL) { +#else + if (ro->ro_rt == NULL) { +#endif + /* + * Need a route to cache. + */ + SCTP_RTALLOC(ro, vrf_id, inp->fibnum); + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (ro->ro_nh == NULL) { +#else + if (ro->ro_rt == NULL) { +#endif + return (NULL); + } +#if defined(_WIN32) + /* On Windows the sa_family is U_SHORT or ADDRESS_FAMILY */ + fam = (sa_family_t)ro->ro_dst.sa_family; +#else + fam = ro->ro_dst.sa_family; +#endif + dest_is_priv = dest_is_loop = 0; + /* Setup our scopes for the destination */ + switch (fam) { +#ifdef INET + case AF_INET: + /* Scope based on outbound address */ + if (IN4_ISLOOPBACK_ADDRESS(&to->sin_addr)) { + dest_is_loop = 1; + if (net != NULL) { + /* mark it as local */ + net->addr_is_local = 1; + } + } else if ((IN4_ISPRIVATE_ADDRESS(&to->sin_addr))) { + dest_is_priv = 1; + } + break; +#endif +#ifdef INET6 + case AF_INET6: + /* Scope based on outbound address */ +#if defined(_WIN32) + if (IN6_IS_ADDR_LOOPBACK(&to6->sin6_addr)) { +#else + if (IN6_IS_ADDR_LOOPBACK(&to6->sin6_addr) || + SCTP_ROUTE_IS_REAL_LOOP(ro)) { +#endif + /* + * If the address is a loopback address, which + * consists of "::1" OR "fe80::1%lo0", we are loopback + * scope. But we don't use dest_is_priv (link local + * addresses). + */ + dest_is_loop = 1; + if (net != NULL) { + /* mark it as local */ + net->addr_is_local = 1; + } + } else if (IN6_IS_ADDR_LINKLOCAL(&to6->sin6_addr)) { + dest_is_priv = 1; + } + break; +#endif + } + SCTPDBG(SCTP_DEBUG_OUTPUT2, "Select source addr for:"); + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)&ro->ro_dst); + SCTP_IPI_ADDR_RLOCK(); + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + /* + * Bound all case + */ + answer = sctp_choose_boundall(inp, stcb, net, ro, vrf_id, + dest_is_priv, dest_is_loop, + non_asoc_addr_ok, fam); + SCTP_IPI_ADDR_RUNLOCK(); + return (answer); + } + /* + * Subset bound case + */ + if (stcb) { + answer = sctp_choose_boundspecific_stcb(inp, stcb, ro, + vrf_id, dest_is_priv, + dest_is_loop, + non_asoc_addr_ok, fam); + } else { + answer = sctp_choose_boundspecific_inp(inp, ro, vrf_id, + non_asoc_addr_ok, + dest_is_priv, + dest_is_loop, fam); + } + SCTP_IPI_ADDR_RUNLOCK(); + return (answer); +} + +static bool +sctp_find_cmsg(int c_type, void *data, struct mbuf *control, size_t cpsize) +{ +#if defined(_WIN32) + WSACMSGHDR cmh; +#else + struct cmsghdr cmh; +#endif + struct sctp_sndinfo sndinfo; + struct sctp_prinfo prinfo; + struct sctp_authinfo authinfo; + int tot_len, rem_len, cmsg_data_len, cmsg_data_off, off; + bool found; + + /* + * Independent of how many mbufs, find the c_type inside the control + * structure and copy out the data. + */ + found = false; + tot_len = SCTP_BUF_LEN(control); + for (off = 0; off < tot_len; off += CMSG_ALIGN(cmh.cmsg_len)) { + rem_len = tot_len - off; + if (rem_len < (int)CMSG_ALIGN(sizeof(cmh))) { + /* There is not enough room for one more. */ + return (found); + } + m_copydata(control, off, sizeof(cmh), (caddr_t)&cmh); + if (cmh.cmsg_len < CMSG_ALIGN(sizeof(cmh))) { + /* We dont't have a complete CMSG header. */ + return (found); + } + if ((cmh.cmsg_len > INT_MAX) || ((int)cmh.cmsg_len > rem_len)) { + /* We don't have the complete CMSG. */ + return (found); + } + cmsg_data_len = (int)cmh.cmsg_len - CMSG_ALIGN(sizeof(cmh)); + cmsg_data_off = off + CMSG_ALIGN(sizeof(cmh)); + if ((cmh.cmsg_level == IPPROTO_SCTP) && + ((c_type == cmh.cmsg_type) || + ((c_type == SCTP_SNDRCV) && + ((cmh.cmsg_type == SCTP_SNDINFO) || + (cmh.cmsg_type == SCTP_PRINFO) || + (cmh.cmsg_type == SCTP_AUTHINFO))))) { + if (c_type == cmh.cmsg_type) { + if (cpsize > INT_MAX) { + return (found); + } + if (cmsg_data_len < (int)cpsize) { + return (found); + } + /* It is exactly what we want. Copy it out. */ + m_copydata(control, cmsg_data_off, (int)cpsize, (caddr_t)data); + return (1); + } else { + struct sctp_sndrcvinfo *sndrcvinfo; + + sndrcvinfo = (struct sctp_sndrcvinfo *)data; + if (!found) { + if (cpsize < sizeof(struct sctp_sndrcvinfo)) { + return (found); + } + memset(sndrcvinfo, 0, sizeof(struct sctp_sndrcvinfo)); + } + switch (cmh.cmsg_type) { + case SCTP_SNDINFO: + if (cmsg_data_len < (int)sizeof(struct sctp_sndinfo)) { + return (found); + } + m_copydata(control, cmsg_data_off, sizeof(struct sctp_sndinfo), (caddr_t)&sndinfo); + sndrcvinfo->sinfo_stream = sndinfo.snd_sid; + sndrcvinfo->sinfo_flags = sndinfo.snd_flags; + sndrcvinfo->sinfo_ppid = sndinfo.snd_ppid; + sndrcvinfo->sinfo_context = sndinfo.snd_context; + sndrcvinfo->sinfo_assoc_id = sndinfo.snd_assoc_id; + break; + case SCTP_PRINFO: + if (cmsg_data_len < (int)sizeof(struct sctp_prinfo)) { + return (found); + } + m_copydata(control, cmsg_data_off, sizeof(struct sctp_prinfo), (caddr_t)&prinfo); + if (prinfo.pr_policy != SCTP_PR_SCTP_NONE) { + sndrcvinfo->sinfo_timetolive = prinfo.pr_value; + } else { + sndrcvinfo->sinfo_timetolive = 0; + } + sndrcvinfo->sinfo_flags |= prinfo.pr_policy; + break; + case SCTP_AUTHINFO: + if (cmsg_data_len < (int)sizeof(struct sctp_authinfo)) { + return (found); + } + m_copydata(control, cmsg_data_off, sizeof(struct sctp_authinfo), (caddr_t)&authinfo); + sndrcvinfo->sinfo_keynumber_valid = 1; + sndrcvinfo->sinfo_keynumber = authinfo.auth_keynumber; + break; + default: + return (found); + } + found = true; + } + } + } + return (found); +} + +static int +sctp_process_cmsgs_for_init(struct sctp_tcb *stcb, struct mbuf *control, int *error) +{ +#if defined(_WIN32) + WSACMSGHDR cmh; +#else + struct cmsghdr cmh; +#endif + struct sctp_initmsg initmsg; +#ifdef INET + struct sockaddr_in sin; +#endif +#ifdef INET6 + struct sockaddr_in6 sin6; +#endif + int tot_len, rem_len, cmsg_data_len, cmsg_data_off, off; + + tot_len = SCTP_BUF_LEN(control); + for (off = 0; off < tot_len; off += CMSG_ALIGN(cmh.cmsg_len)) { + rem_len = tot_len - off; + if (rem_len < (int)CMSG_ALIGN(sizeof(cmh))) { + /* There is not enough room for one more. */ + *error = EINVAL; + return (1); + } + m_copydata(control, off, sizeof(cmh), (caddr_t)&cmh); + if (cmh.cmsg_len < CMSG_ALIGN(sizeof(cmh))) { + /* We dont't have a complete CMSG header. */ + *error = EINVAL; + return (1); + } + if ((cmh.cmsg_len > INT_MAX) || ((int)cmh.cmsg_len > rem_len)) { + /* We don't have the complete CMSG. */ + *error = EINVAL; + return (1); + } + cmsg_data_len = (int)cmh.cmsg_len - CMSG_ALIGN(sizeof(cmh)); + cmsg_data_off = off + CMSG_ALIGN(sizeof(cmh)); + if (cmh.cmsg_level == IPPROTO_SCTP) { + switch (cmh.cmsg_type) { + case SCTP_INIT: + if (cmsg_data_len < (int)sizeof(struct sctp_initmsg)) { + *error = EINVAL; + return (1); + } + m_copydata(control, cmsg_data_off, sizeof(struct sctp_initmsg), (caddr_t)&initmsg); + if (initmsg.sinit_max_attempts) + stcb->asoc.max_init_times = initmsg.sinit_max_attempts; + if (initmsg.sinit_num_ostreams) + stcb->asoc.pre_open_streams = initmsg.sinit_num_ostreams; + if (initmsg.sinit_max_instreams) + stcb->asoc.max_inbound_streams = initmsg.sinit_max_instreams; + if (initmsg.sinit_max_init_timeo) + stcb->asoc.initial_init_rto_max = initmsg.sinit_max_init_timeo; + if (stcb->asoc.streamoutcnt < stcb->asoc.pre_open_streams) { + struct sctp_stream_out *tmp_str; + unsigned int i; +#if defined(SCTP_DETAILED_STR_STATS) + int j; +#endif + + /* Default is NOT correct */ + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Ok, default:%d pre_open:%d\n", + stcb->asoc.streamoutcnt, stcb->asoc.pre_open_streams); + SCTP_TCB_UNLOCK(stcb); + SCTP_MALLOC(tmp_str, + struct sctp_stream_out *, + (stcb->asoc.pre_open_streams * sizeof(struct sctp_stream_out)), + SCTP_M_STRMO); + SCTP_TCB_LOCK(stcb); + if (tmp_str != NULL) { + SCTP_FREE(stcb->asoc.strmout, SCTP_M_STRMO); + stcb->asoc.strmout = tmp_str; + stcb->asoc.strm_realoutsize = stcb->asoc.streamoutcnt = stcb->asoc.pre_open_streams; + } else { + stcb->asoc.pre_open_streams = stcb->asoc.streamoutcnt; + } + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + TAILQ_INIT(&stcb->asoc.strmout[i].outqueue); + stcb->asoc.ss_functions.sctp_ss_init_stream(stcb, &stcb->asoc.strmout[i], NULL); + stcb->asoc.strmout[i].chunks_on_queues = 0; +#if defined(SCTP_DETAILED_STR_STATS) + for (j = 0; j < SCTP_PR_SCTP_MAX + 1; j++) { + stcb->asoc.strmout[i].abandoned_sent[j] = 0; + stcb->asoc.strmout[i].abandoned_unsent[j] = 0; + } +#else + stcb->asoc.strmout[i].abandoned_sent[0] = 0; + stcb->asoc.strmout[i].abandoned_unsent[0] = 0; +#endif + stcb->asoc.strmout[i].next_mid_ordered = 0; + stcb->asoc.strmout[i].next_mid_unordered = 0; + stcb->asoc.strmout[i].sid = i; + stcb->asoc.strmout[i].last_msg_incomplete = 0; + stcb->asoc.strmout[i].state = SCTP_STREAM_OPENING; + } + } + break; +#ifdef INET + case SCTP_DSTADDRV4: + if (cmsg_data_len < (int)sizeof(struct in_addr)) { + *error = EINVAL; + return (1); + } + memset(&sin, 0, sizeof(struct sockaddr_in)); + sin.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + sin.sin_len = sizeof(struct sockaddr_in); +#endif + sin.sin_port = stcb->rport; + m_copydata(control, cmsg_data_off, sizeof(struct in_addr), (caddr_t)&sin.sin_addr); + if ((sin.sin_addr.s_addr == INADDR_ANY) || + (sin.sin_addr.s_addr == INADDR_BROADCAST) || + IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { + *error = EINVAL; + return (1); + } + if (sctp_add_remote_addr(stcb, (struct sockaddr *)&sin, NULL, stcb->asoc.port, + SCTP_DONOT_SETSCOPE, SCTP_ADDR_IS_CONFIRMED)) { + *error = ENOBUFS; + return (1); + } + break; +#endif +#ifdef INET6 + case SCTP_DSTADDRV6: + if (cmsg_data_len < (int)sizeof(struct in6_addr)) { + *error = EINVAL; + return (1); + } + memset(&sin6, 0, sizeof(struct sockaddr_in6)); + sin6.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6.sin6_len = sizeof(struct sockaddr_in6); +#endif + sin6.sin6_port = stcb->rport; + m_copydata(control, cmsg_data_off, sizeof(struct in6_addr), (caddr_t)&sin6.sin6_addr); + if (IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr) || + IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) { + *error = EINVAL; + return (1); + } +#ifdef INET + if (IN6_IS_ADDR_V4MAPPED(&sin6.sin6_addr)) { + in6_sin6_2_sin(&sin, &sin6); + if ((sin.sin_addr.s_addr == INADDR_ANY) || + (sin.sin_addr.s_addr == INADDR_BROADCAST) || + IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { + *error = EINVAL; + return (1); + } + if (sctp_add_remote_addr(stcb, (struct sockaddr *)&sin, NULL, stcb->asoc.port, + SCTP_DONOT_SETSCOPE, SCTP_ADDR_IS_CONFIRMED)) { + *error = ENOBUFS; + return (1); + } + } else +#endif + if (sctp_add_remote_addr(stcb, (struct sockaddr *)&sin6, NULL, stcb->asoc.port, + SCTP_DONOT_SETSCOPE, SCTP_ADDR_IS_CONFIRMED)) { + *error = ENOBUFS; + return (1); + } + break; +#endif + default: + break; + } + } + } + return (0); +} + +#if defined(INET) || defined(INET6) +static struct sctp_tcb * +sctp_findassociation_cmsgs(struct sctp_inpcb **inp_p, + uint16_t port, + struct mbuf *control, + struct sctp_nets **net_p, + int *error) +{ +#if defined(_WIN32) + WSACMSGHDR cmh; +#else + struct cmsghdr cmh; +#endif + struct sctp_tcb *stcb; + struct sockaddr *addr; +#ifdef INET + struct sockaddr_in sin; +#endif +#ifdef INET6 + struct sockaddr_in6 sin6; +#endif + int tot_len, rem_len, cmsg_data_len, cmsg_data_off, off; + + tot_len = SCTP_BUF_LEN(control); + for (off = 0; off < tot_len; off += CMSG_ALIGN(cmh.cmsg_len)) { + rem_len = tot_len - off; + if (rem_len < (int)CMSG_ALIGN(sizeof(cmh))) { + /* There is not enough room for one more. */ + *error = EINVAL; + return (NULL); + } + m_copydata(control, off, sizeof(cmh), (caddr_t)&cmh); + if (cmh.cmsg_len < CMSG_ALIGN(sizeof(cmh))) { + /* We dont't have a complete CMSG header. */ + *error = EINVAL; + return (NULL); + } + if ((cmh.cmsg_len > INT_MAX) || ((int)cmh.cmsg_len > rem_len)) { + /* We don't have the complete CMSG. */ + *error = EINVAL; + return (NULL); + } + cmsg_data_len = (int)cmh.cmsg_len - CMSG_ALIGN(sizeof(cmh)); + cmsg_data_off = off + CMSG_ALIGN(sizeof(cmh)); + if (cmh.cmsg_level == IPPROTO_SCTP) { + switch (cmh.cmsg_type) { +#ifdef INET + case SCTP_DSTADDRV4: + if (cmsg_data_len < (int)sizeof(struct in_addr)) { + *error = EINVAL; + return (NULL); + } + memset(&sin, 0, sizeof(struct sockaddr_in)); + sin.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + sin.sin_len = sizeof(struct sockaddr_in); +#endif + sin.sin_port = port; + m_copydata(control, cmsg_data_off, sizeof(struct in_addr), (caddr_t)&sin.sin_addr); + addr = (struct sockaddr *)&sin; + break; +#endif +#ifdef INET6 + case SCTP_DSTADDRV6: + if (cmsg_data_len < (int)sizeof(struct in6_addr)) { + *error = EINVAL; + return (NULL); + } + memset(&sin6, 0, sizeof(struct sockaddr_in6)); + sin6.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6.sin6_len = sizeof(struct sockaddr_in6); +#endif + sin6.sin6_port = port; + m_copydata(control, cmsg_data_off, sizeof(struct in6_addr), (caddr_t)&sin6.sin6_addr); +#ifdef INET + if (IN6_IS_ADDR_V4MAPPED(&sin6.sin6_addr)) { + in6_sin6_2_sin(&sin, &sin6); + addr = (struct sockaddr *)&sin; + } else +#endif + addr = (struct sockaddr *)&sin6; + break; +#endif + default: + addr = NULL; + break; + } + if (addr) { + stcb = sctp_findassociation_ep_addr(inp_p, addr, net_p, NULL, NULL); + if (stcb != NULL) { + return (stcb); + } + } + } + } + return (NULL); +} +#endif + +static struct mbuf * +sctp_add_cookie(struct mbuf *init, int init_offset, + struct mbuf *initack, int initack_offset, struct sctp_state_cookie *stc_in, uint8_t **signature) +{ + struct mbuf *copy_init, *copy_initack, *m_at, *sig, *mret; + struct sctp_state_cookie *stc; + struct sctp_paramhdr *ph; + uint16_t cookie_sz; + + mret = sctp_get_mbuf_for_msg((sizeof(struct sctp_state_cookie) + + sizeof(struct sctp_paramhdr)), 0, + M_NOWAIT, 1, MT_DATA); + if (mret == NULL) { + return (NULL); + } + copy_init = SCTP_M_COPYM(init, init_offset, M_COPYALL, M_NOWAIT); + if (copy_init == NULL) { + sctp_m_freem(mret); + return (NULL); + } +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(copy_init, SCTP_MBUF_ICOPY); + } +#endif + copy_initack = SCTP_M_COPYM(initack, initack_offset, M_COPYALL, + M_NOWAIT); + if (copy_initack == NULL) { + sctp_m_freem(mret); + sctp_m_freem(copy_init); + return (NULL); + } +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(copy_initack, SCTP_MBUF_ICOPY); + } +#endif + /* easy side we just drop it on the end */ + ph = mtod(mret, struct sctp_paramhdr *); + SCTP_BUF_LEN(mret) = sizeof(struct sctp_state_cookie) + + sizeof(struct sctp_paramhdr); + stc = (struct sctp_state_cookie *)((caddr_t)ph + + sizeof(struct sctp_paramhdr)); + ph->param_type = htons(SCTP_STATE_COOKIE); + ph->param_length = 0; /* fill in at the end */ + /* Fill in the stc cookie data */ + memcpy(stc, stc_in, sizeof(struct sctp_state_cookie)); + + /* tack the INIT and then the INIT-ACK onto the chain */ + cookie_sz = 0; + for (m_at = mret; m_at; m_at = SCTP_BUF_NEXT(m_at)) { + cookie_sz += SCTP_BUF_LEN(m_at); + if (SCTP_BUF_NEXT(m_at) == NULL) { + SCTP_BUF_NEXT(m_at) = copy_init; + break; + } + } + for (m_at = copy_init; m_at; m_at = SCTP_BUF_NEXT(m_at)) { + cookie_sz += SCTP_BUF_LEN(m_at); + if (SCTP_BUF_NEXT(m_at) == NULL) { + SCTP_BUF_NEXT(m_at) = copy_initack; + break; + } + } + for (m_at = copy_initack; m_at; m_at = SCTP_BUF_NEXT(m_at)) { + cookie_sz += SCTP_BUF_LEN(m_at); + if (SCTP_BUF_NEXT(m_at) == NULL) { + break; + } + } + sig = sctp_get_mbuf_for_msg(SCTP_SIGNATURE_SIZE, 0, M_NOWAIT, 1, MT_DATA); + if (sig == NULL) { + /* no space, so free the entire chain */ + sctp_m_freem(mret); + return (NULL); + } + SCTP_BUF_NEXT(m_at) = sig; + SCTP_BUF_LEN(sig) = SCTP_SIGNATURE_SIZE; + cookie_sz += SCTP_SIGNATURE_SIZE; + ph->param_length = htons(cookie_sz); + *signature = (uint8_t *)mtod(sig, caddr_t); + memset(*signature, 0, SCTP_SIGNATURE_SIZE); + return (mret); +} + +static uint8_t +sctp_get_ect(struct sctp_tcb *stcb) +{ + if ((stcb != NULL) && (stcb->asoc.ecn_supported == 1)) { + return (SCTP_ECT0_BIT); + } else { + return (0); + } +} + +#if defined(INET) || defined(INET6) +static void +sctp_handle_no_route(struct sctp_tcb *stcb, + struct sctp_nets *net, + int so_locked) +{ + SCTPDBG(SCTP_DEBUG_OUTPUT1, "dropped packet - no valid source addr\n"); + + if (net) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Destination was "); + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT1, &net->ro._l_addr.sa); + if (net->dest_state & SCTP_ADDR_CONFIRMED) { + if ((net->dest_state & SCTP_ADDR_REACHABLE) && stcb) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "no route takes interface %p down\n", (void *)net); + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, + stcb, 0, + (void *)net, + so_locked); + net->dest_state &= ~SCTP_ADDR_REACHABLE; + net->dest_state &= ~SCTP_ADDR_PF; + } + } + if (stcb) { + if (net == stcb->asoc.primary_destination) { + /* need a new primary */ + struct sctp_nets *alt; + + alt = sctp_find_alternate_net(stcb, net, 0); + if (alt != net) { + if (stcb->asoc.alternate) { + sctp_free_remote_addr(stcb->asoc.alternate); + } + stcb->asoc.alternate = alt; + atomic_add_int(&stcb->asoc.alternate->ref_count, 1); + if (net->ro._s_addr) { + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + } + net->src_addr_selected = 0; + } + } + } + } +} +#endif + +static int +sctp_lowlevel_chunk_output(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, /* may be NULL */ + struct sctp_nets *net, + struct sockaddr *to, + struct mbuf *m, + uint32_t auth_offset, + struct sctp_auth_chunk *auth, + uint16_t auth_keyid, + int nofragment_flag, + int ecn_ok, + int out_of_asoc_ok, + uint16_t src_port, + uint16_t dest_port, + uint32_t v_tag, + uint16_t port, + union sctp_sockstore *over_addr, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, +#endif + bool use_zero_crc, + int so_locked) +/* nofragment_flag to tell if IP_DF should be set (IPv4 only) */ +{ + /** + * Given a mbuf chain (via SCTP_BUF_NEXT()) that holds a packet header + * WITH an SCTPHDR but no IP header, endpoint inp and sa structure: + * - fill in the HMAC digest of any AUTH chunk in the packet. + * - calculate and fill in the SCTP checksum. + * - prepend an IP address header. + * - if boundall use INADDR_ANY. + * - if boundspecific do source address selection. + * - set fragmentation option for ipV4. + * - On return from IP output, check/adjust mtu size of output + * interface and smallest_mtu size as well. + */ + /* Will need ifdefs around this */ + struct mbuf *newm; + struct sctphdr *sctphdr; + int packet_length; + int ret; +#if defined(INET) || defined(INET6) + uint32_t vrf_id; +#endif +#if defined(INET) || defined(INET6) + struct mbuf *o_pak; + sctp_route_t *ro = NULL; + struct udphdr *udp = NULL; +#endif + uint8_t tos_value; +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so = NULL; +#endif + +#if defined(__APPLE__) && !defined(__Userspace__) + if (so_locked) { + sctp_lock_assert(SCTP_INP_SO(inp)); + SCTP_TCB_LOCK_ASSERT(stcb); + } else { + sctp_unlock_assert(SCTP_INP_SO(inp)); + } +#endif + if ((net) && (net->dest_state & SCTP_ADDR_OUT_OF_SCOPE)) { + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EFAULT); + sctp_m_freem(m); + return (EFAULT); + } +#if defined(INET) || defined(INET6) + if (stcb) { + vrf_id = stcb->asoc.vrf_id; + } else { + vrf_id = inp->def_vrf_id; + } +#endif + /* fill in the HMAC digest for any AUTH chunk in the packet */ + if ((auth != NULL) && (stcb != NULL)) { + sctp_fill_hmac_digest_m(m, auth_offset, auth, stcb, auth_keyid); + } + + if (net) { + tos_value = net->dscp; + } else if (stcb) { + tos_value = stcb->asoc.default_dscp; + } else { + tos_value = inp->sctp_ep.default_dscp; + } + + switch (to->sa_family) { +#ifdef INET + case AF_INET: + { + struct ip *ip = NULL; + sctp_route_t iproute; + int len; + + len = SCTP_MIN_V4_OVERHEAD; + if (port) { + len += sizeof(struct udphdr); + } + newm = sctp_get_mbuf_for_msg(len, 1, M_NOWAIT, 1, MT_DATA); + if (newm == NULL) { + sctp_m_freem(m); + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + SCTP_ALIGN_TO_END(newm, len); + SCTP_BUF_LEN(newm) = len; + SCTP_BUF_NEXT(newm) = m; + m = newm; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (net != NULL) { + m->m_pkthdr.flowid = net->flowid; + M_HASHTYPE_SET(m, net->flowtype); + } else { + m->m_pkthdr.flowid = mflowid; + M_HASHTYPE_SET(m, mflowtype); + } +#endif + packet_length = sctp_calculate_len(m); + ip = mtod(m, struct ip *); + ip->ip_v = IPVERSION; + ip->ip_hl = (sizeof(struct ip) >> 2); + if (tos_value == 0) { + /* + * This means especially, that it is not set at the + * SCTP layer. So use the value from the IP layer. + */ + tos_value = inp->ip_inp.inp.inp_ip_tos; + } + tos_value &= 0xfc; + if (ecn_ok) { + tos_value |= sctp_get_ect(stcb); + } + if ((nofragment_flag) && (port == 0)) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + ip->ip_off = htons(IP_DF); +#elif defined(WITH_CONVERT_IP_OFF) || defined(__APPLE__) + ip->ip_off = IP_DF; +#else + ip->ip_off = htons(IP_DF); +#endif + } else { +#if defined(__FreeBSD__) && !defined(__Userspace__) + ip->ip_off = htons(0); +#else + ip->ip_off = 0; +#endif + } +#if defined(__Userspace__) + ip->ip_id = htons(SCTP_IP_ID(inp)++); +#elif defined(__FreeBSD__) + /* FreeBSD has a function for ip_id's */ + ip_fillid(ip); +#elif defined(__APPLE__) +#if RANDOM_IP_ID + ip->ip_id = ip_randomid(); +#else + ip->ip_id = htons(ip_id++); +#endif +#else + ip->ip_id = SCTP_IP_ID(inp)++; +#endif + + ip->ip_ttl = inp->ip_inp.inp.inp_ip_ttl; +#if defined(__FreeBSD__) && !defined(__Userspace__) + ip->ip_len = htons(packet_length); +#else + ip->ip_len = packet_length; +#endif + ip->ip_tos = tos_value; + if (port) { + ip->ip_p = IPPROTO_UDP; + } else { + ip->ip_p = IPPROTO_SCTP; + } + ip->ip_sum = 0; + if (net == NULL) { + ro = &iproute; + memset(&iproute, 0, sizeof(iproute)); +#ifdef HAVE_SA_LEN + memcpy(&ro->ro_dst, to, to->sa_len); +#else + memcpy(&ro->ro_dst, to, sizeof(struct sockaddr_in)); +#endif + } else { + ro = (sctp_route_t *)&net->ro; + } + /* Now the address selection part */ + ip->ip_dst.s_addr = ((struct sockaddr_in *)to)->sin_addr.s_addr; + + /* call the routine to select the src address */ + if (net && out_of_asoc_ok == 0) { + if (net->ro._s_addr && (net->ro._s_addr->localifa_flags & (SCTP_BEING_DELETED|SCTP_ADDR_IFA_UNUSEABLE))) { + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + net->src_addr_selected = 0; +#if defined(__FreeBSD__) && !defined(__Userspace__) + RO_NHFREE(ro); +#else + if (ro->ro_rt) { + RTFREE(ro->ro_rt); + ro->ro_rt = NULL; + } +#endif + } + if (net->src_addr_selected == 0) { + /* Cache the source address */ + net->ro._s_addr = sctp_source_address_selection(inp,stcb, + ro, net, 0, + vrf_id); + net->src_addr_selected = 1; + } + if (net->ro._s_addr == NULL) { + /* No route to host */ + net->src_addr_selected = 0; + sctp_handle_no_route(stcb, net, so_locked); + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EHOSTUNREACH); + sctp_m_freem(m); + return (EHOSTUNREACH); + } + ip->ip_src = net->ro._s_addr->address.sin.sin_addr; + } else { + if (over_addr == NULL) { + struct sctp_ifa *_lsrc; + + _lsrc = sctp_source_address_selection(inp, stcb, ro, + net, + out_of_asoc_ok, + vrf_id); + if (_lsrc == NULL) { + sctp_handle_no_route(stcb, net, so_locked); + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EHOSTUNREACH); + sctp_m_freem(m); + return (EHOSTUNREACH); + } + ip->ip_src = _lsrc->address.sin.sin_addr; + sctp_free_ifa(_lsrc); + } else { + ip->ip_src = over_addr->sin.sin_addr; + SCTP_RTALLOC(ro, vrf_id, inp->fibnum); + } + } + if (port) { + if (htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)) == 0) { + sctp_handle_no_route(stcb, net, so_locked); + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EHOSTUNREACH); + sctp_m_freem(m); + return (EHOSTUNREACH); + } + udp = (struct udphdr *)((caddr_t)ip + sizeof(struct ip)); + udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); + udp->uh_dport = port; + udp->uh_ulen = htons((uint16_t)(packet_length - sizeof(struct ip))); +#if !defined(__Userspace__) +#if defined(__FreeBSD__) + if (V_udp_cksum) { + udp->uh_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, udp->uh_ulen + htons(IPPROTO_UDP)); + } else { + udp->uh_sum = 0; + } +#else + udp->uh_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, udp->uh_ulen + htons(IPPROTO_UDP)); +#endif +#else + udp->uh_sum = 0; +#endif + sctphdr = (struct sctphdr *)((caddr_t)udp + sizeof(struct udphdr)); + } else { + sctphdr = (struct sctphdr *)((caddr_t)ip + sizeof(struct ip)); + } + + sctphdr->src_port = src_port; + sctphdr->dest_port = dest_port; + sctphdr->v_tag = v_tag; + sctphdr->checksum = 0; + + /* + * If source address selection fails and we find no route + * then the ip_output should fail as well with a + * NO_ROUTE_TO_HOST type error. We probably should catch + * that somewhere and abort the association right away + * (assuming this is an INIT being sent). + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (ro->ro_nh == NULL) { +#else + if (ro->ro_rt == NULL) { +#endif + /* + * src addr selection failed to find a route (or + * valid source addr), so we can't get there from + * here (yet)! + */ + sctp_handle_no_route(stcb, net, so_locked); + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EHOSTUNREACH); + sctp_m_freem(m); + return (EHOSTUNREACH); + } + if (ro != &iproute) { + memcpy(&iproute, ro, sizeof(*ro)); + } + SCTPDBG(SCTP_DEBUG_OUTPUT3, "Calling ipv4 output routine from low level src addr:%x\n", + (uint32_t) (ntohl(ip->ip_src.s_addr))); + SCTPDBG(SCTP_DEBUG_OUTPUT3, "Destination is %x\n", + (uint32_t)(ntohl(ip->ip_dst.s_addr))); +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTPDBG(SCTP_DEBUG_OUTPUT3, "RTP route is %p through\n", + (void *)ro->ro_nh); +#else + SCTPDBG(SCTP_DEBUG_OUTPUT3, "RTP route is %p through\n", + (void *)ro->ro_rt); +#endif + + if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { + /* failed to prepend data, give up */ + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + sctp_m_freem(m); + return (ENOMEM); + } + SCTP_ATTACH_CHAIN(o_pak, m, packet_length); + if (port) { + if (use_zero_crc) { + SCTP_STAT_INCR(sctps_sendzerocrc); + } else { + sctphdr->checksum = sctp_calculate_cksum(m, sizeof(struct ip) + sizeof(struct udphdr)); + SCTP_STAT_INCR(sctps_sendswcrc); + } +#if !defined(__Userspace__) +#if defined(__FreeBSD__) + if (V_udp_cksum) { + SCTP_ENABLE_UDP_CSUM(o_pak); + } +#else + SCTP_ENABLE_UDP_CSUM(o_pak); +#endif +#endif + } else { + if (use_zero_crc) { + SCTP_STAT_INCR(sctps_sendzerocrc); + } else { +#if defined(__FreeBSD__) && !defined(__Userspace__) + m->m_pkthdr.csum_flags = CSUM_SCTP; + m->m_pkthdr.csum_data = offsetof(struct sctphdr, checksum); + SCTP_STAT_INCR(sctps_sendhwcrc); +#else + if (!(SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) && + (stcb) && (stcb->asoc.scope.loopback_scope))) { + sctphdr->checksum = sctp_calculate_cksum(m, sizeof(struct ip)); + SCTP_STAT_INCR(sctps_sendswcrc); + } else { + SCTP_STAT_INCR(sctps_sendhwcrc); + } +#endif + } + } +#ifdef SCTP_PACKET_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) + sctp_packet_log(o_pak); +#endif + /* send it out. table id is taken from stcb */ +#if defined(__APPLE__) && !defined(__Userspace__) + if ((SCTP_BASE_SYSCTL(sctp_output_unlocked)) && (so_locked)) { + so = SCTP_INP_SO(inp); + SCTP_SOCKET_UNLOCK(so, 0); + } +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_PROBE5(send, NULL, stcb, ip, stcb, sctphdr); +#endif + SCTP_IP_OUTPUT(ret, o_pak, ro, inp, vrf_id); +#if defined(__APPLE__) && !defined(__Userspace__) + if ((SCTP_BASE_SYSCTL(sctp_output_unlocked)) && (so_locked)) { + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 0); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (port) { + UDPSTAT_INC(udps_opackets); + } +#endif + SCTP_STAT_INCR(sctps_sendpackets); + SCTP_STAT_INCR_COUNTER64(sctps_outpackets); + if (ret) + SCTP_STAT_INCR(sctps_senderrors); + + SCTPDBG(SCTP_DEBUG_OUTPUT3, "IP output returns %d\n", ret); + if (net == NULL) { + /* free tempy routes */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + RO_NHFREE(ro); +#else + if (ro->ro_rt) { + RTFREE(ro->ro_rt); + ro->ro_rt = NULL; + } +#endif + } else { +#if defined(__FreeBSD__) && !defined(__Userspace__) + if ((ro->ro_nh != NULL) && (net->ro._s_addr) && +#else + if ((ro->ro_rt != NULL) && (net->ro._s_addr) && +#endif + ((net->dest_state & SCTP_ADDR_NO_PMTUD) == 0)) { + uint32_t mtu; + +#if defined(__FreeBSD__) && !defined(__Userspace__) + mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, ro->ro_nh); +#else + mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, ro->ro_rt); +#endif + if (mtu > 0) { + if (net->port) { + mtu -= sizeof(struct udphdr); + } + if (mtu < net->mtu) { + net->mtu = mtu; + if ((stcb != NULL) && (stcb->asoc.smallest_mtu > mtu)) { + sctp_pathmtu_adjustment(stcb, mtu, true); + } + } + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + } else if (ro->ro_nh == NULL) { +#else + } else if (ro->ro_rt == NULL) { +#endif + /* route was freed */ + if (net->ro._s_addr && + net->src_addr_selected) { + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + } + net->src_addr_selected = 0; + } + } + return (ret); + } +#endif +#ifdef INET6 + case AF_INET6: + { + uint32_t flowlabel, flowinfo; + struct ip6_hdr *ip6h; + struct route_in6 ip6route; +#if !defined(__Userspace__) + struct ifnet *ifp; +#endif + struct sockaddr_in6 *sin6, tmp, *lsa6, lsa6_tmp; + int prev_scope = 0; +#ifdef SCTP_EMBEDDED_V6_SCOPE + struct sockaddr_in6 lsa6_storage; + int error; +#endif + u_short prev_port = 0; + int len; + + if (net) { + flowlabel = net->flowlabel; + } else if (stcb) { + flowlabel = stcb->asoc.default_flowlabel; + } else { + flowlabel = inp->sctp_ep.default_flowlabel; + } + if (flowlabel == 0) { + /* + * This means especially, that it is not set at the + * SCTP layer. So use the value from the IP layer. + */ +#if defined(__APPLE__) && !defined(__Userspace__) && (!defined(APPLE_LEOPARD) && !defined(APPLE_SNOWLEOPARD) && !defined(APPLE_LION) && !defined(APPLE_MOUNTAINLION)) + flowlabel = ntohl(inp->ip_inp.inp.inp_flow); +#else + flowlabel = ntohl(((struct inpcb *)inp)->inp_flow); +#endif + } + flowlabel &= 0x000fffff; + len = SCTP_MIN_OVERHEAD; + if (port) { + len += sizeof(struct udphdr); + } + newm = sctp_get_mbuf_for_msg(len, 1, M_NOWAIT, 1, MT_DATA); + if (newm == NULL) { + sctp_m_freem(m); + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + SCTP_ALIGN_TO_END(newm, len); + SCTP_BUF_LEN(newm) = len; + SCTP_BUF_NEXT(newm) = m; + m = newm; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (net != NULL) { + m->m_pkthdr.flowid = net->flowid; + M_HASHTYPE_SET(m, net->flowtype); + } else { + m->m_pkthdr.flowid = mflowid; + M_HASHTYPE_SET(m, mflowtype); + } +#endif + packet_length = sctp_calculate_len(m); + + ip6h = mtod(m, struct ip6_hdr *); + /* protect *sin6 from overwrite */ + sin6 = (struct sockaddr_in6 *)to; + tmp = *sin6; + sin6 = &tmp; + +#ifdef SCTP_EMBEDDED_V6_SCOPE + /* KAME hack: embed scopeid */ +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) + if (in6_embedscope(&sin6->sin6_addr, sin6, NULL, NULL) != 0) +#else + if (in6_embedscope(&sin6->sin6_addr, sin6, NULL, NULL, NULL) != 0) +#endif +#elif defined(SCTP_KAME) + if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) +#else + if (in6_embedscope(&sin6->sin6_addr, sin6) != 0) +#endif + { + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); + sctp_m_freem(m); + return (EINVAL); + } +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + if (net == NULL) { + memset(&ip6route, 0, sizeof(ip6route)); + ro = (sctp_route_t *)&ip6route; +#ifdef HAVE_SIN6_LEN + memcpy(&ro->ro_dst, sin6, sin6->sin6_len); +#else + memcpy(&ro->ro_dst, sin6, sizeof(struct sockaddr_in6)); +#endif + } else { + ro = (sctp_route_t *)&net->ro; + } + /* + * We assume here that inp_flow is in host byte order within + * the TCB! + */ + if (tos_value == 0) { + /* + * This means especially, that it is not set at the + * SCTP layer. So use the value from the IP layer. + */ +#if defined(__APPLE__) && !defined(__Userspace__) && (!defined(APPLE_LEOPARD) && !defined(APPLE_SNOWLEOPARD) && !defined(APPLE_LION) && !defined(APPLE_MOUNTAINLION)) + tos_value = (ntohl(inp->ip_inp.inp.inp_flow) >> 20) & 0xff; +#else + tos_value = (ntohl(((struct inpcb *)inp)->inp_flow) >> 20) & 0xff; +#endif + } + tos_value &= 0xfc; + if (ecn_ok) { + tos_value |= sctp_get_ect(stcb); + } + flowinfo = 0x06; + flowinfo <<= 8; + flowinfo |= tos_value; + flowinfo <<= 20; + flowinfo |= flowlabel; + ip6h->ip6_flow = htonl(flowinfo); + if (port) { + ip6h->ip6_nxt = IPPROTO_UDP; + } else { + ip6h->ip6_nxt = IPPROTO_SCTP; + } + ip6h->ip6_plen = htons((uint16_t)(packet_length - sizeof(struct ip6_hdr))); + ip6h->ip6_dst = sin6->sin6_addr; + + /* + * Add SRC address selection here: we can only reuse to a + * limited degree the kame src-addr-sel, since we can try + * their selection but it may not be bound. + */ + memset(&lsa6_tmp, 0, sizeof(lsa6_tmp)); + lsa6_tmp.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + lsa6_tmp.sin6_len = sizeof(lsa6_tmp); +#endif + lsa6 = &lsa6_tmp; + if (net && out_of_asoc_ok == 0) { + if (net->ro._s_addr && (net->ro._s_addr->localifa_flags & (SCTP_BEING_DELETED|SCTP_ADDR_IFA_UNUSEABLE))) { + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + net->src_addr_selected = 0; +#if defined(__FreeBSD__) && !defined(__Userspace__) + RO_NHFREE(ro); +#else + if (ro->ro_rt) { + RTFREE(ro->ro_rt); + ro->ro_rt = NULL; + } +#endif + } + if (net->src_addr_selected == 0) { +#ifdef SCTP_EMBEDDED_V6_SCOPE + sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; + /* KAME hack: embed scopeid */ +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) + if (in6_embedscope(&sin6->sin6_addr, sin6, NULL, NULL) != 0) +#else + if (in6_embedscope(&sin6->sin6_addr, sin6, NULL, NULL, NULL) != 0) +#endif +#elif defined(SCTP_KAME) + if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) +#else + if (in6_embedscope(&sin6->sin6_addr, sin6) != 0) +#endif + { + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); + sctp_m_freem(m); + return (EINVAL); + } +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + /* Cache the source address */ + net->ro._s_addr = sctp_source_address_selection(inp, + stcb, + ro, + net, + 0, + vrf_id); +#ifdef SCTP_EMBEDDED_V6_SCOPE +#ifdef SCTP_KAME + (void)sa6_recoverscope(sin6); +#else + (void)in6_recoverscope(sin6, &sin6->sin6_addr, NULL); +#endif /* SCTP_KAME */ +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + net->src_addr_selected = 1; + } + if (net->ro._s_addr == NULL) { + SCTPDBG(SCTP_DEBUG_OUTPUT3, "V6:No route to host\n"); + net->src_addr_selected = 0; + sctp_handle_no_route(stcb, net, so_locked); + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EHOSTUNREACH); + sctp_m_freem(m); + return (EHOSTUNREACH); + } + lsa6->sin6_addr = net->ro._s_addr->address.sin6.sin6_addr; + } else { +#ifdef SCTP_EMBEDDED_V6_SCOPE + sin6 = (struct sockaddr_in6 *)&ro->ro_dst; + /* KAME hack: embed scopeid */ +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) + if (in6_embedscope(&sin6->sin6_addr, sin6, NULL, NULL) != 0) +#else + if (in6_embedscope(&sin6->sin6_addr, sin6, NULL, NULL, NULL) != 0) +#endif +#elif defined(SCTP_KAME) + if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) +#else + if (in6_embedscope(&sin6->sin6_addr, sin6) != 0) +#endif + { + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); + sctp_m_freem(m); + return (EINVAL); + } +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + if (over_addr == NULL) { + struct sctp_ifa *_lsrc; + + _lsrc = sctp_source_address_selection(inp, stcb, ro, + net, + out_of_asoc_ok, + vrf_id); + if (_lsrc == NULL) { + sctp_handle_no_route(stcb, net, so_locked); + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EHOSTUNREACH); + sctp_m_freem(m); + return (EHOSTUNREACH); + } + lsa6->sin6_addr = _lsrc->address.sin6.sin6_addr; + sctp_free_ifa(_lsrc); + } else { + lsa6->sin6_addr = over_addr->sin6.sin6_addr; + SCTP_RTALLOC(ro, vrf_id, inp->fibnum); + } +#ifdef SCTP_EMBEDDED_V6_SCOPE +#ifdef SCTP_KAME + (void)sa6_recoverscope(sin6); +#else + (void)in6_recoverscope(sin6, &sin6->sin6_addr, NULL); +#endif /* SCTP_KAME */ +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + } + lsa6->sin6_port = inp->sctp_lport; + +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (ro->ro_nh == NULL) { +#else + if (ro->ro_rt == NULL) { +#endif + /* + * src addr selection failed to find a route (or + * valid source addr), so we can't get there from + * here! + */ + sctp_handle_no_route(stcb, net, so_locked); + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EHOSTUNREACH); + sctp_m_freem(m); + return (EHOSTUNREACH); + } +#ifndef SCOPEDROUTING +#ifdef SCTP_EMBEDDED_V6_SCOPE + /* + * XXX: sa6 may not have a valid sin6_scope_id in the + * non-SCOPEDROUTING case. + */ + memset(&lsa6_storage, 0, sizeof(lsa6_storage)); + lsa6_storage.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + lsa6_storage.sin6_len = sizeof(lsa6_storage); +#endif +#ifdef SCTP_KAME + lsa6_storage.sin6_addr = lsa6->sin6_addr; + if ((error = sa6_recoverscope(&lsa6_storage)) != 0) { +#else + if ((error = in6_recoverscope(&lsa6_storage, &lsa6->sin6_addr, + NULL)) != 0) { +#endif /* SCTP_KAME */ + SCTPDBG(SCTP_DEBUG_OUTPUT3, "recover scope fails error %d\n", error); + sctp_m_freem(m); + return (error); + } + /* XXX */ + lsa6_storage.sin6_addr = lsa6->sin6_addr; + lsa6_storage.sin6_port = inp->sctp_lport; + lsa6 = &lsa6_storage; +#endif /* SCTP_EMBEDDED_V6_SCOPE */ +#endif /* SCOPEDROUTING */ + ip6h->ip6_src = lsa6->sin6_addr; + + if (port) { + if (htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)) == 0) { + sctp_handle_no_route(stcb, net, so_locked); + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EHOSTUNREACH); + sctp_m_freem(m); + return (EHOSTUNREACH); + } + udp = (struct udphdr *)((caddr_t)ip6h + sizeof(struct ip6_hdr)); + udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); + udp->uh_dport = port; + udp->uh_ulen = htons((uint16_t)(packet_length - sizeof(struct ip6_hdr))); + udp->uh_sum = 0; + sctphdr = (struct sctphdr *)((caddr_t)udp + sizeof(struct udphdr)); + } else { + sctphdr = (struct sctphdr *)((caddr_t)ip6h + sizeof(struct ip6_hdr)); + } + + sctphdr->src_port = src_port; + sctphdr->dest_port = dest_port; + sctphdr->v_tag = v_tag; + sctphdr->checksum = 0; + + /* + * We set the hop limit now since there is a good chance + * that our ro pointer is now filled + */ + ip6h->ip6_hlim = SCTP_GET_HLIM(inp, ro); +#if !defined(__Userspace__) + ifp = SCTP_GET_IFN_VOID_FROM_ROUTE(ro); +#endif + +#ifdef SCTP_DEBUG + /* Copy to be sure something bad is not happening */ + sin6->sin6_addr = ip6h->ip6_dst; + lsa6->sin6_addr = ip6h->ip6_src; +#endif + + SCTPDBG(SCTP_DEBUG_OUTPUT3, "Calling ipv6 output routine from low level\n"); + SCTPDBG(SCTP_DEBUG_OUTPUT3, "src: "); + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT3, (struct sockaddr *)lsa6); + SCTPDBG(SCTP_DEBUG_OUTPUT3, "dst: "); + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT3, (struct sockaddr *)sin6); + if (net) { + sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; + /* preserve the port and scope for link local send */ + prev_scope = sin6->sin6_scope_id; + prev_port = sin6->sin6_port; + } + + if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { + /* failed to prepend data, give up */ + sctp_m_freem(m); + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + SCTP_ATTACH_CHAIN(o_pak, m, packet_length); + if (port) { + sctphdr->checksum = sctp_calculate_cksum(m, sizeof(struct ip6_hdr) + sizeof(struct udphdr)); + SCTP_STAT_INCR(sctps_sendswcrc); +#if !defined(__Userspace__) +#if defined(_WIN32) + udp->uh_sum = 0; +#else + if ((udp->uh_sum = in6_cksum(o_pak, IPPROTO_UDP, sizeof(struct ip6_hdr), packet_length - sizeof(struct ip6_hdr))) == 0) { + udp->uh_sum = 0xffff; + } +#endif +#endif + } else { +#if defined(__FreeBSD__) && !defined(__Userspace__) + m->m_pkthdr.csum_flags = CSUM_SCTP_IPV6; + m->m_pkthdr.csum_data = offsetof(struct sctphdr, checksum); + SCTP_STAT_INCR(sctps_sendhwcrc); +#else + if (!(SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) && + (stcb) && (stcb->asoc.scope.loopback_scope))) { + sctphdr->checksum = sctp_calculate_cksum(m, sizeof(struct ip6_hdr)); + SCTP_STAT_INCR(sctps_sendswcrc); + } else { + SCTP_STAT_INCR(sctps_sendhwcrc); + } +#endif + } + /* send it out. table id is taken from stcb */ +#if defined(__APPLE__) && !defined(__Userspace__) + if ((SCTP_BASE_SYSCTL(sctp_output_unlocked)) && (so_locked)) { + so = SCTP_INP_SO(inp); + SCTP_SOCKET_UNLOCK(so, 0); + } +#endif +#ifdef SCTP_PACKET_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) + sctp_packet_log(o_pak); +#endif +#if !defined(__Userspace__) +#if defined(__FreeBSD__) + SCTP_PROBE5(send, NULL, stcb, ip6h, stcb, sctphdr); +#endif + SCTP_IP6_OUTPUT(ret, o_pak, (struct route_in6 *)ro, &ifp, inp, vrf_id); +#else + SCTP_IP6_OUTPUT(ret, o_pak, (struct route_in6 *)ro, NULL, inp, vrf_id); +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + if ((SCTP_BASE_SYSCTL(sctp_output_unlocked)) && (so_locked)) { + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 0); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } +#endif + if (net) { + /* for link local this must be done */ + sin6->sin6_scope_id = prev_scope; + sin6->sin6_port = prev_port; + } + SCTPDBG(SCTP_DEBUG_OUTPUT3, "return from send is %d\n", ret); +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (port) { + UDPSTAT_INC(udps_opackets); + } +#endif + SCTP_STAT_INCR(sctps_sendpackets); + SCTP_STAT_INCR_COUNTER64(sctps_outpackets); + if (ret) { + SCTP_STAT_INCR(sctps_senderrors); + } + if (net == NULL) { + /* Now if we had a temp route free it */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + RO_NHFREE(ro); +#else + if (ro->ro_rt) { + RTFREE(ro->ro_rt); + ro->ro_rt = NULL; + } +#endif + } else { + /* PMTU check versus smallest asoc MTU goes here */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (ro->ro_nh == NULL) { +#else + if (ro->ro_rt == NULL) { +#endif + /* Route was freed */ + if (net->ro._s_addr && + net->src_addr_selected) { + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + } + net->src_addr_selected = 0; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if ((ro->ro_nh != NULL) && (net->ro._s_addr) && +#else + if ((ro->ro_rt != NULL) && (net->ro._s_addr) && +#endif + ((net->dest_state & SCTP_ADDR_NO_PMTUD) == 0)) { + uint32_t mtu; + +#if defined(__FreeBSD__) && !defined(__Userspace__) + mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, ro->ro_nh); +#else + mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, ro->ro_rt); +#endif + if (mtu > 0) { + if (net->port) { + mtu -= sizeof(struct udphdr); + } + if (mtu < net->mtu) { + net->mtu = mtu; + if ((stcb != NULL) && (stcb->asoc.smallest_mtu > mtu)) { + sctp_pathmtu_adjustment(stcb, mtu, false); + } + } + } + } +#if !defined(__Userspace__) + else if (ifp != NULL) { +#if defined(_WIN32) +#define ND_IFINFO(ifp) (ifp) +#define linkmtu if_mtu +#endif + if ((ND_IFINFO(ifp)->linkmtu > 0) && + (stcb->asoc.smallest_mtu > ND_IFINFO(ifp)->linkmtu)) { + sctp_pathmtu_adjustment(stcb, ND_IFINFO(ifp)->linkmtu, false); + } + } +#endif + } + return (ret); + } +#endif +#if defined(__Userspace__) + case AF_CONN: + { + char *buffer; + struct sockaddr_conn *sconn; + int len; + + sconn = (struct sockaddr_conn *)to; + len = sizeof(struct sctphdr); + newm = sctp_get_mbuf_for_msg(len, 1, M_NOWAIT, 1, MT_DATA); + if (newm == NULL) { + sctp_m_freem(m); + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + SCTP_ALIGN_TO_END(newm, len); + SCTP_BUF_LEN(newm) = len; + SCTP_BUF_NEXT(newm) = m; + m = newm; + packet_length = sctp_calculate_len(m); + m->m_pkthdr.len = packet_length; + sctphdr = mtod(m, struct sctphdr *); + sctphdr->src_port = src_port; + sctphdr->dest_port = dest_port; + sctphdr->v_tag = v_tag; + sctphdr->checksum = 0; + if (SCTP_BASE_VAR(crc32c_offloaded) == 0) { + sctphdr->checksum = sctp_calculate_cksum(m, 0); + SCTP_STAT_INCR(sctps_sendswcrc); + } else { + SCTP_STAT_INCR(sctps_sendhwcrc); + } + if (tos_value == 0) { + tos_value = inp->ip_inp.inp.inp_ip_tos; + } + tos_value &= 0xfc; + if (ecn_ok) { + tos_value |= sctp_get_ect(stcb); + } + /* Don't alloc/free for each packet */ + if ((buffer = malloc(packet_length)) != NULL) { + m_copydata(m, 0, packet_length, buffer); + ret = SCTP_BASE_VAR(conn_output)(sconn->sconn_addr, buffer, packet_length, tos_value, nofragment_flag); + free(buffer); + } else { + ret = ENOMEM; + } + sctp_m_freem(m); + return (ret); + } +#endif + default: + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Unknown protocol (TSNH) type %d\n", + ((struct sockaddr *)to)->sa_family); + sctp_m_freem(m); + SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EFAULT); + return (EFAULT); + } +} + +void +sctp_send_initiate(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int so_locked) +{ + struct mbuf *m, *m_last; + struct sctp_nets *net; + struct sctp_init_chunk *init; + struct sctp_supported_addr_param *sup_addr; + struct sctp_adaptation_layer_indication *ali; + struct sctp_supported_chunk_types_param *pr_supported; + struct sctp_paramhdr *ph; + int cnt_inits_to = 0; + int error; + uint16_t num_ext, chunk_len, padding_len, parameter_len; + +#if defined(__APPLE__) && !defined(__Userspace__) + if (so_locked) { + sctp_lock_assert(SCTP_INP_SO(inp)); + } else { + sctp_unlock_assert(SCTP_INP_SO(inp)); + } +#endif + /* INIT's always go to the primary (and usually ONLY address) */ + net = stcb->asoc.primary_destination; + if (net == NULL) { + net = TAILQ_FIRST(&stcb->asoc.nets); + if (net == NULL) { + /* TSNH */ + return; + } + /* we confirm any address we send an INIT to */ + net->dest_state &= ~SCTP_ADDR_UNCONFIRMED; + (void)sctp_set_primary_addr(stcb, NULL, net); + } else { + /* we confirm any address we send an INIT to */ + net->dest_state &= ~SCTP_ADDR_UNCONFIRMED; + } + SCTPDBG(SCTP_DEBUG_OUTPUT4, "Sending INIT\n"); +#ifdef INET6 + if (net->ro._l_addr.sa.sa_family == AF_INET6) { + /* + * special hook, if we are sending to link local it will not + * show up in our private address count. + */ + if (IN6_IS_ADDR_LINKLOCAL(&net->ro._l_addr.sin6.sin6_addr)) + cnt_inits_to = 1; + } +#endif + if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { + /* This case should not happen */ + SCTPDBG(SCTP_DEBUG_OUTPUT4, "Sending INIT - failed timer?\n"); + return; + } + /* start the INIT timer */ + sctp_timer_start(SCTP_TIMER_TYPE_INIT, inp, stcb, net); + + m = sctp_get_mbuf_for_msg(MCLBYTES, 1, M_NOWAIT, 1, MT_DATA); + if (m == NULL) { + /* No memory, INIT timer will re-attempt. */ + SCTPDBG(SCTP_DEBUG_OUTPUT4, "Sending INIT - mbuf?\n"); + return; + } + chunk_len = (uint16_t)sizeof(struct sctp_init_chunk); + padding_len = 0; + /* Now lets put the chunk header in place */ + init = mtod(m, struct sctp_init_chunk *); + /* now the chunk header */ + init->ch.chunk_type = SCTP_INITIATION; + init->ch.chunk_flags = 0; + /* fill in later from mbuf we build */ + init->ch.chunk_length = 0; + /* place in my tag */ + init->init.initiate_tag = htonl(stcb->asoc.my_vtag); + /* set up some of the credits. */ + init->init.a_rwnd = htonl(max(inp->sctp_socket?SCTP_SB_LIMIT_RCV(inp->sctp_socket):0, + SCTP_MINIMAL_RWND)); + init->init.num_outbound_streams = htons(stcb->asoc.pre_open_streams); + init->init.num_inbound_streams = htons(stcb->asoc.max_inbound_streams); + init->init.initial_tsn = htonl(stcb->asoc.init_seq_number); + + /* Adaptation layer indication parameter */ + if (inp->sctp_ep.adaptation_layer_indicator_provided) { + parameter_len = (uint16_t)sizeof(struct sctp_adaptation_layer_indication); + ali = (struct sctp_adaptation_layer_indication *)(mtod(m, caddr_t) + chunk_len); + ali->ph.param_type = htons(SCTP_ULP_ADAPTATION); + ali->ph.param_length = htons(parameter_len); + ali->indication = htonl(inp->sctp_ep.adaptation_layer_indicator); + chunk_len += parameter_len; + } + + /* ECN parameter */ + if (stcb->asoc.ecn_supported == 1) { + parameter_len = (uint16_t)sizeof(struct sctp_paramhdr); + ph = (struct sctp_paramhdr *)(mtod(m, caddr_t) + chunk_len); + ph->param_type = htons(SCTP_ECN_CAPABLE); + ph->param_length = htons(parameter_len); + chunk_len += parameter_len; + } + + /* PR-SCTP supported parameter */ + if (stcb->asoc.prsctp_supported == 1) { + parameter_len = (uint16_t)sizeof(struct sctp_paramhdr); + ph = (struct sctp_paramhdr *)(mtod(m, caddr_t) + chunk_len); + ph->param_type = htons(SCTP_PRSCTP_SUPPORTED); + ph->param_length = htons(parameter_len); + chunk_len += parameter_len; + } + + /* Zero checksum acceptable parameter */ + if (stcb->asoc.zero_checksum > 0) { + parameter_len = (uint16_t)sizeof(struct sctp_paramhdr); + ph = (struct sctp_paramhdr *)(mtod(m, caddr_t) + chunk_len); + ph->param_type = htons(SCTP_ZERO_CHECKSUM_ACCEPTABLE); + ph->param_length = htons(parameter_len); + chunk_len += parameter_len; + } + + /* Add NAT friendly parameter. */ + if (SCTP_BASE_SYSCTL(sctp_inits_include_nat_friendly)) { + parameter_len = (uint16_t)sizeof(struct sctp_paramhdr); + ph = (struct sctp_paramhdr *)(mtod(m, caddr_t) + chunk_len); + ph->param_type = htons(SCTP_HAS_NAT_SUPPORT); + ph->param_length = htons(parameter_len); + chunk_len += parameter_len; + } + + /* And now tell the peer which extensions we support */ + num_ext = 0; + pr_supported = (struct sctp_supported_chunk_types_param *)(mtod(m, caddr_t) + chunk_len); + if (stcb->asoc.prsctp_supported == 1) { + pr_supported->chunk_types[num_ext++] = SCTP_FORWARD_CUM_TSN; + if (stcb->asoc.idata_supported) { + pr_supported->chunk_types[num_ext++] = SCTP_IFORWARD_CUM_TSN; + } + } + if (stcb->asoc.auth_supported == 1) { + pr_supported->chunk_types[num_ext++] = SCTP_AUTHENTICATION; + } + if (stcb->asoc.asconf_supported == 1) { + pr_supported->chunk_types[num_ext++] = SCTP_ASCONF; + pr_supported->chunk_types[num_ext++] = SCTP_ASCONF_ACK; + } + if (stcb->asoc.reconfig_supported == 1) { + pr_supported->chunk_types[num_ext++] = SCTP_STREAM_RESET; + } + if (stcb->asoc.idata_supported) { + pr_supported->chunk_types[num_ext++] = SCTP_IDATA; + } + if (stcb->asoc.nrsack_supported == 1) { + pr_supported->chunk_types[num_ext++] = SCTP_NR_SELECTIVE_ACK; + } + if (stcb->asoc.pktdrop_supported == 1) { + pr_supported->chunk_types[num_ext++] = SCTP_PACKET_DROPPED; + } + if (num_ext > 0) { + parameter_len = (uint16_t)sizeof(struct sctp_supported_chunk_types_param) + num_ext; + pr_supported->ph.param_type = htons(SCTP_SUPPORTED_CHUNK_EXT); + pr_supported->ph.param_length = htons(parameter_len); + padding_len = SCTP_SIZE32(parameter_len) - parameter_len; + chunk_len += parameter_len; + } + /* add authentication parameters */ + if (stcb->asoc.auth_supported) { + /* attach RANDOM parameter, if available */ + if (stcb->asoc.authinfo.random != NULL) { + struct sctp_auth_random *randp; + + if (padding_len > 0) { + memset(mtod(m, caddr_t) + chunk_len, 0, padding_len); + chunk_len += padding_len; + padding_len = 0; + } + randp = (struct sctp_auth_random *)(mtod(m, caddr_t) + chunk_len); + parameter_len = (uint16_t)sizeof(struct sctp_auth_random) + stcb->asoc.authinfo.random_len; + /* random key already contains the header */ + memcpy(randp, stcb->asoc.authinfo.random->key, parameter_len); + padding_len = SCTP_SIZE32(parameter_len) - parameter_len; + chunk_len += parameter_len; + } + /* add HMAC_ALGO parameter */ + if (stcb->asoc.local_hmacs != NULL) { + struct sctp_auth_hmac_algo *hmacs; + + if (padding_len > 0) { + memset(mtod(m, caddr_t) + chunk_len, 0, padding_len); + chunk_len += padding_len; + padding_len = 0; + } + hmacs = (struct sctp_auth_hmac_algo *)(mtod(m, caddr_t) + chunk_len); + parameter_len = (uint16_t)(sizeof(struct sctp_auth_hmac_algo) + + stcb->asoc.local_hmacs->num_algo * sizeof(uint16_t)); + hmacs->ph.param_type = htons(SCTP_HMAC_LIST); + hmacs->ph.param_length = htons(parameter_len); + sctp_serialize_hmaclist(stcb->asoc.local_hmacs, (uint8_t *)hmacs->hmac_ids); + padding_len = SCTP_SIZE32(parameter_len) - parameter_len; + chunk_len += parameter_len; + } + /* add CHUNKS parameter */ + if (stcb->asoc.local_auth_chunks != NULL) { + struct sctp_auth_chunk_list *chunks; + + if (padding_len > 0) { + memset(mtod(m, caddr_t) + chunk_len, 0, padding_len); + chunk_len += padding_len; + padding_len = 0; + } + chunks = (struct sctp_auth_chunk_list *)(mtod(m, caddr_t) + chunk_len); + parameter_len = (uint16_t)(sizeof(struct sctp_auth_chunk_list) + + sctp_auth_get_chklist_size(stcb->asoc.local_auth_chunks)); + chunks->ph.param_type = htons(SCTP_CHUNK_LIST); + chunks->ph.param_length = htons(parameter_len); + sctp_serialize_auth_chunks(stcb->asoc.local_auth_chunks, chunks->chunk_types); + padding_len = SCTP_SIZE32(parameter_len) - parameter_len; + chunk_len += parameter_len; + } + } + + /* now any cookie time extensions */ + if (stcb->asoc.cookie_preserve_req > 0) { + struct sctp_cookie_perserve_param *cookie_preserve; + + if (padding_len > 0) { + memset(mtod(m, caddr_t) + chunk_len, 0, padding_len); + chunk_len += padding_len; + padding_len = 0; + } + parameter_len = (uint16_t)sizeof(struct sctp_cookie_perserve_param); + cookie_preserve = (struct sctp_cookie_perserve_param *)(mtod(m, caddr_t) + chunk_len); + cookie_preserve->ph.param_type = htons(SCTP_COOKIE_PRESERVE); + cookie_preserve->ph.param_length = htons(parameter_len); + cookie_preserve->time = htonl(stcb->asoc.cookie_preserve_req); + stcb->asoc.cookie_preserve_req = 0; + chunk_len += parameter_len; + } + + if (stcb->asoc.scope.ipv4_addr_legal || stcb->asoc.scope.ipv6_addr_legal) { + uint8_t i; + + if (padding_len > 0) { + memset(mtod(m, caddr_t) + chunk_len, 0, padding_len); + chunk_len += padding_len; + padding_len = 0; + } + parameter_len = (uint16_t)sizeof(struct sctp_paramhdr); + if (stcb->asoc.scope.ipv4_addr_legal) { + parameter_len += (uint16_t)sizeof(uint16_t); + } + if (stcb->asoc.scope.ipv6_addr_legal) { + parameter_len += (uint16_t)sizeof(uint16_t); + } + sup_addr = (struct sctp_supported_addr_param *)(mtod(m, caddr_t) + chunk_len); + sup_addr->ph.param_type = htons(SCTP_SUPPORTED_ADDRTYPE); + sup_addr->ph.param_length = htons(parameter_len); + i = 0; + if (stcb->asoc.scope.ipv4_addr_legal) { + sup_addr->addr_type[i++] = htons(SCTP_IPV4_ADDRESS); + } + if (stcb->asoc.scope.ipv6_addr_legal) { + sup_addr->addr_type[i++] = htons(SCTP_IPV6_ADDRESS); + } + padding_len = 4 - 2 * i; + chunk_len += parameter_len; + } + + SCTP_BUF_LEN(m) = chunk_len; + /* now the addresses */ + /* To optimize this we could put the scoping stuff + * into a structure and remove the individual uint8's from + * the assoc structure. Then we could just sifa in the + * address within the stcb. But for now this is a quick + * hack to get the address stuff teased apart. + */ + m_last = sctp_add_addresses_to_i_ia(inp, stcb, &stcb->asoc.scope, + m, cnt_inits_to, + &padding_len, &chunk_len); + + init->ch.chunk_length = htons(chunk_len); + if (padding_len > 0) { + if (sctp_add_pad_tombuf(m_last, padding_len) == NULL) { + sctp_m_freem(m); + return; + } + } + SCTPDBG(SCTP_DEBUG_OUTPUT4, "Sending INIT - calls lowlevel_output\n"); + if ((error = sctp_lowlevel_chunk_output(inp, stcb, net, + (struct sockaddr *)&net->ro._l_addr, + m, 0, NULL, 0, 0, 0, 0, + inp->sctp_lport, stcb->rport, htonl(0), + net->port, NULL, +#if defined(__FreeBSD__) && !defined(__Userspace__) + 0, 0, +#endif + false, so_locked))) { + SCTPDBG(SCTP_DEBUG_OUTPUT4, "Gak send error %d\n", error); + if (error == ENOBUFS) { + stcb->asoc.ifp_had_enobuf = 1; + SCTP_STAT_INCR(sctps_lowlevelerr); + } + } else { + stcb->asoc.ifp_had_enobuf = 0; + } + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); + (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); +} + +struct mbuf * +sctp_arethere_unrecognized_parameters(struct mbuf *in_initpkt, + int param_offset, int *abort_processing, + struct sctp_chunkhdr *cp, + int *nat_friendly, + int *cookie_found) +{ + /* + * Given a mbuf containing an INIT or INIT-ACK with the param_offset + * being equal to the beginning of the params i.e. (iphlen + + * sizeof(struct sctp_init_msg) parse through the parameters to the + * end of the mbuf verifying that all parameters are known. + * + * For unknown parameters build and return a mbuf with + * UNRECOGNIZED_PARAMETER errors. If the flags indicate to stop + * processing this chunk stop, and set *abort_processing to 1. + * + * By having param_offset be pre-set to where parameters begin it is + * hoped that this routine may be reused in the future by new + * features. + */ + struct sctp_paramhdr *phdr, params; + + struct mbuf *mat, *m_tmp, *op_err, *op_err_last; + int at, limit, pad_needed; + uint16_t ptype, plen, padded_size; + + *abort_processing = 0; + if (cookie_found != NULL) { + *cookie_found = 0; + } + mat = in_initpkt; + limit = ntohs(cp->chunk_length) - sizeof(struct sctp_init_chunk); + at = param_offset; + op_err = NULL; + op_err_last = NULL; + pad_needed = 0; + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Check for unrecognized param's\n"); + phdr = sctp_get_next_param(mat, at, ¶ms, sizeof(params)); + while ((phdr != NULL) && ((size_t)limit >= sizeof(struct sctp_paramhdr))) { + ptype = ntohs(phdr->param_type); + plen = ntohs(phdr->param_length); + if ((plen > limit) || (plen < sizeof(struct sctp_paramhdr))) { + /* wacked parameter */ + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error %d\n", plen); + goto invalid_size; + } + limit -= SCTP_SIZE32(plen); + /*- + * All parameters for all chunks that we know/understand are + * listed here. We process them other places and make + * appropriate stop actions per the upper bits. However this + * is the generic routine processor's can call to get back + * an operr.. to either incorporate (init-ack) or send. + */ + padded_size = SCTP_SIZE32(plen); + switch (ptype) { + /* Param's with variable size */ + case SCTP_HEARTBEAT_INFO: + case SCTP_UNRECOG_PARAM: + case SCTP_ERROR_CAUSE_IND: + /* ok skip fwd */ + at += padded_size; + break; + case SCTP_STATE_COOKIE: + if (cookie_found != NULL) { + *cookie_found = 1; + } + at += padded_size; + break; + /* Param's with variable size within a range */ + case SCTP_CHUNK_LIST: + case SCTP_SUPPORTED_CHUNK_EXT: + if (padded_size > (sizeof(struct sctp_supported_chunk_types_param) + (sizeof(uint8_t) * SCTP_MAX_SUPPORTED_EXT))) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error chklist %d\n", plen); + goto invalid_size; + } + at += padded_size; + break; + case SCTP_SUPPORTED_ADDRTYPE: + if (padded_size > SCTP_MAX_ADDR_PARAMS_SIZE) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error supaddrtype %d\n", plen); + goto invalid_size; + } + at += padded_size; + break; + case SCTP_RANDOM: + if (padded_size > (sizeof(struct sctp_auth_random) + SCTP_RANDOM_MAX_SIZE)) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error random %d\n", plen); + goto invalid_size; + } + at += padded_size; + break; + case SCTP_SET_PRIM_ADDR: + case SCTP_DEL_IP_ADDRESS: + case SCTP_ADD_IP_ADDRESS: + if ((padded_size != sizeof(struct sctp_asconf_addrv4_param)) && + (padded_size != sizeof(struct sctp_asconf_addr_param))) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error setprim %d\n", plen); + goto invalid_size; + } + at += padded_size; + break; + /* Param's with a fixed size */ + case SCTP_IPV4_ADDRESS: + if (padded_size != sizeof(struct sctp_ipv4addr_param)) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error ipv4 addr %d\n", plen); + goto invalid_size; + } + at += padded_size; + break; + case SCTP_IPV6_ADDRESS: + if (padded_size != sizeof(struct sctp_ipv6addr_param)) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error ipv6 addr %d\n", plen); + goto invalid_size; + } + at += padded_size; + break; + case SCTP_COOKIE_PRESERVE: + if (padded_size != sizeof(struct sctp_cookie_perserve_param)) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error cookie-preserve %d\n", plen); + goto invalid_size; + } + at += padded_size; + break; + case SCTP_HAS_NAT_SUPPORT: + *nat_friendly = 1; + /* fall through */ + case SCTP_PRSCTP_SUPPORTED: + if (padded_size != sizeof(struct sctp_paramhdr)) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error prsctp/nat support %d\n", plen); + goto invalid_size; + } + at += padded_size; + break; + case SCTP_ECN_CAPABLE: + if (padded_size != sizeof(struct sctp_paramhdr)) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error ecn %d\n", plen); + goto invalid_size; + } + at += padded_size; + break; + case SCTP_ULP_ADAPTATION: + if (padded_size != sizeof(struct sctp_adaptation_layer_indication)) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error adapatation %d\n", plen); + goto invalid_size; + } + at += padded_size; + break; + case SCTP_SUCCESS_REPORT: + if (padded_size != sizeof(struct sctp_asconf_paramhdr)) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error success %d\n", plen); + goto invalid_size; + } + at += padded_size; + break; + case SCTP_HOSTNAME_ADDRESS: + { + /* Hostname parameters are deprecated. */ + struct sctp_gen_error_cause *cause; + int l_len; + + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Can't handle hostname addresses.. abort processing\n"); + *abort_processing = 1; + sctp_m_freem(op_err); + op_err = NULL; + op_err_last = NULL; +#ifdef INET6 + l_len = SCTP_MIN_OVERHEAD; +#else + l_len = SCTP_MIN_V4_OVERHEAD; +#endif + l_len += sizeof(struct sctp_chunkhdr); + l_len += sizeof(struct sctp_gen_error_cause); + op_err = sctp_get_mbuf_for_msg(l_len, 0, M_NOWAIT, 1, MT_DATA); + if (op_err != NULL) { + /* + * Pre-reserve space for IP, SCTP, and + * chunk header. + */ +#ifdef INET6 + SCTP_BUF_RESV_UF(op_err, sizeof(struct ip6_hdr)); +#else + SCTP_BUF_RESV_UF(op_err, sizeof(struct ip)); +#endif + SCTP_BUF_RESV_UF(op_err, sizeof(struct sctphdr)); + SCTP_BUF_RESV_UF(op_err, sizeof(struct sctp_chunkhdr)); + SCTP_BUF_LEN(op_err) = sizeof(struct sctp_gen_error_cause); + cause = mtod(op_err, struct sctp_gen_error_cause *); + cause->code = htons(SCTP_CAUSE_UNRESOLVABLE_ADDR); + cause->length = htons((uint16_t)(sizeof(struct sctp_gen_error_cause) + plen)); + SCTP_BUF_NEXT(op_err) = SCTP_M_COPYM(mat, at, plen, M_NOWAIT); + if (SCTP_BUF_NEXT(op_err) == NULL) { + sctp_m_freem(op_err); + op_err = NULL; + op_err_last = NULL; + } + } + return (op_err); + } + default: + /* + * we do not recognize the parameter figure out what + * we do. + */ + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Hit default param %x\n", ptype); + if ((ptype & 0x4000) == 0x4000) { + /* Report bit is set?? */ + SCTPDBG(SCTP_DEBUG_OUTPUT1, "report op err\n"); + if (op_err == NULL) { + int l_len; + /* Ok need to try to get an mbuf */ +#ifdef INET6 + l_len = SCTP_MIN_OVERHEAD; +#else + l_len = SCTP_MIN_V4_OVERHEAD; +#endif + l_len += sizeof(struct sctp_chunkhdr); + l_len += sizeof(struct sctp_paramhdr); + op_err = sctp_get_mbuf_for_msg(l_len, 0, M_NOWAIT, 1, MT_DATA); + if (op_err) { + SCTP_BUF_LEN(op_err) = 0; +#ifdef INET6 + SCTP_BUF_RESV_UF(op_err, sizeof(struct ip6_hdr)); +#else + SCTP_BUF_RESV_UF(op_err, sizeof(struct ip)); +#endif + SCTP_BUF_RESV_UF(op_err, sizeof(struct sctphdr)); + SCTP_BUF_RESV_UF(op_err, sizeof(struct sctp_chunkhdr)); + op_err_last = op_err; + } + } + if (op_err != NULL) { + /* If we have space */ + struct sctp_paramhdr *param; + + if (pad_needed > 0) { + op_err_last = sctp_add_pad_tombuf(op_err_last, pad_needed); + } + if (op_err_last == NULL) { + sctp_m_freem(op_err); + op_err = NULL; + op_err_last = NULL; + goto more_processing; + } + if (M_TRAILINGSPACE(op_err_last) < (int)sizeof(struct sctp_paramhdr)) { + m_tmp = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr), 0, M_NOWAIT, 1, MT_DATA); + if (m_tmp == NULL) { + sctp_m_freem(op_err); + op_err = NULL; + op_err_last = NULL; + goto more_processing; + } + SCTP_BUF_LEN(m_tmp) = 0; + SCTP_BUF_NEXT(m_tmp) = NULL; + SCTP_BUF_NEXT(op_err_last) = m_tmp; + op_err_last = m_tmp; + } + param = (struct sctp_paramhdr *)(mtod(op_err_last, caddr_t) + SCTP_BUF_LEN(op_err_last)); + param->param_type = htons(SCTP_UNRECOG_PARAM); + param->param_length = htons((uint16_t)sizeof(struct sctp_paramhdr) + plen); + SCTP_BUF_LEN(op_err_last) += sizeof(struct sctp_paramhdr); + SCTP_BUF_NEXT(op_err_last) = SCTP_M_COPYM(mat, at, plen, M_NOWAIT); + if (SCTP_BUF_NEXT(op_err_last) == NULL) { + sctp_m_freem(op_err); + op_err = NULL; + op_err_last = NULL; + goto more_processing; + } else { + while (SCTP_BUF_NEXT(op_err_last) != NULL) { + op_err_last = SCTP_BUF_NEXT(op_err_last); + } + } + if (plen % 4 != 0) { + pad_needed = 4 - (plen % 4); + } else { + pad_needed = 0; + } + } + } + more_processing: + if ((ptype & 0x8000) == 0x0000) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "stop proc\n"); + return (op_err); + } else { + /* skip this chunk and continue processing */ + SCTPDBG(SCTP_DEBUG_OUTPUT1, "move on\n"); + at += SCTP_SIZE32(plen); + } + break; + } + phdr = sctp_get_next_param(mat, at, ¶ms, sizeof(params)); + } + return (op_err); + invalid_size: + SCTPDBG(SCTP_DEBUG_OUTPUT1, "abort flag set\n"); + *abort_processing = 1; + sctp_m_freem(op_err); + op_err = NULL; + op_err_last = NULL; + if (phdr != NULL) { + struct sctp_paramhdr *param; + int l_len; +#ifdef INET6 + l_len = SCTP_MIN_OVERHEAD; +#else + l_len = SCTP_MIN_V4_OVERHEAD; +#endif + l_len += sizeof(struct sctp_chunkhdr); + l_len += (2 * sizeof(struct sctp_paramhdr)); + op_err = sctp_get_mbuf_for_msg(l_len, 0, M_NOWAIT, 1, MT_DATA); + if (op_err) { + SCTP_BUF_LEN(op_err) = 0; +#ifdef INET6 + SCTP_BUF_RESV_UF(op_err, sizeof(struct ip6_hdr)); +#else + SCTP_BUF_RESV_UF(op_err, sizeof(struct ip)); +#endif + SCTP_BUF_RESV_UF(op_err, sizeof(struct sctphdr)); + SCTP_BUF_RESV_UF(op_err, sizeof(struct sctp_chunkhdr)); + SCTP_BUF_LEN(op_err) = 2 * sizeof(struct sctp_paramhdr); + param = mtod(op_err, struct sctp_paramhdr *); + param->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); + param->param_length = htons(2 * sizeof(struct sctp_paramhdr)); + param++; + param->param_type = htons(ptype); + param->param_length = htons(plen); + } + } + return (op_err); +} + +/* + * Given a INIT chunk, look through the parameters to verify that there + * are no new addresses. + * Return true, if there is a new address or there is a problem parsing + the parameters. Provide an optional error cause used when sending an ABORT. + * Return false, if there are no new addresses and there is no problem in + parameter processing. + */ +static bool +sctp_are_there_new_addresses(struct sctp_association *asoc, + struct mbuf *in_initpkt, int offset, int limit, struct sockaddr *src, + struct mbuf **op_err) +{ + struct sockaddr *sa_touse; + struct sockaddr *sa; + struct sctp_paramhdr *phdr, params; + struct sctp_nets *net; +#ifdef INET + struct sockaddr_in sin4, *sa4; +#endif +#ifdef INET6 + struct sockaddr_in6 sin6, *sa6; +#endif +#if defined(__Userspace__) + struct sockaddr_conn *sac; +#endif + uint16_t ptype, plen; + bool fnd, check_src; + + *op_err = NULL; +#ifdef INET + memset(&sin4, 0, sizeof(sin4)); + sin4.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + sin4.sin_len = sizeof(sin4); +#endif +#endif +#ifdef INET6 + memset(&sin6, 0, sizeof(sin6)); + sin6.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6.sin6_len = sizeof(sin6); +#endif +#endif + /* First what about the src address of the pkt ? */ + check_src = false; + switch (src->sa_family) { +#ifdef INET + case AF_INET: + if (asoc->scope.ipv4_addr_legal) { + check_src = true; + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (asoc->scope.ipv6_addr_legal) { + check_src = true; + } + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + if (asoc->scope.conn_addr_legal) { + check_src = true; + } + break; +#endif + default: + /* TSNH */ + break; + } + if (check_src) { + fnd = false; + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + sa = (struct sockaddr *)&net->ro._l_addr; + if (sa->sa_family == src->sa_family) { +#ifdef INET + if (sa->sa_family == AF_INET) { + struct sockaddr_in *src4; + + sa4 = (struct sockaddr_in *)sa; + src4 = (struct sockaddr_in *)src; + if (sa4->sin_addr.s_addr == src4->sin_addr.s_addr) { + fnd = true; + break; + } + } +#endif +#ifdef INET6 + if (sa->sa_family == AF_INET6) { + struct sockaddr_in6 *src6; + + sa6 = (struct sockaddr_in6 *)sa; + src6 = (struct sockaddr_in6 *)src; + if (SCTP6_ARE_ADDR_EQUAL(sa6, src6)) { + fnd = true; + break; + } + } +#endif +#if defined(__Userspace__) + if (sa->sa_family == AF_CONN) { + struct sockaddr_conn *srcc; + + sac = (struct sockaddr_conn *)sa; + srcc = (struct sockaddr_conn *)src; + if (sac->sconn_addr == srcc->sconn_addr) { + fnd = true; + break; + } + } +#endif + } + } + if (!fnd) { + /* + * If sending an ABORT in case of an additional address, + * don't use the new address error cause. + * This looks no different than if no listener was + * present. + */ + *op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), "Address added"); + return (true); + } + } + /* Ok so far lets munge through the rest of the packet */ + offset += sizeof(struct sctp_init_chunk); + phdr = sctp_get_next_param(in_initpkt, offset, ¶ms, sizeof(params)); + while (phdr) { + sa_touse = NULL; + ptype = ntohs(phdr->param_type); + plen = ntohs(phdr->param_length); + if (offset + plen > limit) { + *op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, "Partial parameter"); + return (true); + } + if (plen < sizeof(struct sctp_paramhdr)) { + *op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, "Parameter length too small"); + return (true); + } + switch (ptype) { +#ifdef INET + case SCTP_IPV4_ADDRESS: + { + struct sctp_ipv4addr_param *p4, p4_buf; + + if (plen != sizeof(struct sctp_ipv4addr_param)) { + *op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, "Parameter length illegal"); + return (true); + } + phdr = sctp_get_next_param(in_initpkt, offset, + (struct sctp_paramhdr *)&p4_buf, sizeof(p4_buf)); + if (phdr == NULL) { + *op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, ""); + return (true); + } + if (asoc->scope.ipv4_addr_legal) { + p4 = (struct sctp_ipv4addr_param *)phdr; + sin4.sin_addr.s_addr = p4->addr; + sa_touse = (struct sockaddr *)&sin4; + } + break; + } +#endif +#ifdef INET6 + case SCTP_IPV6_ADDRESS: + { + struct sctp_ipv6addr_param *p6, p6_buf; + + if (plen != sizeof(struct sctp_ipv6addr_param)) { + *op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, "Parameter length illegal"); + return (true); + } + phdr = sctp_get_next_param(in_initpkt, offset, + (struct sctp_paramhdr *)&p6_buf, sizeof(p6_buf)); + if (phdr == NULL) { + *op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION, ""); + return (true); + } + if (asoc->scope.ipv6_addr_legal) { + p6 = (struct sctp_ipv6addr_param *)phdr; + memcpy((caddr_t)&sin6.sin6_addr, p6->addr, + sizeof(p6->addr)); + sa_touse = (struct sockaddr *)&sin6; + } + break; + } +#endif + default: + sa_touse = NULL; + break; + } + if (sa_touse) { + /* ok, sa_touse points to one to check */ + fnd = false; + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + sa = (struct sockaddr *)&net->ro._l_addr; + if (sa->sa_family != sa_touse->sa_family) { + continue; + } +#ifdef INET + if (sa->sa_family == AF_INET) { + sa4 = (struct sockaddr_in *)sa; + if (sa4->sin_addr.s_addr == + sin4.sin_addr.s_addr) { + fnd = true; + break; + } + } +#endif +#ifdef INET6 + if (sa->sa_family == AF_INET6) { + sa6 = (struct sockaddr_in6 *)sa; + if (SCTP6_ARE_ADDR_EQUAL( + sa6, &sin6)) { + fnd = true; + break; + } + } +#endif + } + if (!fnd) { + /* + * If sending an ABORT in case of an additional + * address, don't use the new address error + * cause. + * This looks no different than if no listener + * was present. + */ + *op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), "Address added"); + return (true); + } + } + offset += SCTP_SIZE32(plen); + if (offset >= limit) { + break; + } + phdr = sctp_get_next_param(in_initpkt, offset, ¶ms, sizeof(params)); + } + return (false); +} + +/* + * Given a MBUF chain that was sent into us containing an INIT. Build a + * INIT-ACK with COOKIE and send back. We assume that the in_initpkt has done + * a pullup to include IPv6/4header, SCTP header and initial part of INIT + * message (i.e. the struct sctp_init_msg). + */ +void +sctp_send_initiate_ack(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct sctp_nets *src_net, struct mbuf *init_pkt, + int iphlen, int offset, + struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, struct sctp_init_chunk *init_chk, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, +#endif + uint32_t vrf_id, uint16_t port) +{ + struct sctp_association *asoc; + struct mbuf *m, *m_tmp, *m_last, *m_cookie, *op_err; + struct sctp_init_ack_chunk *initack; + struct sctp_adaptation_layer_indication *ali; + struct sctp_supported_chunk_types_param *pr_supported; + struct sctp_paramhdr *ph; + union sctp_sockstore *over_addr; + struct sctp_scoping scp; + struct timeval now; +#ifdef INET + struct sockaddr_in *dst4 = (struct sockaddr_in *)dst; + struct sockaddr_in *src4 = (struct sockaddr_in *)src; + struct sockaddr_in *sin; +#endif +#ifdef INET6 + struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst; + struct sockaddr_in6 *src6 = (struct sockaddr_in6 *)src; + struct sockaddr_in6 *sin6; +#endif +#if defined(__Userspace__) + struct sockaddr_conn *dstconn = (struct sockaddr_conn *)dst; + struct sockaddr_conn *srcconn = (struct sockaddr_conn *)src; + struct sockaddr_conn *sconn; +#endif + struct sockaddr *to; + struct sctp_state_cookie stc; + struct sctp_nets *net = NULL; + uint8_t *signature = NULL; + int cnt_inits_to = 0; + uint16_t his_limit, i_want; + int abort_flag; + int nat_friendly = 0; + int error; + struct socket *so; + uint16_t num_ext, chunk_len, padding_len, parameter_len; + + if (stcb) { + asoc = &stcb->asoc; + } else { + asoc = NULL; + } + if ((asoc != NULL) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_COOKIE_WAIT)) { + if (sctp_are_there_new_addresses(asoc, init_pkt, offset, offset + ntohs(init_chk->ch.chunk_length), src, &op_err)) { + /* + * new addresses, out of here in non-cookie-wait states + */ + sctp_send_abort(init_pkt, iphlen, src, dst, sh, 0, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, port); + return; + } + if (src_net != NULL && (src_net->port != port)) { + /* + * change of remote encapsulation port, out of here in + * non-cookie-wait states + * + * Send an ABORT, without an specific error cause. + * This looks no different than if no listener + * was present. + */ + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + "Remote encapsulation port changed"); + sctp_send_abort(init_pkt, iphlen, src, dst, sh, 0, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, port); + return; + } + } + abort_flag = 0; + op_err = sctp_arethere_unrecognized_parameters(init_pkt, + (offset + sizeof(struct sctp_init_chunk)), + &abort_flag, + (struct sctp_chunkhdr *)init_chk, + &nat_friendly, NULL); + if (abort_flag) { + do_a_abort: + if (op_err == NULL) { + char msg[SCTP_DIAG_INFO_LEN]; + + SCTP_SNPRINTF(msg, sizeof(msg), "%s:%d at %s", __FILE__, __LINE__, __func__); + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + msg); + } + sctp_send_abort(init_pkt, iphlen, src, dst, sh, + init_chk->init.initiate_tag, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, port); + return; + } + m = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_NOWAIT, 1, MT_DATA); + if (m == NULL) { + /* No memory, INIT timer will re-attempt. */ + sctp_m_freem(op_err); + return; + } + chunk_len = (uint16_t)sizeof(struct sctp_init_ack_chunk); + padding_len = 0; + + /* + * We might not overwrite the identification[] completely and on + * some platforms time_entered will contain some padding. + * Therefore zero out the cookie to avoid putting + * uninitialized memory on the wire. + */ + memset(&stc, 0, sizeof(struct sctp_state_cookie)); + + /* the time I built cookie */ + (void)SCTP_GETTIME_TIMEVAL(&now); + stc.time_entered.tv_sec = now.tv_sec; + stc.time_entered.tv_usec = now.tv_usec; + + /* populate any tie tags */ + if (asoc != NULL) { + /* unlock before tag selections */ + stc.tie_tag_my_vtag = asoc->my_vtag_nonce; + stc.tie_tag_peer_vtag = asoc->peer_vtag_nonce; + stc.cookie_life = asoc->cookie_life; + net = asoc->primary_destination; + } else { + stc.tie_tag_my_vtag = 0; + stc.tie_tag_peer_vtag = 0; + /* life I will award this cookie */ + stc.cookie_life = inp->sctp_ep.def_cookie_life; + } + + /* copy in the ports for later check */ + stc.myport = sh->dest_port; + stc.peerport = sh->src_port; + + /* + * If we wanted to honor cookie life extensions, we would add to + * stc.cookie_life. For now we should NOT honor any extension + */ + stc.site_scope = stc.local_scope = stc.loopback_scope = 0; + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + stc.ipv6_addr_legal = 1; + if (SCTP_IPV6_V6ONLY(inp)) { + stc.ipv4_addr_legal = 0; + } else { + stc.ipv4_addr_legal = 1; + } +#if defined(__Userspace__) + stc.conn_addr_legal = 0; +#endif + } else { + stc.ipv6_addr_legal = 0; +#if defined(__Userspace__) + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_CONN) { + stc.conn_addr_legal = 1; + stc.ipv4_addr_legal = 0; + } else { + stc.conn_addr_legal = 0; + stc.ipv4_addr_legal = 1; + } +#else + stc.ipv4_addr_legal = 1; +#endif + } + stc.ipv4_scope = 0; + if (net == NULL) { + to = src; + switch (dst->sa_family) { +#ifdef INET + case AF_INET: + { + /* lookup address */ + stc.address[0] = src4->sin_addr.s_addr; + stc.address[1] = 0; + stc.address[2] = 0; + stc.address[3] = 0; + stc.addr_type = SCTP_IPV4_ADDRESS; + /* local from address */ + stc.laddress[0] = dst4->sin_addr.s_addr; + stc.laddress[1] = 0; + stc.laddress[2] = 0; + stc.laddress[3] = 0; + stc.laddr_type = SCTP_IPV4_ADDRESS; + /* scope_id is only for v6 */ + stc.scope_id = 0; + if ((IN4_ISPRIVATE_ADDRESS(&src4->sin_addr)) || + (IN4_ISPRIVATE_ADDRESS(&dst4->sin_addr))) { + stc.ipv4_scope = 1; + } + /* Must use the address in this case */ + if (sctp_is_address_on_local_host(src, vrf_id)) { + stc.loopback_scope = 1; + stc.ipv4_scope = 1; + stc.site_scope = 1; + stc.local_scope = 0; + } + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + stc.addr_type = SCTP_IPV6_ADDRESS; + memcpy(&stc.address, &src6->sin6_addr, sizeof(struct in6_addr)); +#if defined(__FreeBSD__) && !defined(__Userspace__) + stc.scope_id = ntohs(in6_getscope(&src6->sin6_addr)); +#else + stc.scope_id = 0; +#endif + if (sctp_is_address_on_local_host(src, vrf_id)) { + stc.loopback_scope = 1; + stc.local_scope = 0; + stc.site_scope = 1; + stc.ipv4_scope = 1; + } else if (IN6_IS_ADDR_LINKLOCAL(&src6->sin6_addr) || + IN6_IS_ADDR_LINKLOCAL(&dst6->sin6_addr)) { + /* + * If the new destination or source is a + * LINK_LOCAL we must have common both site and + * local scope. Don't set local scope though + * since we must depend on the source to be + * added implicitly. We cannot assure just + * because we share one link that all links are + * common. + */ +#if defined(__APPLE__) && !defined(__Userspace__) + /* Mac OS X currently doesn't have in6_getscope() */ + stc.scope_id = src6->sin6_addr.s6_addr16[1]; +#endif + stc.local_scope = 0; + stc.site_scope = 1; + stc.ipv4_scope = 1; + /* + * we start counting for the private address + * stuff at 1. since the link local we + * source from won't show up in our scoped + * count. + */ + cnt_inits_to = 1; + /* pull out the scope_id from incoming pkt */ + } else if (IN6_IS_ADDR_SITELOCAL(&src6->sin6_addr) || + IN6_IS_ADDR_SITELOCAL(&dst6->sin6_addr)) { + /* + * If the new destination or source is + * SITE_LOCAL then we must have site scope in + * common. + */ + stc.site_scope = 1; + } + memcpy(&stc.laddress, &dst6->sin6_addr, sizeof(struct in6_addr)); + stc.laddr_type = SCTP_IPV6_ADDRESS; + break; + } +#endif +#if defined(__Userspace__) + case AF_CONN: + { + /* lookup address */ + stc.address[0] = 0; + stc.address[1] = 0; + stc.address[2] = 0; + stc.address[3] = 0; + memcpy(&stc.address, &srcconn->sconn_addr, sizeof(void *)); + stc.addr_type = SCTP_CONN_ADDRESS; + /* local from address */ + stc.laddress[0] = 0; + stc.laddress[1] = 0; + stc.laddress[2] = 0; + stc.laddress[3] = 0; + memcpy(&stc.laddress, &dstconn->sconn_addr, sizeof(void *)); + stc.laddr_type = SCTP_CONN_ADDRESS; + /* scope_id is only for v6 */ + stc.scope_id = 0; + break; + } +#endif + default: + /* TSNH */ + goto do_a_abort; + break; + } + } else { + /* set the scope per the existing tcb */ + +#ifdef INET6 + struct sctp_nets *lnet; +#endif + + stc.loopback_scope = asoc->scope.loopback_scope; + stc.ipv4_scope = asoc->scope.ipv4_local_scope; + stc.site_scope = asoc->scope.site_scope; + stc.local_scope = asoc->scope.local_scope; +#ifdef INET6 + /* Why do we not consider IPv4 LL addresses? */ + TAILQ_FOREACH(lnet, &asoc->nets, sctp_next) { + if (lnet->ro._l_addr.sin6.sin6_family == AF_INET6) { + if (IN6_IS_ADDR_LINKLOCAL(&lnet->ro._l_addr.sin6.sin6_addr)) { + /* + * if we have a LL address, start + * counting at 1. + */ + cnt_inits_to = 1; + } + } + } +#endif + /* use the net pointer */ + to = (struct sockaddr *)&net->ro._l_addr; + switch (to->sa_family) { +#ifdef INET + case AF_INET: + sin = (struct sockaddr_in *)to; + stc.address[0] = sin->sin_addr.s_addr; + stc.address[1] = 0; + stc.address[2] = 0; + stc.address[3] = 0; + stc.addr_type = SCTP_IPV4_ADDRESS; + if (net->src_addr_selected == 0) { + /* + * strange case here, the INIT should have + * did the selection. + */ + net->ro._s_addr = sctp_source_address_selection(inp, + stcb, (sctp_route_t *)&net->ro, + net, 0, vrf_id); + if (net->ro._s_addr == NULL) { + sctp_m_freem(op_err); + sctp_m_freem(m); + return; + } + + net->src_addr_selected = 1; + } + stc.laddress[0] = net->ro._s_addr->address.sin.sin_addr.s_addr; + stc.laddress[1] = 0; + stc.laddress[2] = 0; + stc.laddress[3] = 0; + stc.laddr_type = SCTP_IPV4_ADDRESS; + /* scope_id is only for v6 */ + stc.scope_id = 0; + break; +#endif +#ifdef INET6 + case AF_INET6: + sin6 = (struct sockaddr_in6 *)to; + memcpy(&stc.address, &sin6->sin6_addr, + sizeof(struct in6_addr)); + stc.addr_type = SCTP_IPV6_ADDRESS; + stc.scope_id = sin6->sin6_scope_id; + if (net->src_addr_selected == 0) { + /* + * strange case here, the INIT should have + * done the selection. + */ + net->ro._s_addr = sctp_source_address_selection(inp, + stcb, (sctp_route_t *)&net->ro, + net, 0, vrf_id); + if (net->ro._s_addr == NULL) { + sctp_m_freem(op_err); + sctp_m_freem(m); + return; + } + + net->src_addr_selected = 1; + } + memcpy(&stc.laddress, &net->ro._s_addr->address.sin6.sin6_addr, + sizeof(struct in6_addr)); + stc.laddr_type = SCTP_IPV6_ADDRESS; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + sconn = (struct sockaddr_conn *)to; + stc.address[0] = 0; + stc.address[1] = 0; + stc.address[2] = 0; + stc.address[3] = 0; + memcpy(&stc.address, &sconn->sconn_addr, sizeof(void *)); + stc.addr_type = SCTP_CONN_ADDRESS; + stc.laddress[0] = 0; + stc.laddress[1] = 0; + stc.laddress[2] = 0; + stc.laddress[3] = 0; + memcpy(&stc.laddress, &sconn->sconn_addr, sizeof(void *)); + stc.laddr_type = SCTP_CONN_ADDRESS; + stc.scope_id = 0; + break; +#endif + } + } + if (asoc != NULL) { + stc.zero_checksum = asoc->zero_checksum > 0 ? 1 : 0; + } else { + stc.zero_checksum = inp->zero_checksum; + } + /* Now lets put the SCTP header in place */ + initack = mtod(m, struct sctp_init_ack_chunk *); + /* Save it off for quick ref */ + stc.peers_vtag = ntohl(init_chk->init.initiate_tag); + /* who are we */ + memcpy(stc.identification, SCTP_VERSION_STRING, + min(strlen(SCTP_VERSION_STRING), sizeof(stc.identification))); + memset(stc.reserved, 0, SCTP_RESERVE_SPACE); + /* now the chunk header */ + initack->ch.chunk_type = SCTP_INITIATION_ACK; + initack->ch.chunk_flags = 0; + /* fill in later from mbuf we build */ + initack->ch.chunk_length = 0; + /* place in my tag */ + if ((asoc != NULL) && + ((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_INUSE) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED))) { + /* re-use the v-tags and init-seq here */ + initack->init.initiate_tag = htonl(asoc->my_vtag); + initack->init.initial_tsn = htonl(asoc->init_seq_number); + } else { + uint32_t vtag, itsn; + + if (asoc) { + atomic_add_int(&asoc->refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + new_tag: + SCTP_INP_INFO_RLOCK(); + vtag = sctp_select_a_tag(inp, inp->sctp_lport, sh->src_port, 1); + SCTP_INP_INFO_RUNLOCK(); + if ((asoc->peer_supports_nat) && (vtag == asoc->my_vtag)) { + /* Got a duplicate vtag on some guy behind a nat + * make sure we don't use it. + */ + goto new_tag; + } + initack->init.initiate_tag = htonl(vtag); + /* get a TSN to use too */ + itsn = sctp_select_initial_TSN(&inp->sctp_ep); + initack->init.initial_tsn = htonl(itsn); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&asoc->refcnt, 1); + } else { + SCTP_INP_INCR_REF(inp); + SCTP_INP_RUNLOCK(inp); + SCTP_INP_INFO_RLOCK(); + vtag = sctp_select_a_tag(inp, inp->sctp_lport, sh->src_port, 1); + SCTP_INP_INFO_RUNLOCK(); + initack->init.initiate_tag = htonl(vtag); + /* get a TSN to use too */ + initack->init.initial_tsn = htonl(sctp_select_initial_TSN(&inp->sctp_ep)); + SCTP_INP_RLOCK(inp); + SCTP_INP_DECR_REF(inp); + } + } + /* save away my tag to */ + stc.my_vtag = initack->init.initiate_tag; + + /* set up some of the credits. */ + so = inp->sctp_socket; + if (so == NULL) { + /* memory problem */ + sctp_m_freem(op_err); + sctp_m_freem(m); + return; + } else { + initack->init.a_rwnd = htonl(max(SCTP_SB_LIMIT_RCV(so), SCTP_MINIMAL_RWND)); + } + /* set what I want */ + his_limit = ntohs(init_chk->init.num_inbound_streams); + /* choose what I want */ + if (asoc != NULL) { + if (asoc->streamoutcnt > asoc->pre_open_streams) { + i_want = asoc->streamoutcnt; + } else { + i_want = asoc->pre_open_streams; + } + } else { + i_want = inp->sctp_ep.pre_open_stream_count; + } + if (his_limit < i_want) { + /* I Want more :< */ + initack->init.num_outbound_streams = init_chk->init.num_inbound_streams; + } else { + /* I can have what I want :> */ + initack->init.num_outbound_streams = htons(i_want); + } + /* tell him his limit. */ + initack->init.num_inbound_streams = + htons(inp->sctp_ep.max_open_streams_intome); + + /* adaptation layer indication parameter */ + if (inp->sctp_ep.adaptation_layer_indicator_provided) { + parameter_len = (uint16_t)sizeof(struct sctp_adaptation_layer_indication); + ali = (struct sctp_adaptation_layer_indication *)(mtod(m, caddr_t) + chunk_len); + ali->ph.param_type = htons(SCTP_ULP_ADAPTATION); + ali->ph.param_length = htons(parameter_len); + ali->indication = htonl(inp->sctp_ep.adaptation_layer_indicator); + chunk_len += parameter_len; + } + + /* ECN parameter */ + if (((asoc != NULL) && (asoc->ecn_supported == 1)) || + ((asoc == NULL) && (inp->ecn_supported == 1))) { + parameter_len = (uint16_t)sizeof(struct sctp_paramhdr); + ph = (struct sctp_paramhdr *)(mtod(m, caddr_t) + chunk_len); + ph->param_type = htons(SCTP_ECN_CAPABLE); + ph->param_length = htons(parameter_len); + chunk_len += parameter_len; + } + + /* PR-SCTP supported parameter */ + if (((asoc != NULL) && (asoc->prsctp_supported == 1)) || + ((asoc == NULL) && (inp->prsctp_supported == 1))) { + parameter_len = (uint16_t)sizeof(struct sctp_paramhdr); + ph = (struct sctp_paramhdr *)(mtod(m, caddr_t) + chunk_len); + ph->param_type = htons(SCTP_PRSCTP_SUPPORTED); + ph->param_length = htons(parameter_len); + chunk_len += parameter_len; + } + + /* Zero checksum acceptable parameter */ + if (((asoc != NULL) && (asoc->zero_checksum > 0)) || + ((asoc == NULL) && (inp->zero_checksum == 1))) { + parameter_len = (uint16_t)sizeof(struct sctp_paramhdr); + ph = (struct sctp_paramhdr *)(mtod(m, caddr_t) + chunk_len); + ph->param_type = htons(SCTP_ZERO_CHECKSUM_ACCEPTABLE); + ph->param_length = htons(parameter_len); + chunk_len += parameter_len; + } + + /* Add NAT friendly parameter */ + if (nat_friendly) { + parameter_len = (uint16_t)sizeof(struct sctp_paramhdr); + ph = (struct sctp_paramhdr *)(mtod(m, caddr_t) + chunk_len); + ph->param_type = htons(SCTP_HAS_NAT_SUPPORT); + ph->param_length = htons(parameter_len); + chunk_len += parameter_len; + } + + /* And now tell the peer which extensions we support */ + num_ext = 0; + pr_supported = (struct sctp_supported_chunk_types_param *)(mtod(m, caddr_t) + chunk_len); + if (((asoc != NULL) && (asoc->prsctp_supported == 1)) || + ((asoc == NULL) && (inp->prsctp_supported == 1))) { + pr_supported->chunk_types[num_ext++] = SCTP_FORWARD_CUM_TSN; + if (((asoc != NULL) && (asoc->idata_supported == 1)) || + ((asoc == NULL) && (inp->idata_supported == 1))) { + pr_supported->chunk_types[num_ext++] = SCTP_IFORWARD_CUM_TSN; + } + } + if (((asoc != NULL) && (asoc->auth_supported == 1)) || + ((asoc == NULL) && (inp->auth_supported == 1))) { + pr_supported->chunk_types[num_ext++] = SCTP_AUTHENTICATION; + } + if (((asoc != NULL) && (asoc->asconf_supported == 1)) || + ((asoc == NULL) && (inp->asconf_supported == 1))) { + pr_supported->chunk_types[num_ext++] = SCTP_ASCONF; + pr_supported->chunk_types[num_ext++] = SCTP_ASCONF_ACK; + } + if (((asoc != NULL) && (asoc->reconfig_supported == 1)) || + ((asoc == NULL) && (inp->reconfig_supported == 1))) { + pr_supported->chunk_types[num_ext++] = SCTP_STREAM_RESET; + } + if (((asoc != NULL) && (asoc->idata_supported == 1)) || + ((asoc == NULL) && (inp->idata_supported == 1))) { + pr_supported->chunk_types[num_ext++] = SCTP_IDATA; + } + if (((asoc != NULL) && (asoc->nrsack_supported == 1)) || + ((asoc == NULL) && (inp->nrsack_supported == 1))) { + pr_supported->chunk_types[num_ext++] = SCTP_NR_SELECTIVE_ACK; + } + if (((asoc != NULL) && (asoc->pktdrop_supported == 1)) || + ((asoc == NULL) && (inp->pktdrop_supported == 1))) { + pr_supported->chunk_types[num_ext++] = SCTP_PACKET_DROPPED; + } + if (num_ext > 0) { + parameter_len = (uint16_t)sizeof(struct sctp_supported_chunk_types_param) + num_ext; + pr_supported->ph.param_type = htons(SCTP_SUPPORTED_CHUNK_EXT); + pr_supported->ph.param_length = htons(parameter_len); + padding_len = SCTP_SIZE32(parameter_len) - parameter_len; + chunk_len += parameter_len; + } + + /* add authentication parameters */ + if (((asoc != NULL) && (asoc->auth_supported == 1)) || + ((asoc == NULL) && (inp->auth_supported == 1))) { + struct sctp_auth_random *randp; + struct sctp_auth_hmac_algo *hmacs; + struct sctp_auth_chunk_list *chunks; + + if (padding_len > 0) { + memset(mtod(m, caddr_t) + chunk_len, 0, padding_len); + chunk_len += padding_len; + padding_len = 0; + } + /* generate and add RANDOM parameter */ + randp = (struct sctp_auth_random *)(mtod(m, caddr_t) + chunk_len); + parameter_len = (uint16_t)sizeof(struct sctp_auth_random) + + SCTP_AUTH_RANDOM_SIZE_DEFAULT; + randp->ph.param_type = htons(SCTP_RANDOM); + randp->ph.param_length = htons(parameter_len); + SCTP_READ_RANDOM(randp->random_data, SCTP_AUTH_RANDOM_SIZE_DEFAULT); + padding_len = SCTP_SIZE32(parameter_len) - parameter_len; + chunk_len += parameter_len; + + if (padding_len > 0) { + memset(mtod(m, caddr_t) + chunk_len, 0, padding_len); + chunk_len += padding_len; + padding_len = 0; + } + /* add HMAC_ALGO parameter */ + hmacs = (struct sctp_auth_hmac_algo *)(mtod(m, caddr_t) + chunk_len); + parameter_len = (uint16_t)sizeof(struct sctp_auth_hmac_algo) + + sctp_serialize_hmaclist(inp->sctp_ep.local_hmacs, + (uint8_t *)hmacs->hmac_ids); + hmacs->ph.param_type = htons(SCTP_HMAC_LIST); + hmacs->ph.param_length = htons(parameter_len); + padding_len = SCTP_SIZE32(parameter_len) - parameter_len; + chunk_len += parameter_len; + + if (padding_len > 0) { + memset(mtod(m, caddr_t) + chunk_len, 0, padding_len); + chunk_len += padding_len; + padding_len = 0; + } + /* add CHUNKS parameter */ + chunks = (struct sctp_auth_chunk_list *)(mtod(m, caddr_t) + chunk_len); + parameter_len = (uint16_t)sizeof(struct sctp_auth_chunk_list) + + sctp_serialize_auth_chunks(inp->sctp_ep.local_auth_chunks, + chunks->chunk_types); + chunks->ph.param_type = htons(SCTP_CHUNK_LIST); + chunks->ph.param_length = htons(parameter_len); + padding_len = SCTP_SIZE32(parameter_len) - parameter_len; + chunk_len += parameter_len; + } + SCTP_BUF_LEN(m) = chunk_len; + m_last = m; + /* now the addresses */ + /* To optimize this we could put the scoping stuff + * into a structure and remove the individual uint8's from + * the stc structure. Then we could just sifa in the + * address within the stc.. but for now this is a quick + * hack to get the address stuff teased apart. + */ + scp.ipv4_addr_legal = stc.ipv4_addr_legal; + scp.ipv6_addr_legal = stc.ipv6_addr_legal; +#if defined(__Userspace__) + scp.conn_addr_legal = stc.conn_addr_legal; +#endif + scp.loopback_scope = stc.loopback_scope; + scp.ipv4_local_scope = stc.ipv4_scope; + scp.local_scope = stc.local_scope; + scp.site_scope = stc.site_scope; + m_last = sctp_add_addresses_to_i_ia(inp, stcb, &scp, m_last, + cnt_inits_to, + &padding_len, &chunk_len); + /* padding_len can only be positive, if no addresses have been added */ + if (padding_len > 0) { + memset(mtod(m, caddr_t) + chunk_len, 0, padding_len); + chunk_len += padding_len; + SCTP_BUF_LEN(m) += padding_len; + padding_len = 0; + } + + /* tack on the operational error if present */ + if (op_err) { + parameter_len = 0; + for (m_tmp = op_err; m_tmp != NULL; m_tmp = SCTP_BUF_NEXT(m_tmp)) { + parameter_len += SCTP_BUF_LEN(m_tmp); + } + padding_len = SCTP_SIZE32(parameter_len) - parameter_len; + SCTP_BUF_NEXT(m_last) = op_err; + while (SCTP_BUF_NEXT(m_last) != NULL) { + m_last = SCTP_BUF_NEXT(m_last); + } + chunk_len += parameter_len; + } + if (padding_len > 0) { + m_last = sctp_add_pad_tombuf(m_last, padding_len); + if (m_last == NULL) { + /* Houston we have a problem, no space */ + sctp_m_freem(m); + return; + } + chunk_len += padding_len; + padding_len = 0; + } + /* Now we must build a cookie */ + m_cookie = sctp_add_cookie(init_pkt, offset, m, 0, &stc, &signature); + if (m_cookie == NULL) { + /* memory problem */ + sctp_m_freem(m); + return; + } + /* Now append the cookie to the end and update the space/size */ + SCTP_BUF_NEXT(m_last) = m_cookie; + parameter_len = 0; + for (m_tmp = m_cookie; m_tmp != NULL; m_tmp = SCTP_BUF_NEXT(m_tmp)) { + parameter_len += SCTP_BUF_LEN(m_tmp); + if (SCTP_BUF_NEXT(m_tmp) == NULL) { + m_last = m_tmp; + } + } + padding_len = SCTP_SIZE32(parameter_len) - parameter_len; + chunk_len += parameter_len; + + /* Place in the size, but we don't include + * the last pad (if any) in the INIT-ACK. + */ + initack->ch.chunk_length = htons(chunk_len); + + /* Time to sign the cookie, we don't sign over the cookie + * signature though thus we set trailer. + */ + (void)sctp_hmac_m(SCTP_HMAC, + (uint8_t *)inp->sctp_ep.secret_key[(int)(inp->sctp_ep.current_secret_number)], + SCTP_SECRET_SIZE, m_cookie, sizeof(struct sctp_paramhdr), + (uint8_t *)signature, SCTP_SIGNATURE_SIZE); +#if defined(__Userspace__) + /* + * Don't put AF_CONN addresses on the wire, in case this is critical + * for the application. However, they are protected by the HMAC and + * need to be reconstructed before checking the HMAC. + * Clearing is only done in the mbuf chain, since the local stc is + * not used anymore. + */ + if (stc.addr_type == SCTP_CONN_ADDRESS) { + const void *p = NULL; + + m_copyback(m_cookie, sizeof(struct sctp_paramhdr) + offsetof(struct sctp_state_cookie, address), + (int)sizeof(void *), (caddr_t)&p); + } + if (stc.laddr_type == SCTP_CONN_ADDRESS) { + const void *p = NULL; + + m_copyback(m_cookie, sizeof(struct sctp_paramhdr) + offsetof(struct sctp_state_cookie, laddress), + (int)sizeof(void *), (caddr_t)&p); + } +#endif + /* + * We sifa 0 here to NOT set IP_DF if its IPv4, we ignore the return + * here since the timer will drive a retranmission. + */ + if (padding_len > 0) { + if (sctp_add_pad_tombuf(m_last, padding_len) == NULL) { + sctp_m_freem(m); + return; + } + } + if (stc.loopback_scope) { + over_addr = (union sctp_sockstore *)dst; + } else { + over_addr = NULL; + } + + if ((error = sctp_lowlevel_chunk_output(inp, NULL, NULL, to, m, 0, NULL, 0, 0, + 0, 0, + inp->sctp_lport, sh->src_port, init_chk->init.initiate_tag, + port, over_addr, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, +#endif + false, /* XXXMT: Improve this! */ + SCTP_SO_NOT_LOCKED))) { + SCTPDBG(SCTP_DEBUG_OUTPUT4, "Gak send error %d\n", error); + if (error == ENOBUFS) { + if (asoc != NULL) { + asoc->ifp_had_enobuf = 1; + } + SCTP_STAT_INCR(sctps_lowlevelerr); + } + } else { + if (asoc != NULL) { + asoc->ifp_had_enobuf = 0; + } + } + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); +} + +static void +sctp_prune_prsctp(struct sctp_tcb *stcb, + struct sctp_association *asoc, + struct sctp_nonpad_sndrcvinfo *srcv, + int dataout) +{ + int freed_spc = 0; + struct sctp_tmit_chunk *chk, *nchk; + + SCTP_TCB_LOCK_ASSERT(stcb); + if ((asoc->prsctp_supported) && + (asoc->sent_queue_cnt_removeable > 0)) { + TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { + /* + * Look for chunks marked with the PR_SCTP flag AND + * the buffer space flag. If the one being sent is + * equal or greater priority then purge the old one + * and free some space. + */ + if (PR_SCTP_BUF_ENABLED(chk->flags)) { + /* + * This one is PR-SCTP AND buffer space + * limited type + */ + if (chk->rec.data.timetodrop.tv_sec > (long)srcv->sinfo_timetolive) { + /* + * Lower numbers equates to higher + * priority. So if the one we are + * looking at has a larger priority, + * we want to drop the data and NOT + * retransmit it. + */ + if (chk->data) { + /* + * We release the book_size + * if the mbuf is here + */ + int ret_spc; + uint8_t sent; + + if (chk->sent > SCTP_DATAGRAM_UNSENT) + sent = 1; + else + sent = 0; + ret_spc = sctp_release_pr_sctp_chunk(stcb, chk, + sent, + SCTP_SO_LOCKED); + freed_spc += ret_spc; + if (freed_spc >= dataout) { + return; + } + } /* if chunk was present */ + } /* if of sufficient priority */ + } /* if chunk has enabled */ + } /* tailqforeach */ + + TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) { + /* Here we must move to the sent queue and mark */ + if (PR_SCTP_BUF_ENABLED(chk->flags)) { + if (chk->rec.data.timetodrop.tv_sec > (long)srcv->sinfo_timetolive) { + if (chk->data) { + /* + * We release the book_size + * if the mbuf is here + */ + int ret_spc; + + ret_spc = sctp_release_pr_sctp_chunk(stcb, chk, + 0, SCTP_SO_LOCKED); + + freed_spc += ret_spc; + if (freed_spc >= dataout) { + return; + } + } /* end if chk->data */ + } /* end if right class */ + } /* end if chk pr-sctp */ + } /* tailqforeachsafe (chk) */ + } /* if enabled in asoc */ +} + +uint32_t +sctp_get_frag_point(struct sctp_tcb *stcb) +{ + struct sctp_association *asoc; + uint32_t frag_point, overhead; + + asoc = &stcb->asoc; + /* Consider IP header and SCTP common header. */ + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + overhead = SCTP_MIN_OVERHEAD; + } else { +#if defined(__Userspace__) + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_CONN) { + overhead = sizeof(struct sctphdr); + } else { + overhead = SCTP_MIN_V4_OVERHEAD; + } +#else + overhead = SCTP_MIN_V4_OVERHEAD; +#endif + } + /* Consider DATA/IDATA chunk header and AUTH header, if needed. */ + if (asoc->idata_supported) { + overhead += sizeof(struct sctp_idata_chunk); + if (sctp_auth_is_required_chunk(SCTP_IDATA, asoc->peer_auth_chunks)) { + overhead += sctp_get_auth_chunk_len(asoc->peer_hmac_id); + } + } else { + overhead += sizeof(struct sctp_data_chunk); + if (sctp_auth_is_required_chunk(SCTP_DATA, asoc->peer_auth_chunks)) { + overhead += sctp_get_auth_chunk_len(asoc->peer_hmac_id); + } + } + KASSERT(overhead % 4 == 0, + ("overhead (%u) not a multiple of 4", overhead)); + /* Consider padding. */ + if (asoc->smallest_mtu % 4 > 0) { + overhead += (asoc->smallest_mtu % 4); + } + KASSERT(asoc->smallest_mtu > overhead, + ("Association MTU (%u) too small for overhead (%u)", + asoc->smallest_mtu, overhead)); + frag_point = asoc->smallest_mtu - overhead; + KASSERT(frag_point % 4 == 0, + ("frag_point (%u) not a multiple of 4", frag_point)); + /* Honor MAXSEG socket option. */ + if ((asoc->sctp_frag_point > 0) && + (asoc->sctp_frag_point < frag_point)) { + frag_point = asoc->sctp_frag_point; + } + return (frag_point); +} + +static void +sctp_set_prsctp_policy(struct sctp_stream_queue_pending *sp) +{ + /* + * We assume that the user wants PR_SCTP_TTL if the user + * provides a positive lifetime but does not specify any + * PR_SCTP policy. + */ + if (PR_SCTP_ENABLED(sp->sinfo_flags)) { + sp->act_flags |= PR_SCTP_POLICY(sp->sinfo_flags); + } else if (sp->timetolive > 0) { + sp->sinfo_flags |= SCTP_PR_SCTP_TTL; + sp->act_flags |= PR_SCTP_POLICY(sp->sinfo_flags); + } else { + return; + } + switch (PR_SCTP_POLICY(sp->sinfo_flags)) { + case CHUNK_FLAGS_PR_SCTP_BUF: + /* + * Time to live is a priority stored in tv_sec when + * doing the buffer drop thing. + */ + sp->ts.tv_sec = sp->timetolive; + sp->ts.tv_usec = 0; + break; + case CHUNK_FLAGS_PR_SCTP_TTL: + { + struct timeval tv; + (void)SCTP_GETTIME_TIMEVAL(&sp->ts); + tv.tv_sec = sp->timetolive / 1000; + tv.tv_usec = (sp->timetolive * 1000) % 1000000; + /* TODO sctp_constants.h needs alternative time macros when + * _KERNEL is undefined. + */ +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + timeradd(&sp->ts, &tv, &sp->ts); +#else + timevaladd(&sp->ts, &tv); +#endif + } + break; + case CHUNK_FLAGS_PR_SCTP_RTX: + /* + * Time to live is a the number or retransmissions + * stored in tv_sec. + */ + sp->ts.tv_sec = sp->timetolive; + sp->ts.tv_usec = 0; + break; + default: + SCTPDBG(SCTP_DEBUG_USRREQ1, + "Unknown PR_SCTP policy %u.\n", + PR_SCTP_POLICY(sp->sinfo_flags)); + break; + } +} + +static int +sctp_msg_append(struct sctp_tcb *stcb, + struct sctp_nets *net, + struct mbuf *m, + struct sctp_nonpad_sndrcvinfo *srcv) +{ + int error = 0; + struct mbuf *at; + struct sctp_stream_queue_pending *sp = NULL; + struct sctp_stream_out *strm; + + SCTP_TCB_LOCK_ASSERT(stcb); + + /* Given an mbuf chain, put it + * into the association send queue and + * place it on the wheel + */ + if (srcv->sinfo_stream >= stcb->asoc.streamoutcnt) { + /* Invalid stream number */ + SCTP_LTRACE_ERR_RET_PKT(m, NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); + error = EINVAL; + goto out_now; + } + if ((stcb->asoc.stream_locked) && + (stcb->asoc.stream_locked_on != srcv->sinfo_stream)) { + SCTP_LTRACE_ERR_RET_PKT(m, NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); + error = EINVAL; + goto out_now; + } + if ((stcb->asoc.strmout[srcv->sinfo_stream].state != SCTP_STREAM_OPEN) && + (stcb->asoc.strmout[srcv->sinfo_stream].state != SCTP_STREAM_OPENING)) { + /* + * Can't queue any data while stream reset is underway. + */ + if (stcb->asoc.strmout[srcv->sinfo_stream].state > SCTP_STREAM_OPEN) { + error = EAGAIN; + } else { + error = EINVAL; + } + goto out_now; + } + /* Now can we send this? */ + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_SENT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_ACK_SENT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED) || + (stcb->asoc.state & SCTP_STATE_SHUTDOWN_PENDING)) { + /* got data while shutting down */ + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EPIPE); + error = EPIPE; + goto out_now; + } + sctp_alloc_a_strmoq(stcb, sp); + if (sp == NULL) { + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + error = ENOMEM; + goto out_now; + } + sp->sinfo_flags = srcv->sinfo_flags; + sp->timetolive = srcv->sinfo_timetolive; + sp->ppid = srcv->sinfo_ppid; + sp->context = srcv->sinfo_context; + sp->fsn = 0; + if (sp->sinfo_flags & SCTP_ADDR_OVER) { + sp->net = net; + atomic_add_int(&sp->net->ref_count, 1); + } else { + sp->net = NULL; + } + (void)SCTP_GETTIME_TIMEVAL(&sp->ts); + sp->sid = srcv->sinfo_stream; + sp->msg_is_complete = 1; + sp->sender_all_done = 1; + sp->some_taken = 0; + sp->data = m; + sp->tail_mbuf = NULL; + sctp_set_prsctp_policy(sp); + /* We could in theory (for sendall) sifa the length + * in, but we would still have to hunt through the + * chain since we need to setup the tail_mbuf + */ + sp->length = 0; + for (at = m; at; at = SCTP_BUF_NEXT(at)) { + if (SCTP_BUF_NEXT(at) == NULL) + sp->tail_mbuf = at; + sp->length += SCTP_BUF_LEN(at); + } + if (srcv->sinfo_keynumber_valid) { + sp->auth_keyid = srcv->sinfo_keynumber; + } else { + sp->auth_keyid = stcb->asoc.authinfo.active_keyid; + } + if (sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks)) { + sctp_auth_key_acquire(stcb, sp->auth_keyid); + sp->holds_key_ref = 1; + } + strm = &stcb->asoc.strmout[srcv->sinfo_stream]; + sctp_snd_sb_alloc(stcb, sp->length); + atomic_add_int(&stcb->asoc.stream_queue_cnt, 1); + TAILQ_INSERT_TAIL(&strm->outqueue, sp, next); + stcb->asoc.ss_functions.sctp_ss_add_to_stream(stcb, &stcb->asoc, strm, sp); + m = NULL; +out_now: + if (m) { + sctp_m_freem(m); + } + return (error); +} + +static struct mbuf * +sctp_copy_mbufchain(struct mbuf *clonechain, + struct mbuf *outchain, + struct mbuf **endofchain, + int can_take_mbuf, + int sizeofcpy, + uint8_t copy_by_ref) +{ + struct mbuf *m; + struct mbuf *appendchain; + caddr_t cp; + int len; + + if (endofchain == NULL) { + /* error */ + error_out: + if (outchain) + sctp_m_freem(outchain); + return (NULL); + } + if (can_take_mbuf) { + appendchain = clonechain; + } else { + if (!copy_by_ref && + (sizeofcpy <= (int)((((SCTP_BASE_SYSCTL(sctp_mbuf_threshold_count) - 1) * MLEN) + MHLEN)))) { + /* Its not in a cluster */ + if (*endofchain == NULL) { + /* lets get a mbuf cluster */ + if (outchain == NULL) { + /* This is the general case */ + new_mbuf: + outchain = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_NOWAIT, 1, MT_HEADER); + if (outchain == NULL) { + goto error_out; + } + SCTP_BUF_LEN(outchain) = 0; + *endofchain = outchain; + /* get the prepend space */ + SCTP_BUF_RESV_UF(outchain, (SCTP_FIRST_MBUF_RESV+4)); + } else { + /* We really should not get a NULL in endofchain */ + /* find end */ + m = outchain; + while (m) { + if (SCTP_BUF_NEXT(m) == NULL) { + *endofchain = m; + break; + } + m = SCTP_BUF_NEXT(m); + } + /* sanity */ + if (*endofchain == NULL) { + /* huh, TSNH XXX maybe we should panic */ + sctp_m_freem(outchain); + goto new_mbuf; + } + } + /* get the new end of length */ + len = (int)M_TRAILINGSPACE(*endofchain); + } else { + /* how much is left at the end? */ + len = (int)M_TRAILINGSPACE(*endofchain); + } + /* Find the end of the data, for appending */ + cp = (mtod((*endofchain), caddr_t) + SCTP_BUF_LEN((*endofchain))); + + /* Now lets copy it out */ + if (len >= sizeofcpy) { + /* It all fits, copy it in */ + m_copydata(clonechain, 0, sizeofcpy, cp); + SCTP_BUF_LEN((*endofchain)) += sizeofcpy; + } else { + /* fill up the end of the chain */ + if (len > 0) { + m_copydata(clonechain, 0, len, cp); + SCTP_BUF_LEN((*endofchain)) += len; + /* now we need another one */ + sizeofcpy -= len; + } + m = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_NOWAIT, 1, MT_HEADER); + if (m == NULL) { + /* We failed */ + goto error_out; + } + SCTP_BUF_NEXT((*endofchain)) = m; + *endofchain = m; + cp = mtod((*endofchain), caddr_t); + m_copydata(clonechain, len, sizeofcpy, cp); + SCTP_BUF_LEN((*endofchain)) += sizeofcpy; + } + return (outchain); + } else { + /* copy the old fashion way */ + appendchain = SCTP_M_COPYM(clonechain, 0, M_COPYALL, M_NOWAIT); +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(appendchain, SCTP_MBUF_ICOPY); + } +#endif + } + } + if (appendchain == NULL) { + /* error */ + if (outchain) + sctp_m_freem(outchain); + return (NULL); + } + if (outchain) { + /* tack on to the end */ + if (*endofchain != NULL) { + SCTP_BUF_NEXT(((*endofchain))) = appendchain; + } else { + m = outchain; + while (m) { + if (SCTP_BUF_NEXT(m) == NULL) { + SCTP_BUF_NEXT(m) = appendchain; + break; + } + m = SCTP_BUF_NEXT(m); + } + } + /* + * save off the end and update the end-chain + * position + */ + m = appendchain; + while (m) { + if (SCTP_BUF_NEXT(m) == NULL) { + *endofchain = m; + break; + } + m = SCTP_BUF_NEXT(m); + } + return (outchain); + } else { + /* save off the end and update the end-chain position */ + m = appendchain; + while (m) { + if (SCTP_BUF_NEXT(m) == NULL) { + *endofchain = m; + break; + } + m = SCTP_BUF_NEXT(m); + } + return (appendchain); + } +} + +static int +sctp_med_chunk_output(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_association *asoc, + int *num_out, + int *reason_code, + int control_only, int from_where, + struct timeval *now, int *now_filled, + uint32_t frag_point, int so_locked); + +static void +sctp_sendall_iterator(struct sctp_inpcb *inp, struct sctp_tcb *stcb, void *ptr, + uint32_t val SCTP_UNUSED) +{ + struct sctp_copy_all *ca; + struct mbuf *m; + int ret = 0; + int added_control = 0; + int un_sent, do_chunk_output = 1; + struct sctp_association *asoc; + struct sctp_nets *net; + + ca = (struct sctp_copy_all *)ptr; + if (ca->m == NULL) { + return; + } + if (ca->inp != inp) { + /* TSNH */ + return; + } + if (ca->sndlen > 0) { + m = SCTP_M_COPYM(ca->m, 0, M_COPYALL, M_NOWAIT); + if (m == NULL) { + /* can't copy so we are done */ + ca->cnt_failed++; + return; + } +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(m, SCTP_MBUF_ICOPY); + } +#endif + } else { + m = NULL; + } + SCTP_TCB_LOCK_ASSERT(stcb); + if (stcb->asoc.alternate) { + net = stcb->asoc.alternate; + } else { + net = stcb->asoc.primary_destination; + } + if (ca->sndrcv.sinfo_flags & SCTP_ABORT) { + /* Abort this assoc with m as the user defined reason */ + if (m != NULL) { + SCTP_BUF_PREPEND(m, sizeof(struct sctp_paramhdr), M_NOWAIT); + } else { + m = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr), + 0, M_NOWAIT, 1, MT_DATA); + SCTP_BUF_LEN(m) = sizeof(struct sctp_paramhdr); + } + if (m != NULL) { + struct sctp_paramhdr *ph; + + ph = mtod(m, struct sctp_paramhdr *); + ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT); + ph->param_length = htons((uint16_t)(sizeof(struct sctp_paramhdr) + ca->sndlen)); + } + /* We add one here to keep the assoc from + * dis-appearing on us. + */ + atomic_add_int(&stcb->asoc.refcnt, 1); + sctp_abort_an_association(inp, stcb, m, false, SCTP_SO_NOT_LOCKED); + /* sctp_abort_an_association calls sctp_free_asoc() + * free association will NOT free it since we + * incremented the refcnt .. we do this to prevent + * it being freed and things getting tricky since + * we could end up (from free_asoc) calling inpcb_free + * which would get a recursive lock call to the + * iterator lock.. But as a consequence of that the + * stcb will return to us un-locked.. since free_asoc + * returns with either no TCB or the TCB unlocked, we + * must relock.. to unlock in the iterator timer :-0 + */ + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + goto no_chunk_output; + } else { + if (m != NULL) { + ret = sctp_msg_append(stcb, net, m, &ca->sndrcv); + } + asoc = &stcb->asoc; + if (ca->sndrcv.sinfo_flags & SCTP_EOF) { + /* shutdown this assoc */ + if (TAILQ_EMPTY(&asoc->send_queue) && + TAILQ_EMPTY(&asoc->sent_queue) && + sctp_is_there_unsent_data(stcb, SCTP_SO_NOT_LOCKED) == 0) { + if ((*asoc->ss_functions.sctp_ss_is_user_msgs_incomplete)(stcb, asoc)) { + goto abort_anyway; + } + /* there is nothing queued to send, so I'm done... */ + if ((SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_RECEIVED) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { + /* only send SHUTDOWN the first time through */ + if (SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_SENT); + sctp_stop_timers_for_shutdown(stcb); + sctp_send_shutdown(stcb, net); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, + net); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, + NULL); + added_control = 1; + do_chunk_output = 0; + } + } else { + /* + * we still got (or just got) data to send, so set + * SHUTDOWN_PENDING + */ + /* + * XXX sockets draft says that SCTP_EOF should be + * sent with no data. currently, we will allow user + * data to be sent first and move to + * SHUTDOWN-PENDING + */ + if ((SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_RECEIVED) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { + if ((*asoc->ss_functions.sctp_ss_is_user_msgs_incomplete)(stcb, asoc)) { + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_PARTIAL_MSG_LEFT); + } + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_SHUTDOWN_PENDING); + if (TAILQ_EMPTY(&asoc->send_queue) && + TAILQ_EMPTY(&asoc->sent_queue) && + (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + abort_anyway: + SCTP_SNPRINTF(msg, sizeof(msg), + "%s:%d at %s", __FILE__, __LINE__, __func__); + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + msg); + atomic_add_int(&stcb->asoc.refcnt, 1); + sctp_abort_an_association(stcb->sctp_ep, stcb, + op_err, false, SCTP_SO_NOT_LOCKED); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + goto no_chunk_output; + } + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, + NULL); + } + } + } + } + un_sent = ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) + + (stcb->asoc.stream_queue_cnt * SCTP_DATA_CHUNK_OVERHEAD(stcb))); + + if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) && + (stcb->asoc.total_flight > 0) && + (un_sent < (int)(stcb->asoc.smallest_mtu - SCTP_MIN_OVERHEAD))) { + do_chunk_output = 0; + } + if (do_chunk_output) + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_NOT_LOCKED); + else if (added_control) { + struct timeval now; + int num_out, reason, now_filled = 0; + + (void)sctp_med_chunk_output(inp, stcb, &stcb->asoc, &num_out, + &reason, 1, 1, &now, &now_filled, + sctp_get_frag_point(stcb), + SCTP_SO_NOT_LOCKED); + } + no_chunk_output: + if (ret) { + ca->cnt_failed++; + } else { + ca->cnt_sent++; + } +} + +static void +sctp_sendall_completes(void *ptr, uint32_t val SCTP_UNUSED) +{ + struct sctp_copy_all *ca; + + ca = (struct sctp_copy_all *)ptr; + /* + * Do a notify here? Kacheong suggests that the notify be done at + * the send time.. so you would push up a notification if any send + * failed. Don't know if this is feasible since the only failures we + * have is "memory" related and if you cannot get an mbuf to send + * the data you surely can't get an mbuf to send up to notify the + * user you can't send the data :-> + */ + + /* now free everything */ + if (ca->inp) { + /* Lets clear the flag to allow others to run. */ + SCTP_INP_WLOCK(ca->inp); + ca->inp->sctp_flags &= ~SCTP_PCB_FLAGS_SND_ITERATOR_UP; + SCTP_INP_WUNLOCK(ca->inp); + } + sctp_m_freem(ca->m); + SCTP_FREE(ca, SCTP_M_COPYAL); +} + +static struct mbuf * +sctp_copy_out_all(struct uio *uio, ssize_t len) +{ + struct mbuf *ret, *at; + ssize_t left, willcpy, cancpy, error; + + ret = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_WAITOK, 1, MT_DATA); + if (ret == NULL) { + /* TSNH */ + return (NULL); + } + left = len; + SCTP_BUF_LEN(ret) = 0; + /* save space for the data chunk header */ + cancpy = (int)M_TRAILINGSPACE(ret); + willcpy = min(cancpy, left); + at = ret; + while (left > 0) { + /* Align data to the end */ + error = uiomove(mtod(at, caddr_t), (int)willcpy, uio); + if (error) { + err_out_now: + sctp_m_freem(at); + return (NULL); + } + SCTP_BUF_LEN(at) = (int)willcpy; + SCTP_BUF_NEXT_PKT(at) = SCTP_BUF_NEXT(at) = 0; + left -= willcpy; + if (left > 0) { + SCTP_BUF_NEXT(at) = sctp_get_mbuf_for_msg((unsigned int)left, 0, M_WAITOK, 1, MT_DATA); + if (SCTP_BUF_NEXT(at) == NULL) { + goto err_out_now; + } + at = SCTP_BUF_NEXT(at); + SCTP_BUF_LEN(at) = 0; + cancpy = (int)M_TRAILINGSPACE(at); + willcpy = min(cancpy, left); + } + } + return (ret); +} + +static int +sctp_sendall(struct sctp_inpcb *inp, struct uio *uio, struct mbuf *m, + struct sctp_nonpad_sndrcvinfo *srcv) +{ + int ret; + struct sctp_copy_all *ca; + +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + if (uio->uio_resid > SCTP_BASE_SYSCTL(sctp_sendall_limit)) { +#else + if (uio_resid(uio) > SCTP_BASE_SYSCTL(sctp_sendall_limit)) { +#endif +#else + if (uio->uio_resid > (ssize_t)SCTP_BASE_SYSCTL(sctp_sendall_limit)) { +#endif + /* You must not be larger than the limit! */ + return (EMSGSIZE); + } + SCTP_MALLOC(ca, struct sctp_copy_all *, sizeof(struct sctp_copy_all), + SCTP_M_COPYAL); + if (ca == NULL) { + sctp_m_freem(m); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + memset(ca, 0, sizeof(struct sctp_copy_all)); + + ca->inp = inp; + if (srcv != NULL) { + memcpy(&ca->sndrcv, srcv, sizeof(struct sctp_nonpad_sndrcvinfo)); + } + + /* Serialize. */ + SCTP_INP_WLOCK(inp); + if ((inp->sctp_flags & SCTP_PCB_FLAGS_SND_ITERATOR_UP) != 0) { + SCTP_INP_WUNLOCK(inp); + sctp_m_freem(m); + SCTP_FREE(ca, SCTP_M_COPYAL); + return (EBUSY); + } + inp->sctp_flags |= SCTP_PCB_FLAGS_SND_ITERATOR_UP; + SCTP_INP_WUNLOCK(inp); + + /* + * take off the sendall flag, it would be bad if we failed to do + * this :-0 + */ + ca->sndrcv.sinfo_flags &= ~SCTP_SENDALL; + /* get length and mbuf chain */ + if (uio) { +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + ca->sndlen = uio->uio_resid; +#else + ca->sndlen = uio_resid(uio); +#endif +#else + ca->sndlen = uio->uio_resid; +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(SCTP_INP_SO(inp), 0); +#endif + ca->m = sctp_copy_out_all(uio, ca->sndlen); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_LOCK(SCTP_INP_SO(inp), 0); +#endif + if (ca->m == NULL) { + SCTP_FREE(ca, SCTP_M_COPYAL); + sctp_m_freem(m); + SCTP_INP_WLOCK(inp); + inp->sctp_flags &= ~SCTP_PCB_FLAGS_SND_ITERATOR_UP; + SCTP_INP_WUNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + } else { + /* Gather the length of the send */ + struct mbuf *mat; + + ca->sndlen = 0; + for (mat = m; mat; mat = SCTP_BUF_NEXT(mat)) { + ca->sndlen += SCTP_BUF_LEN(mat); + } + } + ret = sctp_initiate_iterator(NULL, sctp_sendall_iterator, NULL, + SCTP_PCB_ANY_FLAGS, SCTP_PCB_ANY_FEATURES, + SCTP_ASOC_ANY_STATE, + (void *)ca, 0, + sctp_sendall_completes, inp, 1); + if (ret) { + SCTP_INP_WLOCK(inp); + inp->sctp_flags &= ~SCTP_PCB_FLAGS_SND_ITERATOR_UP; + SCTP_INP_WUNLOCK(inp); + SCTP_FREE(ca, SCTP_M_COPYAL); + SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EFAULT); + return (EFAULT); + } + return (0); +} + +void +sctp_toss_old_cookies(struct sctp_tcb *stcb, struct sctp_association *asoc) +{ + struct sctp_tmit_chunk *chk, *nchk; + + TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) { + if (chk->rec.chunk_id.id == SCTP_COOKIE_ECHO) { + TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); + asoc->ctrl_queue_cnt--; + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + } + } +} + +void +sctp_toss_old_asconf(struct sctp_tcb *stcb) +{ + struct sctp_association *asoc; + struct sctp_tmit_chunk *chk, *nchk; + struct sctp_asconf_chunk *acp; + + asoc = &stcb->asoc; + TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) { + /* find SCTP_ASCONF chunk in queue */ + if (chk->rec.chunk_id.id == SCTP_ASCONF) { + if (chk->data) { + acp = mtod(chk->data, struct sctp_asconf_chunk *); + if (SCTP_TSN_GT(ntohl(acp->serial_number), asoc->asconf_seq_out_acked)) { + /* Not Acked yet */ + break; + } + } + TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next); + asoc->ctrl_queue_cnt--; + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + } + } +} + +static void +sctp_clean_up_datalist(struct sctp_tcb *stcb, + struct sctp_association *asoc, + struct sctp_tmit_chunk **data_list, + int bundle_at, + struct sctp_nets *net) +{ + int i; + struct sctp_tmit_chunk *tp1; + + for (i = 0; i < bundle_at; i++) { + /* off of the send queue */ + TAILQ_REMOVE(&asoc->send_queue, data_list[i], sctp_next); + asoc->send_queue_cnt--; + if (i > 0) { + /* + * Any chunk NOT 0 you zap the time chunk 0 gets + * zapped or set based on if a RTO measurement is + * needed. + */ + data_list[i]->do_rtt = 0; + } + /* record time */ + data_list[i]->sent_rcv_time = net->last_sent_time; + data_list[i]->rec.data.cwnd_at_send = net->cwnd; + data_list[i]->rec.data.fast_retran_tsn = data_list[i]->rec.data.tsn; + if (data_list[i]->whoTo == NULL) { + data_list[i]->whoTo = net; + atomic_add_int(&net->ref_count, 1); + } + /* on to the sent queue */ + tp1 = TAILQ_LAST(&asoc->sent_queue, sctpchunk_listhead); + if ((tp1) && SCTP_TSN_GT(tp1->rec.data.tsn, data_list[i]->rec.data.tsn)) { + struct sctp_tmit_chunk *tpp; + + /* need to move back */ + back_up_more: + tpp = TAILQ_PREV(tp1, sctpchunk_listhead, sctp_next); + if (tpp == NULL) { + TAILQ_INSERT_BEFORE(tp1, data_list[i], sctp_next); + goto all_done; + } + tp1 = tpp; + if (SCTP_TSN_GT(tp1->rec.data.tsn, data_list[i]->rec.data.tsn)) { + goto back_up_more; + } + TAILQ_INSERT_AFTER(&asoc->sent_queue, tp1, data_list[i], sctp_next); + } else { + TAILQ_INSERT_TAIL(&asoc->sent_queue, + data_list[i], + sctp_next); + } + all_done: + /* This does not lower until the cum-ack passes it */ + asoc->sent_queue_cnt++; + if ((asoc->peers_rwnd <= 0) && + (asoc->total_flight == 0) && + (bundle_at == 1)) { + /* Mark the chunk as being a window probe */ + SCTP_STAT_INCR(sctps_windowprobed); + } +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xC2, 3); +#endif + data_list[i]->sent = SCTP_DATAGRAM_SENT; + data_list[i]->snd_count = 1; + data_list[i]->rec.data.chunk_was_revoked = 0; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_UP, + data_list[i]->whoTo->flight_size, + data_list[i]->book_size, + (uint32_t)(uintptr_t)data_list[i]->whoTo, + data_list[i]->rec.data.tsn); + } + sctp_flight_size_increase(data_list[i]); + sctp_total_flight_increase(stcb, data_list[i]); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { + sctp_log_rwnd(SCTP_DECREASE_PEER_RWND, + asoc->peers_rwnd, data_list[i]->send_size, SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)); + } + asoc->peers_rwnd = sctp_sbspace_sub(asoc->peers_rwnd, + (uint32_t) (data_list[i]->send_size + SCTP_BASE_SYSCTL(sctp_peer_chunk_oh))); + if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { + /* SWS sender side engages */ + asoc->peers_rwnd = 0; + } + } + if (asoc->cc_functions.sctp_cwnd_update_packet_transmitted) { + (*asoc->cc_functions.sctp_cwnd_update_packet_transmitted)(stcb, net); + } +} + +static void +sctp_clean_up_ctl(struct sctp_tcb *stcb, struct sctp_association *asoc, int so_locked) +{ + struct sctp_tmit_chunk *chk, *nchk; + + TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) { + if ((chk->rec.chunk_id.id == SCTP_SELECTIVE_ACK) || + (chk->rec.chunk_id.id == SCTP_NR_SELECTIVE_ACK) || /* EY */ + (chk->rec.chunk_id.id == SCTP_HEARTBEAT_REQUEST) || + (chk->rec.chunk_id.id == SCTP_HEARTBEAT_ACK) || + (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN) || + (chk->rec.chunk_id.id == SCTP_SHUTDOWN) || + (chk->rec.chunk_id.id == SCTP_SHUTDOWN_ACK) || + (chk->rec.chunk_id.id == SCTP_OPERATION_ERROR) || + (chk->rec.chunk_id.id == SCTP_PACKET_DROPPED) || + (chk->rec.chunk_id.id == SCTP_COOKIE_ACK) || + (chk->rec.chunk_id.id == SCTP_ECN_CWR) || + (chk->rec.chunk_id.id == SCTP_ASCONF_ACK)) { + /* Stray chunks must be cleaned up */ + clean_up_anyway: + TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); + asoc->ctrl_queue_cnt--; + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + if (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN) { + asoc->fwd_tsn_cnt--; + } + sctp_free_a_chunk(stcb, chk, so_locked); + } else if (chk->rec.chunk_id.id == SCTP_STREAM_RESET) { + /* special handling, we must look into the param */ + if (chk != asoc->str_reset) { + goto clean_up_anyway; + } + } + } +} + +static uint32_t +sctp_can_we_split_this(struct sctp_tcb *stcb, uint32_t length, + uint32_t space_left, uint32_t frag_point, int eeor_on) +{ + /* Make a decision on if I should split a + * msg into multiple parts. This is only asked of + * incomplete messages. + */ + if (eeor_on) { + /* If we are doing EEOR we need to always send + * it if its the entire thing, since it might + * be all the guy is putting in the hopper. + */ + if (space_left >= length) { + /*- + * If we have data outstanding, + * we get another chance when the sack + * arrives to transmit - wait for more data + */ + if (stcb->asoc.total_flight == 0) { + /* If nothing is in flight, we zero + * the packet counter. + */ + return (length); + } + return (0); + + } else { + /* You can fill the rest */ + return (space_left); + } + } + /*- + * For those strange folk that make the send buffer + * smaller than our fragmentation point, we can't + * get a full msg in so we have to allow splitting. + */ + if (SCTP_SB_LIMIT_SND(stcb->sctp_socket) < frag_point) { + return (length); + } + if ((length <= space_left) || + ((length - space_left) < SCTP_BASE_SYSCTL(sctp_min_residual))) { + /* Sub-optimal residual don't split in non-eeor mode. */ + return (0); + } + /* If we reach here length is larger + * than the space_left. Do we wish to split + * it for the sake of packet putting together? + */ + if (space_left >= min(SCTP_BASE_SYSCTL(sctp_min_split_point), frag_point)) { + /* Its ok to split it */ + return (min(space_left, frag_point)); + } + /* Nope, can't split */ + return (0); +} + +static uint32_t +sctp_move_to_outqueue(struct sctp_tcb *stcb, + struct sctp_nets *net, + struct sctp_stream_out *strq, + uint32_t space_left, + uint32_t frag_point, + int *giveup, + int eeor_mode, + int *bail, + int so_locked) +{ + /* Move from the stream to the send_queue keeping track of the total */ + struct sctp_association *asoc; + struct sctp_stream_queue_pending *sp; + struct sctp_tmit_chunk *chk; + struct sctp_data_chunk *dchkh=NULL; + struct sctp_idata_chunk *ndchkh=NULL; + uint32_t to_move, length; + int leading; + uint8_t rcv_flags = 0; + uint8_t some_taken; + + SCTP_TCB_LOCK_ASSERT(stcb); + asoc = &stcb->asoc; +one_more_time: + /*sa_ignore FREED_MEMORY*/ + sp = TAILQ_FIRST(&strq->outqueue); + if (sp == NULL) { + sp = TAILQ_FIRST(&strq->outqueue); + if (sp) { + goto one_more_time; + } + if ((sctp_is_feature_on(stcb->sctp_ep, SCTP_PCB_FLAGS_EXPLICIT_EOR) == 0) && + (stcb->asoc.idata_supported == 0) && + (strq->last_msg_incomplete)) { + SCTP_PRINTF("Huh? Stream:%d lm_in_c=%d but queue is NULL\n", + strq->sid, + strq->last_msg_incomplete); + strq->last_msg_incomplete = 0; + } + to_move = 0; + goto out_of; + } + if ((sp->msg_is_complete) && (sp->length == 0)) { + if (sp->sender_all_done) { + /* We are doing deferred cleanup. Last + * time through when we took all the data + * the sender_all_done was not set. + */ + if ((sp->put_last_out == 0) && (sp->discard_rest == 0)) { + SCTP_PRINTF("Gak, put out entire msg with NO end!-1\n"); + SCTP_PRINTF("sender_done:%d len:%d msg_comp:%d put_last_out:%d\n", + sp->sender_all_done, + sp->length, + sp->msg_is_complete, + sp->put_last_out); + } + atomic_subtract_int(&asoc->stream_queue_cnt, 1); + TAILQ_REMOVE(&strq->outqueue, sp, next); + stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, strq, sp); + if ((strq->state == SCTP_STREAM_RESET_PENDING) && + (strq->chunks_on_queues == 0) && + TAILQ_EMPTY(&strq->outqueue)) { + stcb->asoc.trigger_reset = 1; + } + if (sp->net) { + sctp_free_remote_addr(sp->net); + sp->net = NULL; + } + if (sp->data) { + sctp_m_freem(sp->data); + sp->data = NULL; + } + sctp_free_a_strmoq(stcb, sp, so_locked); + /* back to get the next msg */ + goto one_more_time; + } else { + /* sender just finished this but + * still holds a reference + */ + *giveup = 1; + to_move = 0; + goto out_of; + } + } else { + /* is there some to get */ + if (sp->length == 0) { + /* no */ + *giveup = 1; + to_move = 0; + goto out_of; + } else if (sp->discard_rest) { + /* Whack down the size */ + atomic_subtract_int(&stcb->asoc.total_output_queue_size, sp->length); + if ((stcb->sctp_socket != NULL) && + ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL))) { + atomic_subtract_int(&stcb->sctp_socket->so_snd.sb_cc, sp->length); + } + if (sp->data) { + sctp_m_freem(sp->data); + sp->data = NULL; + sp->tail_mbuf = NULL; + } + sp->length = 0; + sp->some_taken = 1; + *giveup = 1; + to_move = 0; + goto out_of; + } + } + some_taken = sp->some_taken; + length = sp->length; + if (sp->msg_is_complete) { + /* The message is complete */ + to_move = min(length, frag_point); + if (to_move == length) { + /* All of it fits in the MTU */ + if (sp->some_taken) { + rcv_flags |= SCTP_DATA_LAST_FRAG; + } else { + rcv_flags |= SCTP_DATA_NOT_FRAG; + } + sp->put_last_out = 1; + if (sp->sinfo_flags & SCTP_SACK_IMMEDIATELY) { + rcv_flags |= SCTP_DATA_SACK_IMMEDIATELY; + } + } else { + /* Not all of it fits, we fragment */ + if (sp->some_taken == 0) { + rcv_flags |= SCTP_DATA_FIRST_FRAG; + } + sp->some_taken = 1; + } + } else { + to_move = sctp_can_we_split_this(stcb, length, space_left, frag_point, eeor_mode); + if (to_move > 0) { + if (to_move >= length) { + to_move = length; + } + if (sp->some_taken == 0) { + rcv_flags |= SCTP_DATA_FIRST_FRAG; + sp->some_taken = 1; + } + } else { + /* Nothing to take. */ + *giveup = 1; + to_move = 0; + goto out_of; + } + } + + /* If we reach here, we can copy out a chunk */ + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + /* No chunk memory */ + *giveup = 1; + to_move = 0; + goto out_of; + } + /* Setup for unordered if needed by looking + * at the user sent info flags. + */ + if (sp->sinfo_flags & SCTP_UNORDERED) { + rcv_flags |= SCTP_DATA_UNORDERED; + } + if (SCTP_BASE_SYSCTL(sctp_enable_sack_immediately) && + (sp->sinfo_flags & SCTP_EOF) == SCTP_EOF) { + rcv_flags |= SCTP_DATA_SACK_IMMEDIATELY; + } + /* clear out the chunk before setting up */ + memset(chk, 0, sizeof(*chk)); + chk->rec.data.rcv_flags = rcv_flags; + + if (to_move >= length) { + /* we think we can steal the whole thing */ + if (to_move < sp->length) { + /* bail, it changed */ + goto dont_do_it; + } + chk->data = sp->data; + chk->last_mbuf = sp->tail_mbuf; + /* register the stealing */ + sp->data = sp->tail_mbuf = NULL; + } else { + struct mbuf *m; + dont_do_it: + chk->data = SCTP_M_COPYM(sp->data, 0, to_move, M_NOWAIT); + chk->last_mbuf = NULL; + if (chk->data == NULL) { + sp->some_taken = some_taken; + sctp_free_a_chunk(stcb, chk, so_locked); + *bail = 1; + to_move = 0; + goto out_of; + } +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(chk->data, SCTP_MBUF_ICOPY); + } +#endif + /* Pull off the data */ + m_adj(sp->data, to_move); + /* Now lets work our way down and compact it */ + m = sp->data; + while (m && (SCTP_BUF_LEN(m) == 0)) { + sp->data = SCTP_BUF_NEXT(m); + SCTP_BUF_NEXT(m) = NULL; + if (sp->tail_mbuf == m) { + /*- + * Freeing tail? TSNH since + * we supposedly were taking less + * than the sp->length. + */ +#ifdef INVARIANTS + panic("Huh, freeing tail? - TSNH"); +#else + SCTP_PRINTF("Huh, freeing tail? - TSNH\n"); + sp->tail_mbuf = sp->data = NULL; + sp->length = 0; +#endif + } + sctp_m_free(m); + m = sp->data; + } + } + if (SCTP_BUF_IS_EXTENDED(chk->data)) { + chk->copy_by_ref = 1; + } else { + chk->copy_by_ref = 0; + } + /* get last_mbuf and counts of mb usage + * This is ugly but hopefully its only one mbuf. + */ + if (chk->last_mbuf == NULL) { + chk->last_mbuf = chk->data; + while (SCTP_BUF_NEXT(chk->last_mbuf) != NULL) { + chk->last_mbuf = SCTP_BUF_NEXT(chk->last_mbuf); + } + } + + if (to_move > length) { + /*- This should not happen either + * since we always lower to_move to the size + * of sp->length if its larger. + */ +#ifdef INVARIANTS + panic("Huh, how can to_move be larger?"); +#else + SCTP_PRINTF("Huh, how can to_move be larger?\n"); + sp->length = 0; +#endif + } else { + atomic_subtract_int(&sp->length, to_move); + } + leading = SCTP_DATA_CHUNK_OVERHEAD(stcb); + if (M_LEADINGSPACE(chk->data) < leading) { + /* Not enough room for a chunk header, get some */ + struct mbuf *m; + + m = sctp_get_mbuf_for_msg(1, 0, M_NOWAIT, 1, MT_DATA); + if (m == NULL) { + /* + * we're in trouble here. _PREPEND below will free + * all the data if there is no leading space, so we + * must put the data back and restore. + */ + if (sp->data == NULL) { + /* unsteal the data */ + sp->data = chk->data; + sp->tail_mbuf = chk->last_mbuf; + } else { + struct mbuf *m_tmp; + /* reassemble the data */ + m_tmp = sp->data; + sp->data = chk->data; + SCTP_BUF_NEXT(chk->last_mbuf) = m_tmp; + } + sp->some_taken = some_taken; + atomic_add_int(&sp->length, to_move); + chk->data = NULL; + *bail = 1; + sctp_free_a_chunk(stcb, chk, so_locked); + to_move = 0; + goto out_of; + } else { + SCTP_BUF_LEN(m) = 0; + SCTP_BUF_NEXT(m) = chk->data; + chk->data = m; + M_ALIGN(chk->data, 4); + } + } + SCTP_BUF_PREPEND(chk->data, SCTP_DATA_CHUNK_OVERHEAD(stcb), M_NOWAIT); + if (chk->data == NULL) { + /* HELP, TSNH since we assured it would not above? */ +#ifdef INVARIANTS + panic("prepend fails HELP?"); +#else + SCTP_PRINTF("prepend fails HELP?\n"); + sctp_free_a_chunk(stcb, chk, so_locked); +#endif + *bail = 1; + to_move = 0; + goto out_of; + } + sctp_snd_sb_alloc(stcb, SCTP_DATA_CHUNK_OVERHEAD(stcb)); + chk->book_size = chk->send_size = (uint16_t)(to_move + SCTP_DATA_CHUNK_OVERHEAD(stcb)); + chk->book_size_scale = 0; + chk->sent = SCTP_DATAGRAM_UNSENT; + + chk->flags = 0; + chk->asoc = &stcb->asoc; + chk->pad_inplace = 0; + chk->no_fr_allowed = 0; + if (stcb->asoc.idata_supported == 0) { + if (rcv_flags & SCTP_DATA_UNORDERED) { + /* Just use 0. The receiver ignores the values. */ + chk->rec.data.mid = 0; + } else { + chk->rec.data.mid = strq->next_mid_ordered; + if (rcv_flags & SCTP_DATA_LAST_FRAG) { + strq->next_mid_ordered++; + } + } + } else { + if (rcv_flags & SCTP_DATA_UNORDERED) { + chk->rec.data.mid = strq->next_mid_unordered; + if (rcv_flags & SCTP_DATA_LAST_FRAG) { + strq->next_mid_unordered++; + } + } else { + chk->rec.data.mid = strq->next_mid_ordered; + if (rcv_flags & SCTP_DATA_LAST_FRAG) { + strq->next_mid_ordered++; + } + } + } + chk->rec.data.sid = sp->sid; + chk->rec.data.ppid = sp->ppid; + chk->rec.data.context = sp->context; + chk->rec.data.doing_fast_retransmit = 0; + + chk->rec.data.timetodrop = sp->ts; + chk->flags = sp->act_flags; + + if (sp->net) { + chk->whoTo = sp->net; + atomic_add_int(&chk->whoTo->ref_count, 1); + } else + chk->whoTo = NULL; + + if (sp->holds_key_ref) { + chk->auth_keyid = sp->auth_keyid; + sctp_auth_key_acquire(stcb, chk->auth_keyid); + chk->holds_key_ref = 1; + } + stcb->asoc.ss_functions.sctp_ss_scheduled(stcb, net, asoc, strq, to_move); +#if defined(__FreeBSD__) && !defined(__Userspace__) + chk->rec.data.tsn = atomic_fetchadd_int(&asoc->sending_seq, 1); +#else + chk->rec.data.tsn = asoc->sending_seq++; +#endif + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_AT_SEND_2_OUTQ) { + sctp_misc_ints(SCTP_STRMOUT_LOG_SEND, + (uint32_t)(uintptr_t)stcb, sp->length, + (uint32_t)((chk->rec.data.sid << 16) | (0x0000ffff & chk->rec.data.mid)), + chk->rec.data.tsn); + } + if (stcb->asoc.idata_supported == 0) { + dchkh = mtod(chk->data, struct sctp_data_chunk *); + } else { + ndchkh = mtod(chk->data, struct sctp_idata_chunk *); + } + /* + * Put the rest of the things in place now. Size was done + * earlier in previous loop prior to padding. + */ + + SCTP_TCB_LOCK_ASSERT(stcb); +#ifdef SCTP_ASOCLOG_OF_TSNS + if (asoc->tsn_out_at >= SCTP_TSN_LOG_SIZE) { + asoc->tsn_out_at = 0; + asoc->tsn_out_wrapped = 1; + } + asoc->out_tsnlog[asoc->tsn_out_at].tsn = chk->rec.data.tsn; + asoc->out_tsnlog[asoc->tsn_out_at].strm = chk->rec.data.sid; + asoc->out_tsnlog[asoc->tsn_out_at].seq = chk->rec.data.mid; + asoc->out_tsnlog[asoc->tsn_out_at].sz = chk->send_size; + asoc->out_tsnlog[asoc->tsn_out_at].flgs = chk->rec.data.rcv_flags; + asoc->out_tsnlog[asoc->tsn_out_at].stcb = (void *)stcb; + asoc->out_tsnlog[asoc->tsn_out_at].in_pos = asoc->tsn_out_at; + asoc->out_tsnlog[asoc->tsn_out_at].in_out = 2; + asoc->tsn_out_at++; +#endif + if (stcb->asoc.idata_supported == 0) { + dchkh->ch.chunk_type = SCTP_DATA; + dchkh->ch.chunk_flags = chk->rec.data.rcv_flags; + dchkh->dp.tsn = htonl(chk->rec.data.tsn); + dchkh->dp.sid = htons(strq->sid); + dchkh->dp.ssn = htons((uint16_t)chk->rec.data.mid); + dchkh->dp.ppid = chk->rec.data.ppid; + dchkh->ch.chunk_length = htons(chk->send_size); + } else { + ndchkh->ch.chunk_type = SCTP_IDATA; + ndchkh->ch.chunk_flags = chk->rec.data.rcv_flags; + ndchkh->dp.tsn = htonl(chk->rec.data.tsn); + ndchkh->dp.sid = htons(strq->sid); + ndchkh->dp.reserved = htons(0); + ndchkh->dp.mid = htonl(chk->rec.data.mid); + if (sp->fsn == 0) + ndchkh->dp.ppid_fsn.ppid = chk->rec.data.ppid; + else + ndchkh->dp.ppid_fsn.fsn = htonl(sp->fsn); + sp->fsn++; + ndchkh->ch.chunk_length = htons(chk->send_size); + } + /* Now advance the chk->send_size by the actual pad needed. */ + if (chk->send_size < SCTP_SIZE32(chk->book_size)) { + /* need a pad */ + struct mbuf *lm; + int pads; + + pads = SCTP_SIZE32(chk->book_size) - chk->send_size; + lm = sctp_pad_lastmbuf(chk->data, pads, chk->last_mbuf); + if (lm != NULL) { + chk->last_mbuf = lm; + chk->pad_inplace = 1; + } + chk->send_size += pads; + } + if (PR_SCTP_ENABLED(chk->flags)) { + asoc->pr_sctp_cnt++; + } + if (sp->msg_is_complete && (sp->length == 0) && (sp->sender_all_done)) { + /* All done pull and kill the message */ + if (sp->put_last_out == 0) { + SCTP_PRINTF("Gak, put out entire msg with NO end!-2\n"); + SCTP_PRINTF("sender_done:%d len:%d msg_comp:%d put_last_out:%d\n", + sp->sender_all_done, + sp->length, + sp->msg_is_complete, + sp->put_last_out); + } + atomic_subtract_int(&asoc->stream_queue_cnt, 1); + TAILQ_REMOVE(&strq->outqueue, sp, next); + stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, strq, sp); + if ((strq->state == SCTP_STREAM_RESET_PENDING) && + (strq->chunks_on_queues == 0) && + TAILQ_EMPTY(&strq->outqueue)) { + stcb->asoc.trigger_reset = 1; + } + if (sp->net) { + sctp_free_remote_addr(sp->net); + sp->net = NULL; + } + if (sp->data) { + sctp_m_freem(sp->data); + sp->data = NULL; + } + sctp_free_a_strmoq(stcb, sp, so_locked); + } + asoc->chunks_on_out_queue++; + strq->chunks_on_queues++; + TAILQ_INSERT_TAIL(&asoc->send_queue, chk, sctp_next); + asoc->send_queue_cnt++; +out_of: + return (to_move); +} + +static void +sctp_fill_outqueue(struct sctp_tcb *stcb, struct sctp_nets *net, + uint32_t frag_point, int eeor_mode, int *quit_now, + int so_locked) +{ + struct sctp_association *asoc; + struct sctp_stream_out *strq; + uint32_t space_left, moved, total_moved; + int bail, giveup; + + SCTP_TCB_LOCK_ASSERT(stcb); + asoc = &stcb->asoc; + total_moved = 0; + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + space_left = net->mtu - SCTP_MIN_V4_OVERHEAD; + break; +#endif +#ifdef INET6 + case AF_INET6: + space_left = net->mtu - SCTP_MIN_OVERHEAD; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + space_left = net->mtu - sizeof(struct sctphdr); + break; +#endif + default: + /* TSNH */ + space_left = net->mtu; + break; + } + /* Need an allowance for the data chunk header too */ + space_left -= SCTP_DATA_CHUNK_OVERHEAD(stcb); + + /* must make even word boundary */ + space_left &= 0xfffffffc; + strq = stcb->asoc.ss_functions.sctp_ss_select_stream(stcb, net, asoc); + giveup = 0; + bail = 0; + while ((space_left > 0) && (strq != NULL)) { + moved = sctp_move_to_outqueue(stcb, net, strq, space_left, + frag_point, &giveup, eeor_mode, + &bail, so_locked); + if ((giveup != 0) || (bail != 0)) { + break; + } + strq = stcb->asoc.ss_functions.sctp_ss_select_stream(stcb, net, asoc); + total_moved += moved; + if (space_left >= moved) { + space_left -= moved; + } else { + space_left = 0; + } + if (space_left >= SCTP_DATA_CHUNK_OVERHEAD(stcb)) { + space_left -= SCTP_DATA_CHUNK_OVERHEAD(stcb); + } else { + space_left = 0; + } + space_left &= 0xfffffffc; + } + if (bail != 0) + *quit_now = 1; + + stcb->asoc.ss_functions.sctp_ss_packet_done(stcb, net, asoc); + + if (total_moved == 0) { + if ((stcb->asoc.sctp_cmt_on_off == 0) && + (net == stcb->asoc.primary_destination)) { + /* ran dry for primary network net */ + SCTP_STAT_INCR(sctps_primary_randry); + } else if (stcb->asoc.sctp_cmt_on_off > 0) { + /* ran dry with CMT on */ + SCTP_STAT_INCR(sctps_cmt_randry); + } + } +} + +void +sctp_fix_ecn_echo(struct sctp_association *asoc) +{ + struct sctp_tmit_chunk *chk; + + TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { + if (chk->rec.chunk_id.id == SCTP_ECN_ECHO) { + chk->sent = SCTP_DATAGRAM_UNSENT; + } + } +} + +void +sctp_move_chunks_from_net(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + struct sctp_association *asoc; + struct sctp_tmit_chunk *chk; + struct sctp_stream_queue_pending *sp; + unsigned int i; + + if (net == NULL) { + return; + } + asoc = &stcb->asoc; + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) { + if (sp->net == net) { + sctp_free_remote_addr(sp->net); + sp->net = NULL; + } + } + } + TAILQ_FOREACH(chk, &asoc->send_queue, sctp_next) { + if (chk->whoTo == net) { + sctp_free_remote_addr(chk->whoTo); + chk->whoTo = NULL; + } + } +} + +int +sctp_med_chunk_output(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_association *asoc, + int *num_out, + int *reason_code, + int control_only, int from_where, + struct timeval *now, int *now_filled, + uint32_t frag_point, int so_locked) +{ + /** + * Ok this is the generic chunk service queue. we must do the + * following: + * - Service the stream queue that is next, moving any + * message (note I must get a complete message i.e. FIRST/MIDDLE and + * LAST to the out queue in one pass) and assigning TSN's. This + * only applies though if the peer does not support NDATA. For NDATA + * chunks its ok to not send the entire message ;-) + * - Check to see if the cwnd/rwnd allows any output, if so we go ahead and + * formulate and send the low level chunks. Making sure to combine + * any control in the control chunk queue also. + */ + struct sctp_nets *net, *start_at, *sack_goes_to = NULL, *old_start_at = NULL; + struct mbuf *outchain, *endoutchain; + struct sctp_tmit_chunk *chk, *nchk; + + /* temp arrays for unlinking */ + struct sctp_tmit_chunk *data_list[SCTP_MAX_DATA_BUNDLING]; + int no_fragmentflg, error; + unsigned int max_rwnd_per_dest, max_send_per_dest; + int one_chunk, hbflag, skip_data_for_this_net; + int asconf, cookie, no_out_cnt; + int bundle_at, ctl_cnt, no_data_chunks, eeor_mode; + unsigned int mtu, r_mtu, omtu, mx_mtu, to_out; + int tsns_sent = 0; + uint32_t auth_offset; + struct sctp_auth_chunk *auth; + uint16_t auth_keyid; + int override_ok = 1; + int skip_fill_up = 0; + int data_auth_reqd = 0; + /* JRS 5/14/07 - Add flag for whether a heartbeat is sent to + the destination. */ + int quit_now = 0; + bool use_zero_crc; + +#if defined(__APPLE__) && !defined(__Userspace__) + if (so_locked) { + sctp_lock_assert(SCTP_INP_SO(inp)); + } else { + sctp_unlock_assert(SCTP_INP_SO(inp)); + } +#endif + *num_out = 0; + *reason_code = 0; + auth_keyid = stcb->asoc.authinfo.active_keyid; + if ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED) || + (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR))) { + eeor_mode = 1; + } else { + eeor_mode = 0; + } + ctl_cnt = no_out_cnt = asconf = cookie = 0; + /* + * First lets prime the pump. For each destination, if there is room + * in the flight size, attempt to pull an MTU's worth out of the + * stream queues into the general send_queue + */ +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xC2, 2); +#endif + SCTP_TCB_LOCK_ASSERT(stcb); + hbflag = 0; + if (control_only) + no_data_chunks = 1; + else + no_data_chunks = 0; + + /* Nothing to possible to send? */ + if ((TAILQ_EMPTY(&asoc->control_send_queue) || + (asoc->ctrl_queue_cnt == stcb->asoc.ecn_echo_cnt_onq)) && + TAILQ_EMPTY(&asoc->asconf_send_queue) && + TAILQ_EMPTY(&asoc->send_queue) && + sctp_is_there_unsent_data(stcb, so_locked) == 0) { + nothing_to_send: + *reason_code = 9; + return (0); + } + if (asoc->peers_rwnd == 0) { + /* No room in peers rwnd */ + *reason_code = 1; + if (asoc->total_flight > 0) { + /* we are allowed one chunk in flight */ + no_data_chunks = 1; + } + } + if (stcb->asoc.ecn_echo_cnt_onq) { + /* Record where a sack goes, if any */ + if (no_data_chunks && + (asoc->ctrl_queue_cnt == stcb->asoc.ecn_echo_cnt_onq)) { + /* Nothing but ECNe to send - we don't do that */ + goto nothing_to_send; + } + TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { + if ((chk->rec.chunk_id.id == SCTP_SELECTIVE_ACK) || + (chk->rec.chunk_id.id == SCTP_NR_SELECTIVE_ACK)) { + sack_goes_to = chk->whoTo; + break; + } + } + } + max_rwnd_per_dest = ((asoc->peers_rwnd + asoc->total_flight) / asoc->numnets); + if (stcb->sctp_socket) + max_send_per_dest = SCTP_SB_LIMIT_SND(stcb->sctp_socket) / asoc->numnets; + else + max_send_per_dest = 0; + if (no_data_chunks == 0) { + /* How many non-directed chunks are there? */ + TAILQ_FOREACH(chk, &asoc->send_queue, sctp_next) { + if (chk->whoTo == NULL) { + /* We already have non-directed + * chunks on the queue, no need + * to do a fill-up. + */ + skip_fill_up = 1; + break; + } + } + } + if ((no_data_chunks == 0) && + (skip_fill_up == 0) && + (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, asoc))) { + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + /* + * This for loop we are in takes in + * each net, if its's got space in cwnd and + * has data sent to it (when CMT is off) then it + * calls sctp_fill_outqueue for the net. This gets + * data on the send queue for that network. + * + * In sctp_fill_outqueue TSN's are assigned and + * data is copied out of the stream buffers. Note + * mostly copy by reference (we hope). + */ + net->window_probe = 0; + if ((net != stcb->asoc.alternate) && + ((net->dest_state & SCTP_ADDR_PF) || + ((net->dest_state & SCTP_ADDR_REACHABLE) == 0) || + (net->dest_state & SCTP_ADDR_UNCONFIRMED))) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, 1, + SCTP_CWND_LOG_FILL_OUTQ_CALLED); + } + continue; + } + if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) && + (net->flight_size == 0)) { + (*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins)(stcb, net); + } + if (net->flight_size >= net->cwnd) { + /* skip this network, no room - can't fill */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, 3, + SCTP_CWND_LOG_FILL_OUTQ_CALLED); + } + continue; + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, 4, SCTP_CWND_LOG_FILL_OUTQ_CALLED); + } + sctp_fill_outqueue(stcb, net, frag_point, eeor_mode, &quit_now, so_locked); + if (quit_now) { + /* memory alloc failure */ + no_data_chunks = 1; + break; + } + } + } + /* now service each destination and send out what we can for it */ + /* Nothing to send? */ + if (TAILQ_EMPTY(&asoc->control_send_queue) && + TAILQ_EMPTY(&asoc->asconf_send_queue) && + TAILQ_EMPTY(&asoc->send_queue)) { + *reason_code = 8; + return (0); + } + + if (asoc->sctp_cmt_on_off > 0) { + /* get the last start point */ + start_at = asoc->last_net_cmt_send_started; + if (start_at == NULL) { + /* null so to beginning */ + start_at = TAILQ_FIRST(&asoc->nets); + } else { + start_at = TAILQ_NEXT(asoc->last_net_cmt_send_started, sctp_next); + if (start_at == NULL) { + start_at = TAILQ_FIRST(&asoc->nets); + } + } + asoc->last_net_cmt_send_started = start_at; + } else { + start_at = TAILQ_FIRST(&asoc->nets); + } + TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { + if (chk->whoTo == NULL) { + if (asoc->alternate) { + chk->whoTo = asoc->alternate; + } else { + chk->whoTo = asoc->primary_destination; + } + atomic_add_int(&chk->whoTo->ref_count, 1); + } + } + old_start_at = NULL; +again_one_more_time: + for (net = start_at; net != NULL; net = TAILQ_NEXT(net, sctp_next)) { + /* how much can we send? */ + /* SCTPDBG("Examine for sending net:%x\n", (uint32_t)net); */ + if (old_start_at && (old_start_at == net)) { + /* through list completely. */ + break; + } + tsns_sent = 0xa; + if (TAILQ_EMPTY(&asoc->control_send_queue) && + TAILQ_EMPTY(&asoc->asconf_send_queue) && + (net->flight_size >= net->cwnd)) { + /* Nothing on control or asconf and flight is full, we can skip + * even in the CMT case. + */ + continue; + } + bundle_at = 0; + endoutchain = outchain = NULL; + auth = NULL; + auth_offset = 0; + no_fragmentflg = 1; + one_chunk = 0; + if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { + skip_data_for_this_net = 1; + } else { + skip_data_for_this_net = 0; + } + switch (((struct sockaddr *)&net->ro._l_addr)->sa_family) { +#ifdef INET + case AF_INET: + mtu = net->mtu - SCTP_MIN_V4_OVERHEAD; + break; +#endif +#ifdef INET6 + case AF_INET6: + mtu = net->mtu - SCTP_MIN_OVERHEAD; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + mtu = net->mtu - sizeof(struct sctphdr); + break; +#endif + default: + /* TSNH */ + mtu = net->mtu; + break; + } + mx_mtu = mtu; + to_out = 0; + if (mtu > asoc->peers_rwnd) { + if (asoc->total_flight > 0) { + /* We have a packet in flight somewhere */ + r_mtu = asoc->peers_rwnd; + } else { + /* We are always allowed to send one MTU out */ + one_chunk = 1; + r_mtu = mtu; + } + } else { + r_mtu = mtu; + } + error = 0; + /************************/ + /* ASCONF transmission */ + /************************/ + /* Now first lets go through the asconf queue */ + TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) { + if (chk->rec.chunk_id.id != SCTP_ASCONF) { + continue; + } + if (chk->whoTo == NULL) { + if (asoc->alternate == NULL) { + if (asoc->primary_destination != net) { + break; + } + } else { + if (asoc->alternate != net) { + break; + } + } + } else { + if (chk->whoTo != net) { + break; + } + } + if (chk->data == NULL) { + break; + } + if (chk->sent != SCTP_DATAGRAM_UNSENT && + chk->sent != SCTP_DATAGRAM_RESEND) { + break; + } + /* + * if no AUTH is yet included and this chunk + * requires it, make sure to account for it. We + * don't apply the size until the AUTH chunk is + * actually added below in case there is no room for + * this chunk. NOTE: we overload the use of "omtu" + * here + */ + if ((auth == NULL) && + sctp_auth_is_required_chunk(chk->rec.chunk_id.id, + stcb->asoc.peer_auth_chunks)) { + omtu = sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); + } else + omtu = 0; + /* Here we do NOT factor the r_mtu */ + if ((chk->send_size < (int)(mtu - omtu)) || + (chk->flags & CHUNK_FLAGS_FRAGMENT_OK)) { + /* + * We probably should glom the mbuf chain + * from the chk->data for control but the + * problem is it becomes yet one more level + * of tracking to do if for some reason + * output fails. Then I have got to + * reconstruct the merged control chain.. el + * yucko.. for now we take the easy way and + * do the copy + */ + /* + * Add an AUTH chunk, if chunk requires it + * save the offset into the chain for AUTH + */ + if ((auth == NULL) && + (sctp_auth_is_required_chunk(chk->rec.chunk_id.id, + stcb->asoc.peer_auth_chunks))) { + outchain = sctp_add_auth_chunk(outchain, + &endoutchain, + &auth, + &auth_offset, + stcb, + chk->rec.chunk_id.id); + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); + } + outchain = sctp_copy_mbufchain(chk->data, outchain, &endoutchain, + (int)chk->rec.chunk_id.can_take_data, + chk->send_size, chk->copy_by_ref); + if (outchain == NULL) { + *reason_code = 8; + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); + /* update our MTU size */ + if (mtu > (chk->send_size + omtu)) + mtu -= (chk->send_size + omtu); + else + mtu = 0; + to_out += (chk->send_size + omtu); + /* Do clear IP_DF ? */ + if (chk->flags & CHUNK_FLAGS_FRAGMENT_OK) { + no_fragmentflg = 0; + } + if (chk->rec.chunk_id.can_take_data) + chk->data = NULL; + /* + * set hb flag since we can + * use these for RTO + */ + hbflag = 1; + asconf = 1; + /* + * should sysctl this: don't + * bundle data with ASCONF + * since it requires AUTH + */ + no_data_chunks = 1; + chk->sent = SCTP_DATAGRAM_SENT; + if (chk->whoTo == NULL) { + chk->whoTo = net; + atomic_add_int(&net->ref_count, 1); + } + chk->snd_count++; + if (mtu == 0) { + /* + * Ok we are out of room but we can + * output without effecting the + * flight size since this little guy + * is a control only packet. + */ + sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, stcb, net); + /* + * do NOT clear the asconf + * flag as it is used to do + * appropriate source address + * selection. + */ + if (*now_filled == 0) { + (void)SCTP_GETTIME_TIMEVAL(now); + *now_filled = 1; + } + net->last_sent_time = *now; + hbflag = 0; + if ((error = sctp_lowlevel_chunk_output(inp, stcb, net, + (struct sockaddr *)&net->ro._l_addr, + outchain, auth_offset, auth, + stcb->asoc.authinfo.active_keyid, + no_fragmentflg, 0, asconf, + inp->sctp_lport, stcb->rport, + htonl(stcb->asoc.peer_vtag), + net->port, NULL, +#if defined(__FreeBSD__) && !defined(__Userspace__) + 0, 0, +#endif + false, so_locked))) { + /* error, we could not output */ + SCTPDBG(SCTP_DEBUG_OUTPUT3, "Gak send error %d\n", error); + if (from_where == 0) { + SCTP_STAT_INCR(sctps_lowlevelerrusr); + } + if (error == ENOBUFS) { + asoc->ifp_had_enobuf = 1; + SCTP_STAT_INCR(sctps_lowlevelerr); + } + /* error, could not output */ + if (error == EHOSTUNREACH) { + /* + * Destination went + * unreachable + * during this send + */ + sctp_move_chunks_from_net(stcb, net); + } + asconf = 0; + *reason_code = 7; + break; + } else { + asoc->ifp_had_enobuf = 0; + } + /* + * increase the number we sent, if a + * cookie is sent we don't tell them + * any was sent out. + */ + outchain = endoutchain = NULL; + auth = NULL; + auth_offset = 0; + asconf = 0; + if (!no_out_cnt) + *num_out += ctl_cnt; + /* recalc a clean slate and setup */ + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + mtu = net->mtu - SCTP_MIN_V4_OVERHEAD; + break; +#endif +#ifdef INET6 + case AF_INET6: + mtu = net->mtu - SCTP_MIN_OVERHEAD; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + mtu = net->mtu - sizeof(struct sctphdr); + break; +#endif + default: + /* TSNH */ + mtu = net->mtu; + break; + } + to_out = 0; + no_fragmentflg = 1; + } + } + } + if (error != 0) { + /* try next net */ + continue; + } + /************************/ + /* Control transmission */ + /************************/ + /* Now first lets go through the control queue */ + TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) { + if ((sack_goes_to) && + (chk->rec.chunk_id.id == SCTP_ECN_ECHO) && + (chk->whoTo != sack_goes_to)) { + /* + * if we have a sack in queue, and we are looking at an + * ecn echo that is NOT queued to where the sack is going.. + */ + if (chk->whoTo == net) { + /* Don't transmit it to where its going (current net) */ + continue; + } else if (sack_goes_to == net) { + /* But do transmit it to this address */ + goto skip_net_check; + } + } + if (chk->whoTo == NULL) { + if (asoc->alternate == NULL) { + if (asoc->primary_destination != net) { + continue; + } + } else { + if (asoc->alternate != net) { + continue; + } + } + } else { + if (chk->whoTo != net) { + continue; + } + } + skip_net_check: + if (chk->data == NULL) { + continue; + } + if (chk->sent != SCTP_DATAGRAM_UNSENT) { + /* + * It must be unsent. Cookies and ASCONF's + * hang around but there timers will force + * when marked for resend. + */ + continue; + } + /* + * if no AUTH is yet included and this chunk + * requires it, make sure to account for it. We + * don't apply the size until the AUTH chunk is + * actually added below in case there is no room for + * this chunk. NOTE: we overload the use of "omtu" + * here + */ + if ((auth == NULL) && + sctp_auth_is_required_chunk(chk->rec.chunk_id.id, + stcb->asoc.peer_auth_chunks)) { + omtu = sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); + } else + omtu = 0; + /* Here we do NOT factor the r_mtu */ + if ((chk->send_size <= (int)(mtu - omtu)) || + (chk->flags & CHUNK_FLAGS_FRAGMENT_OK)) { + /* + * We probably should glom the mbuf chain + * from the chk->data for control but the + * problem is it becomes yet one more level + * of tracking to do if for some reason + * output fails. Then I have got to + * reconstruct the merged control chain.. el + * yucko.. for now we take the easy way and + * do the copy + */ + /* + * Add an AUTH chunk, if chunk requires it + * save the offset into the chain for AUTH + */ + if ((auth == NULL) && + (sctp_auth_is_required_chunk(chk->rec.chunk_id.id, + stcb->asoc.peer_auth_chunks))) { + outchain = sctp_add_auth_chunk(outchain, + &endoutchain, + &auth, + &auth_offset, + stcb, + chk->rec.chunk_id.id); + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); + } + outchain = sctp_copy_mbufchain(chk->data, outchain, &endoutchain, + (int)chk->rec.chunk_id.can_take_data, + chk->send_size, chk->copy_by_ref); + if (outchain == NULL) { + *reason_code = 8; + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); + /* update our MTU size */ + if (mtu > (chk->send_size + omtu)) + mtu -= (chk->send_size + omtu); + else + mtu = 0; + to_out += (chk->send_size + omtu); + /* Do clear IP_DF ? */ + if (chk->flags & CHUNK_FLAGS_FRAGMENT_OK) { + no_fragmentflg = 0; + } + if (chk->rec.chunk_id.can_take_data) + chk->data = NULL; + /* Mark things to be removed, if needed */ + if ((chk->rec.chunk_id.id == SCTP_SELECTIVE_ACK) || + (chk->rec.chunk_id.id == SCTP_NR_SELECTIVE_ACK) || /* EY */ + (chk->rec.chunk_id.id == SCTP_HEARTBEAT_REQUEST) || + (chk->rec.chunk_id.id == SCTP_HEARTBEAT_ACK) || + (chk->rec.chunk_id.id == SCTP_SHUTDOWN) || + (chk->rec.chunk_id.id == SCTP_SHUTDOWN_ACK) || + (chk->rec.chunk_id.id == SCTP_OPERATION_ERROR) || + (chk->rec.chunk_id.id == SCTP_COOKIE_ACK) || + (chk->rec.chunk_id.id == SCTP_ECN_CWR) || + (chk->rec.chunk_id.id == SCTP_PACKET_DROPPED) || + (chk->rec.chunk_id.id == SCTP_ASCONF_ACK)) { + if (chk->rec.chunk_id.id == SCTP_HEARTBEAT_REQUEST) { + hbflag = 1; + } + /* remove these chunks at the end */ + if ((chk->rec.chunk_id.id == SCTP_SELECTIVE_ACK) || + (chk->rec.chunk_id.id == SCTP_NR_SELECTIVE_ACK)) { + /* turn off the timer */ + if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) { + sctp_timer_stop(SCTP_TIMER_TYPE_RECV, + inp, stcb, NULL, + SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_1); + } + } + ctl_cnt++; + } else { + /* + * Other chunks, since they have + * timers running (i.e. COOKIE) + * we just "trust" that it + * gets sent or retransmitted. + */ + ctl_cnt++; + if (chk->rec.chunk_id.id == SCTP_COOKIE_ECHO) { + cookie = 1; + no_out_cnt = 1; + } else if (chk->rec.chunk_id.id == SCTP_ECN_ECHO) { + /* + * Increment ecne send count here + * this means we may be over-zealous in + * our counting if the send fails, but its + * the best place to do it (we used to do + * it in the queue of the chunk, but that did + * not tell how many times it was sent. + */ + SCTP_STAT_INCR(sctps_sendecne); + } + chk->sent = SCTP_DATAGRAM_SENT; + if (chk->whoTo == NULL) { + chk->whoTo = net; + atomic_add_int(&net->ref_count, 1); + } + chk->snd_count++; + } + if (mtu == 0) { + /* + * Ok we are out of room but we can + * output without effecting the + * flight size since this little guy + * is a control only packet. + */ + use_zero_crc = asoc->zero_checksum == 2; + if (asconf) { + sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, stcb, net); + use_zero_crc = false; + /* + * do NOT clear the asconf + * flag as it is used to do + * appropriate source address + * selection. + */ + } + if (cookie) { + sctp_timer_start(SCTP_TIMER_TYPE_COOKIE, inp, stcb, net); + use_zero_crc = false; + cookie = 0; + } + /* Only HB or ASCONF advances time */ + if (hbflag) { + if (*now_filled == 0) { + (void)SCTP_GETTIME_TIMEVAL(now); + *now_filled = 1; + } + net->last_sent_time = *now; + hbflag = 0; + } + if ((error = sctp_lowlevel_chunk_output(inp, stcb, net, + (struct sockaddr *)&net->ro._l_addr, + outchain, + auth_offset, auth, + stcb->asoc.authinfo.active_keyid, + no_fragmentflg, 0, asconf, + inp->sctp_lport, stcb->rport, + htonl(stcb->asoc.peer_vtag), + net->port, NULL, +#if defined(__FreeBSD__) && !defined(__Userspace__) + 0, 0, +#endif + use_zero_crc, so_locked))) { + /* error, we could not output */ + SCTPDBG(SCTP_DEBUG_OUTPUT3, "Gak send error %d\n", error); + if (from_where == 0) { + SCTP_STAT_INCR(sctps_lowlevelerrusr); + } + if (error == ENOBUFS) { + asoc->ifp_had_enobuf = 1; + SCTP_STAT_INCR(sctps_lowlevelerr); + } + if (error == EHOSTUNREACH) { + /* + * Destination went + * unreachable + * during this send + */ + sctp_move_chunks_from_net(stcb, net); + } + asconf = 0; + *reason_code = 7; + break; + } else { + asoc->ifp_had_enobuf = 0; + } + /* + * increase the number we sent, if a + * cookie is sent we don't tell them + * any was sent out. + */ + outchain = endoutchain = NULL; + auth = NULL; + auth_offset = 0; + asconf = 0; + if (!no_out_cnt) + *num_out += ctl_cnt; + /* recalc a clean slate and setup */ + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + mtu = net->mtu - SCTP_MIN_V4_OVERHEAD; + break; +#endif +#ifdef INET6 + case AF_INET6: + mtu = net->mtu - SCTP_MIN_OVERHEAD; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + mtu = net->mtu - sizeof(struct sctphdr); + break; +#endif + default: + /* TSNH */ + mtu = net->mtu; + break; + } + to_out = 0; + no_fragmentflg = 1; + } + } + } + if (error != 0) { + /* try next net */ + continue; + } + /* JRI: if dest is in PF state, do not send data to it */ + if ((asoc->sctp_cmt_on_off > 0) && + (net != stcb->asoc.alternate) && + (net->dest_state & SCTP_ADDR_PF)) { + goto no_data_fill; + } + if (net->flight_size >= net->cwnd) { + goto no_data_fill; + } + if ((asoc->sctp_cmt_on_off > 0) && + (SCTP_BASE_SYSCTL(sctp_buffer_splitting) & SCTP_RECV_BUFFER_SPLITTING) && + (net->flight_size > max_rwnd_per_dest)) { + goto no_data_fill; + } + /* + * We need a specific accounting for the usage of the + * send buffer. We also need to check the number of messages + * per net. For now, this is better than nothing and it + * disabled by default... + */ + if ((asoc->sctp_cmt_on_off > 0) && + (SCTP_BASE_SYSCTL(sctp_buffer_splitting) & SCTP_SEND_BUFFER_SPLITTING) && + (max_send_per_dest > 0) && + (net->flight_size > max_send_per_dest)) { + goto no_data_fill; + } + /*********************/ + /* Data transmission */ + /*********************/ + /* + * if AUTH for DATA is required and no AUTH has been added + * yet, account for this in the mtu now... if no data can be + * bundled, this adjustment won't matter anyways since the + * packet will be going out... + */ + data_auth_reqd = sctp_auth_is_required_chunk(SCTP_DATA, + stcb->asoc.peer_auth_chunks); + if (data_auth_reqd && (auth == NULL)) { + mtu -= sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); + } + /* now lets add any data within the MTU constraints */ + switch (((struct sockaddr *)&net->ro._l_addr)->sa_family) { +#ifdef INET + case AF_INET: + if (net->mtu > SCTP_MIN_V4_OVERHEAD) + omtu = net->mtu - SCTP_MIN_V4_OVERHEAD; + else + omtu = 0; + break; +#endif +#ifdef INET6 + case AF_INET6: + if (net->mtu > SCTP_MIN_OVERHEAD) + omtu = net->mtu - SCTP_MIN_OVERHEAD; + else + omtu = 0; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + if (net->mtu > sizeof(struct sctphdr)) { + omtu = net->mtu - sizeof(struct sctphdr); + } else { + omtu = 0; + } + break; +#endif + default: + /* TSNH */ + omtu = 0; + break; + } + if ((((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) && + (skip_data_for_this_net == 0)) || + (cookie)) { + TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) { + if (no_data_chunks) { + /* let only control go out */ + *reason_code = 1; + break; + } + if (net->flight_size >= net->cwnd) { + /* skip this net, no room for data */ + *reason_code = 2; + break; + } + if ((chk->whoTo != NULL) && + (chk->whoTo != net)) { + /* Don't send the chunk on this net */ + continue; + } + + if (asoc->sctp_cmt_on_off == 0) { + if ((asoc->alternate) && + (asoc->alternate != net) && + (chk->whoTo == NULL)) { + continue; + } else if ((net != asoc->primary_destination) && + (asoc->alternate == NULL) && + (chk->whoTo == NULL)) { + continue; + } + } + if ((chk->send_size > omtu) && ((chk->flags & CHUNK_FLAGS_FRAGMENT_OK) == 0)) { + /*- + * strange, we have a chunk that is + * to big for its destination and + * yet no fragment ok flag. + * Something went wrong when the + * PMTU changed...we did not mark + * this chunk for some reason?? I + * will fix it here by letting IP + * fragment it for now and printing + * a warning. This really should not + * happen ... + */ + SCTP_PRINTF("Warning chunk of %d bytes > mtu:%d and yet PMTU disc missed\n", + chk->send_size, mtu); + chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; + } + if (SCTP_BASE_SYSCTL(sctp_enable_sack_immediately) && + (asoc->state & SCTP_STATE_SHUTDOWN_PENDING)) { + struct sctp_data_chunk *dchkh; + + dchkh = mtod(chk->data, struct sctp_data_chunk *); + dchkh->ch.chunk_flags |= SCTP_DATA_SACK_IMMEDIATELY; + } + if (((chk->send_size <= mtu) && (chk->send_size <= r_mtu)) || + ((chk->flags & CHUNK_FLAGS_FRAGMENT_OK) && (chk->send_size <= asoc->peers_rwnd))) { + /* ok we will add this one */ + + /* + * Add an AUTH chunk, if chunk + * requires it, save the offset into + * the chain for AUTH + */ + if (data_auth_reqd) { + if (auth == NULL) { + outchain = sctp_add_auth_chunk(outchain, + &endoutchain, + &auth, + &auth_offset, + stcb, + SCTP_DATA); + auth_keyid = chk->auth_keyid; + override_ok = 0; + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); + } else if (override_ok) { + /* use this data's keyid */ + auth_keyid = chk->auth_keyid; + override_ok = 0; + } else if (auth_keyid != chk->auth_keyid) { + /* different keyid, so done bundling */ + break; + } + } + outchain = sctp_copy_mbufchain(chk->data, outchain, &endoutchain, 0, + chk->send_size, chk->copy_by_ref); + if (outchain == NULL) { + SCTPDBG(SCTP_DEBUG_OUTPUT3, "No memory?\n"); + if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { + sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); + } + *reason_code = 3; + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + /* update our MTU size */ + /* Do clear IP_DF ? */ + if (chk->flags & CHUNK_FLAGS_FRAGMENT_OK) { + no_fragmentflg = 0; + } + /* unsigned subtraction of mtu */ + if (mtu > chk->send_size) + mtu -= chk->send_size; + else + mtu = 0; + /* unsigned subtraction of r_mtu */ + if (r_mtu > chk->send_size) + r_mtu -= chk->send_size; + else + r_mtu = 0; + + to_out += chk->send_size; + if ((to_out > mx_mtu) && no_fragmentflg) { +#ifdef INVARIANTS + panic("Exceeding mtu of %d out size is %d", mx_mtu, to_out); +#else + SCTP_PRINTF("Exceeding mtu of %d out size is %d\n", + mx_mtu, to_out); +#endif + } + chk->window_probe = 0; + data_list[bundle_at++] = chk; + if (bundle_at >= SCTP_MAX_DATA_BUNDLING) { + break; + } + if (chk->sent == SCTP_DATAGRAM_UNSENT) { + if ((chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == 0) { + SCTP_STAT_INCR_COUNTER64(sctps_outorderchunks); + } else { + SCTP_STAT_INCR_COUNTER64(sctps_outunorderchunks); + } + if (((chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) == SCTP_DATA_LAST_FRAG) && + ((chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) == 0)) + /* Count number of user msg's that were fragmented + * we do this by counting when we see a LAST fragment + * only. + */ + SCTP_STAT_INCR_COUNTER64(sctps_fragusrmsgs); + } + if ((mtu == 0) || (r_mtu == 0) || (one_chunk)) { + if ((one_chunk) && (stcb->asoc.total_flight == 0)) { + data_list[0]->window_probe = 1; + net->window_probe = 1; + } + break; + } + } else { + /* + * Must be sent in order of the + * TSN's (on a network) + */ + break; + } + } /* for (chunk gather loop for this net) */ + } /* if asoc.state OPEN */ + no_data_fill: + /* Is there something to send for this destination? */ + if (outchain) { + /* We may need to start a control timer or two */ + use_zero_crc = asoc->zero_checksum == 2; + if (asconf) { + sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, + stcb, net); + use_zero_crc = false; + /* + * do NOT clear the asconf flag as it is used + * to do appropriate source address selection. + */ + } + if (cookie) { + sctp_timer_start(SCTP_TIMER_TYPE_COOKIE, inp, stcb, net); + use_zero_crc = false; + cookie = 0; + } + /* must start a send timer if data is being sent */ + if (bundle_at && (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer))) { + /* + * no timer running on this destination + * restart it. + */ + sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); + } + if (bundle_at || hbflag) { + /* For data/asconf and hb set time */ + if (*now_filled == 0) { + (void)SCTP_GETTIME_TIMEVAL(now); + *now_filled = 1; + } + net->last_sent_time = *now; + } + /* Now send it, if there is anything to send :> */ + if ((error = sctp_lowlevel_chunk_output(inp, + stcb, + net, + (struct sockaddr *)&net->ro._l_addr, + outchain, + auth_offset, + auth, + auth_keyid, + no_fragmentflg, + bundle_at, + asconf, + inp->sctp_lport, stcb->rport, + htonl(stcb->asoc.peer_vtag), + net->port, NULL, +#if defined(__FreeBSD__) && !defined(__Userspace__) + 0, 0, +#endif + use_zero_crc, + so_locked))) { + /* error, we could not output */ + SCTPDBG(SCTP_DEBUG_OUTPUT3, "Gak send error %d\n", error); + if (from_where == 0) { + SCTP_STAT_INCR(sctps_lowlevelerrusr); + } + if (error == ENOBUFS) { + asoc->ifp_had_enobuf = 1; + SCTP_STAT_INCR(sctps_lowlevelerr); + } + if (error == EHOSTUNREACH) { + /* + * Destination went unreachable + * during this send + */ + sctp_move_chunks_from_net(stcb, net); + } + asconf = 0; + *reason_code = 6; + /*- + * I add this line to be paranoid. As far as + * I can tell the continue, takes us back to + * the top of the for, but just to make sure + * I will reset these again here. + */ + ctl_cnt = 0; + continue; /* This takes us back to the for() for the nets. */ + } else { + asoc->ifp_had_enobuf = 0; + } + endoutchain = NULL; + auth = NULL; + auth_offset = 0; + asconf = 0; + if (!no_out_cnt) { + *num_out += (ctl_cnt + bundle_at); + } + if (bundle_at) { + /* setup for a RTO measurement */ + tsns_sent = data_list[0]->rec.data.tsn; + /* fill time if not already filled */ + if (*now_filled == 0) { + (void)SCTP_GETTIME_TIMEVAL(&asoc->time_last_sent); + *now_filled = 1; + *now = asoc->time_last_sent; + } else { + asoc->time_last_sent = *now; + } + if (net->rto_needed) { + data_list[0]->do_rtt = 1; + net->rto_needed = 0; + } + SCTP_STAT_INCR_BY(sctps_senddata, bundle_at); + sctp_clean_up_datalist(stcb, asoc, data_list, bundle_at, net); + } + if (one_chunk) { + break; + } + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, tsns_sent, SCTP_CWND_LOG_FROM_SEND); + } + } + if (old_start_at == NULL) { + old_start_at = start_at; + start_at = TAILQ_FIRST(&asoc->nets); + if (old_start_at) + goto again_one_more_time; + } + + /* + * At the end there should be no NON timed chunks hanging on this + * queue. + */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, *num_out, SCTP_CWND_LOG_FROM_SEND); + } + if ((*num_out == 0) && (*reason_code == 0)) { + *reason_code = 4; + } else { + *reason_code = 5; + } + sctp_clean_up_ctl(stcb, asoc, so_locked); + return (0); +} + +void +sctp_queue_op_err(struct sctp_tcb *stcb, struct mbuf *op_err) +{ + /*- + * Prepend a OPERATIONAL_ERROR chunk header and put on the end of + * the control chunk queue. + */ + struct sctp_chunkhdr *hdr; + struct sctp_tmit_chunk *chk; + struct mbuf *mat, *last_mbuf; + uint32_t chunk_length; + uint16_t padding_length; + + SCTP_TCB_LOCK_ASSERT(stcb); + SCTP_BUF_PREPEND(op_err, sizeof(struct sctp_chunkhdr), M_NOWAIT); + if (op_err == NULL) { + return; + } + last_mbuf = NULL; + chunk_length = 0; + for (mat = op_err; mat != NULL; mat = SCTP_BUF_NEXT(mat)) { + chunk_length += SCTP_BUF_LEN(mat); + if (SCTP_BUF_NEXT(mat) == NULL) { + last_mbuf = mat; + } + } + if (chunk_length > SCTP_MAX_CHUNK_LENGTH) { + sctp_m_freem(op_err); + return; + } + padding_length = chunk_length % 4; + if (padding_length != 0) { + padding_length = 4 - padding_length; + } + if (padding_length != 0) { + if (sctp_add_pad_tombuf(last_mbuf, padding_length) == NULL) { + sctp_m_freem(op_err); + return; + } + } + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + /* no memory */ + sctp_m_freem(op_err); + return; + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_OPERATION_ERROR; + chk->rec.chunk_id.can_take_data = 0; + chk->flags = 0; + chk->send_size = (uint16_t)chunk_length; + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + chk->asoc = &stcb->asoc; + chk->data = op_err; + chk->whoTo = NULL; + hdr = mtod(op_err, struct sctp_chunkhdr *); + hdr->chunk_type = SCTP_OPERATION_ERROR; + hdr->chunk_flags = 0; + hdr->chunk_length = htons(chk->send_size); + TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); + chk->asoc->ctrl_queue_cnt++; +} + +int +sctp_send_cookie_echo(struct mbuf *m, + int offset, int limit, + struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + /*- + * pull out the cookie and put it at the front of the control chunk + * queue. + */ + int at; + struct mbuf *cookie; + struct sctp_paramhdr param, *phdr; + struct sctp_chunkhdr *hdr; + struct sctp_tmit_chunk *chk; + uint16_t ptype, plen; + + SCTP_TCB_LOCK_ASSERT(stcb); + /* First find the cookie in the param area */ + cookie = NULL; + at = offset + sizeof(struct sctp_init_chunk); + for (;;) { + phdr = sctp_get_next_param(m, at, ¶m, sizeof(param)); + if (phdr == NULL) { + return (-3); + } + ptype = ntohs(phdr->param_type); + plen = ntohs(phdr->param_length); + if (plen < sizeof(struct sctp_paramhdr)) { + return (-6); + } + if (ptype == SCTP_STATE_COOKIE) { + int pad; + + /* found the cookie */ + if (at + plen > limit) { + return (-7); + } + cookie = SCTP_M_COPYM(m, at, plen, M_NOWAIT); + if (cookie == NULL) { + /* No memory */ + return (-2); + } + if ((pad = (plen % 4)) > 0) { + pad = 4 - pad; + } + if (pad > 0) { + if (sctp_pad_lastmbuf(cookie, pad, NULL) == NULL) { + return (-8); + } + } +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(cookie, SCTP_MBUF_ICOPY); + } +#endif + break; + } + at += SCTP_SIZE32(plen); + } + /* ok, we got the cookie lets change it into a cookie echo chunk */ + /* first the change from param to cookie */ + hdr = mtod(cookie, struct sctp_chunkhdr *); + hdr->chunk_type = SCTP_COOKIE_ECHO; + hdr->chunk_flags = 0; + /* get the chunk stuff now and place it in the FRONT of the queue */ + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + /* no memory */ + sctp_m_freem(cookie); + return (-5); + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_COOKIE_ECHO; + chk->rec.chunk_id.can_take_data = 0; + chk->flags = CHUNK_FLAGS_FRAGMENT_OK; + chk->send_size = SCTP_SIZE32(plen); + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + chk->asoc = &stcb->asoc; + chk->data = cookie; + chk->whoTo = net; + atomic_add_int(&chk->whoTo->ref_count, 1); + TAILQ_INSERT_HEAD(&chk->asoc->control_send_queue, chk, sctp_next); + chk->asoc->ctrl_queue_cnt++; + return (0); +} + +void +sctp_send_heartbeat_ack(struct sctp_tcb *stcb, + struct mbuf *m, + int offset, + int chk_length, + struct sctp_nets *net) +{ + /* + * take a HB request and make it into a HB ack and send it. + */ + struct mbuf *outchain; + struct sctp_chunkhdr *chdr; + struct sctp_tmit_chunk *chk; + + if (net == NULL) + /* must have a net pointer */ + return; + + outchain = SCTP_M_COPYM(m, offset, chk_length, M_NOWAIT); + if (outchain == NULL) { + /* gak out of memory */ + return; + } +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(outchain, SCTP_MBUF_ICOPY); + } +#endif + chdr = mtod(outchain, struct sctp_chunkhdr *); + chdr->chunk_type = SCTP_HEARTBEAT_ACK; + chdr->chunk_flags = 0; + if (chk_length % 4 != 0) { + sctp_pad_lastmbuf(outchain, 4 - (chk_length % 4), NULL); + } + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + /* no memory */ + sctp_m_freem(outchain); + return; + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_HEARTBEAT_ACK; + chk->rec.chunk_id.can_take_data = 1; + chk->flags = 0; + chk->send_size = chk_length; + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + chk->asoc = &stcb->asoc; + chk->data = outchain; + chk->whoTo = net; + atomic_add_int(&chk->whoTo->ref_count, 1); + TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); + chk->asoc->ctrl_queue_cnt++; +} + +void +sctp_send_cookie_ack(struct sctp_tcb *stcb) +{ + /* formulate and queue a cookie-ack back to sender */ + struct mbuf *cookie_ack; + struct sctp_chunkhdr *hdr; + struct sctp_tmit_chunk *chk; + + SCTP_TCB_LOCK_ASSERT(stcb); + + cookie_ack = sctp_get_mbuf_for_msg(sizeof(struct sctp_chunkhdr), 0, M_NOWAIT, 1, MT_HEADER); + if (cookie_ack == NULL) { + /* no mbuf's */ + return; + } + SCTP_BUF_RESV_UF(cookie_ack, SCTP_MIN_OVERHEAD); + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + /* no memory */ + sctp_m_freem(cookie_ack); + return; + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_COOKIE_ACK; + chk->rec.chunk_id.can_take_data = 1; + chk->flags = 0; + chk->send_size = sizeof(struct sctp_chunkhdr); + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + chk->asoc = &stcb->asoc; + chk->data = cookie_ack; + if (chk->asoc->last_control_chunk_from != NULL) { + chk->whoTo = chk->asoc->last_control_chunk_from; + atomic_add_int(&chk->whoTo->ref_count, 1); + } else { + chk->whoTo = NULL; + } + hdr = mtod(cookie_ack, struct sctp_chunkhdr *); + hdr->chunk_type = SCTP_COOKIE_ACK; + hdr->chunk_flags = 0; + hdr->chunk_length = htons(chk->send_size); + SCTP_BUF_LEN(cookie_ack) = chk->send_size; + TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); + chk->asoc->ctrl_queue_cnt++; + return; +} + +void +sctp_send_shutdown_ack(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + /* formulate and queue a SHUTDOWN-ACK back to the sender */ + struct mbuf *m_shutdown_ack; + struct sctp_shutdown_ack_chunk *ack_cp; + struct sctp_tmit_chunk *chk; + + m_shutdown_ack = sctp_get_mbuf_for_msg(sizeof(struct sctp_shutdown_ack_chunk), 0, M_NOWAIT, 1, MT_HEADER); + if (m_shutdown_ack == NULL) { + /* no mbuf's */ + return; + } + SCTP_BUF_RESV_UF(m_shutdown_ack, SCTP_MIN_OVERHEAD); + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + /* no memory */ + sctp_m_freem(m_shutdown_ack); + return; + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_SHUTDOWN_ACK; + chk->rec.chunk_id.can_take_data = 1; + chk->flags = 0; + chk->send_size = sizeof(struct sctp_chunkhdr); + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + chk->asoc = &stcb->asoc; + chk->data = m_shutdown_ack; + chk->whoTo = net; + if (chk->whoTo) { + atomic_add_int(&chk->whoTo->ref_count, 1); + } + ack_cp = mtod(m_shutdown_ack, struct sctp_shutdown_ack_chunk *); + ack_cp->ch.chunk_type = SCTP_SHUTDOWN_ACK; + ack_cp->ch.chunk_flags = 0; + ack_cp->ch.chunk_length = htons(chk->send_size); + SCTP_BUF_LEN(m_shutdown_ack) = chk->send_size; + TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); + chk->asoc->ctrl_queue_cnt++; + return; +} + +void +sctp_send_shutdown(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + /* formulate and queue a SHUTDOWN to the sender */ + struct mbuf *m_shutdown; + struct sctp_shutdown_chunk *shutdown_cp; + struct sctp_tmit_chunk *chk; + + TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) { + if (chk->rec.chunk_id.id == SCTP_SHUTDOWN) { + /* We already have a SHUTDOWN queued. Reuse it. */ + if (chk->whoTo) { + sctp_free_remote_addr(chk->whoTo); + chk->whoTo = NULL; + } + break; + } + } + if (chk == NULL) { + m_shutdown = sctp_get_mbuf_for_msg(sizeof(struct sctp_shutdown_chunk), 0, M_NOWAIT, 1, MT_HEADER); + if (m_shutdown == NULL) { + /* no mbuf's */ + return; + } + SCTP_BUF_RESV_UF(m_shutdown, SCTP_MIN_OVERHEAD); + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + /* no memory */ + sctp_m_freem(m_shutdown); + return; + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_SHUTDOWN; + chk->rec.chunk_id.can_take_data = 1; + chk->flags = 0; + chk->send_size = sizeof(struct sctp_shutdown_chunk); + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + chk->asoc = &stcb->asoc; + chk->data = m_shutdown; + chk->whoTo = net; + if (chk->whoTo) { + atomic_add_int(&chk->whoTo->ref_count, 1); + } + shutdown_cp = mtod(m_shutdown, struct sctp_shutdown_chunk *); + shutdown_cp->ch.chunk_type = SCTP_SHUTDOWN; + shutdown_cp->ch.chunk_flags = 0; + shutdown_cp->ch.chunk_length = htons(chk->send_size); + shutdown_cp->cumulative_tsn_ack = htonl(stcb->asoc.cumulative_tsn); + SCTP_BUF_LEN(m_shutdown) = chk->send_size; + TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); + chk->asoc->ctrl_queue_cnt++; + } else { + TAILQ_REMOVE(&stcb->asoc.control_send_queue, chk, sctp_next); + chk->whoTo = net; + if (chk->whoTo) { + atomic_add_int(&chk->whoTo->ref_count, 1); + } + shutdown_cp = mtod(chk->data, struct sctp_shutdown_chunk *); + shutdown_cp->cumulative_tsn_ack = htonl(stcb->asoc.cumulative_tsn); + TAILQ_INSERT_TAIL(&stcb->asoc.control_send_queue, chk, sctp_next); + } + return; +} + +void +sctp_send_asconf(struct sctp_tcb *stcb, struct sctp_nets *net, int addr_locked) +{ + /* + * formulate and queue an ASCONF to the peer. + * ASCONF parameters should be queued on the assoc queue. + */ + struct sctp_tmit_chunk *chk; + struct mbuf *m_asconf; + int len; + + SCTP_TCB_LOCK_ASSERT(stcb); + + if ((!TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) && + (!sctp_is_feature_on(stcb->sctp_ep, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS))) { + /* can't send a new one if there is one in flight already */ + return; + } + + /* compose an ASCONF chunk, maximum length is PMTU */ + m_asconf = sctp_compose_asconf(stcb, &len, addr_locked); + if (m_asconf == NULL) { + return; + } + + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + /* no memory */ + sctp_m_freem(m_asconf); + return; + } + + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_ASCONF; + chk->rec.chunk_id.can_take_data = 0; + chk->flags = CHUNK_FLAGS_FRAGMENT_OK; + chk->data = m_asconf; + chk->send_size = len; + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + chk->asoc = &stcb->asoc; + chk->whoTo = net; + if (chk->whoTo) { + atomic_add_int(&chk->whoTo->ref_count, 1); + } + TAILQ_INSERT_TAIL(&chk->asoc->asconf_send_queue, chk, sctp_next); + chk->asoc->ctrl_queue_cnt++; + return; +} + +void +sctp_send_asconf_ack(struct sctp_tcb *stcb) +{ + /* + * formulate and queue a asconf-ack back to sender. + * the asconf-ack must be stored in the tcb. + */ + struct sctp_tmit_chunk *chk; + struct sctp_asconf_ack *ack, *latest_ack; + struct mbuf *m_ack; + struct sctp_nets *net = NULL; + + SCTP_TCB_LOCK_ASSERT(stcb); + /* Get the latest ASCONF-ACK */ + latest_ack = TAILQ_LAST(&stcb->asoc.asconf_ack_sent, sctp_asconf_ackhead); + if (latest_ack == NULL) { + return; + } + if (latest_ack->last_sent_to != NULL && + latest_ack->last_sent_to == stcb->asoc.last_control_chunk_from) { + /* we're doing a retransmission */ + net = sctp_find_alternate_net(stcb, stcb->asoc.last_control_chunk_from, 0); + if (net == NULL) { + /* no alternate */ + if (stcb->asoc.last_control_chunk_from == NULL) { + if (stcb->asoc.alternate) { + net = stcb->asoc.alternate; + } else { + net = stcb->asoc.primary_destination; + } + } else { + net = stcb->asoc.last_control_chunk_from; + } + } + } else { + /* normal case */ + if (stcb->asoc.last_control_chunk_from == NULL) { + if (stcb->asoc.alternate) { + net = stcb->asoc.alternate; + } else { + net = stcb->asoc.primary_destination; + } + } else { + net = stcb->asoc.last_control_chunk_from; + } + } + latest_ack->last_sent_to = net; + + TAILQ_FOREACH(ack, &stcb->asoc.asconf_ack_sent, next) { + if (ack->data == NULL) { + continue; + } + + /* copy the asconf_ack */ + m_ack = SCTP_M_COPYM(ack->data, 0, M_COPYALL, M_NOWAIT); + if (m_ack == NULL) { + /* couldn't copy it */ + return; + } +#ifdef SCTP_MBUF_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(m_ack, SCTP_MBUF_ICOPY); + } +#endif + + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + /* no memory */ + if (m_ack) + sctp_m_freem(m_ack); + return; + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_ASCONF_ACK; + chk->rec.chunk_id.can_take_data = 1; + chk->flags = CHUNK_FLAGS_FRAGMENT_OK; + chk->whoTo = net; + if (chk->whoTo) { + atomic_add_int(&chk->whoTo->ref_count, 1); + } + chk->data = m_ack; + chk->send_size = ack->len; + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + chk->asoc = &stcb->asoc; + + TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); + chk->asoc->ctrl_queue_cnt++; + } + return; +} + +static int +sctp_chunk_retransmission(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_association *asoc, + int *cnt_out, struct timeval *now, int *now_filled, int *fr_done, int so_locked) +{ + /*- + * send out one MTU of retransmission. If fast_retransmit is + * happening we ignore the cwnd. Otherwise we obey the cwnd and + * rwnd. For a Cookie or Asconf in the control chunk queue we + * retransmit them by themselves. + * + * For data chunks we will pick out the lowest TSN's in the sent_queue + * marked for resend and bundle them all together (up to a MTU of + * destination). The address to send to should have been + * selected/changed where the retransmission was marked (i.e. in FR + * or t3-timeout routines). + */ + struct sctp_tmit_chunk *data_list[SCTP_MAX_DATA_BUNDLING]; + struct sctp_tmit_chunk *chk, *fwd; + struct mbuf *m, *endofchain; + struct sctp_nets *net = NULL; + uint32_t tsns_sent = 0; + int no_fragmentflg, bundle_at; + unsigned int mtu; + int error, i, one_chunk, fwd_tsn, ctl_cnt, tmr_started; + struct sctp_auth_chunk *auth = NULL; + uint32_t auth_offset = 0; + uint16_t auth_keyid; + int override_ok = 1; + int data_auth_reqd = 0; + uint32_t dmtu = 0; + bool use_zero_crc; + +#if defined(__APPLE__) && !defined(__Userspace__) + if (so_locked) { + sctp_lock_assert(SCTP_INP_SO(inp)); + } else { + sctp_unlock_assert(SCTP_INP_SO(inp)); + } +#endif + SCTP_TCB_LOCK_ASSERT(stcb); + tmr_started = ctl_cnt = 0; + no_fragmentflg = 1; + fwd_tsn = 0; + *cnt_out = 0; + fwd = NULL; + endofchain = m = NULL; + auth_keyid = stcb->asoc.authinfo.active_keyid; +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xC3, 1); +#endif + if ((TAILQ_EMPTY(&asoc->sent_queue)) && + (TAILQ_EMPTY(&asoc->control_send_queue))) { + SCTPDBG(SCTP_DEBUG_OUTPUT1,"SCTP hits empty queue with cnt set to %d?\n", + asoc->sent_queue_retran_cnt); + asoc->sent_queue_cnt = 0; + asoc->sent_queue_cnt_removeable = 0; + /* send back 0/0 so we enter normal transmission */ + *cnt_out = 0; + return (0); + } + TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { + if ((chk->rec.chunk_id.id == SCTP_COOKIE_ECHO) || + (chk->rec.chunk_id.id == SCTP_STREAM_RESET) || + (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN)) { + if (chk->sent != SCTP_DATAGRAM_RESEND) { + continue; + } + if (chk->rec.chunk_id.id == SCTP_STREAM_RESET) { + if (chk != asoc->str_reset) { + /* + * not eligible for retran if its + * not ours + */ + continue; + } + } + ctl_cnt++; + if (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN) { + fwd_tsn = 1; + } + /* + * Add an AUTH chunk, if chunk requires it save the + * offset into the chain for AUTH + */ + if ((auth == NULL) && + (sctp_auth_is_required_chunk(chk->rec.chunk_id.id, + stcb->asoc.peer_auth_chunks))) { + m = sctp_add_auth_chunk(m, &endofchain, + &auth, &auth_offset, + stcb, + chk->rec.chunk_id.id); + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); + } + m = sctp_copy_mbufchain(chk->data, m, &endofchain, 0, chk->send_size, chk->copy_by_ref); + break; + } + } + one_chunk = 0; + /* do we have control chunks to retransmit? */ + if (m != NULL) { + /* Start a timer no matter if we succeed or fail */ + use_zero_crc = asoc->zero_checksum == 2; + if (chk->rec.chunk_id.id == SCTP_COOKIE_ECHO) { + sctp_timer_start(SCTP_TIMER_TYPE_COOKIE, inp, stcb, chk->whoTo); + use_zero_crc = false; + } else if (chk->rec.chunk_id.id == SCTP_ASCONF) { + /* XXXMT: Can this happen? */ + sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, stcb, chk->whoTo); + use_zero_crc = false; + } + chk->snd_count++; /* update our count */ + if ((error = sctp_lowlevel_chunk_output(inp, stcb, chk->whoTo, + (struct sockaddr *)&chk->whoTo->ro._l_addr, m, + auth_offset, auth, stcb->asoc.authinfo.active_keyid, + no_fragmentflg, 0, 0, + inp->sctp_lport, stcb->rport, htonl(stcb->asoc.peer_vtag), + chk->whoTo->port, NULL, +#if defined(__FreeBSD__) && !defined(__Userspace__) + 0, 0, +#endif + use_zero_crc, + so_locked))) { + SCTPDBG(SCTP_DEBUG_OUTPUT3, "Gak send error %d\n", error); + if (error == ENOBUFS) { + asoc->ifp_had_enobuf = 1; + SCTP_STAT_INCR(sctps_lowlevelerr); + } + return (error); + } else { + asoc->ifp_had_enobuf = 0; + } + endofchain = NULL; + auth = NULL; + auth_offset = 0; + /* + * We don't want to mark the net->sent time here since this + * we use this for HB and retrans cannot measure RTT + */ + /* (void)SCTP_GETTIME_TIMEVAL(&chk->whoTo->last_sent_time); */ + *cnt_out += 1; + chk->sent = SCTP_DATAGRAM_SENT; + sctp_ucount_decr(stcb->asoc.sent_queue_retran_cnt); + if (fwd_tsn == 0) { + return (0); + } else { + /* Clean up the fwd-tsn list */ + sctp_clean_up_ctl(stcb, asoc, so_locked); + return (0); + } + } + /* + * Ok, it is just data retransmission we need to do or that and a + * fwd-tsn with it all. + */ + if (TAILQ_EMPTY(&asoc->sent_queue)) { + return (SCTP_RETRAN_DONE); + } + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT)) { + /* not yet open, resend the cookie and that is it */ + return (1); + } +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(20, inp, stcb, NULL); +#endif + data_auth_reqd = sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks); + TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { + if (chk->sent != SCTP_DATAGRAM_RESEND) { + /* No, not sent to this net or not ready for rtx */ + continue; + } + if (chk->data == NULL) { + SCTP_PRINTF("TSN:%x chk->snd_count:%d chk->sent:%d can't retran - no data\n", + chk->rec.data.tsn, chk->snd_count, chk->sent); + continue; + } + if ((SCTP_BASE_SYSCTL(sctp_max_retran_chunk)) && + (chk->snd_count >= SCTP_BASE_SYSCTL(sctp_max_retran_chunk))) { + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + SCTP_SNPRINTF(msg, sizeof(msg), "TSN %8.8x retransmitted %d times, giving up", + chk->rec.data.tsn, chk->snd_count); + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + msg); + atomic_add_int(&stcb->asoc.refcnt, 1); + sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, + false, so_locked); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + return (SCTP_RETRAN_EXIT); + } + /* pick up the net */ + net = chk->whoTo; + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + mtu = net->mtu - SCTP_MIN_V4_OVERHEAD; + break; +#endif +#ifdef INET6 + case AF_INET6: + mtu = net->mtu - SCTP_MIN_OVERHEAD; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + mtu = net->mtu - sizeof(struct sctphdr); + break; +#endif + default: + /* TSNH */ + mtu = net->mtu; + break; + } + + if ((asoc->peers_rwnd < mtu) && (asoc->total_flight > 0)) { + /* No room in peers rwnd */ + uint32_t tsn; + + tsn = asoc->last_acked_seq + 1; + if (tsn == chk->rec.data.tsn) { + /* + * we make a special exception for this + * case. The peer has no rwnd but is missing + * the lowest chunk.. which is probably what + * is holding up the rwnd. + */ + goto one_chunk_around; + } + return (1); + } + one_chunk_around: + if (asoc->peers_rwnd < mtu) { + one_chunk = 1; + if ((asoc->peers_rwnd == 0) && + (asoc->total_flight == 0)) { + chk->window_probe = 1; + chk->whoTo->window_probe = 1; + } + } +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xC3, 2); +#endif + bundle_at = 0; + m = NULL; + net->fast_retran_ip = 0; + if (chk->rec.data.doing_fast_retransmit == 0) { + /* + * if no FR in progress skip destination that have + * flight_size > cwnd. + */ + if (net->flight_size >= net->cwnd) { + continue; + } + } else { + /* + * Mark the destination net to have FR recovery + * limits put on it. + */ + *fr_done = 1; + net->fast_retran_ip = 1; + } + + /* + * if no AUTH is yet included and this chunk requires it, + * make sure to account for it. We don't apply the size + * until the AUTH chunk is actually added below in case + * there is no room for this chunk. + */ + if (data_auth_reqd && (auth == NULL)) { + dmtu = sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); + } else + dmtu = 0; + + if ((chk->send_size <= (mtu - dmtu)) || + (chk->flags & CHUNK_FLAGS_FRAGMENT_OK)) { + /* ok we will add this one */ + if (data_auth_reqd) { + if (auth == NULL) { + m = sctp_add_auth_chunk(m, + &endofchain, + &auth, + &auth_offset, + stcb, + SCTP_DATA); + auth_keyid = chk->auth_keyid; + override_ok = 0; + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); + } else if (override_ok) { + auth_keyid = chk->auth_keyid; + override_ok = 0; + } else if (chk->auth_keyid != auth_keyid) { + /* different keyid, so done bundling */ + break; + } + } + m = sctp_copy_mbufchain(chk->data, m, &endofchain, 0, chk->send_size, chk->copy_by_ref); + if (m == NULL) { + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + /* Do clear IP_DF ? */ + if (chk->flags & CHUNK_FLAGS_FRAGMENT_OK) { + no_fragmentflg = 0; + } + /* update our MTU size */ + if (mtu > (chk->send_size + dmtu)) + mtu -= (chk->send_size + dmtu); + else + mtu = 0; + data_list[bundle_at++] = chk; + if (one_chunk && (asoc->total_flight <= 0)) { + SCTP_STAT_INCR(sctps_windowprobed); + } + } + if (one_chunk == 0) { + /* + * now are there anymore forward from chk to pick + * up? + */ + for (fwd = TAILQ_NEXT(chk, sctp_next); fwd != NULL; fwd = TAILQ_NEXT(fwd, sctp_next)) { + if (fwd->sent != SCTP_DATAGRAM_RESEND) { + /* Nope, not for retran */ + continue; + } + if (fwd->whoTo != net) { + /* Nope, not the net in question */ + continue; + } + if (data_auth_reqd && (auth == NULL)) { + dmtu = sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); + } else + dmtu = 0; + if (fwd->send_size <= (mtu - dmtu)) { + if (data_auth_reqd) { + if (auth == NULL) { + m = sctp_add_auth_chunk(m, + &endofchain, + &auth, + &auth_offset, + stcb, + SCTP_DATA); + auth_keyid = fwd->auth_keyid; + override_ok = 0; + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); + } else if (override_ok) { + auth_keyid = fwd->auth_keyid; + override_ok = 0; + } else if (fwd->auth_keyid != auth_keyid) { + /* different keyid, so done bundling */ + break; + } + } + m = sctp_copy_mbufchain(fwd->data, m, &endofchain, 0, fwd->send_size, fwd->copy_by_ref); + if (m == NULL) { + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + /* Do clear IP_DF ? */ + if (fwd->flags & CHUNK_FLAGS_FRAGMENT_OK) { + no_fragmentflg = 0; + } + /* update our MTU size */ + if (mtu > (fwd->send_size + dmtu)) + mtu -= (fwd->send_size + dmtu); + else + mtu = 0; + data_list[bundle_at++] = fwd; + if (bundle_at >= SCTP_MAX_DATA_BUNDLING) { + break; + } + } else { + /* can't fit so we are done */ + break; + } + } + } + /* Is there something to send for this destination? */ + if (m) { + /* + * No matter if we fail/or succeed we should start a + * timer. A failure is like a lost IP packet :-) + */ + if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { + /* + * no timer running on this destination + * restart it. + */ + sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); + tmr_started = 1; + } + /* Now lets send it, if there is anything to send :> */ + if ((error = sctp_lowlevel_chunk_output(inp, stcb, net, + (struct sockaddr *)&net->ro._l_addr, m, + auth_offset, auth, auth_keyid, + no_fragmentflg, 0, 0, + inp->sctp_lport, stcb->rport, htonl(stcb->asoc.peer_vtag), + net->port, NULL, +#if defined(__FreeBSD__) && !defined(__Userspace__) + 0, 0, +#endif + asoc->zero_checksum == 2, + so_locked))) { + /* error, we could not output */ + SCTPDBG(SCTP_DEBUG_OUTPUT3, "Gak send error %d\n", error); + if (error == ENOBUFS) { + asoc->ifp_had_enobuf = 1; + SCTP_STAT_INCR(sctps_lowlevelerr); + } + return (error); + } else { + asoc->ifp_had_enobuf = 0; + } + endofchain = NULL; + auth = NULL; + auth_offset = 0; + /* For HB's */ + /* + * We don't want to mark the net->sent time here + * since this we use this for HB and retrans cannot + * measure RTT + */ + /* (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); */ + + /* For auto-close */ + if (*now_filled == 0) { + (void)SCTP_GETTIME_TIMEVAL(&asoc->time_last_sent); + *now = asoc->time_last_sent; + *now_filled = 1; + } else { + asoc->time_last_sent = *now; + } + *cnt_out += bundle_at; +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xC4, bundle_at); +#endif + if (bundle_at) { + tsns_sent = data_list[0]->rec.data.tsn; + } + for (i = 0; i < bundle_at; i++) { + SCTP_STAT_INCR(sctps_sendretransdata); + data_list[i]->sent = SCTP_DATAGRAM_SENT; + /* + * When we have a revoked data, and we + * retransmit it, then we clear the revoked + * flag since this flag dictates if we + * subtracted from the fs + */ + if (data_list[i]->rec.data.chunk_was_revoked) { + /* Deflate the cwnd */ + data_list[i]->whoTo->cwnd -= data_list[i]->book_size; + data_list[i]->rec.data.chunk_was_revoked = 0; + } + data_list[i]->snd_count++; + sctp_ucount_decr(asoc->sent_queue_retran_cnt); + /* record the time */ + data_list[i]->sent_rcv_time = asoc->time_last_sent; + if (data_list[i]->book_size_scale) { + /* + * need to double the book size on + * this one + */ + data_list[i]->book_size_scale = 0; + /* Since we double the booksize, we must + * also double the output queue size, since this + * get shrunk when we free by this amount. + */ + atomic_add_int(&((asoc)->total_output_queue_size), data_list[i]->book_size); + data_list[i]->book_size *= 2; + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { + sctp_log_rwnd(SCTP_DECREASE_PEER_RWND, + asoc->peers_rwnd, data_list[i]->send_size, SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)); + } + asoc->peers_rwnd = sctp_sbspace_sub(asoc->peers_rwnd, + (uint32_t) (data_list[i]->send_size + + SCTP_BASE_SYSCTL(sctp_peer_chunk_oh))); + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_UP_RSND, + data_list[i]->whoTo->flight_size, + data_list[i]->book_size, + (uint32_t)(uintptr_t)data_list[i]->whoTo, + data_list[i]->rec.data.tsn); + } + sctp_flight_size_increase(data_list[i]); + sctp_total_flight_increase(stcb, data_list[i]); + if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { + /* SWS sender side engages */ + asoc->peers_rwnd = 0; + } + if ((i == 0) && + (data_list[i]->rec.data.doing_fast_retransmit)) { + SCTP_STAT_INCR(sctps_sendfastretrans); + if ((data_list[i] == TAILQ_FIRST(&asoc->sent_queue)) && + (tmr_started == 0)) { + /*- + * ok we just fast-retrans'd + * the lowest TSN, i.e the + * first on the list. In + * this case we want to give + * some more time to get a + * SACK back without a + * t3-expiring. + */ + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, inp, stcb, net, + SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_2); + sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); + } + } + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, net, tsns_sent, SCTP_CWND_LOG_FROM_RESEND); + } +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(21, inp, stcb, NULL); +#endif + } else { + /* None will fit */ + return (1); + } + if (asoc->sent_queue_retran_cnt <= 0) { + /* all done we have no more to retran */ + asoc->sent_queue_retran_cnt = 0; + break; + } + if (one_chunk) { + /* No more room in rwnd */ + return (1); + } + /* stop the for loop here. we sent out a packet */ + break; + } + return (0); +} + +static void +sctp_timer_validation(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_association *asoc) +{ + struct sctp_nets *net; + + /* Validate that a timer is running somewhere */ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { + /* Here is a timer */ + return; + } + } + SCTP_TCB_LOCK_ASSERT(stcb); + /* Gak, we did not have a timer somewhere */ + SCTPDBG(SCTP_DEBUG_OUTPUT3, "Deadlock avoided starting timer on a dest at retran\n"); + if (asoc->alternate) { + sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, asoc->alternate); + } else { + sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, asoc->primary_destination); + } + return; +} + +void +sctp_chunk_output(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + int from_where, + int so_locked) +{ + /*- + * Ok this is the generic chunk service queue. we must do the + * following: + * - See if there are retransmits pending, if so we must + * do these first. + * - Service the stream queue that is next, moving any + * message (note I must get a complete message i.e. + * FIRST/MIDDLE and LAST to the out queue in one pass) and assigning + * TSN's + * - Check to see if the cwnd/rwnd allows any output, if so we + * go ahead and formulate and send the low level chunks. Making sure + * to combine any control in the control chunk queue also. + */ + struct sctp_association *asoc; + struct sctp_nets *net; + int error = 0, num_out, tot_out = 0, ret = 0, reason_code; + unsigned int burst_cnt = 0; + struct timeval now; + int now_filled = 0; + int nagle_on; + uint32_t frag_point = sctp_get_frag_point(stcb); + int un_sent = 0; + int fr_done; + unsigned int tot_frs = 0; + +#if defined(__APPLE__) && !defined(__Userspace__) + if (so_locked) { + sctp_lock_assert(SCTP_INP_SO(inp)); + } else { + sctp_unlock_assert(SCTP_INP_SO(inp)); + } +#endif + asoc = &stcb->asoc; +do_it_again: + /* The Nagle algorithm is only applied when handling a send call. */ + if (from_where == SCTP_OUTPUT_FROM_USR_SEND) { + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NODELAY)) { + nagle_on = 0; + } else { + nagle_on = 1; + } + } else { + nagle_on = 0; + } + SCTP_TCB_LOCK_ASSERT(stcb); + + un_sent = (stcb->asoc.total_output_queue_size - stcb->asoc.total_flight); + + if ((un_sent <= 0) && + (TAILQ_EMPTY(&asoc->control_send_queue)) && + (TAILQ_EMPTY(&asoc->asconf_send_queue)) && + (asoc->sent_queue_retran_cnt == 0) && + (asoc->trigger_reset == 0)) { + /* Nothing to do unless there is something to be sent left */ + return; + } + /* Do we have something to send, data or control AND + * a sack timer running, if so piggy-back the sack. + */ + if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) { + sctp_send_sack(stcb, so_locked); + sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, + SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_3); + } + while (asoc->sent_queue_retran_cnt) { + /*- + * Ok, it is retransmission time only, we send out only ONE + * packet with a single call off to the retran code. + */ + if (from_where == SCTP_OUTPUT_FROM_COOKIE_ACK) { + /*- + * Special hook for handling cookies discarded + * by peer that carried data. Send cookie-ack only + * and then the next call with get the retran's. + */ + (void)sctp_med_chunk_output(inp, stcb, asoc, &num_out, &reason_code, 1, + from_where, + &now, &now_filled, frag_point, so_locked); + return; + } else if (from_where != SCTP_OUTPUT_FROM_HB_TMR) { + /* if its not from a HB then do it */ + fr_done = 0; + ret = sctp_chunk_retransmission(inp, stcb, asoc, &num_out, &now, &now_filled, &fr_done, so_locked); + if (fr_done) { + tot_frs++; + } + } else { + /* + * its from any other place, we don't allow retran + * output (only control) + */ + ret = 1; + } + if (ret > 0) { + /* Can't send anymore */ + /*- + * now lets push out control by calling med-level + * output once. this assures that we WILL send HB's + * if queued too. + */ + (void)sctp_med_chunk_output(inp, stcb, asoc, &num_out, &reason_code, 1, + from_where, + &now, &now_filled, frag_point, so_locked); +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(8, inp, stcb, NULL); +#endif + sctp_timer_validation(inp, stcb, asoc); + return; + } + if (ret < 0) { + /*- + * The count was off.. retran is not happening so do + * the normal retransmission. + */ +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(9, inp, stcb, NULL); +#endif + if (ret == SCTP_RETRAN_EXIT) { + return; + } + break; + } + if (from_where == SCTP_OUTPUT_FROM_T3) { + /* Only one transmission allowed out of a timeout */ +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(10, inp, stcb, NULL); +#endif + /* Push out any control */ + (void)sctp_med_chunk_output(inp, stcb, asoc, &num_out, &reason_code, 1, from_where, + &now, &now_filled, frag_point, so_locked); + return; + } + if ((asoc->fr_max_burst > 0) && (tot_frs >= asoc->fr_max_burst)) { + /* Hit FR burst limit */ + return; + } + if ((num_out == 0) && (ret == 0)) { + /* No more retrans to send */ + break; + } + } +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(12, inp, stcb, NULL); +#endif + /* Check for bad destinations, if they exist move chunks around. */ + TAILQ_FOREACH(net, &asoc->nets, sctp_next) { + if ((net->dest_state & SCTP_ADDR_REACHABLE) == 0) { + /*- + * if possible move things off of this address we + * still may send below due to the dormant state but + * we try to find an alternate address to send to + * and if we have one we move all queued data on the + * out wheel to this alternate address. + */ + if (net->ref_count > 1) + sctp_move_chunks_from_net(stcb, net); + } else { + /*- + * if ((asoc->sat_network) || (net->addr_is_local)) + * { burst_limit = asoc->max_burst * + * SCTP_SAT_NETWORK_BURST_INCR; } + */ + if (asoc->max_burst > 0) { + if (SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst)) { + if ((net->flight_size + (asoc->max_burst * net->mtu)) < net->cwnd) { + /* JRS - Use the congestion control given in the congestion control module */ + asoc->cc_functions.sctp_cwnd_update_after_output(stcb, net, asoc->max_burst); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_MAXBURST_ENABLE) { + sctp_log_maxburst(stcb, net, 0, asoc->max_burst, SCTP_MAX_BURST_APPLIED); + } + SCTP_STAT_INCR(sctps_maxburstqueued); + } + net->fast_retran_ip = 0; + } else { + if (net->flight_size == 0) { + /* Should be decaying the cwnd here */ + ; + } + } + } + } + } + burst_cnt = 0; + do { + error = sctp_med_chunk_output(inp, stcb, asoc, &num_out, + &reason_code, 0, from_where, + &now, &now_filled, frag_point, so_locked); + if (error) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Error %d was returned from med-c-op\n", error); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_MAXBURST_ENABLE) { + sctp_log_maxburst(stcb, asoc->primary_destination, error, burst_cnt, SCTP_MAX_BURST_ERROR_STOP); + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, NULL, error, SCTP_SEND_NOW_COMPLETES); + sctp_log_cwnd(stcb, NULL, 0xdeadbeef, SCTP_SEND_NOW_COMPLETES); + } + break; + } + SCTPDBG(SCTP_DEBUG_OUTPUT3, "m-c-o put out %d\n", num_out); + + tot_out += num_out; + burst_cnt++; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, NULL, num_out, SCTP_SEND_NOW_COMPLETES); + if (num_out == 0) { + sctp_log_cwnd(stcb, NULL, reason_code, SCTP_SEND_NOW_COMPLETES); + } + } + if (nagle_on) { + /* + * When the Nagle algorithm is used, look at how much + * is unsent, then if its smaller than an MTU and we + * have data in flight we stop, except if we are + * handling a fragmented user message. + */ + un_sent = stcb->asoc.total_output_queue_size - stcb->asoc.total_flight; + if ((un_sent < (int)(stcb->asoc.smallest_mtu - SCTP_MIN_OVERHEAD)) && + (stcb->asoc.total_flight > 0)) { +/* && sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR))) {*/ + break; + } + } + if (TAILQ_EMPTY(&asoc->control_send_queue) && + TAILQ_EMPTY(&asoc->send_queue) && + sctp_is_there_unsent_data(stcb, so_locked) == 0) { + /* Nothing left to send */ + break; + } + if ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) <= 0) { + /* Nothing left to send */ + break; + } + } while (num_out && + ((asoc->max_burst == 0) || + SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst) || + (burst_cnt < asoc->max_burst))); + + if (SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst) == 0) { + if ((asoc->max_burst > 0) && (burst_cnt >= asoc->max_burst)) { + SCTP_STAT_INCR(sctps_maxburstqueued); + asoc->burst_limit_applied = 1; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_MAXBURST_ENABLE) { + sctp_log_maxburst(stcb, asoc->primary_destination, 0, burst_cnt, SCTP_MAX_BURST_APPLIED); + } + } else { + asoc->burst_limit_applied = 0; + } + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + sctp_log_cwnd(stcb, NULL, tot_out, SCTP_SEND_NOW_COMPLETES); + } + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Ok, we have put out %d chunks\n", + tot_out); + + /*- + * Now we need to clean up the control chunk chain if a ECNE is on + * it. It must be marked as UNSENT again so next call will continue + * to send it until such time that we get a CWR, to remove it. + */ + if (stcb->asoc.ecn_echo_cnt_onq) + sctp_fix_ecn_echo(asoc); + + if (stcb->asoc.trigger_reset) { + if (sctp_send_stream_reset_out_if_possible(stcb, so_locked) == 0) { + goto do_it_again; + } + } + return; +} + +int +sctp_output( + struct sctp_inpcb *inp, + struct mbuf *m, + struct sockaddr *addr, + struct mbuf *control, +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct thread *p, +#elif defined(_WIN32) && !defined(__Userspace__) + PKTHREAD p, +#else +#if defined(__APPLE__) && !defined(__Userspace__) + struct proc *p SCTP_UNUSED, +#else + struct proc *p, +#endif +#endif + int flags) +{ + if (inp == NULL) { + SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); + return (EINVAL); + } + + if (inp->sctp_socket == NULL) { + SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); + return (EINVAL); + } + return (sctp_sosend(inp->sctp_socket, + addr, + (struct uio *)NULL, + m, + control, +#if defined(__APPLE__) && !defined(__Userspace__) + flags +#else + flags, p +#endif + )); +} + +void +send_forward_tsn(struct sctp_tcb *stcb, + struct sctp_association *asoc) +{ + struct sctp_tmit_chunk *chk, *at, *tp1, *last; + struct sctp_forward_tsn_chunk *fwdtsn; + struct sctp_strseq *strseq; + struct sctp_strseq_mid *strseq_m; + uint32_t advance_peer_ack_point; + unsigned int cnt_of_space, i, ovh; + unsigned int space_needed; + unsigned int cnt_of_skipped = 0; + + SCTP_TCB_LOCK_ASSERT(stcb); + TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { + if (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN) { + /* mark it to unsent */ + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + /* Do we correct its output location? */ + if (chk->whoTo) { + sctp_free_remote_addr(chk->whoTo); + chk->whoTo = NULL; + } + goto sctp_fill_in_rest; + } + } + /* Ok if we reach here we must build one */ + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + return; + } + asoc->fwd_tsn_cnt++; + chk->copy_by_ref = 0; + /* + * We don't do the old thing here since + * this is used not for on-wire but to + * tell if we are sending a fwd-tsn by + * the stack during output. And if its + * a IFORWARD or a FORWARD it is a fwd-tsn. + */ + chk->rec.chunk_id.id = SCTP_FORWARD_CUM_TSN; + chk->rec.chunk_id.can_take_data = 0; + chk->flags = 0; + chk->asoc = asoc; + chk->whoTo = NULL; + chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_NOWAIT, 1, MT_DATA); + if (chk->data == NULL) { + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + return; + } + SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + TAILQ_INSERT_TAIL(&asoc->control_send_queue, chk, sctp_next); + asoc->ctrl_queue_cnt++; +sctp_fill_in_rest: + /*- + * Here we go through and fill out the part that deals with + * stream/seq of the ones we skip. + */ + SCTP_BUF_LEN(chk->data) = 0; + TAILQ_FOREACH(at, &asoc->sent_queue, sctp_next) { + if ((at->sent != SCTP_FORWARD_TSN_SKIP) && + (at->sent != SCTP_DATAGRAM_NR_ACKED)) { + /* no more to look at */ + break; + } + if (!asoc->idata_supported && (at->rec.data.rcv_flags & SCTP_DATA_UNORDERED)) { + /* We don't report these */ + continue; + } + cnt_of_skipped++; + } + if (asoc->idata_supported) { + space_needed = (sizeof(struct sctp_forward_tsn_chunk) + + (cnt_of_skipped * sizeof(struct sctp_strseq_mid))); + } else { + space_needed = (sizeof(struct sctp_forward_tsn_chunk) + + (cnt_of_skipped * sizeof(struct sctp_strseq))); + } + cnt_of_space = (unsigned int)M_TRAILINGSPACE(chk->data); + + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + ovh = SCTP_MIN_OVERHEAD; + } else { + ovh = SCTP_MIN_V4_OVERHEAD; + } + if (cnt_of_space > (asoc->smallest_mtu - ovh)) { + /* trim to a mtu size */ + cnt_of_space = asoc->smallest_mtu - ovh; + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_TRY_ADVANCE) { + sctp_misc_ints(SCTP_FWD_TSN_CHECK, + 0xff, 0, cnt_of_skipped, + asoc->advanced_peer_ack_point); + } + advance_peer_ack_point = asoc->advanced_peer_ack_point; + if (cnt_of_space < space_needed) { + /*- + * ok we must trim down the chunk by lowering the + * advance peer ack point. + */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_TRY_ADVANCE) { + sctp_misc_ints(SCTP_FWD_TSN_CHECK, + 0xff, 0xff, cnt_of_space, + space_needed); + } + cnt_of_skipped = cnt_of_space - sizeof(struct sctp_forward_tsn_chunk); + if (asoc->idata_supported) { + cnt_of_skipped /= sizeof(struct sctp_strseq_mid); + } else { + cnt_of_skipped /= sizeof(struct sctp_strseq); + } + /*- + * Go through and find the TSN that will be the one + * we report. + */ + at = TAILQ_FIRST(&asoc->sent_queue); + if (at != NULL) { + for (i = 0; i < cnt_of_skipped; i++) { + tp1 = TAILQ_NEXT(at, sctp_next); + if (tp1 == NULL) { + break; + } + at = tp1; + } + } + if (at && SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_TRY_ADVANCE) { + sctp_misc_ints(SCTP_FWD_TSN_CHECK, + 0xff, cnt_of_skipped, at->rec.data.tsn, + asoc->advanced_peer_ack_point); + } + last = at; + /*- + * last now points to last one I can report, update + * peer ack point + */ + if (last) { + advance_peer_ack_point = last->rec.data.tsn; + } + if (asoc->idata_supported) { + space_needed = sizeof(struct sctp_forward_tsn_chunk) + + cnt_of_skipped * sizeof(struct sctp_strseq_mid); + } else { + space_needed = sizeof(struct sctp_forward_tsn_chunk) + + cnt_of_skipped * sizeof(struct sctp_strseq); + } + } + chk->send_size = space_needed; + /* Setup the chunk */ + fwdtsn = mtod(chk->data, struct sctp_forward_tsn_chunk *); + fwdtsn->ch.chunk_length = htons(chk->send_size); + fwdtsn->ch.chunk_flags = 0; + if (asoc->idata_supported) { + fwdtsn->ch.chunk_type = SCTP_IFORWARD_CUM_TSN; + } else { + fwdtsn->ch.chunk_type = SCTP_FORWARD_CUM_TSN; + } + fwdtsn->new_cumulative_tsn = htonl(advance_peer_ack_point); + SCTP_BUF_LEN(chk->data) = chk->send_size; + fwdtsn++; + /*- + * Move pointer to after the fwdtsn and transfer to the + * strseq pointer. + */ + if (asoc->idata_supported) { + strseq_m = (struct sctp_strseq_mid *)fwdtsn; + strseq = NULL; + } else { + strseq = (struct sctp_strseq *)fwdtsn; + strseq_m = NULL; + } + /*- + * Now populate the strseq list. This is done blindly + * without pulling out duplicate stream info. This is + * inefficient but won't harm the process since the peer will + * look at these in sequence and will thus release anything. + * It could mean we exceed the PMTU and chop off some that + * we could have included.. but this is unlikely (aka 1432/4 + * would mean 300+ stream seq's would have to be reported in + * one FWD-TSN. With a bit of work we can later FIX this to + * optimize and pull out duplicates.. but it does add more + * overhead. So for now... not! + */ + i = 0; + TAILQ_FOREACH(at, &asoc->sent_queue, sctp_next) { + if (i >= cnt_of_skipped) { + break; + } + if (!asoc->idata_supported && (at->rec.data.rcv_flags & SCTP_DATA_UNORDERED)) { + /* We don't report these */ + continue; + } + if (at->rec.data.tsn == advance_peer_ack_point) { + at->rec.data.fwd_tsn_cnt = 0; + } + if (asoc->idata_supported) { + strseq_m->sid = htons(at->rec.data.sid); + if (at->rec.data.rcv_flags & SCTP_DATA_UNORDERED) { + strseq_m->flags = htons(PR_SCTP_UNORDERED_FLAG); + } else { + strseq_m->flags = 0; + } + strseq_m->mid = htonl(at->rec.data.mid); + strseq_m++; + } else { + strseq->sid = htons(at->rec.data.sid); + strseq->ssn = htons((uint16_t)at->rec.data.mid); + strseq++; + } + i++; + } + return; +} + +void +sctp_send_sack(struct sctp_tcb *stcb, int so_locked) +{ + /*- + * Queue up a SACK or NR-SACK in the control queue. + * We must first check to see if a SACK or NR-SACK is + * somehow on the control queue. + * If so, we will take and and remove the old one. + */ + struct sctp_association *asoc; + struct sctp_tmit_chunk *chk, *a_chk; + struct sctp_sack_chunk *sack; + struct sctp_nr_sack_chunk *nr_sack; + struct sctp_gap_ack_block *gap_descriptor; + const struct sack_track *selector; + int mergeable = 0; + int offset; + caddr_t limit; + uint32_t *dup; + int limit_reached = 0; + unsigned int i, siz, j; + unsigned int num_gap_blocks = 0, num_nr_gap_blocks = 0, space; + int num_dups = 0; + int space_req; + uint32_t highest_tsn; + uint8_t flags; + uint8_t type; + uint8_t tsn_map; + + if (stcb->asoc.nrsack_supported == 1) { + type = SCTP_NR_SELECTIVE_ACK; + } else { + type = SCTP_SELECTIVE_ACK; + } + a_chk = NULL; + asoc = &stcb->asoc; + SCTP_TCB_LOCK_ASSERT(stcb); + if (asoc->last_data_chunk_from == NULL) { + /* Hmm we never received anything */ + return; + } + sctp_slide_mapping_arrays(stcb); + sctp_set_rwnd(stcb, asoc); + TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { + if (chk->rec.chunk_id.id == type) { + /* Hmm, found a sack already on queue, remove it */ + TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); + asoc->ctrl_queue_cnt--; + a_chk = chk; + if (a_chk->data) { + sctp_m_freem(a_chk->data); + a_chk->data = NULL; + } + if (a_chk->whoTo) { + sctp_free_remote_addr(a_chk->whoTo); + a_chk->whoTo = NULL; + } + break; + } + } + if (a_chk == NULL) { + sctp_alloc_a_chunk(stcb, a_chk); + if (a_chk == NULL) { + /* No memory so we drop the idea, and set a timer */ + if (stcb->asoc.delayed_ack) { + sctp_timer_stop(SCTP_TIMER_TYPE_RECV, + stcb->sctp_ep, stcb, NULL, + SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_4); + sctp_timer_start(SCTP_TIMER_TYPE_RECV, + stcb->sctp_ep, stcb, NULL); + } else { + stcb->asoc.send_sack = 1; + } + return; + } + a_chk->copy_by_ref = 0; + a_chk->rec.chunk_id.id = type; + a_chk->rec.chunk_id.can_take_data = 1; + } + /* Clear our pkt counts */ + asoc->data_pkts_seen = 0; + + a_chk->flags = 0; + a_chk->asoc = asoc; + a_chk->snd_count = 0; + a_chk->send_size = 0; /* fill in later */ + a_chk->sent = SCTP_DATAGRAM_UNSENT; + a_chk->whoTo = NULL; + + if ((asoc->last_data_chunk_from->dest_state & SCTP_ADDR_REACHABLE) == 0) { + /*- + * Ok, the destination for the SACK is unreachable, lets see if + * we can select an alternate to asoc->last_data_chunk_from + */ + a_chk->whoTo = sctp_find_alternate_net(stcb, asoc->last_data_chunk_from, 0); + if (a_chk->whoTo == NULL) { + /* Nope, no alternate */ + a_chk->whoTo = asoc->last_data_chunk_from; + } + } else { + a_chk->whoTo = asoc->last_data_chunk_from; + } + if (a_chk->whoTo) { + atomic_add_int(&a_chk->whoTo->ref_count, 1); + } + if (SCTP_TSN_GT(asoc->highest_tsn_inside_map, asoc->highest_tsn_inside_nr_map)) { + highest_tsn = asoc->highest_tsn_inside_map; + } else { + highest_tsn = asoc->highest_tsn_inside_nr_map; + } + if (highest_tsn == asoc->cumulative_tsn) { + /* no gaps */ + if (type == SCTP_SELECTIVE_ACK) { + space_req = sizeof(struct sctp_sack_chunk); + } else { + space_req = sizeof(struct sctp_nr_sack_chunk); + } + } else { + /* gaps get a cluster */ + space_req = MCLBYTES; + } + /* Ok now lets formulate a MBUF with our sack */ + a_chk->data = sctp_get_mbuf_for_msg(space_req, 0, M_NOWAIT, 1, MT_DATA); + if ((a_chk->data == NULL) || + (a_chk->whoTo == NULL)) { + /* rats, no mbuf memory */ + if (a_chk->data) { + /* was a problem with the destination */ + sctp_m_freem(a_chk->data); + a_chk->data = NULL; + } + sctp_free_a_chunk(stcb, a_chk, so_locked); + /* sa_ignore NO_NULL_CHK */ + if (stcb->asoc.delayed_ack) { + sctp_timer_stop(SCTP_TIMER_TYPE_RECV, + stcb->sctp_ep, stcb, NULL, + SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_5); + sctp_timer_start(SCTP_TIMER_TYPE_RECV, + stcb->sctp_ep, stcb, NULL); + } else { + stcb->asoc.send_sack = 1; + } + return; + } + /* ok, lets go through and fill it in */ + SCTP_BUF_RESV_UF(a_chk->data, SCTP_MIN_OVERHEAD); + space = (unsigned int)M_TRAILINGSPACE(a_chk->data); + if (space > (a_chk->whoTo->mtu - SCTP_MIN_OVERHEAD)) { + space = (a_chk->whoTo->mtu - SCTP_MIN_OVERHEAD); + } + limit = mtod(a_chk->data, caddr_t); + limit += space; + + flags = 0; + + if ((asoc->sctp_cmt_on_off > 0) && + SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) { + /*- + * CMT DAC algorithm: If 2 (i.e., 0x10) packets have been + * received, then set high bit to 1, else 0. Reset + * pkts_rcvd. + */ + flags |= (asoc->cmt_dac_pkts_rcvd << 6); + asoc->cmt_dac_pkts_rcvd = 0; + } +#ifdef SCTP_ASOCLOG_OF_TSNS + stcb->asoc.cumack_logsnt[stcb->asoc.cumack_log_atsnt] = asoc->cumulative_tsn; + stcb->asoc.cumack_log_atsnt++; + if (stcb->asoc.cumack_log_atsnt >= SCTP_TSN_LOG_SIZE) { + stcb->asoc.cumack_log_atsnt = 0; + } +#endif + /* reset the readers interpretation */ + stcb->freed_by_sorcv_sincelast = 0; + + if (type == SCTP_SELECTIVE_ACK) { + sack = mtod(a_chk->data, struct sctp_sack_chunk *); + nr_sack = NULL; + gap_descriptor = (struct sctp_gap_ack_block *)((caddr_t)sack + sizeof(struct sctp_sack_chunk)); + if (highest_tsn > asoc->mapping_array_base_tsn) { + siz = (((highest_tsn - asoc->mapping_array_base_tsn) + 1) + 7) / 8; + } else { + siz = (((MAX_TSN - asoc->mapping_array_base_tsn) + 1) + highest_tsn + 7) / 8; + } + } else { + sack = NULL; + nr_sack = mtod(a_chk->data, struct sctp_nr_sack_chunk *); + gap_descriptor = (struct sctp_gap_ack_block *)((caddr_t)nr_sack + sizeof(struct sctp_nr_sack_chunk)); + if (asoc->highest_tsn_inside_map > asoc->mapping_array_base_tsn) { + siz = (((asoc->highest_tsn_inside_map - asoc->mapping_array_base_tsn) + 1) + 7) / 8; + } else { + siz = (((MAX_TSN - asoc->mapping_array_base_tsn) + 1) + asoc->highest_tsn_inside_map + 7) / 8; + } + } + + if (SCTP_TSN_GT(asoc->mapping_array_base_tsn, asoc->cumulative_tsn)) { + offset = 1; + } else { + offset = asoc->mapping_array_base_tsn - asoc->cumulative_tsn; + } + if (((type == SCTP_SELECTIVE_ACK) && + SCTP_TSN_GT(highest_tsn, asoc->cumulative_tsn)) || + ((type == SCTP_NR_SELECTIVE_ACK) && + SCTP_TSN_GT(asoc->highest_tsn_inside_map, asoc->cumulative_tsn))) { + /* we have a gap .. maybe */ + for (i = 0; i < siz; i++) { + tsn_map = asoc->mapping_array[i]; + if (type == SCTP_SELECTIVE_ACK) { + tsn_map |= asoc->nr_mapping_array[i]; + } + if (i == 0) { + /* + * Clear all bits corresponding to TSNs + * smaller or equal to the cumulative TSN. + */ + tsn_map &= (~0U << (1 - offset)); + } + selector = &sack_array[tsn_map]; + if (mergeable && selector->right_edge) { + /* + * Backup, left and right edges were ok to + * merge. + */ + num_gap_blocks--; + gap_descriptor--; + } + if (selector->num_entries == 0) + mergeable = 0; + else { + for (j = 0; j < selector->num_entries; j++) { + if (mergeable && selector->right_edge) { + /* + * do a merge by NOT setting + * the left side + */ + mergeable = 0; + } else { + /* + * no merge, set the left + * side + */ + mergeable = 0; + gap_descriptor->start = htons((selector->gaps[j].start + offset)); + } + gap_descriptor->end = htons((selector->gaps[j].end + offset)); + num_gap_blocks++; + gap_descriptor++; + if (((caddr_t)gap_descriptor + sizeof(struct sctp_gap_ack_block)) > limit) { + /* no more room */ + limit_reached = 1; + break; + } + } + if (selector->left_edge) { + mergeable = 1; + } + } + if (limit_reached) { + /* Reached the limit stop */ + break; + } + offset += 8; + } + } + if ((type == SCTP_NR_SELECTIVE_ACK) && + (limit_reached == 0)) { + mergeable = 0; + + if (asoc->highest_tsn_inside_nr_map > asoc->mapping_array_base_tsn) { + siz = (((asoc->highest_tsn_inside_nr_map - asoc->mapping_array_base_tsn) + 1) + 7) / 8; + } else { + siz = (((MAX_TSN - asoc->mapping_array_base_tsn) + 1) + asoc->highest_tsn_inside_nr_map + 7) / 8; + } + + if (SCTP_TSN_GT(asoc->mapping_array_base_tsn, asoc->cumulative_tsn)) { + offset = 1; + } else { + offset = asoc->mapping_array_base_tsn - asoc->cumulative_tsn; + } + if (SCTP_TSN_GT(asoc->highest_tsn_inside_nr_map, asoc->cumulative_tsn)) { + /* we have a gap .. maybe */ + for (i = 0; i < siz; i++) { + tsn_map = asoc->nr_mapping_array[i]; + if (i == 0) { + /* + * Clear all bits corresponding to TSNs + * smaller or equal to the cumulative TSN. + */ + tsn_map &= (~0U << (1 - offset)); + } + selector = &sack_array[tsn_map]; + if (mergeable && selector->right_edge) { + /* + * Backup, left and right edges were ok to + * merge. + */ + num_nr_gap_blocks--; + gap_descriptor--; + } + if (selector->num_entries == 0) + mergeable = 0; + else { + for (j = 0; j < selector->num_entries; j++) { + if (mergeable && selector->right_edge) { + /* + * do a merge by NOT setting + * the left side + */ + mergeable = 0; + } else { + /* + * no merge, set the left + * side + */ + mergeable = 0; + gap_descriptor->start = htons((selector->gaps[j].start + offset)); + } + gap_descriptor->end = htons((selector->gaps[j].end + offset)); + num_nr_gap_blocks++; + gap_descriptor++; + if (((caddr_t)gap_descriptor + sizeof(struct sctp_gap_ack_block)) > limit) { + /* no more room */ + limit_reached = 1; + break; + } + } + if (selector->left_edge) { + mergeable = 1; + } + } + if (limit_reached) { + /* Reached the limit stop */ + break; + } + offset += 8; + } + } + } + /* now we must add any dups we are going to report. */ + if ((limit_reached == 0) && (asoc->numduptsns)) { + dup = (uint32_t *) gap_descriptor; + for (i = 0; i < asoc->numduptsns; i++) { + *dup = htonl(asoc->dup_tsns[i]); + dup++; + num_dups++; + if (((caddr_t)dup + sizeof(uint32_t)) > limit) { + /* no more room */ + break; + } + } + asoc->numduptsns = 0; + } + /* + * now that the chunk is prepared queue it to the control chunk + * queue. + */ + if (type == SCTP_SELECTIVE_ACK) { + a_chk->send_size = (uint16_t)(sizeof(struct sctp_sack_chunk) + + (num_gap_blocks + num_nr_gap_blocks) * sizeof(struct sctp_gap_ack_block) + + num_dups * sizeof(int32_t)); + SCTP_BUF_LEN(a_chk->data) = a_chk->send_size; + sack->sack.cum_tsn_ack = htonl(asoc->cumulative_tsn); + sack->sack.a_rwnd = htonl(asoc->my_rwnd); + sack->sack.num_gap_ack_blks = htons(num_gap_blocks); + sack->sack.num_dup_tsns = htons(num_dups); + sack->ch.chunk_type = type; + sack->ch.chunk_flags = flags; + sack->ch.chunk_length = htons(a_chk->send_size); + } else { + a_chk->send_size = (uint16_t)(sizeof(struct sctp_nr_sack_chunk) + + (num_gap_blocks + num_nr_gap_blocks) * sizeof(struct sctp_gap_ack_block) + + num_dups * sizeof(int32_t)); + SCTP_BUF_LEN(a_chk->data) = a_chk->send_size; + nr_sack->nr_sack.cum_tsn_ack = htonl(asoc->cumulative_tsn); + nr_sack->nr_sack.a_rwnd = htonl(asoc->my_rwnd); + nr_sack->nr_sack.num_gap_ack_blks = htons(num_gap_blocks); + nr_sack->nr_sack.num_nr_gap_ack_blks = htons(num_nr_gap_blocks); + nr_sack->nr_sack.num_dup_tsns = htons(num_dups); + nr_sack->nr_sack.reserved = 0; + nr_sack->ch.chunk_type = type; + nr_sack->ch.chunk_flags = flags; + nr_sack->ch.chunk_length = htons(a_chk->send_size); + } + TAILQ_INSERT_TAIL(&asoc->control_send_queue, a_chk, sctp_next); + asoc->my_last_reported_rwnd = asoc->my_rwnd; + asoc->ctrl_queue_cnt++; + asoc->send_sack = 0; + SCTP_STAT_INCR(sctps_sendsacks); + return; +} + +void +sctp_send_abort_tcb(struct sctp_tcb *stcb, struct mbuf *operr, int so_locked) +{ + struct mbuf *m_abort, *m, *m_last; + struct mbuf *m_out, *m_end = NULL; + struct sctp_abort_chunk *abort; + struct sctp_auth_chunk *auth = NULL; + struct sctp_nets *net; + uint32_t vtag; + uint32_t auth_offset = 0; + int error; + uint16_t cause_len, chunk_len, padding_len; + +#if defined(__APPLE__) && !defined(__Userspace__) + if (so_locked) { + sctp_lock_assert(SCTP_INP_SO(stcb->sctp_ep)); + } else { + sctp_unlock_assert(SCTP_INP_SO(stcb->sctp_ep)); + } +#endif + SCTP_TCB_LOCK_ASSERT(stcb); + /*- + * Add an AUTH chunk, if chunk requires it and save the offset into + * the chain for AUTH + */ + if (sctp_auth_is_required_chunk(SCTP_ABORT_ASSOCIATION, + stcb->asoc.peer_auth_chunks)) { + m_out = sctp_add_auth_chunk(NULL, &m_end, &auth, &auth_offset, + stcb, SCTP_ABORT_ASSOCIATION); + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); + } else { + m_out = NULL; + } + m_abort = sctp_get_mbuf_for_msg(sizeof(struct sctp_abort_chunk), 0, M_NOWAIT, 1, MT_HEADER); + if (m_abort == NULL) { + if (m_out) { + sctp_m_freem(m_out); + } + if (operr) { + sctp_m_freem(operr); + } + return; + } + /* link in any error */ + SCTP_BUF_NEXT(m_abort) = operr; + cause_len = 0; + m_last = NULL; + for (m = operr; m; m = SCTP_BUF_NEXT(m)) { + cause_len += (uint16_t)SCTP_BUF_LEN(m); + if (SCTP_BUF_NEXT(m) == NULL) { + m_last = m; + } + } + SCTP_BUF_LEN(m_abort) = sizeof(struct sctp_abort_chunk); + chunk_len = (uint16_t)sizeof(struct sctp_abort_chunk) + cause_len; + padding_len = SCTP_SIZE32(chunk_len) - chunk_len; + if (m_out == NULL) { + /* NO Auth chunk prepended, so reserve space in front */ + SCTP_BUF_RESV_UF(m_abort, SCTP_MIN_OVERHEAD); + m_out = m_abort; + } else { + /* Put AUTH chunk at the front of the chain */ + SCTP_BUF_NEXT(m_end) = m_abort; + } + if (stcb->asoc.alternate) { + net = stcb->asoc.alternate; + } else { + net = stcb->asoc.primary_destination; + } + /* Fill in the ABORT chunk header. */ + abort = mtod(m_abort, struct sctp_abort_chunk *); + abort->ch.chunk_type = SCTP_ABORT_ASSOCIATION; + if (stcb->asoc.peer_vtag == 0) { + /* This happens iff the assoc is in COOKIE-WAIT state. */ + vtag = stcb->asoc.my_vtag; + abort->ch.chunk_flags = SCTP_HAD_NO_TCB; + } else { + vtag = stcb->asoc.peer_vtag; + abort->ch.chunk_flags = 0; + } + abort->ch.chunk_length = htons(chunk_len); + /* Add padding, if necessary. */ + if (padding_len > 0) { + if ((m_last == NULL) || + (sctp_add_pad_tombuf(m_last, padding_len) == NULL)) { + sctp_m_freem(m_out); + return; + } + } + if ((error = sctp_lowlevel_chunk_output(stcb->sctp_ep, stcb, net, + (struct sockaddr *)&net->ro._l_addr, + m_out, auth_offset, auth, stcb->asoc.authinfo.active_keyid, 1, 0, 0, + stcb->sctp_ep->sctp_lport, stcb->rport, htonl(vtag), + stcb->asoc.primary_destination->port, NULL, +#if defined(__FreeBSD__) && !defined(__Userspace__) + 0, 0, +#endif + stcb->asoc.zero_checksum == 2, + so_locked))) { + SCTPDBG(SCTP_DEBUG_OUTPUT3, "Gak send error %d\n", error); + if (error == ENOBUFS) { + stcb->asoc.ifp_had_enobuf = 1; + SCTP_STAT_INCR(sctps_lowlevelerr); + } + } else { + stcb->asoc.ifp_had_enobuf = 0; + } + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); +} + +void +sctp_send_shutdown_complete(struct sctp_tcb *stcb, + struct sctp_nets *net, + int reflect_vtag) +{ + /* formulate and SEND a SHUTDOWN-COMPLETE */ + struct mbuf *m_shutdown_comp; + struct sctp_shutdown_complete_chunk *shutdown_complete; + uint32_t vtag; + int error; + uint8_t flags; + + m_shutdown_comp = sctp_get_mbuf_for_msg(sizeof(struct sctp_chunkhdr), 0, M_NOWAIT, 1, MT_HEADER); + if (m_shutdown_comp == NULL) { + /* no mbuf's */ + return; + } + if (reflect_vtag) { + flags = SCTP_HAD_NO_TCB; + vtag = stcb->asoc.my_vtag; + } else { + flags = 0; + vtag = stcb->asoc.peer_vtag; + } + shutdown_complete = mtod(m_shutdown_comp, struct sctp_shutdown_complete_chunk *); + shutdown_complete->ch.chunk_type = SCTP_SHUTDOWN_COMPLETE; + shutdown_complete->ch.chunk_flags = flags; + shutdown_complete->ch.chunk_length = htons(sizeof(struct sctp_shutdown_complete_chunk)); + SCTP_BUF_LEN(m_shutdown_comp) = sizeof(struct sctp_shutdown_complete_chunk); + if ((error = sctp_lowlevel_chunk_output(stcb->sctp_ep, stcb, net, + (struct sockaddr *)&net->ro._l_addr, + m_shutdown_comp, 0, NULL, 0, 1, 0, 0, + stcb->sctp_ep->sctp_lport, stcb->rport, + htonl(vtag), + net->port, NULL, +#if defined(__FreeBSD__) && !defined(__Userspace__) + 0, 0, +#endif + stcb->asoc.zero_checksum == 2, + SCTP_SO_NOT_LOCKED))) { + SCTPDBG(SCTP_DEBUG_OUTPUT3, "Gak send error %d\n", error); + if (error == ENOBUFS) { + stcb->asoc.ifp_had_enobuf = 1; + SCTP_STAT_INCR(sctps_lowlevelerr); + } + } else { + stcb->asoc.ifp_had_enobuf = 0; + } + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); + return; +} + +#if defined(__FreeBSD__) && !defined(__Userspace__) +static void +sctp_send_resp_msg(struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, uint32_t vtag, + uint8_t type, struct mbuf *cause, + uint8_t mflowtype, uint32_t mflowid, uint16_t fibnum, + uint32_t vrf_id, uint16_t port) +#else +static void +sctp_send_resp_msg(struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, uint32_t vtag, + uint8_t type, struct mbuf *cause, + uint32_t vrf_id SCTP_UNUSED, uint16_t port) +#endif +{ + struct mbuf *o_pak; + struct mbuf *mout; + struct sctphdr *shout; + struct sctp_chunkhdr *ch; +#if defined(INET) || defined(INET6) + struct udphdr *udp; +#endif + int ret, len, cause_len, padding_len; +#ifdef INET +#if defined(__APPLE__) && !defined(__Userspace__) + sctp_route_t ro; +#endif + struct sockaddr_in *src_sin, *dst_sin; + struct ip *ip; +#endif +#ifdef INET6 + struct sockaddr_in6 *src_sin6, *dst_sin6; + struct ip6_hdr *ip6; +#endif + + /* Compute the length of the cause and add final padding. */ + cause_len = 0; + if (cause != NULL) { + struct mbuf *m_at, *m_last = NULL; + + for (m_at = cause; m_at; m_at = SCTP_BUF_NEXT(m_at)) { + if (SCTP_BUF_NEXT(m_at) == NULL) + m_last = m_at; + cause_len += SCTP_BUF_LEN(m_at); + } + padding_len = cause_len % 4; + if (padding_len != 0) { + padding_len = 4 - padding_len; + } + if (padding_len != 0) { + if (sctp_add_pad_tombuf(m_last, padding_len) == NULL) { + sctp_m_freem(cause); + return; + } + } + } else { + padding_len = 0; + } + /* Get an mbuf for the header. */ + len = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); + switch (dst->sa_family) { +#ifdef INET + case AF_INET: + len += sizeof(struct ip); + break; +#endif +#ifdef INET6 + case AF_INET6: + len += sizeof(struct ip6_hdr); + break; +#endif + default: + break; + } +#if defined(INET) || defined(INET6) + if (port) { + len += sizeof(struct udphdr); + } +#endif +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) + mout = sctp_get_mbuf_for_msg(len + max_linkhdr, 1, M_NOWAIT, 1, MT_DATA); +#else + mout = sctp_get_mbuf_for_msg(len + SCTP_MAX_LINKHDR, 1, M_NOWAIT, 1, MT_DATA); +#endif +#else + mout = sctp_get_mbuf_for_msg(len + max_linkhdr, 1, M_NOWAIT, 1, MT_DATA); +#endif + if (mout == NULL) { + if (cause) { + sctp_m_freem(cause); + } + return; + } +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) + SCTP_BUF_RESV_UF(mout, max_linkhdr); +#else + SCTP_BUF_RESV_UF(mout, SCTP_MAX_LINKHDR); +#endif +#else + SCTP_BUF_RESV_UF(mout, max_linkhdr); +#endif + SCTP_BUF_LEN(mout) = len; + SCTP_BUF_NEXT(mout) = cause; +#if defined(__FreeBSD__) && !defined(__Userspace__) + M_SETFIB(mout, fibnum); + mout->m_pkthdr.flowid = mflowid; + M_HASHTYPE_SET(mout, mflowtype); +#endif +#ifdef INET + ip = NULL; +#endif +#ifdef INET6 + ip6 = NULL; +#endif + switch (dst->sa_family) { +#ifdef INET + case AF_INET: + src_sin = (struct sockaddr_in *)src; + dst_sin = (struct sockaddr_in *)dst; + ip = mtod(mout, struct ip *); + ip->ip_v = IPVERSION; + ip->ip_hl = (sizeof(struct ip) >> 2); + ip->ip_tos = 0; +#if defined(__FreeBSD__) && !defined(__Userspace__) + ip->ip_off = htons(IP_DF); +#elif defined(WITH_CONVERT_IP_OFF) || defined(__APPLE__) + ip->ip_off = IP_DF; +#else + ip->ip_off = htons(IP_DF); +#endif +#if defined(__Userspace__) + ip->ip_id = htons(ip_id++); +#elif defined(__FreeBSD__) + ip_fillid(ip); +#elif defined(__APPLE__) +#if RANDOM_IP_ID + ip->ip_id = ip_randomid(); +#else + ip->ip_id = htons(ip_id++); +#endif +#else + ip->ip_id = ip_id++; +#endif + ip->ip_ttl = MODULE_GLOBAL(ip_defttl); + if (port) { + ip->ip_p = IPPROTO_UDP; + } else { + ip->ip_p = IPPROTO_SCTP; + } + ip->ip_src.s_addr = dst_sin->sin_addr.s_addr; + ip->ip_dst.s_addr = src_sin->sin_addr.s_addr; + ip->ip_sum = 0; + len = sizeof(struct ip); + shout = (struct sctphdr *)((caddr_t)ip + len); + break; +#endif +#ifdef INET6 + case AF_INET6: + src_sin6 = (struct sockaddr_in6 *)src; + dst_sin6 = (struct sockaddr_in6 *)dst; + ip6 = mtod(mout, struct ip6_hdr *); + ip6->ip6_flow = htonl(0x60000000); +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (V_ip6_auto_flowlabel) { + ip6->ip6_flow |= (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); + } +#endif +#if defined(__Userspace__) + ip6->ip6_hlim = IPv6_HOP_LIMIT; +#else + ip6->ip6_hlim = MODULE_GLOBAL(ip6_defhlim); +#endif + if (port) { + ip6->ip6_nxt = IPPROTO_UDP; + } else { + ip6->ip6_nxt = IPPROTO_SCTP; + } + ip6->ip6_src = dst_sin6->sin6_addr; + ip6->ip6_dst = src_sin6->sin6_addr; + len = sizeof(struct ip6_hdr); + shout = (struct sctphdr *)((caddr_t)ip6 + len); + break; +#endif + default: + len = 0; + shout = mtod(mout, struct sctphdr *); + break; + } +#if defined(INET) || defined(INET6) + if (port) { + if (htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)) == 0) { + sctp_m_freem(mout); + return; + } + udp = (struct udphdr *)shout; + udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); + udp->uh_dport = port; + udp->uh_sum = 0; + udp->uh_ulen = htons((uint16_t)(sizeof(struct udphdr) + + sizeof(struct sctphdr) + + sizeof(struct sctp_chunkhdr) + + cause_len + padding_len)); + len += sizeof(struct udphdr); + shout = (struct sctphdr *)((caddr_t)shout + sizeof(struct udphdr)); + } else { + udp = NULL; + } +#endif + shout->src_port = sh->dest_port; + shout->dest_port = sh->src_port; + shout->checksum = 0; + if (vtag) { + shout->v_tag = htonl(vtag); + } else { + shout->v_tag = sh->v_tag; + } + len += sizeof(struct sctphdr); + ch = (struct sctp_chunkhdr *)((caddr_t)shout + sizeof(struct sctphdr)); + ch->chunk_type = type; + if (vtag) { + ch->chunk_flags = 0; + } else { + ch->chunk_flags = SCTP_HAD_NO_TCB; + } + ch->chunk_length = htons((uint16_t)(sizeof(struct sctp_chunkhdr) + cause_len)); + len += sizeof(struct sctp_chunkhdr); + len += cause_len + padding_len; + + if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { + sctp_m_freem(mout); + return; + } + SCTP_ATTACH_CHAIN(o_pak, mout, len); + switch (dst->sa_family) { +#ifdef INET + case AF_INET: +#if defined(__APPLE__) && !defined(__Userspace__) + /* zap the stack pointer to the route */ + memset(&ro, 0, sizeof(sctp_route_t)); +#endif + if (port) { +#if !defined(_WIN32) && !defined(__Userspace__) +#if defined(__FreeBSD__) + if (V_udp_cksum) { + udp->uh_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, udp->uh_ulen + htons(IPPROTO_UDP)); + } else { + udp->uh_sum = 0; + } +#else + udp->uh_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, udp->uh_ulen + htons(IPPROTO_UDP)); +#endif +#else + udp->uh_sum = 0; +#endif + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + ip->ip_len = htons(len); +#elif defined(__APPLE__) || defined(__Userspace__) + ip->ip_len = len; +#else + ip->ip_len = htons(len); +#endif + if (port) { + shout->checksum = sctp_calculate_cksum(mout, sizeof(struct ip) + sizeof(struct udphdr)); + SCTP_STAT_INCR(sctps_sendswcrc); +#if !defined(_WIN32) && !defined(__Userspace__) +#if defined(__FreeBSD__) + if (V_udp_cksum) { + SCTP_ENABLE_UDP_CSUM(o_pak); + } +#else + SCTP_ENABLE_UDP_CSUM(o_pak); +#endif +#endif + } else { +#if defined(__FreeBSD__) && !defined(__Userspace__) + mout->m_pkthdr.csum_flags = CSUM_SCTP; + mout->m_pkthdr.csum_data = offsetof(struct sctphdr, checksum); + SCTP_STAT_INCR(sctps_sendhwcrc); +#else + shout->checksum = sctp_calculate_cksum(mout, sizeof(struct ip)); + SCTP_STAT_INCR(sctps_sendswcrc); +#endif + } +#ifdef SCTP_PACKET_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) { + sctp_packet_log(o_pak); + } +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_IP_OUTPUT(ret, o_pak, &ro, NULL, vrf_id); + /* Free the route if we got one back */ + if (ro.ro_rt) { + RTFREE(ro.ro_rt); + ro.ro_rt = NULL; + } +#else +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_PROBE5(send, NULL, NULL, ip, NULL, shout); +#endif + SCTP_IP_OUTPUT(ret, o_pak, NULL, NULL, vrf_id); +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: + ip6->ip6_plen = htons((uint16_t)(len - sizeof(struct ip6_hdr))); + if (port) { + shout->checksum = sctp_calculate_cksum(mout, sizeof(struct ip6_hdr) + sizeof(struct udphdr)); + SCTP_STAT_INCR(sctps_sendswcrc); +#if !defined(__Userspace__) +#if defined(_WIN32) + udp->uh_sum = 0; +#else + if ((udp->uh_sum = in6_cksum(o_pak, IPPROTO_UDP, sizeof(struct ip6_hdr), len - sizeof(struct ip6_hdr))) == 0) { + udp->uh_sum = 0xffff; + } +#endif +#endif + } else { +#if defined(__FreeBSD__) && !defined(__Userspace__) + mout->m_pkthdr.csum_flags = CSUM_SCTP_IPV6; + mout->m_pkthdr.csum_data = offsetof(struct sctphdr, checksum); + SCTP_STAT_INCR(sctps_sendhwcrc); +#else + shout->checksum = sctp_calculate_cksum(mout, sizeof(struct ip6_hdr)); + SCTP_STAT_INCR(sctps_sendswcrc); +#endif + } +#ifdef SCTP_PACKET_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) { + sctp_packet_log(o_pak); + } +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_PROBE5(send, NULL, NULL, ip6, NULL, shout); +#endif + SCTP_IP6_OUTPUT(ret, o_pak, NULL, NULL, NULL, vrf_id); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + { + char *buffer; + struct sockaddr_conn *sconn; + + sconn = (struct sockaddr_conn *)src; + if (SCTP_BASE_VAR(crc32c_offloaded) == 0) { + shout->checksum = sctp_calculate_cksum(o_pak, 0); + SCTP_STAT_INCR(sctps_sendswcrc); + } else { + SCTP_STAT_INCR(sctps_sendhwcrc); + } +#ifdef SCTP_PACKET_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) { + sctp_packet_log(mout); + } +#endif + /* Don't alloc/free for each packet */ + if ((buffer = malloc(len)) != NULL) { + m_copydata(o_pak, 0, len, buffer); + ret = SCTP_BASE_VAR(conn_output)(sconn->sconn_addr, buffer, len, 0, 0); + free(buffer); + } else { + ret = ENOMEM; + } + sctp_m_freem(o_pak); + break; + } +#endif + default: + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Unknown protocol (TSNH) type %d\n", + dst->sa_family); + sctp_m_freem(mout); + SCTP_LTRACE_ERR_RET_PKT(mout, NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EFAULT); + return; + } + SCTPDBG(SCTP_DEBUG_OUTPUT3, "return from send is %d\n", ret); +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (port) { + UDPSTAT_INC(udps_opackets); + } +#endif + SCTP_STAT_INCR(sctps_sendpackets); + SCTP_STAT_INCR_COUNTER64(sctps_outpackets); + SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); + if (ret) { + SCTP_STAT_INCR(sctps_senderrors); + } + return; +} + +void +sctp_send_shutdown_complete2(struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, uint16_t fibnum, +#endif + uint32_t vrf_id, uint16_t port) +{ + sctp_send_resp_msg(src, dst, sh, 0, SCTP_SHUTDOWN_COMPLETE, NULL, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, fibnum, +#endif + vrf_id, port); +} + +void +sctp_send_hb(struct sctp_tcb *stcb, struct sctp_nets *net,int so_locked) +{ + struct sctp_tmit_chunk *chk; + struct sctp_heartbeat_chunk *hb; + struct timeval now; + + SCTP_TCB_LOCK_ASSERT(stcb); + if (net == NULL) { + return; + } + (void)SCTP_GETTIME_TIMEVAL(&now); + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + break; +#endif +#ifdef INET6 + case AF_INET6: + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + break; +#endif + default: + return; + } + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + SCTPDBG(SCTP_DEBUG_OUTPUT4, "Gak, can't get a chunk for hb\n"); + return; + } + + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_HEARTBEAT_REQUEST; + chk->rec.chunk_id.can_take_data = 1; + chk->flags = 0; + chk->asoc = &stcb->asoc; + chk->send_size = sizeof(struct sctp_heartbeat_chunk); + + chk->data = sctp_get_mbuf_for_msg(chk->send_size, 0, M_NOWAIT, 1, MT_HEADER); + if (chk->data == NULL) { + sctp_free_a_chunk(stcb, chk, so_locked); + return; + } + SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); + SCTP_BUF_LEN(chk->data) = chk->send_size; + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + chk->whoTo = net; + atomic_add_int(&chk->whoTo->ref_count, 1); + /* Now we have a mbuf that we can fill in with the details */ + hb = mtod(chk->data, struct sctp_heartbeat_chunk *); + memset(hb, 0, sizeof(struct sctp_heartbeat_chunk)); + /* fill out chunk header */ + hb->ch.chunk_type = SCTP_HEARTBEAT_REQUEST; + hb->ch.chunk_flags = 0; + hb->ch.chunk_length = htons(chk->send_size); + /* Fill out hb parameter */ + hb->heartbeat.hb_info.ph.param_type = htons(SCTP_HEARTBEAT_INFO); + hb->heartbeat.hb_info.ph.param_length = htons(sizeof(struct sctp_heartbeat_info_param)); + hb->heartbeat.hb_info.time_value_1 = (uint32_t)now.tv_sec; + hb->heartbeat.hb_info.time_value_2 = now.tv_usec; + /* Did our user request this one, put it in */ + hb->heartbeat.hb_info.addr_family = (uint8_t)net->ro._l_addr.sa.sa_family; +#ifdef HAVE_SA_LEN + hb->heartbeat.hb_info.addr_len = net->ro._l_addr.sa.sa_len; +#else + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + hb->heartbeat.hb_info.addr_len = sizeof(struct sockaddr_in); + break; +#endif +#ifdef INET6 + case AF_INET6: + hb->heartbeat.hb_info.addr_len = sizeof(struct sockaddr_in6); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + hb->heartbeat.hb_info.addr_len = sizeof(struct sockaddr_conn); + break; +#endif + default: + hb->heartbeat.hb_info.addr_len = 0; + break; + } +#endif + if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { + /* + * we only take from the entropy pool if the address is not + * confirmed. + */ + net->heartbeat_random1 = hb->heartbeat.hb_info.random_value1 = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep); + net->heartbeat_random2 = hb->heartbeat.hb_info.random_value2 = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep); + } else { + net->heartbeat_random1 = hb->heartbeat.hb_info.random_value1 = 0; + net->heartbeat_random2 = hb->heartbeat.hb_info.random_value2 = 0; + } + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + memcpy(hb->heartbeat.hb_info.address, + &net->ro._l_addr.sin.sin_addr, + sizeof(net->ro._l_addr.sin.sin_addr)); + break; +#endif +#ifdef INET6 + case AF_INET6: + memcpy(hb->heartbeat.hb_info.address, + &net->ro._l_addr.sin6.sin6_addr, + sizeof(net->ro._l_addr.sin6.sin6_addr)); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + memcpy(hb->heartbeat.hb_info.address, + &net->ro._l_addr.sconn.sconn_addr, + sizeof(net->ro._l_addr.sconn.sconn_addr)); + break; +#endif + default: + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + sctp_free_a_chunk(stcb, chk, so_locked); + return; + break; + } + net->hb_responded = 0; + TAILQ_INSERT_TAIL(&stcb->asoc.control_send_queue, chk, sctp_next); + stcb->asoc.ctrl_queue_cnt++; + SCTP_STAT_INCR(sctps_sendheartbeat); + return; +} + +void +sctp_send_ecn_echo(struct sctp_tcb *stcb, struct sctp_nets *net, + uint32_t high_tsn) +{ + struct sctp_association *asoc; + struct sctp_ecne_chunk *ecne; + struct sctp_tmit_chunk *chk; + + if (net == NULL) { + return; + } + asoc = &stcb->asoc; + SCTP_TCB_LOCK_ASSERT(stcb); + TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { + if ((chk->rec.chunk_id.id == SCTP_ECN_ECHO) && (net == chk->whoTo)) { + /* found a previous ECN_ECHO update it if needed */ + uint32_t cnt, ctsn; + ecne = mtod(chk->data, struct sctp_ecne_chunk *); + ctsn = ntohl(ecne->tsn); + if (SCTP_TSN_GT(high_tsn, ctsn)) { + ecne->tsn = htonl(high_tsn); + SCTP_STAT_INCR(sctps_queue_upd_ecne); + } + cnt = ntohl(ecne->num_pkts_since_cwr); + cnt++; + ecne->num_pkts_since_cwr = htonl(cnt); + return; + } + } + /* nope could not find one to update so we must build one */ + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + return; + } + SCTP_STAT_INCR(sctps_queue_upd_ecne); + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_ECN_ECHO; + chk->rec.chunk_id.can_take_data = 0; + chk->flags = 0; + chk->asoc = &stcb->asoc; + chk->send_size = sizeof(struct sctp_ecne_chunk); + chk->data = sctp_get_mbuf_for_msg(chk->send_size, 0, M_NOWAIT, 1, MT_HEADER); + if (chk->data == NULL) { + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + return; + } + SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); + SCTP_BUF_LEN(chk->data) = chk->send_size; + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + chk->whoTo = net; + atomic_add_int(&chk->whoTo->ref_count, 1); + + stcb->asoc.ecn_echo_cnt_onq++; + ecne = mtod(chk->data, struct sctp_ecne_chunk *); + ecne->ch.chunk_type = SCTP_ECN_ECHO; + ecne->ch.chunk_flags = 0; + ecne->ch.chunk_length = htons(sizeof(struct sctp_ecne_chunk)); + ecne->tsn = htonl(high_tsn); + ecne->num_pkts_since_cwr = htonl(1); + TAILQ_INSERT_HEAD(&stcb->asoc.control_send_queue, chk, sctp_next); + asoc->ctrl_queue_cnt++; +} + +void +sctp_send_packet_dropped(struct sctp_tcb *stcb, struct sctp_nets *net, + struct mbuf *m, int len, int iphlen, int bad_crc) +{ + struct sctp_association *asoc; + struct sctp_pktdrop_chunk *drp; + struct sctp_tmit_chunk *chk; + uint8_t *datap; + int was_trunc = 0; + int fullsz = 0; + long spc; + int offset; + struct sctp_chunkhdr *ch, chunk_buf; + unsigned int chk_length; + + if (!stcb) { + return; + } + asoc = &stcb->asoc; + SCTP_TCB_LOCK_ASSERT(stcb); + if (asoc->pktdrop_supported == 0) { + /*- + * peer must declare support before I send one. + */ + return; + } + if (stcb->sctp_socket == NULL) { + return; + } + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + return; + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_PACKET_DROPPED; + chk->rec.chunk_id.can_take_data = 1; + chk->flags = 0; + len -= iphlen; + chk->send_size = len; + /* Validate that we do not have an ABORT in here. */ + offset = iphlen + sizeof(struct sctphdr); + ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, + sizeof(*ch), (uint8_t *) & chunk_buf); + while (ch != NULL) { + chk_length = ntohs(ch->chunk_length); + if (chk_length < sizeof(*ch)) { + /* break to abort land */ + break; + } + switch (ch->chunk_type) { + case SCTP_PACKET_DROPPED: + case SCTP_ABORT_ASSOCIATION: + case SCTP_INITIATION_ACK: + /** + * We don't respond with an PKT-DROP to an ABORT + * or PKT-DROP. We also do not respond to an + * INIT-ACK, because we can't know if the initiation + * tag is correct or not. + */ + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + return; + default: + break; + } + offset += SCTP_SIZE32(chk_length); + ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, + sizeof(*ch), (uint8_t *) & chunk_buf); + } + + if ((len + SCTP_MAX_OVERHEAD + sizeof(struct sctp_pktdrop_chunk)) > + min(stcb->asoc.smallest_mtu, MCLBYTES)) { + /* only send 1 mtu worth, trim off the + * excess on the end. + */ + fullsz = len; + len = min(stcb->asoc.smallest_mtu, MCLBYTES) - SCTP_MAX_OVERHEAD; + was_trunc = 1; + } + chk->asoc = &stcb->asoc; + chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_NOWAIT, 1, MT_DATA); + if (chk->data == NULL) { +jump_out: + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + return; + } + SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); + drp = mtod(chk->data, struct sctp_pktdrop_chunk *); + if (drp == NULL) { + sctp_m_freem(chk->data); + chk->data = NULL; + goto jump_out; + } + chk->book_size = SCTP_SIZE32((chk->send_size + sizeof(struct sctp_pktdrop_chunk) + + sizeof(struct sctphdr) + SCTP_MED_OVERHEAD)); + chk->book_size_scale = 0; + if (was_trunc) { + drp->ch.chunk_flags = SCTP_PACKET_TRUNCATED; + drp->trunc_len = htons(fullsz); + /* Len is already adjusted to size minus overhead above + * take out the pkt_drop chunk itself from it. + */ + chk->send_size = (uint16_t)(len - sizeof(struct sctp_pktdrop_chunk)); + len = chk->send_size; + } else { + /* no truncation needed */ + drp->ch.chunk_flags = 0; + drp->trunc_len = htons(0); + } + if (bad_crc) { + drp->ch.chunk_flags |= SCTP_BADCRC; + } + chk->send_size += sizeof(struct sctp_pktdrop_chunk); + SCTP_BUF_LEN(chk->data) = chk->send_size; + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + if (net) { + /* we should hit here */ + chk->whoTo = net; + atomic_add_int(&chk->whoTo->ref_count, 1); + } else { + chk->whoTo = NULL; + } + drp->ch.chunk_type = SCTP_PACKET_DROPPED; + drp->ch.chunk_length = htons(chk->send_size); + spc = SCTP_SB_LIMIT_RCV(stcb->sctp_socket); + if (spc < 0) { + spc = 0; + } + drp->bottle_bw = htonl(spc); + if (asoc->my_rwnd) { + drp->current_onq = htonl(asoc->size_on_reasm_queue + + asoc->size_on_all_streams + + asoc->my_rwnd_control_len + + SCTP_SBAVAIL(&stcb->sctp_socket->so_rcv)); + } else { + /*- + * If my rwnd is 0, possibly from mbuf depletion as well as + * space used, tell the peer there is NO space aka onq == bw + */ + drp->current_onq = htonl(spc); + } + drp->reserved = 0; + datap = drp->data; + m_copydata(m, iphlen, len, (caddr_t)datap); + TAILQ_INSERT_TAIL(&stcb->asoc.control_send_queue, chk, sctp_next); + asoc->ctrl_queue_cnt++; +} + +void +sctp_send_cwr(struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t high_tsn, uint8_t override) +{ + struct sctp_association *asoc; + struct sctp_cwr_chunk *cwr; + struct sctp_tmit_chunk *chk; + + SCTP_TCB_LOCK_ASSERT(stcb); + if (net == NULL) { + return; + } + asoc = &stcb->asoc; + TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { + if ((chk->rec.chunk_id.id == SCTP_ECN_CWR) && (net == chk->whoTo)) { + /* found a previous CWR queued to same destination update it if needed */ + uint32_t ctsn; + cwr = mtod(chk->data, struct sctp_cwr_chunk *); + ctsn = ntohl(cwr->tsn); + if (SCTP_TSN_GT(high_tsn, ctsn)) { + cwr->tsn = htonl(high_tsn); + } + if (override & SCTP_CWR_REDUCE_OVERRIDE) { + /* Make sure override is carried */ + cwr->ch.chunk_flags |= SCTP_CWR_REDUCE_OVERRIDE; + } + return; + } + } + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + return; + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_ECN_CWR; + chk->rec.chunk_id.can_take_data = 1; + chk->flags = 0; + chk->asoc = asoc; + chk->send_size = sizeof(struct sctp_cwr_chunk); + chk->data = sctp_get_mbuf_for_msg(chk->send_size, 0, M_NOWAIT, 1, MT_HEADER); + if (chk->data == NULL) { + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + return; + } + SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); + SCTP_BUF_LEN(chk->data) = chk->send_size; + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + chk->whoTo = net; + atomic_add_int(&chk->whoTo->ref_count, 1); + cwr = mtod(chk->data, struct sctp_cwr_chunk *); + cwr->ch.chunk_type = SCTP_ECN_CWR; + cwr->ch.chunk_flags = override; + cwr->ch.chunk_length = htons(sizeof(struct sctp_cwr_chunk)); + cwr->tsn = htonl(high_tsn); + TAILQ_INSERT_TAIL(&asoc->control_send_queue, chk, sctp_next); + asoc->ctrl_queue_cnt++; +} + +static int +sctp_add_stream_reset_out(struct sctp_tcb *stcb, struct sctp_tmit_chunk *chk, + uint32_t seq, uint32_t resp_seq, uint32_t last_sent) +{ + uint16_t len, old_len, i; + struct sctp_stream_reset_out_request *req_out; + struct sctp_chunkhdr *ch; + int at; + int number_entries=0; + + ch = mtod(chk->data, struct sctp_chunkhdr *); + old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); + /* get to new offset for the param. */ + req_out = (struct sctp_stream_reset_out_request *)((caddr_t)ch + len); + /* now how long will this param be? */ + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + if ((stcb->asoc.strmout[i].state == SCTP_STREAM_RESET_PENDING) && + (stcb->asoc.strmout[i].chunks_on_queues == 0) && + TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) { + number_entries++; + } + } + if (number_entries == 0) { + return (0); + } + if (number_entries == stcb->asoc.streamoutcnt) { + number_entries = 0; + } + if (number_entries > SCTP_MAX_STREAMS_AT_ONCE_RESET) { + number_entries = SCTP_MAX_STREAMS_AT_ONCE_RESET; + } + len = (uint16_t)(sizeof(struct sctp_stream_reset_out_request) + (sizeof(uint16_t) * number_entries)); + req_out->ph.param_type = htons(SCTP_STR_RESET_OUT_REQUEST); + req_out->ph.param_length = htons(len); + req_out->request_seq = htonl(seq); + req_out->response_seq = htonl(resp_seq); + req_out->send_reset_at_tsn = htonl(last_sent); + at = 0; + if (number_entries) { + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + if ((stcb->asoc.strmout[i].state == SCTP_STREAM_RESET_PENDING) && + (stcb->asoc.strmout[i].chunks_on_queues == 0) && + TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) { + req_out->list_of_streams[at] = htons(i); + at++; + stcb->asoc.strmout[i].state = SCTP_STREAM_RESET_IN_FLIGHT; + if (at >= number_entries) { + break; + } + } + } + } else { + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + stcb->asoc.strmout[i].state = SCTP_STREAM_RESET_IN_FLIGHT; + } + } + if (SCTP_SIZE32(len) > len) { + /*- + * Need to worry about the pad we may end up adding to the + * end. This is easy since the struct is either aligned to 4 + * bytes or 2 bytes off. + */ + req_out->list_of_streams[number_entries] = 0; + } + /* now fix the chunk length */ + ch->chunk_length = htons(len + old_len); + chk->book_size = len + old_len; + chk->book_size_scale = 0; + chk->send_size = SCTP_SIZE32(chk->book_size); + SCTP_BUF_LEN(chk->data) = chk->send_size; + return (1); +} + +static void +sctp_add_stream_reset_in(struct sctp_tmit_chunk *chk, + int number_entries, uint16_t *list, + uint32_t seq) +{ + uint16_t len, old_len, i; + struct sctp_stream_reset_in_request *req_in; + struct sctp_chunkhdr *ch; + + ch = mtod(chk->data, struct sctp_chunkhdr *); + old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); + + /* get to new offset for the param. */ + req_in = (struct sctp_stream_reset_in_request *)((caddr_t)ch + len); + /* now how long will this param be? */ + len = (uint16_t)(sizeof(struct sctp_stream_reset_in_request) + (sizeof(uint16_t) * number_entries)); + req_in->ph.param_type = htons(SCTP_STR_RESET_IN_REQUEST); + req_in->ph.param_length = htons(len); + req_in->request_seq = htonl(seq); + if (number_entries) { + for (i = 0; i < number_entries; i++) { + req_in->list_of_streams[i] = htons(list[i]); + } + } + if (SCTP_SIZE32(len) > len) { + /*- + * Need to worry about the pad we may end up adding to the + * end. This is easy since the struct is either aligned to 4 + * bytes or 2 bytes off. + */ + req_in->list_of_streams[number_entries] = 0; + } + /* now fix the chunk length */ + ch->chunk_length = htons(len + old_len); + chk->book_size = len + old_len; + chk->book_size_scale = 0; + chk->send_size = SCTP_SIZE32(chk->book_size); + SCTP_BUF_LEN(chk->data) = chk->send_size; + return; +} + +static void +sctp_add_stream_reset_tsn(struct sctp_tmit_chunk *chk, + uint32_t seq) +{ + uint16_t len, old_len; + struct sctp_stream_reset_tsn_request *req_tsn; + struct sctp_chunkhdr *ch; + + ch = mtod(chk->data, struct sctp_chunkhdr *); + old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); + + /* get to new offset for the param. */ + req_tsn = (struct sctp_stream_reset_tsn_request *)((caddr_t)ch + len); + /* now how long will this param be? */ + len = sizeof(struct sctp_stream_reset_tsn_request); + req_tsn->ph.param_type = htons(SCTP_STR_RESET_TSN_REQUEST); + req_tsn->ph.param_length = htons(len); + req_tsn->request_seq = htonl(seq); + + /* now fix the chunk length */ + ch->chunk_length = htons(len + old_len); + chk->send_size = len + old_len; + chk->book_size = SCTP_SIZE32(chk->send_size); + chk->book_size_scale = 0; + SCTP_BUF_LEN(chk->data) = SCTP_SIZE32(chk->send_size); + return; +} + +void +sctp_add_stream_reset_result(struct sctp_tmit_chunk *chk, + uint32_t resp_seq, uint32_t result) +{ + uint16_t len, old_len; + struct sctp_stream_reset_response *resp; + struct sctp_chunkhdr *ch; + + ch = mtod(chk->data, struct sctp_chunkhdr *); + old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); + + /* get to new offset for the param. */ + resp = (struct sctp_stream_reset_response *)((caddr_t)ch + len); + /* now how long will this param be? */ + len = sizeof(struct sctp_stream_reset_response); + resp->ph.param_type = htons(SCTP_STR_RESET_RESPONSE); + resp->ph.param_length = htons(len); + resp->response_seq = htonl(resp_seq); + resp->result = ntohl(result); + + /* now fix the chunk length */ + ch->chunk_length = htons(len + old_len); + chk->book_size = len + old_len; + chk->book_size_scale = 0; + chk->send_size = SCTP_SIZE32(chk->book_size); + SCTP_BUF_LEN(chk->data) = chk->send_size; + return; +} + +void +sctp_send_deferred_reset_response(struct sctp_tcb *stcb, + struct sctp_stream_reset_list *ent, + int response) +{ + struct sctp_association *asoc; + struct sctp_tmit_chunk *chk; + struct sctp_chunkhdr *ch; + + asoc = &stcb->asoc; + + /* + * Reset our last reset action to the new one IP -> response + * (PERFORMED probably). This assures that if we fail to send, a + * retran from the peer will get the new response. + */ + asoc->last_reset_action[0] = response; + if (asoc->stream_reset_outstanding) { + return; + } + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return; + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_STREAM_RESET; + chk->rec.chunk_id.can_take_data = 0; + chk->flags = 0; + chk->asoc = &stcb->asoc; + chk->book_size = sizeof(struct sctp_chunkhdr); + chk->send_size = SCTP_SIZE32(chk->book_size); + chk->book_size_scale = 0; + chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_NOWAIT, 1, MT_DATA); + if (chk->data == NULL) { + sctp_free_a_chunk(stcb, chk, SCTP_SO_LOCKED); + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return; + } + SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); + /* setup chunk parameters */ + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + if (stcb->asoc.alternate) { + chk->whoTo = stcb->asoc.alternate; + } else { + chk->whoTo = stcb->asoc.primary_destination; + } + ch = mtod(chk->data, struct sctp_chunkhdr *); + ch->chunk_type = SCTP_STREAM_RESET; + ch->chunk_flags = 0; + ch->chunk_length = htons(chk->book_size); + atomic_add_int(&chk->whoTo->ref_count, 1); + SCTP_BUF_LEN(chk->data) = chk->send_size; + sctp_add_stream_reset_result(chk, ent->seq, response); + /* insert the chunk for sending */ + TAILQ_INSERT_TAIL(&asoc->control_send_queue, + chk, + sctp_next); + asoc->ctrl_queue_cnt++; +} + +void +sctp_add_stream_reset_result_tsn(struct sctp_tmit_chunk *chk, + uint32_t resp_seq, uint32_t result, + uint32_t send_una, uint32_t recv_next) +{ + uint16_t len, old_len; + struct sctp_stream_reset_response_tsn *resp; + struct sctp_chunkhdr *ch; + + ch = mtod(chk->data, struct sctp_chunkhdr *); + old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); + + /* get to new offset for the param. */ + resp = (struct sctp_stream_reset_response_tsn *)((caddr_t)ch + len); + /* now how long will this param be? */ + len = sizeof(struct sctp_stream_reset_response_tsn); + resp->ph.param_type = htons(SCTP_STR_RESET_RESPONSE); + resp->ph.param_length = htons(len); + resp->response_seq = htonl(resp_seq); + resp->result = htonl(result); + resp->senders_next_tsn = htonl(send_una); + resp->receivers_next_tsn = htonl(recv_next); + + /* now fix the chunk length */ + ch->chunk_length = htons(len + old_len); + chk->book_size = len + old_len; + chk->send_size = SCTP_SIZE32(chk->book_size); + chk->book_size_scale = 0; + SCTP_BUF_LEN(chk->data) = chk->send_size; + return; +} + +static void +sctp_add_an_out_stream(struct sctp_tmit_chunk *chk, + uint32_t seq, + uint16_t adding) +{ + uint16_t len, old_len; + struct sctp_chunkhdr *ch; + struct sctp_stream_reset_add_strm *addstr; + + ch = mtod(chk->data, struct sctp_chunkhdr *); + old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); + + /* get to new offset for the param. */ + addstr = (struct sctp_stream_reset_add_strm *)((caddr_t)ch + len); + /* now how long will this param be? */ + len = sizeof(struct sctp_stream_reset_add_strm); + + /* Fill it out. */ + addstr->ph.param_type = htons(SCTP_STR_RESET_ADD_OUT_STREAMS); + addstr->ph.param_length = htons(len); + addstr->request_seq = htonl(seq); + addstr->number_of_streams = htons(adding); + addstr->reserved = 0; + + /* now fix the chunk length */ + ch->chunk_length = htons(len + old_len); + chk->send_size = len + old_len; + chk->book_size = SCTP_SIZE32(chk->send_size); + chk->book_size_scale = 0; + SCTP_BUF_LEN(chk->data) = SCTP_SIZE32(chk->send_size); + return; +} + +static void +sctp_add_an_in_stream(struct sctp_tmit_chunk *chk, + uint32_t seq, + uint16_t adding) +{ + uint16_t len, old_len; + struct sctp_chunkhdr *ch; + struct sctp_stream_reset_add_strm *addstr; + + ch = mtod(chk->data, struct sctp_chunkhdr *); + old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); + + /* get to new offset for the param. */ + addstr = (struct sctp_stream_reset_add_strm *)((caddr_t)ch + len); + /* now how long will this param be? */ + len = sizeof(struct sctp_stream_reset_add_strm); + /* Fill it out. */ + addstr->ph.param_type = htons(SCTP_STR_RESET_ADD_IN_STREAMS); + addstr->ph.param_length = htons(len); + addstr->request_seq = htonl(seq); + addstr->number_of_streams = htons(adding); + addstr->reserved = 0; + + /* now fix the chunk length */ + ch->chunk_length = htons(len + old_len); + chk->send_size = len + old_len; + chk->book_size = SCTP_SIZE32(chk->send_size); + chk->book_size_scale = 0; + SCTP_BUF_LEN(chk->data) = SCTP_SIZE32(chk->send_size); + return; +} + +int +sctp_send_stream_reset_out_if_possible(struct sctp_tcb *stcb, int so_locked) +{ + struct sctp_association *asoc; + struct sctp_tmit_chunk *chk; + struct sctp_chunkhdr *ch; + uint32_t seq; + + asoc = &stcb->asoc; + asoc->trigger_reset = 0; + if (asoc->stream_reset_outstanding) { + return (EALREADY); + } + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_STREAM_RESET; + chk->rec.chunk_id.can_take_data = 0; + chk->flags = 0; + chk->asoc = &stcb->asoc; + chk->book_size = sizeof(struct sctp_chunkhdr); + chk->send_size = SCTP_SIZE32(chk->book_size); + chk->book_size_scale = 0; + chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_NOWAIT, 1, MT_DATA); + if (chk->data == NULL) { + sctp_free_a_chunk(stcb, chk, so_locked); + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); + + /* setup chunk parameters */ + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + if (stcb->asoc.alternate) { + chk->whoTo = stcb->asoc.alternate; + } else { + chk->whoTo = stcb->asoc.primary_destination; + } + ch = mtod(chk->data, struct sctp_chunkhdr *); + ch->chunk_type = SCTP_STREAM_RESET; + ch->chunk_flags = 0; + ch->chunk_length = htons(chk->book_size); + atomic_add_int(&chk->whoTo->ref_count, 1); + SCTP_BUF_LEN(chk->data) = chk->send_size; + seq = stcb->asoc.str_reset_seq_out; + if (sctp_add_stream_reset_out(stcb, chk, seq, (stcb->asoc.str_reset_seq_in - 1), (stcb->asoc.sending_seq - 1))) { + seq++; + asoc->stream_reset_outstanding++; + } else { + m_freem(chk->data); + chk->data = NULL; + sctp_free_a_chunk(stcb, chk, so_locked); + return (ENOENT); + } + asoc->str_reset = chk; + /* insert the chunk for sending */ + TAILQ_INSERT_TAIL(&asoc->control_send_queue, + chk, + sctp_next); + asoc->ctrl_queue_cnt++; + + if (stcb->asoc.send_sack) { + sctp_send_sack(stcb, so_locked); + } + sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, stcb->sctp_ep, stcb, chk->whoTo); + return (0); +} + +int +sctp_send_str_reset_req(struct sctp_tcb *stcb, + uint16_t number_entries, uint16_t *list, + uint8_t send_in_req, + uint8_t send_tsn_req, + uint8_t add_stream, + uint16_t adding_o, + uint16_t adding_i, uint8_t peer_asked) +{ + struct sctp_association *asoc; + struct sctp_tmit_chunk *chk; + struct sctp_chunkhdr *ch; + int can_send_out_req=0; + uint32_t seq; + + SCTP_TCB_LOCK_ASSERT(stcb); + + asoc = &stcb->asoc; + if (asoc->stream_reset_outstanding) { + /*- + * Already one pending, must get ACK back to clear the flag. + */ + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EBUSY); + return (EBUSY); + } + if ((send_in_req == 0) && (send_tsn_req == 0) && + (add_stream == 0)) { + /* nothing to do */ + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); + return (EINVAL); + } + if (send_tsn_req && send_in_req) { + /* error, can't do that */ + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); + return (EINVAL); + } else if (send_in_req) { + can_send_out_req = 1; + } + if (number_entries > (MCLBYTES - + SCTP_MIN_OVERHEAD - + sizeof(struct sctp_chunkhdr) - + sizeof(struct sctp_stream_reset_out_request)) / + sizeof(uint16_t)) { + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + chk->copy_by_ref = 0; + chk->rec.chunk_id.id = SCTP_STREAM_RESET; + chk->rec.chunk_id.can_take_data = 0; + chk->flags = 0; + chk->asoc = &stcb->asoc; + chk->book_size = sizeof(struct sctp_chunkhdr); + chk->send_size = SCTP_SIZE32(chk->book_size); + chk->book_size_scale = 0; + chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_NOWAIT, 1, MT_DATA); + if (chk->data == NULL) { + sctp_free_a_chunk(stcb, chk, SCTP_SO_LOCKED); + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + return (ENOMEM); + } + SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); + + /* setup chunk parameters */ + chk->sent = SCTP_DATAGRAM_UNSENT; + chk->snd_count = 0; + if (stcb->asoc.alternate) { + chk->whoTo = stcb->asoc.alternate; + } else { + chk->whoTo = stcb->asoc.primary_destination; + } + atomic_add_int(&chk->whoTo->ref_count, 1); + ch = mtod(chk->data, struct sctp_chunkhdr *); + ch->chunk_type = SCTP_STREAM_RESET; + ch->chunk_flags = 0; + ch->chunk_length = htons(chk->book_size); + SCTP_BUF_LEN(chk->data) = chk->send_size; + + seq = stcb->asoc.str_reset_seq_out; + if (can_send_out_req) { + int ret; + + ret = sctp_add_stream_reset_out(stcb, chk, seq, (stcb->asoc.str_reset_seq_in - 1), (stcb->asoc.sending_seq - 1)); + if (ret) { + seq++; + asoc->stream_reset_outstanding++; + } + } + if ((add_stream & 1) && + ((stcb->asoc.strm_realoutsize - stcb->asoc.streamoutcnt) < adding_o)) { + /* Need to allocate more */ + struct sctp_stream_out *oldstream; + struct sctp_stream_queue_pending *sp, *nsp; + int i; +#if defined(SCTP_DETAILED_STR_STATS) + int j; +#endif + + oldstream = stcb->asoc.strmout; + /* get some more */ + SCTP_MALLOC(stcb->asoc.strmout, struct sctp_stream_out *, + (stcb->asoc.streamoutcnt + adding_o) * sizeof(struct sctp_stream_out), + SCTP_M_STRMO); + if (stcb->asoc.strmout == NULL) { + uint8_t x; + stcb->asoc.strmout = oldstream; + /* Turn off the bit */ + x = add_stream & 0xfe; + add_stream = x; + goto skip_stuff; + } + /* Ok now we proceed with copying the old out stuff and + * initializing the new stuff. + */ + stcb->asoc.ss_functions.sctp_ss_clear(stcb, &stcb->asoc, false); + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + TAILQ_INIT(&stcb->asoc.strmout[i].outqueue); + /* FIX ME FIX ME */ + /* This should be a SS_COPY operation FIX ME STREAM SCHEDULER EXPERT */ + stcb->asoc.ss_functions.sctp_ss_init_stream(stcb, &stcb->asoc.strmout[i], &oldstream[i]); + stcb->asoc.strmout[i].chunks_on_queues = oldstream[i].chunks_on_queues; +#if defined(SCTP_DETAILED_STR_STATS) + for (j = 0; j < SCTP_PR_SCTP_MAX + 1; j++) { + stcb->asoc.strmout[i].abandoned_sent[j] = oldstream[i].abandoned_sent[j]; + stcb->asoc.strmout[i].abandoned_unsent[j] = oldstream[i].abandoned_unsent[j]; + } +#else + stcb->asoc.strmout[i].abandoned_sent[0] = oldstream[i].abandoned_sent[0]; + stcb->asoc.strmout[i].abandoned_unsent[0] = oldstream[i].abandoned_unsent[0]; +#endif + stcb->asoc.strmout[i].next_mid_ordered = oldstream[i].next_mid_ordered; + stcb->asoc.strmout[i].next_mid_unordered = oldstream[i].next_mid_unordered; + stcb->asoc.strmout[i].last_msg_incomplete = oldstream[i].last_msg_incomplete; + stcb->asoc.strmout[i].sid = i; + stcb->asoc.strmout[i].state = oldstream[i].state; + /* now anything on those queues? */ + TAILQ_FOREACH_SAFE(sp, &oldstream[i].outqueue, next, nsp) { + TAILQ_REMOVE(&oldstream[i].outqueue, sp, next); + TAILQ_INSERT_TAIL(&stcb->asoc.strmout[i].outqueue, sp, next); + } + } + /* now the new streams */ + stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc); + for (i = stcb->asoc.streamoutcnt; i < (stcb->asoc.streamoutcnt + adding_o); i++) { + TAILQ_INIT(&stcb->asoc.strmout[i].outqueue); + stcb->asoc.strmout[i].chunks_on_queues = 0; +#if defined(SCTP_DETAILED_STR_STATS) + for (j = 0; j < SCTP_PR_SCTP_MAX + 1; j++) { + stcb->asoc.strmout[i].abandoned_sent[j] = 0; + stcb->asoc.strmout[i].abandoned_unsent[j] = 0; + } +#else + stcb->asoc.strmout[i].abandoned_sent[0] = 0; + stcb->asoc.strmout[i].abandoned_unsent[0] = 0; +#endif + stcb->asoc.strmout[i].next_mid_ordered = 0; + stcb->asoc.strmout[i].next_mid_unordered = 0; + stcb->asoc.strmout[i].sid = i; + stcb->asoc.strmout[i].last_msg_incomplete = 0; + stcb->asoc.ss_functions.sctp_ss_init_stream(stcb, &stcb->asoc.strmout[i], NULL); + stcb->asoc.strmout[i].state = SCTP_STREAM_CLOSED; + } + stcb->asoc.strm_realoutsize = stcb->asoc.streamoutcnt + adding_o; + SCTP_FREE(oldstream, SCTP_M_STRMO); + } +skip_stuff: + if ((add_stream & 1) && (adding_o > 0)) { + asoc->strm_pending_add_size = adding_o; + asoc->peer_req_out = peer_asked; + sctp_add_an_out_stream(chk, seq, adding_o); + seq++; + asoc->stream_reset_outstanding++; + } + if ((add_stream & 2) && (adding_i > 0)) { + sctp_add_an_in_stream(chk, seq, adding_i); + seq++; + asoc->stream_reset_outstanding++; + } + if (send_in_req) { + sctp_add_stream_reset_in(chk, number_entries, list, seq); + seq++; + asoc->stream_reset_outstanding++; + } + if (send_tsn_req) { + sctp_add_stream_reset_tsn(chk, seq); + asoc->stream_reset_outstanding++; + } + asoc->str_reset = chk; + /* insert the chunk for sending */ + TAILQ_INSERT_TAIL(&asoc->control_send_queue, + chk, + sctp_next); + asoc->ctrl_queue_cnt++; + if (stcb->asoc.send_sack) { + sctp_send_sack(stcb, SCTP_SO_LOCKED); + } + sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, stcb->sctp_ep, stcb, chk->whoTo); + return (0); +} + +void +sctp_send_abort(struct mbuf *m, int iphlen, struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, uint32_t vtag, struct mbuf *cause, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, uint16_t fibnum, +#endif + uint32_t vrf_id, uint16_t port) +{ + /* Don't respond to an ABORT with an ABORT. */ + if (sctp_is_there_an_abort_here(m, iphlen, &vtag)) { + if (cause) + sctp_m_freem(cause); + return; + } + sctp_send_resp_msg(src, dst, sh, vtag, SCTP_ABORT_ASSOCIATION, cause, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, fibnum, +#endif + vrf_id, port); + return; +} + +void +sctp_send_operr_to(struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, uint32_t vtag, struct mbuf *cause, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, uint16_t fibnum, +#endif + uint32_t vrf_id, uint16_t port) +{ + sctp_send_resp_msg(src, dst, sh, vtag, SCTP_OPERATION_ERROR, cause, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, fibnum, +#endif + vrf_id, port); + return; +} + +static struct mbuf * +sctp_copy_resume(struct uio *uio, + int max_send_len, +#if defined(__FreeBSD__) || defined(__Userspace__) + int user_marks_eor, +#endif + int *error, + uint32_t *sndout, + struct mbuf **new_tail) +{ +#if defined(__FreeBSD__) || defined(__Userspace__) + struct mbuf *m; + + m = m_uiotombuf(uio, M_WAITOK, max_send_len, 0, + (M_PKTHDR | (user_marks_eor ? M_EOR : 0))); + if (m == NULL) { + /* The only possible error is EFAULT. */ + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EFAULT); + *error = EFAULT; + } else { + *sndout = m_length(m, NULL); + *new_tail = m_last(m); + } + return (m); +#else + int left, cancpy, willcpy; + struct mbuf *m, *head; + +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + left = (int)min(uio->uio_resid, max_send_len); +#else + left = (int)min(uio_resid(uio), max_send_len); +#endif +#else + left = (int)min(uio->uio_resid, max_send_len); +#endif + /* Always get a header just in case */ + head = sctp_get_mbuf_for_msg(left, 0, M_WAITOK, 0, MT_DATA); + if (head == NULL) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOBUFS); + *error = ENOBUFS; + return (NULL); + } + cancpy = (int)M_TRAILINGSPACE(head); + willcpy = min(cancpy, left); + *error = uiomove(mtod(head, caddr_t), willcpy, uio); + if (*error != 0) { + sctp_m_freem(head); + return (NULL); + } + *sndout += willcpy; + left -= willcpy; + SCTP_BUF_LEN(head) = willcpy; + m = head; + *new_tail = head; + while (left > 0) { + /* move in user data */ + SCTP_BUF_NEXT(m) = sctp_get_mbuf_for_msg(left, 0, M_WAITOK, 0, MT_DATA); + if (SCTP_BUF_NEXT(m) == NULL) { + sctp_m_freem(head); + *new_tail = NULL; + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOBUFS); + *error = ENOBUFS; + return (NULL); + } + m = SCTP_BUF_NEXT(m); + cancpy = (int)M_TRAILINGSPACE(m); + willcpy = min(cancpy, left); + *error = uiomove(mtod(m, caddr_t), willcpy, uio); + if (*error != 0) { + sctp_m_freem(head); + *new_tail = NULL; + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, *error); + return (NULL); + } + SCTP_BUF_LEN(m) = willcpy; + left -= willcpy; + *sndout += willcpy; + *new_tail = m; + if (left == 0) { + SCTP_BUF_NEXT(m) = NULL; + } + } + return (head); +#endif +} + +static int +sctp_copy_one(struct sctp_stream_queue_pending *sp, + struct uio *uio, + int resv_upfront) +{ +#if defined(__FreeBSD__) || defined(__Userspace__) + sp->data = m_uiotombuf(uio, M_WAITOK, sp->length, resv_upfront, 0); + if (sp->data == NULL) { + /* The only possible error is EFAULT. */ + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EFAULT); + return (EFAULT); + } + sp->tail_mbuf = m_last(sp->data); + return (0); +#else + int left; + int cancpy, willcpy, error; + struct mbuf *m, *head; + int cpsz = 0; + + /* First one gets a header */ + left = sp->length; + head = m = sctp_get_mbuf_for_msg((left + resv_upfront), 0, M_WAITOK, 0, MT_DATA); + if (m == NULL) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOBUFS); + return (ENOBUFS); + } + /*- + * Add this one for m in now, that way if the alloc fails we won't + * have a bad cnt. + */ + SCTP_BUF_RESV_UF(m, resv_upfront); + cancpy = (int)M_TRAILINGSPACE(m); + willcpy = min(cancpy, left); + while (left > 0) { + /* move in user data */ + error = uiomove(mtod(m, caddr_t), willcpy, uio); + if (error) { + sctp_m_freem(head); + return (error); + } + SCTP_BUF_LEN(m) = willcpy; + left -= willcpy; + cpsz += willcpy; + if (left > 0) { + SCTP_BUF_NEXT(m) = sctp_get_mbuf_for_msg(left, 0, M_WAITOK, 0, MT_DATA); + if (SCTP_BUF_NEXT(m) == NULL) { + /* + * the head goes back to caller, he can free + * the rest + */ + sctp_m_freem(head); + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOBUFS); + return (ENOBUFS); + } + m = SCTP_BUF_NEXT(m); + cancpy = (int)M_TRAILINGSPACE(m); + willcpy = min(cancpy, left); + } else { + sp->tail_mbuf = m; + SCTP_BUF_NEXT(m) = NULL; + } + } + sp->data = head; + sp->length = cpsz; + return (0); +#endif +} + +static struct sctp_stream_queue_pending * +sctp_copy_it_in(struct sctp_tcb *stcb, + struct sctp_association *asoc, + struct sctp_nonpad_sndrcvinfo *srcv, + struct uio *uio, + struct sctp_nets *net, + ssize_t max_send_len, + int user_marks_eor, + int *error) + +{ + /*- + * This routine must be very careful in its work. Protocol + * processing is up and running so care must be taken to spl...() + * when you need to do something that may effect the stcb/asoc. The + * sb is locked however. When data is copied the protocol processing + * should be enabled since this is a slower operation... + */ + struct sctp_stream_queue_pending *sp; + int resv_in_first; + + *error = 0; + sctp_alloc_a_strmoq(stcb, sp); + if (sp == NULL) { + SCTP_LTRACE_ERR_RET(NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOMEM); + *error = ENOMEM; + goto out_now; + } + sp->act_flags = 0; + sp->sender_all_done = 0; + sp->sinfo_flags = srcv->sinfo_flags; + sp->timetolive = srcv->sinfo_timetolive; + sp->ppid = srcv->sinfo_ppid; + sp->context = srcv->sinfo_context; + sp->fsn = 0; + (void)SCTP_GETTIME_TIMEVAL(&sp->ts); + sp->sid = srcv->sinfo_stream; +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + sp->length = (uint32_t)min(uio->uio_resid, max_send_len); +#else + sp->length = (uint32_t)min(uio_resid(uio), max_send_len); +#endif +#else + sp->length = (uint32_t)min(uio->uio_resid, max_send_len); +#endif +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + if ((sp->length == (uint32_t)uio->uio_resid) && +#else + if ((sp->length == (uint32_t)uio_resid(uio)) && +#endif +#else + if ((sp->length == (uint32_t)uio->uio_resid) && +#endif + ((user_marks_eor == 0) || + (srcv->sinfo_flags & SCTP_EOF) || + (user_marks_eor && (srcv->sinfo_flags & SCTP_EOR)))) { + sp->msg_is_complete = 1; + } else { + sp->msg_is_complete = 0; + } + sp->sender_all_done = 0; + sp->some_taken = 0; + sp->put_last_out = 0; + resv_in_first = SCTP_DATA_CHUNK_OVERHEAD(stcb); + sp->data = sp->tail_mbuf = NULL; + if (sp->length == 0) { + goto skip_copy; + } + if (srcv->sinfo_keynumber_valid) { + sp->auth_keyid = srcv->sinfo_keynumber; + } else { + sp->auth_keyid = stcb->asoc.authinfo.active_keyid; + } + if (sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks)) { + sctp_auth_key_acquire(stcb, sp->auth_keyid); + sp->holds_key_ref = 1; + } +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(SCTP_INP_SO(stcb->sctp_ep), 0); +#endif + *error = sctp_copy_one(sp, uio, resv_in_first); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_LOCK(SCTP_INP_SO(stcb->sctp_ep), 0); +#endif + skip_copy: + if (*error) { +#if defined(__Userspace__) + SCTP_TCB_LOCK(stcb); +#endif + sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED); +#if defined(__Userspace__) + SCTP_TCB_UNLOCK(stcb); +#endif + sp = NULL; + } else { + if (sp->sinfo_flags & SCTP_ADDR_OVER) { + sp->net = net; + atomic_add_int(&sp->net->ref_count, 1); + } else { + sp->net = NULL; + } + sctp_set_prsctp_policy(sp); + } +out_now: + return (sp); +} + +int +sctp_sosend(struct socket *so, + struct sockaddr *addr, + struct uio *uio, + struct mbuf *top, + struct mbuf *control, +#if defined(__APPLE__) && !defined(__Userspace__) + int flags) +#else + int flags, +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct thread *p) +#elif defined(_WIN32) && !defined(__Userspace__) + PKTHREAD p) +#else +#if defined(__Userspace__) + /* + * proc is a dummy in __Userspace__ and will not be passed + * to sctp_lower_sosend + */ +#endif + struct proc *p) +#endif +#endif +{ + struct sctp_sndrcvinfo sndrcvninfo; +#if defined(INET) && defined(INET6) + struct sockaddr_in sin; +#endif + struct sockaddr *addr_to_use; +#if defined(__APPLE__) && !defined(__Userspace__) + struct proc *p = current_proc(); +#endif + int error; + bool use_sndinfo; + + if (control != NULL) { + /* process cmsg snd/rcv info (maybe a assoc-id) */ + use_sndinfo = sctp_find_cmsg(SCTP_SNDRCV, (void *)&sndrcvninfo, control, sizeof(sndrcvninfo)); + } else { + use_sndinfo = false; + } +#if defined(INET) && defined(INET6) + if ((addr != NULL) && (addr->sa_family == AF_INET6)) { + struct sockaddr_in6 *sin6; + +#ifdef HAVE_SA_LEN + if (addr->sa_len != sizeof(struct sockaddr_in6)) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); + return (EINVAL); + } +#endif + sin6 = (struct sockaddr_in6 *)addr; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + in6_sin6_2_sin(&sin, sin6); + addr_to_use = (struct sockaddr *)&sin; + } else { + addr_to_use = addr; + } + } else { + addr_to_use = addr; + } +#else + addr_to_use = addr; +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_LOCK(so, 1); +#endif + error = sctp_lower_sosend(so, addr_to_use, uio, top, control, flags, +#if defined(__Userspace__) + use_sndinfo ? &sndrcvninfo : NULL); +#else + use_sndinfo ? &sndrcvninfo : NULL, p); +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + return (error); +} + +int +sctp_lower_sosend(struct socket *so, + struct sockaddr *addr, + struct uio *uio, + struct mbuf *top, + struct mbuf *control, + int flags, +#if defined(__Userspace__) + struct sctp_sndrcvinfo *srcv) +#else + struct sctp_sndrcvinfo *srcv, +#if defined(__FreeBSD__) + struct thread *p) +#elif defined(_WIN32) + PKTHREAD p) +#else + struct proc *p) +#endif +#endif +{ + struct sctp_nonpad_sndrcvinfo sndrcvninfo_buf; +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif + struct timeval now; + struct sctp_block_entry be; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb = NULL; + struct sctp_nets *net; + struct sctp_association *asoc; + struct sctp_inpcb *t_inp; + struct sctp_nonpad_sndrcvinfo *sndrcvninfo; + ssize_t sndlen = 0, max_len, local_add_more; + ssize_t local_soresv = 0; + sctp_assoc_t sinfo_assoc_id; + int user_marks_eor; + int nagle_applies = 0; + int error; + int queue_only = 0, queue_only_for_init = 0; + int un_sent; + int now_filled = 0; + unsigned int inqueue_bytes = 0; + uint16_t port; + uint16_t sinfo_flags; + uint16_t sinfo_stream; + bool create_lock_applied = false; + bool free_cnt_applied = false; + bool some_on_control; + bool got_all_of_the_send = false; + bool non_blocking = false; + + error = 0; + net = NULL; + stcb = NULL; + +#if defined(__APPLE__) && !defined(__Userspace__) + sctp_lock_assert(so); +#endif + if ((uio == NULL) && (top == NULL)) { + error = EINVAL; + goto out_unlocked; + } + if (addr != NULL) { + union sctp_sockstore *raddr = (union sctp_sockstore *)addr; + + switch (raddr->sa.sa_family) { +#ifdef INET + case AF_INET: +#ifdef HAVE_SIN_LEN + if (raddr->sin.sin_len != sizeof(struct sockaddr_in)) { + error = EINVAL; + goto out_unlocked; + } +#endif + port = raddr->sin.sin_port; + break; +#endif +#ifdef INET6 + case AF_INET6: +#ifdef HAVE_SIN6_LEN + if (raddr->sin6.sin6_len != sizeof(struct sockaddr_in6)) { + error = EINVAL; + goto out_unlocked; + } +#endif + port = raddr->sin6.sin6_port; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: +#ifdef HAVE_SCONN_LEN + if (raddr->sconn.sconn_len != sizeof(struct sockaddr_conn)) { + error = EINVAL; + goto out_unlocked; + } +#endif + port = raddr->sconn.sconn_port; + break; +#endif + default: + error = EAFNOSUPPORT; + goto out_unlocked; + } + } else { + port = 0; + } + if (uio != NULL) { +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + if (uio->uio_resid < 0) { +#else + if (uio_resid(uio) < 0) { +#endif +#else + if (uio->uio_resid < 0) { +#endif + error = EINVAL; + goto out_unlocked; + } +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + sndlen = uio->uio_resid; +#else + sndlen = uio_resid(uio); +#endif +#else + sndlen = uio->uio_resid; +#endif + } else { + sndlen = SCTP_HEADER_LEN(top); + } + SCTPDBG(SCTP_DEBUG_OUTPUT1, "Send called addr:%p send length %zd\n", + (void *)addr, sndlen); + + t_inp = inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + error = EINVAL; + goto out_unlocked; + } + user_marks_eor = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR); + if ((uio == NULL) && (user_marks_eor != 0)) { + /*- + * We do not support eeor mode for + * sending with mbuf chains (like sendfile). + */ + error = EINVAL; + goto out_unlocked; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && + SCTP_IS_LISTENING(inp)) { + /* The listener can NOT send. */ + error = EINVAL; + goto out_unlocked; + } + atomic_add_int(&inp->total_sends, 1); + + if (srcv != NULL) { + sndrcvninfo = (struct sctp_nonpad_sndrcvinfo *)srcv; + sinfo_assoc_id = sndrcvninfo->sinfo_assoc_id; + sinfo_flags = sndrcvninfo->sinfo_flags; + if (INVALID_SINFO_FLAG(sinfo_flags) || + PR_SCTP_INVALID_POLICY(sinfo_flags)) { + error = EINVAL; + goto out_unlocked; + } + if (sinfo_flags != 0) { + SCTP_STAT_INCR(sctps_sends_with_flags); + } + } else { + sndrcvninfo = NULL; + sinfo_flags = inp->def_send.sinfo_flags; + sinfo_assoc_id = inp->def_send.sinfo_assoc_id; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (flags & MSG_EOR) { + sinfo_flags |= SCTP_EOR; + } + if (flags & MSG_EOF) { + sinfo_flags |= SCTP_EOF; + } +#endif + if ((sinfo_flags & SCTP_ADDR_OVER) && (addr == NULL)) { + error = EINVAL; + goto out_unlocked; + } + SCTP_INP_RLOCK(inp); + if ((sinfo_flags & SCTP_SENDALL) && + (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE)) { + SCTP_INP_RUNLOCK(inp); + error = sctp_sendall(inp, uio, top, sndrcvninfo); + top = NULL; + goto out_unlocked; + } + /* Now we must find the association. */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb != NULL) { + SCTP_TCB_LOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } else if (sinfo_assoc_id > SCTP_ALL_ASSOC) { + stcb = sctp_findasoc_ep_asocid_locked(inp, sinfo_assoc_id, 1); + SCTP_INP_RUNLOCK(inp); + if (stcb != NULL) { + SCTP_TCB_LOCK_ASSERT(stcb); + } + } else if (addr != NULL) { + /*- + * Since we did not use findep we must + * increment it, and if we don't find a tcb + * decrement it. + */ + SCTP_INP_INCR_REF(inp); + SCTP_INP_RUNLOCK(inp); + stcb = sctp_findassociation_ep_addr(&t_inp, addr, &net, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_TCB_LOCK_ASSERT(stcb); + } + } else { + SCTP_INP_RUNLOCK(inp); + } + +#ifdef INVARIANTS + if (stcb != NULL) { + SCTP_TCB_LOCK_ASSERT(stcb); + } +#endif + + if ((stcb == NULL) && (addr != NULL)) { + /* Possible implicit send? */ + SCTP_ASOC_CREATE_LOCK(inp); + create_lock_applied = true; + if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { + error = EINVAL; + goto out_unlocked; + } + if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) && + (addr->sa_family == AF_INET6)) { + error = EINVAL; + goto out_unlocked; + } + SCTP_INP_WLOCK(inp); + SCTP_INP_INCR_REF(inp); + SCTP_INP_WUNLOCK(inp); + /* With the lock applied look again */ + stcb = sctp_findassociation_ep_addr(&t_inp, addr, &net, NULL, NULL); +#if defined(INET) || defined(INET6) + if ((stcb == NULL) && (control != NULL) && (port > 0)) { + stcb = sctp_findassociation_cmsgs(&t_inp, port, control, &net, &error); + } +#endif + if (stcb == NULL) { + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_TCB_LOCK_ASSERT(stcb); + SCTP_ASOC_CREATE_UNLOCK(inp); + create_lock_applied = false; + } + if (error != 0) { + goto out_unlocked; + } + if (t_inp != inp) { + error = ENOTCONN; + goto out_unlocked; + } + } + if (stcb == NULL) { + if (addr == NULL) { + error = ENOENT; + goto out_unlocked; + } else { + /* We must go ahead and start the INIT process */ + uint32_t vrf_id; + + if ((sinfo_flags & SCTP_ABORT) || + ((sinfo_flags & SCTP_EOF) && (sndlen == 0))) { + /*- + * User asks to abort a non-existent assoc, + * or EOF a non-existent assoc with no data + */ + error = ENOENT; + goto out_unlocked; + } + /* get an asoc/stcb struct */ + vrf_id = inp->def_vrf_id; + KASSERT(create_lock_applied, ("create_lock_applied is false")); + stcb = sctp_aloc_assoc_connected(inp, addr, &error, 0, 0, vrf_id, + inp->sctp_ep.pre_open_stream_count, + inp->sctp_ep.port, +#if !defined(__Userspace__) + p, +#else + (struct proc *)NULL, +#endif + SCTP_INITIALIZE_AUTH_PARAMS); + if (stcb == NULL) { + /* error is setup for us in the call. */ + KASSERT(error != 0, ("error is 0 although stcb is NULL")); + goto out_unlocked; + } + SCTP_TCB_LOCK_ASSERT(stcb); + SCTP_ASOC_CREATE_UNLOCK(inp); + create_lock_applied = false; + /* Turn on queue only flag to prevent data from being sent */ + queue_only = 1; + SCTP_SET_STATE(stcb, SCTP_STATE_COOKIE_WAIT); + (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); + if (control != NULL) { + if (sctp_process_cmsgs_for_init(stcb, control, &error)) { + sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_6); + stcb = NULL; + KASSERT(error != 0, + ("error is 0 although sctp_process_cmsgs_for_init() indicated an error")); + goto out_unlocked; + } + } + /* out with the INIT */ + queue_only_for_init = 1; + /*- + * we may want to dig in after this call and adjust the MTU + * value. It defaulted to 1500 (constant) but the ro + * structure may now have an update and thus we may need to + * change it BEFORE we append the message. + */ + } + } + + KASSERT(!create_lock_applied, ("create_lock_applied is true")); + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + + asoc = &stcb->asoc; + if ((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) || + (asoc->state & SCTP_STATE_WAS_ABORTED)) { + if (asoc->state & SCTP_STATE_WAS_ABORTED) { + /* XXX: Could also be ECONNABORTED, not enough info. */ + error = ECONNRESET; + } else { + error = ENOTCONN; + } + goto out_unlocked; + } + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) { + queue_only = 1; + } + /* Keep the stcb from being freed under our feet. */ + atomic_add_int(&asoc->refcnt, 1); + free_cnt_applied = true; + if (sndrcvninfo == NULL) { + /* Use a local copy to have a consistent view. */ + sndrcvninfo_buf = asoc->def_send; + sndrcvninfo = &sndrcvninfo_buf; + sinfo_flags = sndrcvninfo->sinfo_flags; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (flags & MSG_EOR) { + sinfo_flags |= SCTP_EOR; + } + if (flags & MSG_EOF) { + sinfo_flags |= SCTP_EOF; + } +#endif + } + /* Are we aborting? */ + if (sinfo_flags & SCTP_ABORT) { + struct mbuf *mm; + struct sctp_paramhdr *ph; + ssize_t tot_demand, tot_out = 0, max_out; + + SCTP_STAT_INCR(sctps_sends_with_abort); + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) { + /* It has to be up before we abort. */ + error = EINVAL; + goto out_unlocked; + } + /* How big is the user initiated abort? */ + if (top != NULL) { + struct mbuf *cntm; + + if (sndlen != 0) { + for (cntm = top; cntm; cntm = SCTP_BUF_NEXT(cntm)) { + tot_out += SCTP_BUF_LEN(cntm); + } + } + mm = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr), 0, M_NOWAIT, 1, MT_DATA); + } else { + /* Must fit in a MTU */ + tot_out = sndlen; + tot_demand = (tot_out + sizeof(struct sctp_paramhdr)); + if (tot_demand > SCTP_DEFAULT_ADD_MORE) { + error = EMSGSIZE; + goto out_unlocked; + } + mm = sctp_get_mbuf_for_msg((unsigned int)tot_demand, 0, M_NOWAIT, 1, MT_DATA); + } + if (mm == NULL) { + error = ENOMEM; + goto out_unlocked; + } + max_out = asoc->smallest_mtu - sizeof(struct sctp_paramhdr); + max_out -= sizeof(struct sctp_abort_msg); + if (tot_out > max_out) { + tot_out = max_out; + } + ph = mtod(mm, struct sctp_paramhdr *); + ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT); + ph->param_length = htons((uint16_t)(sizeof(struct sctp_paramhdr) + tot_out)); + ph++; + SCTP_BUF_LEN(mm) = (int)(tot_out + sizeof(struct sctp_paramhdr)); + if (top == NULL) { + SCTP_TCB_UNLOCK(stcb); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 0); +#endif + error = uiomove((caddr_t)ph, (int)tot_out, uio); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_LOCK(so, 0); +#endif + SCTP_TCB_LOCK(stcb); + if ((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) || + (asoc->state & SCTP_STATE_WAS_ABORTED)) { + sctp_m_freem(mm); + if (asoc->state & SCTP_STATE_WAS_ABORTED) { + /* XXX: Could also be ECONNABORTED, not enough info. */ + error = ECONNRESET; + } else { + error = ENOTCONN; + } + goto out_unlocked; + } + if (error != 0) { + /*- + * Here if we can't get his data we + * still abort we just don't get to + * send the users note :-0 + */ + sctp_m_freem(mm); + mm = NULL; + error = 0; + } + } else { + if (sndlen != 0) { + SCTP_BUF_NEXT(mm) = top; + } + } + atomic_subtract_int(&asoc->refcnt, 1); + free_cnt_applied = false; + /* release this lock, otherwise we hang on ourselves */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + sctp_abort_an_association(stcb->sctp_ep, stcb, mm, false, SCTP_SO_LOCKED); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + stcb = NULL; + /* In this case top is already chained to mm + * avoid double free, since we free it below if + * top != NULL and driver would free it after sending + * the packet out + */ + if (sndlen != 0) { + top = NULL; + } + goto out_unlocked; + } + + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + KASSERT((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0, + ("Association about to be freed")); + KASSERT((asoc->state & SCTP_STATE_WAS_ABORTED) == 0, + ("Association was aborted")); + + if (sinfo_flags & SCTP_ADDR_OVER) { + if (addr != NULL) { + net = sctp_findnet(stcb, addr); + } else { + net = NULL; + } + if ((net == NULL) || + ((port != 0) && (port != stcb->rport))) { + error = EINVAL; + goto out_unlocked; + } + } else { + if (asoc->alternate != NULL) { + net = asoc->alternate; + } else { + net = asoc->primary_destination; + } + } + if (sndlen == 0) { + if (sinfo_flags & SCTP_EOF) { + got_all_of_the_send = true; + goto dataless_eof; + } else { + error = EINVAL; + goto out_unlocked; + } + } + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NO_FRAGMENT)) { + if (sndlen > (ssize_t)asoc->smallest_mtu) { + error = EMSGSIZE; + goto out_unlocked; + } + } + sinfo_stream = sndrcvninfo->sinfo_stream; + /* Is the stream no. valid? */ + if (sinfo_stream >= asoc->streamoutcnt) { + /* Invalid stream number */ + error = EINVAL; + goto out_unlocked; + } + if ((asoc->strmout[sinfo_stream].state != SCTP_STREAM_OPEN) && + (asoc->strmout[sinfo_stream].state != SCTP_STREAM_OPENING)) { + /* + * Can't queue any data while stream reset is underway. + */ + if (asoc->strmout[sinfo_stream].state > SCTP_STREAM_OPEN) { + error = EAGAIN; + } else { + error = EINVAL; + } + goto out_unlocked; + } + atomic_add_int(&stcb->total_sends, 1); +#if defined(__Userspace__) + if (inp->recv_callback != NULL) { + non_blocking = true; + } +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (SCTP_SO_IS_NBIO(so) || (flags & (MSG_NBIO | MSG_DONTWAIT)) != 0) { +#else + if (SCTP_SO_IS_NBIO(so)) { +#endif + non_blocking = true; + } + if (non_blocking) { + ssize_t amount; + + inqueue_bytes = asoc->total_output_queue_size - (asoc->chunks_on_out_queue * SCTP_DATA_CHUNK_OVERHEAD(stcb)); + if (user_marks_eor == 0) { + amount = sndlen; + } else { + amount = 1; + } + if ((SCTP_SB_LIMIT_SND(so) < (amount + inqueue_bytes + asoc->sb_send_resv)) || + (asoc->chunks_on_out_queue >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue))) { + if ((sndlen > (ssize_t)SCTP_SB_LIMIT_SND(so)) && + (user_marks_eor == 0)) { + error = EMSGSIZE; + } else { + error = EWOULDBLOCK; + } + goto out_unlocked; + } + } + atomic_add_int(&asoc->sb_send_resv, (int)sndlen); + local_soresv = sndlen; + + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + KASSERT((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0, + ("Association about to be freed")); + KASSERT((asoc->state & SCTP_STATE_WAS_ABORTED) == 0, + ("Association was aborted")); + + /* Ok, we will attempt a msgsnd :> */ +#if !(defined(_WIN32) || defined(__Userspace__)) + if (p != NULL) { +#if defined(__FreeBSD__) + p->td_ru.ru_msgsnd++; +#else + p->p_stats->p_ru.ru_msgsnd++; +#endif + } +#endif + /* Calculate the maximum we can send */ + inqueue_bytes = asoc->total_output_queue_size - (asoc->chunks_on_out_queue * SCTP_DATA_CHUNK_OVERHEAD(stcb)); + if (SCTP_SB_LIMIT_SND(so) > inqueue_bytes) { + max_len = SCTP_SB_LIMIT_SND(so) - inqueue_bytes; + } else { + max_len = 0; + } + /* Unless E_EOR mode is on, we must make a send FIT in one call. */ + if ((user_marks_eor == 0) && + (sndlen > (ssize_t)SCTP_SB_LIMIT_SND(stcb->sctp_socket))) { + /* It will NEVER fit. */ + error = EMSGSIZE; + goto out_unlocked; + } + if (user_marks_eor != 0) { + local_add_more = (ssize_t)min(SCTP_SB_LIMIT_SND(so), SCTP_BASE_SYSCTL(sctp_add_more_threshold)); + } else { + /*- + * For non-eeor the whole message must fit in + * the socket send buffer. + */ + local_add_more = sndlen; + } + if (non_blocking) { + goto skip_preblock; + } + if (((max_len <= local_add_more) && ((ssize_t)SCTP_SB_LIMIT_SND(so) >= local_add_more)) || + (max_len == 0) || + ((asoc->chunks_on_out_queue + asoc->stream_queue_cnt) >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue))) { + /* No room right now! */ + inqueue_bytes = asoc->total_output_queue_size - (asoc->chunks_on_out_queue * SCTP_DATA_CHUNK_OVERHEAD(stcb)); + SOCKBUF_LOCK(&so->so_snd); + while ((SCTP_SB_LIMIT_SND(so) < (inqueue_bytes + local_add_more)) || + ((asoc->stream_queue_cnt + asoc->chunks_on_out_queue) >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue))) { + SCTPDBG(SCTP_DEBUG_OUTPUT1,"pre_block limit:%u <(inq:%d + %zd) || (%d+%d > %d)\n", + (unsigned int)SCTP_SB_LIMIT_SND(so), + inqueue_bytes, + local_add_more, + asoc->stream_queue_cnt, + asoc->chunks_on_out_queue, + SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue)); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { + sctp_log_block(SCTP_BLOCK_LOG_INTO_BLKA, asoc, sndlen); + } + be.error = 0; +#if !(defined(_WIN32) && !defined(__Userspace__)) + stcb->block_entry = &be; +#endif + SCTP_TCB_UNLOCK(stcb); +#if defined(__FreeBSD__) && !defined(__Userspace__) + error = sbwait(so, SO_SND); +#else + error = sbwait(&so->so_snd); +#endif + if (error == 0) { + if (so->so_error != 0) { + error = so->so_error; + } + if (be.error != 0) { + error = be.error; + } + } + SOCKBUF_UNLOCK(&so->so_snd); + SCTP_TCB_LOCK(stcb); + stcb->block_entry = NULL; + if (error != 0) { + goto out_unlocked; + } + if ((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) || + (asoc->state & SCTP_STATE_WAS_ABORTED)) { + if (asoc->state & SCTP_STATE_WAS_ABORTED) { + /* XXX: Could also be ECONNABORTED, not enough info. */ + error = ECONNRESET; + } else { + error = ENOTCONN; + } + goto out_unlocked; + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { + sctp_log_block(SCTP_BLOCK_LOG_OUTOF_BLK, + asoc, asoc->total_output_queue_size); + } + inqueue_bytes = asoc->total_output_queue_size - (asoc->chunks_on_out_queue * SCTP_DATA_CHUNK_OVERHEAD(stcb)); + SOCKBUF_LOCK(&so->so_snd); + } + if (SCTP_SB_LIMIT_SND(so) > inqueue_bytes) { + max_len = SCTP_SB_LIMIT_SND(so) - inqueue_bytes; + } else { + max_len = 0; + } + SOCKBUF_UNLOCK(&so->so_snd); + } + +skip_preblock: + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + KASSERT((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0, + ("Association about to be freed")); + KASSERT((asoc->state & SCTP_STATE_WAS_ABORTED) == 0, + ("Association was aborted")); + +#if defined(__APPLE__) && !defined(__Userspace__) + error = sblock(&so->so_snd, SBLOCKWAIT(flags)); + if (error != 0) { + goto out_unlocked; + } +#endif + /* sndlen covers for mbuf case + * uio_resid covers for the non-mbuf case + * NOTE: uio will be null when top/mbuf is passed + */ + if (top == NULL) { + struct sctp_stream_queue_pending *sp; + struct sctp_stream_out *strm; + uint32_t sndout; + + if ((asoc->stream_locked) && + (asoc->stream_locked_on != sinfo_stream)) { + error = EINVAL; + goto out; + } + strm = &asoc->strmout[sinfo_stream]; + if (strm->last_msg_incomplete == 0) { + do_a_copy_in: + SCTP_TCB_UNLOCK(stcb); + sp = sctp_copy_it_in(stcb, asoc, sndrcvninfo, uio, net, max_len, user_marks_eor, &error); + SCTP_TCB_LOCK(stcb); + if ((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) || + (asoc->state & SCTP_STATE_WAS_ABORTED)) { + if (asoc->state & SCTP_STATE_WAS_ABORTED) { + /* XXX: Could also be ECONNABORTED, not enough info. */ + error = ECONNRESET; + } else { + error = ENOTCONN; + } + goto out; + } + if (error != 0) { + goto out; + } + /* + * Reject the sending of a new user message, if the + * association is about to be shut down. + */ + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_SENT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_ACK_SENT) || + (asoc->state & SCTP_STATE_SHUTDOWN_PENDING)) { + if (sp->data != 0) { + sctp_m_freem(sp->data); + sp->data = NULL; + sp->tail_mbuf = NULL; + sp->length = 0; + } + if (sp->net != NULL) { + sctp_free_remote_addr(sp->net); + sp->net = NULL; + } + sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED); + error = EPIPE; + goto out_unlocked; + } + /* The out streams might be reallocated. */ + strm = &asoc->strmout[sinfo_stream]; + if (sp->msg_is_complete) { + strm->last_msg_incomplete = 0; + asoc->stream_locked = 0; + } else { + /* Just got locked to this guy in + * case of an interrupt. + */ + strm->last_msg_incomplete = 1; + if (asoc->idata_supported == 0) { + asoc->stream_locked = 1; + asoc->stream_locked_on = sinfo_stream; + } + sp->sender_all_done = 0; + } + sctp_snd_sb_alloc(stcb, sp->length); + atomic_add_int(&asoc->stream_queue_cnt, 1); + if (sinfo_flags & SCTP_UNORDERED) { + SCTP_STAT_INCR(sctps_sends_with_unord); + } + sp->processing = 1; + TAILQ_INSERT_TAIL(&strm->outqueue, sp, next); + asoc->ss_functions.sctp_ss_add_to_stream(stcb, asoc, strm, sp); + } else { + sp = TAILQ_LAST(&strm->outqueue, sctp_streamhead); + if (sp == NULL) { + /* ???? Huh ??? last msg is gone */ +#ifdef INVARIANTS + panic("Warning: Last msg marked incomplete, yet nothing left?"); +#else + SCTP_PRINTF("Warning: Last msg marked incomplete, yet nothing left?\n"); + strm->last_msg_incomplete = 0; +#endif + goto do_a_copy_in; + } + if (sp->processing != 0) { + error = EINVAL; + goto out; + } else { + sp->processing = 1; + } + } + + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + KASSERT((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0, + ("Association about to be freed")); + KASSERT((asoc->state & SCTP_STATE_WAS_ABORTED) == 0, + ("Association was aborted")); + +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + while (uio->uio_resid > 0) { +#else + while (uio_resid(uio) > 0) { +#endif +#else + while (uio->uio_resid > 0) { +#endif + /* How much room do we have? */ + struct mbuf *new_tail, *mm; + + inqueue_bytes = asoc->total_output_queue_size - (asoc->chunks_on_out_queue * SCTP_DATA_CHUNK_OVERHEAD(stcb)); + if (SCTP_SB_LIMIT_SND(so) > inqueue_bytes) { + max_len = SCTP_SB_LIMIT_SND(so) - inqueue_bytes; + } else { + max_len = 0; + } + if ((max_len > (ssize_t)SCTP_BASE_SYSCTL(sctp_add_more_threshold)) || + ((max_len > 0 ) && (SCTP_SB_LIMIT_SND(so) < SCTP_BASE_SYSCTL(sctp_add_more_threshold))) || +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + (uio->uio_resid <= max_len)) { +#else + (uio_resid(uio) <= max_len)) { +#endif +#else + (uio->uio_resid <= max_len)) { +#endif + SCTP_TCB_UNLOCK(stcb); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 0); +#endif + sndout = 0; + new_tail = NULL; +#if defined(__FreeBSD__) || defined(__Userspace__) + mm = sctp_copy_resume(uio, (int)max_len, user_marks_eor, &error, &sndout, &new_tail); +#else + mm = sctp_copy_resume(uio, (int)max_len, &error, &sndout, &new_tail); +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_LOCK(so, 0); +#endif + SCTP_TCB_LOCK(stcb); + if ((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) || + (asoc->state & SCTP_STATE_WAS_ABORTED)) { + /* We need to get out. + * Peer probably aborted. + */ + sctp_m_freem(mm); + if (asoc->state & SCTP_STATE_WAS_ABORTED) { + /* XXX: Could also be ECONNABORTED, not enough info. */ + error = ECONNRESET; + } else { + error = ENOTCONN; + } + goto out; + } + if ((mm == NULL) || (error != 0)) { + if (mm != NULL) { + sctp_m_freem(mm); + } + if (sp != NULL) { + sp->processing = 0; + } + goto out; + } + /* Update the mbuf and count */ + if (sp->tail_mbuf != NULL) { + /* Tack it to the end. */ + SCTP_BUF_NEXT(sp->tail_mbuf) = mm; + } else { + /* A stolen mbuf. */ + sp->data = mm; + } + sp->tail_mbuf = new_tail; + sctp_snd_sb_alloc(stcb, sndout); + atomic_add_int(&sp->length, sndout); + if (sinfo_flags & SCTP_SACK_IMMEDIATELY) { + sp->sinfo_flags |= SCTP_SACK_IMMEDIATELY; + } + + /* Did we reach EOR? */ +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + if ((uio->uio_resid == 0) && +#else + if ((uio_resid(uio) == 0) && +#endif +#else + if ((uio->uio_resid == 0) && +#endif + ((user_marks_eor == 0) || + (sinfo_flags & SCTP_EOF) || + (user_marks_eor && (sinfo_flags & SCTP_EOR)))) { + sp->msg_is_complete = 1; + } else { + sp->msg_is_complete = 0; + } + } + + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + KASSERT((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0, + ("Association about to be freed")); + KASSERT((asoc->state & SCTP_STATE_WAS_ABORTED) == 0, + ("Association was aborted")); + +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + if (uio->uio_resid == 0) { +#else + if (uio_resid(uio) == 0) { +#endif +#else + if (uio->uio_resid == 0) { +#endif + /* got it all? */ + continue; + } + /* PR-SCTP? */ + if ((asoc->prsctp_supported) && (asoc->sent_queue_cnt_removeable > 0)) { + /* This is ugly but we must assure locking order */ + sctp_prune_prsctp(stcb, asoc, sndrcvninfo, (int)sndlen); + inqueue_bytes = asoc->total_output_queue_size - (asoc->chunks_on_out_queue * SCTP_DATA_CHUNK_OVERHEAD(stcb)); + if (SCTP_SB_LIMIT_SND(so) > inqueue_bytes) + max_len = SCTP_SB_LIMIT_SND(so) - inqueue_bytes; + else + max_len = 0; + if (max_len > 0) { + continue; + } + } + /* wait for space now */ + if (non_blocking) { + /* Non-blocking io in place out */ + if (sp != NULL) { + sp->processing = 0; + } + goto skip_out_eof; + } + /* What about the INIT, send it maybe */ + if (queue_only_for_init) { + if (SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) { + /* a collision took us forward? */ + queue_only = 0; + } else { +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + SCTP_SET_STATE(stcb, SCTP_STATE_COOKIE_WAIT); + queue_only = 1; + } + } + if ((net->flight_size > net->cwnd) && + (asoc->sctp_cmt_on_off == 0)) { + SCTP_STAT_INCR(sctps_send_cwnd_avoid); + queue_only = 1; + } else if (asoc->ifp_had_enobuf) { + SCTP_STAT_INCR(sctps_ifnomemqueued); + if (net->flight_size > (2 * net->mtu)) { + queue_only = 1; + } + asoc->ifp_had_enobuf = 0; + } + un_sent = asoc->total_output_queue_size - asoc->total_flight; + if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) && + (asoc->total_flight > 0) && + (asoc->stream_queue_cnt < SCTP_MAX_DATA_BUNDLING) && + (un_sent < (int)(asoc->smallest_mtu - SCTP_MIN_OVERHEAD))) { + /*- + * Ok, Nagle is set on and we have data outstanding. + * Don't send anything and let SACKs drive out the + * data unless we have a "full" segment to send. + */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_NAGLE_LOGGING_ENABLE) { + sctp_log_nagle_event(stcb, SCTP_NAGLE_APPLIED); + } + SCTP_STAT_INCR(sctps_naglequeued); + nagle_applies = 1; + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_NAGLE_LOGGING_ENABLE) { + if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) + sctp_log_nagle_event(stcb, SCTP_NAGLE_SKIPPED); + } + SCTP_STAT_INCR(sctps_naglesent); + nagle_applies = 0; + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_CWNDLOG_PRESEND, queue_only_for_init, queue_only, + nagle_applies, un_sent); + sctp_misc_ints(SCTP_CWNDLOG_PRESEND, asoc->total_output_queue_size, + asoc->total_flight, + asoc->chunks_on_out_queue, asoc->total_flight_count); + } + if (queue_only_for_init) { + queue_only_for_init = 0; + } + if ((queue_only == 0) && (nagle_applies == 0)) { + /*- + * need to start chunk output + * before blocking.. note that if + * a lock is already applied, then + * the input via the net is happening + * and I don't need to start output :-D + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + sctp_chunk_output(inp, stcb, + SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + } + /*- + * This is a bit strange, but I think it will + * work. The total_output_queue_size is locked and + * protected by the TCB_LOCK, which we just released. + * There is a race that can occur between releasing it + * above, and me getting the socket lock, where sacks + * come in but we have not put the SB_WAIT on the + * so_snd buffer to get the wakeup. After the LOCK + * is applied the sack_processing will also need to + * LOCK the so->so_snd to do the actual sowwakeup(). So + * once we have the socket buffer lock if we recheck the + * size we KNOW we will get to sleep safely with the + * wakeup flag in place. + */ + inqueue_bytes = asoc->total_output_queue_size - (asoc->chunks_on_out_queue * SCTP_DATA_CHUNK_OVERHEAD(stcb)); + SOCKBUF_LOCK(&so->so_snd); + if (SCTP_SB_LIMIT_SND(so) <= (inqueue_bytes + + min(SCTP_BASE_SYSCTL(sctp_add_more_threshold), SCTP_SB_LIMIT_SND(so)))) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + sctp_log_block(SCTP_BLOCK_LOG_INTO_BLK, + asoc, uio->uio_resid); +#else + sctp_log_block(SCTP_BLOCK_LOG_INTO_BLK, + asoc, uio_resid(uio)); +#endif +#else + sctp_log_block(SCTP_BLOCK_LOG_INTO_BLK, + asoc, uio->uio_resid); +#endif + } + be.error = 0; +#if !(defined(_WIN32) && !defined(__Userspace__)) + stcb->block_entry = &be; +#endif + SCTP_TCB_UNLOCK(stcb); +#if defined(__APPLE__) && !defined(__Userspace__) + sbunlock(&so->so_snd, 1); +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + error = sbwait(so, SO_SND); +#else + error = sbwait(&so->so_snd); +#endif + if (error == 0) { + if (so->so_error != 0) + error = so->so_error; + if (be.error != 0) { + error = be.error; + } + } + SOCKBUF_UNLOCK(&so->so_snd); + SCTP_TCB_LOCK(stcb); + stcb->block_entry = NULL; + if ((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) || + (asoc->state & SCTP_STATE_WAS_ABORTED)) { + if (asoc->state & SCTP_STATE_WAS_ABORTED) { + /* XXX: Could also be ECONNABORTED, not enough info. */ + error = ECONNRESET; + } else { + error = ENOTCONN; + } + goto out_unlocked; + } + if (error != 0) { + if (sp != NULL) { + sp->processing = 0; + } + goto out_unlocked; + } +#if defined(__APPLE__) && !defined(__Userspace__) + error = sblock(&so->so_snd, SBLOCKWAIT(flags)); + if (error != 0) { + goto out_unlocked; + } +#endif + } else { + SOCKBUF_UNLOCK(&so->so_snd); + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { + sctp_log_block(SCTP_BLOCK_LOG_OUTOF_BLK, + asoc, asoc->total_output_queue_size); + } + } + + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + KASSERT((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0, + ("Association about to be freed")); + KASSERT((asoc->state & SCTP_STATE_WAS_ABORTED) == 0, + ("Association was aborted")); + + /* The out streams might be reallocated. */ + strm = &asoc->strmout[sinfo_stream]; + if (sp != NULL) { + if (sp->msg_is_complete == 0) { + strm->last_msg_incomplete = 1; + if (asoc->idata_supported == 0) { + asoc->stream_locked = 1; + asoc->stream_locked_on = sinfo_stream; + } + } else { + sp->sender_all_done = 1; + strm->last_msg_incomplete = 0; + asoc->stream_locked = 0; + } + sp->processing = 0; + } else { + SCTP_PRINTF("Huh no sp TSNH?\n"); + strm->last_msg_incomplete = 0; + asoc->stream_locked = 0; + } +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + if (uio->uio_resid == 0) { +#else + if (uio_resid(uio) == 0) { +#endif +#else + if (uio->uio_resid == 0) { +#endif + got_all_of_the_send = true; + } + } else { + error = sctp_msg_append(stcb, net, top, sndrcvninfo); + top = NULL; + if ((sinfo_flags & SCTP_EOF) != 0) { + got_all_of_the_send = true; + } + } + if (error != 0) { + goto out; + } + +dataless_eof: + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + KASSERT((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0, + ("Association about to be freed")); + KASSERT((asoc->state & SCTP_STATE_WAS_ABORTED) == 0, + ("Association was aborted")); + + /* EOF thing ? */ + if ((sinfo_flags & SCTP_EOF) && got_all_of_the_send) { + SCTP_STAT_INCR(sctps_sends_with_eof); + error = 0; + if (TAILQ_EMPTY(&asoc->send_queue) && + TAILQ_EMPTY(&asoc->sent_queue) && + sctp_is_there_unsent_data(stcb, SCTP_SO_LOCKED) == 0) { + if ((*asoc->ss_functions.sctp_ss_is_user_msgs_incomplete)(stcb, asoc)) { + goto abort_anyway; + } + /* there is nothing queued to send, so I'm done... */ + if ((SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_RECEIVED) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { + struct sctp_nets *netp; + + /* only send SHUTDOWN the first time through */ + if (SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_SENT); + sctp_stop_timers_for_shutdown(stcb); + if (asoc->alternate != NULL) { + netp = asoc->alternate; + } else { + netp = asoc->primary_destination; + } + sctp_send_shutdown(stcb, netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, + netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, + NULL); + } + } else { + /*- + * we still got (or just got) data to send, so set + * SHUTDOWN_PENDING + */ + /*- + * XXX sockets draft says that SCTP_EOF should be + * sent with no data. currently, we will allow user + * data to be sent first and move to + * SHUTDOWN-PENDING + */ + if ((SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_RECEIVED) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { + if ((*asoc->ss_functions.sctp_ss_is_user_msgs_incomplete)(stcb, asoc)) { + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_PARTIAL_MSG_LEFT); + } + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_SHUTDOWN_PENDING); + if (TAILQ_EMPTY(&asoc->send_queue) && + TAILQ_EMPTY(&asoc->sent_queue) && + (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + abort_anyway: + if (free_cnt_applied) { + atomic_subtract_int(&asoc->refcnt, 1); + free_cnt_applied = false; + } + SCTP_SNPRINTF(msg, sizeof(msg), + "%s:%d at %s", __FILE__, __LINE__, __func__); + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + msg); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + sctp_abort_an_association(stcb->sctp_ep, stcb, + op_err, false, SCTP_SO_LOCKED); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + stcb = NULL; + error = ECONNABORTED; + goto out; + } + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, + NULL); + sctp_feature_off(inp, SCTP_PCB_FLAGS_NODELAY); + } + } + } + +skip_out_eof: + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + KASSERT((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0, + ("Association about to be freed")); + KASSERT((asoc->state & SCTP_STATE_WAS_ABORTED) == 0, + ("Association was aborted")); + + some_on_control = !TAILQ_EMPTY(&asoc->control_send_queue); + if (queue_only_for_init) { + if (SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) { + /* a collision took us forward? */ + queue_only = 0; + } else { +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + SCTP_SET_STATE(stcb, SCTP_STATE_COOKIE_WAIT); + queue_only = 1; + } + } + + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + KASSERT((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0, + ("Association about to be freed")); + KASSERT((asoc->state & SCTP_STATE_WAS_ABORTED) == 0, + ("Association was aborted")); + + if ((net->flight_size > net->cwnd) && + (asoc->sctp_cmt_on_off == 0)) { + SCTP_STAT_INCR(sctps_send_cwnd_avoid); + queue_only = 1; + } else if (asoc->ifp_had_enobuf) { + SCTP_STAT_INCR(sctps_ifnomemqueued); + if (net->flight_size > (2 * net->mtu)) { + queue_only = 1; + } + asoc->ifp_had_enobuf = 0; + } + un_sent = asoc->total_output_queue_size - asoc->total_flight; + if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) && + (asoc->total_flight > 0) && + (asoc->stream_queue_cnt < SCTP_MAX_DATA_BUNDLING) && + (un_sent < (int)(asoc->smallest_mtu - SCTP_MIN_OVERHEAD))) { + /*- + * Ok, Nagle is set on and we have data outstanding. + * Don't send anything and let SACKs drive out the + * data unless wen have a "full" segment to send. + */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_NAGLE_LOGGING_ENABLE) { + sctp_log_nagle_event(stcb, SCTP_NAGLE_APPLIED); + } + SCTP_STAT_INCR(sctps_naglequeued); + nagle_applies = 1; + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_NAGLE_LOGGING_ENABLE) { + if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) + sctp_log_nagle_event(stcb, SCTP_NAGLE_SKIPPED); + } + SCTP_STAT_INCR(sctps_naglesent); + nagle_applies = 0; + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_CWNDLOG_PRESEND, queue_only_for_init, queue_only, + nagle_applies, un_sent); + sctp_misc_ints(SCTP_CWNDLOG_PRESEND, asoc->total_output_queue_size, + asoc->total_flight, + asoc->chunks_on_out_queue, asoc->total_flight_count); + } + + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + KASSERT((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0, + ("Association about to be freed")); + KASSERT((asoc->state & SCTP_STATE_WAS_ABORTED) == 0, + ("Association was aborted")); + +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + if ((queue_only == 0) && (nagle_applies == 0) && (asoc->peers_rwnd && un_sent)) { + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); + } else if ((queue_only == 0) && + (asoc->peers_rwnd == 0) && + (asoc->total_flight == 0)) { + /* We get to have a probe outstanding */ + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); + } else if (some_on_control) { + int num_out, reason; + + /* Here we do control only */ + (void)sctp_med_chunk_output(inp, stcb, asoc, &num_out, + &reason, 1, 1, &now, &now_filled, + sctp_get_frag_point(stcb), + SCTP_SO_LOCKED); + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + SCTPDBG(SCTP_DEBUG_OUTPUT1, "USR Send complete qo:%d prw:%d unsent:%d tf:%d cooq:%d toqs:%d err:%d\n", + queue_only, asoc->peers_rwnd, un_sent, + asoc->total_flight, asoc->chunks_on_out_queue, + asoc->total_output_queue_size, error); + + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + KASSERT((asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0, + ("Association about to be freed")); + KASSERT((asoc->state & SCTP_STATE_WAS_ABORTED) == 0, + ("Association was aborted")); + +out: +#if defined(__APPLE__) && !defined(__Userspace__) + sbunlock(&so->so_snd, 1); +#endif +out_unlocked: + if (create_lock_applied) { + SCTP_ASOC_CREATE_UNLOCK(inp); + } + if (stcb != NULL) { + if (local_soresv) { + atomic_subtract_int(&asoc->sb_send_resv, (int)sndlen); + } + if (free_cnt_applied) { + atomic_subtract_int(&asoc->refcnt, 1); + } + SCTP_TCB_UNLOCK(stcb); + } + if (top != NULL) { + sctp_m_freem(top); + } + if (control != NULL) { + sctp_m_freem(control); + } + SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, error); + return (error); +} + +/* + * generate an AUTHentication chunk, if required + */ +struct mbuf * +sctp_add_auth_chunk(struct mbuf *m, struct mbuf **m_end, + struct sctp_auth_chunk **auth_ret, uint32_t * offset, + struct sctp_tcb *stcb, uint8_t chunk) +{ + struct mbuf *m_auth; + struct sctp_auth_chunk *auth; + int chunk_len; + struct mbuf *cn; + + if ((m_end == NULL) || (auth_ret == NULL) || (offset == NULL) || + (stcb == NULL)) + return (m); + + if (stcb->asoc.auth_supported == 0) { + return (m); + } + /* does the requested chunk require auth? */ + if (!sctp_auth_is_required_chunk(chunk, stcb->asoc.peer_auth_chunks)) { + return (m); + } + m_auth = sctp_get_mbuf_for_msg(sizeof(*auth), 0, M_NOWAIT, 1, MT_HEADER); + if (m_auth == NULL) { + /* no mbuf's */ + return (m); + } + /* reserve some space if this will be the first mbuf */ + if (m == NULL) + SCTP_BUF_RESV_UF(m_auth, SCTP_MIN_OVERHEAD); + /* fill in the AUTH chunk details */ + auth = mtod(m_auth, struct sctp_auth_chunk *); + memset(auth, 0, sizeof(*auth)); + auth->ch.chunk_type = SCTP_AUTHENTICATION; + auth->ch.chunk_flags = 0; + chunk_len = sizeof(*auth) + + sctp_get_hmac_digest_len(stcb->asoc.peer_hmac_id); + auth->ch.chunk_length = htons(chunk_len); + auth->hmac_id = htons(stcb->asoc.peer_hmac_id); + /* key id and hmac digest will be computed and filled in upon send */ + + /* save the offset where the auth was inserted into the chain */ + *offset = 0; + for (cn = m; cn; cn = SCTP_BUF_NEXT(cn)) { + *offset += SCTP_BUF_LEN(cn); + } + + /* update length and return pointer to the auth chunk */ + SCTP_BUF_LEN(m_auth) = chunk_len; + m = sctp_copy_mbufchain(m_auth, m, m_end, 1, chunk_len, 0); + if (auth_ret != NULL) + *auth_ret = auth; + + return (m); +} + +#if (defined(__FreeBSD__) || defined(__APPLE__)) && !defined(__Userspace__) +#ifdef INET6 +int +sctp_v6src_match_nexthop(struct sockaddr_in6 *src6, sctp_route_t *ro) +{ + struct nd_prefix *pfx = NULL; + struct nd_pfxrouter *pfxrtr = NULL; + struct sockaddr_in6 gw6; + +#if defined(__FreeBSD__) + if (ro == NULL || ro->ro_nh == NULL || src6->sin6_family != AF_INET6) +#else + if (ro == NULL || ro->ro_rt == NULL || src6->sin6_family != AF_INET6) +#endif + return (0); + + /* get prefix entry of address */ +#if defined(__FreeBSD__) + ND6_RLOCK(); +#endif + LIST_FOREACH(pfx, &MODULE_GLOBAL(nd_prefix), ndpr_entry) { + if (pfx->ndpr_stateflags & NDPRF_DETACHED) + continue; + if (IN6_ARE_MASKED_ADDR_EQUAL(&pfx->ndpr_prefix.sin6_addr, + &src6->sin6_addr, &pfx->ndpr_mask)) + break; + } + /* no prefix entry in the prefix list */ + if (pfx == NULL) { +#if defined(__FreeBSD__) + ND6_RUNLOCK(); +#endif + SCTPDBG(SCTP_DEBUG_OUTPUT2, "No prefix entry for "); + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)src6); + return (0); + } + + SCTPDBG(SCTP_DEBUG_OUTPUT2, "v6src_match_nexthop(), Prefix entry is "); + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)src6); + + /* search installed gateway from prefix entry */ + LIST_FOREACH(pfxrtr, &pfx->ndpr_advrtrs, pfr_entry) { + memset(&gw6, 0, sizeof(struct sockaddr_in6)); + gw6.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + gw6.sin6_len = sizeof(struct sockaddr_in6); +#endif + memcpy(&gw6.sin6_addr, &pfxrtr->router->rtaddr, + sizeof(struct in6_addr)); + SCTPDBG(SCTP_DEBUG_OUTPUT2, "prefix router is "); + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)&gw6); + SCTPDBG(SCTP_DEBUG_OUTPUT2, "installed router is "); +#if defined(__FreeBSD__) + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &ro->ro_nh->gw_sa); +#else + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, ro->ro_rt->rt_gateway); +#endif +#if defined(__FreeBSD__) + if (sctp_cmpaddr((struct sockaddr *)&gw6, &ro->ro_nh->gw_sa)) { + ND6_RUNLOCK(); +#else + if (sctp_cmpaddr((struct sockaddr *)&gw6, ro->ro_rt->rt_gateway)) { +#endif + SCTPDBG(SCTP_DEBUG_OUTPUT2, "pfxrouter is installed\n"); + return (1); + } + } +#if defined(__FreeBSD__) + ND6_RUNLOCK(); +#endif + SCTPDBG(SCTP_DEBUG_OUTPUT2, "pfxrouter is not installed\n"); + return (0); +} +#endif + +int +sctp_v4src_match_nexthop(struct sctp_ifa *sifa, sctp_route_t *ro) +{ +#ifdef INET + struct sockaddr_in *sin, *mask; + struct ifaddr *ifa; + struct in_addr srcnetaddr, gwnetaddr; + +#if defined(__FreeBSD__) + if (ro == NULL || ro->ro_nh == NULL || +#else + if (ro == NULL || ro->ro_rt == NULL || +#endif + sifa->address.sa.sa_family != AF_INET) { + return (0); + } + ifa = (struct ifaddr *)sifa->ifa; + mask = (struct sockaddr_in *)(ifa->ifa_netmask); + sin = &sifa->address.sin; + srcnetaddr.s_addr = (sin->sin_addr.s_addr & mask->sin_addr.s_addr); + SCTPDBG(SCTP_DEBUG_OUTPUT1, "match_nexthop4: src address is "); + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &sifa->address.sa); + SCTPDBG(SCTP_DEBUG_OUTPUT1, "network address is %x\n", srcnetaddr.s_addr); + +#if defined(__FreeBSD__) + sin = &ro->ro_nh->gw4_sa; +#else + sin = (struct sockaddr_in *)ro->ro_rt->rt_gateway; +#endif + gwnetaddr.s_addr = (sin->sin_addr.s_addr & mask->sin_addr.s_addr); + SCTPDBG(SCTP_DEBUG_OUTPUT1, "match_nexthop4: nexthop is "); +#if defined(__FreeBSD__) + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &ro->ro_nh->gw_sa); +#else + SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, ro->ro_rt->rt_gateway); +#endif + SCTPDBG(SCTP_DEBUG_OUTPUT1, "network address is %x\n", gwnetaddr.s_addr); + if (srcnetaddr.s_addr == gwnetaddr.s_addr) { + return (1); + } +#endif + return (0); +} +#elif defined(__Userspace__) +/* TODO __Userspace__ versions of sctp_vXsrc_match_nexthop(). */ +int +sctp_v6src_match_nexthop(struct sockaddr_in6 *src6, sctp_route_t *ro) +{ + return (0); +} +int +sctp_v4src_match_nexthop(struct sctp_ifa *sifa, sctp_route_t *ro) +{ + return (0); +} + +#endif diff --git a/netwerk/sctp/src/netinet/sctp_output.h b/netwerk/sctp/src/netinet/sctp_output.h new file mode 100644 index 0000000000..56858d880e --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_output.h @@ -0,0 +1,240 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_OUTPUT_H_ +#define _NETINET_SCTP_OUTPUT_H_ + +#include <netinet/sctp_header.h> + +#if defined(_KERNEL) || defined(__Userspace__) + +struct mbuf * +sctp_add_addresses_to_i_ia(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_scoping *scope, + struct mbuf *m_at, + int cnt_inits_to, + uint16_t *padding_len, uint16_t *chunk_len); + +int sctp_is_addr_restricted(struct sctp_tcb *, struct sctp_ifa *); + +int +sctp_is_address_in_scope(struct sctp_ifa *ifa, + struct sctp_scoping *scope, + int do_update); + +int +sctp_is_addr_in_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa); + +struct sctp_ifa * +sctp_source_address_selection(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + sctp_route_t *ro, struct sctp_nets *net, + int non_asoc_addr_ok, uint32_t vrf_id); + +#if defined(__FreeBSD__) || defined(__APPLE__) || defined(__Userspace__) +int sctp_v6src_match_nexthop(struct sockaddr_in6 *src6, sctp_route_t *ro); + +int sctp_v4src_match_nexthop(struct sctp_ifa *sifa, sctp_route_t *ro); +#endif + +void sctp_send_initiate(struct sctp_inpcb *, struct sctp_tcb *, int); + +void +sctp_send_initiate_ack(struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *, struct mbuf *, + int, int, + struct sockaddr *, struct sockaddr *, + struct sctphdr *, struct sctp_init_chunk *, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t, uint32_t, +#endif + uint32_t, uint16_t); + +struct mbuf * +sctp_arethere_unrecognized_parameters(struct mbuf *, int, int *, + struct sctp_chunkhdr *, int *, int *); +void sctp_queue_op_err(struct sctp_tcb *, struct mbuf *); + +int +sctp_send_cookie_echo(struct mbuf *, int, int, struct sctp_tcb *, + struct sctp_nets *); + +void sctp_send_cookie_ack(struct sctp_tcb *); + +void +sctp_send_heartbeat_ack(struct sctp_tcb *, struct mbuf *, int, int, + struct sctp_nets *); + +void +sctp_remove_from_wheel(struct sctp_tcb *stcb, + struct sctp_association *asoc, + struct sctp_stream_out *strq, int holds_lock); + +void sctp_send_shutdown(struct sctp_tcb *, struct sctp_nets *); + +void sctp_send_shutdown_ack(struct sctp_tcb *, struct sctp_nets *); + +void sctp_send_shutdown_complete(struct sctp_tcb *, struct sctp_nets *, int); + +void sctp_send_shutdown_complete2(struct sockaddr *, struct sockaddr *, + struct sctphdr *, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t, uint32_t, uint16_t, +#endif + uint32_t, uint16_t); + +void sctp_send_asconf(struct sctp_tcb *, struct sctp_nets *, int addr_locked); + +void sctp_send_asconf_ack(struct sctp_tcb *); + +uint32_t sctp_get_frag_point(struct sctp_tcb *); + +void sctp_toss_old_cookies(struct sctp_tcb *, struct sctp_association *); + +void sctp_toss_old_asconf(struct sctp_tcb *); + +void sctp_fix_ecn_echo(struct sctp_association *); + +void sctp_move_chunks_from_net(struct sctp_tcb *stcb, struct sctp_nets *net); + +#define SCTP_DATA_CHUNK_OVERHEAD(stcb) ((stcb)->asoc.idata_supported ? \ + sizeof(struct sctp_idata_chunk) : \ + sizeof(struct sctp_data_chunk)) + +#if defined(__FreeBSD__) && !defined(__Userspace__) +int +sctp_output(struct sctp_inpcb *, struct mbuf *, struct sockaddr *, + struct mbuf *, struct thread *, int); +#elif defined(_WIN32) && !defined(__Userspace__) +sctp_output(struct sctp_inpcb *, struct mbuf *, struct sockaddr *, + struct mbuf *, PKTHREAD, int); +#else +#if defined(__Userspace__) +/* sctp_output is called bu sctp_sendm. Not using sctp_sendm for __Userspace__ */ +#endif +int +sctp_output(struct sctp_inpcb *, + struct mbuf *, + struct sockaddr *, + struct mbuf *, + struct proc *, int); +#endif + +void sctp_chunk_output(struct sctp_inpcb *, struct sctp_tcb *, int, int); + +void sctp_send_abort_tcb(struct sctp_tcb *, struct mbuf *, int); + +void send_forward_tsn(struct sctp_tcb *, struct sctp_association *); + +void sctp_send_sack(struct sctp_tcb *, int); + +void sctp_send_hb(struct sctp_tcb *, struct sctp_nets *, int); + +void sctp_send_ecn_echo(struct sctp_tcb *, struct sctp_nets *, uint32_t); + +void +sctp_send_packet_dropped(struct sctp_tcb *, struct sctp_nets *, struct mbuf *, + int, int, int); + +void sctp_send_cwr(struct sctp_tcb *, struct sctp_nets *, uint32_t, uint8_t); + +void +sctp_add_stream_reset_result(struct sctp_tmit_chunk *, uint32_t, uint32_t); + +void +sctp_send_deferred_reset_response(struct sctp_tcb *, + struct sctp_stream_reset_list *, + int); + +void +sctp_add_stream_reset_result_tsn(struct sctp_tmit_chunk *, + uint32_t, uint32_t, uint32_t, uint32_t); +int +sctp_send_stream_reset_out_if_possible(struct sctp_tcb *, int); + +int +sctp_send_str_reset_req(struct sctp_tcb *, uint16_t , uint16_t *, + uint8_t, uint8_t, uint8_t, uint16_t, uint16_t, uint8_t); + +void +sctp_send_abort(struct mbuf *, int, struct sockaddr *, struct sockaddr *, + struct sctphdr *, uint32_t, struct mbuf *, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t, uint32_t, uint16_t, +#endif + uint32_t, uint16_t); + +void +sctp_send_operr_to(struct sockaddr *, struct sockaddr *, + struct sctphdr *, uint32_t, struct mbuf *, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t, uint32_t, uint16_t, +#endif + uint32_t, uint16_t); + +#endif /* _KERNEL || __Userspace__ */ + +#if defined(_KERNEL) || defined(__Userspace__) +int +sctp_sosend(struct socket *so, + struct sockaddr *addr, + struct uio *uio, + struct mbuf *top, + struct mbuf *control, +#if defined(__APPLE__) && !defined(__Userspace__) + int flags +#else + int flags, +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct thread *p +#elif defined(_WIN32) && !defined(__Userspace__) + PKTHREAD p +#else +#if defined(__Userspace__) + /* proc is a dummy in __Userspace__ and will not be passed to sctp_lower_sosend */ +#endif + struct proc *p +#endif +#endif +); + +#endif +#endif + diff --git a/netwerk/sctp/src/netinet/sctp_pcb.c b/netwerk/sctp/src/netinet/sctp_pcb.c new file mode 100644 index 0000000000..233ba5dbf3 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_pcb.c @@ -0,0 +1,8131 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/proc.h> +#endif +#include <netinet/sctp_var.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctputil.h> +#include <netinet/sctp.h> +#include <netinet/sctp_header.h> +#include <netinet/sctp_asconf.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_timer.h> +#include <netinet/sctp_bsd_addr.h> +#if defined(INET) || defined(INET6) +#if !defined(_WIN32) +#include <netinet/udp.h> +#endif +#endif +#ifdef INET6 +#if defined(__Userspace__) +#include "user_ip6_var.h" +#else +#include <netinet6/ip6_var.h> +#endif +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/sched.h> +#include <sys/smp.h> +#include <sys/unistd.h> +#endif +#if defined(__Userspace__) +#include <user_socketvar.h> +#include <user_atomic.h> +#if !defined(_WIN32) +#include <netdb.h> +#endif +#endif + +#if !defined(__FreeBSD__) || defined(__Userspace__) +struct sctp_base_info system_base_info; + +#endif +/* FIX: we don't handle multiple link local scopes */ +/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */ +#ifdef INET6 +int +SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b) +{ +#ifdef SCTP_EMBEDDED_V6_SCOPE +#if defined(__APPLE__) && !defined(__Userspace__) + struct in6_addr tmp_a, tmp_b; + + tmp_a = a->sin6_addr; +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) + if (in6_embedscope(&tmp_a, a, NULL, NULL) != 0) { +#else + if (in6_embedscope(&tmp_a, a, NULL, NULL, NULL) != 0) { +#endif + return (0); + } + tmp_b = b->sin6_addr; +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) + if (in6_embedscope(&tmp_b, b, NULL, NULL) != 0) { +#else + if (in6_embedscope(&tmp_b, b, NULL, NULL, NULL) != 0) { +#endif + return (0); + } + return (IN6_ARE_ADDR_EQUAL(&tmp_a, &tmp_b)); +#elif defined(SCTP_KAME) + struct sockaddr_in6 tmp_a, tmp_b; + + memcpy(&tmp_a, a, sizeof(struct sockaddr_in6)); + if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) { + return (0); + } + memcpy(&tmp_b, b, sizeof(struct sockaddr_in6)); + if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) { + return (0); + } + return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr)); +#else + struct in6_addr tmp_a, tmp_b; + + tmp_a = a->sin6_addr; + if (in6_embedscope(&tmp_a, a) != 0) { + return (0); + } + tmp_b = b->sin6_addr; + if (in6_embedscope(&tmp_b, b) != 0) { + return (0); + } + return (IN6_ARE_ADDR_EQUAL(&tmp_a, &tmp_b)); +#endif +#else + return (IN6_ARE_ADDR_EQUAL(&(a->sin6_addr), &(b->sin6_addr))); +#endif /* SCTP_EMBEDDED_V6_SCOPE */ +} +#endif + +void +sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb) +{ + /* + * We really don't need to lock this, but I will just because it + * does not hurt. + */ + SCTP_INP_INFO_RLOCK(); + spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep); + spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc); + spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr); + spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr); + spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk); + spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq); + spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq); + spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks); + SCTP_INP_INFO_RUNLOCK(); +} + +/*- + * Addresses are added to VRF's (Virtual Router's). For BSD we + * have only the default VRF 0. We maintain a hash list of + * VRF's. Each VRF has its own list of sctp_ifn's. Each of + * these has a list of addresses. When we add a new address + * to a VRF we lookup the ifn/ifn_index, if the ifn does + * not exist we create it and add it to the list of IFN's + * within the VRF. Once we have the sctp_ifn, we add the + * address to the list. So we look something like: + * + * hash-vrf-table + * vrf-> ifn-> ifn -> ifn + * vrf | + * ... +--ifa-> ifa -> ifa + * vrf + * + * We keep these separate lists since the SCTP subsystem will + * point to these from its source address selection nets structure. + * When an address is deleted it does not happen right away on + * the SCTP side, it gets scheduled. What we do when a + * delete happens is immediately remove the address from + * the master list and decrement the refcount. As our + * addip iterator works through and frees the src address + * selection pointing to the sctp_ifa, eventually the refcount + * will reach 0 and we will delete it. Note that it is assumed + * that any locking on system level ifn/ifa is done at the + * caller of these functions and these routines will only + * lock the SCTP structures as they add or delete things. + * + * Other notes on VRF concepts. + * - An endpoint can be in multiple VRF's + * - An association lives within a VRF and only one VRF. + * - Any incoming packet we can deduce the VRF for by + * looking at the mbuf/pak inbound (for BSD its VRF=0 :D) + * - Any downward send call or connect call must supply the + * VRF via ancillary data or via some sort of set default + * VRF socket option call (again for BSD no brainer since + * the VRF is always 0). + * - An endpoint may add multiple VRF's to it. + * - Listening sockets can accept associations in any + * of the VRF's they are in but the assoc will end up + * in only one VRF (gotten from the packet or connect/send). + * + */ + +struct sctp_vrf * +sctp_allocate_vrf(int vrf_id) +{ + struct sctp_vrf *vrf = NULL; + struct sctp_vrflist *bucket; + + /* First allocate the VRF structure */ + vrf = sctp_find_vrf(vrf_id); + if (vrf) { + /* Already allocated */ + return (vrf); + } + SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf), + SCTP_M_VRF); + if (vrf == NULL) { + /* No memory */ +#ifdef INVARIANTS + panic("No memory for VRF:%d", vrf_id); +#endif + return (NULL); + } + /* setup the VRF */ + memset(vrf, 0, sizeof(struct sctp_vrf)); + vrf->vrf_id = vrf_id; + LIST_INIT(&vrf->ifnlist); + vrf->total_ifa_count = 0; + vrf->refcount = 0; + /* now also setup table ids */ + SCTP_INIT_VRF_TABLEID(vrf); + /* Init the HASH of addresses */ + vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE, + &vrf->vrf_addr_hashmark); + if (vrf->vrf_addr_hash == NULL) { + /* No memory */ +#ifdef INVARIANTS + panic("No memory for VRF:%d", vrf_id); +#endif + SCTP_FREE(vrf, SCTP_M_VRF); + return (NULL); + } + + /* Add it to the hash table */ + bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))]; + LIST_INSERT_HEAD(bucket, vrf, next_vrf); + atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1); + return (vrf); +} + +struct sctp_ifn * +sctp_find_ifn(void *ifn, uint32_t ifn_index) +{ + struct sctp_ifn *sctp_ifnp; + struct sctp_ifnlist *hash_ifn_head; + + /* We assume the lock is held for the addresses + * if that's wrong problems could occur :-) + */ + SCTP_IPI_ADDR_LOCK_ASSERT(); + hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))]; + LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) { + if (sctp_ifnp->ifn_index == ifn_index) { + return (sctp_ifnp); + } + if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) { + return (sctp_ifnp); + } + } + return (NULL); +} + +struct sctp_vrf * +sctp_find_vrf(uint32_t vrf_id) +{ + struct sctp_vrflist *bucket; + struct sctp_vrf *liste; + + bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))]; + LIST_FOREACH(liste, bucket, next_vrf) { + if (vrf_id == liste->vrf_id) { + return (liste); + } + } + return (NULL); +} + +void +sctp_free_vrf(struct sctp_vrf *vrf) +{ + if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) { + if (vrf->vrf_addr_hash) { + SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark); + vrf->vrf_addr_hash = NULL; + } + /* We zero'd the count */ + LIST_REMOVE(vrf, next_vrf); + SCTP_FREE(vrf, SCTP_M_VRF); + atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1); + } +} + +void +sctp_free_ifn(struct sctp_ifn *sctp_ifnp) +{ + if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) { + /* We zero'd the count */ + if (sctp_ifnp->vrf) { + sctp_free_vrf(sctp_ifnp->vrf); + } + SCTP_FREE(sctp_ifnp, SCTP_M_IFN); + atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1); + } +} + +void +sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu) +{ + struct sctp_ifn *sctp_ifnp; + + sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index); + if (sctp_ifnp != NULL) { + sctp_ifnp->ifn_mtu = mtu; + } +} + +void +sctp_free_ifa(struct sctp_ifa *sctp_ifap) +{ + if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) { + /* We zero'd the count */ + if (sctp_ifap->ifn_p) { + sctp_free_ifn(sctp_ifap->ifn_p); + } + SCTP_FREE(sctp_ifap, SCTP_M_IFA); + atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1); + } +} + +static void +sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock) +{ + struct sctp_ifn *found; + + found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index); + if (found == NULL) { + /* Not in the list.. sorry */ + return; + } + if (hold_addr_lock == 0) { + SCTP_IPI_ADDR_WLOCK(); + } else { + SCTP_IPI_ADDR_WLOCK_ASSERT(); + } + LIST_REMOVE(sctp_ifnp, next_bucket); + LIST_REMOVE(sctp_ifnp, next_ifn); + if (hold_addr_lock == 0) { + SCTP_IPI_ADDR_WUNLOCK(); + } + /* Take away the reference, and possibly free it */ + sctp_free_ifn(sctp_ifnp); +} + +void +sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr, + const char *if_name, uint32_t ifn_index) +{ + struct sctp_vrf *vrf; + struct sctp_ifa *sctp_ifap; + + SCTP_IPI_ADDR_RLOCK(); + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) { + SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); + goto out; + } + sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); + if (sctp_ifap == NULL) { + SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n"); + goto out; + } + if (sctp_ifap->ifn_p == NULL) { + SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unusable\n"); + goto out; + } + if (if_name) { + if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) { + SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n", + sctp_ifap->ifn_p->ifn_name, if_name); + goto out; + } + } else { + if (sctp_ifap->ifn_p->ifn_index != ifn_index) { + SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n", + sctp_ifap->ifn_p->ifn_index, ifn_index); + goto out; + } + } + + sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID); + sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE; + out: + SCTP_IPI_ADDR_RUNLOCK(); +} + +void +sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr, + const char *if_name, uint32_t ifn_index) +{ + struct sctp_vrf *vrf; + struct sctp_ifa *sctp_ifap; + + SCTP_IPI_ADDR_RLOCK(); + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) { + SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); + goto out; + } + sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); + if (sctp_ifap == NULL) { + SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n"); + goto out; + } + if (sctp_ifap->ifn_p == NULL) { + SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unusable\n"); + goto out; + } + if (if_name) { + if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) { + SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n", + sctp_ifap->ifn_p->ifn_name, if_name); + goto out; + } + } else { + if (sctp_ifap->ifn_p->ifn_index != ifn_index) { + SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n", + sctp_ifap->ifn_p->ifn_index, ifn_index); + goto out; + } + } + + sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE); + sctp_ifap->localifa_flags |= SCTP_ADDR_VALID; + out: + SCTP_IPI_ADDR_RUNLOCK(); +} + +/*- + * Add an ifa to an ifn. + * Register the interface as necessary. + * NOTE: ADDR write lock MUST be held. + */ +static void +sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap) +{ + int ifa_af; + + LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa); + sctp_ifap->ifn_p = sctp_ifnp; + atomic_add_int(&sctp_ifap->ifn_p->refcount, 1); + /* update address counts */ + sctp_ifnp->ifa_count++; + ifa_af = sctp_ifap->address.sa.sa_family; + switch (ifa_af) { +#ifdef INET + case AF_INET: + sctp_ifnp->num_v4++; + break; +#endif +#ifdef INET6 + case AF_INET6: + sctp_ifnp->num_v6++; + break; +#endif + default: + break; + } + if (sctp_ifnp->ifa_count == 1) { + /* register the new interface */ + sctp_ifnp->registered_af = ifa_af; + } +} + +/*- + * Remove an ifa from its ifn. + * If no more addresses exist, remove the ifn too. Otherwise, re-register + * the interface based on the remaining address families left. + * NOTE: ADDR write lock MUST be held. + */ +static void +sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap) +{ + LIST_REMOVE(sctp_ifap, next_ifa); + if (sctp_ifap->ifn_p) { + /* update address counts */ + sctp_ifap->ifn_p->ifa_count--; + switch (sctp_ifap->address.sa.sa_family) { +#ifdef INET + case AF_INET: + sctp_ifap->ifn_p->num_v4--; + break; +#endif +#ifdef INET6 + case AF_INET6: + sctp_ifap->ifn_p->num_v6--; + break; +#endif + default: + break; + } + + if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) { + /* remove the ifn, possibly freeing it */ + sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED); + } else { + /* re-register address family type, if needed */ + if ((sctp_ifap->ifn_p->num_v6 == 0) && + (sctp_ifap->ifn_p->registered_af == AF_INET6)) { + sctp_ifap->ifn_p->registered_af = AF_INET; + } else if ((sctp_ifap->ifn_p->num_v4 == 0) && + (sctp_ifap->ifn_p->registered_af == AF_INET)) { + sctp_ifap->ifn_p->registered_af = AF_INET6; + } + /* free the ifn refcount */ + sctp_free_ifn(sctp_ifap->ifn_p); + } + sctp_ifap->ifn_p = NULL; + } +} + +struct sctp_ifa * +sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index, + uint32_t ifn_type, const char *if_name, void *ifa, + struct sockaddr *addr, uint32_t ifa_flags, + int dynamic_add) +{ + struct sctp_vrf *vrf; + struct sctp_ifn *sctp_ifnp, *new_sctp_ifnp; + struct sctp_ifa *sctp_ifap, *new_sctp_ifap; + struct sctp_ifalist *hash_addr_head; + struct sctp_ifnlist *hash_ifn_head; + uint32_t hash_of_addr; + int new_ifn_af = 0; + +#ifdef SCTP_DEBUG + SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id); + SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr); +#endif + SCTP_MALLOC(new_sctp_ifnp, struct sctp_ifn *, + sizeof(struct sctp_ifn), SCTP_M_IFN); + if (new_sctp_ifnp == NULL) { +#ifdef INVARIANTS + panic("No memory for IFN"); +#endif + return (NULL); + } + SCTP_MALLOC(new_sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA); + if (new_sctp_ifap == NULL) { +#ifdef INVARIANTS + panic("No memory for IFA"); +#endif + SCTP_FREE(new_sctp_ifnp, SCTP_M_IFN); + return (NULL); + } + + SCTP_IPI_ADDR_WLOCK(); + sctp_ifnp = sctp_find_ifn(ifn, ifn_index); + if (sctp_ifnp) { + vrf = sctp_ifnp->vrf; + } else { + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) { + vrf = sctp_allocate_vrf(vrf_id); + if (vrf == NULL) { + SCTP_IPI_ADDR_WUNLOCK(); + SCTP_FREE(new_sctp_ifnp, SCTP_M_IFN); + SCTP_FREE(new_sctp_ifap, SCTP_M_IFA); + return (NULL); + } + } + } + if (sctp_ifnp == NULL) { + /* build one and add it, can't hold lock + * until after malloc done though. + */ + sctp_ifnp = new_sctp_ifnp; + new_sctp_ifnp = NULL; + memset(sctp_ifnp, 0, sizeof(struct sctp_ifn)); + sctp_ifnp->ifn_index = ifn_index; + sctp_ifnp->ifn_p = ifn; + sctp_ifnp->ifn_type = ifn_type; + sctp_ifnp->refcount = 0; + sctp_ifnp->vrf = vrf; + atomic_add_int(&vrf->refcount, 1); + sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index); + if (if_name != NULL) { + SCTP_SNPRINTF(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name); + } else { + SCTP_SNPRINTF(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown"); + } + hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))]; + LIST_INIT(&sctp_ifnp->ifalist); + LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket); + LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn); + atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1); + new_ifn_af = 1; + } + sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); + if (sctp_ifap) { + /* Hmm, it already exists? */ + if ((sctp_ifap->ifn_p) && + (sctp_ifap->ifn_p->ifn_index == ifn_index)) { + SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n", + sctp_ifap->ifn_p->ifn_name, ifn_index, + (void *)sctp_ifap); + if (new_ifn_af) { + /* Remove the created one that we don't want */ + sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED); + } + if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) { + /* easy to solve, just switch back to active */ + SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n"); + sctp_ifap->localifa_flags = SCTP_ADDR_VALID; + sctp_ifap->ifn_p = sctp_ifnp; + atomic_add_int(&sctp_ifap->ifn_p->refcount, 1); + } + exit_stage_left: + SCTP_IPI_ADDR_WUNLOCK(); + if (new_sctp_ifnp != NULL) { + SCTP_FREE(new_sctp_ifnp, SCTP_M_IFN); + } + SCTP_FREE(new_sctp_ifap, SCTP_M_IFA); + return (sctp_ifap); + } else { + if (sctp_ifap->ifn_p) { + /* + * The last IFN gets the address, remove the + * old one + */ + SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n", + (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name, + sctp_ifap->ifn_p->ifn_index, if_name, + ifn_index); + /* remove the address from the old ifn */ + sctp_remove_ifa_from_ifn(sctp_ifap); + /* move the address over to the new ifn */ + sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap); + goto exit_stage_left; + } else { + /* repair ifnp which was NULL ? */ + sctp_ifap->localifa_flags = SCTP_ADDR_VALID; + SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n", + (void *)sctp_ifnp, (void *)sctp_ifap); + sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap); + } + goto exit_stage_left; + } + } + sctp_ifap = new_sctp_ifap; + memset(sctp_ifap, 0, sizeof(struct sctp_ifa)); + sctp_ifap->ifn_p = sctp_ifnp; + atomic_add_int(&sctp_ifnp->refcount, 1); + sctp_ifap->vrf_id = vrf_id; + sctp_ifap->ifa = ifa; +#ifdef HAVE_SA_LEN + memcpy(&sctp_ifap->address, addr, addr->sa_len); +#else + switch (addr->sa_family) { +#ifdef INET + case AF_INET: + memcpy(&sctp_ifap->address, addr, sizeof(struct sockaddr_in)); + break; +#endif +#ifdef INET6 + case AF_INET6: + memcpy(&sctp_ifap->address, addr, sizeof(struct sockaddr_in6)); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + memcpy(&sctp_ifap->address, addr, sizeof(struct sockaddr_conn)); + break; +#endif + default: + /* TSNH */ + break; + } +#endif + sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE; + sctp_ifap->flags = ifa_flags; + /* Set scope */ + switch (sctp_ifap->address.sa.sa_family) { +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin; + + sin = &sctp_ifap->address.sin; + if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) || + (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) { + sctp_ifap->src_is_loop = 1; + } + if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { + sctp_ifap->src_is_priv = 1; + } + sctp_ifnp->num_v4++; + if (new_ifn_af) + new_ifn_af = AF_INET; + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + /* ok to use deprecated addresses? */ + struct sockaddr_in6 *sin6; + + sin6 = &sctp_ifap->address.sin6; + if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) || + (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) { + sctp_ifap->src_is_loop = 1; + } + if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { + sctp_ifap->src_is_priv = 1; + } + sctp_ifnp->num_v6++; + if (new_ifn_af) + new_ifn_af = AF_INET6; + break; + } +#endif +#if defined(__Userspace__) + case AF_CONN: + if (new_ifn_af) + new_ifn_af = AF_CONN; + break; +#endif + default: + new_ifn_af = 0; + break; + } + hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa); + + if ((sctp_ifap->src_is_priv == 0) && + (sctp_ifap->src_is_loop == 0)) { + sctp_ifap->src_is_glob = 1; + } + hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)]; + LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket); + sctp_ifap->refcount = 1; + LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa); + sctp_ifnp->ifa_count++; + vrf->total_ifa_count++; + atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1); + if (new_ifn_af) { + sctp_ifnp->registered_af = new_ifn_af; + } + SCTP_IPI_ADDR_WUNLOCK(); + if (new_sctp_ifnp != NULL) { + SCTP_FREE(new_sctp_ifnp, SCTP_M_IFN); + } + + if (dynamic_add) { + /* Bump up the refcount so that when the timer + * completes it will drop back down. + */ + struct sctp_laddr *wi; + + atomic_add_int(&sctp_ifap->refcount, 1); + wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); + if (wi == NULL) { + /* + * Gak, what can we do? We have lost an address + * change can you say HOSED? + */ + SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n"); + /* Opps, must decrement the count */ + sctp_del_addr_from_vrf(vrf_id, addr, ifn_index, + if_name); + return (NULL); + } + SCTP_INCR_LADDR_COUNT(); + memset(wi, 0, sizeof(*wi)); + (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); + wi->ifa = sctp_ifap; + wi->action = SCTP_ADD_IP_ADDRESS; + + SCTP_WQ_ADDR_LOCK(); + LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); + sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, + (struct sctp_inpcb *)NULL, + (struct sctp_tcb *)NULL, + (struct sctp_nets *)NULL); + SCTP_WQ_ADDR_UNLOCK(); + } else { + /* it's ready for use */ + sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE; + } + return (sctp_ifap); +} + +void +sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr, + uint32_t ifn_index, const char *if_name) +{ + struct sctp_vrf *vrf; + struct sctp_ifa *sctp_ifap = NULL; + + SCTP_IPI_ADDR_WLOCK(); + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) { + SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); + goto out_now; + } + +#ifdef SCTP_DEBUG + SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id); + SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr); +#endif + sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); + if (sctp_ifap) { + /* Validate the delete */ + if (sctp_ifap->ifn_p) { + int valid = 0; + /*- + * The name has priority over the ifn_index + * if its given. + */ + if (if_name) { + if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) { + /* They match its a correct delete */ + valid = 1; + } + } + if (!valid) { + /* last ditch check ifn_index */ + if (ifn_index == sctp_ifap->ifn_p->ifn_index) { + valid = 1; + } + } + if (!valid) { + SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n", + ifn_index, ((if_name == NULL) ? "NULL" : if_name)); + SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n", + sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name); + SCTP_IPI_ADDR_WUNLOCK(); + return; + } + } + SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap); + sctp_ifap->localifa_flags &= SCTP_ADDR_VALID; + /* + * We don't set the flag. This means that the structure will + * hang around in EP's that have bound specific to it until + * they close. This gives us TCP like behavior if someone + * removes an address (or for that matter adds it right back). + */ + /* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */ + vrf->total_ifa_count--; + LIST_REMOVE(sctp_ifap, next_bucket); + sctp_remove_ifa_from_ifn(sctp_ifap); + } +#ifdef SCTP_DEBUG + else { + SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:", + ifn_index); + SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr); + } +#endif + + out_now: + SCTP_IPI_ADDR_WUNLOCK(); + if (sctp_ifap) { + struct sctp_laddr *wi; + + wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); + if (wi == NULL) { + /* + * Gak, what can we do? We have lost an address + * change can you say HOSED? + */ + SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n"); + + /* Oops, must decrement the count */ + sctp_free_ifa(sctp_ifap); + return; + } + SCTP_INCR_LADDR_COUNT(); + memset(wi, 0, sizeof(*wi)); + (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); + wi->ifa = sctp_ifap; + wi->action = SCTP_DEL_IP_ADDRESS; + SCTP_WQ_ADDR_LOCK(); + /* + * Should this really be a tailq? As it is we will process the + * newest first :-0 + */ + LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); + sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, + (struct sctp_inpcb *)NULL, + (struct sctp_tcb *)NULL, + (struct sctp_nets *)NULL); + SCTP_WQ_ADDR_UNLOCK(); + } + return; +} + +static int +sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to) +{ + int loopback_scope; +#if defined(INET) + int ipv4_local_scope, ipv4_addr_legal; +#endif +#if defined(INET6) + int local_scope, site_scope, ipv6_addr_legal; +#endif +#if defined(__Userspace__) + int conn_addr_legal; +#endif + struct sctp_vrf *vrf; + struct sctp_ifn *sctp_ifn; + struct sctp_ifa *sctp_ifa; + + loopback_scope = stcb->asoc.scope.loopback_scope; +#if defined(INET) + ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope; + ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal; +#endif +#if defined(INET6) + local_scope = stcb->asoc.scope.local_scope; + site_scope = stcb->asoc.scope.site_scope; + ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal; +#endif +#if defined(__Userspace__) + conn_addr_legal = stcb->asoc.scope.conn_addr_legal; +#endif + + SCTP_IPI_ADDR_RLOCK(); + vrf = sctp_find_vrf(stcb->asoc.vrf_id); + if (vrf == NULL) { + /* no vrf, no addresses */ + SCTP_IPI_ADDR_RUNLOCK(); + return (0); + } + + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { + if ((loopback_scope == 0) && + SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { + continue; + } + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { + if (sctp_is_addr_restricted(stcb, sctp_ifa) && + (!sctp_is_addr_pending(stcb, sctp_ifa))) { + /* We allow pending addresses, where we + * have sent an asconf-add to be considered + * valid. + */ + continue; + } + if (sctp_ifa->address.sa.sa_family != to->sa_family) { + continue; + } + switch (sctp_ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: + if (ipv4_addr_legal) { + struct sockaddr_in *sin, *rsin; + + sin = &sctp_ifa->address.sin; + rsin = (struct sockaddr_in *)to; + if ((ipv4_local_scope == 0) && + IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred, + &sin->sin_addr) != 0) { + continue; + } +#endif + if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { + SCTP_IPI_ADDR_RUNLOCK(); + return (1); + } + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (ipv6_addr_legal) { + struct sockaddr_in6 *sin6, *rsin6; +#if defined(SCTP_EMBEDDED_V6_SCOPE) && !defined(SCTP_KAME) + struct sockaddr_in6 lsa6; +#endif + sin6 = &sctp_ifa->address.sin6; + rsin6 = (struct sockaddr_in6 *)to; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred, + &sin6->sin6_addr) != 0) { + continue; + } +#endif + if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { + if (local_scope == 0) + continue; +#if defined(SCTP_EMBEDDED_V6_SCOPE) + if (sin6->sin6_scope_id == 0) { +#ifdef SCTP_KAME + if (sa6_recoverscope(sin6) != 0) + continue; +#else + lsa6 = *sin6; + if (in6_recoverscope(&lsa6, + &lsa6.sin6_addr, + NULL)) + continue; + sin6 = &lsa6; +#endif /* SCTP_KAME */ + } +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + } + if ((site_scope == 0) && + (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { + continue; + } + if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { + SCTP_IPI_ADDR_RUNLOCK(); + return (1); + } + } + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + if (conn_addr_legal) { + struct sockaddr_conn *sconn, *rsconn; + + sconn = &sctp_ifa->address.sconn; + rsconn = (struct sockaddr_conn *)to; + if (sconn->sconn_addr == rsconn->sconn_addr) { + SCTP_IPI_ADDR_RUNLOCK(); + return (1); + } + } + break; +#endif + default: + /* TSNH */ + break; + } + } + } + } else { + struct sctp_laddr *laddr; + + LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { + SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n"); + continue; + } + if (sctp_is_addr_restricted(stcb, laddr->ifa) && + (!sctp_is_addr_pending(stcb, laddr->ifa))) { + /* We allow pending addresses, where we + * have sent an asconf-add to be considered + * valid. + */ + continue; + } + if (laddr->ifa->address.sa.sa_family != to->sa_family) { + continue; + } + switch (to->sa_family) { +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin, *rsin; + + sin = &laddr->ifa->address.sin; + rsin = (struct sockaddr_in *)to; + if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { + SCTP_IPI_ADDR_RUNLOCK(); + return (1); + } + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6, *rsin6; + + sin6 = &laddr->ifa->address.sin6; + rsin6 = (struct sockaddr_in6 *)to; + if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { + SCTP_IPI_ADDR_RUNLOCK(); + return (1); + } + break; + } + +#endif +#if defined(__Userspace__) + case AF_CONN: + { + struct sockaddr_conn *sconn, *rsconn; + + sconn = &laddr->ifa->address.sconn; + rsconn = (struct sockaddr_conn *)to; + if (sconn->sconn_addr == rsconn->sconn_addr) { + SCTP_IPI_ADDR_RUNLOCK(); + return (1); + } + break; + } +#endif + default: + /* TSNH */ + break; + } + } + } + SCTP_IPI_ADDR_RUNLOCK(); + return (0); +} + +static struct sctp_tcb * +sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from, + struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id) +{ + /**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */ + /* + * If we support the TCP model, then we must now dig through to see + * if we can find our endpoint in the list of tcp ep's. + */ + uint16_t lport, rport; + struct sctppcbhead *ephead; + struct sctp_inpcb *inp; + struct sctp_laddr *laddr; + struct sctp_tcb *stcb; + struct sctp_nets *net; +#ifdef SCTP_MVRF + int fnd, i; +#endif + + if ((to == NULL) || (from == NULL)) { + return (NULL); + } + + switch (to->sa_family) { +#ifdef INET + case AF_INET: + if (from->sa_family == AF_INET) { + lport = ((struct sockaddr_in *)to)->sin_port; + rport = ((struct sockaddr_in *)from)->sin_port; + } else { + return (NULL); + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (from->sa_family == AF_INET6) { + lport = ((struct sockaddr_in6 *)to)->sin6_port; + rport = ((struct sockaddr_in6 *)from)->sin6_port; + } else { + return (NULL); + } + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + if (from->sa_family == AF_CONN) { + lport = ((struct sockaddr_conn *)to)->sconn_port; + rport = ((struct sockaddr_conn *)from)->sconn_port; + } else { + return (NULL); + } + break; +#endif + default: + return (NULL); + } + ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))]; + /* + * Ok now for each of the guys in this bucket we must look and see: + * - Does the remote port match. - Does there single association's + * addresses match this address (to). If so we update p_ep to point + * to this ep and return the tcb from it. + */ + LIST_FOREACH(inp, ephead, sctp_hash) { + SCTP_INP_RLOCK(inp); + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { + SCTP_INP_RUNLOCK(inp); + continue; + } + if (lport != inp->sctp_lport) { + SCTP_INP_RUNLOCK(inp); + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + switch (to->sa_family) { +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin; + + sin = (struct sockaddr_in *)to; + if (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sin->sin_addr) != 0) { + SCTP_INP_RUNLOCK(inp); + continue; + } + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)to; + if (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sin6->sin6_addr) != 0) { + SCTP_INP_RUNLOCK(inp); + continue; + } + break; + } +#endif + default: + SCTP_INP_RUNLOCK(inp); + continue; + } +#endif +#ifdef SCTP_MVRF + fnd = 0; + for (i = 0; i < inp->num_vrfs; i++) { + if (inp->m_vrf_ids[i] == vrf_id) { + fnd = 1; + break; + } + } + if (fnd == 0) { + SCTP_INP_RUNLOCK(inp); + continue; + } +#else + if (inp->def_vrf_id != vrf_id) { + SCTP_INP_RUNLOCK(inp); + continue; + } +#endif + /* check to see if the ep has one of the addresses */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { + /* We are NOT bound all, so look further */ + int match = 0; + + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa == NULL) { + SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __func__); + continue; + } + if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { + SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n"); + continue; + } + if (laddr->ifa->address.sa.sa_family == + to->sa_family) { + /* see if it matches */ +#ifdef INET + if (from->sa_family == AF_INET) { + struct sockaddr_in *intf_addr, *sin; + + intf_addr = &laddr->ifa->address.sin; + sin = (struct sockaddr_in *)to; + if (sin->sin_addr.s_addr == + intf_addr->sin_addr.s_addr) { + match = 1; + break; + } + } +#endif +#ifdef INET6 + if (from->sa_family == AF_INET6) { + struct sockaddr_in6 *intf_addr6; + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *) + to; + intf_addr6 = &laddr->ifa->address.sin6; + + if (SCTP6_ARE_ADDR_EQUAL(sin6, + intf_addr6)) { + match = 1; + break; + } + } +#endif +#if defined(__Userspace__) + if (from->sa_family == AF_CONN) { + struct sockaddr_conn *intf_addr, *sconn; + + intf_addr = &laddr->ifa->address.sconn; + sconn = (struct sockaddr_conn *)to; + if (sconn->sconn_addr == + intf_addr->sconn_addr) { + match = 1; + break; + } + } +#endif + } + } + if (match == 0) { + /* This endpoint does not have this address */ + SCTP_INP_RUNLOCK(inp); + continue; + } + } + /* + * Ok if we hit here the ep has the address, does it hold + * the tcb? + */ + /* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */ + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb == NULL) { + SCTP_INP_RUNLOCK(inp); + continue; + } + SCTP_TCB_LOCK(stcb); + if (!sctp_does_stcb_own_this_addr(stcb, to)) { + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_RUNLOCK(inp); + continue; + } + if (stcb->rport != rport) { + /* remote port does not match. */ + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_RUNLOCK(inp); + continue; + } + if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_RUNLOCK(inp); + continue; + } + if (!sctp_does_stcb_own_this_addr(stcb, to)) { + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_RUNLOCK(inp); + continue; + } + /* Does this TCB have a matching address? */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if (net->ro._l_addr.sa.sa_family != from->sa_family) { + /* not the same family, can't be a match */ + continue; + } + switch (from->sa_family) { +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin, *rsin; + + sin = (struct sockaddr_in *)&net->ro._l_addr; + rsin = (struct sockaddr_in *)from; + if (sin->sin_addr.s_addr == + rsin->sin_addr.s_addr) { + /* found it */ + if (netp != NULL) { + *netp = net; + } + /* Update the endpoint pointer */ + *inp_p = inp; + SCTP_INP_RUNLOCK(inp); + return (stcb); + } + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6, *rsin6; + + sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; + rsin6 = (struct sockaddr_in6 *)from; + if (SCTP6_ARE_ADDR_EQUAL(sin6, + rsin6)) { + /* found it */ + if (netp != NULL) { + *netp = net; + } + /* Update the endpoint pointer */ + *inp_p = inp; + SCTP_INP_RUNLOCK(inp); + return (stcb); + } + break; + } +#endif +#if defined(__Userspace__) + case AF_CONN: + { + struct sockaddr_conn *sconn, *rsconn; + + sconn = (struct sockaddr_conn *)&net->ro._l_addr; + rsconn = (struct sockaddr_conn *)from; + if (sconn->sconn_addr == rsconn->sconn_addr) { + /* found it */ + if (netp != NULL) { + *netp = net; + } + /* Update the endpoint pointer */ + *inp_p = inp; + SCTP_INP_RUNLOCK(inp); + return (stcb); + } + break; + } +#endif + default: + /* TSNH */ + break; + } + } + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_RUNLOCK(inp); + } + return (NULL); +} + +/* + * rules for use + * + * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an + * stcb, both will be locked (locked_tcb and stcb) but decrement will be done + * (if locked == NULL). 3) Decrement happens on return ONLY if locked == + * NULL. + */ + +struct sctp_tcb * +sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote, + struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb) +{ + struct sctpasochead *head; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb = NULL; + struct sctp_nets *net; + uint16_t rport; + + inp = *inp_p; + switch (remote->sa_family) { +#ifdef INET + case AF_INET: + rport = (((struct sockaddr_in *)remote)->sin_port); + break; +#endif +#ifdef INET6 + case AF_INET6: + rport = (((struct sockaddr_in6 *)remote)->sin6_port); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + rport = (((struct sockaddr_conn *)remote)->sconn_port); + break; +#endif + default: + return (NULL); + } + if (locked_tcb) { + /* + * UN-lock so we can do proper locking here this occurs when + * called from load_addresses_from_init. + */ + atomic_add_int(&locked_tcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(locked_tcb); + } + SCTP_INP_INFO_RLOCK(); + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { + /*- + * Now either this guy is our listener or it's the + * connector. If it is the one that issued the connect, then + * it's only chance is to be the first TCB in the list. If + * it is the acceptor, then do the special_lookup to hash + * and find the real inp. + */ + if ((inp->sctp_socket) && SCTP_IS_LISTENING(inp)) { + /* to is peer addr, from is my addr */ +#ifndef SCTP_MVRF + stcb = sctp_tcb_special_locate(inp_p, remote, local, + netp, inp->def_vrf_id); + if ((stcb != NULL) && (locked_tcb == NULL)) { + /* we have a locked tcb, lower refcount */ + SCTP_INP_DECR_REF(inp); + } + if ((locked_tcb != NULL) && (locked_tcb != stcb)) { + SCTP_INP_RLOCK(locked_tcb->sctp_ep); + SCTP_TCB_LOCK(locked_tcb); + atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); + SCTP_INP_RUNLOCK(locked_tcb->sctp_ep); + } +#else + /*- + * MVRF is tricky, we must look in every VRF + * the endpoint has. + */ + int i; + + for (i = 0; i < inp->num_vrfs; i++) { + stcb = sctp_tcb_special_locate(inp_p, remote, local, + netp, inp->m_vrf_ids[i]); + if ((stcb != NULL) && (locked_tcb == NULL)) { + /* we have a locked tcb, lower refcount */ + SCTP_INP_DECR_REF(inp); + break; + } + if ((locked_tcb != NULL) && (locked_tcb != stcb)) { + SCTP_INP_RLOCK(locked_tcb->sctp_ep); + SCTP_TCB_LOCK(locked_tcb); + atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); + SCTP_INP_RUNLOCK(locked_tcb->sctp_ep); + break; + } + } +#endif + SCTP_INP_INFO_RUNLOCK(); + return (stcb); + } else { + SCTP_INP_WLOCK(inp); + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { + goto null_return; + } + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb == NULL) { + goto null_return; + } + SCTP_TCB_LOCK(stcb); + + if (stcb->rport != rport) { + /* remote port does not match. */ + SCTP_TCB_UNLOCK(stcb); + goto null_return; + } + if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + SCTP_TCB_UNLOCK(stcb); + goto null_return; + } + if (local && !sctp_does_stcb_own_this_addr(stcb, local)) { + SCTP_TCB_UNLOCK(stcb); + goto null_return; + } + /* now look at the list of remote addresses */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { +#ifdef INVARIANTS + if (net == (TAILQ_NEXT(net, sctp_next))) { + panic("Corrupt net list"); + } +#endif + if (net->ro._l_addr.sa.sa_family != + remote->sa_family) { + /* not the same family */ + continue; + } + switch (remote->sa_family) { +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin, *rsin; + + sin = (struct sockaddr_in *) + &net->ro._l_addr; + rsin = (struct sockaddr_in *)remote; + if (sin->sin_addr.s_addr == + rsin->sin_addr.s_addr) { + /* found it */ + if (netp != NULL) { + *netp = net; + } + if (locked_tcb == NULL) { + SCTP_INP_DECR_REF(inp); + } else if (locked_tcb != stcb) { + SCTP_TCB_LOCK(locked_tcb); + } + if (locked_tcb) { + atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); + } + + SCTP_INP_WUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + return (stcb); + } + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6, *rsin6; + + sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; + rsin6 = (struct sockaddr_in6 *)remote; + if (SCTP6_ARE_ADDR_EQUAL(sin6, + rsin6)) { + /* found it */ + if (netp != NULL) { + *netp = net; + } + if (locked_tcb == NULL) { + SCTP_INP_DECR_REF(inp); + } else if (locked_tcb != stcb) { + SCTP_TCB_LOCK(locked_tcb); + } + if (locked_tcb) { + atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); + } + SCTP_INP_WUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + return (stcb); + } + break; + } +#endif +#if defined(__Userspace__) + case AF_CONN: + { + struct sockaddr_conn *sconn, *rsconn; + + sconn = (struct sockaddr_conn *)&net->ro._l_addr; + rsconn = (struct sockaddr_conn *)remote; + if (sconn->sconn_addr == rsconn->sconn_addr) { + /* found it */ + if (netp != NULL) { + *netp = net; + } + if (locked_tcb == NULL) { + SCTP_INP_DECR_REF(inp); + } else if (locked_tcb != stcb) { + SCTP_TCB_LOCK(locked_tcb); + } + if (locked_tcb) { + atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); + } + SCTP_INP_WUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + return (stcb); + } + break; + } +#endif + default: + /* TSNH */ + break; + } + } + SCTP_TCB_UNLOCK(stcb); + } + } else { + SCTP_INP_WLOCK(inp); + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { + goto null_return; + } + head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport, + inp->sctp_hashmark)]; + LIST_FOREACH(stcb, head, sctp_tcbhash) { + if (stcb->rport != rport) { + /* remote port does not match */ + continue; + } + SCTP_TCB_LOCK(stcb); + if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + SCTP_TCB_UNLOCK(stcb); + continue; + } + if (local && !sctp_does_stcb_own_this_addr(stcb, local)) { + SCTP_TCB_UNLOCK(stcb); + continue; + } + /* now look at the list of remote addresses */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { +#ifdef INVARIANTS + if (net == (TAILQ_NEXT(net, sctp_next))) { + panic("Corrupt net list"); + } +#endif + if (net->ro._l_addr.sa.sa_family != + remote->sa_family) { + /* not the same family */ + continue; + } + switch (remote->sa_family) { +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin, *rsin; + + sin = (struct sockaddr_in *) + &net->ro._l_addr; + rsin = (struct sockaddr_in *)remote; + if (sin->sin_addr.s_addr == + rsin->sin_addr.s_addr) { + /* found it */ + if (netp != NULL) { + *netp = net; + } + if (locked_tcb == NULL) { + SCTP_INP_DECR_REF(inp); + } else if (locked_tcb != stcb) { + SCTP_TCB_LOCK(locked_tcb); + } + if (locked_tcb) { + atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); + } + SCTP_INP_WUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + return (stcb); + } + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6, *rsin6; + + sin6 = (struct sockaddr_in6 *) + &net->ro._l_addr; + rsin6 = (struct sockaddr_in6 *)remote; + if (SCTP6_ARE_ADDR_EQUAL(sin6, + rsin6)) { + /* found it */ + if (netp != NULL) { + *netp = net; + } + if (locked_tcb == NULL) { + SCTP_INP_DECR_REF(inp); + } else if (locked_tcb != stcb) { + SCTP_TCB_LOCK(locked_tcb); + } + if (locked_tcb) { + atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); + } + SCTP_INP_WUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + return (stcb); + } + break; + } +#endif +#if defined(__Userspace__) + case AF_CONN: + { + struct sockaddr_conn *sconn, *rsconn; + + sconn = (struct sockaddr_conn *)&net->ro._l_addr; + rsconn = (struct sockaddr_conn *)remote; + if (sconn->sconn_addr == rsconn->sconn_addr) { + /* found it */ + if (netp != NULL) { + *netp = net; + } + if (locked_tcb == NULL) { + SCTP_INP_DECR_REF(inp); + } else if (locked_tcb != stcb) { + SCTP_TCB_LOCK(locked_tcb); + } + if (locked_tcb) { + atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); + } + SCTP_INP_WUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + return (stcb); + } + break; + } +#endif + default: + /* TSNH */ + break; + } + } + SCTP_TCB_UNLOCK(stcb); + } + } +null_return: + /* clean up for returning null */ + if (locked_tcb) { + SCTP_TCB_LOCK(locked_tcb); + atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); + } + SCTP_INP_WUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + /* not found */ + return (NULL); +} + +/* + * Find an association for a specific endpoint using the association id given + * out in the COMM_UP notification + */ +struct sctp_tcb * +sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock) +{ + /* + * Use my the assoc_id to find a endpoint + */ + struct sctpasochead *head; + struct sctp_tcb *stcb; + uint32_t id; + + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { + SCTP_PRINTF("TSNH ep_associd0\n"); + return (NULL); + } + id = (uint32_t)asoc_id; + head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)]; + if (head == NULL) { + /* invalid id TSNH */ + SCTP_PRINTF("TSNH ep_associd1\n"); + return (NULL); + } + LIST_FOREACH(stcb, head, sctp_tcbasocidhash) { + if (stcb->asoc.assoc_id == id) { + if (inp != stcb->sctp_ep) { + /* + * some other guy has the same id active (id + * collision ??). + */ + SCTP_PRINTF("TSNH ep_associd2\n"); + continue; + } + if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + continue; + } + if (want_lock) { + SCTP_TCB_LOCK(stcb); + } + return (stcb); + } + } + return (NULL); +} + +struct sctp_tcb * +sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock) +{ + struct sctp_tcb *stcb; + + SCTP_INP_RLOCK(inp); + stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock); + SCTP_INP_RUNLOCK(inp); + return (stcb); +} + +/* + * Endpoint probe expects that the INP_INFO is locked. + */ +static struct sctp_inpcb * +sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head, + uint16_t lport, uint32_t vrf_id) +{ + struct sctp_inpcb *inp; + struct sctp_laddr *laddr; +#ifdef INET + struct sockaddr_in *sin; +#endif +#ifdef INET6 + struct sockaddr_in6 *sin6; + struct sockaddr_in6 *intf_addr6; +#endif +#if defined(__Userspace__) + struct sockaddr_conn *sconn; +#endif +#ifdef SCTP_MVRF + int i; +#endif + int fnd; + +#ifdef INET + sin = NULL; +#endif +#ifdef INET6 + sin6 = NULL; +#endif +#if defined(__Userspace__) + sconn = NULL; +#endif + switch (nam->sa_family) { +#ifdef INET + case AF_INET: + sin = (struct sockaddr_in *)nam; + break; +#endif +#ifdef INET6 + case AF_INET6: + sin6 = (struct sockaddr_in6 *)nam; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + sconn = (struct sockaddr_conn *)nam; + break; +#endif + default: + /* unsupported family */ + return (NULL); + } + + if (head == NULL) + return (NULL); + + LIST_FOREACH(inp, head, sctp_hash) { + SCTP_INP_RLOCK(inp); + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { + SCTP_INP_RUNLOCK(inp); + continue; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) && + (inp->sctp_lport == lport)) { + /* got it */ + switch (nam->sa_family) { +#ifdef INET + case AF_INET: + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(inp)) { + /* IPv4 on a IPv6 socket with ONLY IPv6 set */ + SCTP_INP_RUNLOCK(inp); + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sin->sin_addr) != 0) { + SCTP_INP_RUNLOCK(inp); + continue; + } +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: + /* A V6 address and the endpoint is NOT bound V6 */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { + SCTP_INP_RUNLOCK(inp); + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sin6->sin6_addr) != 0) { + SCTP_INP_RUNLOCK(inp); + continue; + } +#endif + break; +#endif + default: + break; + } + /* does a VRF id match? */ + fnd = 0; +#ifdef SCTP_MVRF + for (i = 0; i < inp->num_vrfs; i++) { + if (inp->m_vrf_ids[i] == vrf_id) { + fnd = 1; + break; + } + } +#else + if (inp->def_vrf_id == vrf_id) + fnd = 1; +#endif + + SCTP_INP_RUNLOCK(inp); + if (!fnd) + continue; + return (inp); + } + SCTP_INP_RUNLOCK(inp); + } + switch (nam->sa_family) { +#ifdef INET + case AF_INET: + if (sin->sin_addr.s_addr == INADDR_ANY) { + /* Can't hunt for one that has no address specified */ + return (NULL); + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + /* Can't hunt for one that has no address specified */ + return (NULL); + } + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + if (sconn->sconn_addr == NULL) { + return (NULL); + } + break; +#endif + default: + break; + } + /* + * ok, not bound to all so see if we can find a EP bound to this + * address. + */ + LIST_FOREACH(inp, head, sctp_hash) { + SCTP_INP_RLOCK(inp); + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { + SCTP_INP_RUNLOCK(inp); + continue; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) { + SCTP_INP_RUNLOCK(inp); + continue; + } + /* + * Ok this could be a likely candidate, look at all of its + * addresses + */ + if (inp->sctp_lport != lport) { + SCTP_INP_RUNLOCK(inp); + continue; + } + /* does a VRF id match? */ + fnd = 0; +#ifdef SCTP_MVRF + for (i = 0; i < inp->num_vrfs; i++) { + if (inp->m_vrf_ids[i] == vrf_id) { + fnd = 1; + break; + } + } +#else + if (inp->def_vrf_id == vrf_id) + fnd = 1; + +#endif + if (!fnd) { + SCTP_INP_RUNLOCK(inp); + continue; + } + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa == NULL) { + SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", + __func__); + continue; + } + SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ", + (void *)laddr->ifa); + if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { + SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n"); + continue; + } + if (laddr->ifa->address.sa.sa_family == nam->sa_family) { + /* possible, see if it matches */ + switch (nam->sa_family) { +#ifdef INET + case AF_INET: +#if defined(__APPLE__) && !defined(__Userspace__) + if (sin == NULL) { + /* TSNH */ + break; + } +#endif + if (sin->sin_addr.s_addr == + laddr->ifa->address.sin.sin_addr.s_addr) { + SCTP_INP_RUNLOCK(inp); + return (inp); + } + break; +#endif +#ifdef INET6 + case AF_INET6: + intf_addr6 = &laddr->ifa->address.sin6; + if (SCTP6_ARE_ADDR_EQUAL(sin6, + intf_addr6)) { + SCTP_INP_RUNLOCK(inp); + return (inp); + } + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + if (sconn->sconn_addr == laddr->ifa->address.sconn.sconn_addr) { + SCTP_INP_RUNLOCK(inp); + return (inp); + } + break; +#endif + } + } + } + SCTP_INP_RUNLOCK(inp); + } + return (NULL); +} + +static struct sctp_inpcb * +sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id) +{ + struct sctppcbhead *head; + struct sctp_inpcb *t_inp; +#ifdef SCTP_MVRF + int i; +#endif + int fnd; + + head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, + SCTP_BASE_INFO(hashmark))]; + LIST_FOREACH(t_inp, head, sctp_hash) { + if (t_inp->sctp_lport != lport) { + continue; + } + /* is it in the VRF in question */ + fnd = 0; +#ifdef SCTP_MVRF + for (i = 0; i < inp->num_vrfs; i++) { + if (t_inp->m_vrf_ids[i] == vrf_id) { + fnd = 1; + break; + } + } +#else + if (t_inp->def_vrf_id == vrf_id) + fnd = 1; +#endif + if (!fnd) + continue; + + /* This one is in use. */ + /* check the v6/v4 binding issue */ + if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(t_inp)) { + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + /* collision in V6 space */ + return (t_inp); + } else { + /* inp is BOUND_V4 no conflict */ + continue; + } + } else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + /* t_inp is bound v4 and v6, conflict always */ + return (t_inp); + } else { + /* t_inp is bound only V4 */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(inp)) { + /* no conflict */ + continue; + } + /* else fall through to conflict */ + } + return (t_inp); + } + return (NULL); +} + +int +sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp) +{ + /* For 1-2-1 with port reuse */ + struct sctppcbhead *head; + struct sctp_inpcb *tinp, *ninp; + + SCTP_INP_INFO_WLOCK_ASSERT(); + SCTP_INP_WLOCK_ASSERT(inp); + + if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) { + /* only works with port reuse on */ + return (-1); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) { + return (0); + } + SCTP_INP_WUNLOCK(inp); + head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, + SCTP_BASE_INFO(hashmark))]; + /* Kick out all non-listeners to the TCP hash */ + LIST_FOREACH_SAFE(tinp, head, sctp_hash, ninp) { + if (tinp->sctp_lport != inp->sctp_lport) { + continue; + } + if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { + continue; + } + if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + continue; + } + if (SCTP_IS_LISTENING(tinp)) { + continue; + } + SCTP_INP_WLOCK(tinp); + LIST_REMOVE(tinp, sctp_hash); + head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))]; + tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL; + LIST_INSERT_HEAD(head, tinp, sctp_hash); + SCTP_INP_WUNLOCK(tinp); + } + SCTP_INP_WLOCK(inp); + /* Pull from where he was */ + LIST_REMOVE(inp, sctp_hash); + inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL; + head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))]; + LIST_INSERT_HEAD(head, inp, sctp_hash); + return (0); +} + +struct sctp_inpcb * +sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock, + uint32_t vrf_id) +{ + /* + * First we check the hash table to see if someone has this port + * bound with just the port. + */ + struct sctp_inpcb *inp; + struct sctppcbhead *head; + int lport; + unsigned int i; +#ifdef INET + struct sockaddr_in *sin; +#endif +#ifdef INET6 + struct sockaddr_in6 *sin6; +#endif +#if defined(__Userspace__) + struct sockaddr_conn *sconn; +#endif + + switch (nam->sa_family) { +#ifdef INET + case AF_INET: + sin = (struct sockaddr_in *)nam; + lport = sin->sin_port; + break; +#endif +#ifdef INET6 + case AF_INET6: + sin6 = (struct sockaddr_in6 *)nam; + lport = sin6->sin6_port; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + sconn = (struct sockaddr_conn *)nam; + lport = sconn->sconn_port; + break; +#endif + default: + return (NULL); + } + /* + * I could cheat here and just cast to one of the types but we will + * do it right. It also provides the check against an Unsupported + * type too. + */ + /* Find the head of the ALLADDR chain */ + if (have_lock == 0) { + SCTP_INP_INFO_RLOCK(); + } + head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, + SCTP_BASE_INFO(hashmark))]; + inp = sctp_endpoint_probe(nam, head, lport, vrf_id); + + /* + * If the TCP model exists it could be that the main listening + * endpoint is gone but there still exists a connected socket for this + * guy. If so we can return the first one that we find. This may NOT + * be the correct one so the caller should be wary on the returned INP. + * Currently the only caller that sets find_tcp_pool is in bindx where + * we are verifying that a user CAN bind the address. He either + * has bound it already, or someone else has, or its open to bind, + * so this is good enough. + */ + if (inp == NULL && find_tcp_pool) { + for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) { + head = &SCTP_BASE_INFO(sctp_tcpephash)[i]; + inp = sctp_endpoint_probe(nam, head, lport, vrf_id); + if (inp) { + break; + } + } + } + if (inp) { + SCTP_INP_INCR_REF(inp); + } + if (have_lock == 0) { + SCTP_INP_INFO_RUNLOCK(); + } + return (inp); +} + +/* + * Find an association for an endpoint with the pointer to whom you want to + * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may + * need to change the *to to some other struct like a mbuf... + */ +struct sctp_tcb * +sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to, + struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool, + uint32_t vrf_id) +{ + struct sctp_inpcb *inp = NULL; + struct sctp_tcb *stcb; + + SCTP_INP_INFO_RLOCK(); + if (find_tcp_pool) { + if (inp_p != NULL) { + stcb = sctp_tcb_special_locate(inp_p, from, to, netp, + vrf_id); + } else { + stcb = sctp_tcb_special_locate(&inp, from, to, netp, + vrf_id); + } + if (stcb != NULL) { + SCTP_INP_INFO_RUNLOCK(); + return (stcb); + } + } + inp = sctp_pcb_findep(to, 0, 1, vrf_id); + if (inp_p != NULL) { + *inp_p = inp; + } + SCTP_INP_INFO_RUNLOCK(); + if (inp == NULL) { + return (NULL); + } + /* + * ok, we have an endpoint, now lets find the assoc for it (if any) + * we now place the source address or from in the to of the find + * endpoint call. Since in reality this chain is used from the + * inbound packet side. + */ + if (inp_p != NULL) { + stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to, + NULL); + } else { + stcb = sctp_findassociation_ep_addr(&inp, from, netp, to, + NULL); + } + return (stcb); +} + +/* + * This routine will grub through the mbuf that is a INIT or INIT-ACK and + * find all addresses that the sender has specified in any address list. Each + * address will be used to lookup the TCB and see if one exits. + */ +static struct sctp_tcb * +sctp_findassociation_special_addr(struct mbuf *m, int offset, + struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, + struct sockaddr *dst) +{ + struct sctp_paramhdr *phdr, param_buf; +#if defined(INET) || defined(INET6) + struct sctp_tcb *stcb; + uint16_t ptype; +#endif + uint16_t plen; +#ifdef INET + struct sockaddr_in sin4; +#endif +#ifdef INET6 + struct sockaddr_in6 sin6; +#endif + +#ifdef INET + memset(&sin4, 0, sizeof(sin4)); +#ifdef HAVE_SIN_LEN + sin4.sin_len = sizeof(sin4); +#endif + sin4.sin_family = AF_INET; + sin4.sin_port = sh->src_port; +#endif +#ifdef INET6 + memset(&sin6, 0, sizeof(sin6)); +#ifdef HAVE_SIN6_LEN + sin6.sin6_len = sizeof(sin6); +#endif + sin6.sin6_family = AF_INET6; + sin6.sin6_port = sh->src_port; +#endif + + offset += sizeof(struct sctp_init_chunk); + + phdr = sctp_get_next_param(m, offset, ¶m_buf, sizeof(param_buf)); + while (phdr != NULL) { + /* now we must see if we want the parameter */ +#if defined(INET) || defined(INET6) + ptype = ntohs(phdr->param_type); +#endif + plen = ntohs(phdr->param_length); + if (plen == 0) { + break; + } +#ifdef INET + if (ptype == SCTP_IPV4_ADDRESS && + plen == sizeof(struct sctp_ipv4addr_param)) { + /* Get the rest of the address */ + struct sctp_ipv4addr_param ip4_param, *p4; + + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)&ip4_param, sizeof(ip4_param)); + if (phdr == NULL) { + return (NULL); + } + p4 = (struct sctp_ipv4addr_param *)phdr; + memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr)); + /* look it up */ + stcb = sctp_findassociation_ep_addr(inp_p, + (struct sockaddr *)&sin4, netp, dst, NULL); + if (stcb != NULL) { + return (stcb); + } + } +#endif +#ifdef INET6 + if (ptype == SCTP_IPV6_ADDRESS && + plen == sizeof(struct sctp_ipv6addr_param)) { + /* Get the rest of the address */ + struct sctp_ipv6addr_param ip6_param, *p6; + + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)&ip6_param, sizeof(ip6_param)); + if (phdr == NULL) { + return (NULL); + } + p6 = (struct sctp_ipv6addr_param *)phdr; + memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr)); + /* look it up */ + stcb = sctp_findassociation_ep_addr(inp_p, + (struct sockaddr *)&sin6, netp, dst, NULL); + if (stcb != NULL) { + return (stcb); + } + } +#endif + offset += SCTP_SIZE32(plen); + phdr = sctp_get_next_param(m, offset, ¶m_buf, + sizeof(param_buf)); + } + return (NULL); +} + +static struct sctp_tcb * +sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag, + struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport, + uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag) +{ + /* + * Use my vtag to hash. If we find it we then verify the source addr + * is in the assoc. If all goes well we save a bit on rec of a + * packet. + */ + struct sctpasochead *head; + struct sctp_nets *net; + struct sctp_tcb *stcb; +#ifdef SCTP_MVRF + unsigned int i; +#endif + + SCTP_INP_INFO_RLOCK(); + head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag, + SCTP_BASE_INFO(hashasocmark))]; + LIST_FOREACH(stcb, head, sctp_asocs) { + SCTP_INP_RLOCK(stcb->sctp_ep); + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { + SCTP_INP_RUNLOCK(stcb->sctp_ep); + continue; + } +#ifdef SCTP_MVRF + for (i = 0; i < stcb->sctp_ep->num_vrfs; i++) { + if (stcb->sctp_ep->m_vrf_ids[i] == vrf_id) { + break; + } + } + if (i == stcb->sctp_ep->num_vrfs) { + SCTP_INP_RUNLOCK(inp); + continue; + } +#else + if (stcb->sctp_ep->def_vrf_id != vrf_id) { + SCTP_INP_RUNLOCK(stcb->sctp_ep); + continue; + } +#endif + SCTP_TCB_LOCK(stcb); + SCTP_INP_RUNLOCK(stcb->sctp_ep); + if (stcb->asoc.my_vtag == vtag) { + /* candidate */ + if (stcb->rport != rport) { + SCTP_TCB_UNLOCK(stcb); + continue; + } + if (stcb->sctp_ep->sctp_lport != lport) { + SCTP_TCB_UNLOCK(stcb); + continue; + } + if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + SCTP_TCB_UNLOCK(stcb); + continue; + } + /* RRS:Need toaddr check here */ + if (sctp_does_stcb_own_this_addr(stcb, to) == 0) { + /* Endpoint does not own this address */ + SCTP_TCB_UNLOCK(stcb); + continue; + } + if (remote_tag) { + /* If we have both vtags that's all we match on */ + if (stcb->asoc.peer_vtag == remote_tag) { + /* If both tags match we consider it conclusive + * and check NO source/destination addresses + */ + goto conclusive; + } + } + if (skip_src_check) { + conclusive: + if (from) { + *netp = sctp_findnet(stcb, from); + } else { + *netp = NULL; /* unknown */ + } + if (inp_p) + *inp_p = stcb->sctp_ep; + SCTP_INP_INFO_RUNLOCK(); + return (stcb); + } + net = sctp_findnet(stcb, from); + if (net) { + /* yep its him. */ + *netp = net; + SCTP_STAT_INCR(sctps_vtagexpress); + *inp_p = stcb->sctp_ep; + SCTP_INP_INFO_RUNLOCK(); + return (stcb); + } else { + /* + * not him, this should only happen in rare + * cases so I peg it. + */ + SCTP_STAT_INCR(sctps_vtagbogus); + } + } + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_INFO_RUNLOCK(); + return (NULL); +} + +/* + * Find an association with the pointer to the inbound IP packet. This can be + * a IPv4 or IPv6 packet. + */ +struct sctp_tcb * +sctp_findassociation_addr(struct mbuf *m, int offset, + struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, struct sctp_chunkhdr *ch, + struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id) +{ + struct sctp_tcb *stcb; + struct sctp_inpcb *inp; + + if (sh->v_tag) { + /* we only go down this path if vtag is non-zero */ + stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag), + inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0); + if (stcb) { + return (stcb); + } + } + + if (inp_p) { + stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp, + 1, vrf_id); + inp = *inp_p; + } else { + stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp, + 1, vrf_id); + } + SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp); + if (stcb == NULL && inp) { + /* Found a EP but not this address */ + if ((ch->chunk_type == SCTP_INITIATION) || + (ch->chunk_type == SCTP_INITIATION_ACK)) { + /*- + * special hook, we do NOT return linp or an + * association that is linked to an existing + * association that is under the TCP pool (i.e. no + * listener exists). The endpoint finding routine + * will always find a listener before examining the + * TCP pool. + */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) { + if (inp_p) { + *inp_p = NULL; + } + return (NULL); + } + stcb = sctp_findassociation_special_addr(m, + offset, sh, &inp, netp, dst); + if (inp_p != NULL) { + *inp_p = inp; + } + } + } + SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb); + return (stcb); +} + +/* + * lookup an association by an ASCONF lookup address. + * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup + */ +struct sctp_tcb * +sctp_findassociation_ep_asconf(struct mbuf *m, int offset, + struct sockaddr *dst, struct sctphdr *sh, + struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id) +{ + struct sctp_tcb *stcb; + union sctp_sockstore remote_store; + struct sctp_paramhdr param_buf, *phdr; + int ptype; + int zero_address = 0; +#ifdef INET + struct sockaddr_in *sin; +#endif +#ifdef INET6 + struct sockaddr_in6 *sin6; +#endif + + memset(&remote_store, 0, sizeof(remote_store)); + phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk), + ¶m_buf, sizeof(struct sctp_paramhdr)); + if (phdr == NULL) { + SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n", + __func__); + return NULL; + } + ptype = (int)((uint32_t) ntohs(phdr->param_type)); + /* get the correlation address */ + switch (ptype) { +#ifdef INET6 + case SCTP_IPV6_ADDRESS: + { + /* ipv6 address param */ + struct sctp_ipv6addr_param *p6, p6_buf; + + if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) { + return NULL; + } + p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m, + offset + sizeof(struct sctp_asconf_chunk), + &p6_buf.ph, sizeof(p6_buf)); + if (p6 == NULL) { + SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n", + __func__); + return (NULL); + } + sin6 = &remote_store.sin6; + sin6->sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6->sin6_len = sizeof(*sin6); +#endif + sin6->sin6_port = sh->src_port; + memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr)); + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) + zero_address = 1; + break; + } +#endif +#ifdef INET + case SCTP_IPV4_ADDRESS: + { + /* ipv4 address param */ + struct sctp_ipv4addr_param *p4, p4_buf; + + if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) { + return NULL; + } + p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m, + offset + sizeof(struct sctp_asconf_chunk), + &p4_buf.ph, sizeof(p4_buf)); + if (p4 == NULL) { + SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n", + __func__); + return (NULL); + } + sin = &remote_store.sin; + sin->sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + sin->sin_len = sizeof(*sin); +#endif + sin->sin_port = sh->src_port; + memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr)); + if (sin->sin_addr.s_addr == INADDR_ANY) + zero_address = 1; + break; + } +#endif + default: + /* invalid address param type */ + return NULL; + } + + if (zero_address) { + stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p, + netp, sh->src_port, sh->dest_port, 1, vrf_id, 0); + if (stcb != NULL) { + SCTP_INP_DECR_REF(*inp_p); + } + } else { + stcb = sctp_findassociation_ep_addr(inp_p, + &remote_store.sa, netp, + dst, NULL); + } + return (stcb); +} + +/* + * allocate a sctp_inpcb and setup a temporary binding to a port/all + * addresses. This way if we don't get a bind we by default pick a ephemeral + * port with all addresses bound. + */ +int +sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id) +{ + /* + * we get called when a new endpoint starts up. We need to allocate + * the sctp_inpcb structure from the zone and init it. Mark it as + * unbound and find a port that we can use as an ephemeral with + * INADDR_ANY. If the user binds later no problem we can then add in + * the specific addresses. And setup the default parameters for the + * EP. + */ + int i, error; + struct sctp_inpcb *inp; + struct sctp_pcb *m; + struct timeval time; + sctp_sharedkey_t *null_key; + + error = 0; + + SCTP_INP_INFO_WLOCK(); + inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb); + if (inp == NULL) { + SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n"); + SCTP_INP_INFO_WUNLOCK(); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); + return (ENOBUFS); + } + /* zap it */ + memset(inp, 0, sizeof(*inp)); + + /* bump generations */ +#if defined(__APPLE__) && !defined(__Userspace__) + inp->ip_inp.inp.inp_state = INPCB_STATE_INUSE; +#endif + /* setup socket pointers */ + inp->sctp_socket = so; + inp->ip_inp.inp.inp_socket = so; +#if defined(__FreeBSD__) && !defined(__Userspace__) + inp->ip_inp.inp.inp_cred = crhold(so->so_cred); +#endif +#ifdef INET6 +#if !defined(__Userspace__) && !defined(_WIN32) + if (INP_SOCKAF(so) == AF_INET6) { + if (MODULE_GLOBAL(ip6_auto_flowlabel)) { + inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL; + } + if (MODULE_GLOBAL(ip6_v6only)) { + inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY; + } + } +#endif +#endif + inp->sctp_associd_counter = 1; + inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT; + inp->sctp_frag_point = 0; + inp->max_cwnd = 0; + inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off); + inp->ecn_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_ecn_enable); + inp->prsctp_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_pr_enable); + inp->auth_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_auth_enable); + inp->asconf_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_asconf_enable); + inp->reconfig_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_reconfig_enable); + inp->nrsack_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_nrsack_enable); + inp->pktdrop_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_pktdrop_enable); + inp->idata_supported = 0; + inp->zero_checksum = 0; + +#if defined(__FreeBSD__) && !defined(__Userspace__) + inp->fibnum = so->so_fibnum; +#else + inp->fibnum = 0; +#endif +#if defined(__Userspace__) + inp->ulp_info = NULL; + inp->recv_callback = NULL; + inp->send_callback = NULL; + inp->send_sb_threshold = 0; +#endif + /* init the small hash table we use to track asocid <-> tcb */ + inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark); + if (inp->sctp_asocidhash == NULL) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + crfree(inp->ip_inp.inp.inp_cred); +#endif + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); + SCTP_INP_INFO_WUNLOCK(); + return (ENOBUFS); + } + SCTP_INCR_EP_COUNT(); + inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl); + SCTP_INP_INFO_WUNLOCK(); + + so->so_pcb = (caddr_t)inp; + + if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) { + /* UDP style socket */ + inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE | + SCTP_PCB_FLAGS_UNBOUND); + /* Be sure it is NON-BLOCKING IO for UDP */ + /* SCTP_SET_SO_NBIO(so); */ + } else if (SCTP_SO_TYPE(so) == SOCK_STREAM) { + /* TCP style socket */ + inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE | + SCTP_PCB_FLAGS_UNBOUND); + /* Be sure we have blocking IO by default */ + SOCK_LOCK(so); + SCTP_CLEAR_SO_NBIO(so); + SOCK_UNLOCK(so); + } else { + /* + * unsupported socket type (RAW, etc)- in case we missed it + * in protosw + */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP); + so->so_pcb = NULL; +#if defined(__FreeBSD__) && !defined(__Userspace__) + crfree(inp->ip_inp.inp.inp_cred); +#endif + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); + return (EOPNOTSUPP); + } + if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); + sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); + } else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); + sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); + } else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) { + sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); + sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); + } + inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize), + &inp->sctp_hashmark); + if (inp->sctp_tcbhash == NULL) { + SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n"); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); + so->so_pcb = NULL; +#if defined(__FreeBSD__) && !defined(__Userspace__) + crfree(inp->ip_inp.inp.inp_cred); +#endif + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); + return (ENOBUFS); + } +#ifdef SCTP_MVRF + inp->vrf_size = SCTP_DEFAULT_VRF_SIZE; + SCTP_MALLOC(inp->m_vrf_ids, uint32_t *, + (sizeof(uint32_t) * inp->vrf_size), SCTP_M_MVRF); + if (inp->m_vrf_ids == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); + so->so_pcb = NULL; + SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark); +#if defined(__FreeBSD__) && !defined(__Userspace__) + crfree(inp->ip_inp.inp.inp_cred); +#endif + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); + return (ENOBUFS); + } + inp->m_vrf_ids[0] = vrf_id; + inp->num_vrfs = 1; +#endif + inp->def_vrf_id = vrf_id; + +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) + inp->ip_inp.inp.inpcb_mtx = lck_mtx_alloc_init(SCTP_BASE_INFO(sctbinfo).mtx_grp, SCTP_BASE_INFO(sctbinfo).mtx_attr); + if (inp->ip_inp.inp.inpcb_mtx == NULL) { + SCTP_PRINTF("in_pcballoc: can't alloc mutex! so=%p\n", (void *)so); +#ifdef SCTP_MVRF + SCTP_FREE(inp->m_vrf_ids, SCTP_M_MVRF); +#endif + SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark); + so->so_pcb = NULL; + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); + SCTP_UNLOCK_EXC(SCTP_BASE_INFO(sctbinfo).ipi_lock); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM); + return (ENOMEM); + } +#elif defined(APPLE_LION) || defined(APPLE_MOUNTAINLION) + lck_mtx_init(&inp->ip_inp.inp.inpcb_mtx, SCTP_BASE_INFO(sctbinfo).mtx_grp, SCTP_BASE_INFO(sctbinfo).mtx_attr); +#else + lck_mtx_init(&inp->ip_inp.inp.inpcb_mtx, SCTP_BASE_INFO(sctbinfo).ipi_lock_grp, SCTP_BASE_INFO(sctbinfo).ipi_lock_attr); +#endif +#endif + SCTP_INP_INFO_WLOCK(); + SCTP_INP_LOCK_INIT(inp); +#if defined(__FreeBSD__) && !defined(__Userspace__) + rw_init_flags(&inp->ip_inp.inp.inp_lock, "sctpinp", + RW_RECURSE | RW_DUPOK); +#endif + SCTP_INP_READ_INIT(inp); + SCTP_ASOC_CREATE_LOCK_INIT(inp); + /* lock the new ep */ + SCTP_INP_WLOCK(inp); + + /* add it to the info area */ + LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list); +#if defined(__APPLE__) && !defined(__Userspace__) + inp->ip_inp.inp.inp_pcbinfo = &SCTP_BASE_INFO(sctbinfo); +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) || defined(APPLE_LION) || defined(APPLE_MOUNTAINLION) + LIST_INSERT_HEAD(SCTP_BASE_INFO(sctbinfo).listhead, &inp->ip_inp.inp, inp_list); +#else + LIST_INSERT_HEAD(SCTP_BASE_INFO(sctbinfo).ipi_listhead, &inp->ip_inp.inp, inp_list); +#endif +#endif + SCTP_INP_INFO_WUNLOCK(); + + TAILQ_INIT(&inp->read_queue); + LIST_INIT(&inp->sctp_addr_list); + + LIST_INIT(&inp->sctp_asoc_list); + +#ifdef SCTP_TRACK_FREED_ASOCS + /* TEMP CODE */ + LIST_INIT(&inp->sctp_asoc_free_list); +#endif + /* Init the timer structure for signature change */ + SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer); + inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE; + + /* now init the actual endpoint default data */ + m = &inp->sctp_ep; + + /* setup the base timeout information */ + m->sctp_timeoutticks[SCTP_TIMER_SEND] = sctp_secs_to_ticks(SCTP_SEND_SEC); /* needed ? */ + m->sctp_timeoutticks[SCTP_TIMER_INIT] = sctp_secs_to_ticks(SCTP_INIT_SEC); /* needed ? */ + m->sctp_timeoutticks[SCTP_TIMER_RECV] = sctp_msecs_to_ticks(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default)); + m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = sctp_msecs_to_ticks(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default)); + m->sctp_timeoutticks[SCTP_TIMER_PMTU] = sctp_secs_to_ticks(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default)); + m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = sctp_secs_to_ticks(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default)); + m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = sctp_secs_to_ticks(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default)); + /* all max/min max are in ms */ + m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default); + m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default); + m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default); + m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default); + m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default); + m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default); + m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default); + m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default); + m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold); + m->sctp_sws_sender = SCTP_SWS_SENDER_DEF; + m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF; + m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default); + m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default); + + m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module); + m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module); + m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default); + /* number of streams to pre-open on a association */ + m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default); + + m->default_mtu = 0; + /* Add adaptation cookie */ + m->adaptation_layer_indicator = 0; + m->adaptation_layer_indicator_provided = 0; + + /* seed random number generator */ + m->random_counter = 1; + m->store_at = SCTP_SIGNATURE_SIZE; + SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers)); + sctp_fill_random_store(m); + + /* Minimum cookie size */ + m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) + + sizeof(struct sctp_state_cookie); + m->size_of_a_cookie += SCTP_SIGNATURE_SIZE; + + /* Setup the initial secret */ + (void)SCTP_GETTIME_TIMEVAL(&time); + m->time_of_secret_change = (unsigned int)time.tv_sec; + + for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) { + m->secret_key[0][i] = sctp_select_initial_TSN(m); + } + sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL); + + /* How long is a cookie good for ? */ + m->def_cookie_life = sctp_msecs_to_ticks(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default)); + /* + * Initialize authentication parameters + */ + m->local_hmacs = sctp_default_supported_hmaclist(); + m->local_auth_chunks = sctp_alloc_chunklist(); + if (inp->asconf_supported) { + sctp_auth_add_chunk(SCTP_ASCONF, m->local_auth_chunks); + sctp_auth_add_chunk(SCTP_ASCONF_ACK, m->local_auth_chunks); + } + m->default_dscp = 0; +#ifdef INET6 + m->default_flowlabel = 0; +#endif + m->port = 0; /* encapsulation disabled by default */ + LIST_INIT(&m->shared_keys); + /* add default NULL key as key id 0 */ + null_key = sctp_alloc_sharedkey(); + sctp_insert_sharedkey(&m->shared_keys, null_key); + SCTP_INP_WUNLOCK(inp); +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 12); +#endif + return (error); +} + +void +sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp, + struct sctp_tcb *stcb) +{ + struct sctp_nets *net; + uint16_t lport, rport; + struct sctppcbhead *head; + struct sctp_laddr *laddr, *oladdr; + + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_INFO_WLOCK(); + SCTP_INP_WLOCK(old_inp); + SCTP_INP_WLOCK(new_inp); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#ifdef INET6 + if (old_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + new_inp->ip_inp.inp.inp_flags |= old_inp->ip_inp.inp.inp_flags & INP_CONTROLOPTS; + if (old_inp->ip_inp.inp.in6p_outputopts) { + new_inp->ip_inp.inp.in6p_outputopts = ip6_copypktopts(old_inp->ip_inp.inp.in6p_outputopts, M_NOWAIT); + } + } +#endif +#if defined(INET) && defined(INET6) + else +#endif +#ifdef INET + { + new_inp->ip_inp.inp.inp_ip_tos = old_inp->ip_inp.inp.inp_ip_tos; + new_inp->ip_inp.inp.inp_ip_ttl = old_inp->ip_inp.inp.inp_ip_ttl; + } +#endif +#endif + new_inp->sctp_ep.time_of_secret_change = + old_inp->sctp_ep.time_of_secret_change; + memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key, + sizeof(old_inp->sctp_ep.secret_key)); + new_inp->sctp_ep.current_secret_number = + old_inp->sctp_ep.current_secret_number; + new_inp->sctp_ep.last_secret_number = + old_inp->sctp_ep.last_secret_number; + new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie; + + /* make it so new data pours into the new socket */ + stcb->sctp_socket = new_inp->sctp_socket; + stcb->sctp_ep = new_inp; + + /* Copy the port across */ + lport = new_inp->sctp_lport = old_inp->sctp_lport; + rport = stcb->rport; + /* Pull the tcb from the old association */ + LIST_REMOVE(stcb, sctp_tcbhash); + LIST_REMOVE(stcb, sctp_tcblist); + if (stcb->asoc.in_asocid_hash) { + LIST_REMOVE(stcb, sctp_tcbasocidhash); + } + /* Now insert the new_inp into the TCP connected hash */ + head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))]; + + LIST_INSERT_HEAD(head, new_inp, sctp_hash); + /* Its safe to access */ + new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND; + + /* Now move the tcb into the endpoint list */ + LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist); + /* + * Question, do we even need to worry about the ep-hash since we + * only have one connection? Probably not :> so lets get rid of it + * and not suck up any kernel memory in that. + */ + if (stcb->asoc.in_asocid_hash) { + struct sctpasochead *lhd; + lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id, + new_inp->hashasocidmark)]; + LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash); + } + /* Ok. Let's restart timer. */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp, + stcb, net); + } + + SCTP_INP_INFO_WUNLOCK(); + if (new_inp->sctp_tcbhash != NULL) { + SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark); + new_inp->sctp_tcbhash = NULL; + } + if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { + /* Subset bound, so copy in the laddr list from the old_inp */ + LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) { + laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); + if (laddr == NULL) { + /* + * Gak, what can we do? This assoc is really + * HOSED. We probably should send an abort + * here. + */ + SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n"); + continue; + } + SCTP_INCR_LADDR_COUNT(); + memset(laddr, 0, sizeof(*laddr)); + (void)SCTP_GETTIME_TIMEVAL(&laddr->start_time); + laddr->ifa = oladdr->ifa; + atomic_add_int(&laddr->ifa->refcount, 1); + LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr, + sctp_nxt_addr); + new_inp->laddr_count++; + if (oladdr == stcb->asoc.last_used_address) { + stcb->asoc.last_used_address = laddr; + } + } + } + /* Now any running timers need to be adjusted. */ + if (stcb->asoc.dack_timer.ep == old_inp) { + SCTP_INP_DECR_REF(old_inp); + stcb->asoc.dack_timer.ep = new_inp; + SCTP_INP_INCR_REF(new_inp); + } + if (stcb->asoc.asconf_timer.ep == old_inp) { + SCTP_INP_DECR_REF(old_inp); + stcb->asoc.asconf_timer.ep = new_inp; + SCTP_INP_INCR_REF(new_inp); + } + if (stcb->asoc.strreset_timer.ep == old_inp) { + SCTP_INP_DECR_REF(old_inp); + stcb->asoc.strreset_timer.ep = new_inp; + SCTP_INP_INCR_REF(new_inp); + } + if (stcb->asoc.shut_guard_timer.ep == old_inp) { + SCTP_INP_DECR_REF(old_inp); + stcb->asoc.shut_guard_timer.ep = new_inp; + SCTP_INP_INCR_REF(new_inp); + } + if (stcb->asoc.autoclose_timer.ep == old_inp) { + SCTP_INP_DECR_REF(old_inp); + stcb->asoc.autoclose_timer.ep = new_inp; + SCTP_INP_INCR_REF(new_inp); + } + if (stcb->asoc.delete_prim_timer.ep == old_inp) { + SCTP_INP_DECR_REF(old_inp); + stcb->asoc.delete_prim_timer.ep = new_inp; + SCTP_INP_INCR_REF(new_inp); + } + /* now what about the nets? */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if (net->pmtu_timer.ep == old_inp) { + SCTP_INP_DECR_REF(old_inp); + net->pmtu_timer.ep = new_inp; + SCTP_INP_INCR_REF(new_inp); + } + if (net->hb_timer.ep == old_inp) { + SCTP_INP_DECR_REF(old_inp); + net->hb_timer.ep = new_inp; + SCTP_INP_INCR_REF(new_inp); + } + if (net->rxt_timer.ep == old_inp) { + SCTP_INP_DECR_REF(old_inp); + net->rxt_timer.ep = new_inp; + SCTP_INP_INCR_REF(new_inp); + } + } + SCTP_INP_WUNLOCK(new_inp); + SCTP_INP_WUNLOCK(old_inp); +} + +/* + * insert an laddr entry with the given ifa for the desired list + */ +static int +sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act) +{ + struct sctp_laddr *laddr; + + laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); + if (laddr == NULL) { + /* out of memory? */ + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); + return (EINVAL); + } + SCTP_INCR_LADDR_COUNT(); + memset(laddr, 0, sizeof(*laddr)); + (void)SCTP_GETTIME_TIMEVAL(&laddr->start_time); + laddr->ifa = ifa; + laddr->action = act; + atomic_add_int(&ifa->refcount, 1); + /* insert it */ + LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr); + + return (0); +} + +/* + * Remove an laddr entry from the local address list (on an assoc) + */ +static void +sctp_remove_laddr(struct sctp_laddr *laddr) +{ + + /* remove from the list */ + LIST_REMOVE(laddr, sctp_nxt_addr); + sctp_free_ifa(laddr->ifa); + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr); + SCTP_DECR_LADDR_COUNT(); +} +#if !(defined(__FreeBSD__) || defined(__APPLE__) || defined(__Userspace__)) + +/* + * Don't know why, but without this there is an unknown reference when + * compiling NetBSD... hmm + */ +extern void in6_sin6_2_sin(struct sockaddr_in *, struct sockaddr_in6 *sin6); +#endif + +/* + * Bind the socket, with the PCB and global info locks held. Note, if a + * socket address is specified, the PCB lock may be dropped and re-acquired. + * + * sctp_ifap is used to bypass normal local address validation checks. + */ +int +#if defined(__FreeBSD__) && !defined(__Userspace__) +sctp_inpcb_bind_locked(struct sctp_inpcb *inp, struct sockaddr *addr, + struct sctp_ifa *sctp_ifap, struct thread *td) +#elif defined(_WIN32) && !defined(__Userspace__) +sctp_inpcb_bind_locked(struct sctp_inpcb *inp, struct sockaddr *addr, + struct sctp_ifa *sctp_ifap, PKTHREAD p) +#else +sctp_inpcb_bind_locked(struct sctp_inpcb *inp, struct sockaddr *addr, + struct sctp_ifa *sctp_ifap, struct proc *p) +#endif +{ + /* bind a ep to a socket address */ + struct sctppcbhead *head; + struct sctp_inpcb *inp_tmp; +#if (defined(__FreeBSD__) || defined(__APPLE__)) && !defined(__Userspace__) + struct inpcb *ip_inp; +#endif + int port_reuse_active = 0; + int bindall; +#ifdef SCTP_MVRF + int i; +#endif + uint16_t lport; + int error; + uint32_t vrf_id; + +#if defined(__FreeBSD__) && !defined(__Userspace__) + KASSERT(td != NULL, ("%s: null thread", __func__)); + +#endif + error = 0; + lport = 0; + bindall = 1; +#if (defined(__FreeBSD__) || defined(__APPLE__)) && !defined(__Userspace__) + ip_inp = &inp->ip_inp.inp; +#endif + + SCTP_INP_INFO_WLOCK_ASSERT(); + SCTP_INP_WLOCK_ASSERT(inp); + +#ifdef SCTP_DEBUG + if (addr) { + SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n", + ntohs(((struct sockaddr_in *)addr)->sin_port)); + SCTPDBG(SCTP_DEBUG_PCB1, "Addr: "); + SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr); + } +#endif + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) { + error = EINVAL; + /* already did a bind, subsequent binds NOT allowed ! */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } + if (addr != NULL) { + switch (addr->sa_family) { +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin; + + /* IPV6_V6ONLY socket? */ + if (SCTP_IPV6_V6ONLY(inp)) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } +#ifdef HAVE_SA_LEN + if (addr->sa_len != sizeof(*sin)) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } +#endif + + sin = (struct sockaddr_in *)addr; + lport = sin->sin_port; +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* + * For LOOPBACK the prison_local_ip4() call will transmute the ip address + * to the proper value. + */ + if ((error = prison_local_ip4(td->td_ucred, &sin->sin_addr)) != 0) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } +#endif + if (sin->sin_addr.s_addr != INADDR_ANY) { + bindall = 0; + } + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + /* Only for pure IPv6 Address. (No IPv4 Mapped!) */ + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)addr; +#ifdef HAVE_SA_LEN + if (addr->sa_len != sizeof(*sin6)) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } +#endif + lport = sin6->sin6_port; +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* + * For LOOPBACK the prison_local_ip6() call will transmute the ipv6 address + * to the proper value. + */ + if ((error = prison_local_ip6(td->td_ucred, &sin6->sin6_addr, + (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } +#endif + if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + bindall = 0; +#ifdef SCTP_EMBEDDED_V6_SCOPE + /* KAME hack: embed scopeid */ +#if defined(SCTP_KAME) + if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } +#elif defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) + if (in6_embedscope(&sin6->sin6_addr, sin6, ip_inp, NULL) != 0) { +#else + if (in6_embedscope(&sin6->sin6_addr, sin6, ip_inp, NULL, NULL) != 0) { +#endif + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } +#elif defined(__FreeBSD__) && !defined(__Userspace__) + error = scope6_check_id(sin6, MODULE_GLOBAL(ip6_use_defzone)); + if (error != 0) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } +#else + if (in6_embedscope(&sin6->sin6_addr, sin6) != 0) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } +#endif +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + } +#ifndef SCOPEDROUTING + /* this must be cleared for ifa_ifwithaddr() */ + sin6->sin6_scope_id = 0; +#endif /* SCOPEDROUTING */ + break; + } +#endif +#if defined(__Userspace__) + case AF_CONN: + { + struct sockaddr_conn *sconn; + +#ifdef HAVE_SA_LEN + if (addr->sa_len != sizeof(struct sockaddr_conn)) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } +#endif + sconn = (struct sockaddr_conn *)addr; + lport = sconn->sconn_port; + if (sconn->sconn_addr != NULL) { + bindall = 0; + } + break; + } +#endif + default: + error = EAFNOSUPPORT; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } + } + /* Setup a vrf_id to be the default for the non-bind-all case. */ + vrf_id = inp->def_vrf_id; + + if (lport) { + /* + * Did the caller specify a port? if so we must see if an ep + * already has this one bound. + */ + /* got to be root to get at low ports */ +#if !(defined(_WIN32) && !defined(__Userspace__)) + if (ntohs(lport) < IPPORT_RESERVED && +#if defined(__FreeBSD__) && !defined(__Userspace__) + (error = priv_check(td, PRIV_NETINET_RESERVEDPORT)) != 0) { +#elif defined(__APPLE__) && !defined(__Userspace__) + (error = suser(p->p_ucred, &p->p_acflag)) != 0) { +#elif defined(__Userspace__) + /* TODO ensure uid is 0, etc... */ + 0) { +#else + (error = suser(p, 0)) != 0) { +#endif + goto out; + } +#endif + SCTP_INP_INCR_REF(inp); + SCTP_INP_WUNLOCK(inp); + if (bindall) { +#ifdef SCTP_MVRF + for (i = 0; i < inp->num_vrfs; i++) { + vrf_id = inp->m_vrf_ids[i]; +#else + vrf_id = inp->def_vrf_id; +#endif + inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id); + if (inp_tmp != NULL) { + /* + * lock guy returned and lower count + * note that we are not bound so + * inp_tmp should NEVER be inp. And + * it is this inp (inp_tmp) that gets + * the reference bump, so we must + * lower it. + */ + SCTP_INP_DECR_REF(inp_tmp); + /* unlock info */ + if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && + (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { + /* Ok, must be one-2-one and allowing port re-use */ + port_reuse_active = 1; + goto continue_anyway; + } + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + error = EADDRINUSE; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } +#ifdef SCTP_MVRF + } +#endif + } else { + inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id); + if (inp_tmp != NULL) { + /* + * lock guy returned and lower count note + * that we are not bound so inp_tmp should + * NEVER be inp. And it is this inp (inp_tmp) + * that gets the reference bump, so we must + * lower it. + */ + SCTP_INP_DECR_REF(inp_tmp); + /* unlock info */ + if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && + (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { + /* Ok, must be one-2-one and allowing port re-use */ + port_reuse_active = 1; + goto continue_anyway; + } + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + error = EADDRINUSE; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } + } + continue_anyway: + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + if (bindall) { + /* verify that no lport is not used by a singleton */ + if ((port_reuse_active == 0) && + (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) { + /* Sorry someone already has this one bound */ + if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && + (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { + port_reuse_active = 1; + } else { + error = EADDRINUSE; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } + } + } + } else { + uint16_t first, last, candidate; + uint16_t count; + +#if defined(__Userspace__) + first = MODULE_GLOBAL(ipport_firstauto); + last = MODULE_GLOBAL(ipport_lastauto); +#elif defined(_WIN32) + first = 1; + last = 0xffff; +#elif defined(__FreeBSD__) || defined(__APPLE__) + if (ip_inp->inp_flags & INP_HIGHPORT) { + first = MODULE_GLOBAL(ipport_hifirstauto); + last = MODULE_GLOBAL(ipport_hilastauto); + } else if (ip_inp->inp_flags & INP_LOWPORT) { +#if defined(__FreeBSD__) + if ((error = priv_check(td, PRIV_NETINET_RESERVEDPORT)) != 0) { +#else + if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) { +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } + first = MODULE_GLOBAL(ipport_lowfirstauto); + last = MODULE_GLOBAL(ipport_lowlastauto); + } else { + first = MODULE_GLOBAL(ipport_firstauto); + last = MODULE_GLOBAL(ipport_lastauto); + } +#endif + if (first > last) { + uint16_t temp; + + temp = first; + first = last; + last = temp; + } + count = last - first + 1; /* number of candidates */ + candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count); + + for (;;) { +#ifdef SCTP_MVRF + for (i = 0; i < inp->num_vrfs; i++) { + if (sctp_isport_inuse(inp, htons(candidate), inp->m_vrf_ids[i]) != NULL) { + break; + } + } + if (i == inp->num_vrfs) { + lport = htons(candidate); + break; + } +#else + if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) { + lport = htons(candidate); + break; + } +#endif + if (--count == 0) { + error = EADDRINUSE; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } + if (candidate == last) + candidate = first; + else + candidate = candidate + 1; + } + } + if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | + SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { + /* + * this really should not happen. The guy did a non-blocking + * bind and then did a close at the same time. + */ + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } + /* ok we look clear to give out this port, so lets setup the binding */ + if (bindall) { + /* binding to all addresses, so just set in the proper flags */ + inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL; + /* set the automatic addr changes from kernel flag */ + if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) { + sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF); + sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); + } else { + sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF); + sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); + } + if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) { + sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS); + } else { + sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS); + } + /* set the automatic mobility_base from kernel + flag (by micchie) + */ + if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) { + sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE); + sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); + } else { + sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE); + sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); + } + /* set the automatic mobility_fasthandoff from kernel + flag (by micchie) + */ + if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) { + sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF); + sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); + } else { + sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF); + sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); + } + } else { + /* + * bind specific, make sure flags is off and add a new + * address structure to the sctp_addr_list inside the ep + * structure. + * + * We will need to allocate one and insert it at the head. The + * socketopt call can just insert new addresses in there as + * well. It will also have to do the embed scope kame hack + * too (before adding). + */ + struct sctp_ifa *ifa; + union sctp_sockstore store; + + memset(&store, 0, sizeof(store)); + switch (addr->sa_family) { +#ifdef INET + case AF_INET: + memcpy(&store.sin, addr, sizeof(struct sockaddr_in)); + store.sin.sin_port = 0; + break; +#endif +#ifdef INET6 + case AF_INET6: + memcpy(&store.sin6, addr, sizeof(struct sockaddr_in6)); + store.sin6.sin6_port = 0; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + memcpy(&store.sconn, addr, sizeof(struct sockaddr_conn)); + store.sconn.sconn_port = 0; + break; +#endif + default: + break; + } + /* + * first find the interface with the bound address need to + * zero out the port to find the address! yuck! can't do + * this earlier since need port for sctp_pcb_findep() + */ + if (sctp_ifap != NULL) { + ifa = sctp_ifap; + } else { + /* Note for BSD we hit here always other + * O/S's will pass things in via the + * sctp_ifap argument. + */ + ifa = sctp_find_ifa_by_addr(&store.sa, + vrf_id, SCTP_ADDR_NOT_LOCKED); + } + if (ifa == NULL) { + error = EADDRNOTAVAIL; + /* Can't find an interface with that address */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } +#ifdef INET6 + if (addr->sa_family == AF_INET6) { + /* GAK, more FIXME IFA lock? */ + if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { + /* Can't bind a non-existent addr. */ + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); + goto out; + } + } +#endif + /* we're not bound all */ + inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL; + /* allow bindx() to send ASCONF's for binding changes */ + sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF); + /* clear automatic addr changes from kernel flag */ + sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); + + /* add this address to the endpoint list */ + error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0); + if (error != 0) + goto out; + inp->laddr_count++; + } + /* find the bucket */ + if (port_reuse_active) { + /* Put it into tcp 1-2-1 hash */ + head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))]; + inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL; + } else { + head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))]; + } + /* put it in the bucket */ + LIST_INSERT_HEAD(head, inp, sctp_hash); + SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n", + (void *)head, ntohs(lport), port_reuse_active); + /* set in the port */ + inp->sctp_lport = lport; + + /* turn off just the unbound flag */ + KASSERT((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) != 0, + ("%s: inp %p is already bound", __func__, inp)); + inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND; +out: + return (error); +} + +int +#if defined(__FreeBSD__) && !defined(__Userspace__) +sctp_inpcb_bind(struct socket *so, struct sockaddr *addr, + struct sctp_ifa *sctp_ifap, struct thread *td) +#elif defined(_WIN32) && !defined(__Userspace__) +sctp_inpcb_bind(struct socket *so, struct sockaddr *addr, + struct sctp_ifa *sctp_ifap, PKTHREAD p) +#else +sctp_inpcb_bind(struct socket *so, struct sockaddr *addr, + struct sctp_ifa *sctp_ifap, struct proc *p) +#endif +{ + struct sctp_inpcb *inp; + int error; + + inp = so->so_pcb; + SCTP_INP_INFO_WLOCK(); + SCTP_INP_WLOCK(inp); +#if defined(__FreeBSD__) && !defined(__Userspace__) + error = sctp_inpcb_bind_locked(inp, addr, sctp_ifap, td); +#else + error = sctp_inpcb_bind_locked(inp, addr, sctp_ifap, p); +#endif + SCTP_INP_WUNLOCK(inp); + SCTP_INP_INFO_WUNLOCK(); + return (error); +} + +static void +sctp_iterator_inp_being_freed(struct sctp_inpcb *inp) +{ + struct sctp_iterator *it, *nit; + + /* + * We enter with the only the ITERATOR_LOCK in place and a write + * lock on the inp_info stuff. + */ + it = sctp_it_ctl.cur_it; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (it && (it->vn != curvnet)) { + /* Its not looking at our VNET */ + return; + } +#endif + if (it && (it->inp == inp)) { + /* + * This is tricky and we hold the iterator lock, + * but when it returns and gets the lock (when we + * release it) the iterator will try to operate on + * inp. We need to stop that from happening. But + * of course the iterator has a reference on the + * stcb and inp. We can mark it and it will stop. + * + * If its a single iterator situation, we + * set the end iterator flag. Otherwise + * we set the iterator to go to the next inp. + * + */ + if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { + sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT; + } else { + sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP; + } + } + /* Now go through and remove any single reference to + * our inp that may be still pending on the list + */ + SCTP_IPI_ITERATOR_WQ_LOCK(); + TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (it->vn != curvnet) { + continue; + } +#endif + if (it->inp == inp) { + /* This one points to me is it inp specific? */ + if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { + /* Remove and free this one */ + TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, + it, sctp_nxt_itr); + if (it->function_atend != NULL) { + (*it->function_atend) (it->pointer, it->val); + } + SCTP_FREE(it, SCTP_M_ITER); + } else { + it->inp = LIST_NEXT(it->inp, sctp_list); + if (it->inp) { + SCTP_INP_INCR_REF(it->inp); + } + } + /* When its put in the refcnt is incremented so decr it */ + SCTP_INP_DECR_REF(inp); + } + } + SCTP_IPI_ITERATOR_WQ_UNLOCK(); +} + +/* release sctp_inpcb unbind the port */ +void +sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from) +{ + /* + * Here we free a endpoint. We must find it (if it is in the Hash + * table) and remove it from there. Then we must also find it in the + * overall list and remove it from there. After all removals are + * complete then any timer has to be stopped. Then start the actual + * freeing. a) Any local lists. b) Any associations. c) The hash of + * all associations. d) finally the ep itself. + */ + struct sctp_tcb *stcb, *nstcb; + struct sctp_laddr *laddr, *nladdr; + struct inpcb *ip_pcb; + struct socket *so; + int being_refed = 0; + struct sctp_queued_to_read *sq, *nsq; +#if !defined(__Userspace__) +#if !defined(__FreeBSD__) + sctp_rtentry_t *rt; +#endif +#endif + int cnt; + sctp_sharedkey_t *shared_key, *nshared_key; + +#if defined(__APPLE__) && !defined(__Userspace__) + sctp_lock_assert(SCTP_INP_SO(inp)); +#endif +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 0); +#endif + SCTP_ITERATOR_LOCK(); + /* mark any iterators on the list or being processed */ + sctp_iterator_inp_being_freed(inp); + SCTP_ITERATOR_UNLOCK(); + + SCTP_ASOC_CREATE_LOCK(inp); + SCTP_INP_INFO_WLOCK(); + SCTP_INP_WLOCK(inp); + so = inp->sctp_socket; + KASSERT((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) != 0, + ("%s: inp %p still has socket", __func__, inp)); + KASSERT((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) == 0, + ("%s: double free of inp %p", __func__, inp)); + if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) { + inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP; + /* socket is gone, so no more wakeups allowed */ + inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE; + inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT; + inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT; + } + /* First time through we have the socket lock, after that no more. */ + sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL, + SCTP_FROM_SCTP_PCB + SCTP_LOC_1); + + if (inp->control) { + sctp_m_freem(inp->control); + inp->control = NULL; + } + if (inp->pkt) { + sctp_m_freem(inp->pkt); + inp->pkt = NULL; + } + ip_pcb = &inp->ip_inp.inp; /* we could just cast the main pointer + * here but I will be nice :> (i.e. + * ip_pcb = ep;) */ + if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) { + int cnt_in_sd; + + cnt_in_sd = 0; + LIST_FOREACH_SAFE(stcb, &inp->sctp_asoc_list, sctp_tcblist, nstcb) { + SCTP_TCB_LOCK(stcb); + /* Disconnect the socket please. */ + stcb->sctp_socket = NULL; + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_CLOSED_SOCKET); + if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + /* Skip guys being freed */ + cnt_in_sd++; + if (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) { + /* + * Special case - we did not start a kill + * timer on the asoc due to it was not + * closed. So go ahead and start it now. + */ + SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE); + sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); + } + SCTP_TCB_UNLOCK(stcb); + continue; + } + if (((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) && + (stcb->asoc.total_output_queue_size == 0)) { + /* If we have data in queue, we don't want to just + * free since the app may have done, send()/close + * or connect/send/close. And it wants the data + * to get across first. + */ + /* Just abandon things in the front states */ + if (sctp_free_assoc(inp, stcb, SCTP_PCBFREE_NOFORCE, + SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) { + cnt_in_sd++; + } + continue; + } + if ((stcb->asoc.size_on_reasm_queue > 0) || + (stcb->asoc.control_pdapi) || + (stcb->asoc.size_on_all_streams > 0) || + ((so != NULL) && (SCTP_SBAVAIL(&so->so_rcv) > 0))) { + /* Left with Data unread */ + struct mbuf *op_err; + + op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3; + sctp_send_abort_tcb(stcb, op_err, SCTP_SO_LOCKED); + SCTP_STAT_INCR_COUNTER32(sctps_aborted); + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + if (sctp_free_assoc(inp, stcb, + SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) { + cnt_in_sd++; + } + continue; + } else if (TAILQ_EMPTY(&stcb->asoc.send_queue) && + TAILQ_EMPTY(&stcb->asoc.sent_queue) && + (stcb->asoc.stream_queue_cnt == 0)) { + if ((*stcb->asoc.ss_functions.sctp_ss_is_user_msgs_incomplete)(stcb, &stcb->asoc)) { + goto abort_anyway; + } + if ((SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { + struct sctp_nets *netp; + + /* + * there is nothing queued to send, + * so I send shutdown + */ + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_SENT); + sctp_stop_timers_for_shutdown(stcb); + if (stcb->asoc.alternate) { + netp = stcb->asoc.alternate; + } else { + netp = stcb->asoc.primary_destination; + } + sctp_send_shutdown(stcb, netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, + netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, NULL); + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED); + } + } else { + /* mark into shutdown pending */ + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_SHUTDOWN_PENDING); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, NULL); + if ((*stcb->asoc.ss_functions.sctp_ss_is_user_msgs_incomplete)(stcb, &stcb->asoc)) { + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_PARTIAL_MSG_LEFT); + } + if (TAILQ_EMPTY(&stcb->asoc.send_queue) && + TAILQ_EMPTY(&stcb->asoc.sent_queue) && + (stcb->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) { + struct mbuf *op_err; + abort_anyway: + op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5; + sctp_send_abort_tcb(stcb, op_err, SCTP_SO_LOCKED); + SCTP_STAT_INCR_COUNTER32(sctps_aborted); + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + if (sctp_free_assoc(inp, stcb, + SCTP_PCBFREE_NOFORCE, + SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) { + cnt_in_sd++; + } + continue; + } else { + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED); + } + } + cnt_in_sd++; + SCTP_TCB_UNLOCK(stcb); + } + /* now is there some left in our SHUTDOWN state? */ + if (cnt_in_sd) { +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 2); +#endif + inp->sctp_socket = NULL; + SCTP_INP_WUNLOCK(inp); + SCTP_ASOC_CREATE_UNLOCK(inp); + SCTP_INP_INFO_WUNLOCK(); + return; + } + } + inp->sctp_socket = NULL; + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) { + /* + * ok, this guy has been bound. It's port is + * somewhere in the SCTP_BASE_INFO(hash table). Remove + * it! + */ + LIST_REMOVE(inp, sctp_hash); + inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND; + } + + /* If there is a timer running to kill us, + * forget it, since it may have a contest + * on the INP lock.. which would cause us + * to die ... + */ + cnt = 0; + LIST_FOREACH_SAFE(stcb, &inp->sctp_asoc_list, sctp_tcblist, nstcb) { + SCTP_TCB_LOCK(stcb); + if (immediate != SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) { + /* Disconnect the socket please */ + stcb->sctp_socket = NULL; + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_CLOSED_SOCKET); + } + if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + if (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) { + SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE); + sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); + } + cnt++; + SCTP_TCB_UNLOCK(stcb); + continue; + } + /* Free associations that are NOT killing us */ + if ((SCTP_GET_STATE(stcb) != SCTP_STATE_COOKIE_WAIT) && + ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) { + struct mbuf *op_err; + + op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7; + sctp_send_abort_tcb(stcb, op_err, SCTP_SO_LOCKED); + SCTP_STAT_INCR_COUNTER32(sctps_aborted); + } else if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + cnt++; + SCTP_TCB_UNLOCK(stcb); + continue; + } + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + if (sctp_free_assoc(inp, stcb, SCTP_PCBFREE_FORCE, + SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) { + cnt++; + } + } + if (cnt) { + /* Ok we have someone out there that will kill us */ +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 3); +#endif + SCTP_INP_WUNLOCK(inp); + SCTP_ASOC_CREATE_UNLOCK(inp); + SCTP_INP_INFO_WUNLOCK(); + return; + } + if (SCTP_INP_LOCK_CONTENDED(inp)) + being_refed++; + if (SCTP_INP_READ_CONTENDED(inp)) + being_refed++; + if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp)) + being_refed++; + /* NOTE: 0 refcount also means no timers are referencing us. */ + if ((inp->refcount) || + (being_refed) || + (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) { +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 4); +#endif + sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL); + SCTP_INP_WUNLOCK(inp); + SCTP_ASOC_CREATE_UNLOCK(inp); + SCTP_INP_INFO_WUNLOCK(); + return; + } + inp->sctp_ep.signature_change.type = 0; + inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE; + /* Remove it from the list .. last thing we need a + * lock for. + */ + LIST_REMOVE(inp, sctp_list); + SCTP_INP_WUNLOCK(inp); + SCTP_ASOC_CREATE_UNLOCK(inp); + SCTP_INP_INFO_WUNLOCK(); + +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 5); +#endif +#if !(defined(_WIN32) || defined(__Userspace__)) +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + rt = ip_pcb->inp_route.ro_rt; +#endif +#endif + if ((inp->sctp_asocidhash) != NULL) { + SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark); + inp->sctp_asocidhash = NULL; + } + /*sa_ignore FREED_MEMORY*/ + TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) { + /* Its only abandoned if it had data left */ + if (sq->length) + SCTP_STAT_INCR(sctps_left_abandon); + + TAILQ_REMOVE(&inp->read_queue, sq, next); + sctp_free_remote_addr(sq->whoFrom); + if (so) + so->so_rcv.sb_cc -= sq->length; + if (sq->data) { + sctp_m_freem(sq->data); + sq->data = NULL; + } + /* + * no need to free the net count, since at this point all + * assoc's are gone. + */ + sctp_free_a_readq(NULL, sq); + } + /* Now the sctp_pcb things */ + /* + * free each asoc if it is not already closed/free. we can't use the + * macro here since le_next will get freed as part of the + * sctp_free_assoc() call. + */ + if (ip_pcb->inp_options) { + (void)sctp_m_free(ip_pcb->inp_options); + ip_pcb->inp_options = 0; + } +#if !(defined(_WIN32) || defined(__Userspace__)) +#if !defined(__FreeBSD__) + if (rt) { + RTFREE(rt); + ip_pcb->inp_route.ro_rt = 0; + } +#endif +#endif +#ifdef INET6 +#if !(defined(_WIN32) || defined(__Userspace__)) +#if (defined(__FreeBSD__) || defined(__APPLE__) && !defined(__Userspace__)) + if (ip_pcb->inp_vflag & INP_IPV6) { +#else + if (inp->inp_vflag & INP_IPV6) { +#endif + ip6_freepcbopts(ip_pcb->in6p_outputopts); + } +#endif +#endif /* INET6 */ + ip_pcb->inp_vflag = 0; + /* free up authentication fields */ + if (inp->sctp_ep.local_auth_chunks != NULL) + sctp_free_chunklist(inp->sctp_ep.local_auth_chunks); + if (inp->sctp_ep.local_hmacs != NULL) + sctp_free_hmaclist(inp->sctp_ep.local_hmacs); + + LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) { + LIST_REMOVE(shared_key, next); + sctp_free_sharedkey(shared_key); + /*sa_ignore FREED_MEMORY*/ + } + +#if defined(__APPLE__) && !defined(__Userspace__) + inp->ip_inp.inp.inp_state = INPCB_STATE_DEAD; + if (in_pcb_checkstate(&inp->ip_inp.inp, WNT_STOPUSING, 1) != WNT_STOPUSING) { +#ifdef INVARIANTS + panic("sctp_inpcb_free inp = %p couldn't set to STOPUSING", (void *)inp); +#else + SCTP_PRINTF("sctp_inpcb_free inp = %p couldn't set to STOPUSING\n", (void *)inp); +#endif + } + inp->ip_inp.inp.inp_socket->so_flags |= SOF_PCBCLEARING; +#endif + /* + * if we have an address list the following will free the list of + * ifaddr's that are set into this ep. Again macro limitations here, + * since the LIST_FOREACH could be a bad idea. + */ + LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) { + sctp_remove_laddr(laddr); + } + +#ifdef SCTP_TRACK_FREED_ASOCS + /* TEMP CODE */ + LIST_FOREACH_SAFE(stcb, &inp->sctp_asoc_free_list, sctp_tcblist, nstcb) { + LIST_REMOVE(stcb, sctp_tcblist); + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); + SCTP_DECR_ASOC_COUNT(); + } + /* *** END TEMP CODE ****/ +#endif +#ifdef SCTP_MVRF + SCTP_FREE(inp->m_vrf_ids, SCTP_M_MVRF); +#endif + /* Now lets see about freeing the EP hash table. */ + if (inp->sctp_tcbhash != NULL) { + SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark); + inp->sctp_tcbhash = NULL; + } + /* Now we must put the ep memory back into the zone pool */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + crfree(inp->ip_inp.inp.inp_cred); + INP_LOCK_DESTROY(&inp->ip_inp.inp); +#endif + SCTP_INP_LOCK_DESTROY(inp); + SCTP_INP_READ_DESTROY(inp); + SCTP_ASOC_CREATE_LOCK_DESTROY(inp); +#if !(defined(__APPLE__) && !defined(__Userspace__)) + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); + SCTP_DECR_EP_COUNT(); +#else + /* For Tiger, we will do this later... */ +#endif +} + +struct sctp_nets * +sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr) +{ + struct sctp_nets *net; + /* locate the address */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr)) + return (net); + } + return (NULL); +} + +int +sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id) +{ + struct sctp_ifa *sctp_ifa; + sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED); + if (sctp_ifa) { + return (1); + } else { + return (0); + } +} + +/* + * add's a remote endpoint address, done with the INIT/INIT-ACK as well as + * when a ASCONF arrives that adds it. It will also initialize all the cwnd + * stats of stuff. + */ +int +sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr, + struct sctp_nets **netp, uint16_t port, int set_scope, int from) +{ + /* + * The following is redundant to the same lines in the + * sctp_aloc_assoc() but is needed since others call the add + * address function + */ + struct sctp_nets *net, *netfirst; + int addr_inscope; + + SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ", + from); + SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr); + + netfirst = sctp_findnet(stcb, newaddr); + if (netfirst) { + /* + * Lie and return ok, we don't want to make the association + * go away for this behavior. It will happen in the TCP + * model in a connected socket. It does not reach the hash + * table until after the association is built so it can't be + * found. Mark as reachable, since the initial creation will + * have been cleared and the NOT_IN_ASSOC flag will have + * been added... and we don't want to end up removing it + * back out. + */ + if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) { + netfirst->dest_state = (SCTP_ADDR_REACHABLE | + SCTP_ADDR_UNCONFIRMED); + } else { + netfirst->dest_state = SCTP_ADDR_REACHABLE; + } + + return (0); + } + addr_inscope = 1; + switch (newaddr->sa_family) { +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin; + + sin = (struct sockaddr_in *)newaddr; + if (sin->sin_addr.s_addr == 0) { + /* Invalid address */ + return (-1); + } + /* zero out the zero area */ + memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); + + /* assure len is set */ +#ifdef HAVE_SIN_LEN + sin->sin_len = sizeof(struct sockaddr_in); +#endif + if (set_scope) { + if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { + stcb->asoc.scope.ipv4_local_scope = 1; + } + } else { + /* Validate the address is in scope */ + if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) && + (stcb->asoc.scope.ipv4_local_scope == 0)) { + addr_inscope = 0; + } + } + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)newaddr; + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + /* Invalid address */ + return (-1); + } + /* assure len is set */ +#ifdef HAVE_SIN6_LEN + sin6->sin6_len = sizeof(struct sockaddr_in6); +#endif + if (set_scope) { + if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) { + stcb->asoc.scope.loopback_scope = 1; + stcb->asoc.scope.local_scope = 0; + stcb->asoc.scope.ipv4_local_scope = 1; + stcb->asoc.scope.site_scope = 1; + } else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { + /* + * If the new destination is a LINK_LOCAL we + * must have common site scope. Don't set + * the local scope since we may not share + * all links, only loopback can do this. + * Links on the local network would also be + * on our private network for v4 too. + */ + stcb->asoc.scope.ipv4_local_scope = 1; + stcb->asoc.scope.site_scope = 1; + } else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) { + /* + * If the new destination is SITE_LOCAL then + * we must have site scope in common. + */ + stcb->asoc.scope.site_scope = 1; + } + } else { + /* Validate the address is in scope */ + if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) && + (stcb->asoc.scope.loopback_scope == 0)) { + addr_inscope = 0; + } else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) && + (stcb->asoc.scope.local_scope == 0)) { + addr_inscope = 0; + } else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) && + (stcb->asoc.scope.site_scope == 0)) { + addr_inscope = 0; + } + } + break; + } +#endif +#if defined(__Userspace__) + case AF_CONN: + { + struct sockaddr_conn *sconn; + + sconn = (struct sockaddr_conn *)newaddr; + if (sconn->sconn_addr == NULL) { + /* Invalid address */ + return (-1); + } +#ifdef HAVE_SCONN_LEN + sconn->sconn_len = sizeof(struct sockaddr_conn); +#endif + break; + } +#endif + default: + /* not supported family type */ + return (-1); + } + net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets); + if (net == NULL) { + return (-1); + } + SCTP_INCR_RADDR_COUNT(); + memset(net, 0, sizeof(struct sctp_nets)); + (void)SCTP_GETTIME_TIMEVAL(&net->start_time); +#ifdef HAVE_SA_LEN + memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len); +#endif + switch (newaddr->sa_family) { +#ifdef INET + case AF_INET: +#ifndef HAVE_SA_LEN + memcpy(&net->ro._l_addr, newaddr, sizeof(struct sockaddr_in)); +#endif + ((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport; + break; +#endif +#ifdef INET6 + case AF_INET6: +#ifndef HAVE_SA_LEN + memcpy(&net->ro._l_addr, newaddr, sizeof(struct sockaddr_in6)); +#endif + ((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: +#ifndef HAVE_SA_LEN + memcpy(&net->ro._l_addr, newaddr, sizeof(struct sockaddr_conn)); +#endif + ((struct sockaddr_conn *)&net->ro._l_addr)->sconn_port = stcb->rport; + break; +#endif + default: + break; + } + net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id); + if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) { + stcb->asoc.scope.loopback_scope = 1; + stcb->asoc.scope.ipv4_local_scope = 1; + stcb->asoc.scope.local_scope = 0; + stcb->asoc.scope.site_scope = 1; + addr_inscope = 1; + } + net->failure_threshold = stcb->asoc.def_net_failure; + net->pf_threshold = stcb->asoc.def_net_pf_threshold; + if (addr_inscope == 0) { + net->dest_state = (SCTP_ADDR_REACHABLE | + SCTP_ADDR_OUT_OF_SCOPE); + } else { + if (from == SCTP_ADDR_IS_CONFIRMED) + /* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */ + net->dest_state = SCTP_ADDR_REACHABLE; + else + net->dest_state = SCTP_ADDR_REACHABLE | + SCTP_ADDR_UNCONFIRMED; + } + /* We set this to 0, the timer code knows that + * this means its an initial value + */ + net->rto_needed = 1; + net->RTO = 0; + net->RTO_measured = 0; + stcb->asoc.numnets++; + net->ref_count = 1; + net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1; + net->port = port; + net->dscp = stcb->asoc.default_dscp; +#ifdef INET6 + net->flowlabel = stcb->asoc.default_flowlabel; +#endif + if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) { + net->dest_state |= SCTP_ADDR_NOHB; + } else { + net->dest_state &= ~SCTP_ADDR_NOHB; + } + if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) { + net->dest_state |= SCTP_ADDR_NO_PMTUD; + } else { + net->dest_state &= ~SCTP_ADDR_NO_PMTUD; + } + net->heart_beat_delay = stcb->asoc.heart_beat_delay; + /* Init the timer structure */ + SCTP_OS_TIMER_INIT(&net->rxt_timer.timer); + SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer); + SCTP_OS_TIMER_INIT(&net->hb_timer.timer); + + /* Now generate a route for this guy */ +#ifdef INET6 +#ifdef SCTP_EMBEDDED_V6_SCOPE + /* KAME hack: embed scopeid */ + if (newaddr->sa_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) + (void)in6_embedscope(&sin6->sin6_addr, sin6, &stcb->sctp_ep->ip_inp.inp, NULL); +#else + (void)in6_embedscope(&sin6->sin6_addr, sin6, &stcb->sctp_ep->ip_inp.inp, NULL, NULL); +#endif +#elif defined(SCTP_KAME) + (void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)); +#else + (void)in6_embedscope(&sin6->sin6_addr, sin6); +#endif +#ifndef SCOPEDROUTING + sin6->sin6_scope_id = 0; +#endif + } +#endif /* SCTP_EMBEDDED_V6_SCOPE */ +#endif + SCTP_RTALLOC((sctp_route_t *)&net->ro, + stcb->asoc.vrf_id, + stcb->sctp_ep->fibnum); + + net->src_addr_selected = 0; +#if !defined(__Userspace__) + if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) { + /* Get source address */ + net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep, + stcb, + (sctp_route_t *)&net->ro, + net, + 0, + stcb->asoc.vrf_id); + if (stcb->asoc.default_mtu > 0) { + net->mtu = stcb->asoc.default_mtu; + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + net->mtu += SCTP_MIN_V4_OVERHEAD; + break; +#endif +#ifdef INET6 + case AF_INET6: + net->mtu += SCTP_MIN_OVERHEAD; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + net->mtu += sizeof(struct sctphdr); + break; +#endif + default: + break; + } +#if defined(INET) || defined(INET6) + if (net->port) { + net->mtu += (uint32_t)sizeof(struct udphdr); + } +#endif + } else if (net->ro._s_addr != NULL) { + uint32_t imtu, rmtu, hcmtu; + + net->src_addr_selected = 1; + /* Now get the interface MTU */ + if (net->ro._s_addr->ifn_p != NULL) { + /* + * XXX: Should we here just use + * net->ro._s_addr->ifn_p->ifn_mtu + */ + imtu = SCTP_GATHER_MTU_FROM_IFN_INFO(net->ro._s_addr->ifn_p->ifn_p, + net->ro._s_addr->ifn_p->ifn_index); + } else { + imtu = 0; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_nh); + hcmtu = sctp_hc_get_mtu(&net->ro._l_addr, stcb->sctp_ep->fibnum); +#else + rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt); + hcmtu = 0; +#endif + net->mtu = sctp_min_mtu(hcmtu, rmtu, imtu); +#if defined(__FreeBSD__) && !defined(__Userspace__) +#else + if (rmtu == 0) { + /* Start things off to match mtu of interface please. */ + SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa, + net->ro.ro_rt, net->mtu); + } +#endif + } + } +#endif + if (net->mtu == 0) { + if (stcb->asoc.default_mtu > 0) { + net->mtu = stcb->asoc.default_mtu; + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + net->mtu += SCTP_MIN_V4_OVERHEAD; + break; +#endif +#ifdef INET6 + case AF_INET6: + net->mtu += SCTP_MIN_OVERHEAD; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + net->mtu += sizeof(struct sctphdr); + break; +#endif + default: + break; + } +#if defined(INET) || defined(INET6) + if (net->port) { + net->mtu += (uint32_t)sizeof(struct udphdr); + } +#endif + } else { + switch (newaddr->sa_family) { +#ifdef INET + case AF_INET: + net->mtu = SCTP_DEFAULT_MTU; + break; +#endif +#ifdef INET6 + case AF_INET6: + net->mtu = 1280; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + net->mtu = 1280; + break; +#endif + default: + break; + } + } + } +#if defined(INET) || defined(INET6) + if (net->port) { + net->mtu -= (uint32_t)sizeof(struct udphdr); + } +#endif + if (from == SCTP_ALLOC_ASOC) { + stcb->asoc.smallest_mtu = net->mtu; + } + if (stcb->asoc.smallest_mtu > net->mtu) { + sctp_pathmtu_adjustment(stcb, net->mtu, true); + } +#ifdef INET6 +#ifdef SCTP_EMBEDDED_V6_SCOPE + if (newaddr->sa_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; +#ifdef SCTP_KAME + (void)sa6_recoverscope(sin6); +#else + (void)in6_recoverscope(sin6, &sin6->sin6_addr, NULL); +#endif /* SCTP_KAME */ + } +#endif /* SCTP_EMBEDDED_V6_SCOPE */ +#endif + + /* JRS - Use the congestion control given in the CC module */ + if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL) + (*stcb->asoc.cc_functions.sctp_set_initial_cc_param)(stcb, net); + + /* + * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning + * of assoc (2005/06/27, iyengar@cis.udel.edu) + */ + net->find_pseudo_cumack = 1; + net->find_rtx_pseudo_cumack = 1; +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* Choose an initial flowid. */ + net->flowid = stcb->asoc.my_vtag ^ + ntohs(stcb->rport) ^ + ntohs(stcb->sctp_ep->sctp_lport); + net->flowtype = M_HASHTYPE_OPAQUE_HASH; +#endif + if (netp) { + *netp = net; + } + netfirst = TAILQ_FIRST(&stcb->asoc.nets); +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (net->ro.ro_nh == NULL) { +#else + if (net->ro.ro_rt == NULL) { +#endif + /* Since we have no route put it at the back */ + TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next); + } else if (netfirst == NULL) { + /* We are the first one in the pool. */ + TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); +#if defined(__FreeBSD__) && !defined(__Userspace__) + } else if (netfirst->ro.ro_nh == NULL) { +#else + } else if (netfirst->ro.ro_rt == NULL) { +#endif + /* + * First one has NO route. Place this one ahead of the first + * one. + */ + TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); +#if defined(__FreeBSD__) && !defined(__Userspace__) + } else if (net->ro.ro_nh->nh_ifp != netfirst->ro.ro_nh->nh_ifp) { +#else + } else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) { +#endif + /* + * This one has a different interface than the one at the + * top of the list. Place it ahead. + */ + TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); + } else { + /* + * Ok we have the same interface as the first one. Move + * forward until we find either a) one with a NULL route... + * insert ahead of that b) one with a different ifp.. insert + * after that. c) end of the list.. insert at the tail. + */ + struct sctp_nets *netlook; + + do { + netlook = TAILQ_NEXT(netfirst, sctp_next); + if (netlook == NULL) { + /* End of the list */ + TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next); + break; +#if defined(__FreeBSD__) && !defined(__Userspace__) + } else if (netlook->ro.ro_nh == NULL) { +#else + } else if (netlook->ro.ro_rt == NULL) { +#endif + /* next one has NO route */ + TAILQ_INSERT_BEFORE(netfirst, net, sctp_next); + break; +#if defined(__FreeBSD__) && !defined(__Userspace__) + } else if (netlook->ro.ro_nh->nh_ifp != net->ro.ro_nh->nh_ifp) { +#else + } else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) { +#endif + TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook, + net, sctp_next); + break; + } + /* Shift forward */ + netfirst = netlook; + } while (netlook != NULL); + } + + /* got to have a primary set */ + if (stcb->asoc.primary_destination == 0) { + stcb->asoc.primary_destination = net; +#if defined(__FreeBSD__) && !defined(__Userspace__) + } else if ((stcb->asoc.primary_destination->ro.ro_nh == NULL) && + (net->ro.ro_nh) && +#else + } else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) && + (net->ro.ro_rt) && +#endif + ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) { + /* No route to current primary adopt new primary */ + stcb->asoc.primary_destination = net; + } + /* Validate primary is first */ + net = TAILQ_FIRST(&stcb->asoc.nets); + if ((net != stcb->asoc.primary_destination) && + (stcb->asoc.primary_destination)) { + /* first one on the list is NOT the primary + * sctp_cmpaddr() is much more efficient if + * the primary is the first on the list, make it + * so. + */ + TAILQ_REMOVE(&stcb->asoc.nets, + stcb->asoc.primary_destination, sctp_next); + TAILQ_INSERT_HEAD(&stcb->asoc.nets, + stcb->asoc.primary_destination, sctp_next); + } + return (0); +} + +static uint32_t +sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb) +{ + uint32_t id; + struct sctpasochead *head; + struct sctp_tcb *lstcb; + + try_again: + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { + /* TSNH */ + return (0); + } + /* + * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC, + * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC. + */ + if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) { + inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1; + } + id = inp->sctp_associd_counter; + inp->sctp_associd_counter++; + lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t)id, 0); + if (lstcb) { + goto try_again; + } + head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)]; + LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash); + stcb->asoc.in_asocid_hash = 1; + return (id); +} + +/* + * allocate an association and add it to the endpoint. The caller must be + * careful to add all additional addresses once they are know right away or + * else the assoc will be may experience a blackout scenario. + */ +static struct sctp_tcb * +sctp_aloc_assoc_locked(struct sctp_inpcb *inp, struct sockaddr *firstaddr, + int *error, uint32_t override_tag, uint32_t initial_tsn, + uint32_t vrf_id, uint16_t o_streams, uint16_t port, +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct thread *p, +#elif defined(_WIN32) && !defined(__Userspace__) + PKTHREAD p, +#else +#if defined(__Userspace__) + /* __Userspace__ NULL proc is going to be passed here. See sctp_lower_sosend */ +#endif + struct proc *p, +#endif + int initialize_auth_params) +{ + /* note the p argument is only valid in unbound sockets */ + + struct sctp_tcb *stcb; + struct sctp_association *asoc; + struct sctpasochead *head; + uint16_t rport; + int err; + + SCTP_INP_INFO_WLOCK_ASSERT(); + SCTP_INP_WLOCK_ASSERT(inp); + + /* + * Assumption made here: Caller has done a + * sctp_findassociation_ep_addr(ep, addr's); to make sure the + * address does not exist already. + */ + if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) { + /* Hit max assoc, sorry no more */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); + *error = ENOBUFS; + return (NULL); + } + if (firstaddr == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); + *error = EINVAL; + return (NULL); + } + if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); + *error = EINVAL; + return (NULL); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) && + ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) || + (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) { + /* + * If its in the TCP pool, its NOT allowed to create an + * association. The parent listener needs to call + * sctp_aloc_assoc.. or the one-2-many socket. If a peeled + * off, or connected one does this.. its an error. + */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); + *error = EINVAL; + return (NULL); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) || + (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); + *error = EINVAL; + return (NULL); + } + } + SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:"); +#ifdef SCTP_DEBUG + if (firstaddr) { + SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr); + switch (firstaddr->sa_family) { +#ifdef INET + case AF_INET: + SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n", + ntohs(((struct sockaddr_in *)firstaddr)->sin_port)); + break; +#endif +#ifdef INET6 + case AF_INET6: + SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n", + ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port)); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n", + ntohs(((struct sockaddr_conn *)firstaddr)->sconn_port)); + break; +#endif + default: + break; + } + } else { + SCTPDBG(SCTP_DEBUG_PCB3,"None\n"); + } +#endif /* SCTP_DEBUG */ + switch (firstaddr->sa_family) { +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin; + + sin = (struct sockaddr_in *)firstaddr; + if ((ntohs(sin->sin_port) == 0) || + (sin->sin_addr.s_addr == INADDR_ANY) || + (sin->sin_addr.s_addr == INADDR_BROADCAST) || + IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) || +#if defined(__Userspace__) + ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_CONN) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + (SCTP_IPV6_V6ONLY(inp) != 0)))) { +#else + ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + (SCTP_IPV6_V6ONLY(inp) != 0))) { +#endif + /* Invalid address */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); + *error = EINVAL; + return (NULL); + } + rport = sin->sin_port; + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)firstaddr; + if ((ntohs(sin6->sin6_port) == 0) || + IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) || + IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0)) { + /* Invalid address */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); + *error = EINVAL; + return (NULL); + } + rport = sin6->sin6_port; + break; + } +#endif +#if defined(__Userspace__) + case AF_CONN: + { + struct sockaddr_conn *sconn; + + sconn = (struct sockaddr_conn *)firstaddr; + if ((ntohs(sconn->sconn_port) == 0) || + (sconn->sconn_addr == NULL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_CONN) == 0)) { + /* Invalid address */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); + *error = EINVAL; + return (NULL); + } + rport = sconn->sconn_port; + break; + } +#endif + default: + /* not supported family type */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); + *error = EINVAL; + return (NULL); + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { + /* + * If you have not performed a bind, then we need to do the + * ephemeral bind for you. + */ + if ((err = sctp_inpcb_bind_locked(inp, NULL, NULL, p))) { + /* bind error, probably perm */ + *error = err; + return (NULL); + } + } + stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb); + if (stcb == NULL) { + /* out of memory? */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM); + *error = ENOMEM; + return (NULL); + } + SCTP_INCR_ASOC_COUNT(); + + memset(stcb, 0, sizeof(*stcb)); + asoc = &stcb->asoc; + + SCTP_TCB_LOCK_INIT(stcb); + stcb->rport = rport; + /* setup back pointer's */ + stcb->sctp_ep = inp; + stcb->sctp_socket = inp->sctp_socket; + if ((err = sctp_init_asoc(inp, stcb, override_tag, initial_tsn, vrf_id, o_streams))) { + /* failed */ + SCTP_TCB_LOCK_DESTROY(stcb); + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); + SCTP_DECR_ASOC_COUNT(); + *error = err; + return (NULL); + } + SCTP_TCB_LOCK(stcb); + + asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb); + /* now that my_vtag is set, add it to the hash */ + head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))]; + /* put it in the bucket in the vtag hash of assoc's for the system */ + LIST_INSERT_HEAD(head, stcb, sctp_asocs); + + if (sctp_add_remote_addr(stcb, firstaddr, NULL, port, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC)) { + /* failure.. memory error? */ + if (asoc->strmout) { + SCTP_FREE(asoc->strmout, SCTP_M_STRMO); + asoc->strmout = NULL; + } + if (asoc->mapping_array) { + SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); + asoc->mapping_array = NULL; + } + if (asoc->nr_mapping_array) { + SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP); + asoc->nr_mapping_array = NULL; + } + SCTP_DECR_ASOC_COUNT(); + SCTP_TCB_UNLOCK(stcb); + SCTP_TCB_LOCK_DESTROY(stcb); + LIST_REMOVE(stcb, sctp_asocs); + LIST_REMOVE(stcb, sctp_tcbasocidhash); + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); + SCTP_INP_WUNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); + *error = ENOBUFS; + return (NULL); + } + /* Init all the timers */ + SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer); + SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer); + SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer); + SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer); + SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer); + SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer); + + LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist); + /* now file the port under the hash as well */ + if (inp->sctp_tcbhash != NULL) { + head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport, + inp->sctp_hashmark)]; + LIST_INSERT_HEAD(head, stcb, sctp_tcbhash); + } + if (initialize_auth_params == SCTP_INITIALIZE_AUTH_PARAMS) { + sctp_initialize_auth_params(inp, stcb); + } + SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb); + return (stcb); +} + +struct sctp_tcb * +sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr, + int *error, uint32_t override_tag, uint32_t initial_tsn, + uint32_t vrf_id, uint16_t o_streams, uint16_t port, +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct thread *p, +#elif defined(_WIN32) && !defined(__Userspace__) + PKTHREAD p, +#else + struct proc *p, +#endif + int initialize_auth_params) +{ + struct sctp_tcb *stcb; + + SCTP_INP_INFO_WLOCK(); + SCTP_INP_WLOCK(inp); + stcb = sctp_aloc_assoc_locked(inp, firstaddr, error, override_tag, + initial_tsn, vrf_id, o_streams, port, p, initialize_auth_params); + SCTP_INP_INFO_WUNLOCK(); + SCTP_INP_WUNLOCK(inp); + return (stcb); +} + +struct sctp_tcb * +sctp_aloc_assoc_connected(struct sctp_inpcb *inp, struct sockaddr *firstaddr, + int *error, uint32_t override_tag, uint32_t initial_tsn, + uint32_t vrf_id, uint16_t o_streams, uint16_t port, +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct thread *p, +#elif defined(_WIN32) && !defined(__Userspace__) + PKTHREAD p, +#else + struct proc *p, +#endif + int initialize_auth_params) +{ + struct sctp_tcb *stcb; + + SCTP_INP_INFO_WLOCK(); + SCTP_INP_WLOCK(inp); + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && + SCTP_IS_LISTENING(inp)) { + SCTP_INP_INFO_WUNLOCK(); + SCTP_INP_WUNLOCK(inp); + *error = EINVAL; + return (NULL); + } + stcb = sctp_aloc_assoc_locked(inp, firstaddr, error, override_tag, + initial_tsn, vrf_id, o_streams, port, p, initialize_auth_params); + SCTP_INP_INFO_WUNLOCK(); + if (stcb != NULL && (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) { + inp->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED; + soisconnecting(inp->sctp_socket); + } + SCTP_INP_WUNLOCK(inp); + return (stcb); +} + +void +sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net) +{ + struct sctp_inpcb *inp; + struct sctp_association *asoc; + + inp = stcb->sctp_ep; + asoc = &stcb->asoc; + asoc->numnets--; + TAILQ_REMOVE(&asoc->nets, net, sctp_next); + if (net == asoc->primary_destination) { + /* Reset primary */ + struct sctp_nets *lnet; + + lnet = TAILQ_FIRST(&asoc->nets); + /* Mobility adaptation + Ideally, if deleted destination is the primary, it becomes + a fast retransmission trigger by the subsequent SET PRIMARY. + (by micchie) + */ + if (sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_BASE) || + sctp_is_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_FASTHANDOFF)) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n"); + if (asoc->deleted_primary != NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n"); + goto out; + } + asoc->deleted_primary = net; + atomic_add_int(&net->ref_count, 1); + memset(&net->lastsa, 0, sizeof(net->lastsa)); + memset(&net->lastsv, 0, sizeof(net->lastsv)); + sctp_mobility_feature_on(stcb->sctp_ep, + SCTP_MOBILITY_PRIM_DELETED); + sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED, + stcb->sctp_ep, stcb, NULL); + } +out: + /* Try to find a confirmed primary */ + asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0); + } + if (net == asoc->last_data_chunk_from) { + /* Reset primary */ + asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets); + } + if (net == asoc->last_control_chunk_from) { + /* Clear net */ + asoc->last_control_chunk_from = NULL; + } + if (net == asoc->last_net_cmt_send_started) { + /* Clear net */ + asoc->last_net_cmt_send_started = NULL; + } + if (net == stcb->asoc.alternate) { + sctp_free_remote_addr(stcb->asoc.alternate); + stcb->asoc.alternate = NULL; + } + sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, + SCTP_FROM_SCTP_PCB + SCTP_LOC_9); + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, + SCTP_FROM_SCTP_PCB + SCTP_LOC_10); + net->dest_state |= SCTP_ADDR_BEING_DELETED; + sctp_free_remote_addr(net); +} + +/* + * remove a remote endpoint address from an association, it will fail if the + * address does not exist. + */ +int +sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr) +{ + /* + * Here we need to remove a remote address. This is quite simple, we + * first find it in the list of address for the association + * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE + * on that item. Note we do not allow it to be removed if there are + * no other addresses. + */ + struct sctp_association *asoc; + struct sctp_nets *net, *nnet; + + asoc = &stcb->asoc; + + /* locate the address */ + TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) { + if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) { + continue; + } + if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr, + remaddr)) { + /* we found the guy */ + if (asoc->numnets < 2) { + /* Must have at LEAST two remote addresses */ + return (-1); + } else { + sctp_remove_net(stcb, net); + return (0); + } + } + } + /* not found. */ + return (-2); +} + +static bool +sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport, uint32_t now) +{ + struct sctpvtaghead *chain; + struct sctp_tagblock *twait_block; + int i; + + SCTP_INP_INFO_LOCK_ASSERT(); + chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; + LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { + for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { + if ((twait_block->vtag_block[i].tv_sec_at_expire >= now) && + (twait_block->vtag_block[i].v_tag == tag) && + (twait_block->vtag_block[i].lport == lport) && + (twait_block->vtag_block[i].rport == rport)) { + return (true); + } + } + } + return (false); +} + +static void +sctp_set_vtag_block(struct sctp_timewait *vtag_block, uint32_t time, + uint32_t tag, uint16_t lport, uint16_t rport) +{ + vtag_block->tv_sec_at_expire = time; + vtag_block->v_tag = tag; + vtag_block->lport = lport; + vtag_block->rport = rport; +} + +static void +sctp_add_vtag_to_timewait(uint32_t tag, uint16_t lport, uint16_t rport) +{ + struct sctpvtaghead *chain; + struct sctp_tagblock *twait_block; + struct timeval now; + uint32_t time; + int i; + bool set; + + SCTP_INP_INFO_WLOCK_ASSERT(); + (void)SCTP_GETTIME_TIMEVAL(&now); + time = (uint32_t)now.tv_sec + SCTP_BASE_SYSCTL(sctp_vtag_time_wait); + chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; + set = false; + LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { + /* Block(s) present, lets find space, and expire on the fly */ + for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { + if ((twait_block->vtag_block[i].v_tag == 0) && !set) { + sctp_set_vtag_block(twait_block->vtag_block + i, time, tag, lport, rport); + set = true; + continue; + } + if ((twait_block->vtag_block[i].v_tag != 0) && + (twait_block->vtag_block[i].tv_sec_at_expire < (uint32_t)now.tv_sec)) { + if (set) { + /* Audit expires this guy */ + sctp_set_vtag_block(twait_block->vtag_block + i, 0, 0, 0, 0); + } else { + /* Reuse it for the new tag */ + sctp_set_vtag_block(twait_block->vtag_block + i, time, tag, lport, rport); + set = true; + } + } + } + if (set) { + /* + * We only do up to the block where we can + * place our tag for audits + */ + break; + } + } + /* Need to add a new block to chain */ + if (!set) { + SCTP_MALLOC(twait_block, struct sctp_tagblock *, + sizeof(struct sctp_tagblock), SCTP_M_TIMW); + if (twait_block == NULL) { + return; + } + memset(twait_block, 0, sizeof(struct sctp_tagblock)); + LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock); + sctp_set_vtag_block(twait_block->vtag_block, time, tag, lport, rport); + } +} + +void +sctp_clean_up_stream(struct sctp_tcb *stcb, struct sctp_readhead *rh) +{ + struct sctp_tmit_chunk *chk, *nchk; + struct sctp_queued_to_read *control, *ncontrol; + + TAILQ_FOREACH_SAFE(control, rh, next_instrm, ncontrol) { + TAILQ_REMOVE(rh, control, next_instrm); + control->on_strm_q = 0; + if (control->on_read_q == 0) { + sctp_free_remote_addr(control->whoFrom); + if (control->data) { + sctp_m_freem(control->data); + control->data = NULL; + } + } + /* Reassembly free? */ + TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) { + TAILQ_REMOVE(&control->reasm, chk, sctp_next); + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + if (chk->holds_key_ref) + sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); + sctp_free_remote_addr(chk->whoTo); + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); + SCTP_DECR_CHK_COUNT(); + /*sa_ignore FREED_MEMORY*/ + } + /* + * We don't free the address here + * since all the net's were freed + * above. + */ + if (control->on_read_q == 0) { + sctp_free_a_readq(stcb, control); + } + } +} + +/*- + * Free the association after un-hashing the remote port. This + * function ALWAYS returns holding NO LOCK on the stcb. It DOES + * expect that the input to this function IS a locked TCB. + * It will return 0, if it did NOT destroy the association (instead + * it unlocks it. It will return NON-zero if it either destroyed the + * association OR the association is already destroyed. + */ +int +sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location) +{ + int i; + struct sctp_association *asoc; + struct sctp_nets *net, *nnet; + struct sctp_laddr *laddr, *naddr; + struct sctp_tmit_chunk *chk, *nchk; + struct sctp_asconf_addr *aparam, *naparam; + struct sctp_asconf_ack *aack, *naack; + struct sctp_stream_reset_list *strrst, *nstrrst; + struct sctp_queued_to_read *sq, *nsq; + struct sctp_stream_queue_pending *sp, *nsp; + sctp_sharedkey_t *shared_key, *nshared_key; + struct socket *so; + + /* first, lets purge the entry from the hash table. */ +#if defined(__APPLE__) && !defined(__Userspace__) + sctp_lock_assert(SCTP_INP_SO(inp)); +#endif + SCTP_TCB_LOCK_ASSERT(stcb); + +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, stcb, 6); +#endif + if (stcb->asoc.state == 0) { +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 7); +#endif + /* there is no asoc, really TSNH :-0 */ + return (1); + } + if (stcb->asoc.alternate) { + sctp_free_remote_addr(stcb->asoc.alternate); + stcb->asoc.alternate = NULL; + } +#if !(defined(__APPLE__) && !defined(__Userspace__)) + /* TEMP CODE */ + if (stcb->freed_from_where == 0) { + /* Only record the first place free happened from */ + stcb->freed_from_where = from_location; + } + /* TEMP CODE */ +#endif + + asoc = &stcb->asoc; + if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) + /* nothing around */ + so = NULL; + else + so = inp->sctp_socket; + + /* + * We used timer based freeing if a reader or writer is in the way. + * So we first check if we are actually being called from a timer, + * if so we abort early if a reader or writer is still in the way. + */ + if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) && + (from_inpcbfree == SCTP_NORMAL_PROC)) { + /* + * is it the timer driving us? if so are the reader/writers + * gone? + */ + if (stcb->asoc.refcnt) { + /* nope, reader or writer in the way */ + sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); + /* no asoc destroyed */ + SCTP_TCB_UNLOCK(stcb); +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, stcb, 8); +#endif + return (0); + } + } + /* Now clean up any other timers */ + sctp_stop_association_timers(stcb, false); + /* Now the read queue needs to be cleaned up (only once) */ + if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) { + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_ABOUT_TO_BE_FREED); + SCTP_INP_READ_LOCK(inp); + TAILQ_FOREACH(sq, &inp->read_queue, next) { + if (sq->stcb == stcb) { + sq->do_not_ref_stcb = 1; + sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn; + /* If there is no end, there never + * will be now. + */ + if (sq->end_added == 0) { + /* Held for PD-API clear that. */ + sq->pdapi_aborted = 1; + sq->held_length = 0; + if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) { + /* + * Need to add a PD-API aborted indication. + * Setting the control_pdapi assures that it will + * be added right after this msg. + */ + uint32_t strseq; + stcb->asoc.control_pdapi = sq; + strseq = (sq->sinfo_stream << 16) | (sq->mid & 0x0000ffff); + sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION, + stcb, + SCTP_PARTIAL_DELIVERY_ABORTED, + (void *)&strseq, + SCTP_SO_LOCKED); + stcb->asoc.control_pdapi = NULL; + } + } + /* Add an end to wake them */ + sq->end_added = 1; + } + } + SCTP_INP_READ_UNLOCK(inp); + if (stcb->block_entry) { + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET); + stcb->block_entry->error = ECONNRESET; + stcb->block_entry = NULL; + } + } + if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) { + /* Someone holds a reference OR the socket is unaccepted yet. + */ + if ((stcb->asoc.refcnt) || + (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { + SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE); + sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) + /* nothing around */ + so = NULL; + if (so) { + /* Wake any reader/writers */ + sctp_sorwakeup(inp, so); + sctp_sowwakeup(inp, so); + } + SCTP_TCB_UNLOCK(stcb); + +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, stcb, 9); +#endif + /* no asoc destroyed */ + return (0); + } +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, stcb, 10); +#endif + /* When I reach here, no others want + * to kill the assoc yet.. and I own + * the lock. Now its possible an abort + * comes in when I do the lock exchange + * below to grab all the locks to do + * the final take out. to prevent this + * we increment the count, which will + * start a timer and blow out above thus + * assuring us that we hold exclusive + * killing of the asoc. Note that + * after getting back the TCB lock + * we will go ahead and increment the + * counter back up and stop any timer + * a passing stranger may have started :-S + */ + if (from_inpcbfree == SCTP_NORMAL_PROC) { + atomic_add_int(&stcb->asoc.refcnt, 1); + + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_INFO_WLOCK(); + SCTP_INP_WLOCK(inp); + SCTP_TCB_LOCK(stcb); + } + /* Double check the GONE flag */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) + /* nothing around */ + so = NULL; + + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { + /* + * For TCP type we need special handling when we are + * connected. We also include the peel'ed off ones to. + */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { + inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED; + inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED; + if (so) { + SOCKBUF_LOCK(&so->so_rcv); + so->so_state &= ~(SS_ISCONNECTING | + SS_ISDISCONNECTING | + SS_ISCONFIRMING | + SS_ISCONNECTED); + so->so_state |= SS_ISDISCONNECTED; +#if defined(__APPLE__) && !defined(__Userspace__) + socantrcvmore(so); +#else + socantrcvmore_locked(so); +#endif + socantsendmore(so); + sctp_sowwakeup(inp, so); + sctp_sorwakeup(inp, so); + SCTP_SOWAKEUP(so); + } + } + } + + /* Make it invalid too, that way if its + * about to run it will abort and return. + */ + /* re-increment the lock */ + if (from_inpcbfree == SCTP_NORMAL_PROC) { + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } + if (stcb->asoc.refcnt) { + SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE); + sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); + if (from_inpcbfree == SCTP_NORMAL_PROC) { + SCTP_INP_INFO_WUNLOCK(); + SCTP_INP_WUNLOCK(inp); + } + SCTP_TCB_UNLOCK(stcb); + return (0); + } + asoc->state = 0; + if (inp->sctp_tcbhash) { + LIST_REMOVE(stcb, sctp_tcbhash); + } + if (stcb->asoc.in_asocid_hash) { + LIST_REMOVE(stcb, sctp_tcbasocidhash); + } + if (inp->sctp_socket == NULL) { + stcb->sctp_socket = NULL; + } + /* Now lets remove it from the list of ALL associations in the EP */ + LIST_REMOVE(stcb, sctp_tcblist); + if (from_inpcbfree == SCTP_NORMAL_PROC) { + SCTP_INP_INCR_REF(inp); + SCTP_INP_WUNLOCK(inp); + } + /* pull from vtag hash */ + LIST_REMOVE(stcb, sctp_asocs); + sctp_add_vtag_to_timewait(asoc->my_vtag, inp->sctp_lport, stcb->rport); + + /* Now restop the timers to be sure + * this is paranoia at is finest! + */ + sctp_stop_association_timers(stcb, true); + + /* + * The chunk lists and such SHOULD be empty but we check them just + * in case. + */ + /* anything on the wheel needs to be removed */ + for (i = 0; i < asoc->streamoutcnt; i++) { + struct sctp_stream_out *outs; + + outs = &asoc->strmout[i]; + /* now clean up any chunks here */ + TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) { + atomic_subtract_int(&asoc->stream_queue_cnt, 1); + TAILQ_REMOVE(&outs->outqueue, sp, next); + stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, outs, sp); + sctp_free_spbufspace(stcb, asoc, sp); + if (sp->data) { + if (so) { + /* Still an open socket - report */ + sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb, + 0, (void *)sp, SCTP_SO_LOCKED); + } + if (sp->data) { + sctp_m_freem(sp->data); + sp->data = NULL; + sp->tail_mbuf = NULL; + sp->length = 0; + } + } + if (sp->net) { + sctp_free_remote_addr(sp->net); + sp->net = NULL; + } + sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED); + } + } + /*sa_ignore FREED_MEMORY*/ + TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) { + TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp); + SCTP_FREE(strrst, SCTP_M_STRESET); + } + TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) { + TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next); + if (sq->data) { + sctp_m_freem(sq->data); + sq->data = NULL; + } + sctp_free_remote_addr(sq->whoFrom); + sq->whoFrom = NULL; + sq->stcb = NULL; + /* Free the ctl entry */ + sctp_free_a_readq(stcb, sq); + /*sa_ignore FREED_MEMORY*/ + } + TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) { + TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next); + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + if (chk->holds_key_ref) + sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); + SCTP_DECR_CHK_COUNT(); + atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1); + asoc->free_chunk_cnt--; + /*sa_ignore FREED_MEMORY*/ + } + /* pending send queue SHOULD be empty */ + TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) { + if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) { + asoc->strmout[chk->rec.data.sid].chunks_on_queues--; +#ifdef INVARIANTS + } else { + panic("No chunks on the queues for sid %u.", chk->rec.data.sid); +#endif + } + TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next); + if (chk->data) { + if (so) { + /* Still a socket? */ + sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb, + 0, chk, SCTP_SO_LOCKED); + } + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + } + if (chk->holds_key_ref) + sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); + if (chk->whoTo) { + sctp_free_remote_addr(chk->whoTo); + chk->whoTo = NULL; + } + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); + SCTP_DECR_CHK_COUNT(); + /*sa_ignore FREED_MEMORY*/ + } + /* sent queue SHOULD be empty */ + TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) { + if (chk->sent != SCTP_DATAGRAM_NR_ACKED) { + if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) { + asoc->strmout[chk->rec.data.sid].chunks_on_queues--; +#ifdef INVARIANTS + } else { + panic("No chunks on the queues for sid %u.", chk->rec.data.sid); +#endif + } + } + TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next); + if (chk->data) { + if (so) { + /* Still a socket? */ + sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb, + 0, chk, SCTP_SO_LOCKED); + } + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + } + if (chk->holds_key_ref) + sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); + sctp_free_remote_addr(chk->whoTo); + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); + SCTP_DECR_CHK_COUNT(); + /*sa_ignore FREED_MEMORY*/ + } +#ifdef INVARIANTS + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + if (stcb->asoc.strmout[i].chunks_on_queues > 0) { + panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i); + } + } +#endif + /* control queue MAY not be empty */ + TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) { + TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + if (chk->holds_key_ref) + sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); + sctp_free_remote_addr(chk->whoTo); + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); + SCTP_DECR_CHK_COUNT(); + /*sa_ignore FREED_MEMORY*/ + } + /* ASCONF queue MAY not be empty */ + TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) { + TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next); + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + if (chk->holds_key_ref) + sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); + sctp_free_remote_addr(chk->whoTo); + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); + SCTP_DECR_CHK_COUNT(); + /*sa_ignore FREED_MEMORY*/ + } + if (asoc->mapping_array) { + SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); + asoc->mapping_array = NULL; + } + if (asoc->nr_mapping_array) { + SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP); + asoc->nr_mapping_array = NULL; + } + /* the stream outs */ + if (asoc->strmout) { + SCTP_FREE(asoc->strmout, SCTP_M_STRMO); + asoc->strmout = NULL; + } + asoc->strm_realoutsize = asoc->streamoutcnt = 0; + if (asoc->strmin) { + for (i = 0; i < asoc->streamincnt; i++) { + sctp_clean_up_stream(stcb, &asoc->strmin[i].inqueue); + sctp_clean_up_stream(stcb, &asoc->strmin[i].uno_inqueue); + } + SCTP_FREE(asoc->strmin, SCTP_M_STRMI); + asoc->strmin = NULL; + } + asoc->streamincnt = 0; + TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) { +#ifdef INVARIANTS + if (SCTP_BASE_INFO(ipi_count_raddr) == 0) { + panic("no net's left alloc'ed, or list points to itself"); + } +#endif + TAILQ_REMOVE(&asoc->nets, net, sctp_next); + sctp_free_remote_addr(net); + } + LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) { + /*sa_ignore FREED_MEMORY*/ + sctp_remove_laddr(laddr); + } + + /* pending asconf (address) parameters */ + TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) { + /*sa_ignore FREED_MEMORY*/ + TAILQ_REMOVE(&asoc->asconf_queue, aparam, next); + SCTP_FREE(aparam,SCTP_M_ASC_ADDR); + } + TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) { + /*sa_ignore FREED_MEMORY*/ + TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next); + if (aack->data != NULL) { + sctp_m_freem(aack->data); + } + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack); + } + /* clean up auth stuff */ + if (asoc->local_hmacs) + sctp_free_hmaclist(asoc->local_hmacs); + if (asoc->peer_hmacs) + sctp_free_hmaclist(asoc->peer_hmacs); + + if (asoc->local_auth_chunks) + sctp_free_chunklist(asoc->local_auth_chunks); + if (asoc->peer_auth_chunks) + sctp_free_chunklist(asoc->peer_auth_chunks); + + sctp_free_authinfo(&asoc->authinfo); + + LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) { + LIST_REMOVE(shared_key, next); + sctp_free_sharedkey(shared_key); + /*sa_ignore FREED_MEMORY*/ + } + + /* Insert new items here :> */ + + /* Get rid of LOCK */ + SCTP_TCB_UNLOCK(stcb); + SCTP_TCB_LOCK_DESTROY(stcb); + if (from_inpcbfree == SCTP_NORMAL_PROC) { + SCTP_INP_INFO_WUNLOCK(); + SCTP_INP_RLOCK(inp); + } +#if defined(__APPLE__) && !defined(__Userspace__) + /* TEMP CODE */ + stcb->freed_from_where = from_location; +#endif +#ifdef SCTP_TRACK_FREED_ASOCS + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + /* now clean up the tasoc itself */ + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); + SCTP_DECR_ASOC_COUNT(); + } else { + LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist); + } +#else + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); + SCTP_DECR_ASOC_COUNT(); +#endif + if (from_inpcbfree == SCTP_NORMAL_PROC) { + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + /* If its NOT the inp_free calling us AND + * sctp_close as been called, we + * call back... + */ + SCTP_INP_RUNLOCK(inp); + /* This will start the kill timer (if we are + * the last one) since we hold an increment yet. But + * this is the only safe way to do this + * since otherwise if the socket closes + * at the same time we are here we might + * collide in the cleanup. + */ + sctp_inpcb_free(inp, + SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE, + SCTP_CALLED_DIRECTLY_NOCMPSET); + SCTP_INP_DECR_REF(inp); + } else { + /* The socket is still open. */ + SCTP_INP_DECR_REF(inp); + SCTP_INP_RUNLOCK(inp); + } + } + /* destroyed the asoc */ +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 11); +#endif + return (1); +} + +/* + * determine if a destination is "reachable" based upon the addresses bound + * to the current endpoint (e.g. only v4 or v6 currently bound) + */ +/* + * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use + * assoc level v4/v6 flags, as the assoc *may* not have the same address + * types bound as its endpoint + */ +int +sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr) +{ + struct sctp_inpcb *inp; + int answer; + + /* + * No locks here, the TCB, in all cases is already locked and an + * assoc is up. There is either a INP lock by the caller applied (in + * asconf case when deleting an address) or NOT in the HB case, + * however if HB then the INP increment is up and the INP will not + * be removed (on top of the fact that we have a TCB lock). So we + * only want to read the sctp_flags, which is either bound-all or + * not.. no protection needed since once an assoc is up you can't be + * changing your binding. + */ + inp = stcb->sctp_ep; + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + /* if bound all, destination is not restricted */ + /* + * RRS: Question during lock work: Is this correct? If you + * are bound-all you still might need to obey the V4--V6 + * flags??? IMO this bound-all stuff needs to be removed! + */ + return (1); + } + /* NOTE: all "scope" checks are done when local addresses are added */ + switch (destaddr->sa_family) { +#ifdef INET6 + case AF_INET6: + answer = inp->ip_inp.inp.inp_vflag & INP_IPV6; + break; +#endif +#ifdef INET + case AF_INET: + answer = inp->ip_inp.inp.inp_vflag & INP_IPV4; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + answer = inp->ip_inp.inp.inp_vflag & INP_CONN; + break; +#endif + default: + /* invalid family, so it's unreachable */ + answer = 0; + break; + } + return (answer); +} + +/* + * update the inp_vflags on an endpoint + */ +static void +sctp_update_ep_vflag(struct sctp_inpcb *inp) +{ + struct sctp_laddr *laddr; + + /* first clear the flag */ + inp->ip_inp.inp.inp_vflag = 0; + /* set the flag based on addresses on the ep list */ + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa == NULL) { + SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", + __func__); + continue; + } + + if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { + continue; + } + switch (laddr->ifa->address.sa.sa_family) { +#ifdef INET6 + case AF_INET6: + inp->ip_inp.inp.inp_vflag |= INP_IPV6; + break; +#endif +#ifdef INET + case AF_INET: + inp->ip_inp.inp.inp_vflag |= INP_IPV4; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + inp->ip_inp.inp.inp_vflag |= INP_CONN; + break; +#endif + default: + break; + } + } +} + +/* + * Add the address to the endpoint local address list There is nothing to be + * done if we are bound to all addresses + */ +void +sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action) +{ + struct sctp_laddr *laddr; + struct sctp_tcb *stcb; + int fnd, error = 0; + + fnd = 0; + + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + /* You are already bound to all. You have it already */ + return; + } +#ifdef INET6 + if (ifa->address.sa.sa_family == AF_INET6) { + if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { + /* Can't bind a non-useable addr. */ + return; + } + } +#endif + /* first, is it already present? */ + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa == ifa) { + fnd = 1; + break; + } + } + + if (fnd == 0) { + /* Not in the ep list */ + error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action); + if (error != 0) + return; + inp->laddr_count++; + /* update inp_vflag flags */ + switch (ifa->address.sa.sa_family) { +#ifdef INET6 + case AF_INET6: + inp->ip_inp.inp.inp_vflag |= INP_IPV6; + break; +#endif +#ifdef INET + case AF_INET: + inp->ip_inp.inp.inp_vflag |= INP_IPV4; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + inp->ip_inp.inp.inp_vflag |= INP_CONN; + break; +#endif + default: + break; + } + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + sctp_add_local_addr_restricted(stcb, ifa); + } + } + return; +} + +/* + * select a new (hopefully reachable) destination net (should only be used + * when we deleted an ep addr that is the only usable source address to reach + * the destination net) + */ +static void +sctp_select_primary_destination(struct sctp_tcb *stcb) +{ + struct sctp_nets *net; + + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + /* for now, we'll just pick the first reachable one we find */ + if (net->dest_state & SCTP_ADDR_UNCONFIRMED) + continue; + if (sctp_destination_is_reachable(stcb, + (struct sockaddr *)&net->ro._l_addr)) { + /* found a reachable destination */ + stcb->asoc.primary_destination = net; + } + } + /* I can't there from here! ...we're gonna die shortly... */ +} + +/* + * Delete the address from the endpoint local address list. There is nothing + * to be done if we are bound to all addresses + */ +void +sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa) +{ + struct sctp_laddr *laddr; + int fnd; + + fnd = 0; + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + /* You are already bound to all. You have it already */ + return; + } + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa == ifa) { + fnd = 1; + break; + } + } + if (fnd && (inp->laddr_count < 2)) { + /* can't delete unless there are at LEAST 2 addresses */ + return; + } + if (fnd) { + /* + * clean up any use of this address go through our + * associations and clear any last_used_address that match + * this one for each assoc, see if a new primary_destination + * is needed + */ + struct sctp_tcb *stcb; + + /* clean up "next_addr_touse" */ + if (inp->next_addr_touse == laddr) + /* delete this address */ + inp->next_addr_touse = NULL; + + /* clean up "last_used_address" */ + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + struct sctp_nets *net; + + SCTP_TCB_LOCK(stcb); + if (stcb->asoc.last_used_address == laddr) + /* delete this address */ + stcb->asoc.last_used_address = NULL; + /* Now spin through all the nets and purge any ref to laddr */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if (net->ro._s_addr == laddr->ifa) { + /* Yep, purge src address selected */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + RO_NHFREE(&net->ro); +#else + sctp_rtentry_t *rt; + + /* delete this address if cached */ + rt = net->ro.ro_rt; + if (rt != NULL) { + RTFREE(rt); + net->ro.ro_rt = NULL; + } +#endif + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + net->src_addr_selected = 0; + } + } + SCTP_TCB_UNLOCK(stcb); + } /* for each tcb */ + /* remove it from the ep list */ + sctp_remove_laddr(laddr); + inp->laddr_count--; + /* update inp_vflag flags */ + sctp_update_ep_vflag(inp); + } + return; +} + +/* + * Add the address to the TCB local address restricted list. + * This is a "pending" address list (eg. addresses waiting for an + * ASCONF-ACK response) and cannot be used as a valid source address. + */ +void +sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) +{ + struct sctp_laddr *laddr; + struct sctpladdr *list; + + /* + * Assumes TCB is locked.. and possibly the INP. May need to + * confirm/fix that if we need it and is not the case. + */ + list = &stcb->asoc.sctp_restricted_addrs; + +#ifdef INET6 + if (ifa->address.sa.sa_family == AF_INET6) { + if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { + /* Can't bind a non-existent addr. */ + return; + } + } +#endif + /* does the address already exist? */ + LIST_FOREACH(laddr, list, sctp_nxt_addr) { + if (laddr->ifa == ifa) { + return; + } + } + + /* add to the list */ + (void)sctp_insert_laddr(list, ifa, 0); + return; +} + +/* + * Remove a local address from the TCB local address restricted list + */ +void +sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) +{ + struct sctp_inpcb *inp; + struct sctp_laddr *laddr; + + /* + * This is called by asconf work. It is assumed that a) The TCB is + * locked and b) The INP is locked. This is true in as much as I can + * trace through the entry asconf code where I did these locks. + * Again, the ASCONF code is a bit different in that it does lock + * the INP during its work often times. This must be since we don't + * want other proc's looking up things while what they are looking + * up is changing :-D + */ + + inp = stcb->sctp_ep; + /* if subset bound and don't allow ASCONF's, can't delete last */ + if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) && + sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) { + if (stcb->sctp_ep->laddr_count < 2) { + /* can't delete last address */ + return; + } + } + LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) { + /* remove the address if it exists */ + if (laddr->ifa == NULL) + continue; + if (laddr->ifa == ifa) { + sctp_remove_laddr(laddr); + return; + } + } + + /* address not found! */ + return; +} + +#if defined(__FreeBSD__) && !defined(__Userspace__) +/* sysctl */ +static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC; +static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR; +#endif + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SCTP_MCORE_INPUT) && defined(SMP) +struct sctp_mcore_ctrl *sctp_mcore_workers = NULL; +int *sctp_cpuarry = NULL; + +void +sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use) +{ + /* Queue a packet to a processor for the specified core */ + struct sctp_mcore_queue *qent; + struct sctp_mcore_ctrl *wkq; + int need_wake = 0; + + if (sctp_mcore_workers == NULL) { + /* Something went way bad during setup */ + sctp_input_with_port(m, off, 0); + return; + } + SCTP_MALLOC(qent, struct sctp_mcore_queue *, + (sizeof(struct sctp_mcore_queue)), + SCTP_M_MCORE); + if (qent == NULL) { + /* This is trouble */ + sctp_input_with_port(m, off, 0); + return; + } + qent->vn = curvnet; + qent->m = m; + qent->off = off; + qent->v6 = 0; + wkq = &sctp_mcore_workers[cpu_to_use]; + SCTP_MCORE_QLOCK(wkq); + + TAILQ_INSERT_TAIL(&wkq->que, qent, next); + if (wkq->running == 0) { + need_wake = 1; + } + SCTP_MCORE_QUNLOCK(wkq); + if (need_wake) { + wakeup(&wkq->running); + } +} + +static void +sctp_mcore_thread(void *arg) +{ + + struct sctp_mcore_ctrl *wkq; + struct sctp_mcore_queue *qent; + + wkq = (struct sctp_mcore_ctrl *)arg; + struct mbuf *m; + int off, v6; + + /* Wait for first tickle */ + SCTP_MCORE_LOCK(wkq); + wkq->running = 0; + msleep(&wkq->running, + &wkq->core_mtx, + 0, "wait for pkt", 0); + SCTP_MCORE_UNLOCK(wkq); + + /* Bind to our cpu */ + thread_lock(curthread); + sched_bind(curthread, wkq->cpuid); + thread_unlock(curthread); + + /* Now lets start working */ + SCTP_MCORE_LOCK(wkq); + /* Now grab lock and go */ + for (;;) { + SCTP_MCORE_QLOCK(wkq); + skip_sleep: + wkq->running = 1; + qent = TAILQ_FIRST(&wkq->que); + if (qent) { + TAILQ_REMOVE(&wkq->que, qent, next); + SCTP_MCORE_QUNLOCK(wkq); + CURVNET_SET(qent->vn); + m = qent->m; + off = qent->off; + v6 = qent->v6; + SCTP_FREE(qent, SCTP_M_MCORE); + if (v6 == 0) { + sctp_input_with_port(m, off, 0); + } else { + SCTP_PRINTF("V6 not yet supported\n"); + sctp_m_freem(m); + } + CURVNET_RESTORE(); + SCTP_MCORE_QLOCK(wkq); + } + wkq->running = 0; + if (!TAILQ_EMPTY(&wkq->que)) { + goto skip_sleep; + } + SCTP_MCORE_QUNLOCK(wkq); + msleep(&wkq->running, + &wkq->core_mtx, + 0, "wait for pkt", 0); + } +} + +static void +sctp_startup_mcore_threads(void) +{ + int i, cpu; + + if (mp_ncpus == 1) + return; + + if (sctp_mcore_workers != NULL) { + /* Already been here in some previous + * vnet? + */ + return; + } + SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *, + ((mp_maxid+1) * sizeof(struct sctp_mcore_ctrl)), + SCTP_M_MCORE); + if (sctp_mcore_workers == NULL) { + /* TSNH I hope */ + return; + } + memset(sctp_mcore_workers, 0 , ((mp_maxid+1) * + sizeof(struct sctp_mcore_ctrl))); + /* Init the structures */ + for (i = 0; i<=mp_maxid; i++) { + TAILQ_INIT(&sctp_mcore_workers[i].que); + SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]); + SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]); + sctp_mcore_workers[i].cpuid = i; + } + if (sctp_cpuarry == NULL) { + SCTP_MALLOC(sctp_cpuarry, int *, + (mp_ncpus * sizeof(int)), + SCTP_M_MCORE); + i = 0; + CPU_FOREACH(cpu) { + sctp_cpuarry[i] = cpu; + i++; + } + } + /* Now start them all */ + CPU_FOREACH(cpu) { + (void)kproc_create(sctp_mcore_thread, + (void *)&sctp_mcore_workers[cpu], + &sctp_mcore_workers[cpu].thread_proc, + 0, + SCTP_KTHREAD_PAGES, + SCTP_MCORE_NAME); + } +} +#endif +#endif + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SCTP_NOT_YET) +static struct mbuf * +sctp_netisr_hdlr(struct mbuf *m, uintptr_t source) +{ + struct ip *ip; + struct sctphdr *sh; + int offset; + uint32_t flowid, tag; + + /* + * No flow id built by lower layers fix it so we + * create one. + */ + ip = mtod(m, struct ip *); + offset = (ip->ip_hl << 2) + sizeof(struct sctphdr); + if (SCTP_BUF_LEN(m) < offset) { + if ((m = m_pullup(m, offset)) == NULL) { + SCTP_STAT_INCR(sctps_hdrops); + return (NULL); + } + ip = mtod(m, struct ip *); + } + sh = (struct sctphdr *)((caddr_t)ip + (ip->ip_hl << 2)); + tag = htonl(sh->v_tag); + flowid = tag ^ ntohs(sh->dest_port) ^ ntohs(sh->src_port); + m->m_pkthdr.flowid = flowid; + /* FIX ME */ + m->m_flags |= M_FLOWID; + return (m); +} + +#endif +#endif +void +#if defined(__Userspace__) +sctp_pcb_init(int start_threads) +#else +sctp_pcb_init(void) +#endif +{ + /* + * SCTP initialization for the PCB structures should be called by + * the sctp_init() function. + */ + int i; + struct timeval tv; + + if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) { + /* error I was called twice */ + return; + } + SCTP_BASE_VAR(sctp_pcb_initialized) = 1; + +#if defined(SCTP_PROCESS_LEVEL_LOCKS) +#if !defined(_WIN32) + pthread_mutexattr_init(&SCTP_BASE_VAR(mtx_attr)); + pthread_rwlockattr_init(&SCTP_BASE_VAR(rwlock_attr)); +#ifdef INVARIANTS + pthread_mutexattr_settype(&SCTP_BASE_VAR(mtx_attr), PTHREAD_MUTEX_ERRORCHECK); +#endif +#endif +#endif +#if defined(SCTP_LOCAL_TRACE_BUF) +#if defined(_WIN32) && !defined(__Userspace__) + if (SCTP_BASE_SYSCTL(sctp_log) != NULL) { + memset(SCTP_BASE_SYSCTL(sctp_log), 0, sizeof(struct sctp_log)); + } +#else + memset(&SCTP_BASE_SYSCTL(sctp_log), 0, sizeof(struct sctp_log)); +#endif +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT) + SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *, + ((mp_maxid+1) * sizeof(struct sctpstat)), + SCTP_M_MCORE); +#endif +#endif + (void)SCTP_GETTIME_TIMEVAL(&tv); +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT) + memset(SCTP_BASE_STATS, 0, sizeof(struct sctpstat) * (mp_maxid+1)); + SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t)tv.tv_sec; + SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t)tv.tv_usec; +#else + memset(&SCTP_BASE_STATS, 0, sizeof(struct sctpstat)); + SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t)tv.tv_sec; + SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t)tv.tv_usec; +#endif +#else + memset(&SCTP_BASE_STATS, 0, sizeof(struct sctpstat)); + SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t)tv.tv_sec; + SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t)tv.tv_usec; +#endif + /* init the empty list of (All) Endpoints */ + LIST_INIT(&SCTP_BASE_INFO(listhead)); +#if defined(__APPLE__) && !defined(__Userspace__) + LIST_INIT(&SCTP_BASE_INFO(inplisthead)); +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) || defined(APPLE_LION) || defined(APPLE_MOUNTAINLION) + SCTP_BASE_INFO(sctbinfo).listhead = &SCTP_BASE_INFO(inplisthead); + SCTP_BASE_INFO(sctbinfo).mtx_grp_attr = lck_grp_attr_alloc_init(); + lck_grp_attr_setdefault(SCTP_BASE_INFO(sctbinfo).mtx_grp_attr); + SCTP_BASE_INFO(sctbinfo).mtx_grp = lck_grp_alloc_init("sctppcb", SCTP_BASE_INFO(sctbinfo).mtx_grp_attr); + SCTP_BASE_INFO(sctbinfo).mtx_attr = lck_attr_alloc_init(); + lck_attr_setdefault(SCTP_BASE_INFO(sctbinfo).mtx_attr); +#else + SCTP_BASE_INFO(sctbinfo).ipi_listhead = &SCTP_BASE_INFO(inplisthead); + SCTP_BASE_INFO(sctbinfo).ipi_lock_grp_attr = lck_grp_attr_alloc_init(); + lck_grp_attr_setdefault(SCTP_BASE_INFO(sctbinfo).ipi_lock_grp_attr); + SCTP_BASE_INFO(sctbinfo).ipi_lock_grp = lck_grp_alloc_init("sctppcb", SCTP_BASE_INFO(sctbinfo).ipi_lock_grp_attr); + SCTP_BASE_INFO(sctbinfo).ipi_lock_attr = lck_attr_alloc_init(); + lck_attr_setdefault(SCTP_BASE_INFO(sctbinfo).ipi_lock_attr); +#endif +#if !defined(APPLE_LEOPARD) && !defined(APPLE_SNOWLEOPARD) && !defined(APPLE_LION) && !defined(APPLE_MOUNTAINLION) + SCTP_BASE_INFO(sctbinfo).ipi_gc = sctp_gc; + in_pcbinfo_attach(&SCTP_BASE_INFO(sctbinfo)); +#endif +#endif + + /* init the hash table of endpoints */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize)); + TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize)); + TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale)); +#endif + SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31), + &SCTP_BASE_INFO(hashasocmark)); + SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize), + &SCTP_BASE_INFO(hashmark)); + SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize), + &SCTP_BASE_INFO(hashtcpmark)); + SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize); + SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH, + &SCTP_BASE_INFO(hashvrfmark)); + + SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE, + &SCTP_BASE_INFO(vrf_ifn_hashmark)); + /* init the zones */ + /* + * FIX ME: Should check for NULL returns, but if it does fail we are + * doomed to panic anyways... add later maybe. + */ + SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep", + sizeof(struct sctp_inpcb), maxsockets); + + SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc", + sizeof(struct sctp_tcb), sctp_max_number_of_assoc); + + SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr", + sizeof(struct sctp_laddr), + (sctp_max_number_of_assoc * sctp_scale_up_for_address)); + + SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr", + sizeof(struct sctp_nets), + (sctp_max_number_of_assoc * sctp_scale_up_for_address)); + + SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk", + sizeof(struct sctp_tmit_chunk), + (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); + + SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq", + sizeof(struct sctp_queued_to_read), + (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); + + SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out", + sizeof(struct sctp_stream_queue_pending), + (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); + + SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf", + sizeof(struct sctp_asconf), + (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); + + SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack", + sizeof(struct sctp_asconf_ack), + (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); + + /* Master Lock INIT for info structure */ + SCTP_INP_INFO_LOCK_INIT(); + SCTP_STATLOG_INIT_LOCK(); + + SCTP_IPI_COUNT_INIT(); + SCTP_IPI_ADDR_INIT(); +#ifdef SCTP_PACKET_LOGGING + SCTP_IP_PKTLOG_INIT(); +#endif + LIST_INIT(&SCTP_BASE_INFO(addr_wq)); + + SCTP_WQ_ADDR_INIT(); + /* not sure if we need all the counts */ + SCTP_BASE_INFO(ipi_count_ep) = 0; + /* assoc/tcb zone info */ + SCTP_BASE_INFO(ipi_count_asoc) = 0; + /* local addrlist zone info */ + SCTP_BASE_INFO(ipi_count_laddr) = 0; + /* remote addrlist zone info */ + SCTP_BASE_INFO(ipi_count_raddr) = 0; + /* chunk info */ + SCTP_BASE_INFO(ipi_count_chunk) = 0; + + /* socket queue zone info */ + SCTP_BASE_INFO(ipi_count_readq) = 0; + + /* stream out queue cont */ + SCTP_BASE_INFO(ipi_count_strmoq) = 0; + + SCTP_BASE_INFO(ipi_free_strmoq) = 0; + SCTP_BASE_INFO(ipi_free_chunks) = 0; + + SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer)); + + /* Init the TIMEWAIT list */ + for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) { + LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]); + } +#if defined(SCTP_PROCESS_LEVEL_LOCKS) +#if defined(_WIN32) + InitializeConditionVariable(&sctp_it_ctl.iterator_wakeup); +#else + (void)pthread_cond_init(&sctp_it_ctl.iterator_wakeup, NULL); +#endif +#endif + sctp_startup_iterator(); + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SCTP_MCORE_INPUT) && defined(SMP) + sctp_startup_mcore_threads(); +#endif +#endif + + /* + * INIT the default VRF which for BSD is the only one, other O/S's + * may have more. But initially they must start with one and then + * add the VRF's as addresses are added. + */ + sctp_init_vrf_list(SCTP_DEFAULT_VRF); +#if defined(__FreeBSD__) && !defined(__Userspace__) && defined(SCTP_NOT_YET) + if (ip_register_flow_handler(sctp_netisr_hdlr, IPPROTO_SCTP)) { + SCTP_PRINTF("***SCTP- Error can't register netisr handler***\n"); + } +#endif +#if defined(_SCTP_NEEDS_CALLOUT_) || defined(_USER_SCTP_NEEDS_CALLOUT_) + /* allocate the lock for the callout/timer queue */ + SCTP_TIMERQ_LOCK_INIT(); + TAILQ_INIT(&SCTP_BASE_INFO(callqueue)); +#endif +#if defined(__Userspace__) + mbuf_initialize(NULL); + atomic_init(); +#if defined(INET) || defined(INET6) + if (start_threads) + recv_thread_init(); +#endif +#endif +} + +/* + * Assumes that the SCTP_BASE_INFO() lock is NOT held. + */ +void +sctp_pcb_finish(void) +{ + struct sctp_vrflist *vrf_bucket; + struct sctp_vrf *vrf, *nvrf; + struct sctp_ifn *ifn, *nifn; + struct sctp_ifa *ifa, *nifa; + struct sctpvtaghead *chain; + struct sctp_tagblock *twait_block, *prev_twait_block; + struct sctp_laddr *wi, *nwi; + int i; + struct sctp_iterator *it, *nit; + + if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) { + SCTP_PRINTF("%s: race condition on teardown.\n", __func__); + return; + } + SCTP_BASE_VAR(sctp_pcb_initialized) = 0; +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + /* Notify the iterator to exit. */ + SCTP_IPI_ITERATOR_WQ_LOCK(); + sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_MUST_EXIT; + sctp_wakeup_iterator(); + SCTP_IPI_ITERATOR_WQ_UNLOCK(); +#endif +#if defined(__APPLE__) && !defined(__Userspace__) +#if !defined(APPLE_LEOPARD) && !defined(APPLE_SNOWLEOPARD) && !defined(APPLE_LION) && !defined(APPLE_MOUNTAINLION) + in_pcbinfo_detach(&SCTP_BASE_INFO(sctbinfo)); +#endif + SCTP_IPI_ITERATOR_WQ_LOCK(); + do { + msleep(&sctp_it_ctl.iterator_flags, + sctp_it_ctl.ipi_iterator_wq_mtx, + 0, "waiting_for_work", 0); + } while ((sctp_it_ctl.iterator_flags & SCTP_ITERATOR_EXITED) == 0); + thread_deallocate(sctp_it_ctl.thread_proc); + SCTP_IPI_ITERATOR_WQ_UNLOCK(); +#endif +#if defined(_WIN32) && !defined(__Userspace__) + if (sctp_it_ctl.iterator_thread_obj != NULL) { + NTSTATUS status = STATUS_SUCCESS; + + KeSetEvent(&sctp_it_ctl.iterator_wakeup[1], IO_NO_INCREMENT, FALSE); + status = KeWaitForSingleObject(sctp_it_ctl.iterator_thread_obj, + Executive, + KernelMode, + FALSE, + NULL); + ObDereferenceObject(sctp_it_ctl.iterator_thread_obj); + } +#endif +#if defined(__Userspace__) + if (SCTP_BASE_VAR(iterator_thread_started)) { +#if defined(_WIN32) + WaitForSingleObject(sctp_it_ctl.thread_proc, INFINITE); + CloseHandle(sctp_it_ctl.thread_proc); + sctp_it_ctl.thread_proc = NULL; +#else + pthread_join(sctp_it_ctl.thread_proc, NULL); + sctp_it_ctl.thread_proc = 0; +#endif + } +#endif +#if defined(SCTP_PROCESS_LEVEL_LOCKS) +#if defined(_WIN32) + DeleteConditionVariable(&sctp_it_ctl.iterator_wakeup); +#else + pthread_cond_destroy(&sctp_it_ctl.iterator_wakeup); + pthread_mutexattr_destroy(&SCTP_BASE_VAR(mtx_attr)); + pthread_rwlockattr_destroy(&SCTP_BASE_VAR(rwlock_attr)); +#endif +#endif + /* In FreeBSD the iterator thread never exits + * but we do clean up. + * The only way FreeBSD reaches here is if we have VRF's + * but we still add the ifdef to make it compile on old versions. + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) +retry: +#endif + SCTP_IPI_ITERATOR_WQ_LOCK(); +#if defined(__FreeBSD__) && !defined(__Userspace__) + /* + * sctp_iterator_worker() might be working on an it entry without + * holding the lock. We won't find it on the list either and + * continue and free/destroy it. While holding the lock, spin, to + * avoid the race condition as sctp_iterator_worker() will have to + * wait to re-acquire the lock. + */ + if (sctp_it_ctl.iterator_running != 0 || sctp_it_ctl.cur_it != NULL) { + SCTP_IPI_ITERATOR_WQ_UNLOCK(); + SCTP_PRINTF("%s: Iterator running while we held the lock. Retry. " + "cur_it=%p\n", __func__, sctp_it_ctl.cur_it); + DELAY(10); + goto retry; + } +#endif + TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (it->vn != curvnet) { + continue; + } +#endif + TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); + if (it->function_atend != NULL) { + (*it->function_atend) (it->pointer, it->val); + } + SCTP_FREE(it,SCTP_M_ITER); + } + SCTP_IPI_ITERATOR_WQ_UNLOCK(); +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_ITERATOR_LOCK(); + if ((sctp_it_ctl.cur_it) && + (sctp_it_ctl.cur_it->vn == curvnet)) { + sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT; + } + SCTP_ITERATOR_UNLOCK(); +#endif +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + SCTP_IPI_ITERATOR_WQ_DESTROY(); + SCTP_ITERATOR_LOCK_DESTROY(); +#endif + SCTP_OS_TIMER_STOP_DRAIN(&SCTP_BASE_INFO(addr_wq_timer.timer)); + SCTP_WQ_ADDR_LOCK(); + LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) { + LIST_REMOVE(wi, sctp_nxt_addr); + SCTP_DECR_LADDR_COUNT(); + if (wi->action == SCTP_DEL_IP_ADDRESS) { + SCTP_FREE(wi->ifa, SCTP_M_IFA); + } + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi); + } + SCTP_WQ_ADDR_UNLOCK(); + + /* + * free the vrf/ifn/ifa lists and hashes (be sure address monitor + * is destroyed first). + */ + SCTP_IPI_ADDR_WLOCK(); + vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))]; + LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) { + LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) { + LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) { + /* free the ifa */ + LIST_REMOVE(ifa, next_bucket); + LIST_REMOVE(ifa, next_ifa); + SCTP_FREE(ifa, SCTP_M_IFA); + } + /* free the ifn */ + LIST_REMOVE(ifn, next_bucket); + LIST_REMOVE(ifn, next_ifn); + SCTP_FREE(ifn, SCTP_M_IFN); + } + SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark); + /* free the vrf */ + LIST_REMOVE(vrf, next_vrf); + SCTP_FREE(vrf, SCTP_M_VRF); + } + SCTP_IPI_ADDR_WUNLOCK(); + /* free the vrf hashes */ + SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark)); + SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark)); + + /* free the TIMEWAIT list elements malloc'd in the function + * sctp_add_vtag_to_timewait()... + */ + for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) { + chain = &SCTP_BASE_INFO(vtag_timewait)[i]; + if (!LIST_EMPTY(chain)) { + prev_twait_block = NULL; + LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { + if (prev_twait_block) { + SCTP_FREE(prev_twait_block, SCTP_M_TIMW); + } + prev_twait_block = twait_block; + } + SCTP_FREE(prev_twait_block, SCTP_M_TIMW); + } + } + + /* free the locks and mutexes */ +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_TIMERQ_LOCK_DESTROY(); +#endif +#ifdef SCTP_PACKET_LOGGING + SCTP_IP_PKTLOG_DESTROY(); +#endif + SCTP_IPI_ADDR_DESTROY(); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_IPI_COUNT_DESTROY(); +#endif + SCTP_STATLOG_DESTROY(); + SCTP_INP_INFO_LOCK_DESTROY(); + + SCTP_WQ_ADDR_DESTROY(); + +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) || defined(APPLE_LION) || defined(APPLE_MOUNTAINLION) + lck_grp_attr_free(SCTP_BASE_INFO(sctbinfo).mtx_grp_attr); + lck_grp_free(SCTP_BASE_INFO(sctbinfo).mtx_grp); + lck_attr_free(SCTP_BASE_INFO(sctbinfo).mtx_attr); +#else + lck_grp_attr_free(SCTP_BASE_INFO(sctbinfo).ipi_lock_grp_attr); + lck_grp_free(SCTP_BASE_INFO(sctbinfo).ipi_lock_grp); + lck_attr_free(SCTP_BASE_INFO(sctbinfo).ipi_lock_attr); +#endif +#endif +#if defined(__Userspace__) + SCTP_TIMERQ_LOCK_DESTROY(); + SCTP_ZONE_DESTROY(zone_mbuf); + SCTP_ZONE_DESTROY(zone_clust); + SCTP_ZONE_DESTROY(zone_ext_refcnt); +#endif + /* Get rid of other stuff too. */ + if (SCTP_BASE_INFO(sctp_asochash) != NULL) + SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark)); + if (SCTP_BASE_INFO(sctp_ephash) != NULL) + SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark)); + if (SCTP_BASE_INFO(sctp_tcpephash) != NULL) + SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark)); + +#if defined(_WIN32) || defined(__FreeBSD__) || defined(__Userspace__) + SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep)); + SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc)); + SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr)); + SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net)); + SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk)); + SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq)); + SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq)); + SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf)); + SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack)); +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT) + SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE); +#endif +#endif +} + +int +sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m, + int offset, int limit, + struct sockaddr *src, struct sockaddr *dst, + struct sockaddr *altsa, uint16_t port) +{ + /* + * grub through the INIT pulling addresses and loading them to the + * nets structure in the asoc. The from address in the mbuf should + * also be loaded (if it is not already). This routine can be called + * with either INIT or INIT-ACK's as long as the m points to the IP + * packet and the offset points to the beginning of the parameters. + */ + struct sctp_inpcb *inp; + struct sctp_nets *net, *nnet, *net_tmp; + struct sctp_paramhdr *phdr, param_buf; + struct sctp_tcb *stcb_tmp; + uint16_t ptype, plen; + struct sockaddr *sa; + uint8_t random_store[SCTP_PARAM_BUFFER_SIZE]; + struct sctp_auth_random *p_random = NULL; + uint16_t random_len = 0; + uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE]; + struct sctp_auth_hmac_algo *hmacs = NULL; + uint16_t hmacs_len = 0; + uint8_t saw_asconf = 0; + uint8_t saw_asconf_ack = 0; + uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE]; + struct sctp_auth_chunk_list *chunks = NULL; + uint16_t num_chunks = 0; + sctp_key_t *new_key; + uint32_t keylen; + int got_random = 0, got_hmacs = 0, got_chklist = 0; + uint8_t peer_supports_ecn; + uint8_t peer_supports_prsctp; + uint8_t peer_supports_auth; + uint8_t peer_supports_asconf; + uint8_t peer_supports_asconf_ack; + uint8_t peer_supports_reconfig; + uint8_t peer_supports_nrsack; + uint8_t peer_supports_pktdrop; + uint8_t peer_supports_idata; +#ifdef INET + struct sockaddr_in sin; +#endif +#ifdef INET6 + struct sockaddr_in6 sin6; +#endif + + /* First get the destination address setup too. */ +#ifdef INET + memset(&sin, 0, sizeof(sin)); + sin.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + sin.sin_len = sizeof(sin); +#endif + sin.sin_port = stcb->rport; +#endif +#ifdef INET6 + memset(&sin6, 0, sizeof(sin6)); + sin6.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6.sin6_len = sizeof(struct sockaddr_in6); +#endif + sin6.sin6_port = stcb->rport; +#endif + if (altsa) { + sa = altsa; + } else { + sa = src; + } + peer_supports_idata = 0; + peer_supports_ecn = 0; + peer_supports_prsctp = 0; + peer_supports_auth = 0; + peer_supports_asconf = 0; + peer_supports_asconf_ack = 0; + peer_supports_reconfig = 0; + peer_supports_nrsack = 0; + peer_supports_pktdrop = 0; + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + /* mark all addresses that we have currently on the list */ + net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC; + } + /* does the source address already exist? if so skip it */ + inp = stcb->sctp_ep; + atomic_add_int(&stcb->asoc.refcnt, 1); + stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + + if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) { + /* we must add the source address */ + /* no scope set here since we have a tcb already. */ + switch (sa->sa_family) { +#ifdef INET + case AF_INET: + if (stcb->asoc.scope.ipv4_addr_legal) { + if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) { + return (-1); + } + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (stcb->asoc.scope.ipv6_addr_legal) { + if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) { + return (-2); + } + } + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + if (stcb->asoc.scope.conn_addr_legal) { + if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) { + return (-2); + } + } + break; +#endif + default: + break; + } + } else { + if (net_tmp != NULL && stcb_tmp == stcb) { + net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC; + } else if (stcb_tmp != stcb) { + /* It belongs to another association? */ + if (stcb_tmp) + SCTP_TCB_UNLOCK(stcb_tmp); + return (-3); + } + } + if (stcb->asoc.state == 0) { + /* the assoc was freed? */ + return (-4); + } + /* now we must go through each of the params. */ + phdr = sctp_get_next_param(m, offset, ¶m_buf, sizeof(param_buf)); + while (phdr) { + ptype = ntohs(phdr->param_type); + plen = ntohs(phdr->param_length); + /* + * SCTP_PRINTF("ptype => %0x, plen => %d\n", (uint32_t)ptype, + * (int)plen); + */ + if (offset + plen > limit) { + break; + } + if (plen < sizeof(struct sctp_paramhdr)) { + break; + } +#ifdef INET + if (ptype == SCTP_IPV4_ADDRESS) { + if (stcb->asoc.scope.ipv4_addr_legal) { + struct sctp_ipv4addr_param *p4, p4_buf; + + /* ok get the v4 address and check/add */ + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)&p4_buf, + sizeof(p4_buf)); + if (plen != sizeof(struct sctp_ipv4addr_param) || + phdr == NULL) { + return (-5); + } + p4 = (struct sctp_ipv4addr_param *)phdr; + sin.sin_addr.s_addr = p4->addr; + if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { + /* Skip multi-cast addresses */ + goto next_param; + } + if ((sin.sin_addr.s_addr == INADDR_BROADCAST) || + (sin.sin_addr.s_addr == INADDR_ANY)) { + goto next_param; + } + sa = (struct sockaddr *)&sin; + inp = stcb->sctp_ep; + atomic_add_int(&stcb->asoc.refcnt, 1); + stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net, + dst, stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + + if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || + inp == NULL) { + /* we must add the source address */ + /* + * no scope set since we have a tcb + * already + */ + + /* + * we must validate the state again + * here + */ + add_it_now: + if (stcb->asoc.state == 0) { + /* the assoc was freed? */ + return (-7); + } + if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) { + return (-8); + } + } else if (stcb_tmp == stcb) { + if (stcb->asoc.state == 0) { + /* the assoc was freed? */ + return (-10); + } + if (net != NULL) { + /* clear flag */ + net->dest_state &= + ~SCTP_ADDR_NOT_IN_ASSOC; + } + } else { + /* + * strange, address is in another + * assoc? straighten out locks. + */ + if (stcb_tmp) { + if (SCTP_GET_STATE(stcb_tmp) == SCTP_STATE_COOKIE_WAIT) { + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + /* in setup state we abort this guy */ + SCTP_SNPRINTF(msg, sizeof(msg), + "%s:%d at %s", __FILE__, __LINE__, __func__); + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + msg); + sctp_abort_an_association(stcb_tmp->sctp_ep, + stcb_tmp, op_err, false, + SCTP_SO_NOT_LOCKED); + goto add_it_now; + } + SCTP_TCB_UNLOCK(stcb_tmp); + } + + if (stcb->asoc.state == 0) { + /* the assoc was freed? */ + return (-12); + } + return (-13); + } + } + } else +#endif +#ifdef INET6 + if (ptype == SCTP_IPV6_ADDRESS) { + if (stcb->asoc.scope.ipv6_addr_legal) { + /* ok get the v6 address and check/add */ + struct sctp_ipv6addr_param *p6, p6_buf; + + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)&p6_buf, + sizeof(p6_buf)); + if (plen != sizeof(struct sctp_ipv6addr_param) || + phdr == NULL) { + return (-14); + } + p6 = (struct sctp_ipv6addr_param *)phdr; + memcpy((caddr_t)&sin6.sin6_addr, p6->addr, + sizeof(p6->addr)); + if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) { + /* Skip multi-cast addresses */ + goto next_param; + } + if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) { + /* Link local make no sense without scope */ + goto next_param; + } + sa = (struct sockaddr *)&sin6; + inp = stcb->sctp_ep; + atomic_add_int(&stcb->asoc.refcnt, 1); + stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net, + dst, stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb_tmp == NULL && + (inp == stcb->sctp_ep || inp == NULL)) { + /* + * we must validate the state again + * here + */ + add_it_now6: + if (stcb->asoc.state == 0) { + /* the assoc was freed? */ + return (-16); + } + /* + * we must add the address, no scope + * set + */ + if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) { + return (-17); + } + } else if (stcb_tmp == stcb) { + /* + * we must validate the state again + * here + */ + if (stcb->asoc.state == 0) { + /* the assoc was freed? */ + return (-19); + } + if (net != NULL) { + /* clear flag */ + net->dest_state &= + ~SCTP_ADDR_NOT_IN_ASSOC; + } + } else { + /* + * strange, address is in another + * assoc? straighten out locks. + */ + if (stcb_tmp) { + if (SCTP_GET_STATE(stcb_tmp) == SCTP_STATE_COOKIE_WAIT) { + struct mbuf *op_err; + char msg[SCTP_DIAG_INFO_LEN]; + + /* in setup state we abort this guy */ + SCTP_SNPRINTF(msg, sizeof(msg), + "%s:%d at %s", __FILE__, __LINE__, __func__); + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + msg); + sctp_abort_an_association(stcb_tmp->sctp_ep, + stcb_tmp, op_err, false, + SCTP_SO_NOT_LOCKED); + goto add_it_now6; + } + SCTP_TCB_UNLOCK(stcb_tmp); + } + if (stcb->asoc.state == 0) { + /* the assoc was freed? */ + return (-21); + } + return (-22); + } + } + } else +#endif + if (ptype == SCTP_ECN_CAPABLE) { + peer_supports_ecn = 1; + } else if (ptype == SCTP_ULP_ADAPTATION) { + if (stcb->asoc.state != SCTP_STATE_OPEN) { + struct sctp_adaptation_layer_indication ai, *aip; + + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)&ai, sizeof(ai)); + aip = (struct sctp_adaptation_layer_indication *)phdr; + if (aip) { + stcb->asoc.peers_adaptation = ntohl(aip->indication); + stcb->asoc.adaptation_needed = 1; + } + } + } else if (ptype == SCTP_SET_PRIM_ADDR) { + struct sctp_asconf_addr_param lstore, *fee; + int lptype; + struct sockaddr *lsa = NULL; +#ifdef INET + struct sctp_asconf_addrv4_param *fii; +#endif + + if (stcb->asoc.asconf_supported == 0) { + return (-100); + } + if (plen > sizeof(lstore)) { + return (-23); + } + if (plen < sizeof(struct sctp_asconf_addrv4_param)) { + return (-101); + } + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)&lstore, + plen); + if (phdr == NULL) { + return (-24); + } + fee = (struct sctp_asconf_addr_param *)phdr; + lptype = ntohs(fee->addrp.ph.param_type); + switch (lptype) { +#ifdef INET + case SCTP_IPV4_ADDRESS: + if (plen != + sizeof(struct sctp_asconf_addrv4_param)) { + SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n", + (int)sizeof(struct sctp_asconf_addrv4_param), + plen); + } else { + fii = (struct sctp_asconf_addrv4_param *)fee; + sin.sin_addr.s_addr = fii->addrp.addr; + lsa = (struct sockaddr *)&sin; + } + break; +#endif +#ifdef INET6 + case SCTP_IPV6_ADDRESS: + if (plen != + sizeof(struct sctp_asconf_addr_param)) { + SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n", + (int)sizeof(struct sctp_asconf_addr_param), + plen); + } else { + memcpy(sin6.sin6_addr.s6_addr, + fee->addrp.addr, + sizeof(fee->addrp.addr)); + lsa = (struct sockaddr *)&sin6; + } + break; +#endif + default: + break; + } + if (lsa) { + (void)sctp_set_primary_addr(stcb, sa, NULL); + } + } else if (ptype == SCTP_HAS_NAT_SUPPORT) { + stcb->asoc.peer_supports_nat = 1; + } else if (ptype == SCTP_PRSCTP_SUPPORTED) { + /* Peer supports pr-sctp */ + peer_supports_prsctp = 1; + } else if (ptype == SCTP_ZERO_CHECKSUM_ACCEPTABLE) { + /* + * Only send zero checksums if the upper layer has + * also enabled the support for this. + */ + if (stcb->asoc.zero_checksum == 1) { + stcb->asoc.zero_checksum = 2; + } + } else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) { + /* A supported extension chunk */ + struct sctp_supported_chunk_types_param *pr_supported; + uint8_t local_store[SCTP_PARAM_BUFFER_SIZE]; + int num_ent, i; + + if (plen > sizeof(local_store)) { + return (-35); + } + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)&local_store, plen); + if (phdr == NULL) { + return (-25); + } + pr_supported = (struct sctp_supported_chunk_types_param *)phdr; + num_ent = plen - sizeof(struct sctp_paramhdr); + for (i = 0; i < num_ent; i++) { + switch (pr_supported->chunk_types[i]) { + case SCTP_ASCONF: + peer_supports_asconf = 1; + break; + case SCTP_ASCONF_ACK: + peer_supports_asconf_ack = 1; + break; + case SCTP_FORWARD_CUM_TSN: + peer_supports_prsctp = 1; + break; + case SCTP_PACKET_DROPPED: + peer_supports_pktdrop = 1; + break; + case SCTP_NR_SELECTIVE_ACK: + peer_supports_nrsack = 1; + break; + case SCTP_STREAM_RESET: + peer_supports_reconfig = 1; + break; + case SCTP_AUTHENTICATION: + peer_supports_auth = 1; + break; + case SCTP_IDATA: + peer_supports_idata = 1; + break; + default: + /* one I have not learned yet */ + break; + } + } + } else if (ptype == SCTP_RANDOM) { + if (plen > sizeof(random_store)) + break; + if (got_random) { + /* already processed a RANDOM */ + goto next_param; + } + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)random_store, + plen); + if (phdr == NULL) + return (-26); + p_random = (struct sctp_auth_random *)phdr; + random_len = plen - sizeof(*p_random); + /* enforce the random length */ + if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) { + SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n"); + return (-27); + } + got_random = 1; + } else if (ptype == SCTP_HMAC_LIST) { + uint16_t num_hmacs; + uint16_t i; + + if (plen > sizeof(hmacs_store)) + break; + if (got_hmacs) { + /* already processed a HMAC list */ + goto next_param; + } + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)hmacs_store, + plen); + if (phdr == NULL) + return (-28); + hmacs = (struct sctp_auth_hmac_algo *)phdr; + hmacs_len = plen - sizeof(*hmacs); + num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]); + /* validate the hmac list */ + if (sctp_verify_hmac_param(hmacs, num_hmacs)) { + return (-29); + } + if (stcb->asoc.peer_hmacs != NULL) + sctp_free_hmaclist(stcb->asoc.peer_hmacs); + stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs); + if (stcb->asoc.peer_hmacs != NULL) { + for (i = 0; i < num_hmacs; i++) { + (void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs, + ntohs(hmacs->hmac_ids[i])); + } + } + got_hmacs = 1; + } else if (ptype == SCTP_CHUNK_LIST) { + int i; + + if (plen > sizeof(chunks_store)) + break; + if (got_chklist) { + /* already processed a Chunks list */ + goto next_param; + } + phdr = sctp_get_next_param(m, offset, + (struct sctp_paramhdr *)chunks_store, + plen); + if (phdr == NULL) + return (-30); + chunks = (struct sctp_auth_chunk_list *)phdr; + num_chunks = plen - sizeof(*chunks); + if (stcb->asoc.peer_auth_chunks != NULL) + sctp_clear_chunklist(stcb->asoc.peer_auth_chunks); + else + stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist(); + for (i = 0; i < num_chunks; i++) { + (void)sctp_auth_add_chunk(chunks->chunk_types[i], + stcb->asoc.peer_auth_chunks); + /* record asconf/asconf-ack if listed */ + if (chunks->chunk_types[i] == SCTP_ASCONF) + saw_asconf = 1; + if (chunks->chunk_types[i] == SCTP_ASCONF_ACK) + saw_asconf_ack = 1; + } + got_chklist = 1; + } else if ((ptype == SCTP_HEARTBEAT_INFO) || + (ptype == SCTP_STATE_COOKIE) || + (ptype == SCTP_UNRECOG_PARAM) || + (ptype == SCTP_COOKIE_PRESERVE) || + (ptype == SCTP_SUPPORTED_ADDRTYPE) || + (ptype == SCTP_ADD_IP_ADDRESS) || + (ptype == SCTP_DEL_IP_ADDRESS) || + (ptype == SCTP_ERROR_CAUSE_IND) || + (ptype == SCTP_SUCCESS_REPORT)) { + /* don't care */ + } else { + if ((ptype & 0x8000) == 0x0000) { + /* + * must stop processing the rest of the + * param's. Any report bits were handled + * with the call to + * sctp_arethere_unrecognized_parameters() + * when the INIT or INIT-ACK was first seen. + */ + break; + } + } + + next_param: + offset += SCTP_SIZE32(plen); + if (offset >= limit) { + break; + } + phdr = sctp_get_next_param(m, offset, ¶m_buf, + sizeof(param_buf)); + } + /* Now check to see if we need to purge any addresses */ + TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) { + if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) == + SCTP_ADDR_NOT_IN_ASSOC) { + /* This address has been removed from the asoc */ + /* remove and free it */ + stcb->asoc.numnets--; + TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next); + if (net == stcb->asoc.alternate) { + sctp_free_remote_addr(stcb->asoc.alternate); + stcb->asoc.alternate = NULL; + } + if (net == stcb->asoc.primary_destination) { + stcb->asoc.primary_destination = NULL; + sctp_select_primary_destination(stcb); + } + sctp_free_remote_addr(net); + } + } + if ((stcb->asoc.ecn_supported == 1) && + (peer_supports_ecn == 0)) { + stcb->asoc.ecn_supported = 0; + } + if ((stcb->asoc.prsctp_supported == 1) && + (peer_supports_prsctp == 0)) { + stcb->asoc.prsctp_supported = 0; + } + if ((stcb->asoc.auth_supported == 1) && + ((peer_supports_auth == 0) || + (got_random == 0) || (got_hmacs == 0))) { + stcb->asoc.auth_supported = 0; + } + if ((stcb->asoc.asconf_supported == 1) && + ((peer_supports_asconf == 0) || (peer_supports_asconf_ack == 0) || + (stcb->asoc.auth_supported == 0) || + (saw_asconf == 0) || (saw_asconf_ack == 0))) { + stcb->asoc.asconf_supported = 0; + } + if ((stcb->asoc.reconfig_supported == 1) && + (peer_supports_reconfig == 0)) { + stcb->asoc.reconfig_supported = 0; + } + if ((stcb->asoc.idata_supported == 1) && + (peer_supports_idata == 0)) { + stcb->asoc.idata_supported = 0; + } + if ((stcb->asoc.nrsack_supported == 1) && + (peer_supports_nrsack == 0)) { + stcb->asoc.nrsack_supported = 0; + } + if ((stcb->asoc.pktdrop_supported == 1) && + (peer_supports_pktdrop == 0)) { + stcb->asoc.pktdrop_supported = 0; + } + /* validate authentication required parameters */ + if ((peer_supports_auth == 0) && (got_chklist == 1)) { + /* peer does not support auth but sent a chunks list? */ + return (-31); + } + if ((peer_supports_asconf == 1) && (peer_supports_auth == 0)) { + /* peer supports asconf but not auth? */ + return (-32); + } else if ((peer_supports_asconf == 1) && + (peer_supports_auth == 1) && + ((saw_asconf == 0) || (saw_asconf_ack == 0))) { + return (-33); + } + /* concatenate the full random key */ + keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len; + if (chunks != NULL) { + keylen += sizeof(*chunks) + num_chunks; + } + new_key = sctp_alloc_key(keylen); + if (new_key != NULL) { + /* copy in the RANDOM */ + if (p_random != NULL) { + keylen = sizeof(*p_random) + random_len; + memcpy(new_key->key, p_random, keylen); + } else { + keylen = 0; + } + /* append in the AUTH chunks */ + if (chunks != NULL) { + memcpy(new_key->key + keylen, chunks, + sizeof(*chunks) + num_chunks); + keylen += sizeof(*chunks) + num_chunks; + } + /* append in the HMACs */ + if (hmacs != NULL) { + memcpy(new_key->key + keylen, hmacs, + sizeof(*hmacs) + hmacs_len); + } + } else { + /* failed to get memory for the key */ + return (-34); + } + if (stcb->asoc.authinfo.peer_random != NULL) + sctp_free_key(stcb->asoc.authinfo.peer_random); + stcb->asoc.authinfo.peer_random = new_key; + sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid); + sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid); + + return (0); +} + +int +sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa, + struct sctp_nets *net) +{ + /* make sure the requested primary address exists in the assoc */ + if (net == NULL && sa) + net = sctp_findnet(stcb, sa); + + if (net == NULL) { + /* didn't find the requested primary address! */ + return (-1); + } else { + /* set the primary address */ + if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { + /* Must be confirmed, so queue to set */ + net->dest_state |= SCTP_ADDR_REQ_PRIMARY; + return (0); + } + stcb->asoc.primary_destination = net; + if (((net->dest_state & SCTP_ADDR_PF) == 0) && + (stcb->asoc.alternate != NULL)) { + sctp_free_remote_addr(stcb->asoc.alternate); + stcb->asoc.alternate = NULL; + } + net = TAILQ_FIRST(&stcb->asoc.nets); + if (net != stcb->asoc.primary_destination) { + /* first one on the list is NOT the primary + * sctp_cmpaddr() is much more efficient if + * the primary is the first on the list, make it + * so. + */ + TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); + TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); + } + return (0); + } +} + +bool +sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now) +{ + struct sctpasochead *head; + struct sctp_tcb *stcb; + + SCTP_INP_INFO_LOCK_ASSERT(); + + head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag, SCTP_BASE_INFO(hashasocmark))]; + LIST_FOREACH(stcb, head, sctp_asocs) { + /* We choose not to lock anything here. TCB's can't be + * removed since we have the read lock, so they can't + * be freed on us, same thing for the INP. I may + * be wrong with this assumption, but we will go + * with it for now :-) + */ + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { + continue; + } + if (stcb->asoc.my_vtag == tag) { + /* candidate */ + if (stcb->rport != rport) { + continue; + } + if (stcb->sctp_ep->sctp_lport != lport) { + continue; + } + /* The tag is currently used, so don't use it. */ + return (false); + } + } + return (!sctp_is_in_timewait(tag, lport, rport, (uint32_t)now->tv_sec)); +} + +static void +sctp_drain_mbufs(struct sctp_tcb *stcb) +{ + /* + * We must hunt this association for MBUF's past the cumack (i.e. + * out of order data that we can renege on). + */ + struct sctp_association *asoc; + struct sctp_tmit_chunk *chk, *nchk; + uint32_t cumulative_tsn_p1; + struct sctp_queued_to_read *control, *ncontrol; + int cnt, strmat; + uint32_t gap, i; + int fnd = 0; + + /* We look for anything larger than the cum-ack + 1 */ + + asoc = &stcb->asoc; + if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) { + /* none we can reneg on. */ + return; + } + SCTP_STAT_INCR(sctps_protocol_drains_done); + cumulative_tsn_p1 = asoc->cumulative_tsn + 1; + cnt = 0; + /* Ok that was fun, now we will drain all the inbound streams? */ + for (strmat = 0; strmat < asoc->streamincnt; strmat++) { + TAILQ_FOREACH_SAFE(control, &asoc->strmin[strmat].inqueue, next_instrm, ncontrol) { +#ifdef INVARIANTS + if (control->on_strm_q != SCTP_ON_ORDERED) { + panic("Huh control: %p on_q: %d -- not ordered?", + control, control->on_strm_q); + } +#endif + if (SCTP_TSN_GT(control->sinfo_tsn, cumulative_tsn_p1)) { + /* Yep it is above cum-ack */ + cnt++; + SCTP_CALC_TSN_TO_GAP(gap, control->sinfo_tsn, asoc->mapping_array_base_tsn); + KASSERT(control->length > 0, ("control has zero length")); + if (asoc->size_on_all_streams >= control->length) { + asoc->size_on_all_streams -= control->length; + } else { +#ifdef INVARIANTS + panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); +#else + asoc->size_on_all_streams = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_all_streams); + SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); + if (control->on_read_q) { + TAILQ_REMOVE(&stcb->sctp_ep->read_queue, control, next); + control->on_read_q = 0; + } + TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, control, next_instrm); + control->on_strm_q = 0; + if (control->data) { + sctp_m_freem(control->data); + control->data = NULL; + } + sctp_free_remote_addr(control->whoFrom); + /* Now its reasm? */ + TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) { + cnt++; + SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.tsn, asoc->mapping_array_base_tsn); + KASSERT(chk->send_size > 0, ("chunk has zero length")); + if (asoc->size_on_reasm_queue >= chk->send_size) { + asoc->size_on_reasm_queue -= chk->send_size; + } else { +#ifdef INVARIANTS + panic("size_on_reasm_queue = %u smaller than chunk length %u", asoc->size_on_reasm_queue, chk->send_size); +#else + asoc->size_on_reasm_queue = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_reasm_queue); + SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); + TAILQ_REMOVE(&control->reasm, chk, sctp_next); + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + } + sctp_free_a_readq(stcb, control); + } + } + TAILQ_FOREACH_SAFE(control, &asoc->strmin[strmat].uno_inqueue, next_instrm, ncontrol) { +#ifdef INVARIANTS + if (control->on_strm_q != SCTP_ON_UNORDERED) { + panic("Huh control: %p on_q: %d -- not unordered?", + control, control->on_strm_q); + } +#endif + if (SCTP_TSN_GT(control->sinfo_tsn, cumulative_tsn_p1)) { + /* Yep it is above cum-ack */ + cnt++; + SCTP_CALC_TSN_TO_GAP(gap, control->sinfo_tsn, asoc->mapping_array_base_tsn); + KASSERT(control->length > 0, ("control has zero length")); + if (asoc->size_on_all_streams >= control->length) { + asoc->size_on_all_streams -= control->length; + } else { +#ifdef INVARIANTS + panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length); +#else + asoc->size_on_all_streams = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_all_streams); + SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); + if (control->on_read_q) { + TAILQ_REMOVE(&stcb->sctp_ep->read_queue, control, next); + control->on_read_q = 0; + } + TAILQ_REMOVE(&asoc->strmin[strmat].uno_inqueue, control, next_instrm); + control->on_strm_q = 0; + if (control->data) { + sctp_m_freem(control->data); + control->data = NULL; + } + sctp_free_remote_addr(control->whoFrom); + /* Now its reasm? */ + TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) { + cnt++; + SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.tsn, asoc->mapping_array_base_tsn); + KASSERT(chk->send_size > 0, ("chunk has zero length")); + if (asoc->size_on_reasm_queue >= chk->send_size) { + asoc->size_on_reasm_queue -= chk->send_size; + } else { +#ifdef INVARIANTS + panic("size_on_reasm_queue = %u smaller than chunk length %u", asoc->size_on_reasm_queue, chk->send_size); +#else + asoc->size_on_reasm_queue = 0; +#endif + } + sctp_ucount_decr(asoc->cnt_on_reasm_queue); + SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); + TAILQ_REMOVE(&control->reasm, chk, sctp_next); + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + } + sctp_free_a_readq(stcb, control); + } + } + } + if (cnt) { + /* We must back down to see what the new highest is */ + for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) { + SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn); + if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) { + asoc->highest_tsn_inside_map = i; + fnd = 1; + break; + } + } + if (!fnd) { + asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1; + } + + /* + * Question, should we go through the delivery queue? The only + * reason things are on here is the app not reading OR a p-d-api up. + * An attacker COULD send enough in to initiate the PD-API and then + * send a bunch of stuff to other streams... these would wind up on + * the delivery queue.. and then we would not get to them. But in + * order to do this I then have to back-track and un-deliver + * sequence numbers in streams.. el-yucko. I think for now we will + * NOT look at the delivery queue and leave it to be something to + * consider later. An alternative would be to abort the P-D-API with + * a notification and then deliver the data.... Or another method + * might be to keep track of how many times the situation occurs and + * if we see a possible attack underway just abort the association. + */ +#ifdef SCTP_DEBUG + SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt); +#endif + /* + * Now do we need to find a new + * asoc->highest_tsn_inside_map? + */ + asoc->last_revoke_count = cnt; + sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, + SCTP_FROM_SCTP_PCB + SCTP_LOC_11); + /*sa_ignore NO_NULL_CHK*/ + sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED); + sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED); + } + /* + * Another issue, in un-setting the TSN's in the mapping array we + * DID NOT adjust the highest_tsn marker. This will cause one of two + * things to occur. It may cause us to do extra work in checking for + * our mapping array movement. More importantly it may cause us to + * SACK every datagram. This may not be a bad thing though since we + * will recover once we get our cum-ack above and all this stuff we + * dumped recovered. + */ +} + +#if defined(__FreeBSD__) && !defined(__Userspace__) +static void +#else +void +#endif +sctp_drain(void) +{ +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; + VNET_ITERATOR_DECL(vnet_iter); + + NET_EPOCH_ENTER(et); +#else + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; + + SCTP_STAT_INCR(sctps_protocol_drain_calls); + if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) { + return; + } +#endif + /* + * We must walk the PCB lists for ALL associations here. The system + * is LOW on MBUF's and needs help. This is where reneging will + * occur. We really hope this does NOT happen! + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + VNET_LIST_RLOCK_NOSLEEP(); + VNET_FOREACH(vnet_iter) { + CURVNET_SET(vnet_iter); + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; +#endif + +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_STAT_INCR(sctps_protocol_drain_calls); + if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) { +#ifdef VIMAGE + continue; +#else + NET_EPOCH_EXIT(et); + return; +#endif + } +#endif + SCTP_INP_INFO_RLOCK(); + LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) { + /* For each endpoint */ + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + /* For each association */ + SCTP_TCB_LOCK(stcb); + sctp_drain_mbufs(stcb); + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + SCTP_INP_INFO_RUNLOCK(); +#if defined(__FreeBSD__) && !defined(__Userspace__) + CURVNET_RESTORE(); + } + VNET_LIST_RUNLOCK_NOSLEEP(); + NET_EPOCH_EXIT(et); +#endif +} +#if defined(__FreeBSD__) && !defined(__Userspace__) +EVENTHANDLER_DEFINE(vm_lowmem, sctp_drain, NULL, LOWMEM_PRI_DEFAULT); +EVENTHANDLER_DEFINE(mbuf_lowmem, sctp_drain, NULL, LOWMEM_PRI_DEFAULT); +#endif + +/* + * start a new iterator + * iterates through all endpoints and associations based on the pcb_state + * flags and asoc_state. "af" (mandatory) is executed for all matching + * assocs and "ef" (optional) is executed when the iterator completes. + * "inpf" (optional) is executed for each new endpoint as it is being + * iterated through. inpe (optional) is called when the inp completes + * its way through all the stcbs. + */ +int +sctp_initiate_iterator(inp_func inpf, + asoc_func af, + inp_func inpe, + uint32_t pcb_state, + uint32_t pcb_features, + uint32_t asoc_state, + void *argp, + uint32_t argi, + end_func ef, + struct sctp_inpcb *s_inp, + uint8_t chunk_output_off) +{ + struct sctp_iterator *it = NULL; + + if (af == NULL) { + return (-1); + } + if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) { + SCTP_PRINTF("%s: abort on initialize being %d\n", __func__, + SCTP_BASE_VAR(sctp_pcb_initialized)); + return (-1); + } + SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator), + SCTP_M_ITER); + if (it == NULL) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM); + return (-1); + } + memset(it, 0, sizeof(*it)); + it->function_assoc = af; + it->function_inp = inpf; + if (inpf) + it->done_current_ep = 0; + else + it->done_current_ep = 1; + it->function_atend = ef; + it->pointer = argp; + it->val = argi; + it->pcb_flags = pcb_state; + it->pcb_features = pcb_features; + it->asoc_state = asoc_state; + it->function_inp_end = inpe; + it->no_chunk_output = chunk_output_off; +#if defined(__FreeBSD__) && !defined(__Userspace__) + it->vn = curvnet; +#endif + if (s_inp) { + /* Assume lock is held here */ + it->inp = s_inp; + SCTP_INP_INCR_REF(it->inp); + it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP; + } else { + SCTP_INP_INFO_RLOCK(); + it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead)); + if (it->inp) { + SCTP_INP_INCR_REF(it->inp); + } + SCTP_INP_INFO_RUNLOCK(); + it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP; + } + SCTP_IPI_ITERATOR_WQ_LOCK(); + if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) { + SCTP_IPI_ITERATOR_WQ_UNLOCK(); + SCTP_PRINTF("%s: rollback on initialize being %d it=%p\n", __func__, + SCTP_BASE_VAR(sctp_pcb_initialized), it); + SCTP_FREE(it, SCTP_M_ITER); + return (-1); + } + TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); + if (sctp_it_ctl.iterator_running == 0) { + sctp_wakeup_iterator(); + } + SCTP_IPI_ITERATOR_WQ_UNLOCK(); + /* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */ + return (0); +} + +/* + * Atomically add flags to the sctp_flags of an inp. + * To be used when the write lock of the inp is not held. + */ +void +sctp_pcb_add_flags(struct sctp_inpcb *inp, uint32_t flags) +{ + uint32_t old_flags, new_flags; + + do { + old_flags = inp->sctp_flags; + new_flags = old_flags | flags; + } while (atomic_cmpset_int(&inp->sctp_flags, old_flags, new_flags) == 0); +} diff --git a/netwerk/sctp/src/netinet/sctp_pcb.h b/netwerk/sctp/src/netinet/sctp_pcb.h new file mode 100644 index 0000000000..71c21886fe --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_pcb.h @@ -0,0 +1,881 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_PCB_H_ +#define _NETINET_SCTP_PCB_H_ + +#include <netinet/sctp_os.h> +#include <netinet/sctp.h> +#include <netinet/sctp_constants.h> +#include <netinet/sctp_sysctl.h> + +LIST_HEAD(sctppcbhead, sctp_inpcb); +LIST_HEAD(sctpasochead, sctp_tcb); +LIST_HEAD(sctpladdr, sctp_laddr); +LIST_HEAD(sctpvtaghead, sctp_tagblock); +LIST_HEAD(sctp_vrflist, sctp_vrf); +LIST_HEAD(sctp_ifnlist, sctp_ifn); +LIST_HEAD(sctp_ifalist, sctp_ifa); +TAILQ_HEAD(sctp_readhead, sctp_queued_to_read); +TAILQ_HEAD(sctp_streamhead, sctp_stream_queue_pending); + +#include <netinet/sctp_structs.h> +#include <netinet/sctp_auth.h> + +#define SCTP_PCBHASH_ALLADDR(port, mask) (port & mask) +#define SCTP_PCBHASH_ASOC(tag, mask) (tag & mask) + +struct sctp_vrf { + LIST_ENTRY (sctp_vrf) next_vrf; + struct sctp_ifalist *vrf_addr_hash; + struct sctp_ifnlist ifnlist; + uint32_t vrf_id; + uint32_t tbl_id_v4; /* default v4 table id */ + uint32_t tbl_id_v6; /* default v6 table id */ + uint32_t total_ifa_count; + u_long vrf_addr_hashmark; + uint32_t refcount; +}; + +struct sctp_ifn { + struct sctp_ifalist ifalist; + struct sctp_vrf *vrf; + LIST_ENTRY(sctp_ifn) next_ifn; + LIST_ENTRY(sctp_ifn) next_bucket; + void *ifn_p; /* never access without appropriate lock */ + uint32_t ifn_mtu; + uint32_t ifn_type; + uint32_t ifn_index; /* shorthand way to look at ifn for reference */ + uint32_t refcount; /* number of reference held should be >= ifa_count */ + uint32_t ifa_count; /* IFA's we hold (in our list - ifalist)*/ + uint32_t num_v6; /* number of v6 addresses */ + uint32_t num_v4; /* number of v4 addresses */ + uint32_t registered_af; /* registered address family for i/f events */ + char ifn_name[SCTP_IFNAMSIZ]; +}; + +/* SCTP local IFA flags */ +#define SCTP_ADDR_VALID 0x00000001 /* its up and active */ +#define SCTP_BEING_DELETED 0x00000002 /* being deleted, + * when refcount = 0. Note + * that it is pulled from the ifn list + * and ifa_p is nulled right away but + * it cannot be freed until the last *net + * pointing to it is deleted. + */ +#define SCTP_ADDR_DEFER_USE 0x00000004 /* Hold off using this one */ +#define SCTP_ADDR_IFA_UNUSEABLE 0x00000008 + +struct sctp_ifa { + LIST_ENTRY(sctp_ifa) next_ifa; + LIST_ENTRY(sctp_ifa) next_bucket; + struct sctp_ifn *ifn_p; /* back pointer to parent ifn */ + void *ifa; /* pointer to ifa, needed for flag + * update for that we MUST lock + * appropriate locks. This is for V6. + */ + union sctp_sockstore address; + uint32_t refcount; /* number of folks referring to this */ + uint32_t flags; + uint32_t localifa_flags; + uint32_t vrf_id; /* vrf_id of this addr (for deleting) */ + uint8_t src_is_loop; + uint8_t src_is_priv; + uint8_t src_is_glob; + uint8_t resv; +}; + +struct sctp_laddr { + LIST_ENTRY(sctp_laddr) sctp_nxt_addr; /* next in list */ + struct sctp_ifa *ifa; + uint32_t action; /* Used during asconf and adding + * if no-zero src-addr selection will + * not consider this address. + */ + struct timeval start_time; /* time when this address was created */ +}; + +struct sctp_block_entry { + int error; +}; + +struct sctp_timewait { + uint32_t tv_sec_at_expire; /* the seconds from boot to expire */ + uint32_t v_tag; /* the vtag that can not be reused */ + uint16_t lport; /* the local port used in vtag */ + uint16_t rport; /* the remote port used in vtag */ +}; + +struct sctp_tagblock { + LIST_ENTRY(sctp_tagblock) sctp_nxt_tagblock; + struct sctp_timewait vtag_block[SCTP_NUMBER_IN_VTAG_BLOCK]; +}; + +struct sctp_epinfo { +#if defined(__FreeBSD__) && !defined(__Userspace__) +#ifdef INET + struct socket *udp4_tun_socket; +#endif +#ifdef INET6 + struct socket *udp6_tun_socket; +#endif +#endif + struct sctpasochead *sctp_asochash; + u_long hashasocmark; + + struct sctppcbhead *sctp_ephash; + u_long hashmark; + + /*- + * The TCP model represents a substantial overhead in that we get an + * additional hash table to keep explicit connections in. The + * listening TCP endpoint will exist in the usual ephash above and + * accept only INIT's. It will be incapable of sending off an INIT. + * When a dg arrives we must look in the normal ephash. If we find a + * TCP endpoint that will tell us to go to the specific endpoint + * hash and re-hash to find the right assoc/socket. If we find a UDP + * model socket we then must complete the lookup. If this fails, + * i.e. no association can be found then we must continue to see if + * a sctp_peeloff()'d socket is in the tcpephash (a spun off socket + * acts like a TCP model connected socket). + */ + struct sctppcbhead *sctp_tcpephash; + u_long hashtcpmark; + uint32_t hashtblsize; + + struct sctp_vrflist *sctp_vrfhash; + u_long hashvrfmark; + + struct sctp_ifnlist *vrf_ifn_hash; + u_long vrf_ifn_hashmark; + + struct sctppcbhead listhead; + struct sctpladdr addr_wq; + +#if defined(__APPLE__) && !defined(__Userspace__) + struct inpcbhead inplisthead; + struct inpcbinfo sctbinfo; +#endif + /* ep zone info */ + sctp_zone_t ipi_zone_ep; + sctp_zone_t ipi_zone_asoc; + sctp_zone_t ipi_zone_laddr; + sctp_zone_t ipi_zone_net; + sctp_zone_t ipi_zone_chunk; + sctp_zone_t ipi_zone_readq; + sctp_zone_t ipi_zone_strmoq; + sctp_zone_t ipi_zone_asconf; + sctp_zone_t ipi_zone_asconf_ack; + +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct rwlock ipi_ep_mtx; + struct mtx ipi_iterator_wq_mtx; + struct rwlock ipi_addr_mtx; + struct mtx ipi_pktlog_mtx; + struct mtx wq_addr_mtx; +#elif defined(SCTP_PROCESS_LEVEL_LOCKS) + userland_rwlock_t ipi_ep_mtx; + userland_rwlock_t ipi_addr_mtx; + userland_mutex_t ipi_count_mtx; + userland_mutex_t ipi_pktlog_mtx; + userland_mutex_t wq_addr_mtx; +#elif defined(__APPLE__) && !defined(__Userspace__) +#ifdef _KERN_LOCKS_H_ + lck_mtx_t *ipi_addr_mtx; + lck_mtx_t *ipi_count_mtx; + lck_mtx_t *ipi_pktlog_mtx; + lck_mtx_t *logging_mtx; + lck_mtx_t *wq_addr_mtx; +#else + void *ipi_count_mtx; + void *logging_mtx; +#endif /* _KERN_LOCKS_H_ */ +#elif defined(_WIN32) && !defined(__Userspace__) + struct rwlock ipi_ep_lock; + struct rwlock ipi_addr_lock; + struct spinlock ipi_pktlog_mtx; + struct rwlock wq_addr_mtx; +#elif defined(__Userspace__) +#endif + uint32_t ipi_count_ep; + + /* assoc/tcb zone info */ + uint32_t ipi_count_asoc; + + /* local addrlist zone info */ + uint32_t ipi_count_laddr; + + /* remote addrlist zone info */ + uint32_t ipi_count_raddr; + + /* chunk structure list for output */ + uint32_t ipi_count_chunk; + + /* socket queue zone info */ + uint32_t ipi_count_readq; + + /* socket queue zone info */ + uint32_t ipi_count_strmoq; + + /* Number of vrfs */ + uint32_t ipi_count_vrfs; + + /* Number of ifns */ + uint32_t ipi_count_ifns; + + /* Number of ifas */ + uint32_t ipi_count_ifas; + + /* system wide number of free chunks hanging around */ + uint32_t ipi_free_chunks; + uint32_t ipi_free_strmoq; + + struct sctpvtaghead vtag_timewait[SCTP_STACK_VTAG_HASH_SIZE]; + + /* address work queue handling */ + struct sctp_timer addr_wq_timer; + +#if defined(_SCTP_NEEDS_CALLOUT_) || defined(_USER_SCTP_NEEDS_CALLOUT_) + struct calloutlist callqueue; +#endif +}; + +struct sctp_base_info { + /* All static structures that + * anchor the system must be here. + */ + struct sctp_epinfo sctppcbinfo; +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT) + struct sctpstat *sctpstat; +#else + struct sctpstat sctpstat; +#endif +#else + struct sctpstat sctpstat; +#endif + struct sctp_sysctl sctpsysctl; + uint8_t first_time; + char sctp_pcb_initialized; +#if defined(SCTP_PACKET_LOGGING) + int packet_log_writers; + int packet_log_end; + uint8_t packet_log_buffer[SCTP_PACKET_LOG_SIZE]; +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + eventhandler_tag eh_tag; +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + int sctp_main_timer_ticks; +#endif +#if defined(__Userspace__) + userland_mutex_t timer_mtx; + userland_thread_t timer_thread; + int timer_thread_should_exit; + int iterator_thread_started; + int timer_thread_started; +#if !defined(_WIN32) + pthread_mutexattr_t mtx_attr; + pthread_rwlockattr_t rwlock_attr; +#if defined(INET) || defined(INET6) + int userspace_route; + userland_thread_t recvthreadroute; +#endif +#endif +#ifdef INET +#if defined(_WIN32) && !defined(__MINGW32__) + SOCKET userspace_rawsctp; + SOCKET userspace_udpsctp; +#else + int userspace_rawsctp; + int userspace_udpsctp; +#endif + userland_thread_t recvthreadraw; + userland_thread_t recvthreadudp; +#endif +#ifdef INET6 +#if defined(_WIN32) && !defined(__MINGW32__) + SOCKET userspace_rawsctp6; + SOCKET userspace_udpsctp6; +#else + int userspace_rawsctp6; + int userspace_udpsctp6; +#endif + userland_thread_t recvthreadraw6; + userland_thread_t recvthreadudp6; +#endif + int (*conn_output)(void *addr, void *buffer, size_t length, uint8_t tos, uint8_t set_df); + void (*debug_printf)(const char *format, ...); + int crc32c_offloaded; +#endif +}; + +/*- + * Here we have all the relevant information for each SCTP entity created. We + * will need to modify this as approprate. We also need to figure out how to + * access /dev/random. + */ +struct sctp_pcb { + unsigned int time_of_secret_change; /* number of seconds from + * timeval.tv_sec */ + uint32_t secret_key[SCTP_HOW_MANY_SECRETS][SCTP_NUMBER_OF_SECRETS]; + unsigned int size_of_a_cookie; + + uint32_t sctp_timeoutticks[SCTP_NUM_TMRS]; + uint32_t sctp_minrto; + uint32_t sctp_maxrto; + uint32_t initial_rto; + uint32_t initial_init_rto_max; + + unsigned int sctp_sack_freq; + uint32_t sctp_sws_sender; + uint32_t sctp_sws_receiver; + + uint32_t sctp_default_cc_module; + uint32_t sctp_default_ss_module; + /* authentication related fields */ + struct sctp_keyhead shared_keys; + sctp_auth_chklist_t *local_auth_chunks; + sctp_hmaclist_t *local_hmacs; + uint16_t default_keyid; + uint32_t default_mtu; + + /* various thresholds */ + /* Max times I will init at a guy */ + uint16_t max_init_times; + + /* Max times I will send before we consider someone dead */ + uint16_t max_send_times; + + uint16_t def_net_failure; + + uint16_t def_net_pf_threshold; + + /* number of streams to pre-open on a association */ + uint16_t pre_open_stream_count; + uint16_t max_open_streams_intome; + + /* random number generator */ + uint32_t random_counter; + uint8_t random_numbers[SCTP_SIGNATURE_ALOC_SIZE]; + uint8_t random_store[SCTP_SIGNATURE_ALOC_SIZE]; + + /* + * This timer is kept running per endpoint. When it fires it will + * change the secret key. The default is once a hour + */ + struct sctp_timer signature_change; + + uint32_t def_cookie_life; + /* defaults to 0 */ + uint32_t auto_close_time; + uint32_t initial_sequence_debug; + uint32_t adaptation_layer_indicator; + uint8_t adaptation_layer_indicator_provided; + uint32_t store_at; + uint32_t max_burst; + uint32_t fr_max_burst; +#ifdef INET6 + uint32_t default_flowlabel; +#endif + uint8_t default_dscp; + char current_secret_number; + char last_secret_number; + uint16_t port; /* remote UDP encapsulation port */ +}; + +#ifndef SCTP_ALIGNMENT +#define SCTP_ALIGNMENT 32 +#endif + +#ifndef SCTP_ALIGNM1 +#define SCTP_ALIGNM1 (SCTP_ALIGNMENT-1) +#endif + +#define sctp_lport ip_inp.inp.inp_lport + +struct sctp_pcbtsn_rlog { + uint32_t vtag; + uint16_t strm; + uint16_t seq; + uint16_t sz; + uint16_t flgs; +}; +#define SCTP_READ_LOG_SIZE 135 /* we choose the number to make a pcb a page */ + +struct sctp_inpcb { + /*- + * put an inpcb in front of it all, kind of a waste but we need to + * for compatibility with all the other stuff. + */ + union { + struct inpcb inp; + char align[(sizeof(struct inpcb) + SCTP_ALIGNM1) & + ~SCTP_ALIGNM1]; + } ip_inp; +#if defined(__APPLE__) && !defined(__Userspace__) + /* leave some space in case i386 inpcb is bigger than ppc */ + uint8_t padding[128]; +#endif + + /* Socket buffer lock protects read_queue and of course sb_cc */ + struct sctp_readhead read_queue; + + LIST_ENTRY(sctp_inpcb) sctp_list; /* lists all endpoints */ + /* hash of all endpoints for model */ + LIST_ENTRY(sctp_inpcb) sctp_hash; + /* count of local addresses bound, 0 if bound all */ + int laddr_count; + + /* list of addrs in use by the EP, NULL if bound-all */ + struct sctpladdr sctp_addr_list; + /* used for source address selection rotation when we are subset bound */ + struct sctp_laddr *next_addr_touse; + + /* back pointer to our socket */ + struct socket *sctp_socket; + uint64_t sctp_features; /* Feature flags */ + uint32_t sctp_flags; /* INP state flag set */ + uint32_t sctp_mobility_features; /* Mobility Feature flags */ + struct sctp_pcb sctp_ep;/* SCTP ep data */ + /* head of the hash of all associations */ + struct sctpasochead *sctp_tcbhash; + u_long sctp_hashmark; + /* head of the list of all associations */ + struct sctpasochead sctp_asoc_list; +#ifdef SCTP_TRACK_FREED_ASOCS + struct sctpasochead sctp_asoc_free_list; +#endif + uint32_t sctp_frag_point; + uint32_t partial_delivery_point; + uint32_t sctp_context; + uint32_t max_cwnd; + uint8_t local_strreset_support; + uint32_t sctp_cmt_on_off; + uint8_t ecn_supported; + uint8_t prsctp_supported; + uint8_t auth_supported; + uint8_t idata_supported; + uint8_t asconf_supported; + uint8_t reconfig_supported; + uint8_t nrsack_supported; + uint8_t pktdrop_supported; + uint8_t zero_checksum; + struct sctp_nonpad_sndrcvinfo def_send; + /*- + * These three are here for the sosend_dgram + * (pkt, pkt_last and control). + * routine. However, I don't think anyone in + * the current FreeBSD kernel calls this. So + * they are candidates with sctp_sendm for + * de-supporting. + */ + struct mbuf *pkt, *pkt_last; + struct mbuf *control; +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct mtx inp_mtx; + struct mtx inp_create_mtx; + struct mtx inp_rdata_mtx; + int32_t refcount; +#elif defined(SCTP_PROCESS_LEVEL_LOCKS) + userland_mutex_t inp_mtx; + userland_mutex_t inp_create_mtx; + userland_mutex_t inp_rdata_mtx; + int32_t refcount; +#elif defined(__APPLE__) && !defined(__Userspace__) +#if defined(SCTP_APPLE_RWLOCK) + lck_rw_t *inp_mtx; +#else + lck_mtx_t *inp_mtx; +#endif + lck_mtx_t *inp_create_mtx; + lck_mtx_t *inp_rdata_mtx; +#elif defined(_WIN32) && !defined(__Userspace__) + struct rwlock inp_lock; + struct spinlock inp_create_lock; + struct spinlock inp_rdata_lock; + int32_t refcount; +#elif defined(__Userspace__) + int32_t refcount; +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + int32_t refcount; + + uint32_t lock_caller1; + uint32_t lock_caller2; + uint32_t lock_caller3; + uint32_t unlock_caller1; + uint32_t unlock_caller2; + uint32_t unlock_caller3; + uint32_t getlock_caller1; + uint32_t getlock_caller2; + uint32_t getlock_caller3; + uint32_t gen_count; + uint32_t lock_gen_count; + uint32_t unlock_gen_count; + uint32_t getlock_gen_count; + + uint32_t i_am_here_file; + uint32_t i_am_here_line; +#endif + uint32_t def_vrf_id; + uint16_t fibnum; +#ifdef SCTP_MVRF + uint32_t *m_vrf_ids; + uint32_t num_vrfs; + uint32_t vrf_size; +#endif + uint32_t total_sends; + uint32_t total_recvs; + uint32_t last_abort_code; + uint32_t total_nospaces; + struct sctpasochead *sctp_asocidhash; + u_long hashasocidmark; + uint32_t sctp_associd_counter; + +#ifdef SCTP_ASOCLOG_OF_TSNS + struct sctp_pcbtsn_rlog readlog[SCTP_READ_LOG_SIZE]; + uint32_t readlog_index; +#endif +#if defined(__Userspace__) + void *ulp_info; + int (*recv_callback)(struct socket *, union sctp_sockstore, void *, size_t, + struct sctp_rcvinfo, int, void *); + uint32_t send_sb_threshold; + int (*send_callback)(struct socket *, uint32_t, void *); +#endif +}; + +#if defined(__Userspace__) +int register_recv_cb (struct socket *, + int (*)(struct socket *, union sctp_sockstore, void *, size_t, + struct sctp_rcvinfo, int, void *)); +int register_send_cb (struct socket *, uint32_t, int (*)(struct socket *, uint32_t, void *)); +int register_ulp_info (struct socket *, void *); +int retrieve_ulp_info (struct socket *, void **); + +#endif +struct sctp_tcb { + struct socket *sctp_socket; /* back pointer to socket */ + struct sctp_inpcb *sctp_ep; /* back pointer to ep */ + LIST_ENTRY(sctp_tcb) sctp_tcbhash; /* next link in hash + * table */ + LIST_ENTRY(sctp_tcb) sctp_tcblist; /* list of all of the + * TCB's */ + LIST_ENTRY(sctp_tcb) sctp_tcbasocidhash; /* next link in asocid + * hash table + */ + LIST_ENTRY(sctp_tcb) sctp_asocs; /* vtag hash list */ + struct sctp_block_entry *block_entry; /* pointer locked by socket + * send buffer */ + struct sctp_association asoc; + /*- + * freed_by_sorcv_sincelast is protected by the sockbuf_lock NOT the + * tcb_lock. Its special in this way to help avoid extra mutex calls + * in the reading of data. + */ + uint32_t freed_by_sorcv_sincelast; + uint32_t total_sends; + uint32_t total_recvs; + int freed_from_where; + uint16_t rport; /* remote port in network format */ + uint16_t resv; +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct mtx tcb_mtx; +#elif defined(SCTP_PROCESS_LEVEL_LOCKS) + userland_mutex_t tcb_mtx; +#elif defined(__APPLE__) && !defined(__Userspace__) + lck_mtx_t* tcb_mtx; +#elif defined(_WIN32) && !defined(__Userspace__) + struct spinlock tcb_lock; +#elif defined(__Userspace__) +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + uint32_t caller1; + uint32_t caller2; + uint32_t caller3; +#endif +}; + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <netinet/sctp_lock_bsd.h> +#elif defined(__APPLE__) && !defined(__Userspace__) +/* + * Apple MacOS X 10.4 "Tiger" + */ + +#include <netinet/sctp_lock_apple_fg.h> + +#elif defined(SCTP_PROCESS_LEVEL_LOCKS) + +#include <netinet/sctp_process_lock.h> + +#elif defined(_WIN32) && !defined(__Userspace__) + +#include <netinet/sctp_lock_windows.h> + +#elif defined(__Userspace__) + +#include <netinet/sctp_lock_userspace.h> + +#else +/* + * Pre-5.x FreeBSD, and others. + */ +#include <netinet/sctp_lock_empty.h> +#endif + +#if defined(_KERNEL) || defined(__Userspace__) + +/* Attention Julian, this is the extern that + * goes with the base info. sctp_pcb.c has + * the real definition. + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) +VNET_DECLARE(struct sctp_base_info, system_base_info); +#else +extern struct sctp_base_info system_base_info; +#endif + +#ifdef INET6 +int SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b); +#endif + +void sctp_fill_pcbinfo(struct sctp_pcbinfo *); + +struct sctp_ifn * +sctp_find_ifn(void *ifn, uint32_t ifn_index); + +struct sctp_vrf *sctp_allocate_vrf(int vrfid); +struct sctp_vrf *sctp_find_vrf(uint32_t vrfid); +void sctp_free_vrf(struct sctp_vrf *vrf); + +/*- + * Change address state, can be used if + * O/S supports telling transports about + * changes to IFA/IFN's (link layer triggers). + * If a ifn goes down, we will do src-addr-selection + * and NOT use that, as a source address. This does + * not stop the routing system from routing out + * that interface, but we won't put it as a source. + */ +void sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr, const char *if_name, uint32_t ifn_index); +void sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr, const char *if_name, uint32_t ifn_index); + +struct sctp_ifa * +sctp_add_addr_to_vrf(uint32_t vrfid, + void *ifn, uint32_t ifn_index, uint32_t ifn_type, + const char *if_name, + void *ifa, struct sockaddr *addr, uint32_t ifa_flags, + int dynamic_add); + +void sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu); + +void sctp_free_ifn(struct sctp_ifn *sctp_ifnp); +void sctp_free_ifa(struct sctp_ifa *sctp_ifap); + +void sctp_del_addr_from_vrf(uint32_t vrfid, struct sockaddr *addr, + uint32_t ifn_index, const char *if_name); + +struct sctp_nets *sctp_findnet(struct sctp_tcb *, struct sockaddr *); + +struct sctp_inpcb *sctp_pcb_findep(struct sockaddr *, int, int, uint32_t); + +#if defined(__FreeBSD__) && !defined(__Userspace__) +int +sctp_inpcb_bind(struct socket *, struct sockaddr *, + struct sctp_ifa *, struct thread *); +int +sctp_inpcb_bind_locked(struct sctp_inpcb *, struct sockaddr *, + struct sctp_ifa *, struct thread *); +#elif defined(_WIN32) && !defined(__Userspace__) +int +sctp_inpcb_bind(struct socket *, struct sockaddr *, + struct sctp_ifa *, PKTHREAD); +int +sctp_inpcb_bind_locked(struct sctp_inpcb *, struct sockaddr *, + struct sctp_ifa *, PKTHREAD); +#else +/* struct proc is a dummy for __Userspace__ */ +int +sctp_inpcb_bind(struct socket *, struct sockaddr *, + struct sctp_ifa *, struct proc *); +int +sctp_inpcb_bind_locked(struct sctp_inpcb *, struct sockaddr *, + struct sctp_ifa *, struct proc *); +#endif + +struct sctp_tcb * +sctp_findassociation_addr(struct mbuf *, int, + struct sockaddr *, struct sockaddr *, + struct sctphdr *, struct sctp_chunkhdr *, struct sctp_inpcb **, + struct sctp_nets **, uint32_t vrf_id); + +struct sctp_tcb * +sctp_findassociation_addr_sa(struct sockaddr *, + struct sockaddr *, struct sctp_inpcb **, struct sctp_nets **, int, uint32_t); + +void +sctp_move_pcb_and_assoc(struct sctp_inpcb *, struct sctp_inpcb *, + struct sctp_tcb *); + +/*- + * For this call ep_addr, the to is the destination endpoint address of the + * peer (relative to outbound). The from field is only used if the TCP model + * is enabled and helps distingush amongst the subset bound (non-boundall). + * The TCP model MAY change the actual ep field, this is why it is passed. + */ +struct sctp_tcb * +sctp_findassociation_ep_addr(struct sctp_inpcb **, + struct sockaddr *, struct sctp_nets **, struct sockaddr *, + struct sctp_tcb *); + +struct sctp_tcb * +sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock); + +struct sctp_tcb * +sctp_findassociation_ep_asocid(struct sctp_inpcb *, + sctp_assoc_t, int); + +struct sctp_tcb * +sctp_findassociation_ep_asconf(struct mbuf *, int, struct sockaddr *, + struct sctphdr *, struct sctp_inpcb **, struct sctp_nets **, uint32_t vrf_id); + +int sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id); + +int sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id); + +void sctp_inpcb_free(struct sctp_inpcb *, int, int); + +#define SCTP_DONT_INITIALIZE_AUTH_PARAMS 0 +#define SCTP_INITIALIZE_AUTH_PARAMS 1 + +#if defined(__FreeBSD__) && !defined(__Userspace__) +struct sctp_tcb * +sctp_aloc_assoc(struct sctp_inpcb *, struct sockaddr *, + int *, uint32_t, uint32_t, uint32_t, uint16_t, uint16_t, + struct thread *, int); +struct sctp_tcb * +sctp_aloc_assoc_connected(struct sctp_inpcb *, struct sockaddr *, + int *, uint32_t, uint32_t, uint32_t, uint16_t, uint16_t, + struct thread *, int); +#elif defined(_WIN32) && !defined(__Userspace__) +struct sctp_tcb * +sctp_aloc_assoc(struct sctp_inpcb *, struct sockaddr *, int *, uint32_t, + uint32_t, uint32_t, uint16_t, uint16_t, PKTHREAD, int); +struct sctp_tcb * +sctp_aloc_assoc_connected(struct sctp_inpcb *, struct sockaddr *, int *, uint32_t, + uint32_t, uint32_t, uint16_t, uint16_t, PKTHREAD, int); +#else +/* proc will be NULL for __Userspace__ */ +struct sctp_tcb * +sctp_aloc_assoc(struct sctp_inpcb *, struct sockaddr *, int *, uint32_t, + uint32_t, uint32_t, uint16_t, uint16_t, struct proc *, int); +struct sctp_tcb * +sctp_aloc_assoc_connected(struct sctp_inpcb *, struct sockaddr *, int *, uint32_t, + uint32_t, uint32_t, uint16_t, uint16_t, struct proc *, int); +#endif + +int sctp_free_assoc(struct sctp_inpcb *, struct sctp_tcb *, int, int); + +void sctp_add_local_addr_ep(struct sctp_inpcb *, struct sctp_ifa *, uint32_t); + +void sctp_del_local_addr_ep(struct sctp_inpcb *, struct sctp_ifa *); + +int sctp_add_remote_addr(struct sctp_tcb *, struct sockaddr *, struct sctp_nets **, uint16_t, int, int); + +void sctp_remove_net(struct sctp_tcb *, struct sctp_nets *); + +int sctp_del_remote_addr(struct sctp_tcb *, struct sockaddr *); + +#if defined(__Userspace__) +void sctp_pcb_init(int); +#else +void sctp_pcb_init(void); +#endif + +void sctp_pcb_finish(void); + +void sctp_add_local_addr_restricted(struct sctp_tcb *, struct sctp_ifa *); +void sctp_del_local_addr_restricted(struct sctp_tcb *, struct sctp_ifa *); + +int +sctp_load_addresses_from_init(struct sctp_tcb *, struct mbuf *, int, int, + struct sockaddr *, struct sockaddr *, struct sockaddr *, uint16_t); + +int +sctp_set_primary_addr(struct sctp_tcb *, struct sockaddr *, + struct sctp_nets *); + +bool +sctp_is_vtag_good(uint32_t, uint16_t lport, uint16_t rport, struct timeval *); + +int sctp_destination_is_reachable(struct sctp_tcb *, struct sockaddr *); + +int sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp); + +void sctp_clean_up_stream(struct sctp_tcb *stcb, struct sctp_readhead *rh); + +void +sctp_pcb_add_flags(struct sctp_inpcb *, uint32_t); + +/*- + * Null in last arg inpcb indicate run on ALL ep's. Specific inp in last arg + * indicates run on ONLY assoc's of the specified endpoint. + */ +int +sctp_initiate_iterator(inp_func inpf, + asoc_func af, + inp_func inpe, + uint32_t, uint32_t, + uint32_t, void *, + uint32_t, + end_func ef, + struct sctp_inpcb *, + uint8_t co_off); +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SCTP_MCORE_INPUT) && defined(SMP) +void +sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use); + +#endif +#endif + +#endif /* _KERNEL */ +#endif /* !__sctp_pcb_h__ */ diff --git a/netwerk/sctp/src/netinet/sctp_peeloff.c b/netwerk/sctp/src/netinet/sctp_peeloff.c new file mode 100644 index 0000000000..548c3d82b2 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_peeloff.c @@ -0,0 +1,302 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_var.h> +#include <netinet/sctp_var.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp.h> +#include <netinet/sctp_uio.h> +#include <netinet/sctp_peeloff.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_auth.h> + +int +sctp_can_peel_off(struct socket *head, sctp_assoc_t assoc_id) +{ + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; + uint32_t state; + + if (head == NULL) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PEELOFF, EBADF); + return (EBADF); + } + inp = (struct sctp_inpcb *)head->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PEELOFF, EFAULT); + return (EFAULT); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PEELOFF, EOPNOTSUPP); + return (EOPNOTSUPP); + } + stcb = sctp_findassociation_ep_asocid(inp, assoc_id, 1); + if (stcb == NULL) { + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PEELOFF, ENOENT); + return (ENOENT); + } + state = SCTP_GET_STATE(stcb); + if ((state == SCTP_STATE_EMPTY) || + (state == SCTP_STATE_INUSE)) { + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PEELOFF, ENOTCONN); + return (ENOTCONN); + } + SCTP_TCB_UNLOCK(stcb); + /* We are clear to peel this one off */ + return (0); +} + +int +sctp_do_peeloff(struct socket *head, struct socket *so, sctp_assoc_t assoc_id) +{ + struct sctp_inpcb *inp, *n_inp; + struct sctp_tcb *stcb; + uint32_t state; + + inp = (struct sctp_inpcb *)head->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PEELOFF, EFAULT); + return (EFAULT); + } + stcb = sctp_findassociation_ep_asocid(inp, assoc_id, 1); + if (stcb == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PEELOFF, ENOTCONN); + return (ENOTCONN); + } + + state = SCTP_GET_STATE(stcb); + if ((state == SCTP_STATE_EMPTY) || + (state == SCTP_STATE_INUSE)) { + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PEELOFF, ENOTCONN); + return (ENOTCONN); + } + + n_inp = (struct sctp_inpcb *)so->so_pcb; + n_inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE | + SCTP_PCB_FLAGS_CONNECTED | + SCTP_PCB_FLAGS_IN_TCPPOOL | /* Turn on Blocking IO */ + (SCTP_PCB_COPY_FLAGS & inp->sctp_flags)); + n_inp->sctp_socket = so; + n_inp->sctp_features = inp->sctp_features; + n_inp->sctp_mobility_features = inp->sctp_mobility_features; + n_inp->sctp_frag_point = inp->sctp_frag_point; + n_inp->sctp_cmt_on_off = inp->sctp_cmt_on_off; + n_inp->ecn_supported = inp->ecn_supported; + n_inp->prsctp_supported = inp->prsctp_supported; + n_inp->auth_supported = inp->auth_supported; + n_inp->asconf_supported = inp->asconf_supported; + n_inp->reconfig_supported = inp->reconfig_supported; + n_inp->nrsack_supported = inp->nrsack_supported; + n_inp->pktdrop_supported = inp->pktdrop_supported; + n_inp->partial_delivery_point = inp->partial_delivery_point; + n_inp->sctp_context = inp->sctp_context; + n_inp->max_cwnd = inp->max_cwnd; + n_inp->local_strreset_support = inp->local_strreset_support; + /* copy in the authentication parameters from the original endpoint */ + if (n_inp->sctp_ep.local_hmacs) + sctp_free_hmaclist(n_inp->sctp_ep.local_hmacs); + n_inp->sctp_ep.local_hmacs = + sctp_copy_hmaclist(inp->sctp_ep.local_hmacs); + if (n_inp->sctp_ep.local_auth_chunks) + sctp_free_chunklist(n_inp->sctp_ep.local_auth_chunks); + n_inp->sctp_ep.local_auth_chunks = + sctp_copy_chunklist(inp->sctp_ep.local_auth_chunks); + (void)sctp_copy_skeylist(&inp->sctp_ep.shared_keys, + &n_inp->sctp_ep.shared_keys); +#if defined(__Userspace__) + n_inp->ulp_info = inp->ulp_info; + n_inp->recv_callback = inp->recv_callback; + n_inp->send_callback = inp->send_callback; + n_inp->send_sb_threshold = inp->send_sb_threshold; +#endif + /* + * Now we must move it from one hash table to another and get the + * stcb in the right place. + */ + sctp_move_pcb_and_assoc(inp, n_inp, stcb); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + +#if defined(__FreeBSD__) && !defined(__Userspace__) + sctp_pull_off_control_to_new_inp(inp, n_inp, stcb, SBL_WAIT); +#else + sctp_pull_off_control_to_new_inp(inp, n_inp, stcb, M_WAITOK); +#endif + atomic_subtract_int(&stcb->asoc.refcnt, 1); + + return (0); +} + +#if defined(HAVE_SCTP_PEELOFF_SOCKOPT) +struct socket * +sctp_get_peeloff(struct socket *head, sctp_assoc_t assoc_id, int *error) +{ + struct socket *newso; + struct sctp_inpcb *inp, *n_inp; + struct sctp_tcb *stcb; + + SCTPDBG(SCTP_DEBUG_PEEL1, "SCTP peel-off called\n"); + inp = (struct sctp_inpcb *)head->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PEELOFF, EFAULT); + *error = EFAULT; + return (NULL); + } + stcb = sctp_findassociation_ep_asocid(inp, assoc_id, 1); + if (stcb == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PEELOFF, ENOTCONN); + *error = ENOTCONN; + return (NULL); + } + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); +#if defined(__FreeBSD__) && !defined(__Userspace__) + CURVNET_SET(head->so_vnet); +#endif + newso = sonewconn(head, SS_ISCONNECTED +#if defined(__APPLE__) && !defined(__Userspace__) + , NULL +#endif + ); +#if defined(__FreeBSD__) && !defined(__Userspace__) + CURVNET_RESTORE(); +#endif + if (newso == NULL) { + SCTPDBG(SCTP_DEBUG_PEEL1, "sctp_peeloff:sonewconn failed\n"); + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_PEELOFF, ENOMEM); + *error = ENOMEM; + atomic_subtract_int(&stcb->asoc.refcnt, 1); + return (NULL); + + } +#if defined(__APPLE__) && !defined(__Userspace__) + else { + SCTP_SOCKET_LOCK(newso, 1); + } +#endif + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + n_inp = (struct sctp_inpcb *)newso->so_pcb; + SOCK_LOCK(head); + n_inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE | + SCTP_PCB_FLAGS_CONNECTED | + SCTP_PCB_FLAGS_IN_TCPPOOL | /* Turn on Blocking IO */ + (SCTP_PCB_COPY_FLAGS & inp->sctp_flags)); + n_inp->sctp_features = inp->sctp_features; + n_inp->sctp_frag_point = inp->sctp_frag_point; + n_inp->sctp_cmt_on_off = inp->sctp_cmt_on_off; + n_inp->ecn_supported = inp->ecn_supported; + n_inp->prsctp_supported = inp->prsctp_supported; + n_inp->auth_supported = inp->auth_supported; + n_inp->asconf_supported = inp->asconf_supported; + n_inp->reconfig_supported = inp->reconfig_supported; + n_inp->nrsack_supported = inp->nrsack_supported; + n_inp->pktdrop_supported = inp->pktdrop_supported; + n_inp->partial_delivery_point = inp->partial_delivery_point; + n_inp->sctp_context = inp->sctp_context; + n_inp->max_cwnd = inp->max_cwnd; + n_inp->local_strreset_support = inp->local_strreset_support; + n_inp->inp_starting_point_for_iterator = NULL; +#if defined(__Userspace__) + n_inp->ulp_info = inp->ulp_info; + n_inp->recv_callback = inp->recv_callback; + n_inp->send_callback = inp->send_callback; + n_inp->send_sb_threshold = inp->send_sb_threshold; +#endif + + /* copy in the authentication parameters from the original endpoint */ + if (n_inp->sctp_ep.local_hmacs) + sctp_free_hmaclist(n_inp->sctp_ep.local_hmacs); + n_inp->sctp_ep.local_hmacs = + sctp_copy_hmaclist(inp->sctp_ep.local_hmacs); + if (n_inp->sctp_ep.local_auth_chunks) + sctp_free_chunklist(n_inp->sctp_ep.local_auth_chunks); + n_inp->sctp_ep.local_auth_chunks = + sctp_copy_chunklist(inp->sctp_ep.local_auth_chunks); + (void)sctp_copy_skeylist(&inp->sctp_ep.shared_keys, + &n_inp->sctp_ep.shared_keys); + + n_inp->sctp_socket = newso; + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) { + sctp_feature_off(n_inp, SCTP_PCB_FLAGS_AUTOCLOSE); + n_inp->sctp_ep.auto_close_time = 0; + sctp_timer_stop(SCTP_TIMER_TYPE_AUTOCLOSE, n_inp, stcb, NULL, + SCTP_FROM_SCTP_PEELOFF + SCTP_LOC_1); + } + /* Turn off any non-blocking semantic. */ + SOCK_LOCK(newso); + SCTP_CLEAR_SO_NBIO(newso); + newso->so_state |= SS_ISCONNECTED; + SOCK_UNLOCK(newso); + /* We remove it right away */ + +#ifdef SCTP_LOCK_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) { + sctp_log_lock(inp, (struct sctp_tcb *)NULL, SCTP_LOG_LOCK_SOCK); + } +#endif + TAILQ_REMOVE(&head->so_comp, newso, so_list); + head->so_qlen--; + SOCK_UNLOCK(head); + /* + * Now we must move it from one hash table to another and get the + * stcb in the right place. + */ + sctp_move_pcb_and_assoc(inp, n_inp, stcb); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + /* + * And now the final hack. We move data in the pending side i.e. + * head to the new socket buffer. Let the GRUBBING begin :-0 + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + sctp_pull_off_control_to_new_inp(inp, n_inp, stcb, SBL_WAIT); +#else + sctp_pull_off_control_to_new_inp(inp, n_inp, stcb, M_WAITOK); +#endif + atomic_subtract_int(&stcb->asoc.refcnt, 1); + return (newso); +} +#endif diff --git a/netwerk/sctp/src/netinet/sctp_peeloff.h b/netwerk/sctp/src/netinet/sctp_peeloff.h new file mode 100644 index 0000000000..c22bfca929 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_peeloff.h @@ -0,0 +1,70 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_PEELOFF_H_ +#define _NETINET_SCTP_PEELOFF_H_ +#if defined(HAVE_SCTP_PEELOFF_SOCKOPT) +/* socket option peeloff */ +struct sctp_peeloff_opt { +#if !(defined(_WIN32) && !defined(__Userspace__)) + int s; +#else + HANDLE s; +#endif + sctp_assoc_t assoc_id; +#if !(defined(_WIN32) && !defined(__Userspace__)) + int new_sd; +#else + HANDLE new_sd; +#endif +}; +#endif /* HAVE_SCTP_PEELOFF_SOCKOPT */ +#if defined(_KERNEL) +int sctp_can_peel_off(struct socket *, sctp_assoc_t); +int sctp_do_peeloff(struct socket *, struct socket *, sctp_assoc_t); +#if defined(HAVE_SCTP_PEELOFF_SOCKOPT) +struct socket *sctp_get_peeloff(struct socket *, sctp_assoc_t, int *); +int sctp_peeloff_option(struct proc *p, struct sctp_peeloff_opt *peeloff); +#endif /* HAVE_SCTP_PEELOFF_SOCKOPT */ +#endif /* _KERNEL */ +#if defined(__Userspace__) +int sctp_can_peel_off(struct socket *, sctp_assoc_t); +int sctp_do_peeloff(struct socket *, struct socket *, sctp_assoc_t); +#endif /* __Userspace__ */ +#endif /* _NETINET_SCTP_PEELOFF_H_ */ diff --git a/netwerk/sctp/src/netinet/sctp_process_lock.h b/netwerk/sctp/src/netinet/sctp_process_lock.h new file mode 100644 index 0000000000..165c6e4eab --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_process_lock.h @@ -0,0 +1,689 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2011, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2011, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ +#ifndef __sctp_process_lock_h__ +#define __sctp_process_lock_h__ + +/* + * Need to yet define five atomic fuctions or + * their equivalant. + * - atomic_add_int(&foo, val) - add atomically the value + * - atomic_fetchadd_int(&foo, val) - does same as atomic_add_int + * but value it was is returned. + * - atomic_subtract_int(&foo, val) - can be made from atomic_add_int() + * + * - atomic_cmpset_int(&foo, value, newvalue) - Does a set of newvalue + * in foo if and only if + * foo is value. Returns 0 + * on success. + */ + +#ifdef SCTP_PER_SOCKET_LOCKING +/* + * per socket level locking + */ + +#if defined(_WIN32) +/* Lock for INFO stuff */ +#define SCTP_INP_INFO_LOCK_INIT() +#define SCTP_INP_INFO_RLOCK() +#define SCTP_INP_INFO_RUNLOCK() +#define SCTP_INP_INFO_WLOCK() +#define SCTP_INP_INFO_WUNLOCK() +#define SCTP_INP_INFO_LOCK_ASSERT() +#define SCTP_INP_INFO_RLOCK_ASSERT() +#define SCTP_INP_INFO_WLOCK_ASSERT() +#define SCTP_INP_INFO_LOCK_DESTROY() +#define SCTP_IPI_COUNT_INIT() +#define SCTP_IPI_COUNT_DESTROY() +#else +#define SCTP_INP_INFO_LOCK_INIT() +#define SCTP_INP_INFO_RLOCK() +#define SCTP_INP_INFO_RUNLOCK() +#define SCTP_INP_INFO_WLOCK() +#define SCTP_INP_INFO_WUNLOCK() +#define SCTP_INP_INFO_LOCK_ASSERT() +#define SCTP_INP_INFO_RLOCK_ASSERT() +#define SCTP_INP_INFO_WLOCK_ASSERT() +#define SCTP_INP_INFO_LOCK_DESTROY() +#define SCTP_IPI_COUNT_INIT() +#define SCTP_IPI_COUNT_DESTROY() +#endif + +/* Lock for INP */ +#define SCTP_INP_LOCK_INIT(_inp) +#define SCTP_INP_LOCK_DESTROY(_inp) + +#define SCTP_INP_RLOCK(_inp) +#define SCTP_INP_RUNLOCK(_inp) +#define SCTP_INP_WLOCK(_inp) +#define SCTP_INP_WUNLOCK(_inp) +#define SCTP_INP_RLOCK_ASSERT(_inp) +#define SCTP_INP_WLOCK_ASSERT(_inp) +#define SCTP_INP_INCR_REF(_inp) +#define SCTP_INP_DECR_REF(_inp) + +#define SCTP_ASOC_CREATE_LOCK_INIT(_inp) +#define SCTP_ASOC_CREATE_LOCK_DESTROY(_inp) +#define SCTP_ASOC_CREATE_LOCK(_inp) +#define SCTP_ASOC_CREATE_UNLOCK(_inp) + +#define SCTP_INP_READ_INIT(_inp) +#define SCTP_INP_READ_DESTROY(_inp) +#define SCTP_INP_READ_LOCK(_inp) +#define SCTP_INP_READ_UNLOCK(_inp) + +/* Lock for TCB */ +#define SCTP_TCB_LOCK_INIT(_tcb) +#define SCTP_TCB_LOCK_DESTROY(_tcb) +#define SCTP_TCB_LOCK(_tcb) +#define SCTP_TCB_TRYLOCK(_tcb) 1 +#define SCTP_TCB_UNLOCK(_tcb) +#define SCTP_TCB_UNLOCK_IFOWNED(_tcb) +#define SCTP_TCB_LOCK_ASSERT(_tcb) + +#else +/* + * per tcb level locking + */ +#define SCTP_IPI_COUNT_INIT() + +#if defined(_WIN32) +#define SCTP_WQ_ADDR_INIT() \ + InitializeCriticalSection(&SCTP_BASE_INFO(wq_addr_mtx)) +#define SCTP_WQ_ADDR_DESTROY() \ + DeleteCriticalSection(&SCTP_BASE_INFO(wq_addr_mtx)) +#define SCTP_WQ_ADDR_LOCK() \ + EnterCriticalSection(&SCTP_BASE_INFO(wq_addr_mtx)) +#define SCTP_WQ_ADDR_UNLOCK() \ + LeaveCriticalSection(&SCTP_BASE_INFO(wq_addr_mtx)) +#define SCTP_WQ_ADDR_LOCK_ASSERT() + +#if WINVER < 0x0600 +#define SCTP_INP_INFO_LOCK_INIT() \ + InitializeCriticalSection(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_LOCK_DESTROY() \ + DeleteCriticalSection(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_RLOCK() \ + EnterCriticalSection(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_TRYLOCK() \ + TryEnterCriticalSection(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_WLOCK() \ + EnterCriticalSection(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_RUNLOCK() \ + LeaveCriticalSection(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_WUNLOCK() \ + LeaveCriticalSection(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_LOCK_ASSERT() +#define SCTP_INP_INFO_RLOCK_ASSERT() +#define SCTP_INP_INFO_WLOCK_ASSERT() +#else +#define SCTP_INP_INFO_LOCK_INIT() \ + InitializeSRWLock(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_LOCK_DESTROY() +#define SCTP_INP_INFO_RLOCK() \ + AcquireSRWLockShared(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_TRYLOCK() \ + TryAcquireSRWLockShared(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_WLOCK() \ + AcquireSRWLockExclusive(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_RUNLOCK() \ + ReleaseSRWLockShared(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_WUNLOCK() \ + ReleaseSRWLockExclusive(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_LOCK_ASSERT() +#define SCTP_INP_INFO_RLOCK_ASSERT() +#define SCTP_INP_INFO_WLOCK_ASSERT() +#endif + +#define SCTP_IP_PKTLOG_INIT() \ + InitializeCriticalSection(&SCTP_BASE_INFO(ipi_pktlog_mtx)) +#define SCTP_IP_PKTLOG_DESTROY () \ + DeleteCriticalSection(&SCTP_BASE_INFO(ipi_pktlog_mtx)) +#define SCTP_IP_PKTLOG_LOCK() \ + EnterCriticalSection(&SCTP_BASE_INFO(ipi_pktlog_mtx)) +#define SCTP_IP_PKTLOG_UNLOCK() \ + LeaveCriticalSection(&SCTP_BASE_INFO(ipi_pktlog_mtx)) + +/* + * The INP locks we will use for locking an SCTP endpoint, so for example if + * we want to change something at the endpoint level for example random_store + * or cookie secrets we lock the INP level. + */ +#define SCTP_INP_READ_INIT(_inp) \ + InitializeCriticalSection(&(_inp)->inp_rdata_mtx) +#define SCTP_INP_READ_DESTROY(_inp) \ + DeleteCriticalSection(&(_inp)->inp_rdata_mtx) +#define SCTP_INP_READ_LOCK(_inp) \ + EnterCriticalSection(&(_inp)->inp_rdata_mtx) +#define SCTP_INP_READ_UNLOCK(_inp) \ + LeaveCriticalSection(&(_inp)->inp_rdata_mtx) + +#define SCTP_INP_LOCK_INIT(_inp) \ + InitializeCriticalSection(&(_inp)->inp_mtx) +#define SCTP_INP_LOCK_DESTROY(_inp) \ + DeleteCriticalSection(&(_inp)->inp_mtx) +#ifdef SCTP_LOCK_LOGGING +#define SCTP_INP_RLOCK(_inp) do { \ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) \ + sctp_log_lock(_inp, NULL, SCTP_LOG_LOCK_INP); \ + EnterCriticalSection(&(_inp)->inp_mtx); \ +} while (0) +#define SCTP_INP_WLOCK(_inp) do { \ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) \ + sctp_log_lock(_inp, NULL, SCTP_LOG_LOCK_INP); \ + EnterCriticalSection(&(_inp)->inp_mtx); \ +} while (0) +#else +#define SCTP_INP_RLOCK(_inp) \ + EnterCriticalSection(&(_inp)->inp_mtx) +#define SCTP_INP_WLOCK(_inp) \ + EnterCriticalSection(&(_inp)->inp_mtx) +#endif +#define SCTP_INP_RLOCK_ASSERT(_tcb) +#define SCTP_INP_WLOCK_ASSERT(_tcb) + +#define SCTP_INP_INCR_REF(_inp) atomic_add_int(&((_inp)->refcount), 1) +#define SCTP_INP_DECR_REF(_inp) atomic_subtract_int(&((_inp)->refcount), 1) + +#define SCTP_ASOC_CREATE_LOCK_INIT(_inp) \ + InitializeCriticalSection(&(_inp)->inp_create_mtx) +#define SCTP_ASOC_CREATE_LOCK_DESTROY(_inp) \ + DeleteCriticalSection(&(_inp)->inp_create_mtx) +#ifdef SCTP_LOCK_LOGGING +#define SCTP_ASOC_CREATE_LOCK(_inp) do { \ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) \ + sctp_log_lock(_inp, NULL, SCTP_LOG_LOCK_CREATE); \ + EnterCriticalSection(&(_inp)->inp_create_mtx); \ +} while (0) +#else +#define SCTP_ASOC_CREATE_LOCK(_inp) \ + EnterCriticalSection(&(_inp)->inp_create_mtx) +#endif + +#define SCTP_INP_RUNLOCK(_inp) \ + LeaveCriticalSection(&(_inp)->inp_mtx) +#define SCTP_INP_WUNLOCK(_inp) \ + LeaveCriticalSection(&(_inp)->inp_mtx) +#define SCTP_ASOC_CREATE_UNLOCK(_inp) \ + LeaveCriticalSection(&(_inp)->inp_create_mtx) + +/* + * For the majority of things (once we have found the association) we will + * lock the actual association mutex. This will protect all the assoiciation + * level queues and streams and such. We will need to lock the socket layer + * when we stuff data up into the receiving sb_mb. I.e. we will need to do an + * extra SOCKBUF_LOCK(&so->so_rcv) even though the association is locked. + */ + +#define SCTP_TCB_LOCK_INIT(_tcb) \ + InitializeCriticalSection(&(_tcb)->tcb_mtx) +#define SCTP_TCB_LOCK_DESTROY(_tcb) \ + DeleteCriticalSection(&(_tcb)->tcb_mtx) +#ifdef SCTP_LOCK_LOGGING +#define SCTP_TCB_LOCK(_tcb) do { \ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) \ + sctp_log_lock(_tcb->sctp_ep, _tcb, SCTP_LOG_LOCK_TCB); \ + EnterCriticalSection(&(_tcb)->tcb_mtx); \ +} while (0) +#else +#define SCTP_TCB_LOCK(_tcb) \ + EnterCriticalSection(&(_tcb)->tcb_mtx) +#endif +#define SCTP_TCB_TRYLOCK(_tcb) ((TryEnterCriticalSection(&(_tcb)->tcb_mtx))) +#define SCTP_TCB_UNLOCK(_tcb) \ + LeaveCriticalSection(&(_tcb)->tcb_mtx) +#define SCTP_TCB_LOCK_ASSERT(_tcb) + +#else /* all Userspaces except Windows */ +#define SCTP_WQ_ADDR_INIT() \ + (void)pthread_mutex_init(&SCTP_BASE_INFO(wq_addr_mtx), &SCTP_BASE_VAR(mtx_attr)) +#define SCTP_WQ_ADDR_DESTROY() \ + (void)pthread_mutex_destroy(&SCTP_BASE_INFO(wq_addr_mtx)) +#ifdef INVARIANTS +#define SCTP_WQ_ADDR_LOCK() \ + KASSERT(pthread_mutex_lock(&SCTP_BASE_INFO(wq_addr_mtx)) == 0, ("%s:%d: wq_addr_mtx already locked", __FILE__, __LINE__)) +#define SCTP_WQ_ADDR_UNLOCK() \ + KASSERT(pthread_mutex_unlock(&SCTP_BASE_INFO(wq_addr_mtx)) == 0, ("%s:%d: wq_addr_mtx not locked", __FILE__, __LINE__)) +#else +#define SCTP_WQ_ADDR_LOCK() \ + (void)pthread_mutex_lock(&SCTP_BASE_INFO(wq_addr_mtx)) +#define SCTP_WQ_ADDR_UNLOCK() \ + (void)pthread_mutex_unlock(&SCTP_BASE_INFO(wq_addr_mtx)) +#endif +#define SCTP_WQ_ADDR_LOCK_ASSERT() \ + KASSERT(pthread_mutex_trylock(&SCTP_BASE_INFO(wq_addr_mtx)) == EBUSY, ("%s:%d: wq_addr_mtx not locked", __FILE__, __LINE__)) + +#define SCTP_INP_INFO_LOCK_INIT() \ + (void)pthread_rwlock_init(&SCTP_BASE_INFO(ipi_ep_mtx), &SCTP_BASE_VAR(rwlock_attr)) +#define SCTP_INP_INFO_LOCK_DESTROY() \ + (void)pthread_rwlock_destroy(&SCTP_BASE_INFO(ipi_ep_mtx)) +#ifdef INVARIANTS +#define SCTP_INP_INFO_RLOCK() \ + KASSERT(pthread_rwlock_rdlock(&SCTP_BASE_INFO(ipi_ep_mtx)) == 0, ("%s%d: ipi_ep_mtx already locked", __FILE__, __LINE__)) +#define SCTP_INP_INFO_WLOCK() \ + KASSERT(pthread_rwlock_wrlock(&SCTP_BASE_INFO(ipi_ep_mtx)) == 0, ("%s:%d: ipi_ep_mtx already locked", __FILE__, __LINE__)) +#define SCTP_INP_INFO_RUNLOCK() \ + KASSERT(pthread_rwlock_unlock(&SCTP_BASE_INFO(ipi_ep_mtx)) == 0, ("%s:%d: ipi_ep_mtx not locked", __FILE__, __LINE__)) +#define SCTP_INP_INFO_WUNLOCK() \ + KASSERT(pthread_rwlock_unlock(&SCTP_BASE_INFO(ipi_ep_mtx)) == 0, ("%s:%d: ipi_ep_mtx not locked", __FILE__, __LINE__)) +#else +#define SCTP_INP_INFO_RLOCK() \ + (void)pthread_rwlock_rdlock(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_WLOCK() \ + (void)pthread_rwlock_wrlock(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_RUNLOCK() \ + (void)pthread_rwlock_unlock(&SCTP_BASE_INFO(ipi_ep_mtx)) +#define SCTP_INP_INFO_WUNLOCK() \ + (void)pthread_rwlock_unlock(&SCTP_BASE_INFO(ipi_ep_mtx)) +#endif +#define SCTP_INP_INFO_LOCK_ASSERT() +#define SCTP_INP_INFO_RLOCK_ASSERT() +#define SCTP_INP_INFO_WLOCK_ASSERT() +#define SCTP_INP_INFO_TRYLOCK() \ + (!(pthread_rwlock_tryrdlock(&SCTP_BASE_INFO(ipi_ep_mtx)))) + +#define SCTP_IP_PKTLOG_INIT() \ + (void)pthread_mutex_init(&SCTP_BASE_INFO(ipi_pktlog_mtx), &SCTP_BASE_VAR(mtx_attr)) +#define SCTP_IP_PKTLOG_DESTROY() \ + (void)pthread_mutex_destroy(&SCTP_BASE_INFO(ipi_pktlog_mtx)) +#ifdef INVARIANTS +#define SCTP_IP_PKTLOG_LOCK() \ + KASSERT(pthread_mutex_lock(&SCTP_BASE_INFO(ipi_pktlog_mtx)) == 0, ("%s:%d: ipi_pktlog_mtx already locked", __FILE__, __LINE__)) +#define SCTP_IP_PKTLOG_UNLOCK() \ + KASSERT(pthread_mutex_unlock(&SCTP_BASE_INFO(ipi_pktlog_mtx)) == 0, ("%s:%d: ipi_pktlog_mtx not locked", __FILE__, __LINE__)) +#else +#define SCTP_IP_PKTLOG_LOCK() \ + (void)pthread_mutex_lock(&SCTP_BASE_INFO(ipi_pktlog_mtx)) +#define SCTP_IP_PKTLOG_UNLOCK() \ + (void)pthread_mutex_unlock(&SCTP_BASE_INFO(ipi_pktlog_mtx)) +#endif + + +/* + * The INP locks we will use for locking an SCTP endpoint, so for example if + * we want to change something at the endpoint level for example random_store + * or cookie secrets we lock the INP level. + */ +#define SCTP_INP_READ_INIT(_inp) \ + (void)pthread_mutex_init(&(_inp)->inp_rdata_mtx, &SCTP_BASE_VAR(mtx_attr)) +#define SCTP_INP_READ_DESTROY(_inp) \ + (void)pthread_mutex_destroy(&(_inp)->inp_rdata_mtx) +#ifdef INVARIANTS +#define SCTP_INP_READ_LOCK(_inp) \ + KASSERT(pthread_mutex_lock(&(_inp)->inp_rdata_mtx) == 0, ("%s:%d: inp_rdata_mtx already locked", __FILE__, __LINE__)) +#define SCTP_INP_READ_UNLOCK(_inp) \ + KASSERT(pthread_mutex_unlock(&(_inp)->inp_rdata_mtx) == 0, ("%s:%d: inp_rdata_mtx not locked", __FILE__, __LINE__)) +#else +#define SCTP_INP_READ_LOCK(_inp) \ + (void)pthread_mutex_lock(&(_inp)->inp_rdata_mtx) +#define SCTP_INP_READ_UNLOCK(_inp) \ + (void)pthread_mutex_unlock(&(_inp)->inp_rdata_mtx) +#endif + +#define SCTP_INP_LOCK_INIT(_inp) \ + (void)pthread_mutex_init(&(_inp)->inp_mtx, &SCTP_BASE_VAR(mtx_attr)) +#define SCTP_INP_LOCK_DESTROY(_inp) \ + (void)pthread_mutex_destroy(&(_inp)->inp_mtx) +#ifdef INVARIANTS +#ifdef SCTP_LOCK_LOGGING +#define SCTP_INP_RLOCK(_inp) do { \ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) \ + sctp_log_lock(_inp, NULL, SCTP_LOG_LOCK_INP); \ + KASSERT(pthread_mutex_lock(&(_inp)->inp_mtx) == 0, ("%s:%d: inp_mtx already locked", __FILE__, __LINE__)); \ +} while (0) +#define SCTP_INP_WLOCK(_inp) do { \ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) \ + sctp_log_lock(_inp, NULL, SCTP_LOG_LOCK_INP); \ + KASSERT(pthread_mutex_lock(&(_inp)->inp_mtx) == 0, ("%s:%d: inp_mtx already locked", __FILE__, __LINE__)); \ +} while (0) +#else +#define SCTP_INP_RLOCK(_inp) \ + KASSERT(pthread_mutex_lock(&(_inp)->inp_mtx) == 0, ("%s:%d: inp_mtx already locked", __FILE__, __LINE__)) +#define SCTP_INP_WLOCK(_inp) \ + KASSERT(pthread_mutex_lock(&(_inp)->inp_mtx) == 0, ("%s:%d: inp_mtx already locked", __FILE__, __LINE__)) +#endif +#define SCTP_INP_RUNLOCK(_inp) \ + KASSERT(pthread_mutex_unlock(&(_inp)->inp_mtx) == 0, ("%s:%d: inp_mtx not locked", __FILE__, __LINE__)) +#define SCTP_INP_WUNLOCK(_inp) \ + KASSERT(pthread_mutex_unlock(&(_inp)->inp_mtx) == 0, ("%s:%d: inp_mtx not locked", __FILE__, __LINE__)) +#else +#ifdef SCTP_LOCK_LOGGING +#define SCTP_INP_RLOCK(_inp) do { \ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) \ + sctp_log_lock(_inp, NULL, SCTP_LOG_LOCK_INP); \ + (void)pthread_mutex_lock(&(_inp)->inp_mtx); \ +} while (0) +#define SCTP_INP_WLOCK(_inp) do { \ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) \ + sctp_log_lock(_inp, NULL, SCTP_LOG_LOCK_INP); \ + (void)pthread_mutex_lock(&(_inp)->inp_mtx); \ +} while (0) +#else +#define SCTP_INP_RLOCK(_inp) \ + (void)pthread_mutex_lock(&(_inp)->inp_mtx) +#define SCTP_INP_WLOCK(_inp) \ + (void)pthread_mutex_lock(&(_inp)->inp_mtx) +#endif +#define SCTP_INP_RUNLOCK(_inp) \ + (void)pthread_mutex_unlock(&(_inp)->inp_mtx) +#define SCTP_INP_WUNLOCK(_inp) \ + (void)pthread_mutex_unlock(&(_inp)->inp_mtx) +#endif +#define SCTP_INP_RLOCK_ASSERT(_inp) \ + KASSERT(pthread_mutex_trylock(&(_inp)->inp_mtx) == EBUSY, ("%s:%d: inp_mtx not locked", __FILE__, __LINE__)) +#define SCTP_INP_WLOCK_ASSERT(_inp) \ + KASSERT(pthread_mutex_trylock(&(_inp)->inp_mtx) == EBUSY, ("%s:%d: inp_mtx not locked", __FILE__, __LINE__)) +#define SCTP_INP_INCR_REF(_inp) atomic_add_int(&((_inp)->refcount), 1) +#define SCTP_INP_DECR_REF(_inp) atomic_subtract_int(&((_inp)->refcount), 1) + +#define SCTP_ASOC_CREATE_LOCK_INIT(_inp) \ + (void)pthread_mutex_init(&(_inp)->inp_create_mtx, &SCTP_BASE_VAR(mtx_attr)) +#define SCTP_ASOC_CREATE_LOCK_DESTROY(_inp) \ + (void)pthread_mutex_destroy(&(_inp)->inp_create_mtx) +#ifdef INVARIANTS +#ifdef SCTP_LOCK_LOGGING +#define SCTP_ASOC_CREATE_LOCK(_inp) do { \ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) \ + sctp_log_lock(_inp, NULL, SCTP_LOG_LOCK_CREATE); \ + KASSERT(pthread_mutex_lock(&(_inp)->inp_create_mtx) == 0, ("%s:%d: inp_create_mtx already locked", __FILE__, __LINE__)); \ +} while (0) +#else +#define SCTP_ASOC_CREATE_LOCK(_inp) \ + KASSERT(pthread_mutex_lock(&(_inp)->inp_create_mtx) == 0, ("%s:%d: inp_create_mtx already locked", __FILE__, __LINE__)) +#endif +#define SCTP_ASOC_CREATE_UNLOCK(_inp) \ + KASSERT(pthread_mutex_unlock(&(_inp)->inp_create_mtx) == 0, ("%s:%d: inp_create_mtx not locked", __FILE__, __LINE__)) +#else +#ifdef SCTP_LOCK_LOGGING +#define SCTP_ASOC_CREATE_LOCK(_inp) do { \ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) \ + sctp_log_lock(_inp, NULL, SCTP_LOG_LOCK_CREATE); \ + (void)pthread_mutex_lock(&(_inp)->inp_create_mtx); \ +} while (0) +#else +#define SCTP_ASOC_CREATE_LOCK(_inp) \ + (void)pthread_mutex_lock(&(_inp)->inp_create_mtx) +#endif +#define SCTP_ASOC_CREATE_UNLOCK(_inp) \ + (void)pthread_mutex_unlock(&(_inp)->inp_create_mtx) +#endif +/* + * For the majority of things (once we have found the association) we will + * lock the actual association mutex. This will protect all the assoiciation + * level queues and streams and such. We will need to lock the socket layer + * when we stuff data up into the receiving sb_mb. I.e. we will need to do an + * extra SOCKBUF_LOCK(&so->so_rcv) even though the association is locked. + */ + +#define SCTP_TCB_LOCK_INIT(_tcb) \ + (void)pthread_mutex_init(&(_tcb)->tcb_mtx, &SCTP_BASE_VAR(mtx_attr)) +#define SCTP_TCB_LOCK_DESTROY(_tcb) \ + (void)pthread_mutex_destroy(&(_tcb)->tcb_mtx) +#ifdef INVARIANTS +#ifdef SCTP_LOCK_LOGGING +#define SCTP_TCB_LOCK(_tcb) do { \ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) \ + sctp_log_lock(_tcb->sctp_ep, _tcb, SCTP_LOG_LOCK_TCB); \ + KASSERT(pthread_mutex_lock(&(_tcb)->tcb_mtx) == 0, ("%s:%d: tcb_mtx already locked", __FILE__, __LINE__)) \ +} while (0) +#else +#define SCTP_TCB_LOCK(_tcb) \ + KASSERT(pthread_mutex_lock(&(_tcb)->tcb_mtx) == 0, ("%s:%d: tcb_mtx already locked", __FILE__, __LINE__)) +#endif +#define SCTP_TCB_UNLOCK(_tcb) \ + KASSERT(pthread_mutex_unlock(&(_tcb)->tcb_mtx) == 0, ("%s:%d: tcb_mtx not locked", __FILE__, __LINE__)) +#else +#ifdef SCTP_LOCK_LOGGING +#define SCTP_TCB_LOCK(_tcb) do { \ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) \ + sctp_log_lock(_tcb->sctp_ep, _tcb, SCTP_LOG_LOCK_TCB); \ + (void)pthread_mutex_lock(&(_tcb)->tcb_mtx); \ +} while (0) +#else +#define SCTP_TCB_LOCK(_tcb) \ + (void)pthread_mutex_lock(&(_tcb)->tcb_mtx) +#endif +#define SCTP_TCB_UNLOCK(_tcb) (void)pthread_mutex_unlock(&(_tcb)->tcb_mtx) +#endif +#define SCTP_TCB_LOCK_ASSERT(_tcb) \ + KASSERT(pthread_mutex_trylock(&(_tcb)->tcb_mtx) == EBUSY, ("%s:%d: tcb_mtx not locked", __FILE__, __LINE__)) +#define SCTP_TCB_TRYLOCK(_tcb) (!(pthread_mutex_trylock(&(_tcb)->tcb_mtx))) +#endif + +#endif /* SCTP_PER_SOCKET_LOCKING */ + + +/* + * common locks + */ + +/* copied over to compile */ +#define SCTP_INP_LOCK_CONTENDED(_inp) (0) /* Don't know if this is possible */ +#define SCTP_INP_READ_CONTENDED(_inp) (0) /* Don't know if this is possible */ +#define SCTP_ASOC_CREATE_LOCK_CONTENDED(_inp) (0) /* Don't know if this is possible */ + +/* socket locks */ + +#if defined(_WIN32) +#define SOCKBUF_LOCK_ASSERT(_so_buf) +#define SOCKBUF_LOCK(_so_buf) \ + EnterCriticalSection(&(_so_buf)->sb_mtx) +#define SOCKBUF_UNLOCK(_so_buf) \ + LeaveCriticalSection(&(_so_buf)->sb_mtx) +#define SOCK_LOCK(_so) \ + SOCKBUF_LOCK(&(_so)->so_rcv) +#define SOCK_UNLOCK(_so) \ + SOCKBUF_UNLOCK(&(_so)->so_rcv) +#else +#define SOCKBUF_LOCK_ASSERT(_so_buf) \ + KASSERT(pthread_mutex_trylock(SOCKBUF_MTX(_so_buf)) == EBUSY, ("%s:%d: socket buffer not locked", __FILE__, __LINE__)) +#ifdef INVARIANTS +#define SOCKBUF_LOCK(_so_buf) \ + KASSERT(pthread_mutex_lock(SOCKBUF_MTX(_so_buf)) == 0, ("%s:%d: sockbuf_mtx already locked", __FILE__, __LINE__)) +#define SOCKBUF_UNLOCK(_so_buf) \ + KASSERT(pthread_mutex_unlock(SOCKBUF_MTX(_so_buf)) == 0, ("%s:%d: sockbuf_mtx not locked", __FILE__, __LINE__)) +#else +#define SOCKBUF_LOCK(_so_buf) \ + pthread_mutex_lock(SOCKBUF_MTX(_so_buf)) +#define SOCKBUF_UNLOCK(_so_buf) \ + pthread_mutex_unlock(SOCKBUF_MTX(_so_buf)) +#endif +#define SOCK_LOCK(_so) \ + SOCKBUF_LOCK(&(_so)->so_rcv) +#define SOCK_UNLOCK(_so) \ + SOCKBUF_UNLOCK(&(_so)->so_rcv) +#endif + +#define SCTP_STATLOG_INIT_LOCK() +#define SCTP_STATLOG_LOCK() +#define SCTP_STATLOG_UNLOCK() +#define SCTP_STATLOG_DESTROY() + +#if defined(_WIN32) +/* address list locks */ +#if WINVER < 0x0600 +#define SCTP_IPI_ADDR_INIT() \ + InitializeCriticalSection(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_DESTROY() \ + DeleteCriticalSection(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_RLOCK() \ + EnterCriticalSection(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_RUNLOCK() \ + LeaveCriticalSection(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_WLOCK() \ + EnterCriticalSection(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_WUNLOCK() \ + LeaveCriticalSection(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_LOCK_ASSERT() +#define SCTP_IPI_ADDR_WLOCK_ASSERT() +#else +#define SCTP_IPI_ADDR_INIT() \ + InitializeSRWLock(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_DESTROY() +#define SCTP_IPI_ADDR_RLOCK() \ + AcquireSRWLockShared(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_RUNLOCK() \ + ReleaseSRWLockShared(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_WLOCK() \ + AcquireSRWLockExclusive(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_WUNLOCK() \ + ReleaseSRWLockExclusive(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_LOCK_ASSERT() +#define SCTP_IPI_ADDR_WLOCK_ASSERT() +#endif + +/* iterator locks */ +#define SCTP_ITERATOR_LOCK_INIT() \ + InitializeCriticalSection(&sctp_it_ctl.it_mtx) +#define SCTP_ITERATOR_LOCK_DESTROY() \ + DeleteCriticalSection(&sctp_it_ctl.it_mtx) +#define SCTP_ITERATOR_LOCK() \ + EnterCriticalSection(&sctp_it_ctl.it_mtx) +#define SCTP_ITERATOR_UNLOCK() \ + LeaveCriticalSection(&sctp_it_ctl.it_mtx) + +#define SCTP_IPI_ITERATOR_WQ_INIT() \ + InitializeCriticalSection(&sctp_it_ctl.ipi_iterator_wq_mtx) +#define SCTP_IPI_ITERATOR_WQ_DESTROY() \ + DeleteCriticalSection(&sctp_it_ctl.ipi_iterator_wq_mtx) +#define SCTP_IPI_ITERATOR_WQ_LOCK() \ + EnterCriticalSection(&sctp_it_ctl.ipi_iterator_wq_mtx) +#define SCTP_IPI_ITERATOR_WQ_UNLOCK() \ + LeaveCriticalSection(&sctp_it_ctl.ipi_iterator_wq_mtx) + +#else +/* address list locks */ +#define SCTP_IPI_ADDR_INIT() \ + (void)pthread_rwlock_init(&SCTP_BASE_INFO(ipi_addr_mtx), &SCTP_BASE_VAR(rwlock_attr)) +#define SCTP_IPI_ADDR_DESTROY() \ + (void)pthread_rwlock_destroy(&SCTP_BASE_INFO(ipi_addr_mtx)) +#ifdef INVARIANTS +#define SCTP_IPI_ADDR_RLOCK() \ + KASSERT(pthread_rwlock_rdlock(&SCTP_BASE_INFO(ipi_addr_mtx)) == 0, ("%s:%d: ipi_addr_mtx already locked", __FILE__, __LINE__)) +#define SCTP_IPI_ADDR_RUNLOCK() \ + KASSERT(pthread_rwlock_unlock(&SCTP_BASE_INFO(ipi_addr_mtx)) == 0, ("%s:%d: ipi_addr_mtx not locked", __FILE__, __LINE__)) +#define SCTP_IPI_ADDR_WLOCK() \ + KASSERT(pthread_rwlock_wrlock(&SCTP_BASE_INFO(ipi_addr_mtx)) == 0, ("%s:%d: ipi_addr_mtx already locked", __FILE__, __LINE__)) +#define SCTP_IPI_ADDR_WUNLOCK() \ + KASSERT(pthread_rwlock_unlock(&SCTP_BASE_INFO(ipi_addr_mtx)) == 0, ("%s:%d: ipi_addr_mtx not locked", __FILE__, __LINE__)) +#else +#define SCTP_IPI_ADDR_RLOCK() \ + (void)pthread_rwlock_rdlock(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_RUNLOCK() \ + (void)pthread_rwlock_unlock(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_WLOCK() \ + (void)pthread_rwlock_wrlock(&SCTP_BASE_INFO(ipi_addr_mtx)) +#define SCTP_IPI_ADDR_WUNLOCK() \ + (void)pthread_rwlock_unlock(&SCTP_BASE_INFO(ipi_addr_mtx)) +#endif +#define SCTP_IPI_ADDR_LOCK_ASSERT() +#define SCTP_IPI_ADDR_WLOCK_ASSERT() + +/* iterator locks */ +#define SCTP_ITERATOR_LOCK_INIT() \ + (void)pthread_mutex_init(&sctp_it_ctl.it_mtx, &SCTP_BASE_VAR(mtx_attr)) +#define SCTP_ITERATOR_LOCK_DESTROY() \ + (void)pthread_mutex_destroy(&sctp_it_ctl.it_mtx) +#ifdef INVARIANTS +#define SCTP_ITERATOR_LOCK() \ + KASSERT(pthread_mutex_lock(&sctp_it_ctl.it_mtx) == 0, ("%s:%d: it_mtx already locked", __FILE__, __LINE__)) +#define SCTP_ITERATOR_UNLOCK() \ + KASSERT(pthread_mutex_unlock(&sctp_it_ctl.it_mtx) == 0, ("%s:%d: it_mtx not locked", __FILE__, __LINE__)) +#else +#define SCTP_ITERATOR_LOCK() \ + (void)pthread_mutex_lock(&sctp_it_ctl.it_mtx) +#define SCTP_ITERATOR_UNLOCK() \ + (void)pthread_mutex_unlock(&sctp_it_ctl.it_mtx) +#endif + +#define SCTP_IPI_ITERATOR_WQ_INIT() \ + (void)pthread_mutex_init(&sctp_it_ctl.ipi_iterator_wq_mtx, &SCTP_BASE_VAR(mtx_attr)) +#define SCTP_IPI_ITERATOR_WQ_DESTROY() \ + (void)pthread_mutex_destroy(&sctp_it_ctl.ipi_iterator_wq_mtx) +#ifdef INVARIANTS +#define SCTP_IPI_ITERATOR_WQ_LOCK() \ + KASSERT(pthread_mutex_lock(&sctp_it_ctl.ipi_iterator_wq_mtx) == 0, ("%s:%d: ipi_iterator_wq_mtx already locked", __FILE__, __LINE__)) +#define SCTP_IPI_ITERATOR_WQ_UNLOCK() \ + KASSERT(pthread_mutex_unlock(&sctp_it_ctl.ipi_iterator_wq_mtx) == 0, ("%s:%d: ipi_iterator_wq_mtx not locked", __FILE__, __LINE__)) +#else +#define SCTP_IPI_ITERATOR_WQ_LOCK() \ + (void)pthread_mutex_lock(&sctp_it_ctl.ipi_iterator_wq_mtx) +#define SCTP_IPI_ITERATOR_WQ_UNLOCK() \ + (void)pthread_mutex_unlock(&sctp_it_ctl.ipi_iterator_wq_mtx) +#endif +#endif + +#define SCTP_INCR_EP_COUNT() \ + atomic_add_int(&SCTP_BASE_INFO(ipi_count_ep), 1) + +#define SCTP_DECR_EP_COUNT() \ + atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ep), 1) + +#define SCTP_INCR_ASOC_COUNT() \ + atomic_add_int(&SCTP_BASE_INFO(ipi_count_asoc), 1) + +#define SCTP_DECR_ASOC_COUNT() \ + atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_asoc), 1) + +#define SCTP_INCR_LADDR_COUNT() \ + atomic_add_int(&SCTP_BASE_INFO(ipi_count_laddr), 1) + +#define SCTP_DECR_LADDR_COUNT() \ + atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_laddr), 1) + +#define SCTP_INCR_RADDR_COUNT() \ + atomic_add_int(&SCTP_BASE_INFO(ipi_count_raddr), 1) + +#define SCTP_DECR_RADDR_COUNT() \ + atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_raddr), 1) + +#define SCTP_INCR_CHK_COUNT() \ + atomic_add_int(&SCTP_BASE_INFO(ipi_count_chunk), 1) + +#define SCTP_DECR_CHK_COUNT() \ + atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_chunk), 1) + +#define SCTP_INCR_READQ_COUNT() \ + atomic_add_int(&SCTP_BASE_INFO(ipi_count_readq), 1) + +#define SCTP_DECR_READQ_COUNT() \ + atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_readq), 1) + +#define SCTP_INCR_STRMOQ_COUNT() \ + atomic_add_int(&SCTP_BASE_INFO(ipi_count_strmoq), 1) + +#define SCTP_DECR_STRMOQ_COUNT() \ + atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_strmoq), 1) + +#endif diff --git a/netwerk/sctp/src/netinet/sctp_sha1.c b/netwerk/sctp/src/netinet/sctp_sha1.c new file mode 100644 index 0000000000..8472c3a1c0 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_sha1.c @@ -0,0 +1,347 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2013, by Michael Tuexen. All rights reserved. + * Copyright (c) 2013, by Lally Singh. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include <netinet/sctp_sha1.h> + +#if defined(SCTP_USE_NSS_SHA1) +/* A SHA-1 Digest is 160 bits, or 20 bytes */ +#define SHA_DIGEST_LENGTH (20) + +void +sctp_sha1_init(struct sctp_sha1_context *ctx) +{ + ctx->pk11_ctx = PK11_CreateDigestContext(SEC_OID_SHA1); + PK11_DigestBegin(ctx->pk11_ctx); +} + +void +sctp_sha1_update(struct sctp_sha1_context *ctx, const unsigned char *ptr, unsigned int siz) +{ + PK11_DigestOp(ctx->pk11_ctx, ptr, siz); +} + +void +sctp_sha1_final(unsigned char *digest, struct sctp_sha1_context *ctx) +{ + unsigned int output_len = 0; + + PK11_DigestFinal(ctx->pk11_ctx, digest, &output_len, SHA_DIGEST_LENGTH); + PK11_DestroyContext(ctx->pk11_ctx, PR_TRUE); +} + +#elif defined(SCTP_USE_OPENSSL_SHA1) + +void +sctp_sha1_init(struct sctp_sha1_context *ctx) +{ + SHA1_Init(&ctx->sha_ctx); +} + +void +sctp_sha1_update(struct sctp_sha1_context *ctx, const unsigned char *ptr, unsigned int siz) +{ + SHA1_Update(&ctx->sha_ctx, ptr, (unsigned long)siz); +} + +void +sctp_sha1_final(unsigned char *digest, struct sctp_sha1_context *ctx) +{ + SHA1_Final(digest, &ctx->sha_ctx); +} +#elif defined(SCTP_USE_MBEDTLS_SHA1) +void +sctp_sha1_init(struct sctp_sha1_context *ctx) +{ + mbedtls_sha1_init(&ctx->sha1_ctx); + mbedtls_sha1_starts_ret(&ctx->sha1_ctx); +} + +void +sctp_sha1_update(struct sctp_sha1_context *ctx, const unsigned char *ptr, unsigned int siz) +{ + mbedtls_sha1_update_ret(&ctx->sha1_ctx, ptr, siz); +} + +void +sctp_sha1_final(unsigned char *digest, struct sctp_sha1_context *ctx) +{ + mbedtls_sha1_finish_ret(&ctx->sha1_ctx, digest); +} +#else + +#include <string.h> +#if defined(_WIN32) && defined(__Userspace__) +#include <winsock2.h> +#elif !(defined(_WIN32) && !defined(__Userspace__)) +#include <arpa/inet.h> +#endif + +#define F1(B,C,D) (((B & C) | ((~B) & D))) /* 0 <= t <= 19 */ +#define F2(B,C,D) (B ^ C ^ D) /* 20 <= t <= 39 */ +#define F3(B,C,D) ((B & C) | (B & D) | (C & D)) /* 40 <= t <= 59 */ +#define F4(B,C,D) (B ^ C ^ D) /* 600 <= t <= 79 */ + +/* circular shift */ +#define CSHIFT(A,B) ((B << A) | (B >> (32-A))) + +#define K1 0x5a827999 /* 0 <= t <= 19 */ +#define K2 0x6ed9eba1 /* 20 <= t <= 39 */ +#define K3 0x8f1bbcdc /* 40 <= t <= 59 */ +#define K4 0xca62c1d6 /* 60 <= t <= 79 */ + +#define H0INIT 0x67452301 +#define H1INIT 0xefcdab89 +#define H2INIT 0x98badcfe +#define H3INIT 0x10325476 +#define H4INIT 0xc3d2e1f0 + +void +sctp_sha1_init(struct sctp_sha1_context *ctx) +{ + /* Init the SHA-1 context structure */ + ctx->A = 0; + ctx->B = 0; + ctx->C = 0; + ctx->D = 0; + ctx->E = 0; + ctx->H0 = H0INIT; + ctx->H1 = H1INIT; + ctx->H2 = H2INIT; + ctx->H3 = H3INIT; + ctx->H4 = H4INIT; + ctx->TEMP = 0; + memset(ctx->words, 0, sizeof(ctx->words)); + ctx->how_many_in_block = 0; + ctx->running_total = 0; +} + +static void +sctp_sha1_process_a_block(struct sctp_sha1_context *ctx, unsigned int *block) +{ + int i; + + /* init the W0-W15 to the block of words being hashed. */ + /* step a) */ + for (i = 0; i < 16; i++) { + ctx->words[i] = ntohl(block[i]); + } + /* now init the rest based on the SHA-1 formula, step b) */ + for (i = 16; i < 80; i++) { + ctx->words[i] = CSHIFT(1, ((ctx->words[(i - 3)]) ^ + (ctx->words[(i - 8)]) ^ + (ctx->words[(i - 14)]) ^ + (ctx->words[(i - 16)]))); + } + /* step c) */ + ctx->A = ctx->H0; + ctx->B = ctx->H1; + ctx->C = ctx->H2; + ctx->D = ctx->H3; + ctx->E = ctx->H4; + + /* step d) */ + for (i = 0; i < 80; i++) { + if (i < 20) { + ctx->TEMP = ((CSHIFT(5, ctx->A)) + + (F1(ctx->B, ctx->C, ctx->D)) + + (ctx->E) + + ctx->words[i] + + K1); + } else if (i < 40) { + ctx->TEMP = ((CSHIFT(5, ctx->A)) + + (F2(ctx->B, ctx->C, ctx->D)) + + (ctx->E) + + (ctx->words[i]) + + K2); + } else if (i < 60) { + ctx->TEMP = ((CSHIFT(5, ctx->A)) + + (F3(ctx->B, ctx->C, ctx->D)) + + (ctx->E) + + (ctx->words[i]) + + K3); + } else { + ctx->TEMP = ((CSHIFT(5, ctx->A)) + + (F4(ctx->B, ctx->C, ctx->D)) + + (ctx->E) + + (ctx->words[i]) + + K4); + } + ctx->E = ctx->D; + ctx->D = ctx->C; + ctx->C = CSHIFT(30, ctx->B); + ctx->B = ctx->A; + ctx->A = ctx->TEMP; + } + /* step e) */ + ctx->H0 = (ctx->H0) + (ctx->A); + ctx->H1 = (ctx->H1) + (ctx->B); + ctx->H2 = (ctx->H2) + (ctx->C); + ctx->H3 = (ctx->H3) + (ctx->D); + ctx->H4 = (ctx->H4) + (ctx->E); +} + +void +sctp_sha1_update(struct sctp_sha1_context *ctx, const unsigned char *ptr, unsigned int siz) +{ + unsigned int number_left, left_to_fill; + + number_left = siz; + while (number_left > 0) { + left_to_fill = sizeof(ctx->sha_block) - ctx->how_many_in_block; + if (left_to_fill > number_left) { + /* can only partially fill up this one */ + memcpy(&ctx->sha_block[ctx->how_many_in_block], + ptr, number_left); + ctx->how_many_in_block += number_left; + ctx->running_total += number_left; + break; + } else { + /* block is now full, process it */ + memcpy(&ctx->sha_block[ctx->how_many_in_block], + ptr, left_to_fill); + sctp_sha1_process_a_block(ctx, + (unsigned int *)ctx->sha_block); + number_left -= left_to_fill; + ctx->running_total += left_to_fill; + ctx->how_many_in_block = 0; + ptr = (const unsigned char *)(ptr + left_to_fill); + } + } +} + +void +sctp_sha1_final(unsigned char *digest, struct sctp_sha1_context *ctx) +{ + /* + * if any left in block fill with padding and process. Then transfer + * the digest to the pointer. At the last block some special rules + * need to apply. We must add a 1 bit following the message, then we + * pad with 0's. The total size is encoded as a 64 bit number at the + * end. Now if the last buffer has more than 55 octets in it we + * cannot fit the 64 bit number + 10000000 pad on the end and must + * add the 10000000 pad, pad the rest of the message with 0's and + * then create an all 0 message with just the 64 bit size at the end + * and run this block through by itself. Also the 64 bit int must + * be in network byte order. + */ + int left_to_fill; + unsigned int i, *ptr; + + if (ctx->how_many_in_block > 55) { + /* + * special case, we need to process two blocks here. One for + * the current stuff plus possibly the pad. The other for + * the size. + */ + left_to_fill = sizeof(ctx->sha_block) - ctx->how_many_in_block; + if (left_to_fill == 0) { + /* Should not really happen but I am paranoid */ + sctp_sha1_process_a_block(ctx, + (unsigned int *)ctx->sha_block); + /* init last block, a bit different than the rest */ + ctx->sha_block[0] = '\x80'; + for (i = 1; i < sizeof(ctx->sha_block); i++) { + ctx->sha_block[i] = 0x0; + } + } else if (left_to_fill == 1) { + ctx->sha_block[ctx->how_many_in_block] = '\x80'; + sctp_sha1_process_a_block(ctx, + (unsigned int *)ctx->sha_block); + /* init last block */ + memset(ctx->sha_block, 0, sizeof(ctx->sha_block)); + } else { + ctx->sha_block[ctx->how_many_in_block] = '\x80'; + for (i = (ctx->how_many_in_block + 1); + i < sizeof(ctx->sha_block); + i++) { + ctx->sha_block[i] = 0x0; + } + sctp_sha1_process_a_block(ctx, + (unsigned int *)ctx->sha_block); + /* init last block */ + memset(ctx->sha_block, 0, sizeof(ctx->sha_block)); + } + /* This is in bits so multiply by 8 */ + ctx->running_total *= 8; + ptr = (unsigned int *)&ctx->sha_block[60]; + *ptr = htonl(ctx->running_total); + sctp_sha1_process_a_block(ctx, (unsigned int *)ctx->sha_block); + } else { + /* + * easy case, we just pad this message to size - end with 0 + * add the magic 0x80 to the next word and then put the + * network byte order size in the last spot and process the + * block. + */ + ctx->sha_block[ctx->how_many_in_block] = '\x80'; + for (i = (ctx->how_many_in_block + 1); + i < sizeof(ctx->sha_block); + i++) { + ctx->sha_block[i] = 0x0; + } + /* get last int spot */ + ctx->running_total *= 8; + ptr = (unsigned int *)&ctx->sha_block[60]; + *ptr = htonl(ctx->running_total); + sctp_sha1_process_a_block(ctx, (unsigned int *)ctx->sha_block); + } + /* transfer the digest back to the user */ + digest[3] = (ctx->H0 & 0xff); + digest[2] = ((ctx->H0 >> 8) & 0xff); + digest[1] = ((ctx->H0 >> 16) & 0xff); + digest[0] = ((ctx->H0 >> 24) & 0xff); + + digest[7] = (ctx->H1 & 0xff); + digest[6] = ((ctx->H1 >> 8) & 0xff); + digest[5] = ((ctx->H1 >> 16) & 0xff); + digest[4] = ((ctx->H1 >> 24) & 0xff); + + digest[11] = (ctx->H2 & 0xff); + digest[10] = ((ctx->H2 >> 8) & 0xff); + digest[9] = ((ctx->H2 >> 16) & 0xff); + digest[8] = ((ctx->H2 >> 24) & 0xff); + + digest[15] = (ctx->H3 & 0xff); + digest[14] = ((ctx->H3 >> 8) & 0xff); + digest[13] = ((ctx->H3 >> 16) & 0xff); + digest[12] = ((ctx->H3 >> 24) & 0xff); + + digest[19] = (ctx->H4 & 0xff); + digest[18] = ((ctx->H4 >> 8) & 0xff); + digest[17] = ((ctx->H4 >> 16) & 0xff); + digest[16] = ((ctx->H4 >> 24) & 0xff); +} + +#endif diff --git a/netwerk/sctp/src/netinet/sctp_sha1.h b/netwerk/sctp/src/netinet/sctp_sha1.h new file mode 100644 index 0000000000..9ff4ff7bdc --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_sha1.h @@ -0,0 +1,94 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + + +#ifndef __NETINET_SCTP_SHA1_H__ +#define __NETINET_SCTP_SHA1_H__ + +#include <sys/types.h> +#if defined(SCTP_USE_NSS_SHA1) +#include <pk11pub.h> +#elif defined(SCTP_USE_OPENSSL_SHA1) +#include <openssl/sha.h> +#elif defined(SCTP_USE_MBEDTLS_SHA1) +#include <mbedtls/sha1.h> +#endif + +struct sctp_sha1_context { +#if defined(SCTP_USE_NSS_SHA1) + struct PK11Context *pk11_ctx; +#elif defined(SCTP_USE_OPENSSL_SHA1) + SHA_CTX sha_ctx; +#elif defined(SCTP_USE_MBEDTLS_SHA1) + mbedtls_sha1_context sha1_ctx; +#else + unsigned int A; + unsigned int B; + unsigned int C; + unsigned int D; + unsigned int E; + unsigned int H0; + unsigned int H1; + unsigned int H2; + unsigned int H3; + unsigned int H4; + unsigned int words[80]; + unsigned int TEMP; + /* block I am collecting to process */ + char sha_block[64]; + /* collected so far */ + int how_many_in_block; + unsigned int running_total; +#endif +}; + +#if (defined(__APPLE__) && !defined(__Userspace__) && defined(KERNEL)) +#ifndef _KERNEL +#define _KERNEL +#endif +#endif + +#if defined(_KERNEL) || defined(__Userspace__) + +void sctp_sha1_init(struct sctp_sha1_context *); +void sctp_sha1_update(struct sctp_sha1_context *, const unsigned char *, unsigned int); +void sctp_sha1_final(unsigned char *, struct sctp_sha1_context *); + +#endif +#endif diff --git a/netwerk/sctp/src/netinet/sctp_ss_functions.c b/netwerk/sctp/src/netinet/sctp_ss_functions.c new file mode 100644 index 0000000000..0bc083b580 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_ss_functions.c @@ -0,0 +1,1128 @@ +/*- + * SPDX-License-Identifier: BSD-2-Clause-FreeBSD + * + * Copyright (c) 2010-2012, by Michael Tuexen. All rights reserved. + * Copyright (c) 2010-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2010-2012, by Robin Seggelmann. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_pcb.h> +#if defined(__Userspace__) +#include <netinet/sctp_os_userspace.h> +#endif + +/* + * Default simple round-robin algorithm. + * Just iterates the streams in the order they appear. + */ + +static void +sctp_ss_default_add(struct sctp_tcb *, struct sctp_association *, + struct sctp_stream_out *, + struct sctp_stream_queue_pending *); + +static void +sctp_ss_default_remove(struct sctp_tcb *, struct sctp_association *, + struct sctp_stream_out *, + struct sctp_stream_queue_pending *); + +static void +sctp_ss_default_init(struct sctp_tcb *stcb, struct sctp_association *asoc) +{ + uint16_t i; + + SCTP_TCB_LOCK_ASSERT(stcb); + + asoc->ss_data.locked_on_sending = NULL; + asoc->ss_data.last_out_stream = NULL; + TAILQ_INIT(&asoc->ss_data.out.wheel); + /* + * If there is data in the stream queues already, + * the scheduler of an existing association has + * been changed. We need to add all stream queues + * to the wheel. + */ + for (i = 0; i < asoc->streamoutcnt; i++) { + stcb->asoc.ss_functions.sctp_ss_add_to_stream(stcb, asoc, + &asoc->strmout[i], + NULL); + } + return; +} + +static void +sctp_ss_default_clear(struct sctp_tcb *stcb, struct sctp_association *asoc, + bool clear_values SCTP_UNUSED) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + while (!TAILQ_EMPTY(&asoc->ss_data.out.wheel)) { + struct sctp_stream_out *strq; + + strq = TAILQ_FIRST(&asoc->ss_data.out.wheel); + KASSERT(strq->ss_params.scheduled, ("strq %p not scheduled", (void *)strq)); + TAILQ_REMOVE(&asoc->ss_data.out.wheel, strq, ss_params.ss.rr.next_spoke); + strq->ss_params.scheduled = false; + } + asoc->ss_data.last_out_stream = NULL; + return; +} + +static void +sctp_ss_default_init_stream(struct sctp_tcb *stcb, struct sctp_stream_out *strq, struct sctp_stream_out *with_strq) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + if (with_strq != NULL) { + if (stcb->asoc.ss_data.locked_on_sending == with_strq) { + stcb->asoc.ss_data.locked_on_sending = strq; + } + if (stcb->asoc.ss_data.last_out_stream == with_strq) { + stcb->asoc.ss_data.last_out_stream = strq; + } + } + strq->ss_params.scheduled = false; + return; +} + +static void +sctp_ss_default_add(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq, + struct sctp_stream_queue_pending *sp SCTP_UNUSED) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + /* Add to wheel if not already on it and stream queue not empty */ + if (!TAILQ_EMPTY(&strq->outqueue) && !strq->ss_params.scheduled) { + TAILQ_INSERT_TAIL(&asoc->ss_data.out.wheel, + strq, ss_params.ss.rr.next_spoke); + strq->ss_params.scheduled = true; + } + return; +} + +static bool +sctp_ss_default_is_empty(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_association *asoc) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + return (TAILQ_EMPTY(&asoc->ss_data.out.wheel)); +} + +static void +sctp_ss_default_remove(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq, + struct sctp_stream_queue_pending *sp SCTP_UNUSED) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + /* Remove from wheel if stream queue is empty and actually is on the wheel */ + if (TAILQ_EMPTY(&strq->outqueue) && strq->ss_params.scheduled) { + if (asoc->ss_data.last_out_stream == strq) { + asoc->ss_data.last_out_stream = TAILQ_PREV(asoc->ss_data.last_out_stream, + sctpwheel_listhead, + ss_params.ss.rr.next_spoke); + if (asoc->ss_data.last_out_stream == NULL) { + asoc->ss_data.last_out_stream = TAILQ_LAST(&asoc->ss_data.out.wheel, + sctpwheel_listhead); + } + if (asoc->ss_data.last_out_stream == strq) { + asoc->ss_data.last_out_stream = NULL; + } + } + if (asoc->ss_data.locked_on_sending == strq) { + asoc->ss_data.locked_on_sending = NULL; + } + TAILQ_REMOVE(&asoc->ss_data.out.wheel, strq, ss_params.ss.rr.next_spoke); + strq->ss_params.scheduled = false; + } + return; +} + +static struct sctp_stream_out * +sctp_ss_default_select(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_nets *net, + struct sctp_association *asoc) +{ + struct sctp_stream_out *strq, *strqt; + + SCTP_TCB_LOCK_ASSERT(stcb); + + if (asoc->ss_data.locked_on_sending != NULL) { + KASSERT(asoc->ss_data.locked_on_sending->ss_params.scheduled, + ("locked_on_sending %p not scheduled", + (void *)asoc->ss_data.locked_on_sending)); + return (asoc->ss_data.locked_on_sending); + } + strqt = asoc->ss_data.last_out_stream; + KASSERT(strqt == NULL || strqt->ss_params.scheduled, + ("last_out_stream %p not scheduled", (void *)strqt)); +default_again: + /* Find the next stream to use */ + if (strqt == NULL) { + strq = TAILQ_FIRST(&asoc->ss_data.out.wheel); + } else { + strq = TAILQ_NEXT(strqt, ss_params.ss.rr.next_spoke); + if (strq == NULL) { + strq = TAILQ_FIRST(&asoc->ss_data.out.wheel); + } + } + KASSERT(strq == NULL || strq->ss_params.scheduled, + ("strq %p not scheduled", (void *)strq)); + + /* If CMT is off, we must validate that + * the stream in question has the first + * item pointed towards are network destination + * requested by the caller. Note that if we + * turn out to be locked to a stream (assigning + * TSN's then we must stop, since we cannot + * look for another stream with data to send + * to that destination). In CMT's case, by + * skipping this check, we will send one + * data packet towards the requested net. + */ + if (net != NULL && strq != NULL && + SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0) { + if (TAILQ_FIRST(&strq->outqueue) && + TAILQ_FIRST(&strq->outqueue)->net != NULL && + TAILQ_FIRST(&strq->outqueue)->net != net) { + if (strq == asoc->ss_data.last_out_stream) { + return (NULL); + } else { + strqt = strq; + goto default_again; + } + } + } + return (strq); +} + +static void +sctp_ss_default_scheduled(struct sctp_tcb *stcb, + struct sctp_nets *net SCTP_UNUSED, + struct sctp_association *asoc, + struct sctp_stream_out *strq, + int moved_how_much SCTP_UNUSED) +{ + struct sctp_stream_queue_pending *sp; + + KASSERT(strq != NULL, ("strq is NULL")); + KASSERT(strq->ss_params.scheduled, ("strq %p is not scheduled", (void *)strq)); + SCTP_TCB_LOCK_ASSERT(stcb); + + asoc->ss_data.last_out_stream = strq; + if (asoc->idata_supported == 0) { + sp = TAILQ_FIRST(&strq->outqueue); + if ((sp != NULL) && (sp->some_taken == 1)) { + asoc->ss_data.locked_on_sending = strq; + } else { + asoc->ss_data.locked_on_sending = NULL; + } + } else { + asoc->ss_data.locked_on_sending = NULL; + } + return; +} + +static void +sctp_ss_default_packet_done(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_nets *net SCTP_UNUSED, + struct sctp_association *asoc SCTP_UNUSED) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + /* Nothing to be done here */ + return; +} + +static int +sctp_ss_default_get_value(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_association *asoc SCTP_UNUSED, + struct sctp_stream_out *strq SCTP_UNUSED, uint16_t *value SCTP_UNUSED) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + /* Nothing to be done here */ + return (-1); +} + +static int +sctp_ss_default_set_value(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_association *asoc SCTP_UNUSED, + struct sctp_stream_out *strq SCTP_UNUSED, uint16_t value SCTP_UNUSED) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + /* Nothing to be done here */ + return (-1); +} + +static bool +sctp_ss_default_is_user_msgs_incomplete(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_association *asoc) +{ + struct sctp_stream_out *strq; + struct sctp_stream_queue_pending *sp; + + SCTP_TCB_LOCK_ASSERT(stcb); + + if (asoc->stream_queue_cnt != 1) { + return (false); + } + strq = asoc->ss_data.locked_on_sending; + if (strq == NULL) { + return (false); + } + sp = TAILQ_FIRST(&strq->outqueue); + if (sp == NULL) { + return (false); + } + return (sp->msg_is_complete == 0); +} + +/* + * Real round-robin algorithm. + * Always iterates the streams in ascending order. + */ +static void +sctp_ss_rr_add(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq, + struct sctp_stream_queue_pending *sp SCTP_UNUSED) +{ + struct sctp_stream_out *strqt; + + SCTP_TCB_LOCK_ASSERT(stcb); + + if (!TAILQ_EMPTY(&strq->outqueue) && !strq->ss_params.scheduled) { + if (TAILQ_EMPTY(&asoc->ss_data.out.wheel)) { + TAILQ_INSERT_HEAD(&asoc->ss_data.out.wheel, strq, ss_params.ss.rr.next_spoke); + } else { + strqt = TAILQ_FIRST(&asoc->ss_data.out.wheel); + while (strqt != NULL && (strqt->sid < strq->sid)) { + strqt = TAILQ_NEXT(strqt, ss_params.ss.rr.next_spoke); + } + if (strqt != NULL) { + TAILQ_INSERT_BEFORE(strqt, strq, ss_params.ss.rr.next_spoke); + } else { + TAILQ_INSERT_TAIL(&asoc->ss_data.out.wheel, strq, ss_params.ss.rr.next_spoke); + } + } + strq->ss_params.scheduled = true; + } + return; +} + +/* + * Real round-robin per packet algorithm. + * Always iterates the streams in ascending order and + * only fills messages of the same stream in a packet. + */ +static struct sctp_stream_out * +sctp_ss_rrp_select(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_nets *net SCTP_UNUSED, + struct sctp_association *asoc) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + return (asoc->ss_data.last_out_stream); +} + +static void +sctp_ss_rrp_packet_done(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_nets *net, + struct sctp_association *asoc) +{ + struct sctp_stream_out *strq, *strqt; + + SCTP_TCB_LOCK_ASSERT(stcb); + + strqt = asoc->ss_data.last_out_stream; + KASSERT(strqt == NULL || strqt->ss_params.scheduled, + ("last_out_stream %p not scheduled", (void *)strqt)); +rrp_again: + /* Find the next stream to use */ + if (strqt == NULL) { + strq = TAILQ_FIRST(&asoc->ss_data.out.wheel); + } else { + strq = TAILQ_NEXT(strqt, ss_params.ss.rr.next_spoke); + if (strq == NULL) { + strq = TAILQ_FIRST(&asoc->ss_data.out.wheel); + } + } + KASSERT(strq == NULL || strq->ss_params.scheduled, + ("strq %p not scheduled", (void *)strq)); + + /* If CMT is off, we must validate that + * the stream in question has the first + * item pointed towards are network destination + * requested by the caller. Note that if we + * turn out to be locked to a stream (assigning + * TSN's then we must stop, since we cannot + * look for another stream with data to send + * to that destination). In CMT's case, by + * skipping this check, we will send one + * data packet towards the requested net. + */ + if (net != NULL && strq != NULL && + SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0) { + if (TAILQ_FIRST(&strq->outqueue) && + TAILQ_FIRST(&strq->outqueue)->net != NULL && + TAILQ_FIRST(&strq->outqueue)->net != net) { + if (strq == asoc->ss_data.last_out_stream) { + strq = NULL; + } else { + strqt = strq; + goto rrp_again; + } + } + } + asoc->ss_data.last_out_stream = strq; + return; +} + +/* + * Priority algorithm. + * Always prefers streams based on their priority id. + */ +static void +sctp_ss_prio_clear(struct sctp_tcb *stcb, struct sctp_association *asoc, + bool clear_values) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + while (!TAILQ_EMPTY(&asoc->ss_data.out.wheel)) { + struct sctp_stream_out *strq; + + strq = TAILQ_FIRST(&asoc->ss_data.out.wheel); + KASSERT(strq->ss_params.scheduled, ("strq %p not scheduled", (void *)strq)); + if (clear_values) { + strq->ss_params.ss.prio.priority = 0; + } + TAILQ_REMOVE(&asoc->ss_data.out.wheel, strq, ss_params.ss.prio.next_spoke); + strq->ss_params.scheduled = false; + } + asoc->ss_data.last_out_stream = NULL; + return; +} + +static void +sctp_ss_prio_init_stream(struct sctp_tcb *stcb, struct sctp_stream_out *strq, struct sctp_stream_out *with_strq) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + if (with_strq != NULL) { + if (stcb->asoc.ss_data.locked_on_sending == with_strq) { + stcb->asoc.ss_data.locked_on_sending = strq; + } + if (stcb->asoc.ss_data.last_out_stream == with_strq) { + stcb->asoc.ss_data.last_out_stream = strq; + } + } + strq->ss_params.scheduled = false; + if (with_strq != NULL) { + strq->ss_params.ss.prio.priority = with_strq->ss_params.ss.prio.priority; + } else { + strq->ss_params.ss.prio.priority = 0; + } + return; +} + +static void +sctp_ss_prio_add(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq, struct sctp_stream_queue_pending *sp SCTP_UNUSED) +{ + struct sctp_stream_out *strqt; + + SCTP_TCB_LOCK_ASSERT(stcb); + + /* Add to wheel if not already on it and stream queue not empty */ + if (!TAILQ_EMPTY(&strq->outqueue) && !strq->ss_params.scheduled) { + if (TAILQ_EMPTY(&asoc->ss_data.out.wheel)) { + TAILQ_INSERT_HEAD(&asoc->ss_data.out.wheel, strq, ss_params.ss.prio.next_spoke); + } else { + strqt = TAILQ_FIRST(&asoc->ss_data.out.wheel); + while (strqt != NULL && strqt->ss_params.ss.prio.priority < strq->ss_params.ss.prio.priority) { + strqt = TAILQ_NEXT(strqt, ss_params.ss.prio.next_spoke); + } + if (strqt != NULL) { + TAILQ_INSERT_BEFORE(strqt, strq, ss_params.ss.prio.next_spoke); + } else { + TAILQ_INSERT_TAIL(&asoc->ss_data.out.wheel, strq, ss_params.ss.prio.next_spoke); + } + } + strq->ss_params.scheduled = true; + } + return; +} + +static void +sctp_ss_prio_remove(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq, struct sctp_stream_queue_pending *sp SCTP_UNUSED) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + /* Remove from wheel if stream queue is empty and actually is on the wheel */ + if (TAILQ_EMPTY(&strq->outqueue) && strq->ss_params.scheduled) { + if (asoc->ss_data.last_out_stream == strq) { + asoc->ss_data.last_out_stream = TAILQ_PREV(asoc->ss_data.last_out_stream, + sctpwheel_listhead, + ss_params.ss.prio.next_spoke); + if (asoc->ss_data.last_out_stream == NULL) { + asoc->ss_data.last_out_stream = TAILQ_LAST(&asoc->ss_data.out.wheel, + sctpwheel_listhead); + } + if (asoc->ss_data.last_out_stream == strq) { + asoc->ss_data.last_out_stream = NULL; + } + } + if (asoc->ss_data.locked_on_sending == strq) { + asoc->ss_data.locked_on_sending = NULL; + } + TAILQ_REMOVE(&asoc->ss_data.out.wheel, strq, ss_params.ss.prio.next_spoke); + strq->ss_params.scheduled = false; + } + return; +} + +static struct sctp_stream_out* +sctp_ss_prio_select(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_nets *net, + struct sctp_association *asoc) +{ + struct sctp_stream_out *strq, *strqt, *strqn; + + SCTP_TCB_LOCK_ASSERT(stcb); + + if (asoc->ss_data.locked_on_sending != NULL) { + KASSERT(asoc->ss_data.locked_on_sending->ss_params.scheduled, + ("locked_on_sending %p not scheduled", + (void *)asoc->ss_data.locked_on_sending)); + return (asoc->ss_data.locked_on_sending); + } + strqt = asoc->ss_data.last_out_stream; + KASSERT(strqt == NULL || strqt->ss_params.scheduled, + ("last_out_stream %p not scheduled", (void *)strqt)); +prio_again: + /* Find the next stream to use */ + if (strqt == NULL) { + strq = TAILQ_FIRST(&asoc->ss_data.out.wheel); + } else { + strqn = TAILQ_NEXT(strqt, ss_params.ss.prio.next_spoke); + if (strqn != NULL && + strqn->ss_params.ss.prio.priority == strqt->ss_params.ss.prio.priority) { + strq = strqn; + } else { + strq = TAILQ_FIRST(&asoc->ss_data.out.wheel); + } + } + KASSERT(strq == NULL || strq->ss_params.scheduled, + ("strq %p not scheduled", (void *)strq)); + + /* If CMT is off, we must validate that + * the stream in question has the first + * item pointed towards are network destination + * requested by the caller. Note that if we + * turn out to be locked to a stream (assigning + * TSN's then we must stop, since we cannot + * look for another stream with data to send + * to that destination). In CMT's case, by + * skipping this check, we will send one + * data packet towards the requested net. + */ + if (net != NULL && strq != NULL && + SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0) { + if (TAILQ_FIRST(&strq->outqueue) && + TAILQ_FIRST(&strq->outqueue)->net != NULL && + TAILQ_FIRST(&strq->outqueue)->net != net) { + if (strq == asoc->ss_data.last_out_stream) { + return (NULL); + } else { + strqt = strq; + goto prio_again; + } + } + } + return (strq); +} + +static int +sctp_ss_prio_get_value(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_association *asoc SCTP_UNUSED, + struct sctp_stream_out *strq, uint16_t *value) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + if (strq == NULL) { + return (-1); + } + *value = strq->ss_params.ss.prio.priority; + return (1); +} + +static int +sctp_ss_prio_set_value(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq, uint16_t value) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + if (strq == NULL) { + return (-1); + } + strq->ss_params.ss.prio.priority = value; + sctp_ss_prio_remove(stcb, asoc, strq, NULL); + sctp_ss_prio_add(stcb, asoc, strq, NULL); + return (1); +} + +/* + * Fair bandwidth algorithm. + * Maintains an equal throughput per stream. + */ +static void +sctp_ss_fb_clear(struct sctp_tcb *stcb, struct sctp_association *asoc, + bool clear_values) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + while (!TAILQ_EMPTY(&asoc->ss_data.out.wheel)) { + struct sctp_stream_out *strq; + + strq = TAILQ_FIRST(&asoc->ss_data.out.wheel); + KASSERT(strq->ss_params.scheduled, ("strq %p not scheduled", (void *)strq)); + if (clear_values) { + strq->ss_params.ss.fb.rounds = -1; + } + TAILQ_REMOVE(&asoc->ss_data.out.wheel, strq, ss_params.ss.fb.next_spoke); + strq->ss_params.scheduled = false; + } + asoc->ss_data.last_out_stream = NULL; + return; +} + +static void +sctp_ss_fb_init_stream(struct sctp_tcb *stcb, struct sctp_stream_out *strq, struct sctp_stream_out *with_strq) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + if (with_strq != NULL) { + if (stcb->asoc.ss_data.locked_on_sending == with_strq) { + stcb->asoc.ss_data.locked_on_sending = strq; + } + if (stcb->asoc.ss_data.last_out_stream == with_strq) { + stcb->asoc.ss_data.last_out_stream = strq; + } + } + strq->ss_params.scheduled = false; + if (with_strq != NULL) { + strq->ss_params.ss.fb.rounds = with_strq->ss_params.ss.fb.rounds; + } else { + strq->ss_params.ss.fb.rounds = -1; + } + return; +} + +static void +sctp_ss_fb_add(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq, struct sctp_stream_queue_pending *sp SCTP_UNUSED) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + if (!TAILQ_EMPTY(&strq->outqueue) && !strq->ss_params.scheduled) { + if (strq->ss_params.ss.fb.rounds < 0) + strq->ss_params.ss.fb.rounds = TAILQ_FIRST(&strq->outqueue)->length; + TAILQ_INSERT_TAIL(&asoc->ss_data.out.wheel, strq, ss_params.ss.fb.next_spoke); + strq->ss_params.scheduled = true; + } + return; +} + +static void +sctp_ss_fb_remove(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq, struct sctp_stream_queue_pending *sp SCTP_UNUSED) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + /* Remove from wheel if stream queue is empty and actually is on the wheel */ + if (TAILQ_EMPTY(&strq->outqueue) && strq->ss_params.scheduled) { + if (asoc->ss_data.last_out_stream == strq) { + asoc->ss_data.last_out_stream = TAILQ_PREV(asoc->ss_data.last_out_stream, + sctpwheel_listhead, + ss_params.ss.fb.next_spoke); + if (asoc->ss_data.last_out_stream == NULL) { + asoc->ss_data.last_out_stream = TAILQ_LAST(&asoc->ss_data.out.wheel, + sctpwheel_listhead); + } + if (asoc->ss_data.last_out_stream == strq) { + asoc->ss_data.last_out_stream = NULL; + } + } + if (asoc->ss_data.locked_on_sending == strq) { + asoc->ss_data.locked_on_sending = NULL; + } + TAILQ_REMOVE(&asoc->ss_data.out.wheel, strq, ss_params.ss.fb.next_spoke); + strq->ss_params.scheduled = false; + } + return; +} + +static struct sctp_stream_out* +sctp_ss_fb_select(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_nets *net, + struct sctp_association *asoc) +{ + struct sctp_stream_out *strq = NULL, *strqt; + + SCTP_TCB_LOCK_ASSERT(stcb); + + if (asoc->ss_data.locked_on_sending != NULL) { + KASSERT(asoc->ss_data.locked_on_sending->ss_params.scheduled, + ("locked_on_sending %p not scheduled", + (void *)asoc->ss_data.locked_on_sending)); + return (asoc->ss_data.locked_on_sending); + } + if (asoc->ss_data.last_out_stream == NULL || + TAILQ_FIRST(&asoc->ss_data.out.wheel) == TAILQ_LAST(&asoc->ss_data.out.wheel, sctpwheel_listhead)) { + strqt = TAILQ_FIRST(&asoc->ss_data.out.wheel); + } else { + strqt = TAILQ_NEXT(asoc->ss_data.last_out_stream, ss_params.ss.fb.next_spoke); + } + do { + if ((strqt != NULL) && + ((SCTP_BASE_SYSCTL(sctp_cmt_on_off) > 0) || + (SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0 && + (net == NULL || (TAILQ_FIRST(&strqt->outqueue) && TAILQ_FIRST(&strqt->outqueue)->net == NULL) || + (net != NULL && TAILQ_FIRST(&strqt->outqueue) && TAILQ_FIRST(&strqt->outqueue)->net != NULL && + TAILQ_FIRST(&strqt->outqueue)->net == net))))) { + if ((strqt->ss_params.ss.fb.rounds >= 0) && + ((strq == NULL) || + (strqt->ss_params.ss.fb.rounds < strq->ss_params.ss.fb.rounds))) { + strq = strqt; + } + } + if (strqt != NULL) { + strqt = TAILQ_NEXT(strqt, ss_params.ss.fb.next_spoke); + } else { + strqt = TAILQ_FIRST(&asoc->ss_data.out.wheel); + } + } while (strqt != strq); + return (strq); +} + +static void +sctp_ss_fb_scheduled(struct sctp_tcb *stcb, struct sctp_nets *net SCTP_UNUSED, + struct sctp_association *asoc, struct sctp_stream_out *strq, + int moved_how_much SCTP_UNUSED) +{ + struct sctp_stream_queue_pending *sp; + struct sctp_stream_out *strqt; + int subtract; + + SCTP_TCB_LOCK_ASSERT(stcb); + + if (asoc->idata_supported == 0) { + sp = TAILQ_FIRST(&strq->outqueue); + if ((sp != NULL) && (sp->some_taken == 1)) { + asoc->ss_data.locked_on_sending = strq; + } else { + asoc->ss_data.locked_on_sending = NULL; + } + } else { + asoc->ss_data.locked_on_sending = NULL; + } + subtract = strq->ss_params.ss.fb.rounds; + TAILQ_FOREACH(strqt, &asoc->ss_data.out.wheel, ss_params.ss.fb.next_spoke) { + strqt->ss_params.ss.fb.rounds -= subtract; + if (strqt->ss_params.ss.fb.rounds < 0) + strqt->ss_params.ss.fb.rounds = 0; + } + if (TAILQ_FIRST(&strq->outqueue)) { + strq->ss_params.ss.fb.rounds = TAILQ_FIRST(&strq->outqueue)->length; + } else { + strq->ss_params.ss.fb.rounds = -1; + } + asoc->ss_data.last_out_stream = strq; + return; +} + +/* + * First-come, first-serve algorithm. + * Maintains the order provided by the application. + */ +static void +sctp_ss_fcfs_add(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq SCTP_UNUSED, + struct sctp_stream_queue_pending *sp); + +static void +sctp_ss_fcfs_init(struct sctp_tcb *stcb, struct sctp_association *asoc) +{ + uint32_t x, n = 0, add_more = 1; + struct sctp_stream_queue_pending *sp; + uint16_t i; + + SCTP_TCB_LOCK_ASSERT(stcb); + + TAILQ_INIT(&asoc->ss_data.out.list); + /* + * If there is data in the stream queues already, + * the scheduler of an existing association has + * been changed. We can only cycle through the + * stream queues and add everything to the FCFS + * queue. + */ + while (add_more) { + add_more = 0; + for (i = 0; i < asoc->streamoutcnt; i++) { + sp = TAILQ_FIRST(&asoc->strmout[i].outqueue); + x = 0; + /* Find n. message in current stream queue */ + while (sp != NULL && x < n) { + sp = TAILQ_NEXT(sp, next); + x++; + } + if (sp != NULL) { + sctp_ss_fcfs_add(stcb, asoc, &asoc->strmout[i], sp); + add_more = 1; + } + } + n++; + } + return; +} + +static void +sctp_ss_fcfs_clear(struct sctp_tcb *stcb, struct sctp_association *asoc, + bool clear_values SCTP_UNUSED) +{ + struct sctp_stream_queue_pending *sp; + + SCTP_TCB_LOCK_ASSERT(stcb); + + while (!TAILQ_EMPTY(&asoc->ss_data.out.list)) { + sp = TAILQ_FIRST(&asoc->ss_data.out.list); + KASSERT(sp->scheduled, ("sp %p not scheduled", (void *)sp)); + TAILQ_REMOVE(&asoc->ss_data.out.list, sp, ss_next); + sp->scheduled = false; + } + asoc->ss_data.last_out_stream = NULL; + return; +} + +static void +sctp_ss_fcfs_init_stream(struct sctp_tcb *stcb, struct sctp_stream_out *strq, struct sctp_stream_out *with_strq) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + if (with_strq != NULL) { + if (stcb->asoc.ss_data.locked_on_sending == with_strq) { + stcb->asoc.ss_data.locked_on_sending = strq; + } + if (stcb->asoc.ss_data.last_out_stream == with_strq) { + stcb->asoc.ss_data.last_out_stream = strq; + } + } + strq->ss_params.scheduled = false; + return; +} + +static void +sctp_ss_fcfs_add(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq SCTP_UNUSED, struct sctp_stream_queue_pending *sp) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + if (!sp->scheduled) { + TAILQ_INSERT_TAIL(&asoc->ss_data.out.list, sp, ss_next); + sp->scheduled = true; + } + return; +} + +static bool +sctp_ss_fcfs_is_empty(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_association *asoc) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + return (TAILQ_EMPTY(&asoc->ss_data.out.list)); +} + +static void +sctp_ss_fcfs_remove(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq SCTP_UNUSED, struct sctp_stream_queue_pending *sp) +{ + SCTP_TCB_LOCK_ASSERT(stcb); + + if (sp->scheduled) { + TAILQ_REMOVE(&asoc->ss_data.out.list, sp, ss_next); + sp->scheduled = false; + } + return; +} + +static struct sctp_stream_out * +sctp_ss_fcfs_select(struct sctp_tcb *stcb SCTP_UNUSED, struct sctp_nets *net, + struct sctp_association *asoc) +{ + struct sctp_stream_out *strq; + struct sctp_stream_queue_pending *sp; + + SCTP_TCB_LOCK_ASSERT(stcb); + + if (asoc->ss_data.locked_on_sending) { + return (asoc->ss_data.locked_on_sending); + } + sp = TAILQ_FIRST(&asoc->ss_data.out.list); +default_again: + if (sp != NULL) { + strq = &asoc->strmout[sp->sid]; + } else { + strq = NULL; + } + + /* + * If CMT is off, we must validate that + * the stream in question has the first + * item pointed towards are network destination + * requested by the caller. Note that if we + * turn out to be locked to a stream (assigning + * TSN's then we must stop, since we cannot + * look for another stream with data to send + * to that destination). In CMT's case, by + * skipping this check, we will send one + * data packet towards the requested net. + */ + if (net != NULL && strq != NULL && + SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0) { + if (TAILQ_FIRST(&strq->outqueue) && + TAILQ_FIRST(&strq->outqueue)->net != NULL && + TAILQ_FIRST(&strq->outqueue)->net != net) { + sp = TAILQ_NEXT(sp, ss_next); + goto default_again; + } + } + return (strq); +} + +static void +sctp_ss_fcfs_scheduled(struct sctp_tcb *stcb, + struct sctp_nets *net SCTP_UNUSED, + struct sctp_association *asoc, + struct sctp_stream_out *strq, + int moved_how_much SCTP_UNUSED) +{ + struct sctp_stream_queue_pending *sp; + + KASSERT(strq != NULL, ("strq is NULL")); + asoc->ss_data.last_out_stream = strq; + if (asoc->idata_supported == 0) { + sp = TAILQ_FIRST(&strq->outqueue); + if ((sp != NULL) && (sp->some_taken == 1)) { + asoc->ss_data.locked_on_sending = strq; + } else { + asoc->ss_data.locked_on_sending = NULL; + } + } else { + asoc->ss_data.locked_on_sending = NULL; + } + return; +} + +const struct sctp_ss_functions sctp_ss_functions[] = { +/* SCTP_SS_DEFAULT */ +{ +#if defined(_WIN32) + sctp_ss_default_init, + sctp_ss_default_clear, + sctp_ss_default_init_stream, + sctp_ss_default_add, + sctp_ss_default_is_empty, + sctp_ss_default_remove, + sctp_ss_default_select, + sctp_ss_default_scheduled, + sctp_ss_default_packet_done, + sctp_ss_default_get_value, + sctp_ss_default_set_value, + sctp_ss_default_is_user_msgs_incomplete +#else + .sctp_ss_init = sctp_ss_default_init, + .sctp_ss_clear = sctp_ss_default_clear, + .sctp_ss_init_stream = sctp_ss_default_init_stream, + .sctp_ss_add_to_stream = sctp_ss_default_add, + .sctp_ss_is_empty = sctp_ss_default_is_empty, + .sctp_ss_remove_from_stream = sctp_ss_default_remove, + .sctp_ss_select_stream = sctp_ss_default_select, + .sctp_ss_scheduled = sctp_ss_default_scheduled, + .sctp_ss_packet_done = sctp_ss_default_packet_done, + .sctp_ss_get_value = sctp_ss_default_get_value, + .sctp_ss_set_value = sctp_ss_default_set_value, + .sctp_ss_is_user_msgs_incomplete = sctp_ss_default_is_user_msgs_incomplete +#endif +}, +/* SCTP_SS_ROUND_ROBIN */ +{ +#if defined(_WIN32) + sctp_ss_default_init, + sctp_ss_default_clear, + sctp_ss_default_init_stream, + sctp_ss_rr_add, + sctp_ss_default_is_empty, + sctp_ss_default_remove, + sctp_ss_default_select, + sctp_ss_default_scheduled, + sctp_ss_default_packet_done, + sctp_ss_default_get_value, + sctp_ss_default_set_value, + sctp_ss_default_is_user_msgs_incomplete +#else + .sctp_ss_init = sctp_ss_default_init, + .sctp_ss_clear = sctp_ss_default_clear, + .sctp_ss_init_stream = sctp_ss_default_init_stream, + .sctp_ss_add_to_stream = sctp_ss_rr_add, + .sctp_ss_is_empty = sctp_ss_default_is_empty, + .sctp_ss_remove_from_stream = sctp_ss_default_remove, + .sctp_ss_select_stream = sctp_ss_default_select, + .sctp_ss_scheduled = sctp_ss_default_scheduled, + .sctp_ss_packet_done = sctp_ss_default_packet_done, + .sctp_ss_get_value = sctp_ss_default_get_value, + .sctp_ss_set_value = sctp_ss_default_set_value, + .sctp_ss_is_user_msgs_incomplete = sctp_ss_default_is_user_msgs_incomplete +#endif +}, +/* SCTP_SS_ROUND_ROBIN_PACKET */ +{ +#if defined(_WIN32) + sctp_ss_default_init, + sctp_ss_default_clear, + sctp_ss_default_init_stream, + sctp_ss_rr_add, + sctp_ss_default_is_empty, + sctp_ss_default_remove, + sctp_ss_rrp_select, + sctp_ss_default_scheduled, + sctp_ss_rrp_packet_done, + sctp_ss_default_get_value, + sctp_ss_default_set_value, + sctp_ss_default_is_user_msgs_incomplete +#else + .sctp_ss_init = sctp_ss_default_init, + .sctp_ss_clear = sctp_ss_default_clear, + .sctp_ss_init_stream = sctp_ss_default_init_stream, + .sctp_ss_add_to_stream = sctp_ss_rr_add, + .sctp_ss_is_empty = sctp_ss_default_is_empty, + .sctp_ss_remove_from_stream = sctp_ss_default_remove, + .sctp_ss_select_stream = sctp_ss_rrp_select, + .sctp_ss_scheduled = sctp_ss_default_scheduled, + .sctp_ss_packet_done = sctp_ss_rrp_packet_done, + .sctp_ss_get_value = sctp_ss_default_get_value, + .sctp_ss_set_value = sctp_ss_default_set_value, + .sctp_ss_is_user_msgs_incomplete = sctp_ss_default_is_user_msgs_incomplete +#endif +}, +/* SCTP_SS_PRIORITY */ +{ +#if defined(_WIN32) + sctp_ss_default_init, + sctp_ss_prio_clear, + sctp_ss_prio_init_stream, + sctp_ss_prio_add, + sctp_ss_default_is_empty, + sctp_ss_prio_remove, + sctp_ss_prio_select, + sctp_ss_default_scheduled, + sctp_ss_default_packet_done, + sctp_ss_prio_get_value, + sctp_ss_prio_set_value, + sctp_ss_default_is_user_msgs_incomplete +#else + .sctp_ss_init = sctp_ss_default_init, + .sctp_ss_clear = sctp_ss_prio_clear, + .sctp_ss_init_stream = sctp_ss_prio_init_stream, + .sctp_ss_add_to_stream = sctp_ss_prio_add, + .sctp_ss_is_empty = sctp_ss_default_is_empty, + .sctp_ss_remove_from_stream = sctp_ss_prio_remove, + .sctp_ss_select_stream = sctp_ss_prio_select, + .sctp_ss_scheduled = sctp_ss_default_scheduled, + .sctp_ss_packet_done = sctp_ss_default_packet_done, + .sctp_ss_get_value = sctp_ss_prio_get_value, + .sctp_ss_set_value = sctp_ss_prio_set_value, + .sctp_ss_is_user_msgs_incomplete = sctp_ss_default_is_user_msgs_incomplete +#endif +}, +/* SCTP_SS_FAIR_BANDWITH */ +{ +#if defined(_WIN32) + sctp_ss_default_init, + sctp_ss_fb_clear, + sctp_ss_fb_init_stream, + sctp_ss_fb_add, + sctp_ss_default_is_empty, + sctp_ss_fb_remove, + sctp_ss_fb_select, + sctp_ss_fb_scheduled, + sctp_ss_default_packet_done, + sctp_ss_default_get_value, + sctp_ss_default_set_value, + sctp_ss_default_is_user_msgs_incomplete +#else + .sctp_ss_init = sctp_ss_default_init, + .sctp_ss_clear = sctp_ss_fb_clear, + .sctp_ss_init_stream = sctp_ss_fb_init_stream, + .sctp_ss_add_to_stream = sctp_ss_fb_add, + .sctp_ss_is_empty = sctp_ss_default_is_empty, + .sctp_ss_remove_from_stream = sctp_ss_fb_remove, + .sctp_ss_select_stream = sctp_ss_fb_select, + .sctp_ss_scheduled = sctp_ss_fb_scheduled, + .sctp_ss_packet_done = sctp_ss_default_packet_done, + .sctp_ss_get_value = sctp_ss_default_get_value, + .sctp_ss_set_value = sctp_ss_default_set_value, + .sctp_ss_is_user_msgs_incomplete = sctp_ss_default_is_user_msgs_incomplete +#endif +}, +/* SCTP_SS_FIRST_COME */ +{ +#if defined(_WIN32) + sctp_ss_fcfs_init, + sctp_ss_fcfs_clear, + sctp_ss_fcfs_init_stream, + sctp_ss_fcfs_add, + sctp_ss_fcfs_is_empty, + sctp_ss_fcfs_remove, + sctp_ss_fcfs_select, + sctp_ss_fcfs_scheduled, + sctp_ss_default_packet_done, + sctp_ss_default_get_value, + sctp_ss_default_set_value, + sctp_ss_default_is_user_msgs_incomplete +#else + .sctp_ss_init = sctp_ss_fcfs_init, + .sctp_ss_clear = sctp_ss_fcfs_clear, + .sctp_ss_init_stream = sctp_ss_fcfs_init_stream, + .sctp_ss_add_to_stream = sctp_ss_fcfs_add, + .sctp_ss_is_empty = sctp_ss_fcfs_is_empty, + .sctp_ss_remove_from_stream = sctp_ss_fcfs_remove, + .sctp_ss_select_stream = sctp_ss_fcfs_select, + .sctp_ss_scheduled = sctp_ss_fcfs_scheduled, + .sctp_ss_packet_done = sctp_ss_default_packet_done, + .sctp_ss_get_value = sctp_ss_default_get_value, + .sctp_ss_set_value = sctp_ss_default_set_value, + .sctp_ss_is_user_msgs_incomplete = sctp_ss_default_is_user_msgs_incomplete +#endif +} +}; diff --git a/netwerk/sctp/src/netinet/sctp_structs.h b/netwerk/sctp/src/netinet/sctp_structs.h new file mode 100644 index 0000000000..e96c2c0e2a --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_structs.h @@ -0,0 +1,1299 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_STRUCTS_H_ +#define _NETINET_SCTP_STRUCTS_H_ + +#include <netinet/sctp_os.h> +#include <netinet/sctp_header.h> +#include <netinet/sctp_auth.h> + +struct sctp_timer { + sctp_os_timer_t timer; + + int type; + /* + * Depending on the timer type these will be setup and cast with the + * appropriate entity. + */ + void *ep; + void *tcb; + void *net; +#if defined(__FreeBSD__) && !defined(__Userspace__) + void *vnet; +#endif + + /* for sanity checking */ + void *self; + uint32_t ticks; + uint32_t stopped_from; +}; + +struct sctp_foo_stuff { + struct sctp_inpcb *inp; + uint32_t lineno; + uint32_t ticks; + int updown; +}; + +/* + * This is the information we track on each interface that we know about from + * the distant end. + */ +TAILQ_HEAD(sctpnetlisthead, sctp_nets); + +struct sctp_stream_reset_list { + TAILQ_ENTRY(sctp_stream_reset_list) next_resp; + uint32_t seq; + uint32_t tsn; + uint32_t number_entries; + uint16_t list_of_streams[]; +}; + +TAILQ_HEAD(sctp_resethead, sctp_stream_reset_list); + +/* + * Users of the iterator need to malloc a iterator with a call to + * sctp_initiate_iterator(inp_func, assoc_func, inp_func, pcb_flags, pcb_features, + * asoc_state, void-ptr-arg, uint32-arg, end_func, inp); + * + * Use the following two defines if you don't care what pcb flags are on the EP + * and/or you don't care what state the association is in. + * + * Note that if you specify an INP as the last argument then ONLY each + * association of that single INP will be executed upon. Note that the pcb + * flags STILL apply so if the inp you specify has different pcb_flags then + * what you put in pcb_flags nothing will happen. use SCTP_PCB_ANY_FLAGS to + * assure the inp you specify gets treated. + */ +#define SCTP_PCB_ANY_FLAGS 0x00000000 +#define SCTP_PCB_ANY_FEATURES 0x00000000 +#define SCTP_ASOC_ANY_STATE 0x00000000 + +typedef void (*asoc_func) (struct sctp_inpcb *, struct sctp_tcb *, void *ptr, + uint32_t val); +typedef int (*inp_func) (struct sctp_inpcb *, void *ptr, uint32_t val); +typedef void (*end_func) (void *ptr, uint32_t val); + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SCTP_MCORE_INPUT) && defined(SMP) +/* whats on the mcore control struct */ +struct sctp_mcore_queue { + TAILQ_ENTRY(sctp_mcore_queue) next; + struct vnet *vn; + struct mbuf *m; + int off; + int v6; +}; + +TAILQ_HEAD(sctp_mcore_qhead, sctp_mcore_queue); + +struct sctp_mcore_ctrl { + SCTP_PROCESS_STRUCT thread_proc; + struct sctp_mcore_qhead que; + struct mtx core_mtx; + struct mtx que_mtx; + int running; + int cpuid; +}; +#endif +#endif + +/* This struct is here to cut out the compatiabilty + * pad that bulks up both the inp and stcb. The non + * pad portion MUST stay in complete sync with + * sctp_sndrcvinfo... i.e. if sinfo_xxxx is added + * this must be done here too. + */ +struct sctp_nonpad_sndrcvinfo { + uint16_t sinfo_stream; + uint16_t sinfo_ssn; + uint16_t sinfo_flags; + uint32_t sinfo_ppid; + uint32_t sinfo_context; + uint32_t sinfo_timetolive; + uint32_t sinfo_tsn; + uint32_t sinfo_cumtsn; + sctp_assoc_t sinfo_assoc_id; + uint16_t sinfo_keynumber; + uint16_t sinfo_keynumber_valid; +}; + +struct sctp_iterator { + TAILQ_ENTRY(sctp_iterator) sctp_nxt_itr; +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct vnet *vn; +#endif + struct sctp_timer tmr; + struct sctp_inpcb *inp; /* current endpoint */ + struct sctp_tcb *stcb; /* current* assoc */ + struct sctp_inpcb *next_inp; /* special hook to skip to */ + asoc_func function_assoc; /* per assoc function */ + inp_func function_inp; /* per endpoint function */ + inp_func function_inp_end; /* end INP function */ + end_func function_atend; /* iterator completion function */ + void *pointer; /* pointer for apply func to use */ + uint32_t val; /* value for apply func to use */ + uint32_t pcb_flags; /* endpoint flags being checked */ + uint32_t pcb_features; /* endpoint features being checked */ + uint32_t asoc_state; /* assoc state being checked */ + uint32_t iterator_flags; + uint8_t no_chunk_output; + uint8_t done_current_ep; +}; +/* iterator_flags values */ +#define SCTP_ITERATOR_DO_ALL_INP 0x00000001 +#define SCTP_ITERATOR_DO_SINGLE_INP 0x00000002 + +TAILQ_HEAD(sctpiterators, sctp_iterator); + +struct sctp_copy_all { + struct sctp_inpcb *inp; /* ep */ + struct mbuf *m; + struct sctp_nonpad_sndrcvinfo sndrcv; + ssize_t sndlen; + int cnt_sent; + int cnt_failed; +}; + +struct sctp_asconf_iterator { + struct sctpladdr list_of_work; + int cnt; +}; + +struct iterator_control { +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct mtx ipi_iterator_wq_mtx; + struct mtx it_mtx; +#elif defined(__APPLE__) && !defined(__Userspace__) + lck_mtx_t *ipi_iterator_wq_mtx; + lck_mtx_t *it_mtx; +#elif defined(SCTP_PROCESS_LEVEL_LOCKS) +#if defined(__Userspace__) + userland_mutex_t ipi_iterator_wq_mtx; + userland_mutex_t it_mtx; + userland_cond_t iterator_wakeup; +#else + pthread_mutex_t ipi_iterator_wq_mtx; + pthread_mutex_t it_mtx; + pthread_cond_t iterator_wakeup; +#endif +#elif defined(_WIN32) && !defined(__Userspace__) + struct spinlock it_lock; + struct spinlock ipi_iterator_wq_lock; + KEVENT iterator_wakeup[2]; + PFILE_OBJECT iterator_thread_obj; +#else + void *it_mtx; +#endif +#if !(defined(_WIN32) && !defined(__Userspace__)) +#if !defined(__Userspace__) + SCTP_PROCESS_STRUCT thread_proc; +#else + userland_thread_t thread_proc; +#endif +#endif + struct sctpiterators iteratorhead; + struct sctp_iterator *cur_it; + uint32_t iterator_running; + uint32_t iterator_flags; +}; +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) +#define SCTP_ITERATOR_MUST_EXIT 0x00000001 +#define SCTP_ITERATOR_EXITED 0x00000002 +#endif +#define SCTP_ITERATOR_STOP_CUR_IT 0x00000004 +#define SCTP_ITERATOR_STOP_CUR_INP 0x00000008 + +struct sctp_net_route { +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct nhop_object *ro_nh; + struct llentry *ro_lle; + char *ro_prepend; + uint16_t ro_plen; + uint16_t ro_flags; + uint16_t ro_mtu; + uint16_t spare; +#else + sctp_rtentry_t *ro_rt; +#endif +#if defined(__APPLE__) && !defined(__Userspace__) +#if !defined(APPLE_LEOPARD) && !defined(APPLE_SNOWLEOPARD) && !defined(APPLE_LION) && !defined(APPLE_MOUNTAINLION) && !defined(APPLE_ELCAPITAN) + struct llentry *ro_lle; +#endif +#if !defined(APPLE_LEOPARD) && !defined(APPLE_SNOWLEOPARD) && !defined(APPLE_LION) && !defined(APPLE_MOUNTAINLION) + struct ifaddr *ro_srcia; +#endif +#if !defined(APPLE_LEOPARD) + uint32_t ro_flags; +#endif +#endif + union sctp_sockstore _l_addr; /* remote peer addr */ + struct sctp_ifa *_s_addr; /* our selected src addr */ +}; + +struct htcp { + uint16_t alpha; /* Fixed point arith, << 7 */ + uint8_t beta; /* Fixed point arith, << 7 */ + uint8_t modeswitch; /* Delay modeswitch until we had at least one congestion event */ + uint32_t last_cong; /* Time since last congestion event end */ + uint32_t undo_last_cong; + uint16_t bytes_acked; + uint32_t bytecount; + uint32_t minRTT; + uint32_t maxRTT; + + uint32_t undo_maxRTT; + uint32_t undo_old_maxB; + + /* Bandwidth estimation */ + uint32_t minB; + uint32_t maxB; + uint32_t old_maxB; + uint32_t Bi; + uint32_t lasttime; +}; + +struct rtcc_cc { + struct timeval tls; /* The time we started the sending */ + uint64_t lbw; /* Our last estimated bw */ + uint64_t lbw_rtt; /* RTT at bw estimate */ + uint64_t bw_bytes; /* The total bytes since this sending began */ + uint64_t bw_tot_time; /* The total time since sending began */ + uint64_t new_tot_time; /* temp holding the new value */ + uint64_t bw_bytes_at_last_rttc; /* What bw_bytes was at last rtt calc */ + uint32_t cwnd_at_bw_set; /* Cwnd at last bw saved - lbw */ + uint32_t vol_reduce; /* cnt of voluntary reductions */ + uint16_t steady_step; /* The number required to be in steady state*/ + uint16_t step_cnt; /* The current number */ + uint8_t ret_from_eq; /* When all things are equal what do I return 0/1 - 1 no cc advance */ + uint8_t use_dccc_ecn; /* Flag to enable DCCC ECN */ + uint8_t tls_needs_set; /* Flag to indicate we need to set tls 0 or 1 means set at send 2 not */ + uint8_t last_step_state; /* Last state if steady state stepdown is on */ + uint8_t rtt_set_this_sack; /* Flag saying this sack had RTT calc on it */ + uint8_t last_inst_ind; /* Last saved inst indication */ +}; + +struct sctp_nets { + TAILQ_ENTRY(sctp_nets) sctp_next; /* next link */ + + /* + * Things on the top half may be able to be split into a common + * structure shared by all. + */ + struct sctp_timer pmtu_timer; + struct sctp_timer hb_timer; + + /* + * The following two in combination equate to a route entry for v6 + * or v4. + */ + struct sctp_net_route ro; + + /* mtu discovered so far */ + uint32_t mtu; + uint32_t ssthresh; /* not sure about this one for split */ + uint32_t last_cwr_tsn; + uint32_t cwr_window_tsn; + uint32_t ecn_ce_pkt_cnt; + uint32_t lost_cnt; + /* smoothed average things for RTT and RTO itself */ + int lastsa; + int lastsv; + uint64_t rtt; /* last measured rtt value in us */ + uint32_t RTO; + + /* This is used for SHUTDOWN/SHUTDOWN-ACK/SEND or INIT timers */ + struct sctp_timer rxt_timer; + + /* last time in seconds I sent to it */ + struct timeval last_sent_time; + union cc_control_data { + struct htcp htcp_ca; /* JRS - struct used in HTCP algorithm */ + struct rtcc_cc rtcc; /* rtcc module cc stuff */ + } cc_mod; + int ref_count; + + /* Congestion stats per destination */ + /* + * flight size variables and such, sorry Vern, I could not avoid + * this if I wanted performance :> + */ + uint32_t flight_size; + uint32_t cwnd; /* actual cwnd */ + uint32_t prev_cwnd; /* cwnd before any processing */ + uint32_t ecn_prev_cwnd; /* ECN prev cwnd at first ecn_echo seen in new window */ + uint32_t partial_bytes_acked; /* in CA tracks when to incr a MTU */ + /* tracking variables to avoid the aloc/free in sack processing */ + unsigned int net_ack; + unsigned int net_ack2; + + /* + * JRS - 5/8/07 - Variable to track last time + * a destination was active for CMT PF + */ + uint32_t last_active; + + /* + * CMT variables (iyengar@cis.udel.edu) + */ + uint32_t this_sack_highest_newack; /* tracks highest TSN newly + * acked for a given dest in + * the current SACK. Used in + * SFR and HTNA algos */ + uint32_t pseudo_cumack; /* CMT CUC algorithm. Maintains next expected + * pseudo-cumack for this destination */ + uint32_t rtx_pseudo_cumack; /* CMT CUC algorithm. Maintains next + * expected pseudo-cumack for this + * destination */ + + /* CMT fast recovery variables */ + uint32_t fast_recovery_tsn; + uint32_t heartbeat_random1; + uint32_t heartbeat_random2; +#ifdef INET6 + uint32_t flowlabel; +#endif + uint8_t dscp; + + struct timeval start_time; /* time when this net was created */ + uint32_t marked_retrans; /* number or DATA chunks marked for + timer based retransmissions */ + uint32_t marked_fastretrans; + uint32_t heart_beat_delay; /* Heart Beat delay in ms */ + + /* if this guy is ok or not ... status */ + uint16_t dest_state; + /* number of timeouts to consider the destination unreachable */ + uint16_t failure_threshold; + /* number of timeouts to consider the destination potentially failed */ + uint16_t pf_threshold; + /* error stats on the destination */ + uint16_t error_count; + /* UDP port number in case of UDP tunneling */ + uint16_t port; + + uint8_t fast_retran_loss_recovery; + uint8_t will_exit_fast_recovery; + /* Flags that probably can be combined into dest_state */ + uint8_t fast_retran_ip; /* fast retransmit in progress */ + uint8_t hb_responded; + uint8_t saw_newack; /* CMT's SFR algorithm flag */ + uint8_t src_addr_selected; /* if we split we move */ + uint8_t indx_of_eligible_next_to_use; + uint8_t addr_is_local; /* its a local address (if known) could move + * in split */ + + /* + * CMT variables (iyengar@cis.udel.edu) + */ + uint8_t find_pseudo_cumack; /* CMT CUC algorithm. Flag used to + * find a new pseudocumack. This flag + * is set after a new pseudo-cumack + * has been received and indicates + * that the sender should find the + * next pseudo-cumack expected for + * this destination */ + uint8_t find_rtx_pseudo_cumack; /* CMT CUCv2 algorithm. Flag used to + * find a new rtx-pseudocumack. This + * flag is set after a new + * rtx-pseudo-cumack has been received + * and indicates that the sender + * should find the next + * rtx-pseudo-cumack expected for this + * destination */ + uint8_t new_pseudo_cumack; /* CMT CUC algorithm. Flag used to + * indicate if a new pseudo-cumack or + * rtx-pseudo-cumack has been received */ + uint8_t window_probe; /* Doing a window probe? */ + uint8_t RTO_measured; /* Have we done the first measure */ + uint8_t last_hs_used; /* index into the last HS table entry we used */ + uint8_t lan_type; + uint8_t rto_needed; +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint32_t flowid; + uint8_t flowtype; +#endif +}; + +struct sctp_data_chunkrec { + uint32_t tsn; /* the TSN of this transmit */ + uint32_t mid; /* the message identifier of this transmit */ + uint16_t sid; /* the stream number of this guy */ + uint32_t ppid; + uint32_t context; /* from send */ + uint32_t cwnd_at_send; + /* + * part of the Highest sacked algorithm to be able to stroke counts + * on ones that are FR'd. + */ + uint32_t fast_retran_tsn; /* sending_seq at the time of FR */ + struct timeval timetodrop; /* time we drop it from queue */ + uint32_t fsn; /* Fragment Sequence Number */ + uint8_t doing_fast_retransmit; + uint8_t rcv_flags; /* flags pulled from data chunk on inbound for + * outbound holds sending flags for PR-SCTP. + */ + uint8_t state_flags; + uint8_t chunk_was_revoked; + uint8_t fwd_tsn_cnt; +}; + +TAILQ_HEAD(sctpchunk_listhead, sctp_tmit_chunk); + +/* The lower byte is used to enumerate PR_SCTP policies */ +#define CHUNK_FLAGS_PR_SCTP_TTL SCTP_PR_SCTP_TTL +#define CHUNK_FLAGS_PR_SCTP_BUF SCTP_PR_SCTP_BUF +#define CHUNK_FLAGS_PR_SCTP_RTX SCTP_PR_SCTP_RTX + +/* The upper byte is used as a bit mask */ +#define CHUNK_FLAGS_FRAGMENT_OK 0x0100 + +struct chk_id { + uint8_t id; + uint8_t can_take_data; +}; + +struct sctp_tmit_chunk { + union { + struct sctp_data_chunkrec data; + struct chk_id chunk_id; + } rec; + struct sctp_association *asoc; /* bp to asoc this belongs to */ + struct timeval sent_rcv_time; /* filled in if RTT being calculated */ + struct mbuf *data; /* pointer to mbuf chain of data */ + struct mbuf *last_mbuf; /* pointer to last mbuf in chain */ + struct sctp_nets *whoTo; + TAILQ_ENTRY(sctp_tmit_chunk) sctp_next; /* next link */ + int32_t sent; /* the send status */ + uint16_t snd_count; /* number of times I sent */ + uint16_t flags; /* flags, such as FRAGMENT_OK */ + uint16_t send_size; + uint16_t book_size; + uint16_t mbcnt; + uint16_t auth_keyid; + uint8_t holds_key_ref; /* flag if auth keyid refcount is held */ + uint8_t pad_inplace; + uint8_t do_rtt; + uint8_t book_size_scale; + uint8_t no_fr_allowed; + uint8_t copy_by_ref; + uint8_t window_probe; +}; + +struct sctp_queued_to_read { /* sinfo structure Pluse more */ + uint16_t sinfo_stream; /* off the wire */ + uint16_t sinfo_flags; /* SCTP_UNORDERED from wire use SCTP_EOF for + * EOR */ + uint32_t sinfo_ppid; /* off the wire */ + uint32_t sinfo_context; /* pick this up from assoc def context? */ + uint32_t sinfo_timetolive; /* not used by kernel */ + uint32_t sinfo_tsn; /* Use this in reassembly as first TSN */ + uint32_t sinfo_cumtsn; /* Use this in reassembly as last TSN */ + sctp_assoc_t sinfo_assoc_id; /* our assoc id */ + /* Non sinfo stuff */ + uint32_t mid; /* Fragment Index */ + uint32_t length; /* length of data */ + uint32_t held_length; /* length held in sb */ + uint32_t top_fsn; /* Highest FSN in queue */ + uint32_t fsn_included; /* Highest FSN in *data portion */ + struct sctp_nets *whoFrom; /* where it came from */ + struct mbuf *data; /* front of the mbuf chain of data with + * PKT_HDR */ + struct mbuf *tail_mbuf; /* used for multi-part data */ + struct mbuf *aux_data; /* used to hold/cache control if o/s does not take it from us */ + struct sctp_tcb *stcb; /* assoc, used for window update */ + TAILQ_ENTRY(sctp_queued_to_read) next; + TAILQ_ENTRY(sctp_queued_to_read) next_instrm; + struct sctpchunk_listhead reasm; + uint16_t port_from; + uint16_t spec_flags; /* Flags to hold the notification field */ + uint8_t do_not_ref_stcb; + uint8_t end_added; + uint8_t pdapi_aborted; + uint8_t pdapi_started; + uint8_t some_taken; + uint8_t last_frag_seen; + uint8_t first_frag_seen; + uint8_t on_read_q; + uint8_t on_strm_q; +}; + +#define SCTP_ON_ORDERED 1 +#define SCTP_ON_UNORDERED 2 + +/* This data structure will be on the outbound + * stream queues. Data will be pulled off from + * the front of the mbuf data and chunk-ified + * by the output routines. We will custom + * fit every chunk we pull to the send/sent + * queue to make up the next full packet + * if we can. An entry cannot be removed + * from the stream_out queue until + * the msg_is_complete flag is set. This + * means at times data/tail_mbuf MIGHT + * be NULL.. If that occurs it happens + * for one of two reasons. Either the user + * is blocked on a send() call and has not + * awoken to copy more data down... OR + * the user is in the explict MSG_EOR mode + * and wrote some data, but has not completed + * sending. + * ss_next and scheduled are only used by the FCFS stream scheduler. + */ +struct sctp_stream_queue_pending { + struct mbuf *data; + struct mbuf *tail_mbuf; + struct timeval ts; + struct sctp_nets *net; + TAILQ_ENTRY (sctp_stream_queue_pending) next; + TAILQ_ENTRY (sctp_stream_queue_pending) ss_next; + uint32_t fsn; + uint32_t length; + uint32_t timetolive; + uint32_t ppid; + uint32_t context; + uint16_t sinfo_flags; + uint16_t sid; + uint16_t act_flags; + uint16_t auth_keyid; + uint8_t holds_key_ref; + uint8_t msg_is_complete; + uint8_t some_taken; + uint8_t sender_all_done; + uint8_t put_last_out; + uint8_t discard_rest; + uint8_t processing; + bool scheduled; +}; + +/* + * this struct contains info that is used to track inbound stream data and + * help with ordering. + */ +TAILQ_HEAD(sctpwheelunrel_listhead, sctp_stream_in); +struct sctp_stream_in { + struct sctp_readhead inqueue; + struct sctp_readhead uno_inqueue; + uint32_t last_mid_delivered; /* used for re-order */ + uint16_t sid; + uint8_t delivery_started; + uint8_t pd_api_started; +}; + +TAILQ_HEAD(sctpwheel_listhead, sctp_stream_out); +TAILQ_HEAD(sctplist_listhead, sctp_stream_queue_pending); + +/* + * This union holds all data necessary for + * different stream schedulers. + */ +struct scheduling_data { + struct sctp_stream_out *locked_on_sending; + /* circular looking for output selection */ + struct sctp_stream_out *last_out_stream; + union { + struct sctpwheel_listhead wheel; + struct sctplist_listhead list; + } out; +}; + +/* Round-robin schedulers */ +struct ss_rr { + /* next link in wheel */ + TAILQ_ENTRY(sctp_stream_out) next_spoke; +}; + +/* Priority scheduler */ +struct ss_prio { + /* next link in wheel */ + TAILQ_ENTRY(sctp_stream_out) next_spoke; + /* priority id */ + uint16_t priority; +}; + +/* Fair Bandwidth scheduler */ +struct ss_fb { + /* next link in wheel */ + TAILQ_ENTRY(sctp_stream_out) next_spoke; + /* stores message size */ + int32_t rounds; +}; + +/* + * This union holds all parameters per stream + * necessary for different stream schedulers. + */ +struct scheduling_parameters { + union { + struct ss_rr rr; + struct ss_prio prio; + struct ss_fb fb; + } ss; + bool scheduled; +}; + +/* States for outgoing streams */ +#define SCTP_STREAM_CLOSED 0x00 +#define SCTP_STREAM_OPENING 0x01 +#define SCTP_STREAM_OPEN 0x02 +#define SCTP_STREAM_RESET_PENDING 0x03 +#define SCTP_STREAM_RESET_IN_FLIGHT 0x04 + +/* This struct is used to track the traffic on outbound streams */ +struct sctp_stream_out { + struct sctp_streamhead outqueue; + struct scheduling_parameters ss_params; + uint32_t chunks_on_queues; /* send queue and sent queue */ +#if defined(SCTP_DETAILED_STR_STATS) + uint32_t abandoned_unsent[SCTP_PR_SCTP_MAX + 1]; + uint32_t abandoned_sent[SCTP_PR_SCTP_MAX + 1]; +#else + /* Only the aggregation */ + uint32_t abandoned_unsent[1]; + uint32_t abandoned_sent[1]; +#endif + /* For associations using DATA chunks, the lower 16-bit of + * next_mid_ordered are used as the next SSN. + */ + uint32_t next_mid_ordered; + uint32_t next_mid_unordered; + uint16_t sid; + uint8_t last_msg_incomplete; + uint8_t state; +}; + +#define SCTP_MAX_STREAMS_AT_ONCE_RESET 200 + +/* used to keep track of the addresses yet to try to add/delete */ +TAILQ_HEAD(sctp_asconf_addrhead, sctp_asconf_addr); +struct sctp_asconf_addr { + TAILQ_ENTRY(sctp_asconf_addr) next; + struct sctp_asconf_addr_param ap; + struct sctp_ifa *ifa; /* save the ifa for add/del ip */ + uint8_t sent; /* has this been sent yet? */ + uint8_t special_del; /* not to be used in lookup */ +}; + +struct sctp_scoping { + uint8_t ipv4_addr_legal; + uint8_t ipv6_addr_legal; +#if defined(__Userspace__) + uint8_t conn_addr_legal; +#endif + uint8_t loopback_scope; + uint8_t ipv4_local_scope; + uint8_t local_scope; + uint8_t site_scope; +}; + +#define SCTP_TSN_LOG_SIZE 40 + +struct sctp_tsn_log { + void *stcb; + uint32_t tsn; + uint32_t seq; + uint16_t strm; + uint16_t sz; + uint16_t flgs; + uint16_t in_pos; + uint16_t in_out; + uint16_t resv; +}; + +#define SCTP_FS_SPEC_LOG_SIZE 200 +struct sctp_fs_spec_log { + uint32_t sent; + uint32_t total_flight; + uint32_t tsn; + uint16_t book; + uint8_t incr; + uint8_t decr; +}; + +/* + * JRS - Structure to hold function pointers to the functions responsible + * for congestion control. + */ + +struct sctp_cc_functions { + void (*sctp_set_initial_cc_param)(struct sctp_tcb *stcb, struct sctp_nets *net); + void (*sctp_cwnd_update_after_sack)(struct sctp_tcb *stcb, + struct sctp_association *asoc, + int accum_moved ,int reneged_all, int will_exit); + void (*sctp_cwnd_update_exit_pf)(struct sctp_tcb *stcb, struct sctp_nets *net); + void (*sctp_cwnd_update_after_fr)(struct sctp_tcb *stcb, + struct sctp_association *asoc); + void (*sctp_cwnd_update_after_timeout)(struct sctp_tcb *stcb, + struct sctp_nets *net); + void (*sctp_cwnd_update_after_ecn_echo)(struct sctp_tcb *stcb, + struct sctp_nets *net, int in_window, int num_pkt_lost); + void (*sctp_cwnd_update_after_packet_dropped)(struct sctp_tcb *stcb, + struct sctp_nets *net, struct sctp_pktdrop_chunk *cp, + uint32_t *bottle_bw, uint32_t *on_queue); + void (*sctp_cwnd_update_after_output)(struct sctp_tcb *stcb, + struct sctp_nets *net, int burst_limit); + void (*sctp_cwnd_update_packet_transmitted)(struct sctp_tcb *stcb, + struct sctp_nets *net); + void (*sctp_cwnd_update_tsn_acknowledged)(struct sctp_nets *net, + struct sctp_tmit_chunk *); + void (*sctp_cwnd_new_transmission_begins)(struct sctp_tcb *stcb, + struct sctp_nets *net); + void (*sctp_cwnd_prepare_net_for_sack)(struct sctp_tcb *stcb, + struct sctp_nets *net); + int (*sctp_cwnd_socket_option)(struct sctp_tcb *stcb, int set, struct sctp_cc_option *); + void (*sctp_rtt_calculated)(struct sctp_tcb *, struct sctp_nets *, struct timeval *); +}; + +/* + * RS - Structure to hold function pointers to the functions responsible + * for stream scheduling. + */ +struct sctp_ss_functions { + void (*sctp_ss_init)(struct sctp_tcb *stcb, struct sctp_association *asoc); + void (*sctp_ss_clear)(struct sctp_tcb *stcb, struct sctp_association *asoc, + bool clear_values); + void (*sctp_ss_init_stream)(struct sctp_tcb *stcb, struct sctp_stream_out *strq, struct sctp_stream_out *with_strq); + void (*sctp_ss_add_to_stream)(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq, struct sctp_stream_queue_pending *sp); + bool (*sctp_ss_is_empty)(struct sctp_tcb *stcb, struct sctp_association *asoc); + void (*sctp_ss_remove_from_stream)(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq, struct sctp_stream_queue_pending *sp); + struct sctp_stream_out* (*sctp_ss_select_stream)(struct sctp_tcb *stcb, + struct sctp_nets *net, struct sctp_association *asoc); + void (*sctp_ss_scheduled)(struct sctp_tcb *stcb, struct sctp_nets *net, + struct sctp_association *asoc, struct sctp_stream_out *strq, int moved_how_much); + void (*sctp_ss_packet_done)(struct sctp_tcb *stcb, struct sctp_nets *net, + struct sctp_association *asoc); + int (*sctp_ss_get_value)(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq, uint16_t *value); + int (*sctp_ss_set_value)(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_stream_out *strq, uint16_t value); + bool (*sctp_ss_is_user_msgs_incomplete)(struct sctp_tcb *stcb, struct sctp_association *asoc); +}; + +/* used to save ASCONF chunks for retransmission */ +TAILQ_HEAD(sctp_asconf_head, sctp_asconf); +struct sctp_asconf { + TAILQ_ENTRY(sctp_asconf) next; + uint32_t serial_number; + uint16_t snd_count; + struct mbuf *data; + uint16_t len; +}; + +/* used to save ASCONF-ACK chunks for retransmission */ +TAILQ_HEAD(sctp_asconf_ackhead, sctp_asconf_ack); +struct sctp_asconf_ack { + TAILQ_ENTRY(sctp_asconf_ack) next; + uint32_t serial_number; + struct sctp_nets *last_sent_to; + struct mbuf *data; + uint16_t len; +}; + +/* + * Here we have information about each individual association that we track. + * We probably in production would be more dynamic. But for ease of + * implementation we will have a fixed array that we hunt for in a linear + * fashion. + */ +struct sctp_association { + /* association state */ + int state; + + /* queue of pending addrs to add/delete */ + struct sctp_asconf_addrhead asconf_queue; + + struct timeval time_entered; /* time we entered state */ + struct timeval time_last_rcvd; + struct timeval time_last_sent; + struct timeval time_last_sat_advance; + struct sctp_nonpad_sndrcvinfo def_send; + + /* timers and such */ + struct sctp_timer dack_timer; /* Delayed ack timer */ + struct sctp_timer asconf_timer; /* asconf */ + struct sctp_timer strreset_timer; /* stream reset */ + struct sctp_timer shut_guard_timer; /* shutdown guard */ + struct sctp_timer autoclose_timer; /* automatic close timer */ + struct sctp_timer delete_prim_timer; /* deleting primary dst */ + + /* list of restricted local addresses */ + struct sctpladdr sctp_restricted_addrs; + + /* last local address pending deletion (waiting for an address add) */ + struct sctp_ifa *asconf_addr_del_pending; + /* Deleted primary destination (used to stop timer) */ + struct sctp_nets *deleted_primary; + + struct sctpnetlisthead nets; /* remote address list */ + + /* Free chunk list */ + struct sctpchunk_listhead free_chunks; + + /* Control chunk queue */ + struct sctpchunk_listhead control_send_queue; + + /* ASCONF chunk queue */ + struct sctpchunk_listhead asconf_send_queue; + + /* + * Once a TSN hits the wire it is moved to the sent_queue. We + * maintain two counts here (don't know if any but retran_cnt is + * needed). The idea is that the sent_queue_retran_cnt reflects how + * many chunks have been marked for retranmission by either T3-rxt + * or FR. + */ + struct sctpchunk_listhead sent_queue; + struct sctpchunk_listhead send_queue; + + /* Scheduling queues */ + struct scheduling_data ss_data; + + /* If an iterator is looking at me, this is it */ + struct sctp_iterator *stcb_starting_point_for_iterator; + + /* ASCONF save the last ASCONF-ACK so we can resend it if necessary */ + struct sctp_asconf_ackhead asconf_ack_sent; + + /* + * pointer to last stream reset queued to control queue by us with + * requests. + */ + struct sctp_tmit_chunk *str_reset; + /* + * if Source Address Selection happening, this will rotate through + * the link list. + */ + struct sctp_laddr *last_used_address; + + /* stream arrays */ + struct sctp_stream_in *strmin; + struct sctp_stream_out *strmout; + uint8_t *mapping_array; + /* primary destination to use */ + struct sctp_nets *primary_destination; + struct sctp_nets *alternate; /* If primary is down or PF */ + /* For CMT */ + struct sctp_nets *last_net_cmt_send_started; + /* last place I got a data chunk from */ + struct sctp_nets *last_data_chunk_from; + /* last place I got a control from */ + struct sctp_nets *last_control_chunk_from; + + /* + * wait to the point the cum-ack passes req->send_reset_at_tsn for + * any req on the list. + */ + struct sctp_resethead resetHead; + + /* queue of chunks waiting to be sent into the local stack */ + struct sctp_readhead pending_reply_queue; + + /* JRS - the congestion control functions are in this struct */ + struct sctp_cc_functions cc_functions; + /* JRS - value to store the currently loaded congestion control module */ + uint32_t congestion_control_module; + /* RS - the stream scheduling functions are in this struct */ + struct sctp_ss_functions ss_functions; + /* RS - value to store the currently loaded stream scheduling module */ + uint32_t stream_scheduling_module; + + uint32_t vrf_id; + uint32_t cookie_preserve_req; + /* ASCONF next seq I am sending out, inits at init-tsn */ + uint32_t asconf_seq_out; + uint32_t asconf_seq_out_acked; + /* ASCONF last received ASCONF from peer, starts at peer's TSN-1 */ + uint32_t asconf_seq_in; + + /* next seq I am sending in str reset messages */ + uint32_t str_reset_seq_out; + /* next seq I am expecting in str reset messages */ + uint32_t str_reset_seq_in; + + /* various verification tag information */ + uint32_t my_vtag; /* The tag to be used. if assoc is re-initited + * by remote end, and I have unlocked this + * will be regenerated to a new random value. */ + uint32_t peer_vtag; /* The peers last tag */ + + uint32_t my_vtag_nonce; + uint32_t peer_vtag_nonce; + + uint32_t assoc_id; + + /* This is the SCTP fragmentation threshold */ + uint32_t smallest_mtu; + + /* + * Special hook for Fast retransmit, allows us to track the highest + * TSN that is NEW in this SACK if gap ack blocks are present. + */ + uint32_t this_sack_highest_gap; + + /* + * The highest consecutive TSN that has been acked by peer on my + * sends + */ + uint32_t last_acked_seq; + + /* The next TSN that I will use in sending. */ + uint32_t sending_seq; + + /* Original seq number I used ??questionable to keep?? */ + uint32_t init_seq_number; + + /* The Advanced Peer Ack Point, as required by the PR-SCTP */ + /* (A1 in Section 4.2) */ + uint32_t advanced_peer_ack_point; + + /* + * The highest consequetive TSN at the bottom of the mapping array + * (for his sends). + */ + uint32_t cumulative_tsn; + /* + * Used to track the mapping array and its offset bits. This MAY be + * lower then cumulative_tsn. + */ + uint32_t mapping_array_base_tsn; + /* + * used to track highest TSN we have received and is listed in the + * mapping array. + */ + uint32_t highest_tsn_inside_map; + + /* EY - new NR variables used for nr_sack based on mapping_array*/ + uint8_t *nr_mapping_array; + uint32_t highest_tsn_inside_nr_map; + + uint32_t fast_recovery_tsn; + uint32_t sat_t3_recovery_tsn; + uint32_t tsn_last_delivered; + /* + * For the pd-api we should re-write this a bit more efficient. We + * could have multiple sctp_queued_to_read's that we are building at + * once. Now we only do this when we get ready to deliver to the + * socket buffer. Note that we depend on the fact that the struct is + * "stuck" on the read queue until we finish all the pd-api. + */ + struct sctp_queued_to_read *control_pdapi; + + uint32_t tsn_of_pdapi_last_delivered; + uint32_t pdapi_ppid; + uint32_t context; + uint32_t last_reset_action[SCTP_MAX_RESET_PARAMS]; + uint32_t last_sending_seq[SCTP_MAX_RESET_PARAMS]; + uint32_t last_base_tsnsent[SCTP_MAX_RESET_PARAMS]; +#ifdef SCTP_ASOCLOG_OF_TSNS + /* + * special log - This adds considerable size + * to the asoc, but provides a log that you + * can use to detect problems via kgdb. + */ + struct sctp_tsn_log in_tsnlog[SCTP_TSN_LOG_SIZE]; + struct sctp_tsn_log out_tsnlog[SCTP_TSN_LOG_SIZE]; + uint32_t cumack_log[SCTP_TSN_LOG_SIZE]; + uint32_t cumack_logsnt[SCTP_TSN_LOG_SIZE]; + uint16_t tsn_in_at; + uint16_t tsn_out_at; + uint16_t tsn_in_wrapped; + uint16_t tsn_out_wrapped; + uint16_t cumack_log_at; + uint16_t cumack_log_atsnt; +#endif /* SCTP_ASOCLOG_OF_TSNS */ +#ifdef SCTP_FS_SPEC_LOG + struct sctp_fs_spec_log fslog[SCTP_FS_SPEC_LOG_SIZE]; + uint16_t fs_index; +#endif + + /* + * window state information and smallest MTU that I use to bound + * segmentation + */ + uint32_t peers_rwnd; + uint32_t my_rwnd; + uint32_t my_last_reported_rwnd; + uint32_t sctp_frag_point; + + uint32_t total_output_queue_size; + + uint32_t sb_cc; /* shadow of sb_cc */ + uint32_t sb_send_resv; /* amount reserved on a send */ + uint32_t my_rwnd_control_len; /* shadow of sb_mbcnt used for rwnd control */ +#ifdef INET6 + uint32_t default_flowlabel; +#endif + uint32_t pr_sctp_cnt; + int ctrl_queue_cnt; /* could be removed REM - NO IT CAN'T!! RRS */ + /* + * All outbound datagrams queue into this list from the individual + * stream queue. Here they get assigned a TSN and then await + * sending. The stream seq comes when it is first put in the + * individual str queue + */ + unsigned int stream_queue_cnt; + unsigned int send_queue_cnt; + unsigned int sent_queue_cnt; + unsigned int sent_queue_cnt_removeable; + /* + * Number on sent queue that are marked for retran until this value + * is 0 we only send one packet of retran'ed data. + */ + unsigned int sent_queue_retran_cnt; + + unsigned int size_on_reasm_queue; + unsigned int cnt_on_reasm_queue; + unsigned int fwd_tsn_cnt; + /* amount of data (bytes) currently in flight (on all destinations) */ + unsigned int total_flight; + /* Total book size in flight */ + unsigned int total_flight_count; /* count of chunks used with + * book total */ + /* count of destinaton nets and list of destination nets */ + unsigned int numnets; + + /* Total error count on this association */ + unsigned int overall_error_count; + + unsigned int cnt_msg_on_sb; + + /* All stream count of chunks for delivery */ + unsigned int size_on_all_streams; + unsigned int cnt_on_all_streams; + + /* Heart Beat delay in ms */ + uint32_t heart_beat_delay; + + /* autoclose */ + uint32_t sctp_autoclose_ticks; + + /* how many preopen streams we have */ + unsigned int pre_open_streams; + + /* How many streams I support coming into me */ + unsigned int max_inbound_streams; + + /* the cookie life I award for any cookie, in seconds */ + uint32_t cookie_life; + /* time to delay acks for */ + unsigned int delayed_ack; + unsigned int old_delayed_ack; + unsigned int sack_freq; + unsigned int data_pkts_seen; + + unsigned int numduptsns; + int dup_tsns[SCTP_MAX_DUP_TSNS]; + uint32_t initial_init_rto_max; /* initial RTO for INIT's */ + uint32_t initial_rto; /* initial send RTO */ + uint32_t minrto; /* per assoc RTO-MIN */ + uint32_t maxrto; /* per assoc RTO-MAX */ + + /* authentication fields */ + sctp_auth_chklist_t *local_auth_chunks; + sctp_auth_chklist_t *peer_auth_chunks; + sctp_hmaclist_t *local_hmacs; /* local HMACs supported */ + sctp_hmaclist_t *peer_hmacs; /* peer HMACs supported */ + struct sctp_keyhead shared_keys; /* assoc's shared keys */ + sctp_authinfo_t authinfo; /* randoms, cached keys */ + /* + * refcnt to block freeing when a sender or receiver is off coping + * user data in. + */ + uint32_t refcnt; + uint32_t chunks_on_out_queue; /* total chunks floating around, + * locked by send socket buffer */ + uint32_t peers_adaptation; + uint32_t default_mtu; + uint16_t peer_hmac_id; /* peer HMAC id to send */ + + /* + * Being that we have no bag to collect stale cookies, and that we + * really would not want to anyway.. we will count them in this + * counter. We of course feed them to the pigeons right away (I have + * always thought of pigeons as flying rats). + */ + uint16_t stale_cookie_count; + + /* + * For the partial delivery API, if up, invoked this is what last + * TSN I delivered + */ + uint16_t str_of_pdapi; + uint16_t ssn_of_pdapi; + + /* counts of actual built streams. Allocation may be more however */ + /* could re-arrange to optimize space here. */ + uint16_t streamincnt; + uint16_t streamoutcnt; + uint16_t strm_realoutsize; + uint16_t strm_pending_add_size; + /* my maximum number of retrans of INIT and SEND */ + /* copied from SCTP but should be individually setable */ + uint16_t max_init_times; + uint16_t max_send_times; + + uint16_t def_net_failure; + + uint16_t def_net_pf_threshold; + + /* + * lock flag: 0 is ok to send, 1+ (duals as a retran count) is + * awaiting ACK + */ + uint16_t mapping_array_size; + + uint16_t last_strm_seq_delivered; + uint16_t last_strm_no_delivered; + + uint16_t last_revoke_count; + int16_t num_send_timers_up; + + uint16_t stream_locked_on; + uint16_t ecn_echo_cnt_onq; + + uint16_t free_chunk_cnt; + uint8_t stream_locked; + uint8_t authenticated; /* packet authenticated ok */ + /* + * This flag indicates that a SACK need to be sent. + * Initially this is 1 to send the first sACK immediately. + */ + uint8_t send_sack; + + /* max burst of new packets into the network */ + uint32_t max_burst; + /* max burst of fast retransmit packets */ + uint32_t fr_max_burst; + + uint8_t sat_network; /* RTT is in range of sat net or greater */ + uint8_t sat_network_lockout; /* lockout code */ + uint8_t burst_limit_applied; /* Burst limit in effect at last send? */ + /* flag goes on when we are doing a partial delivery api */ + uint8_t hb_random_values[4]; + uint8_t fragmented_delivery_inprogress; + uint8_t fragment_flags; + uint8_t last_flags_delivered; + uint8_t hb_ect_randombit; + uint8_t hb_random_idx; + uint8_t default_dscp; + uint8_t asconf_del_pending; /* asconf delete last addr pending */ + uint8_t trigger_reset; + /* + * This value, plus all other ack'd but above cum-ack is added + * together to cross check against the bit that we have yet to + * define (probably in the SACK). When the cum-ack is updated, this + * sum is updated as well. + */ + + /* Flags whether an extension is supported or not */ + uint8_t ecn_supported; + uint8_t prsctp_supported; + uint8_t auth_supported; + uint8_t asconf_supported; + uint8_t reconfig_supported; + uint8_t nrsack_supported; + uint8_t pktdrop_supported; + uint8_t idata_supported; + + /* Zero checksum supported information: + * 0: disabled + * 1: enabled for rcv + * 2: enabled for snd/rcv + */ + uint8_t zero_checksum; + + /* Did the peer make the stream config (add out) request */ + uint8_t peer_req_out; + + uint8_t local_strreset_support; + uint8_t peer_supports_nat; + + struct sctp_scoping scope; + /* flags to handle send alternate net tracking */ + uint8_t used_alt_asconfack; + uint8_t fast_retran_loss_recovery; + uint8_t sat_t3_loss_recovery; + uint8_t dropped_special_cnt; + uint8_t seen_a_sack_this_pkt; + uint8_t stream_reset_outstanding; + uint8_t stream_reset_out_is_outstanding; + uint8_t delayed_connection; + uint8_t ifp_had_enobuf; + uint8_t saw_sack_with_frags; + uint8_t saw_sack_with_nr_frags; + uint8_t in_asocid_hash; + uint8_t assoc_up_sent; + uint8_t adaptation_needed; + uint8_t adaptation_sent; + /* CMT variables */ + uint8_t cmt_dac_pkts_rcvd; + uint8_t sctp_cmt_on_off; + uint8_t iam_blocking; + uint8_t cookie_how[8]; + /* JRS 5/21/07 - CMT PF variable */ + uint8_t sctp_cmt_pf; + uint8_t use_precise_time; + uint64_t sctp_features; + uint32_t max_cwnd; + uint16_t port; /* remote UDP encapsulation port */ + /* + * The mapping array is used to track out of order sequences above + * last_acked_seq. 0 indicates packet missing 1 indicates packet + * rec'd. We slide it up every time we raise last_acked_seq and 0 + * trailing locactions out. If I get a TSN above the array + * mappingArraySz, I discard the datagram and let retransmit happen. + */ + uint32_t marked_retrans; + uint32_t timoinit; + uint32_t timodata; + uint32_t timosack; + uint32_t timoshutdown; + uint32_t timoheartbeat; + uint32_t timocookie; + uint32_t timoshutdownack; + struct timeval start_time; + struct timeval discontinuity_time; + uint64_t abandoned_unsent[SCTP_PR_SCTP_MAX + 1]; + uint64_t abandoned_sent[SCTP_PR_SCTP_MAX + 1]; +}; + +#endif diff --git a/netwerk/sctp/src/netinet/sctp_sysctl.c b/netwerk/sctp/src/netinet/sctp_sysctl.c new file mode 100644 index 0000000000..ffbe49e080 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_sysctl.c @@ -0,0 +1,1669 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#include <netinet/sctp.h> +#include <netinet/sctp_constants.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_output.h> +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/smp.h> +#include <sys/sysctl.h> +#endif +#if defined(__APPLE__) && !defined(__Userspace__) +#include <netinet/sctp_bsd_addr.h> +#endif + +#if defined(__FreeBSD__) && !defined(__Userspace__) +FEATURE(sctp, "Stream Control Transmission Protocol"); +#endif + +/* + * sysctl tunable variables + */ + +void +sctp_init_sysctls(void) +{ + SCTP_BASE_SYSCTL(sctp_sendspace) = SCTPCTL_MAXDGRAM_DEFAULT; + SCTP_BASE_SYSCTL(sctp_recvspace) = SCTPCTL_RECVSPACE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_auto_asconf) = SCTPCTL_AUTOASCONF_DEFAULT; + SCTP_BASE_SYSCTL(sctp_multiple_asconfs) = SCTPCTL_MULTIPLEASCONFS_DEFAULT; + SCTP_BASE_SYSCTL(sctp_ecn_enable) = SCTPCTL_ECN_ENABLE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_pr_enable) = SCTPCTL_PR_ENABLE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_auth_enable) = SCTPCTL_AUTH_ENABLE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_asconf_enable) = SCTPCTL_ASCONF_ENABLE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_reconfig_enable) = SCTPCTL_RECONFIG_ENABLE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_nrsack_enable) = SCTPCTL_NRSACK_ENABLE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_pktdrop_enable) = SCTPCTL_PKTDROP_ENABLE_DEFAULT; +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) = SCTPCTL_LOOPBACK_NOCSUM_DEFAULT; +#endif + SCTP_BASE_SYSCTL(sctp_peer_chunk_oh) = SCTPCTL_PEER_CHKOH_DEFAULT; + SCTP_BASE_SYSCTL(sctp_max_burst_default) = SCTPCTL_MAXBURST_DEFAULT; + SCTP_BASE_SYSCTL(sctp_fr_max_burst_default) = SCTPCTL_FRMAXBURST_DEFAULT; + SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue) = SCTPCTL_MAXCHUNKS_DEFAULT; +#if defined(__Userspace__) + if (SCTP_BASE_SYSCTL(sctp_hashtblsize) == 0) { + SCTP_BASE_SYSCTL(sctp_hashtblsize) = SCTPCTL_TCBHASHSIZE_DEFAULT; + } +#else + SCTP_BASE_SYSCTL(sctp_hashtblsize) = SCTPCTL_TCBHASHSIZE_DEFAULT; +#endif +#if defined(__Userspace__) + if (SCTP_BASE_SYSCTL(sctp_pcbtblsize) == 0) { + SCTP_BASE_SYSCTL(sctp_pcbtblsize) = SCTPCTL_PCBHASHSIZE_DEFAULT; + } +#else + SCTP_BASE_SYSCTL(sctp_pcbtblsize) = SCTPCTL_PCBHASHSIZE_DEFAULT; +#endif + SCTP_BASE_SYSCTL(sctp_min_split_point) = SCTPCTL_MIN_SPLIT_POINT_DEFAULT; +#if defined(__Userspace__) + if (SCTP_BASE_SYSCTL(sctp_chunkscale) == 0) { + SCTP_BASE_SYSCTL(sctp_chunkscale) = SCTPCTL_CHUNKSCALE_DEFAULT; + } +#else + SCTP_BASE_SYSCTL(sctp_chunkscale) = SCTPCTL_CHUNKSCALE_DEFAULT; +#endif + SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default) = SCTPCTL_DELAYED_SACK_TIME_DEFAULT; + SCTP_BASE_SYSCTL(sctp_sack_freq_default) = SCTPCTL_SACK_FREQ_DEFAULT; + SCTP_BASE_SYSCTL(sctp_system_free_resc_limit) = SCTPCTL_SYS_RESOURCE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_asoc_free_resc_limit) = SCTPCTL_ASOC_RESOURCE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default) = SCTPCTL_HEARTBEAT_INTERVAL_DEFAULT; + SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default) = SCTPCTL_PMTU_RAISE_TIME_DEFAULT; + SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default) = SCTPCTL_SHUTDOWN_GUARD_TIME_DEFAULT; + SCTP_BASE_SYSCTL(sctp_secret_lifetime_default) = SCTPCTL_SECRET_LIFETIME_DEFAULT; + SCTP_BASE_SYSCTL(sctp_rto_max_default) = SCTPCTL_RTO_MAX_DEFAULT; + SCTP_BASE_SYSCTL(sctp_rto_min_default) = SCTPCTL_RTO_MIN_DEFAULT; + SCTP_BASE_SYSCTL(sctp_rto_initial_default) = SCTPCTL_RTO_INITIAL_DEFAULT; + SCTP_BASE_SYSCTL(sctp_init_rto_max_default) = SCTPCTL_INIT_RTO_MAX_DEFAULT; + SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default) = SCTPCTL_VALID_COOKIE_LIFE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_init_rtx_max_default) = SCTPCTL_INIT_RTX_MAX_DEFAULT; + SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default) = SCTPCTL_ASSOC_RTX_MAX_DEFAULT; + SCTP_BASE_SYSCTL(sctp_path_rtx_max_default) = SCTPCTL_PATH_RTX_MAX_DEFAULT; + SCTP_BASE_SYSCTL(sctp_path_pf_threshold) = SCTPCTL_PATH_PF_THRESHOLD_DEFAULT; + SCTP_BASE_SYSCTL(sctp_add_more_threshold) = SCTPCTL_ADD_MORE_ON_OUTPUT_DEFAULT; + SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default) = SCTPCTL_INCOMING_STREAMS_DEFAULT; + SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default) = SCTPCTL_OUTGOING_STREAMS_DEFAULT; + SCTP_BASE_SYSCTL(sctp_cmt_on_off) = SCTPCTL_CMT_ON_OFF_DEFAULT; + SCTP_BASE_SYSCTL(sctp_cmt_use_dac) = SCTPCTL_CMT_USE_DAC_DEFAULT; + SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst) = SCTPCTL_CWND_MAXBURST_DEFAULT; + SCTP_BASE_SYSCTL(sctp_nat_friendly) = SCTPCTL_NAT_FRIENDLY_DEFAULT; + SCTP_BASE_SYSCTL(sctp_L2_abc_variable) = SCTPCTL_ABC_L_VAR_DEFAULT; + SCTP_BASE_SYSCTL(sctp_mbuf_threshold_count) = SCTPCTL_MAX_CHAINED_MBUFS_DEFAULT; + SCTP_BASE_SYSCTL(sctp_do_drain) = SCTPCTL_DO_SCTP_DRAIN_DEFAULT; + SCTP_BASE_SYSCTL(sctp_hb_maxburst) = SCTPCTL_HB_MAX_BURST_DEFAULT; + SCTP_BASE_SYSCTL(sctp_abort_if_one_2_one_hits_limit) = SCTPCTL_ABORT_AT_LIMIT_DEFAULT; + SCTP_BASE_SYSCTL(sctp_min_residual) = SCTPCTL_MIN_RESIDUAL_DEFAULT; + SCTP_BASE_SYSCTL(sctp_max_retran_chunk) = SCTPCTL_MAX_RETRAN_CHUNK_DEFAULT; + SCTP_BASE_SYSCTL(sctp_logging_level) = SCTPCTL_LOGGING_LEVEL_DEFAULT; + SCTP_BASE_SYSCTL(sctp_default_cc_module) = SCTPCTL_DEFAULT_CC_MODULE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_default_ss_module) = SCTPCTL_DEFAULT_SS_MODULE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_default_frag_interleave) = SCTPCTL_DEFAULT_FRAG_INTERLEAVE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_mobility_base) = SCTPCTL_MOBILITY_BASE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) = SCTPCTL_MOBILITY_FASTHANDOFF_DEFAULT; + SCTP_BASE_SYSCTL(sctp_vtag_time_wait) = SCTPCTL_TIME_WAIT_DEFAULT; + SCTP_BASE_SYSCTL(sctp_buffer_splitting) = SCTPCTL_BUFFER_SPLITTING_DEFAULT; + SCTP_BASE_SYSCTL(sctp_initial_cwnd) = SCTPCTL_INITIAL_CWND_DEFAULT; + SCTP_BASE_SYSCTL(sctp_rttvar_bw) = SCTPCTL_RTTVAR_BW_DEFAULT; + SCTP_BASE_SYSCTL(sctp_rttvar_rtt) = SCTPCTL_RTTVAR_RTT_DEFAULT; + SCTP_BASE_SYSCTL(sctp_rttvar_eqret) = SCTPCTL_RTTVAR_EQRET_DEFAULT; + SCTP_BASE_SYSCTL(sctp_steady_step) = SCTPCTL_RTTVAR_STEADYS_DEFAULT; + SCTP_BASE_SYSCTL(sctp_use_dccc_ecn) = SCTPCTL_RTTVAR_DCCCECN_DEFAULT; + SCTP_BASE_SYSCTL(sctp_blackhole) = SCTPCTL_BLACKHOLE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_sendall_limit) = SCTPCTL_SENDALL_LIMIT_DEFAULT; + SCTP_BASE_SYSCTL(sctp_diag_info_code) = SCTPCTL_DIAG_INFO_CODE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_ootb_with_zero_cksum) = SCTPCTL_OOTB_WITH_ZERO_CKSUM_DEFAULT; +#if defined(SCTP_LOCAL_TRACE_BUF) +#if defined(_WIN32) && !defined(__Userspace__) + /* On Windows, the resource for global variables is limited. */ + MALLOC(SCTP_BASE_SYSCTL(sctp_log), struct sctp_log *, sizeof(struct sctp_log), M_SYSCTL, M_ZERO); +#else + memset(&SCTP_BASE_SYSCTL(sctp_log), 0, sizeof(struct sctp_log)); +#endif +#endif + SCTP_BASE_SYSCTL(sctp_udp_tunneling_port) = SCTPCTL_UDP_TUNNELING_PORT_DEFAULT; + SCTP_BASE_SYSCTL(sctp_enable_sack_immediately) = SCTPCTL_SACK_IMMEDIATELY_ENABLE_DEFAULT; + SCTP_BASE_SYSCTL(sctp_inits_include_nat_friendly) = SCTPCTL_NAT_FRIENDLY_INITS_DEFAULT; +#if defined(SCTP_DEBUG) + SCTP_BASE_SYSCTL(sctp_debug_on) = SCTPCTL_DEBUG_DEFAULT; +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_BASE_SYSCTL(sctp_ignore_vmware_interfaces) = SCTPCTL_IGNORE_VMWARE_INTERFACES_DEFAULT; + SCTP_BASE_SYSCTL(sctp_main_timer) = SCTPCTL_MAIN_TIMER_DEFAULT; + SCTP_BASE_SYSCTL(sctp_addr_watchdog_limit) = SCTPCTL_ADDR_WATCHDOG_LIMIT_DEFAULT; + SCTP_BASE_SYSCTL(sctp_vtag_watchdog_limit) = SCTPCTL_VTAG_WATCHDOG_LIMIT_DEFAULT; +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_BASE_SYSCTL(sctp_output_unlocked) = SCTPCTL_OUTPUT_UNLOCKED_DEFAULT; +#endif +} +#if defined(_WIN32) && !defined(__Userspace__) + +void +sctp_finish_sysctls() +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + if (SCTP_BASE_SYSCTL(sctp_log) != NULL) { + FREE(SCTP_BASE_SYSCTL(sctp_log), M_SYSCTL); + SCTP_BASE_SYSCTL(sctp_log) = NULL; + } +#endif +} +#endif + +#if !defined(__Userspace__) +/* It returns an upper limit. No filtering is done here */ +static unsigned int +sctp_sysctl_number_of_addresses(struct sctp_inpcb *inp) +{ + unsigned int cnt; + struct sctp_vrf *vrf; + struct sctp_ifn *sctp_ifn; + struct sctp_ifa *sctp_ifa; + struct sctp_laddr *laddr; + + cnt = 0; + /* neither Mac OS X nor FreeBSD support multiple routing functions */ + if ((vrf = sctp_find_vrf(inp->def_vrf_id)) == NULL) { + return (0); + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { + switch (sctp_ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: +#endif +#ifdef INET6 + case AF_INET6: +#endif + cnt++; + break; + default: + break; + } + } + } + } else { + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + switch (laddr->ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: +#endif +#ifdef INET6 + case AF_INET6: +#endif + cnt++; + break; + default: + break; + } + } + } + return (cnt); +} + +static int +sctp_sysctl_copy_out_local_addresses(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sysctl_req *req) +{ + struct sctp_ifn *sctp_ifn; + struct sctp_ifa *sctp_ifa; + int loopback_scope; +#ifdef INET + int ipv4_local_scope; + int ipv4_addr_legal; +#endif +#ifdef INET6 + int local_scope, site_scope; + int ipv6_addr_legal; +#endif +#if defined(__Userspace__) + int conn_addr_legal; +#endif + struct sctp_vrf *vrf; + struct xsctp_laddr xladdr; + struct sctp_laddr *laddr; + int error; + + /* Turn on all the appropriate scope */ + if (stcb != NULL) { + /* use association specific values */ + loopback_scope = stcb->asoc.scope.loopback_scope; +#ifdef INET + ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope; + ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal; +#endif +#ifdef INET6 + local_scope = stcb->asoc.scope.local_scope; + site_scope = stcb->asoc.scope.site_scope; + ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal; +#endif +#if defined(__Userspace__) + conn_addr_legal = stcb->asoc.scope.conn_addr_legal; +#endif + } else { + /* Use generic values for endpoints. */ + loopback_scope = 1; +#ifdef INET + ipv4_local_scope = 1; +#endif +#ifdef INET6 + local_scope = 1; + site_scope = 1; +#endif + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { +#ifdef INET6 + ipv6_addr_legal = 1; +#endif +#ifdef INET + if (SCTP_IPV6_V6ONLY(inp)) { + ipv4_addr_legal = 0; + } else { + ipv4_addr_legal = 1; + } +#endif +#if defined(__Userspace__) + conn_addr_legal = 0; +#endif + } else { +#ifdef INET6 + ipv6_addr_legal = 0; +#endif +#if defined(__Userspace__) + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_CONN) { + conn_addr_legal = 1; +#ifdef INET + ipv4_addr_legal = 0; +#endif + } else { + conn_addr_legal = 0; +#ifdef INET + ipv4_addr_legal = 1; +#endif + } +#else +#ifdef INET + ipv4_addr_legal = 1; +#endif +#endif + } + } + + /* Neither Mac OS X nor FreeBSD support multiple routing functions. */ + if ((vrf = sctp_find_vrf(inp->def_vrf_id)) == NULL) { + SCTP_INP_RUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + return (ENOENT); + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { + if ((loopback_scope == 0) && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { + /* Skip loopback if loopback_scope not set. */ + continue; + } + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { + if (stcb != NULL) { + /* + * Ignore if blacklisted at + * association level. + */ + if (sctp_is_addr_restricted(stcb, sctp_ifa)) { + continue; + } + } + switch (sctp_ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: + if (ipv4_addr_legal) { + struct sockaddr_in *sin; + + sin = &sctp_ifa->address.sin; + if (sin->sin_addr.s_addr == 0) { + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sin->sin_addr) != 0) { + continue; + } +#endif + if ((ipv4_local_scope == 0) && (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { + continue; + } + } else { + continue; + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (ipv6_addr_legal) { + struct sockaddr_in6 *sin6; + + sin6 = &sctp_ifa->address.sin6; + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sin6->sin6_addr) != 0) { + continue; + } +#endif + if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { + if (local_scope == 0) { + continue; + } + } + if ((site_scope == 0) && (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { + continue; + } + } else { + continue; + } + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + if (!conn_addr_legal) { + continue; + } + break; +#endif + default: + continue; + } + memset((void *)&xladdr, 0, sizeof(struct xsctp_laddr)); + memcpy((void *)&xladdr.address, (const void *)&sctp_ifa->address, sizeof(union sctp_sockstore)); + SCTP_INP_RUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + error = SYSCTL_OUT(req, &xladdr, sizeof(struct xsctp_laddr)); + if (error != 0) { + return (error); + } else { + SCTP_INP_INFO_RLOCK(); + SCTP_INP_RLOCK(inp); + } + } + } + } else { + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + /* ignore if blacklisted at association level */ + if (stcb != NULL && sctp_is_addr_restricted(stcb, laddr->ifa)) + continue; + memset((void *)&xladdr, 0, sizeof(struct xsctp_laddr)); + memcpy((void *)&xladdr.address, (const void *)&laddr->ifa->address, sizeof(union sctp_sockstore)); + xladdr.start_time.tv_sec = (uint32_t)laddr->start_time.tv_sec; + xladdr.start_time.tv_usec = (uint32_t)laddr->start_time.tv_usec; + SCTP_INP_RUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + error = SYSCTL_OUT(req, &xladdr, sizeof(struct xsctp_laddr)); + if (error != 0) { + return (error); + } else { + SCTP_INP_INFO_RLOCK(); + SCTP_INP_RLOCK(inp); + } + } + } + memset((void *)&xladdr, 0, sizeof(struct xsctp_laddr)); + xladdr.last = 1; + SCTP_INP_RUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + error = SYSCTL_OUT(req, &xladdr, sizeof(struct xsctp_laddr)); + + if (error != 0) { + return (error); + } else { + SCTP_INP_INFO_RLOCK(); + SCTP_INP_RLOCK(inp); + return (0); + } +} + +/* + * sysctl functions + */ +#if defined(__APPLE__) && !defined(__Userspace__) +static int +sctp_sysctl_handle_assoclist SYSCTL_HANDLER_ARGS +{ +#pragma unused(oidp, arg1, arg2) +#else +static int +sctp_sysctl_handle_assoclist(SYSCTL_HANDLER_ARGS) +{ +#endif + unsigned int number_of_endpoints; + unsigned int number_of_local_addresses; + unsigned int number_of_associations; + unsigned int number_of_remote_addresses; + unsigned int n; + int error; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; + struct sctp_nets *net; + struct xsctp_inpcb xinpcb; + struct xsctp_tcb xstcb; + struct xsctp_raddr xraddr; + struct socket *so; + + number_of_endpoints = 0; + number_of_local_addresses = 0; + number_of_associations = 0; + number_of_remote_addresses = 0; + + SCTP_INP_INFO_RLOCK(); +#if defined(__APPLE__) && !defined(__Userspace__) + if (req->oldptr == USER_ADDR_NULL) { +#else + if (req->oldptr == NULL) { +#endif + LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) { + SCTP_INP_RLOCK(inp); + number_of_endpoints++; + number_of_local_addresses += sctp_sysctl_number_of_addresses(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + number_of_associations++; + number_of_local_addresses += sctp_sysctl_number_of_addresses(inp); + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + number_of_remote_addresses++; + } + } + SCTP_INP_RUNLOCK(inp); + } + SCTP_INP_INFO_RUNLOCK(); + n = (number_of_endpoints + 1) * sizeof(struct xsctp_inpcb) + + (number_of_local_addresses + number_of_endpoints + number_of_associations) * sizeof(struct xsctp_laddr) + + (number_of_associations + number_of_endpoints) * sizeof(struct xsctp_tcb) + + (number_of_remote_addresses + number_of_associations) * sizeof(struct xsctp_raddr); + + /* request some more memory than needed */ +#if !(defined(_WIN32) && !defined(__Userspace__)) + req->oldidx = (n + n / 8); +#else + req->dataidx = (n + n / 8); +#endif + return (0); + } +#if defined(__APPLE__) && !defined(__Userspace__) + if (req->newptr != USER_ADDR_NULL) { +#else + if (req->newptr != NULL) { +#endif + SCTP_INP_INFO_RUNLOCK(); + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_SYSCTL, EPERM); + return (EPERM); + } + memset(&xinpcb, 0, sizeof(xinpcb)); + memset(&xstcb, 0, sizeof(xstcb)); + memset(&xraddr, 0, sizeof(xraddr)); + LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) { + SCTP_INP_RLOCK(inp); + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { + /* if its allgone it is being freed - skip it */ + goto skip; + } + xinpcb.last = 0; + xinpcb.local_port = ntohs(inp->sctp_lport); + xinpcb.flags = inp->sctp_flags; + xinpcb.features = inp->sctp_features; + xinpcb.total_sends = inp->total_sends; + xinpcb.total_recvs = inp->total_recvs; + xinpcb.total_nospaces = inp->total_nospaces; + xinpcb.fragmentation_point = inp->sctp_frag_point; +#if defined(__FreeBSD__) && !defined(__Userspace__) + xinpcb.socket = (uintptr_t)inp->sctp_socket; +#else + xinpcb.socket = inp->sctp_socket; +#endif + so = inp->sctp_socket; + if ((so == NULL) || + (!SCTP_IS_LISTENING(inp)) || + (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { + xinpcb.qlen = 0; + xinpcb.maxqlen = 0; + } else { +#if defined(__FreeBSD__) && !defined(__Userspace__) + xinpcb.qlen = so->sol_qlen; + xinpcb.qlen_old = so->sol_qlen > USHRT_MAX ? + USHRT_MAX : (uint16_t) so->sol_qlen; + xinpcb.maxqlen = so->sol_qlimit; + xinpcb.maxqlen_old = so->sol_qlimit > USHRT_MAX ? + USHRT_MAX : (uint16_t) so->sol_qlimit; +#else + xinpcb.qlen = so->so_qlen; + xinpcb.maxqlen = so->so_qlimit; +#endif + } + SCTP_INP_INCR_REF(inp); + SCTP_INP_RUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + error = SYSCTL_OUT(req, &xinpcb, sizeof(struct xsctp_inpcb)); + if (error) { + SCTP_INP_DECR_REF(inp); + return (error); + } + SCTP_INP_INFO_RLOCK(); + SCTP_INP_RLOCK(inp); + error = sctp_sysctl_copy_out_local_addresses(inp, NULL, req); + if (error) { + SCTP_INP_DECR_REF(inp); + return (error); + } + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + xstcb.last = 0; + xstcb.local_port = ntohs(inp->sctp_lport); + xstcb.remote_port = ntohs(stcb->rport); + if (stcb->asoc.primary_destination != NULL) + xstcb.primary_addr = stcb->asoc.primary_destination->ro._l_addr; + xstcb.heartbeat_interval = stcb->asoc.heart_beat_delay; + xstcb.state = (uint32_t)sctp_map_assoc_state(stcb->asoc.state); + xstcb.assoc_id = sctp_get_associd(stcb); + xstcb.peers_rwnd = stcb->asoc.peers_rwnd; + xstcb.in_streams = stcb->asoc.streamincnt; + xstcb.out_streams = stcb->asoc.streamoutcnt; + xstcb.max_nr_retrans = stcb->asoc.overall_error_count; + xstcb.primary_process = 0; /* not really supported yet */ + xstcb.T1_expireries = stcb->asoc.timoinit + stcb->asoc.timocookie; + xstcb.T2_expireries = stcb->asoc.timoshutdown + stcb->asoc.timoshutdownack; + xstcb.retransmitted_tsns = stcb->asoc.marked_retrans; + xstcb.start_time.tv_sec = (uint32_t)stcb->asoc.start_time.tv_sec; + xstcb.start_time.tv_usec = (uint32_t)stcb->asoc.start_time.tv_usec; + xstcb.discontinuity_time.tv_sec = (uint32_t)stcb->asoc.discontinuity_time.tv_sec; + xstcb.discontinuity_time.tv_usec = (uint32_t)stcb->asoc.discontinuity_time.tv_usec; + xstcb.total_sends = stcb->total_sends; + xstcb.total_recvs = stcb->total_recvs; + xstcb.local_tag = stcb->asoc.my_vtag; + xstcb.remote_tag = stcb->asoc.peer_vtag; + xstcb.initial_tsn = stcb->asoc.init_seq_number; + xstcb.highest_tsn = stcb->asoc.sending_seq - 1; + xstcb.cumulative_tsn = stcb->asoc.last_acked_seq; + xstcb.cumulative_tsn_ack = stcb->asoc.cumulative_tsn; + xstcb.mtu = stcb->asoc.smallest_mtu; + xstcb.refcnt = stcb->asoc.refcnt; + SCTP_INP_RUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + error = SYSCTL_OUT(req, &xstcb, sizeof(struct xsctp_tcb)); + if (error) { + SCTP_INP_DECR_REF(inp); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + return (error); + } + SCTP_INP_INFO_RLOCK(); + SCTP_INP_RLOCK(inp); + error = sctp_sysctl_copy_out_local_addresses(inp, stcb, req); + if (error) { + SCTP_INP_DECR_REF(inp); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + return (error); + } + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + xraddr.last = 0; + xraddr.address = net->ro._l_addr; + xraddr.active = ((net->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE); + xraddr.confirmed = ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0); + xraddr.heartbeat_enabled = ((net->dest_state & SCTP_ADDR_NOHB) == 0); + xraddr.potentially_failed = ((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF); + xraddr.rto = net->RTO; + xraddr.max_path_rtx = net->failure_threshold; + xraddr.rtx = net->marked_retrans; + xraddr.error_counter = net->error_count; + xraddr.cwnd = net->cwnd; + xraddr.flight_size = net->flight_size; + xraddr.mtu = net->mtu; + xraddr.rtt = net->rtt / 1000; + xraddr.heartbeat_interval = net->heart_beat_delay; + xraddr.ssthresh = net->ssthresh; + xraddr.encaps_port = net->port; + if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { + xraddr.state = SCTP_UNCONFIRMED; + } else if (net->dest_state & SCTP_ADDR_REACHABLE) { + xraddr.state = SCTP_ACTIVE; + } else { + xraddr.state = SCTP_INACTIVE; + } + xraddr.start_time.tv_sec = (uint32_t)net->start_time.tv_sec; + xraddr.start_time.tv_usec = (uint32_t)net->start_time.tv_usec; + SCTP_INP_RUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + error = SYSCTL_OUT(req, &xraddr, sizeof(struct xsctp_raddr)); + if (error) { + SCTP_INP_DECR_REF(inp); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + return (error); + } + SCTP_INP_INFO_RLOCK(); + SCTP_INP_RLOCK(inp); + } + atomic_subtract_int(&stcb->asoc.refcnt, 1); + memset((void *)&xraddr, 0, sizeof(struct xsctp_raddr)); + xraddr.last = 1; + SCTP_INP_RUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + error = SYSCTL_OUT(req, &xraddr, sizeof(struct xsctp_raddr)); + if (error) { + SCTP_INP_DECR_REF(inp); + return (error); + } + SCTP_INP_INFO_RLOCK(); + SCTP_INP_RLOCK(inp); + } + SCTP_INP_DECR_REF(inp); + SCTP_INP_RUNLOCK(inp); + SCTP_INP_INFO_RUNLOCK(); + memset((void *)&xstcb, 0, sizeof(struct xsctp_tcb)); + xstcb.last = 1; + error = SYSCTL_OUT(req, &xstcb, sizeof(struct xsctp_tcb)); + if (error) { + return (error); + } +skip: + SCTP_INP_INFO_RLOCK(); + } + SCTP_INP_INFO_RUNLOCK(); + + memset((void *)&xinpcb, 0, sizeof(struct xsctp_inpcb)); + xinpcb.last = 1; + error = SYSCTL_OUT(req, &xinpcb, sizeof(struct xsctp_inpcb)); + return (error); +} + +#if defined(__APPLE__) && !defined(__Userspace__) +static int +sctp_sysctl_handle_udp_tunneling SYSCTL_HANDLER_ARGS +{ +#pragma unused(arg1, arg2) +#else +static int +sctp_sysctl_handle_udp_tunneling(SYSCTL_HANDLER_ARGS) +{ +#endif + int error; + uint32_t old, new; + + SCTP_INP_INFO_RLOCK(); + old = SCTP_BASE_SYSCTL(sctp_udp_tunneling_port); + SCTP_INP_INFO_RUNLOCK(); + new = old; + error = sysctl_handle_int(oidp, &new, 0, req); + if ((error == 0) && +#if defined(__APPLE__) && !defined(__Userspace__) + (req->newptr != USER_ADDR_NULL)) { +#else + (req->newptr != NULL)) { +#endif +#if defined(_WIN32) && !defined(__Userspace__) + SCTP_INP_INFO_WLOCK(); + sctp_over_udp_restart(); + SCTP_INP_INFO_WUNLOCK(); +#else +#if (SCTPCTL_UDP_TUNNELING_PORT_MIN == 0) + if (new > SCTPCTL_UDP_TUNNELING_PORT_MAX) { +#else + if ((new < SCTPCTL_UDP_TUNNELING_PORT_MIN) || + (new > SCTPCTL_UDP_TUNNELING_PORT_MAX)) { +#endif + error = EINVAL; + } else { + SCTP_INP_INFO_WLOCK(); + SCTP_BASE_SYSCTL(sctp_udp_tunneling_port) = new; + if (old != 0) { + sctp_over_udp_stop(); + } + if (new != 0) { + error = sctp_over_udp_start(); + } + SCTP_INP_INFO_WUNLOCK(); + } +#endif + } + return (error); +} +#if defined(__APPLE__) && !defined(__Userspace__) + +int sctp_is_vmware_interface(struct ifnet *); + +static int +sctp_sysctl_handle_vmware_interfaces SYSCTL_HANDLER_ARGS +{ +#pragma unused(arg1, arg2) + int error; + uint32_t old, new; + + old = SCTP_BASE_SYSCTL(sctp_ignore_vmware_interfaces); + new = old; + error = sysctl_handle_int(oidp, &new, 0, req); + if ((error == 0) && (req->newptr != USER_ADDR_NULL)) { + if ((new < SCTPCTL_IGNORE_VMWARE_INTERFACES_MIN) || + (new > SCTPCTL_IGNORE_VMWARE_INTERFACES_MAX)) { + error = EINVAL; + } else { + if ((old == 1) && (new == 0)) { + sctp_add_or_del_interfaces(sctp_is_vmware_interface, 1); + } + if ((old == 0) && (new == 1)) { + sctp_add_or_del_interfaces(sctp_is_vmware_interface, 0); + } + if (old != new) { + SCTP_BASE_SYSCTL(sctp_ignore_vmware_interfaces) = new; + } + } + } + return (error); +} +#endif + +#if defined(__APPLE__) && !defined(__Userspace__) +static int +sctp_sysctl_handle_auth SYSCTL_HANDLER_ARGS +{ +#pragma unused(arg1, arg2) +#else +static int +sctp_sysctl_handle_auth(SYSCTL_HANDLER_ARGS) +{ +#endif + int error; + uint32_t new; + + new = SCTP_BASE_SYSCTL(sctp_auth_enable); + error = sysctl_handle_int(oidp, &new, 0, req); + if ((error == 0) && +#if defined(__APPLE__) && !defined(__Userspace__) + (req->newptr != USER_ADDR_NULL)) { +#else + (req->newptr != NULL)) { +#endif +#if (SCTPCTL_AUTH_ENABLE_MIN == 0) + if ((new > SCTPCTL_AUTH_ENABLE_MAX) || + ((new == 0) && (SCTP_BASE_SYSCTL(sctp_asconf_enable) == 1))) { +#else + if ((new < SCTPCTL_AUTH_ENABLE_MIN) || + (new > SCTPCTL_AUTH_ENABLE_MAX) || + ((new == 0) && (SCTP_BASE_SYSCTL(sctp_asconf_enable) == 1))) { +#endif + error = EINVAL; + } else { + SCTP_BASE_SYSCTL(sctp_auth_enable) = new; + } + } + return (error); +} + +#if defined(__APPLE__) && !defined(__Userspace__) +static int +sctp_sysctl_handle_asconf SYSCTL_HANDLER_ARGS +{ +#pragma unused(arg1, arg2) +#else +static int +sctp_sysctl_handle_asconf(SYSCTL_HANDLER_ARGS) +{ +#endif + int error; + uint32_t new; + + new = SCTP_BASE_SYSCTL(sctp_asconf_enable); + error = sysctl_handle_int(oidp, &new, 0, req); + if ((error == 0) && +#if defined(__APPLE__) && !defined(__Userspace__) + (req->newptr != USER_ADDR_NULL)) { +#else + (req->newptr != NULL)) { +#endif +#if (SCTPCTL_ASCONF_ENABLE_MIN == 0) + if ((new > SCTPCTL_ASCONF_ENABLE_MAX) || + ((new == 1) && (SCTP_BASE_SYSCTL(sctp_auth_enable) == 0))) { +#else + if ((new < SCTPCTL_ASCONF_ENABLE_MIN) || + (new > SCTPCTL_ASCONF_ENABLE_MAX) || + ((new == 1) && (SCTP_BASE_SYSCTL(sctp_auth_enable) == 0))) { +#endif + error = EINVAL; + } else { + SCTP_BASE_SYSCTL(sctp_asconf_enable) = new; + } + } + return (error); +} + +#if defined(__APPLE__) && !defined(__Userspace__) +static int +sctp_sysctl_handle_stats SYSCTL_HANDLER_ARGS +{ +#pragma unused(oidp, arg1, arg2) +#else +static int +sctp_sysctl_handle_stats(SYSCTL_HANDLER_ARGS) +{ +#endif + int error; +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT) + struct sctpstat *sarry; + struct sctpstat sb; + int cpu; +#endif + struct sctpstat sb_temp; +#endif + +#if defined(__APPLE__) && !defined(__Userspace__) + if ((req->newptr != USER_ADDR_NULL) && +#else + if ((req->newptr != NULL) && +#endif + (req->newlen != sizeof(struct sctpstat))) { + return (EINVAL); + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + memset(&sb_temp, 0, sizeof(struct sctpstat)); + + if (req->newptr != NULL) { + error = SYSCTL_IN(req, &sb_temp, sizeof(struct sctpstat)); + if (error != 0) { + return (error); + } + } +#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT) + memset(&sb, 0, sizeof(sb)); + for (cpu = 0; cpu < mp_maxid; cpu++) { + sarry = &SCTP_BASE_STATS[cpu]; + if (sarry->sctps_discontinuitytime.tv_sec > sb.sctps_discontinuitytime.tv_sec) { + sb.sctps_discontinuitytime.tv_sec = sarry->sctps_discontinuitytime.tv_sec; + sb.sctps_discontinuitytime.tv_usec = sarry->sctps_discontinuitytime.tv_usec; + } + sb.sctps_currestab += sarry->sctps_currestab; + sb.sctps_activeestab += sarry->sctps_activeestab; + sb.sctps_restartestab += sarry->sctps_restartestab; + sb.sctps_collisionestab += sarry->sctps_collisionestab; + sb.sctps_passiveestab += sarry->sctps_passiveestab; + sb.sctps_aborted += sarry->sctps_aborted; + sb.sctps_shutdown += sarry->sctps_shutdown; + sb.sctps_outoftheblue += sarry->sctps_outoftheblue; + sb.sctps_checksumerrors += sarry->sctps_checksumerrors; + sb.sctps_outcontrolchunks += sarry->sctps_outcontrolchunks; + sb.sctps_outorderchunks += sarry->sctps_outorderchunks; + sb.sctps_outunorderchunks += sarry->sctps_outunorderchunks; + sb.sctps_incontrolchunks += sarry->sctps_incontrolchunks; + sb.sctps_inorderchunks += sarry->sctps_inorderchunks; + sb.sctps_inunorderchunks += sarry->sctps_inunorderchunks; + sb.sctps_fragusrmsgs += sarry->sctps_fragusrmsgs; + sb.sctps_reasmusrmsgs += sarry->sctps_reasmusrmsgs; + sb.sctps_outpackets += sarry->sctps_outpackets; + sb.sctps_inpackets += sarry->sctps_inpackets; + sb.sctps_recvpackets += sarry->sctps_recvpackets; + sb.sctps_recvdatagrams += sarry->sctps_recvdatagrams; + sb.sctps_recvpktwithdata += sarry->sctps_recvpktwithdata; + sb.sctps_recvsacks += sarry->sctps_recvsacks; + sb.sctps_recvdata += sarry->sctps_recvdata; + sb.sctps_recvdupdata += sarry->sctps_recvdupdata; + sb.sctps_recvheartbeat += sarry->sctps_recvheartbeat; + sb.sctps_recvheartbeatack += sarry->sctps_recvheartbeatack; + sb.sctps_recvecne += sarry->sctps_recvecne; + sb.sctps_recvauth += sarry->sctps_recvauth; + sb.sctps_recvauthmissing += sarry->sctps_recvauthmissing; + sb.sctps_recvivalhmacid += sarry->sctps_recvivalhmacid; + sb.sctps_recvivalkeyid += sarry->sctps_recvivalkeyid; + sb.sctps_recvauthfailed += sarry->sctps_recvauthfailed; + sb.sctps_recvexpress += sarry->sctps_recvexpress; + sb.sctps_recvexpressm += sarry->sctps_recvexpressm; + sb.sctps_recvswcrc += sarry->sctps_recvswcrc; + sb.sctps_recvhwcrc += sarry->sctps_recvhwcrc; + sb.sctps_sendpackets += sarry->sctps_sendpackets; + sb.sctps_sendsacks += sarry->sctps_sendsacks; + sb.sctps_senddata += sarry->sctps_senddata; + sb.sctps_sendretransdata += sarry->sctps_sendretransdata; + sb.sctps_sendfastretrans += sarry->sctps_sendfastretrans; + sb.sctps_sendmultfastretrans += sarry->sctps_sendmultfastretrans; + sb.sctps_sendheartbeat += sarry->sctps_sendheartbeat; + sb.sctps_sendecne += sarry->sctps_sendecne; + sb.sctps_sendauth += sarry->sctps_sendauth; + sb.sctps_senderrors += sarry->sctps_senderrors; + sb.sctps_sendswcrc += sarry->sctps_sendswcrc; + sb.sctps_sendhwcrc += sarry->sctps_sendhwcrc; + sb.sctps_pdrpfmbox += sarry->sctps_pdrpfmbox; + sb.sctps_pdrpfehos += sarry->sctps_pdrpfehos; + sb.sctps_pdrpmbda += sarry->sctps_pdrpmbda; + sb.sctps_pdrpmbct += sarry->sctps_pdrpmbct; + sb.sctps_pdrpbwrpt += sarry->sctps_pdrpbwrpt; + sb.sctps_pdrpcrupt += sarry->sctps_pdrpcrupt; + sb.sctps_pdrpnedat += sarry->sctps_pdrpnedat; + sb.sctps_pdrppdbrk += sarry->sctps_pdrppdbrk; + sb.sctps_pdrptsnnf += sarry->sctps_pdrptsnnf; + sb.sctps_pdrpdnfnd += sarry->sctps_pdrpdnfnd; + sb.sctps_pdrpdiwnp += sarry->sctps_pdrpdiwnp; + sb.sctps_pdrpdizrw += sarry->sctps_pdrpdizrw; + sb.sctps_pdrpbadd += sarry->sctps_pdrpbadd; + sb.sctps_pdrpmark += sarry->sctps_pdrpmark; + sb.sctps_timoiterator += sarry->sctps_timoiterator; + sb.sctps_timodata += sarry->sctps_timodata; + sb.sctps_timowindowprobe += sarry->sctps_timowindowprobe; + sb.sctps_timoinit += sarry->sctps_timoinit; + sb.sctps_timosack += sarry->sctps_timosack; + sb.sctps_timoshutdown += sarry->sctps_timoshutdown; + sb.sctps_timoheartbeat += sarry->sctps_timoheartbeat; + sb.sctps_timocookie += sarry->sctps_timocookie; + sb.sctps_timosecret += sarry->sctps_timosecret; + sb.sctps_timopathmtu += sarry->sctps_timopathmtu; + sb.sctps_timoshutdownack += sarry->sctps_timoshutdownack; + sb.sctps_timoshutdownguard += sarry->sctps_timoshutdownguard; + sb.sctps_timostrmrst += sarry->sctps_timostrmrst; + sb.sctps_timoearlyfr += sarry->sctps_timoearlyfr; + sb.sctps_timoasconf += sarry->sctps_timoasconf; + sb.sctps_timodelprim += sarry->sctps_timodelprim; + sb.sctps_timoautoclose += sarry->sctps_timoautoclose; + sb.sctps_timoassockill += sarry->sctps_timoassockill; + sb.sctps_timoinpkill += sarry->sctps_timoinpkill; + sb.sctps_hdrops += sarry->sctps_hdrops; + sb.sctps_badsum += sarry->sctps_badsum; + sb.sctps_noport += sarry->sctps_noport; + sb.sctps_badvtag += sarry->sctps_badvtag; + sb.sctps_badsid += sarry->sctps_badsid; + sb.sctps_nomem += sarry->sctps_nomem; + sb.sctps_fastretransinrtt += sarry->sctps_fastretransinrtt; + sb.sctps_markedretrans += sarry->sctps_markedretrans; + sb.sctps_naglesent += sarry->sctps_naglesent; + sb.sctps_naglequeued += sarry->sctps_naglequeued; + sb.sctps_maxburstqueued += sarry->sctps_maxburstqueued; + sb.sctps_ifnomemqueued += sarry->sctps_ifnomemqueued; + sb.sctps_windowprobed += sarry->sctps_windowprobed; + sb.sctps_lowlevelerr += sarry->sctps_lowlevelerr; + sb.sctps_lowlevelerrusr += sarry->sctps_lowlevelerrusr; + sb.sctps_datadropchklmt += sarry->sctps_datadropchklmt; + sb.sctps_datadroprwnd += sarry->sctps_datadroprwnd; + sb.sctps_ecnereducedcwnd += sarry->sctps_ecnereducedcwnd; + sb.sctps_vtagexpress += sarry->sctps_vtagexpress; + sb.sctps_vtagbogus += sarry->sctps_vtagbogus; + sb.sctps_primary_randry += sarry->sctps_primary_randry; + sb.sctps_cmt_randry += sarry->sctps_cmt_randry; + sb.sctps_slowpath_sack += sarry->sctps_slowpath_sack; + sb.sctps_wu_sacks_sent += sarry->sctps_wu_sacks_sent; + sb.sctps_sends_with_flags += sarry->sctps_sends_with_flags; + sb.sctps_sends_with_unord += sarry->sctps_sends_with_unord; + sb.sctps_sends_with_eof += sarry->sctps_sends_with_eof; + sb.sctps_sends_with_abort += sarry->sctps_sends_with_abort; + sb.sctps_protocol_drain_calls += sarry->sctps_protocol_drain_calls; + sb.sctps_protocol_drains_done += sarry->sctps_protocol_drains_done; + sb.sctps_read_peeks += sarry->sctps_read_peeks; + sb.sctps_cached_chk += sarry->sctps_cached_chk; + sb.sctps_cached_strmoq += sarry->sctps_cached_strmoq; + sb.sctps_left_abandon += sarry->sctps_left_abandon; + sb.sctps_send_burst_avoid += sarry->sctps_send_burst_avoid; + sb.sctps_send_cwnd_avoid += sarry->sctps_send_cwnd_avoid; + sb.sctps_fwdtsn_map_over += sarry->sctps_fwdtsn_map_over; + sb.sctps_queue_upd_ecne += sarry->sctps_queue_upd_ecne; + sb.sctps_recvzerocrc += sarry->sctps_recvzerocrc; + sb.sctps_sendzerocrc += sarry->sctps_sendzerocrc; + if (req->newptr != NULL) { + memcpy(sarry, &sb_temp, sizeof(struct sctpstat)); + } + } + error = SYSCTL_OUT(req, &sb, sizeof(struct sctpstat)); +#else + error = SYSCTL_OUT(req, &SCTP_BASE_STATS, sizeof(struct sctpstat)); + if (error != 0) { + return (error); + } + if (req->newptr != NULL) { + memcpy(&SCTP_BASE_STATS, &sb_temp, sizeof(struct sctpstat)); + } +#endif +#else + error = SYSCTL_OUT(req, &SCTP_BASE_STATS, sizeof(struct sctpstat)); +#endif + return (error); +} + +#if defined(SCTP_LOCAL_TRACE_BUF) +#if defined(__APPLE__) && !defined(__Userspace__) +static int +sctp_sysctl_handle_trace_log SYSCTL_HANDLER_ARGS +{ +#pragma unused(arg1, arg2, oidp) +#else +static int +sctp_sysctl_handle_trace_log(SYSCTL_HANDLER_ARGS) +{ +#endif + int error; + +#if defined(_WIN32) && !defined(__Userspace__) + error = SYSCTL_OUT(req, SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log)); +#else + error = SYSCTL_OUT(req, &SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log)); +#endif + return (error); +} + +#if defined(__APPLE__) && !defined(__Userspace__) +static int +sctp_sysctl_handle_trace_log_clear SYSCTL_HANDLER_ARGS +{ +#pragma unused(arg1, arg2, req, oidp) +#else +static int +sctp_sysctl_handle_trace_log_clear(SYSCTL_HANDLER_ARGS) +{ +#endif + int error = 0; +#if defined(_WIN32) && !defined(__Userspace__) + int value = 0; + + if (req->new_data == NULL) { + return (error); + } + error = SYSCTL_IN(req, &value, sizeof(int)); + if (error == 0 && value != 0 && SCTP_BASE_SYSCTL(sctp_log) != NULL) { + memset(SCTP_BASE_SYSCTL(sctp_log), 0, sizeof(struct sctp_log)); + } +#else + + memset(&SCTP_BASE_SYSCTL(sctp_log), 0, sizeof(struct sctp_log)); +#endif + return (error); +} +#endif + +#if (defined(__APPLE__) || defined(__FreeBSD__)) && !defined(__Userspace__) +#if defined(__FreeBSD__) +#define SCTP_UINT_SYSCTL(mib_name, var_name, prefix) \ + static int \ + sctp_sysctl_handle_##mib_name(SYSCTL_HANDLER_ARGS) \ + { \ + int error; \ + uint32_t new; \ + \ + new = SCTP_BASE_SYSCTL(var_name); \ + error = sysctl_handle_int(oidp, &new, 0, req); \ + if ((error == 0) && (req->newptr != NULL)) { \ + if ((new < prefix##_MIN) || \ + (new > prefix##_MAX)) { \ + error = EINVAL; \ + } else { \ + SCTP_BASE_SYSCTL(var_name) = new; \ + } \ + } \ + return (error); \ + } \ + SYSCTL_PROC(_net_inet_sctp, OID_AUTO, mib_name, \ + CTLFLAG_VNET|CTLTYPE_UINT|CTLFLAG_RW, NULL, 0, \ + sctp_sysctl_handle_##mib_name, "UI", prefix##_DESC); +#else +#define SCTP_UINT_SYSCTL(mib_name, var_name, prefix) \ + static int \ + sctp_sysctl_handle_##mib_name(struct sysctl_oid *oidp, \ + void *arg1 __attribute__((unused)), \ + int arg2 __attribute__((unused)), \ + struct sysctl_req *req) \ + { \ + int error; \ + uint32_t new; \ + \ + new = SCTP_BASE_SYSCTL(var_name); \ + error = sysctl_handle_int(oidp, &new, 0, req); \ + if ((error == 0) && (req->newptr != USER_ADDR_NULL)) { \ + if ((new < prefix##_MIN) || \ + (new > prefix##_MAX)) { \ + error = EINVAL; \ + } else { \ + SCTP_BASE_SYSCTL(var_name) = new; \ + } \ + } \ + return (error); \ + } \ + SYSCTL_PROC(_net_inet_sctp, OID_AUTO, mib_name, \ + CTLTYPE_INT | CTLFLAG_RW, NULL, 0, \ + sctp_sysctl_handle_##mib_name, "I", prefix##_DESC); +#define CTLTYPE_UINT CTLTYPE_INT +#define CTLFLAG_VNET 0 +#endif + +/* + * sysctl definitions + */ + +SCTP_UINT_SYSCTL(sendspace, sctp_sendspace, SCTPCTL_MAXDGRAM) +SCTP_UINT_SYSCTL(recvspace, sctp_recvspace, SCTPCTL_RECVSPACE) +SCTP_UINT_SYSCTL(auto_asconf, sctp_auto_asconf, SCTPCTL_AUTOASCONF) +SCTP_UINT_SYSCTL(ecn_enable, sctp_ecn_enable, SCTPCTL_ECN_ENABLE) +SCTP_UINT_SYSCTL(pr_enable, sctp_pr_enable, SCTPCTL_PR_ENABLE) +SYSCTL_PROC(_net_inet_sctp, OID_AUTO, auth_enable, CTLFLAG_VNET|CTLTYPE_UINT|CTLFLAG_RW, + NULL, 0, sctp_sysctl_handle_auth, "IU", SCTPCTL_AUTH_ENABLE_DESC); +SYSCTL_PROC(_net_inet_sctp, OID_AUTO, asconf_enable, CTLFLAG_VNET|CTLTYPE_UINT|CTLFLAG_RW, + NULL, 0, sctp_sysctl_handle_asconf, "IU", SCTPCTL_ASCONF_ENABLE_DESC); +SCTP_UINT_SYSCTL(reconfig_enable, sctp_reconfig_enable, SCTPCTL_RECONFIG_ENABLE) +SCTP_UINT_SYSCTL(nrsack_enable, sctp_nrsack_enable, SCTPCTL_NRSACK_ENABLE) +SCTP_UINT_SYSCTL(pktdrop_enable, sctp_pktdrop_enable, SCTPCTL_PKTDROP_ENABLE) +#if defined(__APPLE__) && !defined(__Userspace__) +SCTP_UINT_SYSCTL(loopback_nocsum, sctp_no_csum_on_loopback, SCTPCTL_LOOPBACK_NOCSUM) +#endif +SCTP_UINT_SYSCTL(peer_chkoh, sctp_peer_chunk_oh, SCTPCTL_PEER_CHKOH) +SCTP_UINT_SYSCTL(maxburst, sctp_max_burst_default, SCTPCTL_MAXBURST) +SCTP_UINT_SYSCTL(fr_maxburst, sctp_fr_max_burst_default, SCTPCTL_FRMAXBURST) +SCTP_UINT_SYSCTL(maxchunks, sctp_max_chunks_on_queue, SCTPCTL_MAXCHUNKS) +SCTP_UINT_SYSCTL(tcbhashsize, sctp_hashtblsize, SCTPCTL_TCBHASHSIZE) +SCTP_UINT_SYSCTL(pcbhashsize, sctp_pcbtblsize, SCTPCTL_PCBHASHSIZE) +SCTP_UINT_SYSCTL(min_split_point, sctp_min_split_point, SCTPCTL_MIN_SPLIT_POINT) +SCTP_UINT_SYSCTL(chunkscale, sctp_chunkscale, SCTPCTL_CHUNKSCALE) +SCTP_UINT_SYSCTL(delayed_sack_time, sctp_delayed_sack_time_default, SCTPCTL_DELAYED_SACK_TIME) +SCTP_UINT_SYSCTL(sack_freq, sctp_sack_freq_default, SCTPCTL_SACK_FREQ) +SCTP_UINT_SYSCTL(sys_resource, sctp_system_free_resc_limit, SCTPCTL_SYS_RESOURCE) +SCTP_UINT_SYSCTL(asoc_resource, sctp_asoc_free_resc_limit, SCTPCTL_ASOC_RESOURCE) +SCTP_UINT_SYSCTL(heartbeat_interval, sctp_heartbeat_interval_default, SCTPCTL_HEARTBEAT_INTERVAL) +SCTP_UINT_SYSCTL(pmtu_raise_time, sctp_pmtu_raise_time_default, SCTPCTL_PMTU_RAISE_TIME) +SCTP_UINT_SYSCTL(shutdown_guard_time, sctp_shutdown_guard_time_default, SCTPCTL_SHUTDOWN_GUARD_TIME) +SCTP_UINT_SYSCTL(secret_lifetime, sctp_secret_lifetime_default, SCTPCTL_SECRET_LIFETIME) +SCTP_UINT_SYSCTL(rto_max, sctp_rto_max_default, SCTPCTL_RTO_MAX) +SCTP_UINT_SYSCTL(rto_min, sctp_rto_min_default, SCTPCTL_RTO_MIN) +SCTP_UINT_SYSCTL(rto_initial, sctp_rto_initial_default, SCTPCTL_RTO_INITIAL) +SCTP_UINT_SYSCTL(init_rto_max, sctp_init_rto_max_default, SCTPCTL_INIT_RTO_MAX) +SCTP_UINT_SYSCTL(valid_cookie_life, sctp_valid_cookie_life_default, SCTPCTL_VALID_COOKIE_LIFE) +SCTP_UINT_SYSCTL(init_rtx_max, sctp_init_rtx_max_default, SCTPCTL_INIT_RTX_MAX) +SCTP_UINT_SYSCTL(assoc_rtx_max, sctp_assoc_rtx_max_default, SCTPCTL_ASSOC_RTX_MAX) +SCTP_UINT_SYSCTL(path_rtx_max, sctp_path_rtx_max_default, SCTPCTL_PATH_RTX_MAX) +SCTP_UINT_SYSCTL(path_pf_threshold, sctp_path_pf_threshold, SCTPCTL_PATH_PF_THRESHOLD) +SCTP_UINT_SYSCTL(add_more_on_output, sctp_add_more_threshold, SCTPCTL_ADD_MORE_ON_OUTPUT) +SCTP_UINT_SYSCTL(incoming_streams, sctp_nr_incoming_streams_default, SCTPCTL_INCOMING_STREAMS) +SCTP_UINT_SYSCTL(outgoing_streams, sctp_nr_outgoing_streams_default, SCTPCTL_OUTGOING_STREAMS) +SCTP_UINT_SYSCTL(cmt_on_off, sctp_cmt_on_off, SCTPCTL_CMT_ON_OFF) +SCTP_UINT_SYSCTL(cmt_use_dac, sctp_cmt_use_dac, SCTPCTL_CMT_USE_DAC) +SCTP_UINT_SYSCTL(cwnd_maxburst, sctp_use_cwnd_based_maxburst, SCTPCTL_CWND_MAXBURST) +SCTP_UINT_SYSCTL(nat_friendly, sctp_nat_friendly, SCTPCTL_NAT_FRIENDLY) +SCTP_UINT_SYSCTL(abc_l_var, sctp_L2_abc_variable, SCTPCTL_ABC_L_VAR) +SCTP_UINT_SYSCTL(max_chained_mbufs, sctp_mbuf_threshold_count, SCTPCTL_MAX_CHAINED_MBUFS) +SCTP_UINT_SYSCTL(do_sctp_drain, sctp_do_drain, SCTPCTL_DO_SCTP_DRAIN) +SCTP_UINT_SYSCTL(hb_max_burst, sctp_hb_maxburst, SCTPCTL_HB_MAX_BURST) +SCTP_UINT_SYSCTL(abort_at_limit, sctp_abort_if_one_2_one_hits_limit, SCTPCTL_ABORT_AT_LIMIT) +SCTP_UINT_SYSCTL(min_residual, sctp_min_residual, SCTPCTL_MIN_RESIDUAL) +SCTP_UINT_SYSCTL(max_retran_chunk, sctp_max_retran_chunk, SCTPCTL_MAX_RETRAN_CHUNK) +SCTP_UINT_SYSCTL(log_level, sctp_logging_level, SCTPCTL_LOGGING_LEVEL) +SCTP_UINT_SYSCTL(default_cc_module, sctp_default_cc_module, SCTPCTL_DEFAULT_CC_MODULE) +SCTP_UINT_SYSCTL(default_ss_module, sctp_default_ss_module, SCTPCTL_DEFAULT_SS_MODULE) +SCTP_UINT_SYSCTL(default_frag_interleave, sctp_default_frag_interleave, SCTPCTL_DEFAULT_FRAG_INTERLEAVE) +SCTP_UINT_SYSCTL(mobility_base, sctp_mobility_base, SCTPCTL_MOBILITY_BASE) +SCTP_UINT_SYSCTL(mobility_fasthandoff, sctp_mobility_fasthandoff, SCTPCTL_MOBILITY_FASTHANDOFF) +#if defined(SCTP_LOCAL_TRACE_BUF) +SYSCTL_PROC(_net_inet_sctp, OID_AUTO, log, CTLFLAG_VNET|CTLTYPE_STRUCT|CTLFLAG_RD, + NULL, 0, sctp_sysctl_handle_trace_log, "S,sctplog", "SCTP logging (struct sctp_log)"); +SYSCTL_PROC(_net_inet_sctp, OID_AUTO, clear_trace, CTLFLAG_VNET|CTLTYPE_UINT | CTLFLAG_RW, + NULL, 0, sctp_sysctl_handle_trace_log_clear, "IU", "Clear SCTP Logging buffer"); +#endif +SYSCTL_PROC(_net_inet_sctp, OID_AUTO, udp_tunneling_port, CTLFLAG_VNET|CTLTYPE_UINT|CTLFLAG_RW, + NULL, 0, sctp_sysctl_handle_udp_tunneling, "IU", SCTPCTL_UDP_TUNNELING_PORT_DESC); +SCTP_UINT_SYSCTL(enable_sack_immediately, sctp_enable_sack_immediately, SCTPCTL_SACK_IMMEDIATELY_ENABLE) +SCTP_UINT_SYSCTL(nat_friendly_init, sctp_inits_include_nat_friendly, SCTPCTL_NAT_FRIENDLY_INITS) +SCTP_UINT_SYSCTL(vtag_time_wait, sctp_vtag_time_wait, SCTPCTL_TIME_WAIT) +SCTP_UINT_SYSCTL(buffer_splitting, sctp_buffer_splitting, SCTPCTL_BUFFER_SPLITTING) +SCTP_UINT_SYSCTL(initial_cwnd, sctp_initial_cwnd, SCTPCTL_INITIAL_CWND) +SCTP_UINT_SYSCTL(rttvar_bw, sctp_rttvar_bw, SCTPCTL_RTTVAR_BW) +SCTP_UINT_SYSCTL(rttvar_rtt, sctp_rttvar_rtt, SCTPCTL_RTTVAR_RTT) +SCTP_UINT_SYSCTL(rttvar_eqret, sctp_rttvar_eqret, SCTPCTL_RTTVAR_EQRET) +SCTP_UINT_SYSCTL(rttvar_steady_step, sctp_steady_step, SCTPCTL_RTTVAR_STEADYS) +SCTP_UINT_SYSCTL(use_dcccecn, sctp_use_dccc_ecn, SCTPCTL_RTTVAR_DCCCECN) +SCTP_UINT_SYSCTL(blackhole, sctp_blackhole, SCTPCTL_BLACKHOLE) +SCTP_UINT_SYSCTL(sendall_limit, sctp_sendall_limit, SCTPCTL_SENDALL_LIMIT) +SCTP_UINT_SYSCTL(diag_info_code, sctp_diag_info_code, SCTPCTL_DIAG_INFO_CODE) +SCTP_UINT_SYSCTL(ootb_with_zero_cksum, sctp_ootb_with_zero_cksum, SCTPCTL_OOTB_WITH_ZERO_CKSUM) +#ifdef SCTP_DEBUG +SCTP_UINT_SYSCTL(debug, sctp_debug_on, SCTPCTL_DEBUG) +#endif +#if defined(__APPLE__) && !defined(__Userspace__) +SCTP_UINT_SYSCTL(main_timer, sctp_main_timer, SCTPCTL_MAIN_TIMER) +SYSCTL_PROC(_net_inet_sctp, OID_AUTO, ignore_vmware_interfaces, CTLTYPE_UINT|CTLFLAG_RW, + NULL, 0, sctp_sysctl_handle_vmware_interfaces, "IU", SCTPCTL_IGNORE_VMWARE_INTERFACES_DESC); +SCTP_UINT_SYSCTL(addr_watchdog_limit, sctp_addr_watchdog_limit, SCTPCTL_ADDR_WATCHDOG_LIMIT) +SCTP_UINT_SYSCTL(vtag_watchdog_limit, sctp_vtag_watchdog_limit, SCTPCTL_VTAG_WATCHDOG_LIMIT) +#endif +#if defined(__APPLE__) && !defined(__Userspace__) +SCTP_UINT_SYSCTL(output_unlocked, sctp_output_unlocked, SCTPCTL_OUTPUT_UNLOCKED) +#endif +SYSCTL_PROC(_net_inet_sctp, OID_AUTO, stats, CTLFLAG_VNET|CTLTYPE_STRUCT|CTLFLAG_RW, + NULL, 0, sctp_sysctl_handle_stats, "S,sctpstat", "SCTP statistics (struct sctp_stat)"); +SYSCTL_PROC(_net_inet_sctp, OID_AUTO, assoclist, CTLFLAG_VNET|CTLTYPE_OPAQUE|CTLFLAG_RD, + NULL, 0, sctp_sysctl_handle_assoclist, "S,xassoc", "List of active SCTP associations"); + +#elif defined(_WIN32) && !defined(__Userspace__) + +#define RANGECHK(var, min, max) \ + if ((var) < (min)) { (var) = (min); } \ + else if ((var) > (max)) { (var) = (max); } + +static int +sctp_sysctl_handle_int(SYSCTL_HANDLER_ARGS) +{ + int error; + + error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req); + if (error == 0) { + RANGECHK(SCTP_BASE_SYSCTL(sctp_sendspace), SCTPCTL_MAXDGRAM_MIN, SCTPCTL_MAXDGRAM_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_recvspace), SCTPCTL_RECVSPACE_MIN, SCTPCTL_RECVSPACE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_auto_asconf), SCTPCTL_AUTOASCONF_MIN, SCTPCTL_AUTOASCONF_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_auto_asconf), SCTPCTL_AUTOASCONF_MIN, SCTPCTL_AUTOASCONF_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_ecn_enable), SCTPCTL_ECN_ENABLE_MIN, SCTPCTL_ECN_ENABLE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_pr_enable), SCTPCTL_PR_ENABLE_MIN, SCTPCTL_PR_ENABLE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_reconfig_enable), SCTPCTL_RECONFIG_ENABLE_MIN, SCTPCTL_RECONFIG_ENABLE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_nrsack_enable), SCTPCTL_NRSACK_ENABLE_MIN, SCTPCTL_NRSACK_ENABLE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_pktdrop_enable), SCTPCTL_PKTDROP_ENABLE_MIN, SCTPCTL_PKTDROP_ENABLE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback), SCTPCTL_LOOPBACK_NOCSUM_MIN, SCTPCTL_LOOPBACK_NOCSUM_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_peer_chunk_oh), SCTPCTL_PEER_CHKOH_MIN, SCTPCTL_PEER_CHKOH_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_max_burst_default), SCTPCTL_MAXBURST_MIN, SCTPCTL_MAXBURST_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_fr_max_burst_default), SCTPCTL_FRMAXBURST_MIN, SCTPCTL_FRMAXBURST_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue), SCTPCTL_MAXCHUNKS_MIN, SCTPCTL_MAXCHUNKS_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_hashtblsize), SCTPCTL_TCBHASHSIZE_MIN, SCTPCTL_TCBHASHSIZE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_pcbtblsize), SCTPCTL_PCBHASHSIZE_MIN, SCTPCTL_PCBHASHSIZE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_min_split_point), SCTPCTL_MIN_SPLIT_POINT_MIN, SCTPCTL_MIN_SPLIT_POINT_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_chunkscale), SCTPCTL_CHUNKSCALE_MIN, SCTPCTL_CHUNKSCALE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default), SCTPCTL_DELAYED_SACK_TIME_MIN, SCTPCTL_DELAYED_SACK_TIME_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_sack_freq_default), SCTPCTL_SACK_FREQ_MIN, SCTPCTL_SACK_FREQ_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_system_free_resc_limit), SCTPCTL_SYS_RESOURCE_MIN, SCTPCTL_SYS_RESOURCE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_asoc_free_resc_limit), SCTPCTL_ASOC_RESOURCE_MIN, SCTPCTL_ASOC_RESOURCE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default), SCTPCTL_HEARTBEAT_INTERVAL_MIN, SCTPCTL_HEARTBEAT_INTERVAL_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default), SCTPCTL_PMTU_RAISE_TIME_MIN, SCTPCTL_PMTU_RAISE_TIME_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default), SCTPCTL_SHUTDOWN_GUARD_TIME_MIN, SCTPCTL_SHUTDOWN_GUARD_TIME_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default), SCTPCTL_SECRET_LIFETIME_MIN, SCTPCTL_SECRET_LIFETIME_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_rto_max_default), SCTPCTL_RTO_MAX_MIN, SCTPCTL_RTO_MAX_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_rto_min_default), SCTPCTL_RTO_MIN_MIN, SCTPCTL_RTO_MIN_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_rto_initial_default), SCTPCTL_RTO_INITIAL_MIN, SCTPCTL_RTO_INITIAL_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_init_rto_max_default), SCTPCTL_INIT_RTO_MAX_MIN, SCTPCTL_INIT_RTO_MAX_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default), SCTPCTL_VALID_COOKIE_LIFE_MIN, SCTPCTL_VALID_COOKIE_LIFE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_init_rtx_max_default), SCTPCTL_INIT_RTX_MAX_MIN, SCTPCTL_INIT_RTX_MAX_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default), SCTPCTL_ASSOC_RTX_MAX_MIN, SCTPCTL_ASSOC_RTX_MAX_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_path_rtx_max_default), SCTPCTL_PATH_RTX_MAX_MIN, SCTPCTL_PATH_RTX_MAX_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_path_pf_threshold), SCTPCTL_PATH_PF_THRESHOLD_MIN, SCTPCTL_PATH_PF_THRESHOLD_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_add_more_threshold), SCTPCTL_ADD_MORE_ON_OUTPUT_MIN, SCTPCTL_ADD_MORE_ON_OUTPUT_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default), SCTPCTL_INCOMING_STREAMS_MIN, SCTPCTL_INCOMING_STREAMS_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default), SCTPCTL_OUTGOING_STREAMS_MIN, SCTPCTL_OUTGOING_STREAMS_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_cmt_on_off), SCTPCTL_CMT_ON_OFF_MIN, SCTPCTL_CMT_ON_OFF_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_cmt_use_dac), SCTPCTL_CMT_USE_DAC_MIN, SCTPCTL_CMT_USE_DAC_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst), SCTPCTL_CWND_MAXBURST_MIN, SCTPCTL_CWND_MAXBURST_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_nat_friendly), SCTPCTL_NAT_FRIENDLY_MIN, SCTPCTL_NAT_FRIENDLY_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_L2_abc_variable), SCTPCTL_ABC_L_VAR_MIN, SCTPCTL_ABC_L_VAR_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_mbuf_threshold_count), SCTPCTL_MAX_CHAINED_MBUFS_MIN, SCTPCTL_MAX_CHAINED_MBUFS_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_do_drain), SCTPCTL_DO_SCTP_DRAIN_MIN, SCTPCTL_DO_SCTP_DRAIN_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_hb_maxburst), SCTPCTL_HB_MAX_BURST_MIN, SCTPCTL_HB_MAX_BURST_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_abort_if_one_2_one_hits_limit), SCTPCTL_ABORT_AT_LIMIT_MIN, SCTPCTL_ABORT_AT_LIMIT_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_min_residual), SCTPCTL_MIN_RESIDUAL_MIN, SCTPCTL_MIN_RESIDUAL_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_max_retran_chunk), SCTPCTL_MAX_RETRAN_CHUNK_MIN, SCTPCTL_MAX_RETRAN_CHUNK_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_logging_level), SCTPCTL_LOGGING_LEVEL_MIN, SCTPCTL_LOGGING_LEVEL_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_default_cc_module), SCTPCTL_DEFAULT_CC_MODULE_MIN, SCTPCTL_DEFAULT_CC_MODULE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_default_ss_module), SCTPCTL_DEFAULT_SS_MODULE_MIN, SCTPCTL_DEFAULT_SS_MODULE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_default_frag_interleave), SCTPCTL_DEFAULT_FRAG_INTERLEAVE_MIN, SCTPCTL_DEFAULT_FRAG_INTERLEAVE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_vtag_time_wait), SCTPCTL_TIME_WAIT_MIN, SCTPCTL_TIME_WAIT_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_buffer_splitting), SCTPCTL_BUFFER_SPLITTING_MIN, SCTPCTL_BUFFER_SPLITTING_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_initial_cwnd), SCTPCTL_INITIAL_CWND_MIN, SCTPCTL_INITIAL_CWND_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_rttvar_bw), SCTPCTL_RTTVAR_BW_MIN, SCTPCTL_RTTVAR_BW_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_rttvar_rtt), SCTPCTL_RTTVAR_RTT_MIN, SCTPCTL_RTTVAR_RTT_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_rttvar_eqret), SCTPCTL_RTTVAR_EQRET_MIN, SCTPCTL_RTTVAR_EQRET_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_steady_step), SCTPCTL_RTTVAR_STEADYS_MIN, SCTPCTL_RTTVAR_STEADYS_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_use_dccc_ecn), SCTPCTL_RTTVAR_DCCCECN_MIN, SCTPCTL_RTTVAR_DCCCECN_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_mobility_base), SCTPCTL_MOBILITY_BASE_MIN, SCTPCTL_MOBILITY_BASE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff), SCTPCTL_MOBILITY_FASTHANDOFF_MIN, SCTPCTL_MOBILITY_FASTHANDOFF_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_enable_sack_immediately), SCTPCTL_SACK_IMMEDIATELY_ENABLE_MIN, SCTPCTL_SACK_IMMEDIATELY_ENABLE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_inits_include_nat_friendly), SCTPCTL_NAT_FRIENDLY_INITS_MIN, SCTPCTL_NAT_FRIENDLY_INITS_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_blackhole), SCTPCTL_BLACKHOLE_MIN, SCTPCTL_BLACKHOLE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_sendall_limit), SCTPCTL_SENDALL_LIMIT_MIN, SCTPCTL_SENDALL_LIMIT_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_diag_info_code), SCTPCTL_DIAG_INFO_CODE_MIN, SCTPCTL_DIAG_INFO_CODE_MAX); + RANGECHK(SCTP_BASE_SYSCTL(sctp_ootb_with_zero_cksum), SCTPCTL_OOTB_WITH_ZERO_CKSUM_MIN, SCTPCTL_OOTB_WITH_ZERO_CKSUM_MAX); +#ifdef SCTP_DEBUG + RANGECHK(SCTP_BASE_SYSCTL(sctp_debug_on), SCTPCTL_DEBUG_MIN, SCTPCTL_DEBUG_MAX); +#endif + } + return (error); +} + +void +sysctl_setup_sctp(void) +{ + sysctl_add_oid(&sysctl_oid_top, "sendspace", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_sendspace), 0, sctp_sysctl_handle_int, + SCTPCTL_MAXDGRAM_DESC); + + sysctl_add_oid(&sysctl_oid_top, "recvspace", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_recvspace), 0, sctp_sysctl_handle_int, + SCTPCTL_RECVSPACE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "auto_asconf", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_auto_asconf), 0, sctp_sysctl_handle_int, + SCTPCTL_AUTOASCONF_DESC); + + sysctl_add_oid(&sysctl_oid_top, "ecn_enable", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_ecn_enable), 0, sctp_sysctl_handle_int, + SCTPCTL_ECN_ENABLE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "pr_enable", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_pr_enable), 0, sctp_sysctl_handle_int, + SCTPCTL_PR_ENABLE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "auth_enable", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_auth_enable), 0, sctp_sysctl_handle_auth, + SCTPCTL_AUTH_ENABLE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "asconf_enable", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_asconf_enable), 0, sctp_sysctl_handle_asconf, + SCTPCTL_ASCONF_ENABLE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "reconfig_enable", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_reconfig_enable), 0, sctp_sysctl_handle_int, + SCTPCTL_RECONFIG_ENABLE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "nrsack_enable", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_nrsack_enable), 0, sctp_sysctl_handle_int, + SCTPCTL_NRSACK_ENABLE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "pktdrop_enable", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_pktdrop_enable), 0, sctp_sysctl_handle_int, + SCTPCTL_PKTDROP_ENABLE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "loopback_nocsum", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback), 0, sctp_sysctl_handle_int, + SCTPCTL_LOOPBACK_NOCSUM_DESC); + + sysctl_add_oid(&sysctl_oid_top, "peer_chkoh", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_peer_chunk_oh), 0, sctp_sysctl_handle_int, + SCTPCTL_PEER_CHKOH_DESC); + + sysctl_add_oid(&sysctl_oid_top, "maxburst", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_max_burst_default), 0, sctp_sysctl_handle_int, + SCTPCTL_MAXBURST_DESC); + + sysctl_add_oid(&sysctl_oid_top, "fr_maxburst", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_fr_max_burst_default), 0, sctp_sysctl_handle_int, + SCTPCTL_FRMAXBURST_DESC); + + sysctl_add_oid(&sysctl_oid_top, "maxchunks", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue), 0, sctp_sysctl_handle_int, + SCTPCTL_MAXCHUNKS_DESC); + + sysctl_add_oid(&sysctl_oid_top, "tcbhashsize", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_hashtblsize), 0, sctp_sysctl_handle_int, + SCTPCTL_TCBHASHSIZE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "pcbhashsize", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_pcbtblsize), 0, sctp_sysctl_handle_int, + SCTPCTL_PCBHASHSIZE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "min_split_point", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_min_split_point), 0, sctp_sysctl_handle_int, + SCTPCTL_MIN_SPLIT_POINT_DESC); + + sysctl_add_oid(&sysctl_oid_top, "chunkscale", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_chunkscale), 0, sctp_sysctl_handle_int, + SCTPCTL_CHUNKSCALE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "delayed_sack_time", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default), 0, sctp_sysctl_handle_int, + SCTPCTL_DELAYED_SACK_TIME_DESC); + + sysctl_add_oid(&sysctl_oid_top, "sack_freq", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_sack_freq_default), 0, sctp_sysctl_handle_int, + SCTPCTL_SACK_FREQ_DESC); + + sysctl_add_oid(&sysctl_oid_top, "sys_resource", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_system_free_resc_limit), 0, sctp_sysctl_handle_int, + SCTPCTL_SYS_RESOURCE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "asoc_resource", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_asoc_free_resc_limit), 0, sctp_sysctl_handle_int, + SCTPCTL_ASOC_RESOURCE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "heartbeat_interval", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default), 0, sctp_sysctl_handle_int, + SCTPCTL_HEARTBEAT_INTERVAL_DESC); + + sysctl_add_oid(&sysctl_oid_top, "pmtu_raise_time", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default), 0, sctp_sysctl_handle_int, + SCTPCTL_PMTU_RAISE_TIME_DESC); + + sysctl_add_oid(&sysctl_oid_top, "shutdown_guard_time", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default), 0, sctp_sysctl_handle_int, + SCTPCTL_SHUTDOWN_GUARD_TIME_DESC); + + sysctl_add_oid(&sysctl_oid_top, "secret_lifetime", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_secret_lifetime_default), 0, sctp_sysctl_handle_int, + SCTPCTL_SECRET_LIFETIME_DESC); + + sysctl_add_oid(&sysctl_oid_top, "rto_max", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_rto_max_default), 0, sctp_sysctl_handle_int, + SCTPCTL_RTO_MAX_DESC); + + sysctl_add_oid(&sysctl_oid_top, "rto_min", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_rto_min_default), 0, sctp_sysctl_handle_int, + SCTPCTL_RTO_MIN_DESC); + + sysctl_add_oid(&sysctl_oid_top, "rto_initial", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_rto_initial_default), 0, sctp_sysctl_handle_int, + SCTPCTL_RTO_INITIAL_DESC); + + sysctl_add_oid(&sysctl_oid_top, "init_rto_max", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_init_rto_max_default), 0, sctp_sysctl_handle_int, + SCTPCTL_INIT_RTO_MAX_DESC); + + sysctl_add_oid(&sysctl_oid_top, "valid_cookie_life", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default), 0, sctp_sysctl_handle_int, + SCTPCTL_VALID_COOKIE_LIFE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "init_rtx_max", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_init_rtx_max_default), 0, sctp_sysctl_handle_int, + SCTPCTL_INIT_RTX_MAX_DESC); + + sysctl_add_oid(&sysctl_oid_top, "assoc_rtx_max", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default), 0, sctp_sysctl_handle_int, + SCTPCTL_ASSOC_RTX_MAX_DESC); + + sysctl_add_oid(&sysctl_oid_top, "path_rtx_max", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_path_rtx_max_default), 0, sctp_sysctl_handle_int, + SCTPCTL_PATH_RTX_MAX_DESC); + + sysctl_add_oid(&sysctl_oid_top, "path_pf_threshold", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_path_pf_threshold), 0, sctp_sysctl_handle_int, + SCTPCTL_PATH_PF_THRESHOLD_DESC); + + sysctl_add_oid(&sysctl_oid_top, "add_more_on_output", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_add_more_threshold), 0, sctp_sysctl_handle_int, + SCTPCTL_ADD_MORE_ON_OUTPUT_DESC); + + sysctl_add_oid(&sysctl_oid_top, "incoming_streams", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default), 0, sctp_sysctl_handle_int, + SCTPCTL_INCOMING_STREAMS_DESC); + + sysctl_add_oid(&sysctl_oid_top, "outgoing_streams", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default), 0, sctp_sysctl_handle_int, + SCTPCTL_OUTGOING_STREAMS_DESC); + + sysctl_add_oid(&sysctl_oid_top, "cmt_on_off", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_cmt_on_off), 0, sctp_sysctl_handle_int, + SCTPCTL_CMT_ON_OFF_DESC); + + sysctl_add_oid(&sysctl_oid_top, "cmt_use_dac", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_cmt_use_dac), 0, sctp_sysctl_handle_int, + SCTPCTL_CMT_USE_DAC_DESC); + + sysctl_add_oid(&sysctl_oid_top, "cwnd_maxburst", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst), 0, sctp_sysctl_handle_int, + SCTPCTL_CWND_MAXBURST_DESC); + + sysctl_add_oid(&sysctl_oid_top, "nat_friendly", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_nat_friendly), 0, sctp_sysctl_handle_int, + SCTPCTL_NAT_FRIENDLY_DESC); + + sysctl_add_oid(&sysctl_oid_top, "abc_l_var", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_L2_abc_variable), 0, sctp_sysctl_handle_int, + SCTPCTL_ABC_L_VAR_DESC); + + sysctl_add_oid(&sysctl_oid_top, "max_chained_mbufs", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_mbuf_threshold_count), 0, sctp_sysctl_handle_int, + SCTPCTL_MAX_CHAINED_MBUFS_DESC); + + sysctl_add_oid(&sysctl_oid_top, "do_sctp_drain", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_do_drain), 0, sctp_sysctl_handle_int, + SCTPCTL_DO_SCTP_DRAIN_DESC); + + sysctl_add_oid(&sysctl_oid_top, "hb_max_burst", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_hb_maxburst), 0, sctp_sysctl_handle_int, + SCTPCTL_HB_MAX_BURST_DESC); + + sysctl_add_oid(&sysctl_oid_top, "abort_at_limit", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_abort_if_one_2_one_hits_limit), 0, sctp_sysctl_handle_int, + SCTPCTL_ABORT_AT_LIMIT_DESC); + + sysctl_add_oid(&sysctl_oid_top, "min_residual", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_min_residual), 0, sctp_sysctl_handle_int, + SCTPCTL_MIN_RESIDUAL_DESC); + + sysctl_add_oid(&sysctl_oid_top, "max_retran_chunk", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_max_retran_chunk), 0, sctp_sysctl_handle_int, + SCTPCTL_MAX_RETRAN_CHUNK_DESC); + + sysctl_add_oid(&sysctl_oid_top, "log_level", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_logging_level), 0, sctp_sysctl_handle_int, + SCTPCTL_LOGGING_LEVEL_DESC); + + sysctl_add_oid(&sysctl_oid_top, "default_cc_module", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_default_cc_module), 0, sctp_sysctl_handle_int, + SCTPCTL_DEFAULT_CC_MODULE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "default_ss_module", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_default_ss_module), 0, sctp_sysctl_handle_int, + SCTPCTL_DEFAULT_SS_MODULE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "default_frag_interleave", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_default_frag_interleave), 0, sctp_sysctl_handle_int, + SCTPCTL_DEFAULT_FRAG_INTERLEAVE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "mobility_base", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_mobility_base), 0, sctp_sysctl_handle_int, + SCTPCTL_MOBILITY_BASE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "mobility_fasthandoff", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff), 0, sctp_sysctl_handle_int, + SCTPCTL_MOBILITY_FASTHANDOFF_DESC); + +#if defined(SCTP_LOCAL_TRACE_BUF) + sysctl_add_oid(&sysctl_oid_top, "sctp_log", CTLTYPE_STRUCT|CTLFLAG_RD, + SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log), NULL, + "SCTP logging (struct sctp_log)"); + + sysctl_add_oid(&sysctl_oid_top, "clear_trace", CTLTYPE_INT|CTLFLAG_WR, + NULL, 0, sctp_sysctl_handle_trace_log_clear, + "Clear SCTP Logging buffer"); +#endif + + sysctl_add_oid(&sysctl_oid_top, "udp_tunneling_port", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_udp_tunneling_port), 0, sctp_sysctl_handle_udp_tunneling, + SCTPCTL_UDP_TUNNELING_PORT_DESC); + + sysctl_add_oid(&sysctl_oid_top, "enable_sack_immediately", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_enable_sack_immediately), 0, sctp_sysctl_handle_int, + SCTPCTL_SACK_IMMEDIATELY_ENABLE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "nat_friendly_init", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_inits_include_nat_friendly), 0, sctp_sysctl_handle_int, + SCTPCTL_NAT_FRIENDLY_DESC); + + sysctl_add_oid(&sysctl_oid_top, "vtag_time_wait", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_vtag_time_wait), 0, sctp_sysctl_handle_int, + SCTPCTL_TIME_WAIT_DESC); + + sysctl_add_oid(&sysctl_oid_top, "buffer_splitting", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_buffer_splitting), 0, sctp_sysctl_handle_int, + SCTPCTL_BUFFER_SPLITTING_DESC); + + sysctl_add_oid(&sysctl_oid_top, "initial_cwnd", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_initial_cwnd), 0, sctp_sysctl_handle_int, + SCTPCTL_INITIAL_CWND_DESC); + + sysctl_add_oid(&sysctl_oid_top, "rttvar_bw", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_rttvar_bw), 0, sctp_sysctl_handle_int, + SCTPCTL_RTTVAR_BW_DESC); + + sysctl_add_oid(&sysctl_oid_top, "rttvar_rtt", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_rttvar_rtt), 0, sctp_sysctl_handle_int, + SCTPCTL_RTTVAR_RTT_DESC); + + sysctl_add_oid(&sysctl_oid_top, "rttvar_eqret", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_rttvar_eqret), 0, sctp_sysctl_handle_int, + SCTPCTL_RTTVAR_EQRET_DESC); + + sysctl_add_oid(&sysctl_oid_top, "rttvar_steady_step", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_steady_step), 0, sctp_sysctl_handle_int, + SCTPCTL_RTTVAR_STEADYS_DESC); + + sysctl_add_oid(&sysctl_oid_top, "use_dcccecn", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_use_dccc_ecn), 0, sctp_sysctl_handle_int, + SCTPCTL_RTTVAR_DCCCECN_DESC); + + sysctl_add_oid(&sysctl_oid_top, "blackhole", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_blackhole), 0, sctp_sysctl_handle_int, + SCTPCTL_BLACKHOLE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "sendall_limit", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_sendall_limit), 0, sctp_sysctl_handle_int, + SCTPCTL_SENDALL_LIMIT_DESC); + + sysctl_add_oid(&sysctl_oid_top, "diag_info_code", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_diag_info_code), 0, sctp_sysctl_handle_int, + SCTPCTL_DIAG_INFO_CODE_DESC); + + sysctl_add_oid(&sysctl_oid_top, "ootb_with_zero_cksum", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_ootb_with_zero_cksum), 0, sctp_sysctl_handle_int, + SCTPCTL_OOTB_WITH_ZERO_CKSUM_DESC); + +#ifdef SCTP_DEBUG + sysctl_add_oid(&sysctl_oid_top, "debug", CTLTYPE_INT|CTLFLAG_RW, + &SCTP_BASE_SYSCTL(sctp_debug_on), 0, sctp_sysctl_handle_int, + SCTPCTL_DEBUG_DESC); +#endif + + sysctl_add_oid(&sysctl_oid_top, "stats", CTLTYPE_STRUCT|CTLFLAG_RW, + &SCTP_BASE_STATS, sizeof(SCTP_BASE_STATS), NULL, + "SCTP statistics (struct sctp_stat)"); + + sysctl_add_oid(&sysctl_oid_top, "assoclist", CTLTYPE_STRUCT|CTLFLAG_RD, + NULL, 0, sctp_assoclist, + "List of active SCTP associations"); +} +#endif +#endif diff --git a/netwerk/sctp/src/netinet/sctp_sysctl.h b/netwerk/sctp/src/netinet/sctp_sysctl.h new file mode 100644 index 0000000000..9dc40c800d --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_sysctl.h @@ -0,0 +1,633 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_SYSCTL_H_ +#define _NETINET_SCTP_SYSCTL_H_ + +#include <netinet/sctp_os.h> +#include <netinet/sctp_constants.h> + +struct sctp_sysctl { + uint32_t sctp_sendspace; + uint32_t sctp_recvspace; + uint32_t sctp_auto_asconf; + uint32_t sctp_multiple_asconfs; + uint32_t sctp_ecn_enable; + uint32_t sctp_pr_enable; + uint32_t sctp_auth_enable; + uint32_t sctp_asconf_enable; + uint32_t sctp_reconfig_enable; + uint32_t sctp_nrsack_enable; + uint32_t sctp_pktdrop_enable; + uint32_t sctp_fr_max_burst_default; +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + uint32_t sctp_no_csum_on_loopback; +#endif + uint32_t sctp_peer_chunk_oh; + uint32_t sctp_max_burst_default; + uint32_t sctp_max_chunks_on_queue; + uint32_t sctp_hashtblsize; + uint32_t sctp_pcbtblsize; + uint32_t sctp_min_split_point; + uint32_t sctp_chunkscale; + uint32_t sctp_delayed_sack_time_default; + uint32_t sctp_sack_freq_default; + uint32_t sctp_system_free_resc_limit; + uint32_t sctp_asoc_free_resc_limit; + uint32_t sctp_heartbeat_interval_default; + uint32_t sctp_pmtu_raise_time_default; + uint32_t sctp_shutdown_guard_time_default; + uint32_t sctp_secret_lifetime_default; + uint32_t sctp_rto_max_default; + uint32_t sctp_rto_min_default; + uint32_t sctp_rto_initial_default; + uint32_t sctp_init_rto_max_default; + uint32_t sctp_valid_cookie_life_default; + uint32_t sctp_init_rtx_max_default; + uint32_t sctp_assoc_rtx_max_default; + uint32_t sctp_path_rtx_max_default; + uint32_t sctp_path_pf_threshold; + uint32_t sctp_add_more_threshold; + uint32_t sctp_nr_incoming_streams_default; + uint32_t sctp_nr_outgoing_streams_default; + uint32_t sctp_cmt_on_off; + uint32_t sctp_cmt_use_dac; + uint32_t sctp_use_cwnd_based_maxburst; + uint32_t sctp_nat_friendly; + uint32_t sctp_L2_abc_variable; + uint32_t sctp_mbuf_threshold_count; + uint32_t sctp_do_drain; + uint32_t sctp_hb_maxburst; + uint32_t sctp_abort_if_one_2_one_hits_limit; + uint32_t sctp_min_residual; + uint32_t sctp_max_retran_chunk; + uint32_t sctp_logging_level; + /* JRS - Variable for default congestion control module */ + uint32_t sctp_default_cc_module; + /* RS - Variable for default stream scheduling module */ + uint32_t sctp_default_ss_module; + uint32_t sctp_default_frag_interleave; + uint32_t sctp_mobility_base; + uint32_t sctp_mobility_fasthandoff; + uint32_t sctp_inits_include_nat_friendly; + uint32_t sctp_rttvar_bw; + uint32_t sctp_rttvar_rtt; + uint32_t sctp_rttvar_eqret; + uint32_t sctp_steady_step; + uint32_t sctp_use_dccc_ecn; +#if defined(SCTP_LOCAL_TRACE_BUF) +#if defined(_WIN32) && !defined(__Userspace__) + struct sctp_log *sctp_log; +#else + struct sctp_log sctp_log; +#endif +#endif + uint32_t sctp_udp_tunneling_port; + uint32_t sctp_enable_sack_immediately; + uint32_t sctp_vtag_time_wait; + uint32_t sctp_buffer_splitting; + uint32_t sctp_initial_cwnd; + uint32_t sctp_blackhole; + uint32_t sctp_sendall_limit; + uint32_t sctp_diag_info_code; + uint32_t sctp_ootb_with_zero_cksum; +#if defined(SCTP_DEBUG) + uint32_t sctp_debug_on; +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + uint32_t sctp_ignore_vmware_interfaces; + uint32_t sctp_main_timer; + uint32_t sctp_addr_watchdog_limit; + uint32_t sctp_vtag_watchdog_limit; +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + uint32_t sctp_output_unlocked; +#endif +}; + +/* + * limits for the sysctl variables + */ +/* maxdgram: Maximum outgoing SCTP buffer size */ +#define SCTPCTL_MAXDGRAM_DESC "Maximum outgoing SCTP buffer size" +#define SCTPCTL_MAXDGRAM_MIN 0 +#define SCTPCTL_MAXDGRAM_MAX 0xFFFFFFFF +#if defined(__Userspace__) +#define SCTPCTL_MAXDGRAM_DEFAULT SB_MAX +#else +#define SCTPCTL_MAXDGRAM_DEFAULT 262144 /* 256k */ +#endif + +/* recvspace: Maximum incoming SCTP buffer size */ +#define SCTPCTL_RECVSPACE_DESC "Maximum incoming SCTP buffer size" +#define SCTPCTL_RECVSPACE_MIN 0 +#define SCTPCTL_RECVSPACE_MAX 0xFFFFFFFF +#if defined(__Userspace__) +#define SCTPCTL_RECVSPACE_DEFAULT SB_RAW +#else +#define SCTPCTL_RECVSPACE_DEFAULT 262144 /* 256k */ +#endif + +/* autoasconf: Enable SCTP Auto-ASCONF */ +#define SCTPCTL_AUTOASCONF_DESC "Enable SCTP Auto-ASCONF" +#define SCTPCTL_AUTOASCONF_MIN 0 +#define SCTPCTL_AUTOASCONF_MAX 1 +#define SCTPCTL_AUTOASCONF_DEFAULT 1 + +/* autoasconf: Enable SCTP Auto-ASCONF */ +#define SCTPCTL_MULTIPLEASCONFS_DESC "Enable SCTP Multiple-ASCONFs" +#define SCTPCTL_MULTIPLEASCONFS_MIN 0 +#define SCTPCTL_MULTIPLEASCONFS_MAX 1 +#define SCTPCTL_MULTIPLEASCONFS_DEFAULT SCTP_DEFAULT_MULTIPLE_ASCONFS + +/* ecn_enable: Enable SCTP ECN */ +#define SCTPCTL_ECN_ENABLE_DESC "Enable SCTP ECN" +#define SCTPCTL_ECN_ENABLE_MIN 0 +#define SCTPCTL_ECN_ENABLE_MAX 1 +#define SCTPCTL_ECN_ENABLE_DEFAULT 1 + +/* pr_enable: Enable PR-SCTP */ +#define SCTPCTL_PR_ENABLE_DESC "Enable PR-SCTP" +#define SCTPCTL_PR_ENABLE_MIN 0 +#define SCTPCTL_PR_ENABLE_MAX 1 +#define SCTPCTL_PR_ENABLE_DEFAULT 1 + +/* auth_enable: Enable SCTP AUTH function */ +#define SCTPCTL_AUTH_ENABLE_DESC "Enable SCTP AUTH function" +#define SCTPCTL_AUTH_ENABLE_MIN 0 +#define SCTPCTL_AUTH_ENABLE_MAX 1 +#define SCTPCTL_AUTH_ENABLE_DEFAULT 1 + +/* asconf_enable: Enable SCTP ASCONF */ +#define SCTPCTL_ASCONF_ENABLE_DESC "Enable SCTP ASCONF" +#define SCTPCTL_ASCONF_ENABLE_MIN 0 +#define SCTPCTL_ASCONF_ENABLE_MAX 1 +#define SCTPCTL_ASCONF_ENABLE_DEFAULT 1 + +/* reconfig_enable: Enable SCTP RE-CONFIG */ +#define SCTPCTL_RECONFIG_ENABLE_DESC "Enable SCTP RE-CONFIG" +#define SCTPCTL_RECONFIG_ENABLE_MIN 0 +#define SCTPCTL_RECONFIG_ENABLE_MAX 1 +#define SCTPCTL_RECONFIG_ENABLE_DEFAULT 1 + +/* nrsack_enable: Enable NR_SACK */ +#define SCTPCTL_NRSACK_ENABLE_DESC "Enable SCTP NR-SACK" +#define SCTPCTL_NRSACK_ENABLE_MIN 0 +#define SCTPCTL_NRSACK_ENABLE_MAX 1 +#define SCTPCTL_NRSACK_ENABLE_DEFAULT 0 + +/* pktdrop_enable: Enable SCTP Packet Drop Reports */ +#define SCTPCTL_PKTDROP_ENABLE_DESC "Enable SCTP PKTDROP" +#define SCTPCTL_PKTDROP_ENABLE_MIN 0 +#define SCTPCTL_PKTDROP_ENABLE_MAX 1 +#define SCTPCTL_PKTDROP_ENABLE_DEFAULT 0 + +/* loopback_nocsum: Enable NO Csum on packets sent on loopback */ +#define SCTPCTL_LOOPBACK_NOCSUM_DESC "Enable NO Csum on packets sent on loopback" +#define SCTPCTL_LOOPBACK_NOCSUM_MIN 0 +#define SCTPCTL_LOOPBACK_NOCSUM_MAX 1 +#define SCTPCTL_LOOPBACK_NOCSUM_DEFAULT 1 + +/* peer_chkoh: Amount to debit peers rwnd per chunk sent */ +#define SCTPCTL_PEER_CHKOH_DESC "Amount to debit peers rwnd per chunk sent" +#define SCTPCTL_PEER_CHKOH_MIN 0 +#define SCTPCTL_PEER_CHKOH_MAX 0xFFFFFFFF +#define SCTPCTL_PEER_CHKOH_DEFAULT 256 + +/* maxburst: Default max burst for sctp endpoints */ +#define SCTPCTL_MAXBURST_DESC "Default max burst for sctp endpoints" +#define SCTPCTL_MAXBURST_MIN 0 +#define SCTPCTL_MAXBURST_MAX 0xFFFFFFFF +#define SCTPCTL_MAXBURST_DEFAULT SCTP_DEF_MAX_BURST + +/* fr_maxburst: Default max burst for sctp endpoints when fast retransmitting */ +#define SCTPCTL_FRMAXBURST_DESC "Default max burst for SCTP endpoints when fast retransmitting" +#define SCTPCTL_FRMAXBURST_MIN 0 +#define SCTPCTL_FRMAXBURST_MAX 0xFFFFFFFF +#define SCTPCTL_FRMAXBURST_DEFAULT SCTP_DEF_FRMAX_BURST + +/* maxchunks: Default max chunks on queue per asoc */ +#define SCTPCTL_MAXCHUNKS_DESC "Default max chunks on queue per asoc" +#define SCTPCTL_MAXCHUNKS_MIN 0 +#define SCTPCTL_MAXCHUNKS_MAX 0xFFFFFFFF +#define SCTPCTL_MAXCHUNKS_DEFAULT SCTP_ASOC_MAX_CHUNKS_ON_QUEUE + +/* tcbhashsize: Tunable for Hash table sizes */ +#define SCTPCTL_TCBHASHSIZE_DESC "Tunable for TCB hash table sizes" +#define SCTPCTL_TCBHASHSIZE_MIN 1 +#define SCTPCTL_TCBHASHSIZE_MAX 0xFFFFFFFF +#define SCTPCTL_TCBHASHSIZE_DEFAULT SCTP_TCBHASHSIZE + +/* pcbhashsize: Tunable for PCB Hash table sizes */ +#define SCTPCTL_PCBHASHSIZE_DESC "Tunable for PCB hash table sizes" +#define SCTPCTL_PCBHASHSIZE_MIN 1 +#define SCTPCTL_PCBHASHSIZE_MAX 0xFFFFFFFF +#define SCTPCTL_PCBHASHSIZE_DEFAULT SCTP_PCBHASHSIZE + +/* min_split_point: Minimum size when splitting a chunk */ +#define SCTPCTL_MIN_SPLIT_POINT_DESC "Minimum size when splitting a chunk" +#define SCTPCTL_MIN_SPLIT_POINT_MIN 0 +#define SCTPCTL_MIN_SPLIT_POINT_MAX 0xFFFFFFFF +#define SCTPCTL_MIN_SPLIT_POINT_DEFAULT SCTP_DEFAULT_SPLIT_POINT_MIN + +/* chunkscale: Tunable for Scaling of number of chunks and messages */ +#define SCTPCTL_CHUNKSCALE_DESC "Tunable for scaling of number of chunks and messages" +#define SCTPCTL_CHUNKSCALE_MIN 1 +#define SCTPCTL_CHUNKSCALE_MAX 0xFFFFFFFF +#define SCTPCTL_CHUNKSCALE_DEFAULT SCTP_CHUNKQUEUE_SCALE + +/* delayed_sack_time: Default delayed SACK timer in ms */ +#define SCTPCTL_DELAYED_SACK_TIME_DESC "Default delayed SACK timer in ms" +#define SCTPCTL_DELAYED_SACK_TIME_MIN 0 +#define SCTPCTL_DELAYED_SACK_TIME_MAX 0xFFFFFFFF +#define SCTPCTL_DELAYED_SACK_TIME_DEFAULT SCTP_RECV_MSEC + +/* sack_freq: Default SACK frequency */ +#define SCTPCTL_SACK_FREQ_DESC "Default SACK frequency" +#define SCTPCTL_SACK_FREQ_MIN 0 +#define SCTPCTL_SACK_FREQ_MAX 0xFFFFFFFF +#define SCTPCTL_SACK_FREQ_DEFAULT SCTP_DEFAULT_SACK_FREQ + +/* sys_resource: Max number of cached resources in the system */ +#define SCTPCTL_SYS_RESOURCE_DESC "Max number of cached resources in the system" +#define SCTPCTL_SYS_RESOURCE_MIN 0 +#define SCTPCTL_SYS_RESOURCE_MAX 0xFFFFFFFF +#define SCTPCTL_SYS_RESOURCE_DEFAULT SCTP_DEF_SYSTEM_RESC_LIMIT + +/* asoc_resource: Max number of cached resources in an asoc */ +#define SCTPCTL_ASOC_RESOURCE_DESC "Max number of cached resources in an asoc" +#define SCTPCTL_ASOC_RESOURCE_MIN 0 +#define SCTPCTL_ASOC_RESOURCE_MAX 0xFFFFFFFF +#define SCTPCTL_ASOC_RESOURCE_DEFAULT SCTP_DEF_ASOC_RESC_LIMIT + +/* heartbeat_interval: Default heartbeat interval in ms */ +#define SCTPCTL_HEARTBEAT_INTERVAL_DESC "Default heartbeat interval in ms" +#define SCTPCTL_HEARTBEAT_INTERVAL_MIN 0 +#define SCTPCTL_HEARTBEAT_INTERVAL_MAX 0xFFFFFFFF +#define SCTPCTL_HEARTBEAT_INTERVAL_DEFAULT SCTP_HB_DEFAULT_MSEC + +/* pmtu_raise_time: Default PMTU raise timer in seconds */ +#define SCTPCTL_PMTU_RAISE_TIME_DESC "Default PMTU raise timer in seconds" +#define SCTPCTL_PMTU_RAISE_TIME_MIN 0 +#define SCTPCTL_PMTU_RAISE_TIME_MAX 0xFFFFFFFF +#define SCTPCTL_PMTU_RAISE_TIME_DEFAULT SCTP_DEF_PMTU_RAISE_SEC + +/* shutdown_guard_time: Default shutdown guard timer in seconds */ +#define SCTPCTL_SHUTDOWN_GUARD_TIME_DESC "Shutdown guard timer in seconds (0 means 5 times RTO.Max)" +#define SCTPCTL_SHUTDOWN_GUARD_TIME_MIN 0 +#define SCTPCTL_SHUTDOWN_GUARD_TIME_MAX 0xFFFFFFFF +#define SCTPCTL_SHUTDOWN_GUARD_TIME_DEFAULT 0 + +/* secret_lifetime: Default secret lifetime in seconds */ +#define SCTPCTL_SECRET_LIFETIME_DESC "Default secret lifetime in seconds" +#define SCTPCTL_SECRET_LIFETIME_MIN 0 +#define SCTPCTL_SECRET_LIFETIME_MAX 0xFFFFFFFF +#define SCTPCTL_SECRET_LIFETIME_DEFAULT SCTP_DEFAULT_SECRET_LIFE_SEC + +/* rto_max: Default maximum retransmission timeout in ms */ +#define SCTPCTL_RTO_MAX_DESC "Default maximum retransmission timeout in ms" +#define SCTPCTL_RTO_MAX_MIN 0 +#define SCTPCTL_RTO_MAX_MAX 0xFFFFFFFF +#define SCTPCTL_RTO_MAX_DEFAULT SCTP_RTO_UPPER_BOUND + +/* rto_min: Default minimum retransmission timeout in ms */ +#define SCTPCTL_RTO_MIN_DESC "Default minimum retransmission timeout in ms" +#define SCTPCTL_RTO_MIN_MIN 0 +#define SCTPCTL_RTO_MIN_MAX 0xFFFFFFFF +#define SCTPCTL_RTO_MIN_DEFAULT SCTP_RTO_LOWER_BOUND + +/* rto_initial: Default initial retransmission timeout in ms */ +#define SCTPCTL_RTO_INITIAL_DESC "Default initial retransmission timeout in ms" +#define SCTPCTL_RTO_INITIAL_MIN 0 +#define SCTPCTL_RTO_INITIAL_MAX 0xFFFFFFFF +#define SCTPCTL_RTO_INITIAL_DEFAULT SCTP_RTO_INITIAL + +/* init_rto_max: Default maximum retransmission timeout during association setup in ms */ +#define SCTPCTL_INIT_RTO_MAX_DESC "Default maximum retransmission timeout during association setup in ms" +#define SCTPCTL_INIT_RTO_MAX_MIN 0 +#define SCTPCTL_INIT_RTO_MAX_MAX 0xFFFFFFFF +#define SCTPCTL_INIT_RTO_MAX_DEFAULT SCTP_RTO_UPPER_BOUND + +/* valid_cookie_life: Default cookie lifetime in ms */ +#define SCTPCTL_VALID_COOKIE_LIFE_DESC "Default cookie lifetime in ms" +#define SCTPCTL_VALID_COOKIE_LIFE_MIN SCTP_MIN_COOKIE_LIFE +#define SCTPCTL_VALID_COOKIE_LIFE_MAX SCTP_MAX_COOKIE_LIFE +#define SCTPCTL_VALID_COOKIE_LIFE_DEFAULT SCTP_DEFAULT_COOKIE_LIFE + +/* init_rtx_max: Default maximum number of retransmission for INIT chunks */ +#define SCTPCTL_INIT_RTX_MAX_DESC "Default maximum number of retransmissions for INIT chunks" +#define SCTPCTL_INIT_RTX_MAX_MIN 0 +#define SCTPCTL_INIT_RTX_MAX_MAX 0xFFFFFFFF +#define SCTPCTL_INIT_RTX_MAX_DEFAULT SCTP_DEF_MAX_INIT + +/* assoc_rtx_max: Default maximum number of retransmissions per association */ +#define SCTPCTL_ASSOC_RTX_MAX_DESC "Default maximum number of retransmissions per association" +#define SCTPCTL_ASSOC_RTX_MAX_MIN 0 +#define SCTPCTL_ASSOC_RTX_MAX_MAX 0xFFFFFFFF +#define SCTPCTL_ASSOC_RTX_MAX_DEFAULT SCTP_DEF_MAX_SEND + +/* path_rtx_max: Default maximum of retransmissions per path */ +#define SCTPCTL_PATH_RTX_MAX_DESC "Default maximum of retransmissions per path" +#define SCTPCTL_PATH_RTX_MAX_MIN 0 +#define SCTPCTL_PATH_RTX_MAX_MAX 0xFFFFFFFF +#define SCTPCTL_PATH_RTX_MAX_DEFAULT SCTP_DEF_MAX_PATH_RTX + +/* path_pf_threshold: threshold for considering the path potentially failed */ +#define SCTPCTL_PATH_PF_THRESHOLD_DESC "Default potentially failed threshold" +#define SCTPCTL_PATH_PF_THRESHOLD_MIN 0 +#define SCTPCTL_PATH_PF_THRESHOLD_MAX 0xFFFF +#define SCTPCTL_PATH_PF_THRESHOLD_DEFAULT SCTPCTL_PATH_PF_THRESHOLD_MAX + +/* add_more_on_output: When space-wise is it worthwhile to try to add more to a socket send buffer */ +#define SCTPCTL_ADD_MORE_ON_OUTPUT_DESC "When space-wise is it worthwhile to try to add more to a socket send buffer" +#define SCTPCTL_ADD_MORE_ON_OUTPUT_MIN 0 +#define SCTPCTL_ADD_MORE_ON_OUTPUT_MAX 0xFFFFFFFF +#define SCTPCTL_ADD_MORE_ON_OUTPUT_DEFAULT SCTP_DEFAULT_ADD_MORE + +/* incoming_streams: Default number of incoming streams */ +#define SCTPCTL_INCOMING_STREAMS_DESC "Default number of incoming streams" +#define SCTPCTL_INCOMING_STREAMS_MIN 1 +#define SCTPCTL_INCOMING_STREAMS_MAX 65535 +#define SCTPCTL_INCOMING_STREAMS_DEFAULT SCTP_ISTREAM_INITIAL + +/* outgoing_streams: Default number of outgoing streams */ +#define SCTPCTL_OUTGOING_STREAMS_DESC "Default number of outgoing streams" +#define SCTPCTL_OUTGOING_STREAMS_MIN 1 +#define SCTPCTL_OUTGOING_STREAMS_MAX 65535 +#define SCTPCTL_OUTGOING_STREAMS_DEFAULT SCTP_OSTREAM_INITIAL + +/* cmt_on_off: CMT on/off flag */ +#define SCTPCTL_CMT_ON_OFF_DESC "CMT settings" +#define SCTPCTL_CMT_ON_OFF_MIN SCTP_CMT_OFF +#define SCTPCTL_CMT_ON_OFF_MAX SCTP_CMT_MAX +#define SCTPCTL_CMT_ON_OFF_DEFAULT SCTP_CMT_OFF + +/* cmt_use_dac: CMT DAC on/off flag */ +#define SCTPCTL_CMT_USE_DAC_DESC "CMT DAC on/off flag" +#define SCTPCTL_CMT_USE_DAC_MIN 0 +#define SCTPCTL_CMT_USE_DAC_MAX 1 +#define SCTPCTL_CMT_USE_DAC_DEFAULT 0 + +/* cwnd_maxburst: Use a CWND adjusting to implement maxburst */ +#define SCTPCTL_CWND_MAXBURST_DESC "Adjust congestion control window to limit maximum burst when sending" +#define SCTPCTL_CWND_MAXBURST_MIN 0 +#define SCTPCTL_CWND_MAXBURST_MAX 1 +#define SCTPCTL_CWND_MAXBURST_DEFAULT 1 + +/* nat_friendly: SCTP NAT friendly operation */ +#define SCTPCTL_NAT_FRIENDLY_DESC "SCTP NAT friendly operation" +#define SCTPCTL_NAT_FRIENDLY_MIN 0 +#define SCTPCTL_NAT_FRIENDLY_MAX 1 +#define SCTPCTL_NAT_FRIENDLY_DEFAULT 1 + +/* abc_l_var: SCTP ABC max increase per SACK (L) */ +#define SCTPCTL_ABC_L_VAR_DESC "SCTP ABC max increase per SACK (L)" +#define SCTPCTL_ABC_L_VAR_MIN 0 +#define SCTPCTL_ABC_L_VAR_MAX 0xFFFFFFFF +#define SCTPCTL_ABC_L_VAR_DEFAULT 2 + +/* max_chained_mbufs: Default max number of small mbufs on a chain */ +#define SCTPCTL_MAX_CHAINED_MBUFS_DESC "Default max number of small mbufs on a chain" +#define SCTPCTL_MAX_CHAINED_MBUFS_MIN 0 +#define SCTPCTL_MAX_CHAINED_MBUFS_MAX 0xFFFFFFFF +#define SCTPCTL_MAX_CHAINED_MBUFS_DEFAULT SCTP_DEFAULT_MBUFS_IN_CHAIN + +/* do_sctp_drain: Should SCTP respond to the drain calls */ +#define SCTPCTL_DO_SCTP_DRAIN_DESC "Should SCTP respond to the drain calls" +#define SCTPCTL_DO_SCTP_DRAIN_MIN 0 +#define SCTPCTL_DO_SCTP_DRAIN_MAX 1 +#define SCTPCTL_DO_SCTP_DRAIN_DEFAULT 1 + +/* hb_max_burst: Confirmation Heartbeat max burst? */ +#define SCTPCTL_HB_MAX_BURST_DESC "Confirmation Heartbeat max burst" +#define SCTPCTL_HB_MAX_BURST_MIN 1 +#define SCTPCTL_HB_MAX_BURST_MAX 0xFFFFFFFF +#define SCTPCTL_HB_MAX_BURST_DEFAULT SCTP_DEF_HBMAX_BURST + +/* abort_at_limit: When one-2-one hits qlimit abort */ +#define SCTPCTL_ABORT_AT_LIMIT_DESC "Abort when one-to-one hits qlimit" +#define SCTPCTL_ABORT_AT_LIMIT_MIN 0 +#define SCTPCTL_ABORT_AT_LIMIT_MAX 1 +#define SCTPCTL_ABORT_AT_LIMIT_DEFAULT 0 + +/* min_residual: min residual in a data fragment leftover */ +#define SCTPCTL_MIN_RESIDUAL_DESC "Minimum residual data chunk in second part of split" +#define SCTPCTL_MIN_RESIDUAL_MIN 20 +#define SCTPCTL_MIN_RESIDUAL_MAX 65535 +#define SCTPCTL_MIN_RESIDUAL_DEFAULT 1452 + +/* max_retran_chunk: max chunk retransmissions */ +#define SCTPCTL_MAX_RETRAN_CHUNK_DESC "Maximum times an unlucky chunk can be retransmitted before assoc abort" +#define SCTPCTL_MAX_RETRAN_CHUNK_MIN 0 +#define SCTPCTL_MAX_RETRAN_CHUNK_MAX 65535 +#define SCTPCTL_MAX_RETRAN_CHUNK_DEFAULT 30 + +/* sctp_logging: This gives us logging when the options are enabled */ +#define SCTPCTL_LOGGING_LEVEL_DESC "Ltrace/KTR trace logging level" +#define SCTPCTL_LOGGING_LEVEL_MIN 0 +#define SCTPCTL_LOGGING_LEVEL_MAX 0xffffffff +#define SCTPCTL_LOGGING_LEVEL_DEFAULT 0 + +/* JRS - default congestion control module sysctl */ +#define SCTPCTL_DEFAULT_CC_MODULE_DESC "Default congestion control module" +#define SCTPCTL_DEFAULT_CC_MODULE_MIN 0 +#define SCTPCTL_DEFAULT_CC_MODULE_MAX 2 +#define SCTPCTL_DEFAULT_CC_MODULE_DEFAULT 0 + +/* RS - default stream scheduling module sysctl */ +#define SCTPCTL_DEFAULT_SS_MODULE_DESC "Default stream scheduling module" +#define SCTPCTL_DEFAULT_SS_MODULE_MIN 0 +#define SCTPCTL_DEFAULT_SS_MODULE_MAX 5 +#define SCTPCTL_DEFAULT_SS_MODULE_DEFAULT 0 + +/* RRS - default fragment interleave */ +#define SCTPCTL_DEFAULT_FRAG_INTERLEAVE_DESC "Default fragment interleave level" +#define SCTPCTL_DEFAULT_FRAG_INTERLEAVE_MIN 0 +#define SCTPCTL_DEFAULT_FRAG_INTERLEAVE_MAX 2 +#define SCTPCTL_DEFAULT_FRAG_INTERLEAVE_DEFAULT 1 + +/* mobility_base: Enable SCTP mobility support */ +#define SCTPCTL_MOBILITY_BASE_DESC "Enable SCTP base mobility" +#define SCTPCTL_MOBILITY_BASE_MIN 0 +#define SCTPCTL_MOBILITY_BASE_MAX 1 +#define SCTPCTL_MOBILITY_BASE_DEFAULT 0 + +/* mobility_fasthandoff: Enable SCTP fast handoff support */ +#define SCTPCTL_MOBILITY_FASTHANDOFF_DESC "Enable SCTP fast handoff" +#define SCTPCTL_MOBILITY_FASTHANDOFF_MIN 0 +#define SCTPCTL_MOBILITY_FASTHANDOFF_MAX 1 +#define SCTPCTL_MOBILITY_FASTHANDOFF_DEFAULT 0 + +/* Enable SCTP/UDP tunneling port */ +#define SCTPCTL_UDP_TUNNELING_PORT_DESC "Set the SCTP/UDP tunneling port" +#define SCTPCTL_UDP_TUNNELING_PORT_MIN 0 +#define SCTPCTL_UDP_TUNNELING_PORT_MAX 65535 +#if defined(__FreeBSD__) && !defined(__Userspace__) +#define SCTPCTL_UDP_TUNNELING_PORT_DEFAULT 0 +#else +#define SCTPCTL_UDP_TUNNELING_PORT_DEFAULT SCTP_OVER_UDP_TUNNELING_PORT +#endif + +/* Enable sending of the SACK-IMMEDIATELY bit */ +#define SCTPCTL_SACK_IMMEDIATELY_ENABLE_DESC "Enable sending of the SACK-IMMEDIATELY-bit" +#define SCTPCTL_SACK_IMMEDIATELY_ENABLE_MIN 0 +#define SCTPCTL_SACK_IMMEDIATELY_ENABLE_MAX 1 +#define SCTPCTL_SACK_IMMEDIATELY_ENABLE_DEFAULT SCTPCTL_SACK_IMMEDIATELY_ENABLE_MAX + +/* Enable sending of the NAT-FRIENDLY message */ +#define SCTPCTL_NAT_FRIENDLY_INITS_DESC "Enable sending of the nat-friendly SCTP option on INITs" +#define SCTPCTL_NAT_FRIENDLY_INITS_MIN 0 +#define SCTPCTL_NAT_FRIENDLY_INITS_MAX 1 +#define SCTPCTL_NAT_FRIENDLY_INITS_DEFAULT SCTPCTL_NAT_FRIENDLY_INITS_MIN + +/* Vtag time wait in seconds */ +#define SCTPCTL_TIME_WAIT_DESC "Vtag time wait time in seconds, 0 disables it" +#define SCTPCTL_TIME_WAIT_MIN 0 +#define SCTPCTL_TIME_WAIT_MAX 0xffffffff +#define SCTPCTL_TIME_WAIT_DEFAULT SCTP_TIME_WAIT + +/* Enable Send/Receive buffer splitting */ +#define SCTPCTL_BUFFER_SPLITTING_DESC "Enable send/receive buffer splitting" +#define SCTPCTL_BUFFER_SPLITTING_MIN 0 +#define SCTPCTL_BUFFER_SPLITTING_MAX 0x3 +#define SCTPCTL_BUFFER_SPLITTING_DEFAULT SCTPCTL_BUFFER_SPLITTING_MIN + +/* Initial congestion window in MTUs */ +#define SCTPCTL_INITIAL_CWND_DESC "Defines the initial congestion window size in MTUs" +#define SCTPCTL_INITIAL_CWND_MIN 0 +#define SCTPCTL_INITIAL_CWND_MAX 0xffffffff +#define SCTPCTL_INITIAL_CWND_DEFAULT 3 + +/* rttvar smooth avg for bw calc */ +#define SCTPCTL_RTTVAR_BW_DESC "Shift amount DCCC uses for bw smoothing on rtt calc" +#define SCTPCTL_RTTVAR_BW_MIN 0 +#define SCTPCTL_RTTVAR_BW_MAX 32 +#define SCTPCTL_RTTVAR_BW_DEFAULT 4 + +/* rttvar smooth avg for bw calc */ +#define SCTPCTL_RTTVAR_RTT_DESC "Shift amount DCCC uses for rtt smoothing on rtt calc" +#define SCTPCTL_RTTVAR_RTT_MIN 0 +#define SCTPCTL_RTTVAR_RTT_MAX 32 +#define SCTPCTL_RTTVAR_RTT_DEFAULT 5 + +#define SCTPCTL_RTTVAR_EQRET_DESC "Whether DCCC increases cwnd when the rtt and bw are unchanged" +#define SCTPCTL_RTTVAR_EQRET_MIN 0 +#define SCTPCTL_RTTVAR_EQRET_MAX 1 +#define SCTPCTL_RTTVAR_EQRET_DEFAULT 0 + +#define SCTPCTL_RTTVAR_STEADYS_DESC "Number of identical bw measurements DCCC takes to try step down of cwnd" +#define SCTPCTL_RTTVAR_STEADYS_MIN 0 +#define SCTPCTL_RTTVAR_STEADYS_MAX 0xFFFF +#define SCTPCTL_RTTVAR_STEADYS_DEFAULT 20 /* 0 means disable feature */ + +#define SCTPCTL_RTTVAR_DCCCECN_DESC "Enable ECN for DCCC." +#define SCTPCTL_RTTVAR_DCCCECN_MIN 0 +#define SCTPCTL_RTTVAR_DCCCECN_MAX 1 +#define SCTPCTL_RTTVAR_DCCCECN_DEFAULT 1 /* 0 means disable feature */ + +#define SCTPCTL_BLACKHOLE_DESC "Enable SCTP blackholing, see blackhole(4) for more details" +#define SCTPCTL_BLACKHOLE_MIN 0 +#define SCTPCTL_BLACKHOLE_MAX 2 +#define SCTPCTL_BLACKHOLE_DEFAULT SCTPCTL_BLACKHOLE_MIN + +/* sendall_limit: Maximum message with SCTP_SENDALL */ +#define SCTPCTL_SENDALL_LIMIT_DESC "Maximum size of a message send with SCTP_SENDALL" +#define SCTPCTL_SENDALL_LIMIT_MIN 0 +#define SCTPCTL_SENDALL_LIMIT_MAX 0xFFFFFFFF +#define SCTPCTL_SENDALL_LIMIT_DEFAULT 1432 + +#define SCTPCTL_DIAG_INFO_CODE_DESC "Diagnostic information error cause code" +#define SCTPCTL_DIAG_INFO_CODE_MIN 0 +#define SCTPCTL_DIAG_INFO_CODE_MAX 65535 +#define SCTPCTL_DIAG_INFO_CODE_DEFAULT 0 + +#define SCTPCTL_OOTB_WITH_ZERO_CKSUM_DESC "Accept OOTB packets with zero checksum" +#define SCTPCTL_OOTB_WITH_ZERO_CKSUM_MIN 0 +#define SCTPCTL_OOTB_WITH_ZERO_CKSUM_MAX 1 +#define SCTPCTL_OOTB_WITH_ZERO_CKSUM_DEFAULT 0 + +#if defined(SCTP_DEBUG) +/* debug: Configure debug output */ +#define SCTPCTL_DEBUG_DESC "Configure debug output" +#define SCTPCTL_DEBUG_MIN 0 +#define SCTPCTL_DEBUG_MAX 0xFFFFFFFF +#define SCTPCTL_DEBUG_DEFAULT 0 +#endif + +#if defined(__APPLE__) && !defined(__Userspace__) +#define SCTPCTL_MAIN_TIMER_DESC "Main timer interval in ms" +#define SCTPCTL_MAIN_TIMER_MIN 1 +#define SCTPCTL_MAIN_TIMER_MAX 0xFFFFFFFF +#define SCTPCTL_MAIN_TIMER_DEFAULT 10 + +#define SCTPCTL_IGNORE_VMWARE_INTERFACES_DESC "Ignore VMware Interfaces" +#define SCTPCTL_IGNORE_VMWARE_INTERFACES_MIN 0 +#define SCTPCTL_IGNORE_VMWARE_INTERFACES_MAX 1 +#define SCTPCTL_IGNORE_VMWARE_INTERFACES_DEFAULT SCTPCTL_IGNORE_VMWARE_INTERFACES_MAX + +#define SCTPCTL_OUTPUT_UNLOCKED_DESC "Unlock socket when sending packets down to IP" +#define SCTPCTL_OUTPUT_UNLOCKED_MIN 0 +#define SCTPCTL_OUTPUT_UNLOCKED_MAX 1 +#define SCTPCTL_OUTPUT_UNLOCKED_DEFAULT SCTPCTL_OUTPUT_UNLOCKED_MIN + +#define SCTPCTL_ADDR_WATCHDOG_LIMIT_DESC "Address watchdog limit" +#define SCTPCTL_ADDR_WATCHDOG_LIMIT_MIN 0 +#define SCTPCTL_ADDR_WATCHDOG_LIMIT_MAX 0xFFFFFFFF +#define SCTPCTL_ADDR_WATCHDOG_LIMIT_DEFAULT SCTPCTL_ADDR_WATCHDOG_LIMIT_MIN + +#define SCTPCTL_VTAG_WATCHDOG_LIMIT_DESC "VTag watchdog limit" +#define SCTPCTL_VTAG_WATCHDOG_LIMIT_MIN 0 +#define SCTPCTL_VTAG_WATCHDOG_LIMIT_MAX 0xFFFFFFFF +#define SCTPCTL_VTAG_WATCHDOG_LIMIT_DEFAULT SCTPCTL_VTAG_WATCHDOG_LIMIT_MIN + +#endif +#if defined(_KERNEL) || defined(__Userspace__) +#if defined(__FreeBSD__) || defined(__APPLE__) || defined(__Userspace__) +#if defined(SYSCTL_DECL) +SYSCTL_DECL(_net_inet_sctp); +#endif +#endif + +void sctp_init_sysctls(void); +#if defined(_WIN32) && !defined(__Userspace__) +void sctp_finish_sysctls(void); +#endif + +#endif /* _KERNEL */ +#endif /* __sctp_sysctl_h__ */ diff --git a/netwerk/sctp/src/netinet/sctp_timer.c b/netwerk/sctp/src/netinet/sctp_timer.c new file mode 100644 index 0000000000..aa44dad688 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_timer.c @@ -0,0 +1,1641 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#define _IP_VHL +#include <netinet/sctp_os.h> +#include <netinet/sctp_pcb.h> +#ifdef INET6 +#if defined(__FreeBSD__) && defined(__Userspace__) +#include <netinet6/sctp6_var.h> +#endif +#endif +#include <netinet/sctp_var.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_timer.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_header.h> +#include <netinet/sctp_indata.h> +#include <netinet/sctp_asconf.h> +#include <netinet/sctp_input.h> +#include <netinet/sctp.h> +#include <netinet/sctp_uio.h> +#if defined(INET) || defined(INET6) +#if !(defined(_WIN32) && defined(__Userspace__)) +#include <netinet/udp.h> +#endif +#endif + +void +sctp_audit_retranmission_queue(struct sctp_association *asoc) +{ + struct sctp_tmit_chunk *chk; + + SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n", + asoc->sent_queue_retran_cnt, + asoc->sent_queue_cnt); + asoc->sent_queue_retran_cnt = 0; + asoc->sent_queue_cnt = 0; + TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { + if (chk->sent == SCTP_DATAGRAM_RESEND) { + sctp_ucount_incr(asoc->sent_queue_retran_cnt); + } + asoc->sent_queue_cnt++; + } + TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { + if (chk->sent == SCTP_DATAGRAM_RESEND) { + sctp_ucount_incr(asoc->sent_queue_retran_cnt); + } + } + TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) { + if (chk->sent == SCTP_DATAGRAM_RESEND) { + sctp_ucount_incr(asoc->sent_queue_retran_cnt); + } + } + SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n", + asoc->sent_queue_retran_cnt, + asoc->sent_queue_cnt); +} + +static int +sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct sctp_nets *net, uint16_t threshold) +{ + if (net) { + net->error_count++; + SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n", + (void *)net, net->error_count, + net->failure_threshold); + if (net->error_count > net->failure_threshold) { + /* We had a threshold failure */ + if (net->dest_state & SCTP_ADDR_REACHABLE) { + net->dest_state &= ~SCTP_ADDR_REACHABLE; + net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY; + net->dest_state &= ~SCTP_ADDR_PF; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, + stcb, 0, + (void *)net, SCTP_SO_NOT_LOCKED); + } + } else if ((net->pf_threshold < net->failure_threshold) && + (net->error_count > net->pf_threshold)) { + if ((net->dest_state & SCTP_ADDR_PF) == 0) { + net->dest_state |= SCTP_ADDR_PF; + net->last_active = sctp_get_tick_count(); + sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED); + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, + inp, stcb, net, + SCTP_FROM_SCTP_TIMER + SCTP_LOC_1); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net); + } + } + } + if (stcb == NULL) + return (0); + + if (net) { + if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { + sctp_misc_ints(SCTP_THRESHOLD_INCR, + stcb->asoc.overall_error_count, + (stcb->asoc.overall_error_count+1), + SCTP_FROM_SCTP_TIMER, + __LINE__); + } + stcb->asoc.overall_error_count++; + } + } else { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) { + sctp_misc_ints(SCTP_THRESHOLD_INCR, + stcb->asoc.overall_error_count, + (stcb->asoc.overall_error_count+1), + SCTP_FROM_SCTP_TIMER, + __LINE__); + } + stcb->asoc.overall_error_count++; + } + SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n", + (void *)&stcb->asoc, stcb->asoc.overall_error_count, + (uint32_t)threshold, + ((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state)); + /* + * We specifically do not do >= to give the assoc one more change + * before we fail it. + */ + if (stcb->asoc.overall_error_count > threshold) { + /* Abort notification sends a ULP notify */ + struct mbuf *op_err; + + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + "Association error counter exceeded"); + inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_2; + sctp_abort_an_association(inp, stcb, op_err, true, SCTP_SO_NOT_LOCKED); + return (1); + } + return (0); +} + +/* + * sctp_find_alternate_net() returns a non-NULL pointer as long as there + * exists nets, which are not being deleted. + */ +struct sctp_nets * +sctp_find_alternate_net(struct sctp_tcb *stcb, + struct sctp_nets *net, + int mode) +{ + /* Find and return an alternate network if possible */ + struct sctp_nets *alt, *mnet, *min_errors_net = NULL , *max_cwnd_net = NULL; + bool looped; + /* JRS 5/14/07 - Initialize min_errors to an impossible value. */ + int min_errors = -1; + uint32_t max_cwnd = 0; + + if (stcb->asoc.numnets == 1) { + /* No selection can be made. */ + return (TAILQ_FIRST(&stcb->asoc.nets)); + } + /* + * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate net algorithm. + * This algorithm chooses the active destination (not in PF state) with the largest + * cwnd value. If all destinations are in PF state, unreachable, or unconfirmed, choose + * the destination that is in PF state with the lowest error count. In case of a tie, + * choose the destination that was most recently active. + */ + if (mode == 2) { + TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) { + /* JRS 5/14/07 - If the destination is unreachable or unconfirmed, skip it. */ + if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) || + (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) { + continue; + } + /* + * JRS 5/14/07 - If the destination is reachable but in PF state, compare + * the error count of the destination to the minimum error count seen thus far. + * Store the destination with the lower error count. If the error counts are + * equal, store the destination that was most recently active. + */ + if (mnet->dest_state & SCTP_ADDR_PF) { + /* + * JRS 5/14/07 - If the destination under consideration is the current + * destination, work as if the error count is one higher. The + * actual error count will not be incremented until later in the + * t3 handler. + */ + if (mnet == net) { + if (min_errors == -1) { + min_errors = mnet->error_count + 1; + min_errors_net = mnet; + } else if (mnet->error_count + 1 < min_errors) { + min_errors = mnet->error_count + 1; + min_errors_net = mnet; + } else if (mnet->error_count + 1 == min_errors + && mnet->last_active > min_errors_net->last_active) { + min_errors_net = mnet; + min_errors = mnet->error_count + 1; + } + continue; + } else { + if (min_errors == -1) { + min_errors = mnet->error_count; + min_errors_net = mnet; + } else if (mnet->error_count < min_errors) { + min_errors = mnet->error_count; + min_errors_net = mnet; + } else if (mnet->error_count == min_errors + && mnet->last_active > min_errors_net->last_active) { + min_errors_net = mnet; + min_errors = mnet->error_count; + } + continue; + } + } + /* + * JRS 5/14/07 - If the destination is reachable and not in PF state, compare the + * cwnd of the destination to the highest cwnd seen thus far. Store the + * destination with the higher cwnd value. If the cwnd values are equal, + * randomly choose one of the two destinations. + */ + if (max_cwnd < mnet->cwnd) { + max_cwnd_net = mnet; + max_cwnd = mnet->cwnd; + } else if (max_cwnd == mnet->cwnd) { + uint32_t rndval; + uint8_t this_random; + + if (stcb->asoc.hb_random_idx > 3) { + rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep); + memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values)); + this_random = stcb->asoc.hb_random_values[0]; + stcb->asoc.hb_random_idx++; + stcb->asoc.hb_ect_randombit = 0; + } else { + this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx]; + stcb->asoc.hb_random_idx++; + stcb->asoc.hb_ect_randombit = 0; + } + if (this_random % 2 == 1) { + max_cwnd_net = mnet; + max_cwnd = mnet->cwnd; /* Useless? */ + } + } + } + if (max_cwnd_net == NULL) { + if (min_errors_net == NULL) { + return (net); + } + return (min_errors_net); + } else { + return (max_cwnd_net); + } + } /* JRS 5/14/07 - If mode is set to 1, use the CMT policy for choosing an alternate net. */ + else if (mode == 1) { + TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) { + if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) || + (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) { + /* + * will skip ones that are not-reachable or + * unconfirmed + */ + continue; + } + if (max_cwnd < mnet->cwnd) { + max_cwnd_net = mnet; + max_cwnd = mnet->cwnd; + } else if (max_cwnd == mnet->cwnd) { + uint32_t rndval; + uint8_t this_random; + + if (stcb->asoc.hb_random_idx > 3) { + rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep); + memcpy(stcb->asoc.hb_random_values, &rndval, + sizeof(stcb->asoc.hb_random_values)); + this_random = stcb->asoc.hb_random_values[0]; + stcb->asoc.hb_random_idx = 0; + stcb->asoc.hb_ect_randombit = 0; + } else { + this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx]; + stcb->asoc.hb_random_idx++; + stcb->asoc.hb_ect_randombit = 0; + } + if (this_random % 2) { + max_cwnd_net = mnet; + max_cwnd = mnet->cwnd; + } + } + } + if (max_cwnd_net) { + return (max_cwnd_net); + } + } + /* Look for an alternate net, which is active. */ + if ((net != NULL) && ((net->dest_state & SCTP_ADDR_BEING_DELETED) == 0)) { + alt = TAILQ_NEXT(net, sctp_next); + } else { + alt = TAILQ_FIRST(&stcb->asoc.nets); + } + looped = false; + for (;;) { + if (alt == NULL) { + if (!looped) { + alt = TAILQ_FIRST(&stcb->asoc.nets); + looped = true; + } + /* Definitely out of candidates. */ + if (alt == NULL) { + break; + } + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (alt->ro.ro_nh == NULL) { +#else + if (alt->ro.ro_rt == NULL) { +#endif + if (alt->ro._s_addr) { + sctp_free_ifa(alt->ro._s_addr); + alt->ro._s_addr = NULL; + } + alt->src_addr_selected = 0; + } + if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) && +#if defined(__FreeBSD__) && !defined(__Userspace__) + (alt->ro.ro_nh != NULL) && +#else + (alt->ro.ro_rt != NULL) && +#endif + ((alt->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) && + (alt != net)) { + /* Found an alternate net, which is reachable. */ + break; + } + alt = TAILQ_NEXT(alt, sctp_next); + } + + if (alt == NULL) { + /* + * In case no active alternate net has been found, look for + * an alternate net, which is confirmed. + */ + if ((net != NULL) && ((net->dest_state & SCTP_ADDR_BEING_DELETED) == 0)) { + alt = TAILQ_NEXT(net, sctp_next); + } else { + alt = TAILQ_FIRST(&stcb->asoc.nets); + } + looped = false; + for (;;) { + if (alt == NULL) { + if (!looped) { + alt = TAILQ_FIRST(&stcb->asoc.nets); + looped = true; + } + /* Definitely out of candidates. */ + if (alt == NULL) { + break; + } + } + if (((alt->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) && + (alt != net)) { + /* Found an alternate net, which is confirmed. */ + break; + } + alt = TAILQ_NEXT(alt, sctp_next); + } + } + if (alt == NULL) { + /* + * In case no confirmed alternate net has been found, just + * return net, if it is not being deleted. In the other case + * just return the first net. + */ + if ((net != NULL) && ((net->dest_state & SCTP_ADDR_BEING_DELETED) == 0)) { + alt = net; + } + if (alt == NULL) { + alt = TAILQ_FIRST(&stcb->asoc.nets); + } + } + return (alt); +} + +static void +sctp_backoff_on_timeout(struct sctp_tcb *stcb, + struct sctp_nets *net, + int win_probe, + int num_marked, int num_abandoned) +{ + if (net->RTO == 0) { + if (net->RTO_measured) { + net->RTO = stcb->asoc.minrto; + } else { + net->RTO = stcb->asoc.initial_rto; + } + } + net->RTO <<= 1; + if (net->RTO > stcb->asoc.maxrto) { + net->RTO = stcb->asoc.maxrto; + } + if ((win_probe == 0) && (num_marked || num_abandoned)) { + /* We don't apply penalty to window probe scenarios */ + /* JRS - Use the congestion control given in the CC module */ + stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net); + } +} + +#ifndef INVARIANTS +static void +sctp_recover_sent_list(struct sctp_tcb *stcb) +{ + struct sctp_tmit_chunk *chk, *nchk; + struct sctp_association *asoc; + + asoc = &stcb->asoc; + TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) { + if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.tsn)) { + SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n", + (void *)chk, chk->rec.data.tsn, asoc->last_acked_seq); + if (chk->sent != SCTP_DATAGRAM_NR_ACKED) { + if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) { + asoc->strmout[chk->rec.data.sid].chunks_on_queues--; + } + } + if ((asoc->strmout[chk->rec.data.sid].chunks_on_queues == 0) && + (asoc->strmout[chk->rec.data.sid].state == SCTP_STREAM_RESET_PENDING) && + TAILQ_EMPTY(&asoc->strmout[chk->rec.data.sid].outqueue)) { + asoc->trigger_reset = 1; + } + TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next); + if (PR_SCTP_ENABLED(chk->flags)) { + if (asoc->pr_sctp_cnt != 0) + asoc->pr_sctp_cnt--; + } + if (chk->data) { + /*sa_ignore NO_NULL_CHK*/ + sctp_free_bufspace(stcb, asoc, chk, 1); + sctp_m_freem(chk->data); + chk->data = NULL; + if (asoc->prsctp_supported && PR_SCTP_BUF_ENABLED(chk->flags)) { + asoc->sent_queue_cnt_removeable--; + } + } + asoc->sent_queue_cnt--; + sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); + } + } + SCTP_PRINTF("after recover order is as follows\n"); + TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { + SCTP_PRINTF("chk:%p TSN:%x\n", (void *)chk, chk->rec.data.tsn); + } +} +#endif + +static int +sctp_mark_all_for_resend(struct sctp_tcb *stcb, + struct sctp_nets *net, + struct sctp_nets *alt, + int window_probe, + int *num_marked, + int *num_abandoned) +{ + + /* + * Mark all chunks (well not all) that were sent to *net for + * retransmission. Move them to alt for there destination as well... + * We only mark chunks that have been outstanding long enough to + * have received feed-back. + */ + struct sctp_tmit_chunk *chk, *nchk; + struct sctp_nets *lnets; + struct timeval now, min_wait, tv; + int cur_rto; + int cnt_abandoned; + int audit_tf, num_mk, fir; + unsigned int cnt_mk; + uint32_t orig_flight, orig_tf; + uint32_t tsnlast, tsnfirst; +#ifndef INVARIANTS + int recovery_cnt = 0; +#endif + + /* none in flight now */ + audit_tf = 0; + fir = 0; + /* + * figure out how long a data chunk must be pending before we can + * mark it .. + */ + (void)SCTP_GETTIME_TIMEVAL(&now); + /* get cur rto in micro-seconds */ + cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv; + cur_rto *= 1000; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(cur_rto, + stcb->asoc.peers_rwnd, + window_probe, + SCTP_FR_T3_MARK_TIME); + sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT); + sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT); + } + tv.tv_sec = cur_rto / 1000000; + tv.tv_usec = cur_rto % 1000000; +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + timersub(&now, &tv, &min_wait); +#else + min_wait = now; + timevalsub(&min_wait, &tv); +#endif + if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) { + /* + * if we hit here, we don't have enough seconds on the clock + * to account for the RTO. We just let the lower seconds be + * the bounds and don't worry about it. This may mean we + * will mark a lot more than we should. + */ + min_wait.tv_sec = min_wait.tv_usec = 0; + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(cur_rto, (uint32_t)now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME); + sctp_log_fr(0, (uint32_t)min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME); + } + /* + * Our rwnd will be incorrect here since we are not adding back the + * cnt * mbuf but we will fix that down below. + */ + orig_flight = net->flight_size; + orig_tf = stcb->asoc.total_flight; + + net->fast_retran_ip = 0; + /* Now on to each chunk */ + cnt_abandoned = 0; + num_mk = cnt_mk = 0; + tsnfirst = tsnlast = 0; +#ifndef INVARIANTS + start_again: +#endif + TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) { + if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.tsn)) { + /* Strange case our list got out of order? */ + SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x\n", + (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.tsn); +#ifdef INVARIANTS + panic("last acked >= chk on sent-Q"); +#else + recovery_cnt++; + SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt); + sctp_recover_sent_list(stcb); + if (recovery_cnt < 10) { + goto start_again; + } else { + SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt); + } +#endif + } + if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) { + /* + * found one to mark: If it is less than + * DATAGRAM_ACKED it MUST not be a skipped or marked + * TSN but instead one that is either already set + * for retransmission OR one that needs + * retransmission. + */ + + /* validate its been outstanding long enough */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(chk->rec.data.tsn, + (uint32_t)chk->sent_rcv_time.tv_sec, + chk->sent_rcv_time.tv_usec, + SCTP_FR_T3_MARK_TIME); + } + if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) { + /* + * we have reached a chunk that was sent + * some seconds past our min.. forget it we + * will find no more to send. + */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(0, + (uint32_t)chk->sent_rcv_time.tv_sec, + chk->sent_rcv_time.tv_usec, + SCTP_FR_T3_STOPPED); + } + continue; + } else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) && + (window_probe == 0)) { + /* + * we must look at the micro seconds to + * know. + */ + if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) { + /* + * ok it was sent after our boundary + * time. + */ + continue; + } + } + if (stcb->asoc.prsctp_supported && PR_SCTP_TTL_ENABLED(chk->flags)) { + /* Is it expired? */ +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + if (timercmp(&now, &chk->rec.data.timetodrop, >)) { +#else + if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) { +#endif + /* Yes so drop it */ + if (chk->data) { + (void)sctp_release_pr_sctp_chunk(stcb, + chk, + 1, + SCTP_SO_NOT_LOCKED); + cnt_abandoned++; + } + continue; + } + } + if (stcb->asoc.prsctp_supported && PR_SCTP_RTX_ENABLED(chk->flags)) { + /* Has it been retransmitted tv_sec times? */ + if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) { + if (chk->data) { + (void)sctp_release_pr_sctp_chunk(stcb, + chk, + 1, + SCTP_SO_NOT_LOCKED); + cnt_abandoned++; + } + continue; + } + } + if (chk->sent < SCTP_DATAGRAM_RESEND) { + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + num_mk++; + if (fir == 0) { + fir = 1; + tsnfirst = chk->rec.data.tsn; + } + tsnlast = chk->rec.data.tsn; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(chk->rec.data.tsn, chk->snd_count, + 0, SCTP_FR_T3_MARKED); + } + + if (chk->rec.data.chunk_was_revoked) { + /* deflate the cwnd */ + chk->whoTo->cwnd -= chk->book_size; + chk->rec.data.chunk_was_revoked = 0; + } + net->marked_retrans++; + stcb->asoc.marked_retrans++; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO, + chk->whoTo->flight_size, + chk->book_size, + (uint32_t)(uintptr_t)chk->whoTo, + chk->rec.data.tsn); + } + sctp_flight_size_decrease(chk); + sctp_total_flight_decrease(stcb, chk); + stcb->asoc.peers_rwnd += chk->send_size; + stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh); + } + chk->sent = SCTP_DATAGRAM_RESEND; + chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; + SCTP_STAT_INCR(sctps_markedretrans); + + /* reset the TSN for striking and other FR stuff */ + chk->rec.data.doing_fast_retransmit = 0; + /* Clear any time so NO RTT is being done */ + + if (chk->do_rtt) { + if (chk->whoTo->rto_needed == 0) { + chk->whoTo->rto_needed = 1; + } + } + chk->do_rtt = 0; + if (alt != net) { + sctp_free_remote_addr(chk->whoTo); + chk->no_fr_allowed = 1; + chk->whoTo = alt; + atomic_add_int(&alt->ref_count, 1); + } else { + chk->no_fr_allowed = 0; + if (TAILQ_EMPTY(&stcb->asoc.send_queue)) { + chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq; + } else { + chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.tsn; + } + } + /* CMT: Do not allow FRs on retransmitted TSNs. + */ + if (stcb->asoc.sctp_cmt_on_off > 0) { + chk->no_fr_allowed = 1; + } +#ifdef THIS_SHOULD_NOT_BE_DONE + } else if (chk->sent == SCTP_DATAGRAM_ACKED) { + /* remember highest acked one */ + could_be_sent = chk; +#endif + } + if (chk->sent == SCTP_DATAGRAM_RESEND) { + cnt_mk++; + } + } + if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) { + /* we did not subtract the same things? */ + audit_tf = 1; + } + + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT); + } +#ifdef SCTP_DEBUG + if (num_mk) { + SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n", + tsnlast); + SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%u\n", + num_mk, + stcb->asoc.peers_rwnd); + } +#endif + *num_marked = num_mk; + *num_abandoned = cnt_abandoned; + /* Now check for a ECN Echo that may be stranded And + * include the cnt_mk'd to have all resends in the + * control queue. + */ + TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) { + if (chk->sent == SCTP_DATAGRAM_RESEND) { + cnt_mk++; + } + if ((chk->whoTo == net) && + (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) { + sctp_free_remote_addr(chk->whoTo); + chk->whoTo = alt; + if (chk->sent != SCTP_DATAGRAM_RESEND) { + chk->sent = SCTP_DATAGRAM_RESEND; + chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + cnt_mk++; + } + atomic_add_int(&alt->ref_count, 1); + } + } +#ifdef THIS_SHOULD_NOT_BE_DONE + if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) { + /* fix it so we retransmit the highest acked anyway */ + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + cnt_mk++; + could_be_sent->sent = SCTP_DATAGRAM_RESEND; + } +#endif + if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) { +#ifdef INVARIANTS + SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n", + cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk); +#endif +#ifndef SCTP_AUDITING_ENABLED + stcb->asoc.sent_queue_retran_cnt = cnt_mk; +#endif + } + if (audit_tf) { + SCTPDBG(SCTP_DEBUG_TIMER4, + "Audit total flight due to negative value net:%p\n", + (void *)net); + stcb->asoc.total_flight = 0; + stcb->asoc.total_flight_count = 0; + /* Clear all networks flight size */ + TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) { + lnets->flight_size = 0; + SCTPDBG(SCTP_DEBUG_TIMER4, + "Net:%p c-f cwnd:%d ssthresh:%d\n", + (void *)lnets, lnets->cwnd, lnets->ssthresh); + } + TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { + if (chk->sent < SCTP_DATAGRAM_RESEND) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_UP, + chk->whoTo->flight_size, + chk->book_size, + (uint32_t)(uintptr_t)chk->whoTo, + chk->rec.data.tsn); + } + + sctp_flight_size_increase(chk); + sctp_total_flight_increase(stcb, chk); + } + } + } + /* We return 1 if we only have a window probe outstanding */ + return (0); +} + +int +sctp_t3rxt_timer(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + struct sctp_nets *alt; + int win_probe, num_mk, num_abandoned; + + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) { + sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT); + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { + struct sctp_nets *lnet; + + TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { + if (net == lnet) { + sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3); + } else { + sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3); + } + } + } + /* Find an alternate and mark those for retransmission */ + if ((stcb->asoc.peers_rwnd == 0) && + (stcb->asoc.total_flight < net->mtu)) { + SCTP_STAT_INCR(sctps_timowindowprobe); + win_probe = 1; + } else { + win_probe = 0; + } + + if (win_probe == 0) { + /* We don't do normal threshold management on window probes */ + if (sctp_threshold_management(inp, stcb, net, + stcb->asoc.max_send_times)) { + /* Association was destroyed */ + return (1); + } else { + if (net != stcb->asoc.primary_destination) { + /* send a immediate HB if our RTO is stale */ + struct timeval now; + uint32_t ms_goneby; + + (void)SCTP_GETTIME_TIMEVAL(&now); + if (net->last_sent_time.tv_sec) { + ms_goneby = (uint32_t)(now.tv_sec - net->last_sent_time.tv_sec) * 1000; + } else { + ms_goneby = 0; + } + if ((net->dest_state & SCTP_ADDR_PF) == 0) { + if ((ms_goneby > net->RTO) || (net->RTO == 0)) { + /* + * no recent feed back in an RTO or + * more, request a RTT update + */ + sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED); + } + } + } + } + } else { + /* + * For a window probe we don't penalize the net's but only + * the association. This may fail it if SACKs are not coming + * back. If sack's are coming with rwnd locked at 0, we will + * continue to hold things waiting for rwnd to raise + */ + if (sctp_threshold_management(inp, stcb, NULL, + stcb->asoc.max_send_times)) { + /* Association was destroyed */ + return (1); + } + } + if (stcb->asoc.sctp_cmt_on_off > 0) { + if (net->pf_threshold < net->failure_threshold) { + alt = sctp_find_alternate_net(stcb, net, 2); + } else { + /* + * CMT: Using RTX_SSTHRESH policy for CMT. + * If CMT is being used, then pick dest with + * largest ssthresh for any retransmission. + */ + alt = sctp_find_alternate_net(stcb, net, 1); + /* + * CUCv2: If a different dest is picked for + * the retransmission, then new + * (rtx-)pseudo_cumack needs to be tracked + * for orig dest. Let CUCv2 track new (rtx-) + * pseudo-cumack always. + */ + net->find_pseudo_cumack = 1; + net->find_rtx_pseudo_cumack = 1; + } + } else { + alt = sctp_find_alternate_net(stcb, net, 0); + } + + num_mk = 0; + num_abandoned = 0; + (void)sctp_mark_all_for_resend(stcb, net, alt, win_probe, + &num_mk, &num_abandoned); + /* FR Loss recovery just ended with the T3. */ + stcb->asoc.fast_retran_loss_recovery = 0; + + /* CMT FR loss recovery ended with the T3 */ + net->fast_retran_loss_recovery = 0; + if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) && + (net->flight_size == 0)) { + (*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins)(stcb, net); + } + + /* + * setup the sat loss recovery that prevents satellite cwnd advance. + */ + stcb->asoc.sat_t3_loss_recovery = 1; + stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq; + + /* Backoff the timer and cwnd */ + sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned); + if (((net->dest_state & SCTP_ADDR_REACHABLE) == 0) || + (net->dest_state & SCTP_ADDR_PF)) { + /* Move all pending over too */ + sctp_move_chunks_from_net(stcb, net); + + /* Get the address that failed, to + * force a new src address selection and + * a route allocation. + */ + if (net->ro._s_addr != NULL) { + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + } + net->src_addr_selected = 0; + + /* Force a route allocation too */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + RO_NHFREE(&net->ro); +#else + if (net->ro.ro_rt != NULL) { + RTFREE(net->ro.ro_rt); + net->ro.ro_rt = NULL; + } +#endif + + /* Was it our primary? */ + if ((stcb->asoc.primary_destination == net) && (alt != net)) { + /* + * Yes, note it as such and find an alternate note: + * this means HB code must use this to resent the + * primary if it goes active AND if someone does a + * change-primary then this flag must be cleared + * from any net structures. + */ + if (stcb->asoc.alternate != NULL) { + sctp_free_remote_addr(stcb->asoc.alternate); + } + stcb->asoc.alternate = alt; + atomic_add_int(&stcb->asoc.alternate->ref_count, 1); + } + } + /* + * Special case for cookie-echo'ed case, we don't do output but must + * await the COOKIE-ACK before retransmission + */ + if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED) { + /* + * Here we just reset the timer and start again since we + * have not established the asoc + */ + sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); + return (0); + } + if (stcb->asoc.prsctp_supported) { + struct sctp_tmit_chunk *lchk; + + lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc); + /* C3. See if we need to send a Fwd-TSN */ + if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) { + send_forward_tsn(stcb, &stcb->asoc); + for (; lchk != NULL; lchk = TAILQ_NEXT(lchk, sctp_next)) { + if (lchk->whoTo != NULL) { + break; + } + } + if (lchk != NULL) { + /* Assure a timer is up */ + sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo); + } + } + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) { + sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX); + } + return (0); +} + +int +sctp_t1init_timer(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + /* bump the thresholds */ + if (stcb->asoc.delayed_connection) { + /* + * special hook for delayed connection. The library did NOT + * complete the rest of its sends. + */ + stcb->asoc.delayed_connection = 0; + sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED); + return (0); + } + if (SCTP_GET_STATE(stcb) != SCTP_STATE_COOKIE_WAIT) { + return (0); + } + if (sctp_threshold_management(inp, stcb, net, + stcb->asoc.max_init_times)) { + /* Association was destroyed */ + return (1); + } + stcb->asoc.dropped_special_cnt = 0; + sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0); + if (stcb->asoc.initial_init_rto_max < net->RTO) { + net->RTO = stcb->asoc.initial_init_rto_max; + } + if (stcb->asoc.numnets > 1) { + /* If we have more than one addr use it */ + struct sctp_nets *alt; + + alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0); + if (alt != stcb->asoc.primary_destination) { + sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination); + stcb->asoc.primary_destination = alt; + } + } + /* Send out a new init */ + sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED); + return (0); +} + +/* + * For cookie and asconf we actually need to find and mark for resend, then + * increment the resend counter (after all the threshold management stuff of + * course). + */ +int +sctp_cookie_timer(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_nets *net SCTP_UNUSED) +{ + struct sctp_nets *alt; + struct sctp_tmit_chunk *cookie; + + /* first before all else we must find the cookie */ + TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) { + if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) { + break; + } + } + if (cookie == NULL) { + if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED) { + /* FOOBAR! */ + struct mbuf *op_err; + + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + "Cookie timer expired, but no cookie"); + inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_3; + sctp_abort_an_association(inp, stcb, op_err, false, SCTP_SO_NOT_LOCKED); + } else { +#ifdef INVARIANTS + panic("Cookie timer expires in wrong state?"); +#else + SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(stcb)); + return (0); +#endif + } + return (0); + } + /* Ok we found the cookie, threshold management next */ + if (sctp_threshold_management(inp, stcb, cookie->whoTo, + stcb->asoc.max_init_times)) { + /* Assoc is over */ + return (1); + } + /* + * Cleared threshold management, now lets backoff the address + * and select an alternate + */ + stcb->asoc.dropped_special_cnt = 0; + sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0); + alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0); + if (alt != cookie->whoTo) { + sctp_free_remote_addr(cookie->whoTo); + cookie->whoTo = alt; + atomic_add_int(&alt->ref_count, 1); + } + /* Now mark the retran info */ + if (cookie->sent != SCTP_DATAGRAM_RESEND) { + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + } + cookie->sent = SCTP_DATAGRAM_RESEND; + cookie->flags |= CHUNK_FLAGS_FRAGMENT_OK; + /* + * Now call the output routine to kick out the cookie again, Note we + * don't mark any chunks for retran so that FR will need to kick in + * to move these (or a send timer). + */ + return (0); +} + +int +sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb) +{ + struct sctp_nets *alt, *net; + struct sctp_tmit_chunk *strrst = NULL, *chk = NULL; + + if (stcb->asoc.stream_reset_outstanding == 0) { + return (0); + } + /* find the existing STRRESET, we use the seq number we sent out on */ + (void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst); + if (strrst == NULL) { + return (0); + } + net = strrst->whoTo; + /* do threshold management */ + if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) { + /* Assoc is over */ + return (1); + } + /* + * Cleared threshold management, now lets backoff the address + * and select an alternate + */ + sctp_backoff_on_timeout(stcb, net, 1, 0, 0); + alt = sctp_find_alternate_net(stcb, net, 0); + strrst->whoTo = alt; + atomic_add_int(&alt->ref_count, 1); + + /* See if a ECN Echo is also stranded */ + TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) { + if ((chk->whoTo == net) && + (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) { + sctp_free_remote_addr(chk->whoTo); + if (chk->sent != SCTP_DATAGRAM_RESEND) { + chk->sent = SCTP_DATAGRAM_RESEND; + chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + } + chk->whoTo = alt; + atomic_add_int(&alt->ref_count, 1); + } + } + if ((net->dest_state & SCTP_ADDR_REACHABLE) == 0) { + /* + * If the address went un-reachable, we need to move to + * alternates for ALL chk's in queue + */ + sctp_move_chunks_from_net(stcb, net); + } + sctp_free_remote_addr(net); + + /* mark the retran info */ + if (strrst->sent != SCTP_DATAGRAM_RESEND) + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + strrst->sent = SCTP_DATAGRAM_RESEND; + strrst->flags |= CHUNK_FLAGS_FRAGMENT_OK; + + /* restart the timer */ + sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, alt); + return (0); +} + +int +sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + struct sctp_nets *alt; + struct sctp_tmit_chunk *asconf, *chk; + + /* is this a first send, or a retransmission? */ + if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) { + /* compose a new ASCONF chunk and send it */ + sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED); + } else { + /* + * Retransmission of the existing ASCONF is needed + */ + + /* find the existing ASCONF */ + asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue); + if (asconf == NULL) { + return (0); + } + net = asconf->whoTo; + /* do threshold management */ + if (sctp_threshold_management(inp, stcb, net, + stcb->asoc.max_send_times)) { + /* Assoc is over */ + return (1); + } + if (asconf->snd_count > stcb->asoc.max_send_times) { + /* + * Something is rotten: our peer is not responding to + * ASCONFs but apparently is to other chunks. i.e. it + * is not properly handling the chunk type upper bits. + * Mark this peer as ASCONF incapable and cleanup. + */ + SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n"); + sctp_asconf_cleanup(stcb); + return (0); + } + /* + * cleared threshold management, so now backoff the net and + * select an alternate + */ + sctp_backoff_on_timeout(stcb, net, 1, 0, 0); + alt = sctp_find_alternate_net(stcb, net, 0); + if (asconf->whoTo != alt) { + asconf->whoTo = alt; + atomic_add_int(&alt->ref_count, 1); + } + + /* See if an ECN Echo is also stranded */ + TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) { + if ((chk->whoTo == net) && + (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) { + sctp_free_remote_addr(chk->whoTo); + chk->whoTo = alt; + if (chk->sent != SCTP_DATAGRAM_RESEND) { + chk->sent = SCTP_DATAGRAM_RESEND; + chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + } + atomic_add_int(&alt->ref_count, 1); + } + } + TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) { + if (chk->whoTo != alt) { + sctp_free_remote_addr(chk->whoTo); + chk->whoTo = alt; + atomic_add_int(&alt->ref_count, 1); + } + if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT) + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + chk->sent = SCTP_DATAGRAM_RESEND; + chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; + } + if ((net->dest_state & SCTP_ADDR_REACHABLE) == 0) { + /* + * If the address went un-reachable, we need to move + * to the alternate for ALL chunks in queue + */ + sctp_move_chunks_from_net(stcb, net); + } + sctp_free_remote_addr(net); + + /* mark the retran info */ + if (asconf->sent != SCTP_DATAGRAM_RESEND) + sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); + asconf->sent = SCTP_DATAGRAM_RESEND; + asconf->flags |= CHUNK_FLAGS_FRAGMENT_OK; + + /* send another ASCONF if any and we can do */ + sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED); + } + return (0); +} + +/* Mobility adaptation */ +void +sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb) +{ + if (stcb->asoc.deleted_primary == NULL) { + SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n"); + sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); + return; + } + SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary "); + SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa); + sctp_free_remote_addr(stcb->asoc.deleted_primary); + stcb->asoc.deleted_primary = NULL; + sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); + return; +} + +/* + * For the shutdown and shutdown-ack, we do not keep one around on the + * control queue. This means we must generate a new one and call the general + * chunk output routine, AFTER having done threshold management. + * It is assumed that net is non-NULL. + */ +int +sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + struct sctp_nets *alt; + + /* first threshold management */ + if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) { + /* Assoc is over */ + return (1); + } + sctp_backoff_on_timeout(stcb, net, 1, 0, 0); + /* second select an alternative */ + alt = sctp_find_alternate_net(stcb, net, 0); + + /* third generate a shutdown into the queue for out net */ + sctp_send_shutdown(stcb, alt); + + /* fourth restart timer */ + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt); + return (0); +} + +int +sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + struct sctp_nets *alt; + + /* first threshold management */ + if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) { + /* Assoc is over */ + return (1); + } + sctp_backoff_on_timeout(stcb, net, 1, 0, 0); + /* second select an alternative */ + alt = sctp_find_alternate_net(stcb, net, 0); + + /* third generate a shutdown into the queue for out net */ + sctp_send_shutdown_ack(stcb, alt); + + /* fourth restart timer */ + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt); + return (0); +} + +static void +sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp, struct sctp_tcb *stcb) +{ + struct sctp_stream_queue_pending *sp; + unsigned int i, chks_in_queue = 0; + int being_filled = 0; + + KASSERT(inp != NULL, ("inp is NULL")); + KASSERT(stcb != NULL, ("stcb is NULL")); + SCTP_TCB_LOCK_ASSERT(stcb); + KASSERT(TAILQ_EMPTY(&stcb->asoc.send_queue), ("send_queue not empty")); + KASSERT(TAILQ_EMPTY(&stcb->asoc.sent_queue), ("sent_queue not empty")); + + if (stcb->asoc.sent_queue_retran_cnt) { + SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n", + stcb->asoc.sent_queue_retran_cnt); + stcb->asoc.sent_queue_retran_cnt = 0; + } + if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) { + /* No stream scheduler information, initialize scheduler */ + stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc); + if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) { + /* yep, we lost a stream or two */ + SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n"); + } else { + /* no streams lost */ + stcb->asoc.total_output_queue_size = 0; + } + } + /* Check to see if some data queued, if so report it */ + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) { + TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) { + if (sp->msg_is_complete) + being_filled++; + chks_in_queue++; + } + } + } + if (chks_in_queue != stcb->asoc.stream_queue_cnt) { + SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n", + stcb->asoc.stream_queue_cnt, chks_in_queue); + } + if (chks_in_queue) { + /* call the output queue function */ + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); + if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) && + (TAILQ_EMPTY(&stcb->asoc.sent_queue))) { + /* + * Probably should go in and make it go back through + * and add fragments allowed + */ + if (being_filled == 0) { + SCTP_PRINTF("Still nothing moved %d chunks are stuck\n", + chks_in_queue); + } + } + } else { + SCTP_PRINTF("Found no chunks on any queue tot:%lu\n", + (u_long)stcb->asoc.total_output_queue_size); + stcb->asoc.total_output_queue_size = 0; + } +} + +int +sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + bool net_was_pf; + + net_was_pf = (net->dest_state & SCTP_ADDR_PF) != 0; + if (net->hb_responded == 0) { + if (net->ro._s_addr != NULL) { + /* Invalidate the src address if we did not get + * a response last time. + */ + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + net->src_addr_selected = 0; + } + sctp_backoff_on_timeout(stcb, net, 1, 0, 0); + if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) { + /* Assoc is over */ + return (1); + } + } + /* Zero PBA, if it needs it */ + if (net->partial_bytes_acked > 0) { + net->partial_bytes_acked = 0; + } + if ((stcb->asoc.total_output_queue_size > 0) && + (TAILQ_EMPTY(&stcb->asoc.send_queue)) && + (TAILQ_EMPTY(&stcb->asoc.sent_queue))) { + sctp_audit_stream_queues_for_size(inp, stcb); + } + if ((((net->dest_state & SCTP_ADDR_NOHB) == 0) || + (net->dest_state & SCTP_ADDR_UNCONFIRMED)) && + (net_was_pf || ((net->dest_state & SCTP_ADDR_PF) == 0))) { + /* When moving to PF during threshold management, a HB has been + queued in that routine. */ + uint32_t ms_gone_by; + + if ((net->last_sent_time.tv_sec > 0) || + (net->last_sent_time.tv_usec > 0)) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct timeval diff; + + SCTP_GETTIME_TIMEVAL(&diff); + timevalsub(&diff, &net->last_sent_time); +#else + struct timeval diff, now; + + SCTP_GETTIME_TIMEVAL(&now); + timersub(&now, &net->last_sent_time, &diff); +#endif + ms_gone_by = (uint32_t)(diff.tv_sec * 1000) + + (uint32_t)(diff.tv_usec / 1000); + } else { + ms_gone_by = 0xffffffff; + } + if ((ms_gone_by >= net->heart_beat_delay) || + (net->dest_state & SCTP_ADDR_UNCONFIRMED) || + (net->dest_state & SCTP_ADDR_PF)) { + sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED); + } + } + return (0); +} + +void +sctp_pathmtu_timer(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + uint32_t next_mtu, mtu; + + next_mtu = sctp_get_next_mtu(net->mtu); + + if ((next_mtu > net->mtu) && (net->port == 0)) { + if ((net->src_addr_selected == 0) || + (net->ro._s_addr == NULL) || + (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) { + if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) { + sctp_free_ifa(net->ro._s_addr); + net->ro._s_addr = NULL; + net->src_addr_selected = 0; + } else if (net->ro._s_addr == NULL) { +#if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE) + if (net->ro._l_addr.sa.sa_family == AF_INET6) { + struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; + /* KAME hack: embed scopeid */ +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) + (void)in6_embedscope(&sin6->sin6_addr, sin6, NULL, NULL); +#else + (void)in6_embedscope(&sin6->sin6_addr, sin6, NULL, NULL, NULL); +#endif +#elif defined(SCTP_KAME) + (void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)); +#else + (void)in6_embedscope(&sin6->sin6_addr, sin6); +#endif + } +#endif + + net->ro._s_addr = sctp_source_address_selection(inp, + stcb, + (sctp_route_t *)&net->ro, + net, 0, stcb->asoc.vrf_id); +#if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE) + if (net->ro._l_addr.sa.sa_family == AF_INET6) { + struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; +#ifdef SCTP_KAME + (void)sa6_recoverscope(sin6); +#else + (void)in6_recoverscope(sin6, &sin6->sin6_addr, NULL); +#endif /* SCTP_KAME */ + } +#endif /* INET6 */ + } + if (net->ro._s_addr) + net->src_addr_selected = 1; + } + if (net->ro._s_addr) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_nh); +#else + mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt); +#endif +#if defined(INET) || defined(INET6) + if (net->port) { + mtu -= sizeof(struct udphdr); + } +#endif + if (mtu > next_mtu) { + net->mtu = next_mtu; + } else { + net->mtu = mtu; + } + } + } + /* restart the timer */ + sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net); +} + +void +sctp_autoclose_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb) +{ + struct timeval tn, *tim_touse; + struct sctp_association *asoc; + uint32_t ticks_gone_by; + + (void)SCTP_GETTIME_TIMEVAL(&tn); + if (stcb->asoc.sctp_autoclose_ticks > 0 && + sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) { + /* Auto close is on */ + asoc = &stcb->asoc; + /* pick the time to use */ + if (asoc->time_last_rcvd.tv_sec > + asoc->time_last_sent.tv_sec) { + tim_touse = &asoc->time_last_rcvd; + } else { + tim_touse = &asoc->time_last_sent; + } + /* Now has long enough transpired to autoclose? */ + ticks_gone_by = sctp_secs_to_ticks((uint32_t)(tn.tv_sec - tim_touse->tv_sec)); + if (ticks_gone_by >= asoc->sctp_autoclose_ticks) { + /* + * autoclose time has hit, call the output routine, + * which should do nothing just to be SURE we don't + * have hanging data. We can then safely check the + * queues and know that we are clear to send + * shutdown + */ + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED); + /* Are we clean? */ + if (TAILQ_EMPTY(&asoc->send_queue) && + TAILQ_EMPTY(&asoc->sent_queue)) { + /* + * there is nothing queued to send, so I'm + * done... + */ + if (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) { + /* only send SHUTDOWN 1st time thru */ + struct sctp_nets *net; + + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_SENT); + sctp_stop_timers_for_shutdown(stcb); + if (stcb->asoc.alternate) { + net = stcb->asoc.alternate; + } else { + net = stcb->asoc.primary_destination; + } + sctp_send_shutdown(stcb, net); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, + stcb->sctp_ep, stcb, net); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, + stcb->sctp_ep, stcb, NULL); + } + } + } else { + /* + * No auto close at this time, reset t-o to check + * later + */ + uint32_t tmp; + + /* fool the timer startup to use the time left */ + tmp = asoc->sctp_autoclose_ticks; + asoc->sctp_autoclose_ticks -= ticks_gone_by; + sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb, NULL); + /* restore the real tick value */ + asoc->sctp_autoclose_ticks = tmp; + } + } +} + diff --git a/netwerk/sctp/src/netinet/sctp_timer.h b/netwerk/sctp/src/netinet/sctp_timer.h new file mode 100644 index 0000000000..0fe9c03b05 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_timer.h @@ -0,0 +1,103 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_TIMER_H_ +#define _NETINET_SCTP_TIMER_H_ + +#if defined(_KERNEL) || defined(__Userspace__) + +#define SCTP_RTT_SHIFT 3 +#define SCTP_RTT_VAR_SHIFT 2 + +struct sctp_nets * +sctp_find_alternate_net(struct sctp_tcb *, struct sctp_nets *, int); + +int +sctp_t3rxt_timer(struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *); + +int +sctp_t1init_timer(struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *); + +int +sctp_shutdown_timer(struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *); + +int +sctp_heartbeat_timer(struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *); + +int +sctp_cookie_timer(struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *); + +void +sctp_pathmtu_timer(struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *); + +int +sctp_shutdownack_timer(struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *); +int +sctp_strreset_timer(struct sctp_inpcb *, struct sctp_tcb *); + +int +sctp_asconf_timer(struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *); + +void +sctp_delete_prim_timer(struct sctp_inpcb *, struct sctp_tcb *); + +void +sctp_autoclose_timer(struct sctp_inpcb *, struct sctp_tcb *); + +void sctp_audit_retranmission_queue(struct sctp_association *); + +void sctp_iterator_timer(struct sctp_iterator *it); + +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD) || defined(APPLE_LION) || defined(APPLE_MOUNTAINLION) +void sctp_slowtimo(void); +#else +void sctp_gc(struct inpcbinfo *); +#endif +#endif +#endif +#endif diff --git a/netwerk/sctp/src/netinet/sctp_uio.h b/netwerk/sctp/src/netinet/sctp_uio.h new file mode 100644 index 0000000000..eee2d6447c --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_uio.h @@ -0,0 +1,1354 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_UIO_H_ +#define _NETINET_SCTP_UIO_H_ + +#if (defined(__APPLE__) && !defined(__Userspace__) && defined(KERNEL)) +#ifndef _KERNEL +#define _KERNEL +#endif +#endif +#if !defined(_WIN32) +#if !defined(_KERNEL) +#include <stdint.h> +#endif +#include <sys/types.h> +#include <sys/socket.h> +#include <netinet/in.h> +#endif +#if defined(_WIN32) && !defined(__Userspace__) +#pragma warning(push) +#pragma warning(disable: 4200) +#if defined(_KERNEL) +#include <sys/types.h> +#include <sys/socket.h> +#include <netinet/in.h> +#endif +#endif + +typedef uint32_t sctp_assoc_t; + +#define SCTP_FUTURE_ASSOC 0 +#define SCTP_CURRENT_ASSOC 1 +#define SCTP_ALL_ASSOC 2 + +struct sctp_event { + sctp_assoc_t se_assoc_id; + uint16_t se_type; + uint8_t se_on; +}; + +/* Compatibility to previous define's */ +#define sctp_stream_reset_events sctp_stream_reset_event + +/* On/Off setup for subscription to events */ +struct sctp_event_subscribe { + uint8_t sctp_data_io_event; + uint8_t sctp_association_event; + uint8_t sctp_address_event; + uint8_t sctp_send_failure_event; + uint8_t sctp_peer_error_event; + uint8_t sctp_shutdown_event; + uint8_t sctp_partial_delivery_event; + uint8_t sctp_adaptation_layer_event; + uint8_t sctp_authentication_event; + uint8_t sctp_sender_dry_event; + uint8_t sctp_stream_reset_event; +}; + +/* ancillary data types */ +#define SCTP_INIT 0x0001 +#define SCTP_SNDRCV 0x0002 +#define SCTP_EXTRCV 0x0003 +#define SCTP_SNDINFO 0x0004 +#define SCTP_RCVINFO 0x0005 +#define SCTP_NXTINFO 0x0006 +#define SCTP_PRINFO 0x0007 +#define SCTP_AUTHINFO 0x0008 +#define SCTP_DSTADDRV4 0x0009 +#define SCTP_DSTADDRV6 0x000a + +/* + * ancillary data structures + */ +struct sctp_initmsg { + uint16_t sinit_num_ostreams; + uint16_t sinit_max_instreams; + uint16_t sinit_max_attempts; + uint16_t sinit_max_init_timeo; +}; + +/* We add 96 bytes to the size of sctp_sndrcvinfo. + * This makes the current structure 128 bytes long + * which is nicely 64 bit aligned but also has room + * for us to add more and keep ABI compatibility. + * For example, already we have the sctp_extrcvinfo + * when enabled which is 48 bytes. + */ + +/* + * The assoc up needs a verfid + * all sendrcvinfo's need a verfid for SENDING only. + */ + +#define SCTP_ALIGN_RESV_PAD 92 +#define SCTP_ALIGN_RESV_PAD_SHORT 76 + +struct sctp_sndrcvinfo { + uint16_t sinfo_stream; + uint16_t sinfo_ssn; + uint16_t sinfo_flags; + uint32_t sinfo_ppid; + uint32_t sinfo_context; + uint32_t sinfo_timetolive; + uint32_t sinfo_tsn; + uint32_t sinfo_cumtsn; + sctp_assoc_t sinfo_assoc_id; + uint16_t sinfo_keynumber; + uint16_t sinfo_keynumber_valid; + uint8_t __reserve_pad[SCTP_ALIGN_RESV_PAD]; +}; + +struct sctp_extrcvinfo { + uint16_t sinfo_stream; + uint16_t sinfo_ssn; + uint16_t sinfo_flags; + uint32_t sinfo_ppid; + uint32_t sinfo_context; + uint32_t sinfo_timetolive; /* should have been sinfo_pr_value */ + uint32_t sinfo_tsn; + uint32_t sinfo_cumtsn; + sctp_assoc_t sinfo_assoc_id; + uint16_t serinfo_next_flags; + uint16_t serinfo_next_stream; + uint32_t serinfo_next_aid; + uint32_t serinfo_next_length; + uint32_t serinfo_next_ppid; + uint16_t sinfo_keynumber; + uint16_t sinfo_keynumber_valid; + uint8_t __reserve_pad[SCTP_ALIGN_RESV_PAD_SHORT]; +}; +#define sinfo_pr_value sinfo_timetolive +#define sreinfo_next_flags serinfo_next_flags +#define sreinfo_next_stream serinfo_next_stream +#define sreinfo_next_aid serinfo_next_aid +#define sreinfo_next_length serinfo_next_length +#define sreinfo_next_ppid serinfo_next_ppid + +struct sctp_sndinfo { + uint16_t snd_sid; + uint16_t snd_flags; + uint32_t snd_ppid; + uint32_t snd_context; + sctp_assoc_t snd_assoc_id; +}; + +struct sctp_prinfo { + uint16_t pr_policy; + uint32_t pr_value; +}; + +struct sctp_default_prinfo { + uint16_t pr_policy; + uint32_t pr_value; + sctp_assoc_t pr_assoc_id; +}; + +struct sctp_authinfo { + uint16_t auth_keynumber; +}; + +struct sctp_rcvinfo { + uint16_t rcv_sid; + uint16_t rcv_ssn; + uint16_t rcv_flags; + uint32_t rcv_ppid; + uint32_t rcv_tsn; + uint32_t rcv_cumtsn; + uint32_t rcv_context; + sctp_assoc_t rcv_assoc_id; +}; + +struct sctp_nxtinfo { + uint16_t nxt_sid; + uint16_t nxt_flags; + uint32_t nxt_ppid; + uint32_t nxt_length; + sctp_assoc_t nxt_assoc_id; +}; + +#define SCTP_NO_NEXT_MSG 0x0000 +#define SCTP_NEXT_MSG_AVAIL 0x0001 +#define SCTP_NEXT_MSG_ISCOMPLETE 0x0002 +#define SCTP_NEXT_MSG_IS_UNORDERED 0x0004 +#define SCTP_NEXT_MSG_IS_NOTIFICATION 0x0008 + +struct sctp_recvv_rn { + struct sctp_rcvinfo recvv_rcvinfo; + struct sctp_nxtinfo recvv_nxtinfo; +}; + +#define SCTP_RECVV_NOINFO 0 +#define SCTP_RECVV_RCVINFO 1 +#define SCTP_RECVV_NXTINFO 2 +#define SCTP_RECVV_RN 3 + +#define SCTP_SENDV_NOINFO 0 +#define SCTP_SENDV_SNDINFO 1 +#define SCTP_SENDV_PRINFO 2 +#define SCTP_SENDV_AUTHINFO 3 +#define SCTP_SENDV_SPA 4 + +struct sctp_sendv_spa { + uint32_t sendv_flags; + struct sctp_sndinfo sendv_sndinfo; + struct sctp_prinfo sendv_prinfo; + struct sctp_authinfo sendv_authinfo; +}; + +#define SCTP_SEND_SNDINFO_VALID 0x00000001 +#define SCTP_SEND_PRINFO_VALID 0x00000002 +#define SCTP_SEND_AUTHINFO_VALID 0x00000004 + +struct sctp_snd_all_completes { + uint16_t sall_stream; + uint16_t sall_flags; + uint32_t sall_ppid; + uint32_t sall_context; + uint32_t sall_num_sent; + uint32_t sall_num_failed; +}; + +/* Flags that go into the sinfo->sinfo_flags field */ +#define SCTP_NOTIFICATION 0x0010 /* next message is a notification */ +#define SCTP_COMPLETE 0x0020 /* next message is complete */ +#define SCTP_EOF 0x0100 /* Start shutdown procedures */ +#define SCTP_ABORT 0x0200 /* Send an ABORT to peer */ +#define SCTP_UNORDERED 0x0400 /* Message is un-ordered */ +#define SCTP_ADDR_OVER 0x0800 /* Override the primary-address */ +#define SCTP_SENDALL 0x1000 /* Send this on all associations */ +#define SCTP_EOR 0x2000 /* end of message signal */ +#define SCTP_SACK_IMMEDIATELY 0x4000 /* Set I-Bit */ + +#define INVALID_SINFO_FLAG(x) (((x) & 0xfffffff0 \ + & ~(SCTP_EOF | SCTP_ABORT | SCTP_UNORDERED |\ + SCTP_ADDR_OVER | SCTP_SENDALL | SCTP_EOR |\ + SCTP_SACK_IMMEDIATELY)) != 0) +/* for the endpoint */ + +/* The lower four bits is an enumeration of PR-SCTP policies */ +#define SCTP_PR_SCTP_NONE 0x0000 /* Reliable transfer */ +#define SCTP_PR_SCTP_TTL 0x0001 /* Time based PR-SCTP */ +#define SCTP_PR_SCTP_PRIO 0x0002 /* Buffer based PR-SCTP */ +#define SCTP_PR_SCTP_BUF SCTP_PR_SCTP_PRIO /* For backwards compatibility */ +#define SCTP_PR_SCTP_RTX 0x0003 /* Number of retransmissions based PR-SCTP */ +#define SCTP_PR_SCTP_MAX SCTP_PR_SCTP_RTX +#define SCTP_PR_SCTP_ALL 0x000f /* Used for aggregated stats */ + +#define PR_SCTP_POLICY(x) ((x) & 0x0f) +#define PR_SCTP_ENABLED(x) ((PR_SCTP_POLICY(x) != SCTP_PR_SCTP_NONE) && \ + (PR_SCTP_POLICY(x) != SCTP_PR_SCTP_ALL)) +#define PR_SCTP_TTL_ENABLED(x) (PR_SCTP_POLICY(x) == SCTP_PR_SCTP_TTL) +#define PR_SCTP_BUF_ENABLED(x) (PR_SCTP_POLICY(x) == SCTP_PR_SCTP_BUF) +#define PR_SCTP_RTX_ENABLED(x) (PR_SCTP_POLICY(x) == SCTP_PR_SCTP_RTX) +#define PR_SCTP_INVALID_POLICY(x) (PR_SCTP_POLICY(x) > SCTP_PR_SCTP_MAX) +#define PR_SCTP_VALID_POLICY(x) (PR_SCTP_POLICY(x) <= SCTP_PR_SCTP_MAX) + +/* Stat's */ +struct sctp_pcbinfo { + uint32_t ep_count; + uint32_t asoc_count; + uint32_t laddr_count; + uint32_t raddr_count; + uint32_t chk_count; + uint32_t readq_count; + uint32_t free_chunks; + uint32_t stream_oque; +}; + +struct sctp_sockstat { + sctp_assoc_t ss_assoc_id; + uint32_t ss_total_sndbuf; + uint32_t ss_total_recv_buf; +}; + +/* + * notification event structures + */ + +/* + * association change event + */ +struct sctp_assoc_change { + uint16_t sac_type; + uint16_t sac_flags; + uint32_t sac_length; + uint16_t sac_state; + uint16_t sac_error; + uint16_t sac_outbound_streams; + uint16_t sac_inbound_streams; + sctp_assoc_t sac_assoc_id; + uint8_t sac_info[]; +}; + +/* sac_state values */ +#define SCTP_COMM_UP 0x0001 +#define SCTP_COMM_LOST 0x0002 +#define SCTP_RESTART 0x0003 +#define SCTP_SHUTDOWN_COMP 0x0004 +#define SCTP_CANT_STR_ASSOC 0x0005 + +/* sac_info values */ +#define SCTP_ASSOC_SUPPORTS_PR 0x01 +#define SCTP_ASSOC_SUPPORTS_AUTH 0x02 +#define SCTP_ASSOC_SUPPORTS_ASCONF 0x03 +#define SCTP_ASSOC_SUPPORTS_MULTIBUF 0x04 +#define SCTP_ASSOC_SUPPORTS_RE_CONFIG 0x05 +#define SCTP_ASSOC_SUPPORTS_INTERLEAVING 0x06 +#define SCTP_ASSOC_SUPPORTS_MAX 0x06 +/* + * Address event + */ +struct sctp_paddr_change { + uint16_t spc_type; + uint16_t spc_flags; + uint32_t spc_length; + struct sockaddr_storage spc_aaddr; + uint32_t spc_state; + uint32_t spc_error; + sctp_assoc_t spc_assoc_id; +}; + +/* paddr state values */ +#define SCTP_ADDR_AVAILABLE 0x0001 +#define SCTP_ADDR_UNREACHABLE 0x0002 +#define SCTP_ADDR_REMOVED 0x0003 +#define SCTP_ADDR_ADDED 0x0004 +#define SCTP_ADDR_MADE_PRIM 0x0005 +#define SCTP_ADDR_CONFIRMED 0x0006 + +#define SCTP_ACTIVE 0x0001 /* SCTP_ADDR_REACHABLE */ +#define SCTP_INACTIVE 0x0002 /* neither SCTP_ADDR_REACHABLE + nor SCTP_ADDR_UNCONFIRMED */ +#define SCTP_UNCONFIRMED 0x0200 /* SCTP_ADDR_UNCONFIRMED */ + +/* remote error events */ +struct sctp_remote_error { + uint16_t sre_type; + uint16_t sre_flags; + uint32_t sre_length; + uint16_t sre_error; + sctp_assoc_t sre_assoc_id; + uint8_t sre_data[]; +}; + +/* data send failure event (deprecated) */ +struct sctp_send_failed { + uint16_t ssf_type; + uint16_t ssf_flags; + uint32_t ssf_length; + uint32_t ssf_error; + struct sctp_sndrcvinfo ssf_info; + sctp_assoc_t ssf_assoc_id; + uint8_t ssf_data[]; +}; + +/* data send failure event (not deprecated) */ +struct sctp_send_failed_event { + uint16_t ssfe_type; + uint16_t ssfe_flags; + uint32_t ssfe_length; + uint32_t ssfe_error; + struct sctp_sndinfo ssfe_info; + sctp_assoc_t ssfe_assoc_id; + uint8_t ssfe_data[]; +}; + +/* flag that indicates state of data */ +#define SCTP_DATA_UNSENT 0x0001 /* inqueue never on wire */ +#define SCTP_DATA_SENT 0x0002 /* on wire at failure */ + +/* shutdown event */ +struct sctp_shutdown_event { + uint16_t sse_type; + uint16_t sse_flags; + uint32_t sse_length; + sctp_assoc_t sse_assoc_id; +}; + +/* Adaptation layer indication stuff */ +struct sctp_adaptation_event { + uint16_t sai_type; + uint16_t sai_flags; + uint32_t sai_length; + uint32_t sai_adaptation_ind; + sctp_assoc_t sai_assoc_id; +}; + +struct sctp_setadaptation { + uint32_t ssb_adaptation_ind; +}; + +/* compatible old spelling */ +struct sctp_adaption_event { + uint16_t sai_type; + uint16_t sai_flags; + uint32_t sai_length; + uint32_t sai_adaption_ind; + sctp_assoc_t sai_assoc_id; +}; + +struct sctp_setadaption { + uint32_t ssb_adaption_ind; +}; + +/* + * Partial Delivery API event + */ +struct sctp_pdapi_event { + uint16_t pdapi_type; + uint16_t pdapi_flags; + uint32_t pdapi_length; + uint32_t pdapi_indication; + uint16_t pdapi_stream; + uint16_t pdapi_seq; + sctp_assoc_t pdapi_assoc_id; +}; + +/* indication values */ +#define SCTP_PARTIAL_DELIVERY_ABORTED 0x0001 + +/* + * authentication key event + */ +struct sctp_authkey_event { + uint16_t auth_type; + uint16_t auth_flags; + uint32_t auth_length; + uint16_t auth_keynumber; + uint16_t auth_altkeynumber; + uint32_t auth_indication; + sctp_assoc_t auth_assoc_id; +}; + +/* indication values */ +#define SCTP_AUTH_NEW_KEY 0x0001 +#define SCTP_AUTH_NEWKEY SCTP_AUTH_NEW_KEY +#define SCTP_AUTH_NO_AUTH 0x0002 +#define SCTP_AUTH_FREE_KEY 0x0003 + +struct sctp_sender_dry_event { + uint16_t sender_dry_type; + uint16_t sender_dry_flags; + uint32_t sender_dry_length; + sctp_assoc_t sender_dry_assoc_id; +}; + +/* + * Stream reset event - subscribe to SCTP_STREAM_RESET_EVENT + */ +struct sctp_stream_reset_event { + uint16_t strreset_type; + uint16_t strreset_flags; + uint32_t strreset_length; + sctp_assoc_t strreset_assoc_id; + uint16_t strreset_stream_list[]; +}; + +/* flags in stream_reset_event (strreset_flags) */ +#define SCTP_STREAM_RESET_INCOMING_SSN 0x0001 +#define SCTP_STREAM_RESET_OUTGOING_SSN 0x0002 +#define SCTP_STREAM_RESET_DENIED 0x0004 +#define SCTP_STREAM_RESET_FAILED 0x0008 + +/* + * Assoc reset event - subscribe to SCTP_ASSOC_RESET_EVENT + */ +struct sctp_assoc_reset_event { + uint16_t assocreset_type; + uint16_t assocreset_flags; + uint32_t assocreset_length; + sctp_assoc_t assocreset_assoc_id; + uint32_t assocreset_local_tsn; + uint32_t assocreset_remote_tsn; +}; + +#define SCTP_ASSOC_RESET_DENIED 0x0004 +#define SCTP_ASSOC_RESET_FAILED 0x0008 + +/* + * Stream change event - subscribe to SCTP_STREAM_CHANGE_EVENT + */ +struct sctp_stream_change_event { + uint16_t strchange_type; + uint16_t strchange_flags; + uint32_t strchange_length; + sctp_assoc_t strchange_assoc_id; + uint16_t strchange_instrms; + uint16_t strchange_outstrms; +}; + +#define SCTP_STREAM_CHANGE_DENIED 0x0004 +#define SCTP_STREAM_CHANGE_FAILED 0x0008 + +/* SCTP notification event */ +struct sctp_tlv { + uint16_t sn_type; + uint16_t sn_flags; + uint32_t sn_length; +}; + +union sctp_notification { + struct sctp_tlv sn_header; + struct sctp_assoc_change sn_assoc_change; + struct sctp_paddr_change sn_paddr_change; + struct sctp_remote_error sn_remote_error; + struct sctp_send_failed sn_send_failed; + struct sctp_shutdown_event sn_shutdown_event; + struct sctp_adaptation_event sn_adaptation_event; + /* compatibility same as above */ + struct sctp_adaption_event sn_adaption_event; + struct sctp_pdapi_event sn_pdapi_event; + struct sctp_authkey_event sn_auth_event; + struct sctp_sender_dry_event sn_sender_dry_event; + struct sctp_send_failed_event sn_send_failed_event; + struct sctp_stream_reset_event sn_strreset_event; + struct sctp_assoc_reset_event sn_assocreset_event; + struct sctp_stream_change_event sn_strchange_event; +}; + +/* notification types */ +#define SCTP_ASSOC_CHANGE 0x0001 +#define SCTP_PEER_ADDR_CHANGE 0x0002 +#define SCTP_REMOTE_ERROR 0x0003 +#define SCTP_SEND_FAILED 0x0004 +#define SCTP_SHUTDOWN_EVENT 0x0005 +#define SCTP_ADAPTATION_INDICATION 0x0006 +/* same as above */ +#define SCTP_ADAPTION_INDICATION 0x0006 +#define SCTP_PARTIAL_DELIVERY_EVENT 0x0007 +#define SCTP_AUTHENTICATION_EVENT 0x0008 +#define SCTP_STREAM_RESET_EVENT 0x0009 +#define SCTP_SENDER_DRY_EVENT 0x000a +#define SCTP_NOTIFICATIONS_STOPPED_EVENT 0x000b /* we don't send this*/ +#define SCTP_ASSOC_RESET_EVENT 0x000c +#define SCTP_STREAM_CHANGE_EVENT 0x000d +#define SCTP_SEND_FAILED_EVENT 0x000e +/* + * socket option structs + */ + +struct sctp_paddrparams { + struct sockaddr_storage spp_address; + sctp_assoc_t spp_assoc_id; + uint32_t spp_hbinterval; + uint32_t spp_pathmtu; + uint32_t spp_flags; + uint32_t spp_ipv6_flowlabel; + uint16_t spp_pathmaxrxt; + uint8_t spp_dscp; +}; +#define spp_ipv4_tos spp_dscp + +#define SPP_HB_ENABLE 0x00000001 +#define SPP_HB_DISABLE 0x00000002 +#define SPP_HB_DEMAND 0x00000004 +#define SPP_PMTUD_ENABLE 0x00000008 +#define SPP_PMTUD_DISABLE 0x00000010 +#define SPP_HB_TIME_IS_ZERO 0x00000080 +#define SPP_IPV6_FLOWLABEL 0x00000100 +#define SPP_DSCP 0x00000200 +#define SPP_IPV4_TOS SPP_DSCP + +struct sctp_paddrthlds { + struct sockaddr_storage spt_address; + sctp_assoc_t spt_assoc_id; + uint16_t spt_pathmaxrxt; + uint16_t spt_pathpfthld; + uint16_t spt_pathcpthld; +}; + +struct sctp_paddrinfo { + struct sockaddr_storage spinfo_address; + sctp_assoc_t spinfo_assoc_id; + int32_t spinfo_state; + uint32_t spinfo_cwnd; + uint32_t spinfo_srtt; + uint32_t spinfo_rto; + uint32_t spinfo_mtu; +}; + +struct sctp_rtoinfo { + sctp_assoc_t srto_assoc_id; + uint32_t srto_initial; + uint32_t srto_max; + uint32_t srto_min; +}; + +struct sctp_assocparams { + sctp_assoc_t sasoc_assoc_id; + uint32_t sasoc_peer_rwnd; + uint32_t sasoc_local_rwnd; + uint32_t sasoc_cookie_life; + uint16_t sasoc_asocmaxrxt; + uint16_t sasoc_number_peer_destinations; +}; + +struct sctp_setprim { + struct sockaddr_storage ssp_addr; + sctp_assoc_t ssp_assoc_id; + uint8_t ssp_padding[4]; +}; + +struct sctp_setpeerprim { + struct sockaddr_storage sspp_addr; + sctp_assoc_t sspp_assoc_id; + uint8_t sspp_padding[4]; +}; + +union sctp_sockstore { + struct sockaddr_in sin; + struct sockaddr_in6 sin6; +#if defined(__Userspace__) + struct sockaddr_conn sconn; +#endif + struct sockaddr sa; +}; + +struct sctp_getaddresses { + sctp_assoc_t sget_assoc_id; + union sctp_sockstore addr[]; +}; + +struct sctp_status { + sctp_assoc_t sstat_assoc_id; + int32_t sstat_state; + uint32_t sstat_rwnd; + uint16_t sstat_unackdata; + uint16_t sstat_penddata; + uint16_t sstat_instrms; + uint16_t sstat_outstrms; + uint32_t sstat_fragmentation_point; + struct sctp_paddrinfo sstat_primary; +}; + +/* + * AUTHENTICATION support + */ +/* SCTP_AUTH_CHUNK */ +struct sctp_authchunk { + uint8_t sauth_chunk; +}; + +/* SCTP_AUTH_KEY */ +struct sctp_authkey { + sctp_assoc_t sca_assoc_id; + uint16_t sca_keynumber; + uint16_t sca_keylength; + uint8_t sca_key[]; +}; + +/* SCTP_HMAC_IDENT */ +struct sctp_hmacalgo { + uint32_t shmac_number_of_idents; + uint16_t shmac_idents[]; +}; + +/* AUTH hmac_id */ +#define SCTP_AUTH_HMAC_ID_RSVD 0x0000 +#define SCTP_AUTH_HMAC_ID_SHA1 0x0001 /* default, mandatory */ +#define SCTP_AUTH_HMAC_ID_SHA256 0x0003 + +/* SCTP_AUTH_ACTIVE_KEY / SCTP_AUTH_DELETE_KEY */ +struct sctp_authkeyid { + sctp_assoc_t scact_assoc_id; + uint16_t scact_keynumber; +}; + +/* SCTP_PEER_AUTH_CHUNKS / SCTP_LOCAL_AUTH_CHUNKS */ +struct sctp_authchunks { + sctp_assoc_t gauth_assoc_id; + uint32_t gauth_number_of_chunks; + uint8_t gauth_chunks[]; +}; + +struct sctp_assoc_value { + sctp_assoc_t assoc_id; + uint32_t assoc_value; +}; + +struct sctp_cc_option { + int option; + struct sctp_assoc_value aid_value; +}; + +struct sctp_stream_value { + sctp_assoc_t assoc_id; + uint16_t stream_id; + uint16_t stream_value; +}; + +struct sctp_assoc_ids { + uint32_t gaids_number_of_ids; + sctp_assoc_t gaids_assoc_id[]; +}; + +struct sctp_sack_info { + sctp_assoc_t sack_assoc_id; + uint32_t sack_delay; + uint32_t sack_freq; +}; + +struct sctp_timeouts { + sctp_assoc_t stimo_assoc_id; + uint32_t stimo_init; + uint32_t stimo_data; + uint32_t stimo_sack; + uint32_t stimo_shutdown; + uint32_t stimo_heartbeat; + uint32_t stimo_cookie; + uint32_t stimo_shutdownack; +}; + +struct sctp_udpencaps { + struct sockaddr_storage sue_address; + sctp_assoc_t sue_assoc_id; + uint16_t sue_port; +}; + +struct sctp_prstatus { + sctp_assoc_t sprstat_assoc_id; + uint16_t sprstat_sid; + uint16_t sprstat_policy; + uint64_t sprstat_abandoned_unsent; + uint64_t sprstat_abandoned_sent; +}; + +struct sctp_cwnd_args { + struct sctp_nets *net; /* network to */ /* FIXME: LP64 issue */ + uint32_t cwnd_new_value;/* cwnd in k */ + uint32_t pseudo_cumack; + uint16_t inflight; /* flightsize in k */ + uint16_t cwnd_augment; /* increment to it */ + uint8_t meets_pseudo_cumack; + uint8_t need_new_pseudo_cumack; + uint8_t cnt_in_send; + uint8_t cnt_in_str; +}; + +struct sctp_blk_args { + uint32_t onsb; /* in 1k bytes */ + uint32_t sndlen; /* len of send being attempted */ + uint32_t peer_rwnd; /* rwnd of peer */ + uint16_t send_sent_qcnt;/* chnk cnt */ + uint16_t stream_qcnt; /* chnk cnt */ + uint16_t chunks_on_oque;/* chunks out */ + uint16_t flight_size; /* flight size in k */ +}; + +/* + * Max we can reset in one setting, note this is dictated not by the define + * but the size of a mbuf cluster so don't change this define and think you + * can specify more. You must do multiple resets if you want to reset more + * than SCTP_MAX_EXPLICIT_STR_RESET. + */ +#define SCTP_MAX_EXPLICT_STR_RESET 1000 + +struct sctp_reset_streams { + sctp_assoc_t srs_assoc_id; + uint16_t srs_flags; + uint16_t srs_number_streams; /* 0 == ALL */ + uint16_t srs_stream_list[];/* list if strrst_num_streams is not 0 */ +}; + +struct sctp_add_streams { + sctp_assoc_t sas_assoc_id; + uint16_t sas_instrms; + uint16_t sas_outstrms; +}; + +struct sctp_get_nonce_values { + sctp_assoc_t gn_assoc_id; + uint32_t gn_peers_tag; + uint32_t gn_local_tag; +}; + +/* Debugging logs */ +struct sctp_str_log { + void *stcb; /* FIXME: LP64 issue */ + uint32_t n_tsn; + uint32_t e_tsn; + uint16_t n_sseq; + uint16_t e_sseq; + uint16_t strm; +}; + +struct sctp_sb_log { + void *stcb; /* FIXME: LP64 issue */ + uint32_t so_sbcc; + uint32_t stcb_sbcc; + uint32_t incr; +}; + +struct sctp_fr_log { + uint32_t largest_tsn; + uint32_t largest_new_tsn; + uint32_t tsn; +}; + +struct sctp_fr_map { + uint32_t base; + uint32_t cum; + uint32_t high; +}; + +struct sctp_rwnd_log { + uint32_t rwnd; + uint32_t send_size; + uint32_t overhead; + uint32_t new_rwnd; +}; + +struct sctp_mbcnt_log { + uint32_t total_queue_size; + uint32_t size_change; + uint32_t total_queue_mb_size; + uint32_t mbcnt_change; +}; + +struct sctp_sack_log { + uint32_t cumack; + uint32_t oldcumack; + uint32_t tsn; + uint16_t numGaps; + uint16_t numDups; +}; + +struct sctp_lock_log { + void *sock; /* FIXME: LP64 issue */ + void *inp; /* FIXME: LP64 issue */ + uint8_t tcb_lock; + uint8_t inp_lock; + uint8_t info_lock; + uint8_t sock_lock; + uint8_t sockrcvbuf_lock; + uint8_t socksndbuf_lock; + uint8_t create_lock; + uint8_t resv; +}; + +struct sctp_rto_log { + void * net; /* FIXME: LP64 issue */ + uint32_t rtt; +}; + +struct sctp_nagle_log { + void *stcb; /* FIXME: LP64 issue */ + uint32_t total_flight; + uint32_t total_in_queue; + uint16_t count_in_queue; + uint16_t count_in_flight; +}; + +struct sctp_sbwake_log { + void *stcb; /* FIXME: LP64 issue */ + uint16_t send_q; + uint16_t sent_q; + uint16_t flight; + uint16_t wake_cnt; + uint8_t stream_qcnt; /* chnk cnt */ + uint8_t chunks_on_oque;/* chunks out */ + uint8_t sbflags; + uint8_t sctpflags; +}; + +struct sctp_misc_info { + uint32_t log1; + uint32_t log2; + uint32_t log3; + uint32_t log4; +}; + +struct sctp_log_closing { + void *inp; /* FIXME: LP64 issue */ + void *stcb; /* FIXME: LP64 issue */ + uint32_t sctp_flags; + uint16_t state; + int16_t loc; +}; + +struct sctp_mbuf_log { + struct mbuf *mp; /* FIXME: LP64 issue */ + caddr_t ext; + caddr_t data; + uint16_t size; + uint8_t refcnt; + uint8_t mbuf_flags; +}; + +struct sctp_cwnd_log { + uint64_t time_event; + uint8_t from; + uint8_t event_type; + uint8_t resv[2]; + union { + struct sctp_log_closing close; + struct sctp_blk_args blk; + struct sctp_cwnd_args cwnd; + struct sctp_str_log strlog; + struct sctp_fr_log fr; + struct sctp_fr_map map; + struct sctp_rwnd_log rwnd; + struct sctp_mbcnt_log mbcnt; + struct sctp_sack_log sack; + struct sctp_lock_log lock; + struct sctp_rto_log rto; + struct sctp_sb_log sb; + struct sctp_nagle_log nagle; + struct sctp_sbwake_log wake; + struct sctp_mbuf_log mb; + struct sctp_misc_info misc; + } x; +}; + +struct sctp_cwnd_log_req { + int32_t num_in_log; /* Number in log */ + int32_t num_ret; /* Number returned */ + int32_t start_at; /* start at this one */ + int32_t end_at; /* end at this one */ + struct sctp_cwnd_log log[]; +}; + +struct sctp_timeval { + uint32_t tv_sec; + uint32_t tv_usec; +}; + +struct sctpstat { + struct sctp_timeval sctps_discontinuitytime; /* sctpStats 18 (TimeStamp) */ + /* MIB according to RFC 3873 */ + uint32_t sctps_currestab; /* sctpStats 1 (Gauge32) */ + uint32_t sctps_activeestab; /* sctpStats 2 (Counter32) */ + uint32_t sctps_restartestab; + uint32_t sctps_collisionestab; + uint32_t sctps_passiveestab; /* sctpStats 3 (Counter32) */ + uint32_t sctps_aborted; /* sctpStats 4 (Counter32) */ + uint32_t sctps_shutdown; /* sctpStats 5 (Counter32) */ + uint32_t sctps_outoftheblue; /* sctpStats 6 (Counter32) */ + uint32_t sctps_checksumerrors; /* sctpStats 7 (Counter32) */ + uint32_t sctps_outcontrolchunks; /* sctpStats 8 (Counter64) */ + uint32_t sctps_outorderchunks; /* sctpStats 9 (Counter64) */ + uint32_t sctps_outunorderchunks; /* sctpStats 10 (Counter64) */ + uint32_t sctps_incontrolchunks; /* sctpStats 11 (Counter64) */ + uint32_t sctps_inorderchunks; /* sctpStats 12 (Counter64) */ + uint32_t sctps_inunorderchunks; /* sctpStats 13 (Counter64) */ + uint32_t sctps_fragusrmsgs; /* sctpStats 14 (Counter64) */ + uint32_t sctps_reasmusrmsgs; /* sctpStats 15 (Counter64) */ + uint32_t sctps_outpackets; /* sctpStats 16 (Counter64) */ + uint32_t sctps_inpackets; /* sctpStats 17 (Counter64) */ + + /* input statistics: */ + uint32_t sctps_recvpackets; /* total input packets */ + uint32_t sctps_recvdatagrams; /* total input datagrams */ + uint32_t sctps_recvpktwithdata; /* total packets that had data */ + uint32_t sctps_recvsacks; /* total input SACK chunks */ + uint32_t sctps_recvdata; /* total input DATA chunks */ + uint32_t sctps_recvdupdata; /* total input duplicate DATA chunks */ + uint32_t sctps_recvheartbeat; /* total input HB chunks */ + uint32_t sctps_recvheartbeatack; /* total input HB-ACK chunks */ + uint32_t sctps_recvecne; /* total input ECNE chunks */ + uint32_t sctps_recvauth; /* total input AUTH chunks */ + uint32_t sctps_recvauthmissing; /* total input chunks missing AUTH */ + uint32_t sctps_recvivalhmacid; /* total number of invalid HMAC ids received */ + uint32_t sctps_recvivalkeyid; /* total number of invalid secret ids received */ + uint32_t sctps_recvauthfailed; /* total number of auth failed */ + uint32_t sctps_recvexpress; /* total fast path receives all one chunk */ + uint32_t sctps_recvexpressm; /* total fast path multi-part data */ + uint32_t sctps_recv_spare; /* formerly sctps_recvnocrc */ + uint32_t sctps_recvswcrc; + uint32_t sctps_recvhwcrc; + + /* output statistics: */ + uint32_t sctps_sendpackets; /* total output packets */ + uint32_t sctps_sendsacks; /* total output SACKs */ + uint32_t sctps_senddata; /* total output DATA chunks */ + uint32_t sctps_sendretransdata; /* total output retransmitted DATA chunks */ + uint32_t sctps_sendfastretrans; /* total output fast retransmitted DATA chunks */ + uint32_t sctps_sendmultfastretrans; /* total FR's that happened more than once + * to same chunk (u-del multi-fr algo). + */ + uint32_t sctps_sendheartbeat; /* total output HB chunks */ + uint32_t sctps_sendecne; /* total output ECNE chunks */ + uint32_t sctps_sendauth; /* total output AUTH chunks FIXME */ + uint32_t sctps_senderrors; /* ip_output error counter */ + uint32_t sctps_send_spare; /* formerly sctps_sendnocrc */ + uint32_t sctps_sendswcrc; + uint32_t sctps_sendhwcrc; + /* PCKDROPREP statistics: */ + uint32_t sctps_pdrpfmbox; /* Packet drop from middle box */ + uint32_t sctps_pdrpfehos; /* P-drop from end host */ + uint32_t sctps_pdrpmbda; /* P-drops with data */ + uint32_t sctps_pdrpmbct; /* P-drops, non-data, non-endhost */ + uint32_t sctps_pdrpbwrpt; /* P-drop, non-endhost, bandwidth rep only */ + uint32_t sctps_pdrpcrupt; /* P-drop, not enough for chunk header */ + uint32_t sctps_pdrpnedat; /* P-drop, not enough data to confirm */ + uint32_t sctps_pdrppdbrk; /* P-drop, where process_chunk_drop said break */ + uint32_t sctps_pdrptsnnf; /* P-drop, could not find TSN */ + uint32_t sctps_pdrpdnfnd; /* P-drop, attempt reverse TSN lookup */ + uint32_t sctps_pdrpdiwnp; /* P-drop, e-host confirms zero-rwnd */ + uint32_t sctps_pdrpdizrw; /* P-drop, midbox confirms no space */ + uint32_t sctps_pdrpbadd; /* P-drop, data did not match TSN */ + uint32_t sctps_pdrpmark; /* P-drop, TSN's marked for Fast Retran */ + /* timeouts */ + uint32_t sctps_timoiterator; /* Number of iterator timers that fired */ + uint32_t sctps_timodata; /* Number of T3 data time outs */ + uint32_t sctps_timowindowprobe; /* Number of window probe (T3) timers that fired */ + uint32_t sctps_timoinit; /* Number of INIT timers that fired */ + uint32_t sctps_timosack; /* Number of sack timers that fired */ + uint32_t sctps_timoshutdown; /* Number of shutdown timers that fired */ + uint32_t sctps_timoheartbeat; /* Number of heartbeat timers that fired */ + uint32_t sctps_timocookie; /* Number of times a cookie timeout fired */ + uint32_t sctps_timosecret; /* Number of times an endpoint changed its cookie secret*/ + uint32_t sctps_timopathmtu; /* Number of PMTU timers that fired */ + uint32_t sctps_timoshutdownack; /* Number of shutdown ack timers that fired */ + uint32_t sctps_timoshutdownguard; /* Number of shutdown guard timers that fired */ + uint32_t sctps_timostrmrst; /* Number of stream reset timers that fired */ + uint32_t sctps_timoearlyfr; /* Number of early FR timers that fired */ + uint32_t sctps_timoasconf; /* Number of times an asconf timer fired */ + uint32_t sctps_timodelprim; /* Number of times a prim_deleted timer fired */ + uint32_t sctps_timoautoclose; /* Number of times auto close timer fired */ + uint32_t sctps_timoassockill; /* Number of asoc free timers expired */ + uint32_t sctps_timoinpkill; /* Number of inp free timers expired */ + /* former early FR counters */ + uint32_t sctps_spare[11]; + /* others */ + uint32_t sctps_hdrops; /* packet shorter than header */ + uint32_t sctps_badsum; /* checksum error */ + uint32_t sctps_noport; /* no endpoint for port */ + uint32_t sctps_badvtag; /* bad v-tag */ + uint32_t sctps_badsid; /* bad SID */ + uint32_t sctps_nomem; /* no memory */ + uint32_t sctps_fastretransinrtt; /* number of multiple FR in a RTT window */ + uint32_t sctps_markedretrans; + uint32_t sctps_naglesent; /* nagle allowed sending */ + uint32_t sctps_naglequeued; /* nagle doesn't allow sending */ + uint32_t sctps_maxburstqueued; /* max burst doesn't allow sending */ + uint32_t sctps_ifnomemqueued; /* look ahead tells us no memory in + * interface ring buffer OR we had a + * send error and are queuing one send. + */ + uint32_t sctps_windowprobed; /* total number of window probes sent */ + uint32_t sctps_lowlevelerr; /* total times an output error causes us + * to clamp down on next user send. + */ + uint32_t sctps_lowlevelerrusr; /* total times sctp_senderrors were caused from + * a user send from a user invoked send not + * a sack response + */ + uint32_t sctps_datadropchklmt; /* Number of in data drops due to chunk limit reached */ + uint32_t sctps_datadroprwnd; /* Number of in data drops due to rwnd limit reached */ + uint32_t sctps_ecnereducedcwnd; /* Number of times a ECN reduced the cwnd */ + uint32_t sctps_vtagexpress; /* Used express lookup via vtag */ + uint32_t sctps_vtagbogus; /* Collision in express lookup. */ + uint32_t sctps_primary_randry; /* Number of times the sender ran dry of user data on primary */ + uint32_t sctps_cmt_randry; /* Same for above */ + uint32_t sctps_slowpath_sack; /* Sacks the slow way */ + uint32_t sctps_wu_sacks_sent; /* Window Update only sacks sent */ + uint32_t sctps_sends_with_flags; /* number of sends with sinfo_flags !=0 */ + uint32_t sctps_sends_with_unord; /* number of unordered sends */ + uint32_t sctps_sends_with_eof; /* number of sends with EOF flag set */ + uint32_t sctps_sends_with_abort; /* number of sends with ABORT flag set */ + uint32_t sctps_protocol_drain_calls; /* number of times protocol drain called */ + uint32_t sctps_protocol_drains_done; /* number of times we did a protocol drain */ + uint32_t sctps_read_peeks; /* Number of times recv was called with peek */ + uint32_t sctps_cached_chk; /* Number of cached chunks used */ + uint32_t sctps_cached_strmoq; /* Number of cached stream oq's used */ + uint32_t sctps_left_abandon; /* Number of unread messages abandoned by close */ + uint32_t sctps_send_burst_avoid; /* Unused */ + uint32_t sctps_send_cwnd_avoid; /* Send cwnd full avoidance, already max burst inflight to net */ + uint32_t sctps_fwdtsn_map_over; /* number of map array over-runs via fwd-tsn's */ + uint32_t sctps_queue_upd_ecne; /* Number of times we queued or updated an ECN chunk on send queue */ + uint32_t sctps_recvzerocrc; /* Number of accepted packets with zero CRC */ + uint32_t sctps_sendzerocrc; /* Number of packets sent with zero CRC */ + uint32_t sctps_reserved[29]; /* Future ABI compat - remove int's from here when adding new */ +}; + +#define SCTP_STAT_INCR(_x) SCTP_STAT_INCR_BY(_x,1) +#define SCTP_STAT_DECR(_x) SCTP_STAT_DECR_BY(_x,1) +#if defined(__FreeBSD__) && !defined(__Userspace__) +#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT) +#define SCTP_STAT_INCR_BY(_x,_d) (SCTP_BASE_STATS[PCPU_GET(cpuid)]._x += _d) +#define SCTP_STAT_DECR_BY(_x,_d) (SCTP_BASE_STATS[PCPU_GET(cpuid)]._x -= _d) +#else +#define SCTP_STAT_INCR_BY(_x,_d) atomic_add_int(&SCTP_BASE_STAT(_x), _d) +#define SCTP_STAT_DECR_BY(_x,_d) atomic_subtract_int(&SCTP_BASE_STAT(_x), _d) +#endif +#else +#define SCTP_STAT_INCR_BY(_x,_d) atomic_add_int(&SCTP_BASE_STAT(_x), _d) +#define SCTP_STAT_DECR_BY(_x,_d) atomic_subtract_int(&SCTP_BASE_STAT(_x), _d) +#endif +/* The following macros are for handling MIB values, */ +#define SCTP_STAT_INCR_COUNTER32(_x) SCTP_STAT_INCR(_x) +#define SCTP_STAT_INCR_COUNTER64(_x) SCTP_STAT_INCR(_x) +#define SCTP_STAT_INCR_GAUGE32(_x) SCTP_STAT_INCR(_x) +#define SCTP_STAT_DECR_COUNTER32(_x) SCTP_STAT_DECR(_x) +#define SCTP_STAT_DECR_COUNTER64(_x) SCTP_STAT_DECR(_x) +#define SCTP_STAT_DECR_GAUGE32(_x) SCTP_STAT_DECR(_x) + +/***********************************/ +/* And something for us old timers */ +/***********************************/ + +#if !(defined(__APPLE__) && !defined(__Userspace__)) +#if !defined(__Userspace__) +#ifndef ntohll +#if defined(__linux__) +#ifndef _BSD_SOURCE +#define _BSD_SOURCE +#endif +#include <endian.h> +#else +#include <sys/endian.h> +#endif +#define ntohll(x) be64toh(x) +#endif + +#ifndef htonll +#if defined(__linux__) +#ifndef _BSD_SOURCE +#define _BSD_SOURCE +#endif +#include <endian.h> +#else +#include <sys/endian.h> +#endif +#define htonll(x) htobe64(x) +#endif +#endif +#endif +/***********************************/ + +struct xsctp_inpcb { + uint32_t last; + uint32_t flags; + uint64_t features; + uint32_t total_sends; + uint32_t total_recvs; + uint32_t total_nospaces; + uint32_t fragmentation_point; + uint16_t local_port; +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint16_t qlen_old; + uint16_t maxqlen_old; +#else + uint16_t qlen; + uint16_t maxqlen; +#endif + uint16_t __spare16; +#if defined(__FreeBSD__) && !defined(__Userspace__) + kvaddr_t socket; +#else + void *socket; +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint32_t qlen; + uint32_t maxqlen; +#endif + uint32_t extra_padding[26]; /* future */ +}; + +struct xsctp_tcb { + union sctp_sockstore primary_addr; /* sctpAssocEntry 5/6 */ + uint32_t last; + uint32_t heartbeat_interval; /* sctpAssocEntry 7 */ + uint32_t state; /* sctpAssocEntry 8 */ + uint32_t in_streams; /* sctpAssocEntry 9 */ + uint32_t out_streams; /* sctpAssocEntry 10 */ + uint32_t max_nr_retrans; /* sctpAssocEntry 11 */ + uint32_t primary_process; /* sctpAssocEntry 12 */ + uint32_t T1_expireries; /* sctpAssocEntry 13 */ + uint32_t T2_expireries; /* sctpAssocEntry 14 */ + uint32_t retransmitted_tsns; /* sctpAssocEntry 15 */ + uint32_t total_sends; + uint32_t total_recvs; + uint32_t local_tag; + uint32_t remote_tag; + uint32_t initial_tsn; + uint32_t highest_tsn; + uint32_t cumulative_tsn; + uint32_t cumulative_tsn_ack; + uint32_t mtu; + uint32_t refcnt; + uint16_t local_port; /* sctpAssocEntry 3 */ + uint16_t remote_port; /* sctpAssocEntry 4 */ + struct sctp_timeval start_time; /* sctpAssocEntry 16 */ + struct sctp_timeval discontinuity_time; /* sctpAssocEntry 17 */ + uint32_t peers_rwnd; + sctp_assoc_t assoc_id; /* sctpAssocEntry 1 */ + uint32_t extra_padding[32]; /* future */ +}; + +struct xsctp_laddr { + union sctp_sockstore address; /* sctpAssocLocalAddrEntry 1/2 */ + uint32_t last; + struct sctp_timeval start_time; /* sctpAssocLocalAddrEntry 3 */ + uint32_t extra_padding[32]; /* future */ +}; + +struct xsctp_raddr { + union sctp_sockstore address; /* sctpAssocLocalRemEntry 1/2 */ + uint32_t last; + uint32_t rto; /* sctpAssocLocalRemEntry 5 */ + uint32_t max_path_rtx; /* sctpAssocLocalRemEntry 6 */ + uint32_t rtx; /* sctpAssocLocalRemEntry 7 */ + uint32_t error_counter; /* */ + uint32_t cwnd; /* */ + uint32_t flight_size; /* */ + uint32_t mtu; /* */ + uint8_t active; /* sctpAssocLocalRemEntry 3 */ + uint8_t confirmed; /* */ + uint8_t heartbeat_enabled; /* sctpAssocLocalRemEntry 4 */ + uint8_t potentially_failed; + struct sctp_timeval start_time; /* sctpAssocLocalRemEntry 8 */ + uint32_t rtt; + uint32_t heartbeat_interval; + uint32_t ssthresh; + uint16_t encaps_port; + uint16_t state; + uint32_t extra_padding[29]; /* future */ +}; + +#define SCTP_MAX_LOGGING_SIZE 30000 +#define SCTP_TRACE_PARAMS 6 /* This number MUST be even */ + +struct sctp_log_entry { + uint64_t timestamp; + uint32_t subsys; + uint32_t padding; + uint32_t params[SCTP_TRACE_PARAMS]; +}; + +struct sctp_log { + struct sctp_log_entry entry[SCTP_MAX_LOGGING_SIZE]; + uint32_t index; + uint32_t padding; +}; + +/* + * Kernel defined for sctp_send + */ +#if defined(_KERNEL) || defined(__Userspace__) +int +sctp_lower_sosend(struct socket *so, + struct sockaddr *addr, + struct uio *uio, + struct mbuf *top, + struct mbuf *control, + int flags, + struct sctp_sndrcvinfo *srcv +#if !defined(__Userspace__) +#if defined(__FreeBSD__) + ,struct thread *p +#elif defined(_WIN32) + , PKTHREAD p +#else + ,struct proc *p +#endif +#endif +); + +int +sctp_sorecvmsg(struct socket *so, + struct uio *uio, + struct mbuf **mp, + struct sockaddr *from, + int fromlen, + int *msg_flags, + struct sctp_sndrcvinfo *sinfo, + int filling_sinfo); +#endif + +/* + * API system calls + */ +#if !(defined(_KERNEL)) && !(defined(__Userspace__)) + +__BEGIN_DECLS +int sctp_peeloff(int, sctp_assoc_t); +int sctp_bindx(int, struct sockaddr *, int, int); +int sctp_connectx(int, const struct sockaddr *, int, sctp_assoc_t *); +int sctp_getaddrlen(sa_family_t); +int sctp_getpaddrs(int, sctp_assoc_t, struct sockaddr **); +void sctp_freepaddrs(struct sockaddr *); +int sctp_getladdrs(int, sctp_assoc_t, struct sockaddr **); +void sctp_freeladdrs(struct sockaddr *); +int sctp_opt_info(int, sctp_assoc_t, int, void *, socklen_t *); + +/* deprecated */ +ssize_t sctp_sendmsg(int, const void *, size_t, const struct sockaddr *, + socklen_t, uint32_t, uint32_t, uint16_t, uint32_t, uint32_t); + +/* deprecated */ +ssize_t sctp_send(int, const void *, size_t, + const struct sctp_sndrcvinfo *, int); + +/* deprecated */ +ssize_t sctp_sendx(int, const void *, size_t, struct sockaddr *, + int, struct sctp_sndrcvinfo *, int); + +/* deprecated */ +ssize_t sctp_sendmsgx(int sd, const void *, size_t, struct sockaddr *, + int, uint32_t, uint32_t, uint16_t, uint32_t, uint32_t); + +sctp_assoc_t sctp_getassocid(int, struct sockaddr *); + +/* deprecated */ +ssize_t sctp_recvmsg(int, void *, size_t, struct sockaddr *, socklen_t *, + struct sctp_sndrcvinfo *, int *); + +ssize_t sctp_sendv(int, const struct iovec *, int, struct sockaddr *, + int, void *, socklen_t, unsigned int, int); + +ssize_t sctp_recvv(int, const struct iovec *, int, struct sockaddr *, + socklen_t *, void *, socklen_t *, unsigned int *, int *); +__END_DECLS + +#endif /* !_KERNEL */ +#endif /* !__sctp_uio_h__ */ diff --git a/netwerk/sctp/src/netinet/sctp_userspace.c b/netwerk/sctp/src/netinet/sctp_userspace.c new file mode 100644 index 0000000000..71888901fe --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_userspace.c @@ -0,0 +1,425 @@ +/*- + * Copyright (c) 2011-2012 Irene Ruengeler + * Copyright (c) 2011-2012 Michael Tuexen + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + */ + + +#ifdef _WIN32 +#include <netinet/sctp_pcb.h> +#include <sys/timeb.h> +#include <iphlpapi.h> +#if !defined(__MINGW32__) +#pragma comment(lib, "iphlpapi.lib") +#endif +#endif +#include <netinet/sctp_os_userspace.h> +#if defined(__FreeBSD__) +#include <pthread_np.h> +#endif + +#if defined(__linux__) +#include <sys/prctl.h> +#endif + +#if defined(_WIN32) +/* Adapter to translate Unix thread start routines to Windows thread start + * routines. + */ +#if defined(__MINGW32__) +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wpedantic" +#endif +static DWORD WINAPI +sctp_create_thread_adapter(void *arg) { + start_routine_t start_routine = (start_routine_t)arg; + return start_routine(NULL) == NULL; +} + +int +sctp_userspace_thread_create(userland_thread_t *thread, start_routine_t start_routine) +{ + *thread = CreateThread(NULL, 0, sctp_create_thread_adapter, + (void *)start_routine, 0, NULL); + if (*thread == NULL) + return GetLastError(); + return 0; +} + +#if defined(__MINGW32__) +#pragma GCC diagnostic pop +#endif + +#else +int +sctp_userspace_thread_create(userland_thread_t *thread, start_routine_t start_routine) +{ + return pthread_create(thread, NULL, start_routine, NULL); +} +#endif + +void +sctp_userspace_set_threadname(const char *name) +{ +#if defined(__APPLE__) + pthread_setname_np(name); +#endif +#if defined(__linux__) + prctl(PR_SET_NAME, name); +#endif +#if defined(__FreeBSD__) + pthread_set_name_np(pthread_self(), name); +#endif +} + +#if !defined(_WIN32) && !defined(__native_client__) +int +sctp_userspace_get_mtu_from_ifn(uint32_t if_index) +{ +#if defined(INET) || defined(INET6) + struct ifreq ifr; + int fd; +#endif + int mtu; + + if (if_index == 0xffffffff) { + mtu = 1280; + } else { + mtu = 0; +#if defined(INET) || defined(INET6) + memset(&ifr, 0, sizeof(struct ifreq)); + if (if_indextoname(if_index, ifr.ifr_name) != NULL) { + /* TODO can I use the raw socket here and not have to open a new one with each query? */ +#if defined(INET) + if ((fd = socket(AF_INET, SOCK_DGRAM, 0)) >= 0) { +#else + if ((fd = socket(AF_INET6, SOCK_DGRAM, 0)) >= 0) { +#endif + if (ioctl(fd, SIOCGIFMTU, &ifr) >= 0) { + mtu = ifr.ifr_mtu; + } + close(fd); + } + } +#endif + } + return (mtu); +} +#endif + +#if defined(__native_client__) +int +sctp_userspace_get_mtu_from_ifn(uint32_t if_index) +{ + return 1280; +} +#endif + +#if defined(__APPLE__) || defined(__DragonFly__) || defined(__linux__) || defined(__native_client__) || defined(__NetBSD__) || defined(__QNX__) || defined(_WIN32) || defined(__Fuchsia__) || defined(__EMSCRIPTEN__) +int +timingsafe_bcmp(const void *b1, const void *b2, size_t n) +{ + const unsigned char *p1 = b1, *p2 = b2; + int ret = 0; + + for (; n > 0; n--) + ret |= *p1++ ^ *p2++; + return (ret != 0); +} +#endif + +#ifdef _WIN32 +int +sctp_userspace_get_mtu_from_ifn(uint32_t if_index) +{ +#if defined(INET) || defined(INET6) + PIP_ADAPTER_ADDRESSES pAdapterAddrs, pAdapt; + DWORD AdapterAddrsSize, Err; +#endif + int mtu; + + if (if_index == 0xffffffff) { + mtu = 1280; + } else { + mtu = 0; +#if defined(INET) || defined(INET6) + AdapterAddrsSize = 0; + pAdapterAddrs = NULL; + if ((Err = GetAdaptersAddresses(AF_UNSPEC, 0, NULL, NULL, &AdapterAddrsSize)) != 0) { + if ((Err != ERROR_BUFFER_OVERFLOW) && (Err != ERROR_INSUFFICIENT_BUFFER)) { + SCTPDBG(SCTP_DEBUG_USR, "GetAdaptersAddresses() sizing failed with error code %d, AdapterAddrsSize = %d\n", Err, AdapterAddrsSize); + mtu = -1; + goto cleanup; + } + } + if ((pAdapterAddrs = (PIP_ADAPTER_ADDRESSES) GlobalAlloc(GPTR, AdapterAddrsSize)) == NULL) { + SCTPDBG(SCTP_DEBUG_USR, "Memory allocation error!\n"); + mtu = -1; + goto cleanup; + } + if ((Err = GetAdaptersAddresses(AF_UNSPEC, 0, NULL, pAdapterAddrs, &AdapterAddrsSize)) != ERROR_SUCCESS) { + SCTPDBG(SCTP_DEBUG_USR, "GetAdaptersAddresses() failed with error code %d\n", Err); + mtu = -1; + goto cleanup; + } + for (pAdapt = pAdapterAddrs; pAdapt; pAdapt = pAdapt->Next) { + if (pAdapt->IfIndex == if_index) { + mtu = pAdapt->Mtu; + break; + } + } + cleanup: + if (pAdapterAddrs != NULL) { + GlobalFree(pAdapterAddrs); + } +#endif + } + return (mtu); +} + +void +getwintimeofday(struct timeval *tv) +{ + FILETIME filetime; + ULARGE_INTEGER ularge; + + GetSystemTimeAsFileTime(&filetime); + ularge.LowPart = filetime.dwLowDateTime; + ularge.HighPart = filetime.dwHighDateTime; + /* Change base from Jan 1 1601 00:00:00 to Jan 1 1970 00:00:00 */ +#if defined(__MINGW32__) + ularge.QuadPart -= 116444736000000000ULL; +#else + ularge.QuadPart -= 116444736000000000UI64; +#endif + /* + * ularge.QuadPart is now the number of 100-nanosecond intervals + * since Jan 1 1970 00:00:00. + */ +#if defined(__MINGW32__) + tv->tv_sec = (long)(ularge.QuadPart / 10000000ULL); + tv->tv_usec = (long)((ularge.QuadPart % 10000000ULL) / 10ULL); +#else + tv->tv_sec = (long)(ularge.QuadPart / 10000000UI64); + tv->tv_usec = (long)((ularge.QuadPart % 10000000UI64) / 10UI64); +#endif +} + +int +win_if_nametoindex(const char *ifname) +{ + IP_ADAPTER_ADDRESSES *addresses, *addr; + ULONG status, size; + int index = 0; + + if (!ifname) { + return 0; + } + + size = 0; + status = GetAdaptersAddresses(AF_UNSPEC, 0, NULL, NULL, &size); + if (status != ERROR_BUFFER_OVERFLOW) { + return 0; + } + addresses = malloc(size); + status = GetAdaptersAddresses(AF_UNSPEC, 0, NULL, addresses, &size); + if (status == ERROR_SUCCESS) { + for (addr = addresses; addr; addr = addr->Next) { + if (addr->AdapterName && !strcmp(ifname, addr->AdapterName)) { + index = addr->IfIndex; + break; + } + } + } + + free(addresses); + return index; +} + +#if WINVER < 0x0600 +/* These functions are written based on the code at + * http://www.cs.wustl.edu/~schmidt/win32-cv-1.html + * Therefore, for the rest of the file the following applies: + * + * + * Copyright and Licensing Information for ACE(TM), TAO(TM), CIAO(TM), + * DAnCE(TM), and CoSMIC(TM) + * + * [1]ACE(TM), [2]TAO(TM), [3]CIAO(TM), DAnCE(TM), and [4]CoSMIC(TM) + * (henceforth referred to as "DOC software") are copyrighted by + * [5]Douglas C. Schmidt and his [6]research group at [7]Washington + * University, [8]University of California, Irvine, and [9]Vanderbilt + * University, Copyright (c) 1993-2012, all rights reserved. Since DOC + * software is open-source, freely available software, you are free to + * use, modify, copy, and distribute--perpetually and irrevocably--the + * DOC software source code and object code produced from the source, as + * well as copy and distribute modified versions of this software. You + * must, however, include this copyright statement along with any code + * built using DOC software that you release. No copyright statement + * needs to be provided if you just ship binary executables of your + * software products. + * + * You can use DOC software in commercial and/or binary software releases + * and are under no obligation to redistribute any of your source code + * that is built using DOC software. Note, however, that you may not + * misappropriate the DOC software code, such as copyrighting it yourself + * or claiming authorship of the DOC software code, in a way that will + * prevent DOC software from being distributed freely using an + * open-source development model. You needn't inform anyone that you're + * using DOC software in your software, though we encourage you to let + * [10]us know so we can promote your project in the [11]DOC software + * success stories. + * + * The [12]ACE, [13]TAO, [14]CIAO, [15]DAnCE, and [16]CoSMIC web sites + * are maintained by the [17]DOC Group at the [18]Institute for Software + * Integrated Systems (ISIS) and the [19]Center for Distributed Object + * Computing of Washington University, St. Louis for the development of + * open-source software as part of the open-source software community. + * Submissions are provided by the submitter ``as is'' with no warranties + * whatsoever, including any warranty of merchantability, noninfringement + * of third party intellectual property, or fitness for any particular + * purpose. In no event shall the submitter be liable for any direct, + * indirect, special, exemplary, punitive, or consequential damages, + * including without limitation, lost profits, even if advised of the + * possibility of such damages. Likewise, DOC software is provided as is + * with no warranties of any kind, including the warranties of design, + * merchantability, and fitness for a particular purpose, + * noninfringement, or arising from a course of dealing, usage or trade + * practice. Washington University, UC Irvine, Vanderbilt University, + * their employees, and students shall have no liability with respect to + * the infringement of copyrights, trade secrets or any patents by DOC + * software or any part thereof. Moreover, in no event will Washington + * University, UC Irvine, or Vanderbilt University, their employees, or + * students be liable for any lost revenue or profits or other special, + * indirect and consequential damages. + * + * DOC software is provided with no support and without any obligation on + * the part of Washington University, UC Irvine, Vanderbilt University, + * their employees, or students to assist in its use, correction, + * modification, or enhancement. A [20]number of companies around the + * world provide commercial support for DOC software, however. DOC + * software is Y2K-compliant, as long as the underlying OS platform is + * Y2K-compliant. Likewise, DOC software is compliant with the new US + * daylight savings rule passed by Congress as "The Energy Policy Act of + * 2005," which established new daylight savings times (DST) rules for + * the United States that expand DST as of March 2007. Since DOC software + * obtains time/date and calendaring information from operating systems + * users will not be affected by the new DST rules as long as they + * upgrade their operating systems accordingly. + * + * The names ACE(TM), TAO(TM), CIAO(TM), DAnCE(TM), CoSMIC(TM), + * Washington University, UC Irvine, and Vanderbilt University, may not + * be used to endorse or promote products or services derived from this + * source without express written permission from Washington University, + * UC Irvine, or Vanderbilt University. This license grants no permission + * to call products or services derived from this source ACE(TM), + * TAO(TM), CIAO(TM), DAnCE(TM), or CoSMIC(TM), nor does it grant + * permission for the name Washington University, UC Irvine, or + * Vanderbilt University to appear in their names. + * + * If you have any suggestions, additions, comments, or questions, please + * let [21]me know. + * + * [22]Douglas C. Schmidt + * + * References + * + * 1. http://www.cs.wustl.edu/~schmidt/ACE.html + * 2. http://www.cs.wustl.edu/~schmidt/TAO.html + * 3. http://www.dre.vanderbilt.edu/CIAO/ + * 4. http://www.dre.vanderbilt.edu/cosmic/ + * 5. http://www.dre.vanderbilt.edu/~schmidt/ + * 6. http://www.cs.wustl.edu/~schmidt/ACE-members.html + * 7. http://www.wustl.edu/ + * 8. http://www.uci.edu/ + * 9. http://www.vanderbilt.edu/ + * 10. mailto:doc_group@cs.wustl.edu + * 11. http://www.cs.wustl.edu/~schmidt/ACE-users.html + * 12. http://www.cs.wustl.edu/~schmidt/ACE.html + * 13. http://www.cs.wustl.edu/~schmidt/TAO.html + * 14. http://www.dre.vanderbilt.edu/CIAO/ + * 15. http://www.dre.vanderbilt.edu/~schmidt/DOC_ROOT/DAnCE/ + * 16. http://www.dre.vanderbilt.edu/cosmic/ + * 17. http://www.dre.vanderbilt.edu/ + * 18. http://www.isis.vanderbilt.edu/ + * 19. http://www.cs.wustl.edu/~schmidt/doc-center.html + * 20. http://www.cs.wustl.edu/~schmidt/commercial-support.html + * 21. mailto:d.schmidt@vanderbilt.edu + * 22. http://www.dre.vanderbilt.edu/~schmidt/ + * 23. http://www.cs.wustl.edu/ACE.html + */ + +void +InitializeXPConditionVariable(userland_cond_t *cv) +{ + cv->waiters_count = 0; + InitializeCriticalSection(&(cv->waiters_count_lock)); + cv->events_[C_SIGNAL] = CreateEvent (NULL, FALSE, FALSE, NULL); + cv->events_[C_BROADCAST] = CreateEvent (NULL, TRUE, FALSE, NULL); +} + +void +DeleteXPConditionVariable(userland_cond_t *cv) +{ + CloseHandle(cv->events_[C_BROADCAST]); + CloseHandle(cv->events_[C_SIGNAL]); + DeleteCriticalSection(&(cv->waiters_count_lock)); +} + +int +SleepXPConditionVariable(userland_cond_t *cv, userland_mutex_t *mtx) +{ + int result, last_waiter; + + EnterCriticalSection(&cv->waiters_count_lock); + cv->waiters_count++; + LeaveCriticalSection(&cv->waiters_count_lock); + LeaveCriticalSection (mtx); + result = WaitForMultipleObjects(2, cv->events_, FALSE, INFINITE); + if (result==-1) { + result = GetLastError(); + } + EnterCriticalSection(&cv->waiters_count_lock); + cv->waiters_count--; + last_waiter = result == (C_SIGNAL + C_BROADCAST && (cv->waiters_count == 0)); + LeaveCriticalSection(&cv->waiters_count_lock); + if (last_waiter) + ResetEvent(cv->events_[C_BROADCAST]); + EnterCriticalSection (mtx); + return result; +} + +void +WakeAllXPConditionVariable(userland_cond_t *cv) +{ + int have_waiters; + EnterCriticalSection(&cv->waiters_count_lock); + have_waiters = cv->waiters_count > 0; + LeaveCriticalSection(&cv->waiters_count_lock); + if (have_waiters) + SetEvent (cv->events_[C_BROADCAST]); +} +#endif +#endif diff --git a/netwerk/sctp/src/netinet/sctp_usrreq.c b/netwerk/sctp/src/netinet/sctp_usrreq.c new file mode 100644 index 0000000000..254aea3669 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_usrreq.c @@ -0,0 +1,9045 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/proc.h> +#endif +#include <netinet/sctp_pcb.h> +#include <netinet/sctp_header.h> +#include <netinet/sctp_var.h> +#ifdef INET6 +#include <netinet6/sctp6_var.h> +#endif +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_uio.h> +#include <netinet/sctp_asconf.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_indata.h> +#include <netinet/sctp_timer.h> +#include <netinet/sctp_auth.h> +#include <netinet/sctp_bsd_addr.h> +#if defined(__Userspace__) +#include <netinet/sctp_callout.h> +#else +#include <netinet/udp.h> +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/eventhandler.h> +#endif +#if defined(HAVE_SCTP_PEELOFF_SOCKOPT) +#include <netinet/sctp_peeloff.h> +#endif /* HAVE_SCTP_PEELOFF_SOCKOPT */ + +extern const struct sctp_cc_functions sctp_cc_functions[]; +extern const struct sctp_ss_functions sctp_ss_functions[]; + +#if defined(__Userspace__) +void +sctp_init(uint16_t port, + int (*conn_output)(void *addr, void *buffer, size_t length, uint8_t tos, uint8_t set_df), + void (*debug_printf)(const char *format, ...), int start_threads) +#elif defined(__APPLE__) && (!defined(APPLE_LEOPARD) && !defined(APPLE_SNOWLEOPARD) &&!defined(APPLE_LION) && !defined(APPLE_MOUNTAINLION)) +void +sctp_init(struct protosw *pp SCTP_UNUSED, struct domain *dp SCTP_UNUSED) +#elif defined(__FreeBSD__) +static void +sctp_init(void *arg SCTP_UNUSED) +#else +void +sctp_init(void) +#endif +{ +#if !defined(__Userspace__) + u_long sb_max_adj; + +#else + init_random(); +#endif + /* Initialize and modify the sysctled variables */ + sctp_init_sysctls(); +#if defined(__Userspace__) + SCTP_BASE_SYSCTL(sctp_udp_tunneling_port) = port; +#else +#if defined(__APPLE__) && !defined(__Userspace__) + sb_max_adj = (u_long)((u_quad_t) (sb_max) * MCLBYTES / (MSIZE + MCLBYTES)); + SCTP_BASE_SYSCTL(sctp_sendspace) = sb_max_adj; +#else + if ((nmbclusters / 8) > SCTP_ASOC_MAX_CHUNKS_ON_QUEUE) + SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue) = (nmbclusters / 8); + /* + * Allow a user to take no more than 1/2 the number of clusters or + * the SB_MAX, whichever is smaller, for the send window. + */ + sb_max_adj = (u_long)((u_quad_t) (SB_MAX) * MCLBYTES / (MSIZE + MCLBYTES)); + SCTP_BASE_SYSCTL(sctp_sendspace) = min(sb_max_adj, + (((uint32_t)nmbclusters / 2) * MCLBYTES)); +#endif + /* + * Now for the recv window, should we take the same amount? or + * should I do 1/2 the SB_MAX instead in the SB_MAX min above. For + * now I will just copy. + */ + SCTP_BASE_SYSCTL(sctp_recvspace) = SCTP_BASE_SYSCTL(sctp_sendspace); +#endif + SCTP_BASE_VAR(first_time) = 0; + SCTP_BASE_VAR(sctp_pcb_initialized) = 0; +#if defined(__Userspace__) +#if !defined(_WIN32) +#if defined(INET) || defined(INET6) + SCTP_BASE_VAR(userspace_route) = -1; +#endif +#endif +#ifdef INET + SCTP_BASE_VAR(userspace_rawsctp) = -1; + SCTP_BASE_VAR(userspace_udpsctp) = -1; +#endif +#ifdef INET6 + SCTP_BASE_VAR(userspace_rawsctp6) = -1; + SCTP_BASE_VAR(userspace_udpsctp6) = -1; +#endif + SCTP_BASE_VAR(timer_thread_should_exit) = 0; + SCTP_BASE_VAR(conn_output) = conn_output; + SCTP_BASE_VAR(debug_printf) = debug_printf; + SCTP_BASE_VAR(crc32c_offloaded) = 0; + SCTP_BASE_VAR(iterator_thread_started) = 0; + SCTP_BASE_VAR(timer_thread_started) = 0; +#endif +#if defined(__Userspace__) + sctp_pcb_init(start_threads); + if (start_threads) { + sctp_start_timer_thread(); + } +#else + sctp_pcb_init(); +#endif +#if defined(SCTP_PACKET_LOGGING) + SCTP_BASE_VAR(packet_log_writers) = 0; + SCTP_BASE_VAR(packet_log_end) = 0; + memset(&SCTP_BASE_VAR(packet_log_buffer), 0, SCTP_PACKET_LOG_SIZE); +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_BASE_VAR(sctp_main_timer_ticks) = 0; + sctp_start_main_timer(); + timeout(sctp_delayed_startup, NULL, 1); +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_BASE_VAR(eh_tag) = EVENTHANDLER_REGISTER(rt_addrmsg, + sctp_addr_change_event_handler, NULL, EVENTHANDLER_PRI_FIRST); +#endif +} +#if defined(__FreeBSD__) && !defined(__Userspace__) +VNET_SYSINIT(sctp_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, sctp_init, NULL); + +#ifdef VIMAGE +static void +sctp_finish(void *unused __unused) +{ + EVENTHANDLER_DEREGISTER(rt_addrmsg, SCTP_BASE_VAR(eh_tag)); + sctp_pcb_finish(); +} +VNET_SYSUNINIT(sctp, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, sctp_finish, NULL); +#endif +#else +void +sctp_finish(void) +{ +#if defined(__APPLE__) && !defined(__Userspace__) + untimeout(sctp_delayed_startup, NULL); + sctp_over_udp_stop(); + sctp_address_monitor_stop(); + sctp_stop_main_timer(); +#endif +#if defined(__Userspace__) +#if defined(INET) || defined(INET6) + recv_thread_destroy(); +#endif + sctp_stop_timer_thread(); +#endif + sctp_pcb_finish(); +#if defined(_WIN32) && !defined(__Userspace__) + sctp_finish_sysctls(); +#endif +#if defined(__Userspace__) + finish_random(); +#endif +} +#endif + +void +sctp_pathmtu_adjustment(struct sctp_tcb *stcb, uint32_t mtu, bool resend) +{ + struct sctp_association *asoc; + struct sctp_tmit_chunk *chk; + uint32_t overhead; + + asoc = &stcb->asoc; + KASSERT(mtu < asoc->smallest_mtu, + ("Currently only reducing association MTU %u supported (MTU %u)", + asoc->smallest_mtu, mtu)); + asoc->smallest_mtu = mtu; + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + overhead = SCTP_MIN_OVERHEAD; + } else { +#if defined(__Userspace__) + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_CONN) { + overhead = sizeof(struct sctphdr); + } else { + overhead = SCTP_MIN_V4_OVERHEAD; + } +#else + overhead = SCTP_MIN_V4_OVERHEAD; +#endif + } + if (asoc->idata_supported) { + if (sctp_auth_is_required_chunk(SCTP_IDATA, asoc->peer_auth_chunks)) { + overhead += sctp_get_auth_chunk_len(asoc->peer_hmac_id); + } + } else { + if (sctp_auth_is_required_chunk(SCTP_DATA, asoc->peer_auth_chunks)) { + overhead += sctp_get_auth_chunk_len(asoc->peer_hmac_id); + } + } + KASSERT(overhead % 4 == 0, + ("overhead (%u) not a multiple of 4", overhead)); + TAILQ_FOREACH(chk, &asoc->send_queue, sctp_next) { + if (((uint32_t)chk->send_size + overhead) > mtu) { + chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; + } + } + TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { + if (((uint32_t)chk->send_size + overhead) > mtu) { + chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; + if (resend && chk->sent < SCTP_DATAGRAM_RESEND) { + /* + * If requested, mark the chunk for immediate + * resend, since we sent it being too big. + */ + sctp_flight_size_decrease(chk); + sctp_total_flight_decrease(stcb, chk); + chk->sent = SCTP_DATAGRAM_RESEND; + sctp_ucount_incr(asoc->sent_queue_retran_cnt); + chk->rec.data.doing_fast_retransmit = 0; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { + sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_PMTU, + chk->whoTo->flight_size, + chk->book_size, + (uint32_t)(uintptr_t)chk->whoTo, + chk->rec.data.tsn); + } + /* Clear any time, so NO RTT is being done. */ + if (chk->do_rtt == 1) { + chk->do_rtt = 0; + chk->whoTo->rto_needed = 1; + } + } + } + } +} + +#ifdef INET +#if !defined(__Userspace__) +void +sctp_notify(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_nets *net, + uint8_t icmp_type, + uint8_t icmp_code, + uint16_t ip_len, + uint32_t next_mtu) +{ +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif + int timer_stopped; + + if (icmp_type != ICMP_UNREACH) { + /* We only care about unreachable */ + SCTP_TCB_UNLOCK(stcb); + return; + } + if ((icmp_code == ICMP_UNREACH_NET) || + (icmp_code == ICMP_UNREACH_HOST) || + (icmp_code == ICMP_UNREACH_NET_UNKNOWN) || + (icmp_code == ICMP_UNREACH_HOST_UNKNOWN) || + (icmp_code == ICMP_UNREACH_ISOLATED) || + (icmp_code == ICMP_UNREACH_NET_PROHIB) || + (icmp_code == ICMP_UNREACH_HOST_PROHIB) || +#if defined(__NetBSD__) + (icmp_code == ICMP_UNREACH_ADMIN_PROHIBIT)) { +#else + (icmp_code == ICMP_UNREACH_FILTER_PROHIB)) { +#endif + /* Mark the net unreachable. */ + if (net->dest_state & SCTP_ADDR_REACHABLE) { + /* OK, that destination is NOT reachable. */ + net->dest_state &= ~SCTP_ADDR_REACHABLE; + net->dest_state &= ~SCTP_ADDR_PF; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, + stcb, 0, + (void *)net, SCTP_SO_NOT_LOCKED); + } + SCTP_TCB_UNLOCK(stcb); + } else if ((icmp_code == ICMP_UNREACH_PROTOCOL) || + (icmp_code == ICMP_UNREACH_PORT)) { + /* Treat it like an ABORT. */ + sctp_abort_notification(stcb, true, false, 0, NULL, SCTP_SO_NOT_LOCKED); +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(inp); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_2); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); + /* SCTP_TCB_UNLOCK(stcb); MT: I think this is not needed.*/ +#endif + /* no need to unlock here, since the TCB is gone */ + } else if (icmp_code == ICMP_UNREACH_NEEDFRAG) { + if (net->dest_state & SCTP_ADDR_NO_PMTUD) { + SCTP_TCB_UNLOCK(stcb); + return; + } + /* Find the next (smaller) MTU */ + if (next_mtu == 0) { + /* + * Old type router that does not tell us what the next + * MTU is. + * Rats we will have to guess (in a educated fashion + * of course). + */ + next_mtu = sctp_get_prev_mtu(ip_len); + } + /* Stop the PMTU timer. */ + if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { + timer_stopped = 1; + sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_1); + } else { + timer_stopped = 0; + } + /* Update the path MTU. */ + if (net->port) { + next_mtu -= sizeof(struct udphdr); + } + if (net->mtu > next_mtu) { + net->mtu = next_mtu; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (net->port) { + sctp_hc_set_mtu(&net->ro._l_addr, inp->fibnum, next_mtu + sizeof(struct udphdr)); + } else { + sctp_hc_set_mtu(&net->ro._l_addr, inp->fibnum, next_mtu); + } +#endif + } + /* Update the association MTU */ + if (stcb->asoc.smallest_mtu > next_mtu) { + sctp_pathmtu_adjustment(stcb, next_mtu, true); + } + /* Finally, start the PMTU timer if it was running before. */ + if (timer_stopped) { + sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net); + } + SCTP_TCB_UNLOCK(stcb); + } else { + SCTP_TCB_UNLOCK(stcb); + } +} +#endif + +#if !defined(__Userspace__) +#if defined(__FreeBSD__) +void sctp_ctlinput(struct icmp *icmp) +{ + struct ip *inner_ip, *outer_ip; + struct sctphdr *sh; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; + struct sctp_nets *net; + struct sctp_init_chunk *ch; + struct sockaddr_in src, dst; + + if (icmp_errmap(icmp) == 0) + return; + + outer_ip = (struct ip *)((caddr_t)icmp - sizeof(struct ip)); + inner_ip = &icmp->icmp_ip; + sh = (struct sctphdr *)((caddr_t)inner_ip + (inner_ip->ip_hl << 2)); + memset(&src, 0, sizeof(struct sockaddr_in)); + src.sin_family = AF_INET; + src.sin_len = sizeof(struct sockaddr_in); + src.sin_port = sh->src_port; + src.sin_addr = inner_ip->ip_src; + memset(&dst, 0, sizeof(struct sockaddr_in)); + dst.sin_family = AF_INET; + dst.sin_len = sizeof(struct sockaddr_in); + dst.sin_port = sh->dest_port; + dst.sin_addr = inner_ip->ip_dst; + /* + * 'dst' holds the dest of the packet that failed to be sent. + * 'src' holds our local endpoint address. Thus we reverse + * the dst and the src in the lookup. + */ + inp = NULL; + net = NULL; + stcb = sctp_findassociation_addr_sa((struct sockaddr *)&dst, + (struct sockaddr *)&src, + &inp, &net, 1, + SCTP_DEFAULT_VRFID); + if ((stcb != NULL) && + (net != NULL) && + (inp != NULL)) { + /* Check the verification tag */ + if (ntohl(sh->v_tag) != 0) { + /* + * This must be the verification tag used for + * sending out packets. We don't consider + * packets reflecting the verification tag. + */ + if (ntohl(sh->v_tag) != stcb->asoc.peer_vtag) { + SCTP_TCB_UNLOCK(stcb); + return; + } + } else { + if (ntohs(outer_ip->ip_len) >= + sizeof(struct ip) + + 8 + (inner_ip->ip_hl << 2) + 20) { + /* + * In this case we can check if we + * got an INIT chunk and if the + * initiate tag matches. + */ + ch = (struct sctp_init_chunk *)(sh + 1); + if ((ch->ch.chunk_type != SCTP_INITIATION) || + (ntohl(ch->init.initiate_tag) != stcb->asoc.my_vtag)) { + SCTP_TCB_UNLOCK(stcb); + return; + } + } else { + SCTP_TCB_UNLOCK(stcb); + return; + } + } + sctp_notify(inp, stcb, net, + icmp->icmp_type, + icmp->icmp_code, + ntohs(inner_ip->ip_len), + (uint32_t)ntohs(icmp->icmp_nextmtu)); + } else { + if ((stcb == NULL) && (inp != NULL)) { + /* reduce ref-count */ + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + SCTP_INP_WUNLOCK(inp); + } + if (stcb) { + SCTP_TCB_UNLOCK(stcb); + } + } +} +#else +void +#if defined(__APPLE__) && !defined(APPLE_LEOPARD) && !defined(APPLE_SNOWLEOPARD) && !defined(APPLE_LION) && !defined(APPLE_MOUNTAINLION) && !defined(APPLE_ELCAPITAN) +sctp_ctlinput(int cmd, struct sockaddr *sa, void *vip, struct ifnet *ifp SCTP_UNUSED) +#else +sctp_ctlinput(int cmd, struct sockaddr *sa, void *vip) +#endif +{ + struct ip *inner_ip; + struct sctphdr *sh; + struct icmp *icmp; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; + struct sctp_nets *net; + struct sockaddr_in src, dst; + +#if !defined(__FreeBSD__) && !defined(__Userspace__) + if (sa->sa_family != AF_INET || + ((struct sockaddr_in *)sa)->sin_addr.s_addr == INADDR_ANY) { + return; + } +#endif + if (PRC_IS_REDIRECT(cmd)) { + vip = NULL; + } else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) { + return; + } + if (vip != NULL) { + inner_ip = (struct ip *)vip; + icmp = (struct icmp *)((caddr_t)inner_ip - + (sizeof(struct icmp) - sizeof(struct ip))); + sh = (struct sctphdr *)((caddr_t)inner_ip + (inner_ip->ip_hl << 2)); + memset(&src, 0, sizeof(struct sockaddr_in)); + src.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + src.sin_len = sizeof(struct sockaddr_in); +#endif + src.sin_port = sh->src_port; + src.sin_addr = inner_ip->ip_src; + memset(&dst, 0, sizeof(struct sockaddr_in)); + dst.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + dst.sin_len = sizeof(struct sockaddr_in); +#endif + dst.sin_port = sh->dest_port; + dst.sin_addr = inner_ip->ip_dst; + /* + * 'dst' holds the dest of the packet that failed to be sent. + * 'src' holds our local endpoint address. Thus we reverse + * the dst and the src in the lookup. + */ + inp = NULL; + net = NULL; + stcb = sctp_findassociation_addr_sa((struct sockaddr *)&dst, + (struct sockaddr *)&src, + &inp, &net, 1, + SCTP_DEFAULT_VRFID); + if ((stcb != NULL) && + (net != NULL) && + (inp != NULL)) { + /* Check the verification tag */ + if (ntohl(sh->v_tag) != 0) { + /* + * This must be the verification tag used for + * sending out packets. We don't consider + * packets reflecting the verification tag. + */ + if (ntohl(sh->v_tag) != stcb->asoc.peer_vtag) { + SCTP_TCB_UNLOCK(stcb); + return; + } + } else { + SCTP_TCB_UNLOCK(stcb); + return; + } + sctp_notify(inp, stcb, net, + icmp->icmp_type, + icmp->icmp_code, + inner_ip->ip_len, + (uint32_t)ntohs(icmp->icmp_nextmtu)); +#if defined(__Userspace__) + if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (stcb->sctp_socket != NULL)) { + struct socket *upcall_socket; + + upcall_socket = stcb->sctp_socket; + SOCK_LOCK(upcall_socket); + soref(upcall_socket); + SOCK_UNLOCK(upcall_socket); + if ((upcall_socket->so_upcall != NULL) && + (upcall_socket->so_error != 0)) { + (*upcall_socket->so_upcall)(upcall_socket, upcall_socket->so_upcallarg, M_NOWAIT); + } + ACCEPT_LOCK(); + SOCK_LOCK(upcall_socket); + sorele(upcall_socket); + } +#endif + } else { + if ((stcb == NULL) && (inp != NULL)) { + /* reduce ref-count */ + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + SCTP_INP_WUNLOCK(inp); + } + if (stcb) { + SCTP_TCB_UNLOCK(stcb); + } + } + } + return; +} +#endif +#endif +#endif + +#if defined(__FreeBSD__) && !defined(__Userspace__) +static int +sctp_getcred(SYSCTL_HANDLER_ARGS) +{ + struct xucred xuc; + struct sockaddr_in addrs[2]; + struct sctp_inpcb *inp; + struct sctp_nets *net; + struct sctp_tcb *stcb; + int error; + uint32_t vrf_id; + + /* FIX, for non-bsd is this right? */ + vrf_id = SCTP_DEFAULT_VRFID; + + error = priv_check(req->td, PRIV_NETINET_GETCRED); + + if (error) + return (error); + + error = SYSCTL_IN(req, addrs, sizeof(addrs)); + if (error) + return (error); + + stcb = sctp_findassociation_addr_sa(sintosa(&addrs[1]), + sintosa(&addrs[0]), + &inp, &net, 1, vrf_id); + if (stcb == NULL || inp == NULL || inp->sctp_socket == NULL) { + if ((inp != NULL) && (stcb == NULL)) { + /* reduce ref-count */ + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + goto cred_can_cont; + } + + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); + error = ENOENT; + goto out; + } + SCTP_TCB_UNLOCK(stcb); + /* We use the write lock here, only + * since in the error leg we need it. + * If we used RLOCK, then we would have + * to wlock/decr/unlock/rlock. Which + * in theory could create a hole. Better + * to use higher wlock. + */ + SCTP_INP_WLOCK(inp); + cred_can_cont: + error = cr_canseesocket(req->td->td_ucred, inp->sctp_socket); + if (error) { + SCTP_INP_WUNLOCK(inp); + goto out; + } + cru2x(inp->sctp_socket->so_cred, &xuc); + SCTP_INP_WUNLOCK(inp); + error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); +out: + return (error); +} + +SYSCTL_PROC(_net_inet_sctp, OID_AUTO, getcred, + CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_NEEDGIANT, + 0, 0, sctp_getcred, "S,ucred", + "Get the ucred of a SCTP connection"); +#endif + +#ifdef INET +#if defined(_WIN32) || defined(__Userspace__) +int +#elif defined(__FreeBSD__) +static void +#else +static int +#endif +sctp_abort(struct socket *so) +{ +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + return; +#else + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); +#endif + } + + SCTP_INP_WLOCK(inp); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 17); +#endif + if (((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0)) { + inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP; +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 16); +#endif + SCTP_INP_WUNLOCK(inp); + sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, + SCTP_CALLED_AFTER_CMPSET_OFCLOSE); + SOCK_LOCK(so); +#if defined(__FreeBSD__) && !defined(__Userspace__) + KASSERT(!SOLISTENING(so), + ("sctp_abort: called on listening socket %p", so)); +#endif + SCTP_SB_CLEAR(so->so_snd); + SCTP_SB_CLEAR(so->so_rcv); +#if defined(__APPLE__) && !defined(__Userspace__) + so->so_usecount--; +#else + /* Now null out the reference, we are completely detached. */ + so->so_pcb = NULL; +#endif + SOCK_UNLOCK(so); + } else { + SCTP_INP_WUNLOCK(inp); + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#else + return (0); +#endif +} + +#if defined(__Userspace__) +int +#else +static int +#endif +#if defined(__Userspace__) +sctp_attach(struct socket *so, int proto SCTP_UNUSED, uint32_t vrf_id) +#elif defined(__FreeBSD__) +sctp_attach(struct socket *so, int proto SCTP_UNUSED, struct thread *p SCTP_UNUSED) +#elif defined(_WIN32) +sctp_attach(struct socket *so, int proto SCTP_UNUSED, PKTHREAD p SCTP_UNUSED) +#else +sctp_attach(struct socket *so, int proto SCTP_UNUSED, struct proc *p SCTP_UNUSED) +#endif +{ + struct sctp_inpcb *inp; + struct inpcb *ip_inp; + int error; +#if !defined(__Userspace__) + uint32_t vrf_id = SCTP_DEFAULT_VRFID; +#endif + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp != NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { + error = SCTP_SORESERVE(so, SCTP_BASE_SYSCTL(sctp_sendspace), SCTP_BASE_SYSCTL(sctp_recvspace)); + if (error) { + return (error); + } + } + error = sctp_inpcb_alloc(so, vrf_id); + if (error) { + return (error); + } + inp = (struct sctp_inpcb *)so->so_pcb; + SCTP_INP_WLOCK(inp); + inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUND_V6; /* I'm not v6! */ + ip_inp = &inp->ip_inp.inp; + ip_inp->inp_vflag |= INP_IPV4; + ip_inp->inp_ip_ttl = MODULE_GLOBAL(ip_defttl); + SCTP_INP_WUNLOCK(inp); + return (0); +} + +#if defined(__Userspace__) +int +sctp_bind(struct socket *so, struct sockaddr *addr) { + void *p = NULL; +#elif defined(__FreeBSD__) +static int +sctp_bind(struct socket *so, struct sockaddr *addr, struct thread *p) +{ +#elif defined(__APPLE__) +static int +sctp_bind(struct socket *so, struct sockaddr *addr, struct proc *p) { +#elif defined(_WIN32) +static int +sctp_bind(struct socket *so, struct sockaddr *addr, PKTHREAD p) { +#else +static int +sctp_bind(struct socket *so, struct mbuf *nam, struct proc *p) +{ + struct sockaddr *addr = nam ? mtod(nam, struct sockaddr *): NULL; + +#endif + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + if (addr != NULL) { +#ifdef HAVE_SA_LEN + if ((addr->sa_family != AF_INET) || + (addr->sa_len != sizeof(struct sockaddr_in))) { +#else + if (addr->sa_family != AF_INET) { +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + } + return (sctp_inpcb_bind(so, addr, NULL, p)); +} + +#endif +#if defined(__Userspace__) + +int +sctpconn_attach(struct socket *so, int proto SCTP_UNUSED, uint32_t vrf_id) +{ + struct sctp_inpcb *inp; + struct inpcb *ip_inp; + int error; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp != NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { + error = SCTP_SORESERVE(so, SCTP_BASE_SYSCTL(sctp_sendspace), SCTP_BASE_SYSCTL(sctp_recvspace)); + if (error) { + return (error); + } + } + error = sctp_inpcb_alloc(so, vrf_id); + if (error) { + return (error); + } + inp = (struct sctp_inpcb *)so->so_pcb; + SCTP_INP_WLOCK(inp); + inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUND_V6; + inp->sctp_flags |= SCTP_PCB_FLAGS_BOUND_CONN; + ip_inp = &inp->ip_inp.inp; + ip_inp->inp_vflag |= INP_CONN; + ip_inp->inp_ip_ttl = MODULE_GLOBAL(ip_defttl); + SCTP_INP_WUNLOCK(inp); + return (0); +} + +int +sctpconn_bind(struct socket *so, struct sockaddr *addr) +{ + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + if (addr != NULL) { +#ifdef HAVE_SA_LEN + if ((addr->sa_family != AF_CONN) || + (addr->sa_len != sizeof(struct sockaddr_conn))) { +#else + if (addr->sa_family != AF_CONN) { +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + } + return (sctp_inpcb_bind(so, addr, NULL, NULL)); +} + +#endif +#if defined(__FreeBSD__) || defined(_WIN32) || defined(__Userspace__) +void +sctp_close(struct socket *so) +{ +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) + return; + + /* Inform all the lower layer assoc that we + * are done. + */ + SCTP_INP_WLOCK(inp); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 17); +#endif + if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) { + inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP; +#if defined(__Userspace__) + if (((so->so_options & SCTP_SO_LINGER) && (so->so_linger == 0)) || + (SCTP_SBAVAIL(&so->so_rcv) > 0)) { +#else + if (((so->so_options & SO_LINGER) && (so->so_linger == 0)) || + (SCTP_SBAVAIL(&so->so_rcv) > 0)) { +#endif +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 13); +#endif + SCTP_INP_WUNLOCK(inp); + sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, + SCTP_CALLED_AFTER_CMPSET_OFCLOSE); + } else { +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 14); +#endif + SCTP_INP_WUNLOCK(inp); + sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE, + SCTP_CALLED_AFTER_CMPSET_OFCLOSE); + } + /* The socket is now detached, no matter what + * the state of the SCTP association. + */ + SOCK_LOCK(so); +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (!SOLISTENING(so)) { + SCTP_SB_CLEAR(so->so_snd); + SCTP_SB_CLEAR(so->so_rcv); + } +#else + SCTP_SB_CLEAR(so->so_snd); + SCTP_SB_CLEAR(so->so_rcv); +#endif +#if !(defined(__APPLE__) && !defined(__Userspace__)) + /* Now null out the reference, we are completely detached. */ + so->so_pcb = NULL; +#endif + SOCK_UNLOCK(so); + } else { + SCTP_INP_WUNLOCK(inp); + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif +} +#else +int +sctp_detach(struct socket *so) +{ + struct sctp_inpcb *inp; + uint32_t flags; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + return; +#else + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); +#endif + } + sctp_must_try_again: + flags = inp->sctp_flags; +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 17); +#endif + if (((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (atomic_cmpset_int(&inp->sctp_flags, flags, (flags | SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP)))) { +#if defined(__Userspace__) + if (((so->so_options & SCTP_SO_LINGER) && (so->so_linger == 0)) || + (SCTP_SBAVAIL(&so->so_rcv) > 0)) { +#else + if (((so->so_options & SO_LINGER) && (so->so_linger == 0)) || + (SCTP_SBAVAIL(&so->so_rcv) > 0)) { +#endif +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 13); +#endif + sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, + SCTP_CALLED_AFTER_CMPSET_OFCLOSE); + } else { +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 13); +#endif + sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE, + SCTP_CALLED_AFTER_CMPSET_OFCLOSE); + } + /* The socket is now detached, no matter what + * the state of the SCTP association. + */ + SCTP_SB_CLEAR(so->so_snd); + /* same for the rcv ones, they are only + * here for the accounting/select. + */ + SCTP_SB_CLEAR(so->so_rcv); +#if !(defined(__APPLE__) && !defined(__Userspace__)) + /* Now disconnect */ + so->so_pcb = NULL; +#endif + } else { + flags = inp->sctp_flags; + if ((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) { + goto sctp_must_try_again; + } + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + return; +#else + return (0); +#endif +} +#endif + +#if defined(__Userspace__) +/* __Userspace__ is not calling sctp_sendm */ +#endif +#if !(defined(_WIN32) && !defined(__Userspace__)) +int +#if defined(__FreeBSD__) && !defined(__Userspace__) +sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, + struct mbuf *control, struct thread *p); + +#else +sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, + struct mbuf *control, struct proc *p); + +#endif +int +#if defined(__FreeBSD__) && !defined(__Userspace__) +sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, + struct mbuf *control, struct thread *p) +{ +#else +sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, + struct mbuf *control, struct proc *p) +{ +#endif + struct sctp_inpcb *inp; + int error; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + if (control) { + sctp_m_freem(control); + control = NULL; + } + SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + sctp_m_freem(m); + return (EINVAL); + } + /* Got to have an to address if we are NOT a connected socket */ + if ((addr == NULL) && + ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) || + (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE))) { + goto connected_type; + } + + error = 0; + if (addr == NULL) { + SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EDESTADDRREQ); + error = EDESTADDRREQ; + } else if (addr->sa_family != AF_INET) { + SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EAFNOSUPPORT); + error = EAFNOSUPPORT; +#if defined(HAVE_SA_LEN) + } else if (addr->sa_len != sizeof(struct sockaddr_in)) { + SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; +#endif + } + if (error != 0) { + sctp_m_freem(m); + if (control) { + sctp_m_freem(control); + control = NULL; + } + return (error); + } +connected_type: + /* now what about control */ + if (control) { + if (inp->control) { + sctp_m_freem(inp->control); + inp->control = NULL; + } + inp->control = control; + } + /* Place the data */ + if (inp->pkt) { + SCTP_BUF_NEXT(inp->pkt_last) = m; + inp->pkt_last = m; + } else { + inp->pkt_last = inp->pkt = m; + } + if ( +#if (defined(__FreeBSD__) || defined(__APPLE__)) && !defined(__Userspace__) + /* FreeBSD uses a flag passed */ + ((flags & PRUS_MORETOCOME) == 0) +#else + 1 /* Open BSD does not have any "more to come" + * indication */ +#endif + ) { + /* + * note with the current version this code will only be used + * by OpenBSD-- NetBSD, FreeBSD, and MacOS have methods for + * re-defining sosend to use the sctp_sosend. One can + * optionally switch back to this code (by changing back the + * definitions) but this is not advisable. This code is used + * by FreeBSD when sending a file with sendfile() though. + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif + int ret; + +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + ret = sctp_output(inp, inp->pkt, addr, inp->control, p, flags); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + inp->pkt = NULL; + inp->control = NULL; + return (ret); + } else { + return (0); + } +} +#endif + +int +sctp_disconnect(struct socket *so) +{ + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); + return (ENOTCONN); + } + SCTP_INP_RLOCK(inp); + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { + if (LIST_EMPTY(&inp->sctp_asoc_list)) { + /* No connection */ + SCTP_INP_RUNLOCK(inp); + return (0); + } else { +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif + struct sctp_association *asoc; + struct sctp_tcb *stcb; + + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb == NULL) { + SCTP_INP_RUNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + SCTP_TCB_LOCK(stcb); + asoc = &stcb->asoc; + if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + /* We are about to be freed, out of here */ + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_RUNLOCK(inp); + return (0); + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif +#if defined(__Userspace__) + if (((so->so_options & SCTP_SO_LINGER) && (so->so_linger == 0)) || + (SCTP_SBAVAIL(&so->so_rcv) > 0)) { +#else + if (((so->so_options & SO_LINGER) && (so->so_linger == 0)) || + (SCTP_SBAVAIL(&so->so_rcv) > 0)) { +#endif + if (SCTP_GET_STATE(stcb) != SCTP_STATE_COOKIE_WAIT) { + /* Left with Data unread */ + struct mbuf *op_err; + + op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); + sctp_send_abort_tcb(stcb, op_err, SCTP_SO_LOCKED); + SCTP_STAT_INCR_COUNTER32(sctps_aborted); + } + SCTP_INP_RUNLOCK(inp); + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_3); + /* No unlock tcb assoc is gone */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + return (0); + } + if (TAILQ_EMPTY(&asoc->send_queue) && + TAILQ_EMPTY(&asoc->sent_queue) && + (asoc->stream_queue_cnt == 0)) { + /* there is nothing queued to send, so done */ + if ((*asoc->ss_functions.sctp_ss_is_user_msgs_incomplete)(stcb, asoc)) { + goto abort_anyway; + } + if ((SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { + /* only send SHUTDOWN 1st time thru */ + struct sctp_nets *netp; + + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_SENT); + sctp_stop_timers_for_shutdown(stcb); + if (stcb->asoc.alternate) { + netp = stcb->asoc.alternate; + } else { + netp = stcb->asoc.primary_destination; + } + sctp_send_shutdown(stcb,netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, + stcb->sctp_ep, stcb, netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, + stcb->sctp_ep, stcb, NULL); + sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_LOCKED); + } + } else { + /* + * we still got (or just got) data to send, + * so set SHUTDOWN_PENDING + */ + /* + * XXX sockets draft says that SCTP_EOF + * should be sent with no data. currently, + * we will allow user data to be sent first + * and move to SHUTDOWN-PENDING + */ + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_SHUTDOWN_PENDING); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, NULL); + if ((*asoc->ss_functions.sctp_ss_is_user_msgs_incomplete)(stcb, asoc)) { + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_PARTIAL_MSG_LEFT); + } + if (TAILQ_EMPTY(&asoc->send_queue) && + TAILQ_EMPTY(&asoc->sent_queue) && + (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { + struct mbuf *op_err; + abort_anyway: + op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_USRREQ + SCTP_LOC_4; + sctp_send_abort_tcb(stcb, op_err, SCTP_SO_LOCKED); + SCTP_STAT_INCR_COUNTER32(sctps_aborted); + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + SCTP_INP_RUNLOCK(inp); + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_5); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + return (0); + } else { + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED); + } + } + soisdisconnecting(so); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_RUNLOCK(inp); + return (0); + } + /* not reached */ + } else { + /* UDP model does not support this */ + SCTP_INP_RUNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); + return (EOPNOTSUPP); + } +} + +#if defined(__FreeBSD__) || defined(_WIN32) || defined(__Userspace__) +int +sctp_flush(struct socket *so, int how) +{ + /* + * We will just clear out the values and let + * subsequent close clear out the data, if any. + * Note if the user did a shutdown(SHUT_RD) they + * will not be able to read the data, the socket + * will block that from happening. + */ + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + SCTP_INP_RLOCK(inp); + /* For the 1 to many model this does nothing */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) { + SCTP_INP_RUNLOCK(inp); + return (0); + } + SCTP_INP_RUNLOCK(inp); + if ((how == PRU_FLUSH_RD) || (how == PRU_FLUSH_RDWR)) { + /* First make sure the sb will be happy, we don't + * use these except maybe the count + */ + SCTP_INP_WLOCK(inp); + SCTP_INP_READ_LOCK(inp); + inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_CANT_READ; + SCTP_INP_READ_UNLOCK(inp); + SCTP_INP_WUNLOCK(inp); + SOCK_LOCK(so); +#if defined(__FreeBSD__) && !defined(__Userspace__) + KASSERT(!SOLISTENING(so), + ("sctp_flush: called on listening socket %p", so)); +#endif + SCTP_SB_CLEAR(so->so_rcv); + SOCK_UNLOCK(so); + } + if ((how == PRU_FLUSH_WR) || (how == PRU_FLUSH_RDWR)) { + /* First make sure the sb will be happy, we don't + * use these except maybe the count + */ + SOCK_LOCK(so); +#if defined(__FreeBSD__) && !defined(__Userspace__) + KASSERT(!SOLISTENING(so), + ("sctp_flush: called on listening socket %p", so)); +#endif + SCTP_SB_CLEAR(so->so_snd); + SOCK_UNLOCK(so); + } + return (0); +} +#endif + +int +sctp_shutdown(struct socket *so) +{ + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + SCTP_INP_RLOCK(inp); + /* For UDP model this is a invalid call */ + if (!((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL))) { + /* Restore the flags that the soshutdown took away. */ +#if (defined(__FreeBSD__) || defined(_WIN32)) && !defined(__Userspace__) + SOCKBUF_LOCK(&so->so_rcv); + so->so_rcv.sb_state &= ~SBS_CANTRCVMORE; + SOCKBUF_UNLOCK(&so->so_rcv); +#else + SOCK_LOCK(so); + so->so_state &= ~SS_CANTRCVMORE; + SOCK_UNLOCK(so); +#endif + /* This proc will wakeup for read and do nothing (I hope) */ + SCTP_INP_RUNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); + return (EOPNOTSUPP); + } else { + /* + * Ok, if we reach here its the TCP model and it is either + * a SHUT_WR or SHUT_RDWR. + * This means we put the shutdown flag against it. + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif + struct sctp_tcb *stcb; + struct sctp_association *asoc; + struct sctp_nets *netp; + + if ((so->so_state & + (SS_ISCONNECTED|SS_ISCONNECTING|SS_ISDISCONNECTING)) == 0) { + SCTP_INP_RUNLOCK(inp); + return (ENOTCONN); + } + socantsendmore(so); + + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb == NULL) { + /* + * Ok, we hit the case that the shutdown call was + * made after an abort or something. Nothing to do + * now. + */ + SCTP_INP_RUNLOCK(inp); + return (0); + } + SCTP_TCB_LOCK(stcb); + asoc = &stcb->asoc; + if (asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) { + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_RUNLOCK(inp); + return (0); + } + if ((SCTP_GET_STATE(stcb) != SCTP_STATE_COOKIE_WAIT) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_COOKIE_ECHOED) && + (SCTP_GET_STATE(stcb) != SCTP_STATE_OPEN)) { + /* If we are not in or before ESTABLISHED, there is + * no protocol action required. + */ + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_RUNLOCK(inp); + return (0); + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + if (stcb->asoc.alternate) { + netp = stcb->asoc.alternate; + } else { + netp = stcb->asoc.primary_destination; + } + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) && + TAILQ_EMPTY(&asoc->send_queue) && + TAILQ_EMPTY(&asoc->sent_queue) && + (asoc->stream_queue_cnt == 0)) { + if ((*asoc->ss_functions.sctp_ss_is_user_msgs_incomplete)(stcb, asoc)) { + goto abort_anyway; + } + /* there is nothing queued to send, so I'm done... */ + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_SENT); + sctp_stop_timers_for_shutdown(stcb); + sctp_send_shutdown(stcb, netp); + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, + stcb->sctp_ep, stcb, netp); + } else { + /* + * We still got (or just got) data to send, so set + * SHUTDOWN_PENDING. + */ + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_SHUTDOWN_PENDING); + if ((*asoc->ss_functions.sctp_ss_is_user_msgs_incomplete)(stcb, asoc)) { + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_PARTIAL_MSG_LEFT); + } + if (TAILQ_EMPTY(&asoc->send_queue) && + TAILQ_EMPTY(&asoc->sent_queue) && + (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { + struct mbuf *op_err; + abort_anyway: + op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); + stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_USRREQ + SCTP_LOC_6; + SCTP_INP_RUNLOCK(inp); + sctp_abort_an_association(stcb->sctp_ep, stcb, + op_err, false, SCTP_SO_LOCKED); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + return (0); + } + } + sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, NULL); + /* XXX: Why do this in the case where we have still data queued? */ + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED); + SCTP_TCB_UNLOCK(stcb); + SCTP_INP_RUNLOCK(inp); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + return (0); + } +} + +/* + * copies a "user" presentable address and removes embedded scope, etc. + * returns 0 on success, 1 on error + */ +static uint32_t +sctp_fill_user_address(struct sockaddr *dst, struct sockaddr *src) +{ +#ifdef INET6 +#if defined(SCTP_EMBEDDED_V6_SCOPE) + struct sockaddr_in6 lsa6; + + src = (struct sockaddr *)sctp_recover_scope((struct sockaddr_in6 *)src, + &lsa6); +#endif +#endif +#ifdef HAVE_SA_LEN + memcpy(dst, src, src->sa_len); +#else + switch (src->sa_family) { +#ifdef INET + case AF_INET: + memcpy(dst, src, sizeof(struct sockaddr_in)); + break; +#endif +#ifdef INET6 + case AF_INET6: + memcpy(dst, src, sizeof(struct sockaddr_in6)); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + memcpy(dst, src, sizeof(struct sockaddr_conn)); + break; +#endif + default: + /* TSNH */ + break; + } +#endif + return (0); +} + +static size_t +sctp_fill_up_addresses_vrf(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + size_t limit, + struct sockaddr *addr, + uint32_t vrf_id) +{ + struct sctp_ifn *sctp_ifn; + struct sctp_ifa *sctp_ifa; + size_t actual; + int loopback_scope; +#if defined(INET) + int ipv4_local_scope, ipv4_addr_legal; +#endif +#if defined(INET6) + int local_scope, site_scope, ipv6_addr_legal; +#endif +#if defined(__Userspace__) + int conn_addr_legal; +#endif + struct sctp_vrf *vrf; + + SCTP_IPI_ADDR_LOCK_ASSERT(); + actual = 0; + if (limit == 0) + return (actual); + + if (stcb) { + /* Turn on all the appropriate scope */ + loopback_scope = stcb->asoc.scope.loopback_scope; +#if defined(INET) + ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope; + ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal; +#endif +#if defined(INET6) + local_scope = stcb->asoc.scope.local_scope; + site_scope = stcb->asoc.scope.site_scope; + ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal; +#endif +#if defined(__Userspace__) + conn_addr_legal = stcb->asoc.scope.conn_addr_legal; +#endif + } else { + /* Use generic values for endpoints. */ + loopback_scope = 1; +#if defined(INET) + ipv4_local_scope = 1; +#endif +#if defined(INET6) + local_scope = 1; + site_scope = 1; +#endif + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { +#if defined(INET6) + ipv6_addr_legal = 1; +#endif +#if defined(INET) + if (SCTP_IPV6_V6ONLY(inp)) { + ipv4_addr_legal = 0; + } else { + ipv4_addr_legal = 1; + } +#endif +#if defined(__Userspace__) + conn_addr_legal = 0; +#endif + } else { +#if defined(INET6) + ipv6_addr_legal = 0; +#endif +#if defined(__Userspace__) + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_CONN) { + conn_addr_legal = 1; +#if defined(INET) + ipv4_addr_legal = 0; +#endif + } else { + conn_addr_legal = 0; +#if defined(INET) + ipv4_addr_legal = 1; +#endif + } +#else +#if defined(INET) + ipv4_addr_legal = 1; +#endif +#endif + } + } + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) { + return (0); + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { + if ((loopback_scope == 0) && + SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { + /* Skip loopback if loopback_scope not set */ + continue; + } + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { + if (stcb) { + /* + * For the BOUND-ALL case, the list + * associated with a TCB is Always + * considered a reverse list.. i.e. + * it lists addresses that are NOT + * part of the association. If this + * is one of those we must skip it. + */ + if (sctp_is_addr_restricted(stcb, + sctp_ifa)) { + continue; + } + } + switch (sctp_ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: + if (ipv4_addr_legal) { + struct sockaddr_in *sin; + + sin = &sctp_ifa->address.sin; + if (sin->sin_addr.s_addr == 0) { + /* + * we skip unspecified + * addresses + */ + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sin->sin_addr) != 0) { + continue; + } +#endif + if ((ipv4_local_scope == 0) && + (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { + continue; + } +#ifdef INET6 + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) { + if (actual + sizeof(struct sockaddr_in6) > limit) { + return (actual); + } + in6_sin_2_v4mapsin6(sin, (struct sockaddr_in6 *)addr); + ((struct sockaddr_in6 *)addr)->sin6_port = inp->sctp_lport; + addr = (struct sockaddr *)((caddr_t)addr + sizeof(struct sockaddr_in6)); + actual += sizeof(struct sockaddr_in6); + } else { +#endif + if (actual + sizeof(struct sockaddr_in) > limit) { + return (actual); + } + memcpy(addr, sin, sizeof(struct sockaddr_in)); + ((struct sockaddr_in *)addr)->sin_port = inp->sctp_lport; + addr = (struct sockaddr *)((caddr_t)addr + sizeof(struct sockaddr_in)); + actual += sizeof(struct sockaddr_in); +#ifdef INET6 + } +#endif + } else { + continue; + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (ipv6_addr_legal) { + struct sockaddr_in6 *sin6; + +#if defined(SCTP_EMBEDDED_V6_SCOPE) && !defined(SCTP_KAME) + struct sockaddr_in6 lsa6; +#endif + sin6 = &sctp_ifa->address.sin6; + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + /* + * we skip unspecified + * addresses + */ + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sin6->sin6_addr) != 0) { + continue; + } +#endif + if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { + if (local_scope == 0) + continue; +#if defined(SCTP_EMBEDDED_V6_SCOPE) + if (sin6->sin6_scope_id == 0) { +#ifdef SCTP_KAME + if (sa6_recoverscope(sin6) != 0) + /* + * bad link + * local + * address + */ + continue; +#else + lsa6 = *sin6; + if (in6_recoverscope(&lsa6, + &lsa6.sin6_addr, + NULL)) + /* + * bad link + * local + * address + */ + continue; + sin6 = &lsa6; +#endif /* SCTP_KAME */ + } +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + } + if ((site_scope == 0) && + (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { + continue; + } + if (actual + sizeof(struct sockaddr_in6) > limit) { + return (actual); + } + memcpy(addr, sin6, sizeof(struct sockaddr_in6)); + ((struct sockaddr_in6 *)addr)->sin6_port = inp->sctp_lport; + addr = (struct sockaddr *)((caddr_t)addr + sizeof(struct sockaddr_in6)); + actual += sizeof(struct sockaddr_in6); + } else { + continue; + } + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + if (conn_addr_legal) { + if (actual + sizeof(struct sockaddr_conn) > limit) { + return (actual); + } + memcpy(addr, &sctp_ifa->address.sconn, sizeof(struct sockaddr_conn)); + ((struct sockaddr_conn *)addr)->sconn_port = inp->sctp_lport; + addr = (struct sockaddr *)((caddr_t)addr + sizeof(struct sockaddr_conn)); + actual += sizeof(struct sockaddr_conn); + } else { + continue; + } +#endif + default: + /* TSNH */ + break; + } + } + } + } else { + struct sctp_laddr *laddr; + size_t sa_len; + + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (stcb) { + if (sctp_is_addr_restricted(stcb, laddr->ifa)) { + continue; + } + } +#ifdef HAVE_SA_LEN + sa_len = laddr->ifa->address.sa.sa_len; +#else + switch (laddr->ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: + sa_len = sizeof(struct sockaddr_in); + break; +#endif +#ifdef INET6 + case AF_INET6: + sa_len = sizeof(struct sockaddr_in6); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + sa_len = sizeof(struct sockaddr_conn); + break; +#endif + default: + /* TSNH */ + sa_len = 0; + break; + } +#endif + if (actual + sa_len > limit) { + return (actual); + } + if (sctp_fill_user_address(addr, &laddr->ifa->address.sa)) + continue; + switch (laddr->ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: + ((struct sockaddr_in *)addr)->sin_port = inp->sctp_lport; + break; +#endif +#ifdef INET6 + case AF_INET6: + ((struct sockaddr_in6 *)addr)->sin6_port = inp->sctp_lport; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + ((struct sockaddr_conn *)addr)->sconn_port = inp->sctp_lport; + break; +#endif + default: + /* TSNH */ + break; + } + addr = (struct sockaddr *)((caddr_t)addr + sa_len); + actual += sa_len; + } + } + return (actual); +} + +static size_t +sctp_fill_up_addresses(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + size_t limit, + struct sockaddr *addr) +{ + size_t size; +#ifdef SCTP_MVRF + uint32_t id; +#endif + + SCTP_IPI_ADDR_RLOCK(); +#ifdef SCTP_MVRF +/* + * FIX ME: ?? this WILL report duplicate addresses if they appear + * in more than one VRF. + */ + /* fill up addresses for all VRFs on the endpoint */ + size = 0; + for (id = 0; (id < inp->num_vrfs) && (size < limit); id++) { + size += sctp_fill_up_addresses_vrf(inp, stcb, limit, addr, + inp->m_vrf_ids[id]); + addr = (struct sockaddr *)((caddr_t)addr + size); + } +#else + /* fill up addresses for the endpoint's default vrf */ + size = sctp_fill_up_addresses_vrf(inp, stcb, limit, addr, + inp->def_vrf_id); +#endif + SCTP_IPI_ADDR_RUNLOCK(); + return (size); +} + +static size_t +sctp_max_size_addresses_vrf(struct sctp_inpcb *inp, uint32_t vrf_id) +{ + struct sctp_vrf *vrf; + size_t size; + + /* + * In both sub-set bound an bound_all cases we return the size of + * the maximum number of addresses that you could get. In reality + * the sub-set bound may have an exclusion list for a given TCB or + * in the bound-all case a TCB may NOT include the loopback or other + * addresses as well. + */ + SCTP_IPI_ADDR_LOCK_ASSERT(); + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) { + return (0); + } + size = 0; + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + struct sctp_ifn *sctp_ifn; + struct sctp_ifa *sctp_ifa; + + LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { + /* Count them if they are the right type */ + switch (sctp_ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: +#ifdef INET6 + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) + size += sizeof(struct sockaddr_in6); + else + size += sizeof(struct sockaddr_in); +#else + size += sizeof(struct sockaddr_in); +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: + size += sizeof(struct sockaddr_in6); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + size += sizeof(struct sockaddr_conn); + break; +#endif + default: + break; + } + } + } + } else { + struct sctp_laddr *laddr; + + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + switch (laddr->ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: +#ifdef INET6 + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) + size += sizeof(struct sockaddr_in6); + else + size += sizeof(struct sockaddr_in); +#else + size += sizeof(struct sockaddr_in); +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: + size += sizeof(struct sockaddr_in6); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + size += sizeof(struct sockaddr_conn); + break; +#endif + default: + break; + } + } + } + return (size); +} + +static size_t +sctp_max_size_addresses(struct sctp_inpcb *inp) +{ + size_t size; +#ifdef SCTP_MVRF + int id; +#endif + + SCTP_IPI_ADDR_RLOCK(); +#ifdef SCTP_MVRF +/* + * FIX ME: ?? this WILL count duplicate addresses if they appear + * in more than one VRF. + */ + /* Maximum size of all addresses for all VRFs on the endpoint */ + size = 0; + for (id = 0; id < inp->num_vrfs; id++) { + size += sctp_max_size_addresses_vrf(inp, inp->m_vrf_ids[id]); + } +#else + /* Maximum size of all addresses for the endpoint's default VRF */ + size = sctp_max_size_addresses_vrf(inp, inp->def_vrf_id); +#endif + SCTP_IPI_ADDR_RUNLOCK(); + return (size); +} + +static int +sctp_do_connect_x(struct socket *so, struct sctp_inpcb *inp, void *optval, + size_t optsize, void *p, int delay) +{ + int error; + int creat_lock_on = 0; + struct sctp_tcb *stcb = NULL; + struct sockaddr *sa; + unsigned int num_v6 = 0, num_v4 = 0, *totaddrp, totaddr; + uint32_t vrf_id; + sctp_assoc_t *a_id; + + SCTPDBG(SCTP_DEBUG_PCB1, "Connectx called\n"); + + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && + (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) { + /* We are already connected AND the TCP model */ + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE); + return (EADDRINUSE); + } + + if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) && + (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE))) { + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + + if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { + SCTP_INP_RLOCK(inp); + stcb = LIST_FIRST(&inp->sctp_asoc_list); + SCTP_INP_RUNLOCK(inp); + } + if (stcb) { + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); + return (EALREADY); + } + SCTP_INP_INCR_REF(inp); + SCTP_ASOC_CREATE_LOCK(inp); + creat_lock_on = 1; + if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EFAULT); + error = EFAULT; + goto out_now; + } + totaddrp = (unsigned int *)optval; + totaddr = *totaddrp; + sa = (struct sockaddr *)(totaddrp + 1); + error = sctp_connectx_helper_find(inp, sa, totaddr, &num_v4, &num_v6, (unsigned int)(optsize - sizeof(int))); + if (error != 0) { + /* Already have or am bring up an association */ + SCTP_ASOC_CREATE_UNLOCK(inp); + creat_lock_on = 0; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + goto out_now; + } +#ifdef INET6 + if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) && + (num_v6 > 0)) { + error = EINVAL; + goto out_now; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + (num_v4 > 0)) { + if (SCTP_IPV6_V6ONLY(inp)) { + /* + * if IPV6_V6ONLY flag, ignore connections destined + * to a v4 addr or v4-mapped addr + */ + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_now; + } + } +#endif /* INET6 */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { + /* Bind a ephemeral port */ + error = sctp_inpcb_bind(so, NULL, NULL, p); + if (error) { + goto out_now; + } + } + + /* FIX ME: do we want to pass in a vrf on the connect call? */ + vrf_id = inp->def_vrf_id; + + /* We are GOOD to go */ + stcb = sctp_aloc_assoc_connected(inp, sa, &error, 0, 0, vrf_id, + inp->sctp_ep.pre_open_stream_count, + inp->sctp_ep.port, +#if defined(__FreeBSD__) && !defined(__Userspace__) + (struct thread *)p, +#elif defined(_WIN32) && !defined(__Userspace__) + (PKTHREAD)p, +#else + (struct proc *)p, +#endif + SCTP_INITIALIZE_AUTH_PARAMS); + if (stcb == NULL) { + /* Gak! no memory */ + goto out_now; + } + SCTP_SET_STATE(stcb, SCTP_STATE_COOKIE_WAIT); + /* move to second address */ + switch (sa->sa_family) { +#ifdef INET + case AF_INET: + sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in)); + break; +#endif +#ifdef INET6 + case AF_INET6: + sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in6)); + break; +#endif + default: + break; + } + + error = 0; + sctp_connectx_helper_add(stcb, sa, (totaddr-1), &error); + /* Fill in the return id */ + if (error) { + goto out_now; + } + a_id = (sctp_assoc_t *)optval; + *a_id = sctp_get_associd(stcb); + + if (delay) { + /* doing delayed connection */ + stcb->asoc.delayed_connection = 1; + sctp_timer_start(SCTP_TIMER_TYPE_INIT, inp, stcb, stcb->asoc.primary_destination); + } else { + (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); + sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); + } + SCTP_TCB_UNLOCK(stcb); + out_now: + if (creat_lock_on) { + SCTP_ASOC_CREATE_UNLOCK(inp); + } + SCTP_INP_DECR_REF(inp); + return (error); +} + +#define SCTP_FIND_STCB(inp, stcb, assoc_id) { \ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||\ + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { \ + SCTP_INP_RLOCK(inp); \ + stcb = LIST_FIRST(&inp->sctp_asoc_list); \ + if (stcb) { \ + SCTP_TCB_LOCK(stcb); \ + } \ + SCTP_INP_RUNLOCK(inp); \ + } else if (assoc_id > SCTP_ALL_ASSOC) { \ + stcb = sctp_findassociation_ep_asocid(inp, assoc_id, 1); \ + if (stcb == NULL) { \ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); \ + error = ENOENT; \ + break; \ + } \ + } else { \ + stcb = NULL; \ + } \ +} + +#define SCTP_CHECK_AND_CAST(destp, srcp, type, size) {\ + if (size < sizeof(type)) { \ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); \ + error = EINVAL; \ + break; \ + } else { \ + destp = (type *)srcp; \ + } \ +} + +#if defined(__Userspace__) +int +#else +static int +#endif +sctp_getopt(struct socket *so, int optname, void *optval, size_t *optsize, + void *p) { + struct sctp_inpcb *inp = NULL; + int error, val = 0; + struct sctp_tcb *stcb = NULL; + + if (optval == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return EINVAL; + } + error = 0; + + switch (optname) { + case SCTP_NODELAY: + case SCTP_AUTOCLOSE: + case SCTP_EXPLICIT_EOR: + case SCTP_AUTO_ASCONF: + case SCTP_DISABLE_FRAGMENTS: + case SCTP_I_WANT_MAPPED_V4_ADDR: + case SCTP_USE_EXT_RCVINFO: + SCTP_INP_RLOCK(inp); + switch (optname) { + case SCTP_DISABLE_FRAGMENTS: + val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NO_FRAGMENT); + break; + case SCTP_I_WANT_MAPPED_V4_ADDR: + val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4); + break; + case SCTP_AUTO_ASCONF: + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + /* only valid for bound all sockets */ + val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto flags_out; + } + break; + case SCTP_EXPLICIT_EOR: + val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR); + break; + case SCTP_NODELAY: + val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NODELAY); + break; + case SCTP_USE_EXT_RCVINFO: + val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO); + break; + case SCTP_AUTOCLOSE: + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) + val = sctp_ticks_to_secs(inp->sctp_ep.auto_close_time); + else + val = 0; + break; + + default: + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); + error = ENOPROTOOPT; + } /* end switch (sopt->sopt_name) */ + if (*optsize < sizeof(val)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + flags_out: + SCTP_INP_RUNLOCK(inp); + if (error == 0) { + /* return the option value */ + *(int *)optval = val; + *optsize = sizeof(val); + } + break; + case SCTP_GET_PACKET_LOG: + { +#ifdef SCTP_PACKET_LOGGING + uint8_t *target; + int ret; + + SCTP_CHECK_AND_CAST(target, optval, uint8_t, *optsize); + ret = sctp_copy_out_packet_log(target , (int)*optsize); + *optsize = ret; +#else + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); + error = EOPNOTSUPP; +#endif + break; + } + case SCTP_REUSE_PORT: + { + uint32_t *value; + + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE)) { + /* Can't do this for a 1-m socket */ + error = EINVAL; + break; + } + SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); + *value = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE); + *optsize = sizeof(uint32_t); + break; + } + case SCTP_PARTIAL_DELIVERY_POINT: + { + uint32_t *value; + + SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); + *value = inp->partial_delivery_point; + *optsize = sizeof(uint32_t); + break; + } + case SCTP_FRAGMENT_INTERLEAVE: + { + uint32_t *value; + + SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE)) { + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS)) { + *value = SCTP_FRAG_LEVEL_2; + } else { + *value = SCTP_FRAG_LEVEL_1; + } + } else { + *value = SCTP_FRAG_LEVEL_0; + } + *optsize = sizeof(uint32_t); + break; + } + case SCTP_INTERLEAVING_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = stcb->asoc.idata_supported; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + if (inp->idata_supported) { + av->assoc_value = 1; + } else { + av->assoc_value = 0; + } + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_CMT_ON_OFF: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + if (stcb) { + av->assoc_value = stcb->asoc.sctp_cmt_on_off; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->sctp_cmt_on_off; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_PLUGGABLE_CC: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + if (stcb) { + av->assoc_value = stcb->asoc.congestion_control_module; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->sctp_ep.sctp_default_cc_module; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_CC_OPTION: + { + struct sctp_cc_option *cc_opt; + + SCTP_CHECK_AND_CAST(cc_opt, optval, struct sctp_cc_option, *optsize); + SCTP_FIND_STCB(inp, stcb, cc_opt->aid_value.assoc_id); + if (stcb == NULL) { + error = EINVAL; + } else { + if (stcb->asoc.cc_functions.sctp_cwnd_socket_option == NULL) { + error = ENOTSUP; + } else { + error = (*stcb->asoc.cc_functions.sctp_cwnd_socket_option)(stcb, 0, cc_opt); + *optsize = sizeof(struct sctp_cc_option); + } + SCTP_TCB_UNLOCK(stcb); + } + break; + } + case SCTP_PLUGGABLE_SS: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + if (stcb) { + av->assoc_value = stcb->asoc.stream_scheduling_module; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->sctp_ep.sctp_default_ss_module; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_SS_VALUE: + { + struct sctp_stream_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_stream_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + if (stcb) { + if ((av->stream_id >= stcb->asoc.streamoutcnt) || + (stcb->asoc.ss_functions.sctp_ss_get_value(stcb, &stcb->asoc, &stcb->asoc.strmout[av->stream_id], + &av->stream_value) < 0)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } else { + *optsize = sizeof(struct sctp_stream_value); + } + SCTP_TCB_UNLOCK(stcb); + } else { + /* Can't get stream value without association */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + break; + } + case SCTP_GET_ADDR_LEN: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + error = EINVAL; +#ifdef INET + if (av->assoc_value == AF_INET) { + av->assoc_value = sizeof(struct sockaddr_in); + error = 0; + } +#endif +#ifdef INET6 + if (av->assoc_value == AF_INET6) { + av->assoc_value = sizeof(struct sockaddr_in6); + error = 0; + } +#endif +#if defined(__Userspace__) + if (av->assoc_value == AF_CONN) { + av->assoc_value = sizeof(struct sockaddr_conn); + error = 0; + } +#endif + if (error) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + } else { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_GET_ASSOC_NUMBER: + { + uint32_t *value, cnt; + + SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); + SCTP_INP_RLOCK(inp); + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { + /* Can't do this for a 1-1 socket */ + error = EINVAL; + SCTP_INP_RUNLOCK(inp); + break; + } + cnt = 0; + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + cnt++; + } + SCTP_INP_RUNLOCK(inp); + *value = cnt; + *optsize = sizeof(uint32_t); + break; + } + case SCTP_GET_ASSOC_ID_LIST: + { + struct sctp_assoc_ids *ids; + uint32_t at; + size_t limit; + + SCTP_CHECK_AND_CAST(ids, optval, struct sctp_assoc_ids, *optsize); + SCTP_INP_RLOCK(inp); + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { + /* Can't do this for a 1-1 socket */ + error = EINVAL; + SCTP_INP_RUNLOCK(inp); + break; + } + at = 0; + limit = (*optsize - sizeof(uint32_t)) / sizeof(sctp_assoc_t); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + if (at < limit) { + ids->gaids_assoc_id[at++] = sctp_get_associd(stcb); + if (at == 0) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } else { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } + SCTP_INP_RUNLOCK(inp); + if (error == 0) { + ids->gaids_number_of_ids = at; + *optsize = ((at * sizeof(sctp_assoc_t)) + sizeof(uint32_t)); + } + break; + } + case SCTP_CONTEXT: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = stcb->asoc.context; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->sctp_context; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_VRF_ID: + { + uint32_t *default_vrfid; + + SCTP_CHECK_AND_CAST(default_vrfid, optval, uint32_t, *optsize); + *default_vrfid = inp->def_vrf_id; + *optsize = sizeof(uint32_t); + break; + } + case SCTP_GET_ASOC_VRF: + { + struct sctp_assoc_value *id; + + SCTP_CHECK_AND_CAST(id, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, id->assoc_id); + if (stcb == NULL) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + } else { + id->assoc_value = stcb->asoc.vrf_id; + SCTP_TCB_UNLOCK(stcb); + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_GET_VRF_IDS: + { +#ifdef SCTP_MVRF + int siz_needed; + uint32_t *vrf_ids; + + SCTP_CHECK_AND_CAST(vrf_ids, optval, uint32_t, *optsize); + siz_needed = inp->num_vrfs * sizeof(uint32_t); + if (*optsize < siz_needed) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + } else { + memcpy(vrf_ids, inp->m_vrf_ids, siz_needed); + *optsize = siz_needed; + } +#else + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); + error = EOPNOTSUPP; +#endif + break; + } + case SCTP_GET_NONCE_VALUES: + { + struct sctp_get_nonce_values *gnv; + + SCTP_CHECK_AND_CAST(gnv, optval, struct sctp_get_nonce_values, *optsize); + SCTP_FIND_STCB(inp, stcb, gnv->gn_assoc_id); + + if (stcb) { + gnv->gn_peers_tag = stcb->asoc.peer_vtag; + gnv->gn_local_tag = stcb->asoc.my_vtag; + SCTP_TCB_UNLOCK(stcb); + *optsize = sizeof(struct sctp_get_nonce_values); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); + error = ENOTCONN; + } + break; + } + case SCTP_DELAYED_SACK: + { + struct sctp_sack_info *sack; + + SCTP_CHECK_AND_CAST(sack, optval, struct sctp_sack_info, *optsize); + SCTP_FIND_STCB(inp, stcb, sack->sack_assoc_id); + if (stcb) { + sack->sack_delay = stcb->asoc.delayed_ack; + sack->sack_freq = stcb->asoc.sack_freq; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (sack->sack_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + sack->sack_delay = sctp_ticks_to_msecs(inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_RECV]); + sack->sack_freq = inp->sctp_ep.sctp_sack_freq; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_sack_info); + } + break; + } + case SCTP_GET_SNDBUF_USE: + { + struct sctp_sockstat *ss; + + SCTP_CHECK_AND_CAST(ss, optval, struct sctp_sockstat, *optsize); + SCTP_FIND_STCB(inp, stcb, ss->ss_assoc_id); + + if (stcb) { + ss->ss_total_sndbuf = stcb->asoc.total_output_queue_size; + ss->ss_total_recv_buf = (stcb->asoc.size_on_reasm_queue + + stcb->asoc.size_on_all_streams); + SCTP_TCB_UNLOCK(stcb); + *optsize = sizeof(struct sctp_sockstat); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); + error = ENOTCONN; + } + break; + } + case SCTP_MAX_BURST: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = stcb->asoc.max_burst; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->sctp_ep.max_burst; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_MAXSEG: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = stcb->asoc.sctp_frag_point; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->sctp_frag_point; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_GET_STAT_LOG: + error = sctp_fill_stat_log(optval, optsize); + break; + case SCTP_EVENTS: + { + struct sctp_event_subscribe *events; + + SCTP_CHECK_AND_CAST(events, optval, struct sctp_event_subscribe, *optsize); + memset(events, 0, sizeof(struct sctp_event_subscribe)); + SCTP_INP_RLOCK(inp); + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) + events->sctp_data_io_event = 1; + + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVASSOCEVNT)) + events->sctp_association_event = 1; + + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVPADDREVNT)) + events->sctp_address_event = 1; + + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVSENDFAILEVNT)) + events->sctp_send_failure_event = 1; + + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVPEERERR)) + events->sctp_peer_error_event = 1; + + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT)) + events->sctp_shutdown_event = 1; + + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT)) + events->sctp_partial_delivery_event = 1; + + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_ADAPTATIONEVNT)) + events->sctp_adaptation_layer_event = 1; + + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTHEVNT)) + events->sctp_authentication_event = 1; + + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_DRYEVNT)) + events->sctp_sender_dry_event = 1; + + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_STREAM_RESETEVNT)) + events->sctp_stream_reset_event = 1; + SCTP_INP_RUNLOCK(inp); + *optsize = sizeof(struct sctp_event_subscribe); + break; + } + case SCTP_ADAPTATION_LAYER: + { + uint32_t *value; + + SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); + + SCTP_INP_RLOCK(inp); + *value = inp->sctp_ep.adaptation_layer_indicator; + SCTP_INP_RUNLOCK(inp); + *optsize = sizeof(uint32_t); + break; + } + case SCTP_SET_INITIAL_DBG_SEQ: + { + uint32_t *value; + + SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); + SCTP_INP_RLOCK(inp); + *value = inp->sctp_ep.initial_sequence_debug; + SCTP_INP_RUNLOCK(inp); + *optsize = sizeof(uint32_t); + break; + } + case SCTP_GET_LOCAL_ADDR_SIZE: + { + uint32_t *value; + + SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); + SCTP_INP_RLOCK(inp); + *value = (uint32_t)sctp_max_size_addresses(inp); + SCTP_INP_RUNLOCK(inp); + *optsize = sizeof(uint32_t); + break; + } + case SCTP_GET_REMOTE_ADDR_SIZE: + { + uint32_t *value; + struct sctp_nets *net; + size_t size; + + SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); + /* FIXME MT: change to sctp_assoc_value? */ + SCTP_FIND_STCB(inp, stcb, (sctp_assoc_t)*value); + + if (stcb != NULL) { + size = 0; + /* Count the sizes */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: +#ifdef INET6 + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) { + size += sizeof(struct sockaddr_in6); + } else { + size += sizeof(struct sockaddr_in); + } +#else + size += sizeof(struct sockaddr_in); +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: + size += sizeof(struct sockaddr_in6); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + size += sizeof(struct sockaddr_conn); + break; +#endif + default: + break; + } + } + SCTP_TCB_UNLOCK(stcb); + *value = (uint32_t)size; + *optsize = sizeof(uint32_t); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((sctp_assoc_t)*value <= SCTP_ALL_ASSOC)) { + error = EINVAL; + } else { + error = ENOENT; + } + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + } + break; + } + case SCTP_GET_PEER_ADDRESSES: + /* + * Get the address information, an array is passed in to + * fill up we pack it. + */ + { + size_t cpsz, left; + struct sockaddr *addr; + struct sctp_nets *net; + struct sctp_getaddresses *saddr; + + SCTP_CHECK_AND_CAST(saddr, optval, struct sctp_getaddresses, *optsize); + SCTP_FIND_STCB(inp, stcb, saddr->sget_assoc_id); + + if (stcb != NULL) { + left = *optsize - offsetof(struct sctp_getaddresses, addr); + *optsize = offsetof(struct sctp_getaddresses, addr); + addr = &saddr->addr[0].sa; + + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: +#ifdef INET6 + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) { + cpsz = sizeof(struct sockaddr_in6); + } else { + cpsz = sizeof(struct sockaddr_in); + } +#else + cpsz = sizeof(struct sockaddr_in); +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: + cpsz = sizeof(struct sockaddr_in6); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + cpsz = sizeof(struct sockaddr_conn); + break; +#endif + default: + cpsz = 0; + break; + } + if (cpsz == 0) { + break; + } + if (left < cpsz) { + /* not enough room. */ + break; + } +#if defined(INET) && defined(INET6) + if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) && + (net->ro._l_addr.sa.sa_family == AF_INET)) { + /* Must map the address */ + in6_sin_2_v4mapsin6(&net->ro._l_addr.sin, + (struct sockaddr_in6 *)addr); + } else { + memcpy(addr, &net->ro._l_addr, cpsz); + } +#else + memcpy(addr, &net->ro._l_addr, cpsz); +#endif + ((struct sockaddr_in *)addr)->sin_port = stcb->rport; + + addr = (struct sockaddr *)((caddr_t)addr + cpsz); + left -= cpsz; + *optsize += cpsz; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (saddr->sget_assoc_id <= SCTP_ALL_ASSOC)) { + error = EINVAL; + } else { + error = ENOENT; + } + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + } + break; + } + case SCTP_GET_LOCAL_ADDRESSES: + { + size_t limit, actual; + struct sctp_getaddresses *saddr; + + SCTP_CHECK_AND_CAST(saddr, optval, struct sctp_getaddresses, *optsize); + SCTP_FIND_STCB(inp, stcb, saddr->sget_assoc_id); + + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((saddr->sget_assoc_id == SCTP_CURRENT_ASSOC) || + (saddr->sget_assoc_id == SCTP_ALL_ASSOC))) { + error = EINVAL; + } else { + limit = *optsize - offsetof(struct sctp_getaddresses, addr); + actual = sctp_fill_up_addresses(inp, stcb, limit, &saddr->addr[0].sa); + *optsize = offsetof(struct sctp_getaddresses, addr) + actual; + } + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + break; + } + case SCTP_PEER_ADDR_PARAMS: + { + struct sctp_paddrparams *paddrp; + struct sctp_nets *net; + struct sockaddr *addr; +#if defined(INET) && defined(INET6) + struct sockaddr_in sin_store; +#endif + + SCTP_CHECK_AND_CAST(paddrp, optval, struct sctp_paddrparams, *optsize); + SCTP_FIND_STCB(inp, stcb, paddrp->spp_assoc_id); + +#if defined(INET) && defined(INET6) + if (paddrp->spp_address.ss_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)&paddrp->spp_address; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + in6_sin6_2_sin(&sin_store, sin6); + addr = (struct sockaddr *)&sin_store; + } else { + addr = (struct sockaddr *)&paddrp->spp_address; + } + } else { + addr = (struct sockaddr *)&paddrp->spp_address; + } +#else + addr = (struct sockaddr *)&paddrp->spp_address; +#endif + if (stcb != NULL) { + net = sctp_findnet(stcb, addr); + } else { + /* We increment here since sctp_findassociation_ep_addr() wil + * do a decrement if it finds the stcb as long as the locked + * tcb (last argument) is NOT a TCB.. aka NULL. + */ + net = NULL; + SCTP_INP_INCR_REF(inp); + stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_DECR_REF(inp); + } + } + if ((stcb != NULL) && (net == NULL)) { +#ifdef INET + if (addr->sa_family == AF_INET) { + struct sockaddr_in *sin; + + sin = (struct sockaddr_in *)addr; + if (sin->sin_addr.s_addr != INADDR_ANY) { + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } else +#endif +#ifdef INET6 + if (addr->sa_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)addr; + if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } else +#endif +#if defined(__Userspace__) + if (addr->sa_family == AF_CONN) { + struct sockaddr_conn *sconn; + + sconn = (struct sockaddr_conn *)addr; + if (sconn->sconn_addr != NULL) { + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } else +#endif + { + error = EAFNOSUPPORT; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } + + if (stcb != NULL) { + /* Applies to the specific association */ + paddrp->spp_flags = 0; + if (net != NULL) { + paddrp->spp_hbinterval = net->heart_beat_delay; + paddrp->spp_pathmaxrxt = net->failure_threshold; + paddrp->spp_pathmtu = net->mtu; + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + paddrp->spp_pathmtu -= SCTP_MIN_V4_OVERHEAD; + break; +#endif +#ifdef INET6 + case AF_INET6: + paddrp->spp_pathmtu -= SCTP_MIN_OVERHEAD; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + paddrp->spp_pathmtu -= sizeof(struct sctphdr); + break; +#endif + default: + break; + } + /* get flags for HB */ + if (net->dest_state & SCTP_ADDR_NOHB) { + paddrp->spp_flags |= SPP_HB_DISABLE; + } else { + paddrp->spp_flags |= SPP_HB_ENABLE; + } + /* get flags for PMTU */ + if (net->dest_state & SCTP_ADDR_NO_PMTUD) { + paddrp->spp_flags |= SPP_PMTUD_DISABLE; + } else { + paddrp->spp_flags |= SPP_PMTUD_ENABLE; + } + if (net->dscp & 0x01) { + paddrp->spp_dscp = net->dscp & 0xfc; + paddrp->spp_flags |= SPP_DSCP; + } +#ifdef INET6 + if ((net->ro._l_addr.sa.sa_family == AF_INET6) && + (net->flowlabel & 0x80000000)) { + paddrp->spp_ipv6_flowlabel = net->flowlabel & 0x000fffff; + paddrp->spp_flags |= SPP_IPV6_FLOWLABEL; + } +#endif + } else { + /* + * No destination so return default + * value + */ + paddrp->spp_pathmaxrxt = stcb->asoc.def_net_failure; + paddrp->spp_pathmtu = stcb->asoc.default_mtu; + if (stcb->asoc.default_dscp & 0x01) { + paddrp->spp_dscp = stcb->asoc.default_dscp & 0xfc; + paddrp->spp_flags |= SPP_DSCP; + } +#ifdef INET6 + if (stcb->asoc.default_flowlabel & 0x80000000) { + paddrp->spp_ipv6_flowlabel = stcb->asoc.default_flowlabel & 0x000fffff; + paddrp->spp_flags |= SPP_IPV6_FLOWLABEL; + } +#endif + /* default settings should be these */ + if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) { + paddrp->spp_flags |= SPP_HB_DISABLE; + } else { + paddrp->spp_flags |= SPP_HB_ENABLE; + } + if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) { + paddrp->spp_flags |= SPP_PMTUD_DISABLE; + } else { + paddrp->spp_flags |= SPP_PMTUD_ENABLE; + } + paddrp->spp_hbinterval = stcb->asoc.heart_beat_delay; + } + paddrp->spp_assoc_id = sctp_get_associd(stcb); + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (paddrp->spp_assoc_id == SCTP_FUTURE_ASSOC))) { + /* Use endpoint defaults */ + SCTP_INP_RLOCK(inp); + paddrp->spp_pathmaxrxt = inp->sctp_ep.def_net_failure; + paddrp->spp_hbinterval = sctp_ticks_to_msecs(inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT]); + paddrp->spp_assoc_id = SCTP_FUTURE_ASSOC; + /* get inp's default */ + if (inp->sctp_ep.default_dscp & 0x01) { + paddrp->spp_dscp = inp->sctp_ep.default_dscp & 0xfc; + paddrp->spp_flags |= SPP_DSCP; + } +#ifdef INET6 + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + (inp->sctp_ep.default_flowlabel & 0x80000000)) { + paddrp->spp_ipv6_flowlabel = inp->sctp_ep.default_flowlabel & 0x000fffff; + paddrp->spp_flags |= SPP_IPV6_FLOWLABEL; + } +#endif + paddrp->spp_pathmtu = inp->sctp_ep.default_mtu; + + if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) { + paddrp->spp_flags |= SPP_HB_ENABLE; + } else { + paddrp->spp_flags |= SPP_HB_DISABLE; + } + if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) { + paddrp->spp_flags |= SPP_PMTUD_ENABLE; + } else { + paddrp->spp_flags |= SPP_PMTUD_DISABLE; + } + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_paddrparams); + } + break; + } + case SCTP_GET_PEER_ADDR_INFO: + { + struct sctp_paddrinfo *paddri; + struct sctp_nets *net; + struct sockaddr *addr; +#if defined(INET) && defined(INET6) + struct sockaddr_in sin_store; +#endif + + SCTP_CHECK_AND_CAST(paddri, optval, struct sctp_paddrinfo, *optsize); + SCTP_FIND_STCB(inp, stcb, paddri->spinfo_assoc_id); + +#if defined(INET) && defined(INET6) + if (paddri->spinfo_address.ss_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)&paddri->spinfo_address; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + in6_sin6_2_sin(&sin_store, sin6); + addr = (struct sockaddr *)&sin_store; + } else { + addr = (struct sockaddr *)&paddri->spinfo_address; + } + } else { + addr = (struct sockaddr *)&paddri->spinfo_address; + } +#else + addr = (struct sockaddr *)&paddri->spinfo_address; +#endif + if (stcb != NULL) { + net = sctp_findnet(stcb, addr); + } else { + /* We increment here since sctp_findassociation_ep_addr() wil + * do a decrement if it finds the stcb as long as the locked + * tcb (last argument) is NOT a TCB.. aka NULL. + */ + net = NULL; + SCTP_INP_INCR_REF(inp); + stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_DECR_REF(inp); + } + } + + if ((stcb != NULL) && (net != NULL)) { + if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { + /* It's unconfirmed */ + paddri->spinfo_state = SCTP_UNCONFIRMED; + } else if (net->dest_state & SCTP_ADDR_REACHABLE) { + /* It's active */ + paddri->spinfo_state = SCTP_ACTIVE; + } else { + /* It's inactive */ + paddri->spinfo_state = SCTP_INACTIVE; + } + paddri->spinfo_cwnd = net->cwnd; + paddri->spinfo_srtt = net->lastsa >> SCTP_RTT_SHIFT; + paddri->spinfo_rto = net->RTO; + paddri->spinfo_assoc_id = sctp_get_associd(stcb); + paddri->spinfo_mtu = net->mtu; + switch (addr->sa_family) { +#if defined(INET) + case AF_INET: + paddri->spinfo_mtu -= SCTP_MIN_V4_OVERHEAD; + break; +#endif +#if defined(INET6) + case AF_INET6: + paddri->spinfo_mtu -= SCTP_MIN_OVERHEAD; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + paddri->spinfo_mtu -= sizeof(struct sctphdr); + break; +#endif + default: + break; + } + SCTP_TCB_UNLOCK(stcb); + *optsize = sizeof(struct sctp_paddrinfo); + } else { + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); + error = ENOENT; + } + break; + } + case SCTP_PCB_STATUS: + { + struct sctp_pcbinfo *spcb; + + SCTP_CHECK_AND_CAST(spcb, optval, struct sctp_pcbinfo, *optsize); + sctp_fill_pcbinfo(spcb); + *optsize = sizeof(struct sctp_pcbinfo); + break; + } + case SCTP_STATUS: + { + struct sctp_nets *net; + struct sctp_status *sstat; + + SCTP_CHECK_AND_CAST(sstat, optval, struct sctp_status, *optsize); + SCTP_FIND_STCB(inp, stcb, sstat->sstat_assoc_id); + + if (stcb == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + sstat->sstat_state = sctp_map_assoc_state(stcb->asoc.state); + sstat->sstat_assoc_id = sctp_get_associd(stcb); + sstat->sstat_rwnd = stcb->asoc.peers_rwnd; + sstat->sstat_unackdata = stcb->asoc.sent_queue_cnt; + /* + * We can't include chunks that have been passed to + * the socket layer. Only things in queue. + */ + sstat->sstat_penddata = (stcb->asoc.cnt_on_reasm_queue + + stcb->asoc.cnt_on_all_streams); + sstat->sstat_instrms = stcb->asoc.streamincnt; + sstat->sstat_outstrms = stcb->asoc.streamoutcnt; + sstat->sstat_fragmentation_point = sctp_get_frag_point(stcb); + net = stcb->asoc.primary_destination; + if (net != NULL) { +#ifdef HAVE_SA_LEN + memcpy(&sstat->sstat_primary.spinfo_address, + &net->ro._l_addr, + ((struct sockaddr *)(&net->ro._l_addr))->sa_len); +#else + switch (stcb->asoc.primary_destination->ro._l_addr.sa.sa_family) { +#if defined(INET) + case AF_INET: + memcpy(&sstat->sstat_primary.spinfo_address, + &net->ro._l_addr, + sizeof(struct sockaddr_in)); + break; +#endif +#if defined(INET6) + case AF_INET6: + memcpy(&sstat->sstat_primary.spinfo_address, + &net->ro._l_addr, + sizeof(struct sockaddr_in6)); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + memcpy(&sstat->sstat_primary.spinfo_address, + &net->ro._l_addr, + sizeof(struct sockaddr_conn)); + break; +#endif + default: + break; + } +#endif + ((struct sockaddr_in *)&sstat->sstat_primary.spinfo_address)->sin_port = stcb->rport; + /* + * Again the user can get info from sctp_constants.h + * for what the state of the network is. + */ + if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { + /* It's unconfirmed */ + sstat->sstat_primary.spinfo_state = SCTP_UNCONFIRMED; + } else if (net->dest_state & SCTP_ADDR_REACHABLE) { + /* It's active */ + sstat->sstat_primary.spinfo_state = SCTP_ACTIVE; + } else { + /* It's inactive */ + sstat->sstat_primary.spinfo_state = SCTP_INACTIVE; + } + sstat->sstat_primary.spinfo_cwnd = net->cwnd; + sstat->sstat_primary.spinfo_srtt = net->lastsa >> SCTP_RTT_SHIFT; + sstat->sstat_primary.spinfo_rto = net->RTO; + sstat->sstat_primary.spinfo_mtu = net->mtu; + switch (stcb->asoc.primary_destination->ro._l_addr.sa.sa_family) { +#if defined(INET) + case AF_INET: + sstat->sstat_primary.spinfo_mtu -= SCTP_MIN_V4_OVERHEAD; + break; +#endif +#if defined(INET6) + case AF_INET6: + sstat->sstat_primary.spinfo_mtu -= SCTP_MIN_OVERHEAD; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + sstat->sstat_primary.spinfo_mtu -= sizeof(struct sctphdr); + break; +#endif + default: + break; + } + } else { + memset(&sstat->sstat_primary, 0, sizeof(struct sctp_paddrinfo)); + } + sstat->sstat_primary.spinfo_assoc_id = sctp_get_associd(stcb); + SCTP_TCB_UNLOCK(stcb); + *optsize = sizeof(struct sctp_status); + break; + } + case SCTP_RTOINFO: + { + struct sctp_rtoinfo *srto; + + SCTP_CHECK_AND_CAST(srto, optval, struct sctp_rtoinfo, *optsize); + SCTP_FIND_STCB(inp, stcb, srto->srto_assoc_id); + + if (stcb) { + srto->srto_initial = stcb->asoc.initial_rto; + srto->srto_max = stcb->asoc.maxrto; + srto->srto_min = stcb->asoc.minrto; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (srto->srto_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + srto->srto_initial = inp->sctp_ep.initial_rto; + srto->srto_max = inp->sctp_ep.sctp_maxrto; + srto->srto_min = inp->sctp_ep.sctp_minrto; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_rtoinfo); + } + break; + } + case SCTP_TIMEOUTS: + { + struct sctp_timeouts *stimo; + + SCTP_CHECK_AND_CAST(stimo, optval, struct sctp_timeouts, *optsize); + SCTP_FIND_STCB(inp, stcb, stimo->stimo_assoc_id); + + if (stcb) { + stimo->stimo_init= stcb->asoc.timoinit; + stimo->stimo_data= stcb->asoc.timodata; + stimo->stimo_sack= stcb->asoc.timosack; + stimo->stimo_shutdown= stcb->asoc.timoshutdown; + stimo->stimo_heartbeat= stcb->asoc.timoheartbeat; + stimo->stimo_cookie= stcb->asoc.timocookie; + stimo->stimo_shutdownack= stcb->asoc.timoshutdownack; + SCTP_TCB_UNLOCK(stcb); + *optsize = sizeof(struct sctp_timeouts); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + break; + } + case SCTP_ASSOCINFO: + { + struct sctp_assocparams *sasoc; + + SCTP_CHECK_AND_CAST(sasoc, optval, struct sctp_assocparams, *optsize); + SCTP_FIND_STCB(inp, stcb, sasoc->sasoc_assoc_id); + + if (stcb) { + sasoc->sasoc_cookie_life = sctp_ticks_to_msecs(stcb->asoc.cookie_life); + sasoc->sasoc_asocmaxrxt = stcb->asoc.max_send_times; + sasoc->sasoc_number_peer_destinations = stcb->asoc.numnets; + sasoc->sasoc_peer_rwnd = stcb->asoc.peers_rwnd; + sasoc->sasoc_local_rwnd = stcb->asoc.my_rwnd; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (sasoc->sasoc_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + sasoc->sasoc_cookie_life = sctp_ticks_to_msecs(inp->sctp_ep.def_cookie_life); + sasoc->sasoc_asocmaxrxt = inp->sctp_ep.max_send_times; + sasoc->sasoc_number_peer_destinations = 0; + sasoc->sasoc_peer_rwnd = 0; + sasoc->sasoc_local_rwnd = (uint32_t)sbspace(&inp->sctp_socket->so_rcv); + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assocparams); + } + break; + } + case SCTP_DEFAULT_SEND_PARAM: + { + struct sctp_sndrcvinfo *s_info; + + SCTP_CHECK_AND_CAST(s_info, optval, struct sctp_sndrcvinfo, *optsize); + SCTP_FIND_STCB(inp, stcb, s_info->sinfo_assoc_id); + + if (stcb) { + memcpy(s_info, &stcb->asoc.def_send, sizeof(stcb->asoc.def_send)); + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (s_info->sinfo_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + memcpy(s_info, &inp->def_send, sizeof(inp->def_send)); + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_sndrcvinfo); + } + break; + } + case SCTP_INITMSG: + { + struct sctp_initmsg *sinit; + + SCTP_CHECK_AND_CAST(sinit, optval, struct sctp_initmsg, *optsize); + SCTP_INP_RLOCK(inp); + sinit->sinit_num_ostreams = inp->sctp_ep.pre_open_stream_count; + sinit->sinit_max_instreams = inp->sctp_ep.max_open_streams_intome; + sinit->sinit_max_attempts = inp->sctp_ep.max_init_times; + sinit->sinit_max_init_timeo = inp->sctp_ep.initial_init_rto_max; + SCTP_INP_RUNLOCK(inp); + *optsize = sizeof(struct sctp_initmsg); + break; + } + case SCTP_PRIMARY_ADDR: + /* we allow a "get" operation on this */ + { + struct sctp_setprim *ssp; + + SCTP_CHECK_AND_CAST(ssp, optval, struct sctp_setprim, *optsize); + SCTP_FIND_STCB(inp, stcb, ssp->ssp_assoc_id); + + if (stcb) { + union sctp_sockstore *addr; + + addr = &stcb->asoc.primary_destination->ro._l_addr; + switch (addr->sa.sa_family) { +#ifdef INET + case AF_INET: +#ifdef INET6 + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) { + in6_sin_2_v4mapsin6(&addr->sin, + (struct sockaddr_in6 *)&ssp->ssp_addr); + } else { + memcpy(&ssp->ssp_addr, &addr->sin, sizeof(struct sockaddr_in)); + } +#else + memcpy(&ssp->ssp_addr, &addr->sin, sizeof(struct sockaddr_in)); +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: + memcpy(&ssp->ssp_addr, &addr->sin6, sizeof(struct sockaddr_in6)); + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + memcpy(&ssp->ssp_addr, &addr->sconn, sizeof(struct sockaddr_conn)); + break; +#endif + default: + break; + } + SCTP_TCB_UNLOCK(stcb); + *optsize = sizeof(struct sctp_setprim); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + break; + } + case SCTP_HMAC_IDENT: + { + struct sctp_hmacalgo *shmac; + sctp_hmaclist_t *hmaclist; + size_t size; + int i; + + SCTP_CHECK_AND_CAST(shmac, optval, struct sctp_hmacalgo, *optsize); + + SCTP_INP_RLOCK(inp); + hmaclist = inp->sctp_ep.local_hmacs; + if (hmaclist == NULL) { + /* no HMACs to return */ + *optsize = sizeof(*shmac); + SCTP_INP_RUNLOCK(inp); + break; + } + /* is there room for all of the hmac ids? */ + size = sizeof(*shmac) + (hmaclist->num_algo * + sizeof(shmac->shmac_idents[0])); + if (*optsize < size) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_INP_RUNLOCK(inp); + break; + } + /* copy in the list */ + shmac->shmac_number_of_idents = hmaclist->num_algo; + for (i = 0; i < hmaclist->num_algo; i++) { + shmac->shmac_idents[i] = hmaclist->hmac[i]; + } + SCTP_INP_RUNLOCK(inp); + *optsize = size; + break; + } + case SCTP_AUTH_ACTIVE_KEY: + { + struct sctp_authkeyid *scact; + + SCTP_CHECK_AND_CAST(scact, optval, struct sctp_authkeyid, *optsize); + SCTP_FIND_STCB(inp, stcb, scact->scact_assoc_id); + + if (stcb) { + /* get the active key on the assoc */ + scact->scact_keynumber = stcb->asoc.authinfo.active_keyid; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (scact->scact_assoc_id == SCTP_FUTURE_ASSOC))) { + /* get the endpoint active key */ + SCTP_INP_RLOCK(inp); + scact->scact_keynumber = inp->sctp_ep.default_keyid; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_authkeyid); + } + break; + } + case SCTP_LOCAL_AUTH_CHUNKS: + { + struct sctp_authchunks *sac; + sctp_auth_chklist_t *chklist = NULL; + size_t size = 0; + + SCTP_CHECK_AND_CAST(sac, optval, struct sctp_authchunks, *optsize); + SCTP_FIND_STCB(inp, stcb, sac->gauth_assoc_id); + + if (stcb) { + /* get off the assoc */ + chklist = stcb->asoc.local_auth_chunks; + /* is there enough space? */ + size = sctp_auth_get_chklist_size(chklist); + if (*optsize < (sizeof(struct sctp_authchunks) + size)) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + } else { + /* copy in the chunks */ + (void)sctp_serialize_auth_chunks(chklist, sac->gauth_chunks); + sac->gauth_number_of_chunks = (uint32_t)size; + *optsize = sizeof(struct sctp_authchunks) + size; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (sac->gauth_assoc_id == SCTP_FUTURE_ASSOC))) { + /* get off the endpoint */ + SCTP_INP_RLOCK(inp); + chklist = inp->sctp_ep.local_auth_chunks; + /* is there enough space? */ + size = sctp_auth_get_chklist_size(chklist); + if (*optsize < (sizeof(struct sctp_authchunks) + size)) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + } else { + /* copy in the chunks */ + (void)sctp_serialize_auth_chunks(chklist, sac->gauth_chunks); + sac->gauth_number_of_chunks = (uint32_t)size; + *optsize = sizeof(struct sctp_authchunks) + size; + } + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_PEER_AUTH_CHUNKS: + { + struct sctp_authchunks *sac; + sctp_auth_chklist_t *chklist = NULL; + size_t size = 0; + + SCTP_CHECK_AND_CAST(sac, optval, struct sctp_authchunks, *optsize); + SCTP_FIND_STCB(inp, stcb, sac->gauth_assoc_id); + + if (stcb) { + /* get off the assoc */ + chklist = stcb->asoc.peer_auth_chunks; + /* is there enough space? */ + size = sctp_auth_get_chklist_size(chklist); + if (*optsize < (sizeof(struct sctp_authchunks) + size)) { + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + } else { + /* copy in the chunks */ + (void)sctp_serialize_auth_chunks(chklist, sac->gauth_chunks); + sac->gauth_number_of_chunks = (uint32_t)size; + *optsize = sizeof(struct sctp_authchunks) + size; + } + SCTP_TCB_UNLOCK(stcb); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); + error = ENOENT; + } + break; + } +#if defined(HAVE_SCTP_PEELOFF_SOCKOPT) + case SCTP_PEELOFF: + { + struct sctp_peeloff_opt *peeloff; + + SCTP_CHECK_AND_CAST(peeloff, optval, struct sctp_peeloff_opt, *optsize); + /* do the peeloff */ + error = sctp_peeloff_option(p, peeloff); + if (error == 0) { + *optsize = sizeof(struct sctp_peeloff_opt); + } + } + break; +#endif /* HAVE_SCTP_PEELOFF_SOCKOPT */ + case SCTP_EVENT: + { + struct sctp_event *event; + uint32_t event_type; + + SCTP_CHECK_AND_CAST(event, optval, struct sctp_event, *optsize); + SCTP_FIND_STCB(inp, stcb, event->se_assoc_id); + + switch (event->se_type) { + case SCTP_ASSOC_CHANGE: + event_type = SCTP_PCB_FLAGS_RECVASSOCEVNT; + break; + case SCTP_PEER_ADDR_CHANGE: + event_type = SCTP_PCB_FLAGS_RECVPADDREVNT; + break; + case SCTP_REMOTE_ERROR: + event_type = SCTP_PCB_FLAGS_RECVPEERERR; + break; + case SCTP_SEND_FAILED: + event_type = SCTP_PCB_FLAGS_RECVSENDFAILEVNT; + break; + case SCTP_SHUTDOWN_EVENT: + event_type = SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT; + break; + case SCTP_ADAPTATION_INDICATION: + event_type = SCTP_PCB_FLAGS_ADAPTATIONEVNT; + break; + case SCTP_PARTIAL_DELIVERY_EVENT: + event_type = SCTP_PCB_FLAGS_PDAPIEVNT; + break; + case SCTP_AUTHENTICATION_EVENT: + event_type = SCTP_PCB_FLAGS_AUTHEVNT; + break; + case SCTP_STREAM_RESET_EVENT: + event_type = SCTP_PCB_FLAGS_STREAM_RESETEVNT; + break; + case SCTP_SENDER_DRY_EVENT: + event_type = SCTP_PCB_FLAGS_DRYEVNT; + break; + case SCTP_NOTIFICATIONS_STOPPED_EVENT: + event_type = 0; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTSUP); + error = ENOTSUP; + break; + case SCTP_ASSOC_RESET_EVENT: + event_type = SCTP_PCB_FLAGS_ASSOC_RESETEVNT; + break; + case SCTP_STREAM_CHANGE_EVENT: + event_type = SCTP_PCB_FLAGS_STREAM_CHANGEEVNT; + break; + case SCTP_SEND_FAILED_EVENT: + event_type = SCTP_PCB_FLAGS_RECVNSENDFAILEVNT; + break; + default: + event_type = 0; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + if (event_type > 0) { + if (stcb) { + event->se_on = sctp_stcb_is_feature_on(inp, stcb, event_type); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (event->se_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + event->se_on = sctp_is_feature_on(inp, event_type); + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + } + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + if (error == 0) { + *optsize = sizeof(struct sctp_event); + } + break; + } + case SCTP_RECVRCVINFO: + if (*optsize < sizeof(int)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } else { + SCTP_INP_RLOCK(inp); + *(int *)optval = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVRCVINFO); + SCTP_INP_RUNLOCK(inp); + *optsize = sizeof(int); + } + break; + case SCTP_RECVNXTINFO: + if (*optsize < sizeof(int)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } else { + SCTP_INP_RLOCK(inp); + *(int *)optval = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVNXTINFO); + SCTP_INP_RUNLOCK(inp); + *optsize = sizeof(int); + } + break; + case SCTP_DEFAULT_SNDINFO: + { + struct sctp_sndinfo *info; + + SCTP_CHECK_AND_CAST(info, optval, struct sctp_sndinfo, *optsize); + SCTP_FIND_STCB(inp, stcb, info->snd_assoc_id); + + if (stcb) { + info->snd_sid = stcb->asoc.def_send.sinfo_stream; + info->snd_flags = stcb->asoc.def_send.sinfo_flags; + info->snd_flags &= 0xfff0; + info->snd_ppid = stcb->asoc.def_send.sinfo_ppid; + info->snd_context = stcb->asoc.def_send.sinfo_context; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (info->snd_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + info->snd_sid = inp->def_send.sinfo_stream; + info->snd_flags = inp->def_send.sinfo_flags; + info->snd_flags &= 0xfff0; + info->snd_ppid = inp->def_send.sinfo_ppid; + info->snd_context = inp->def_send.sinfo_context; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_sndinfo); + } + break; + } + case SCTP_DEFAULT_PRINFO: + { + struct sctp_default_prinfo *info; + + SCTP_CHECK_AND_CAST(info, optval, struct sctp_default_prinfo, *optsize); + SCTP_FIND_STCB(inp, stcb, info->pr_assoc_id); + + if (stcb) { + info->pr_policy = PR_SCTP_POLICY(stcb->asoc.def_send.sinfo_flags); + info->pr_value = stcb->asoc.def_send.sinfo_timetolive; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (info->pr_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + info->pr_policy = PR_SCTP_POLICY(inp->def_send.sinfo_flags); + info->pr_value = inp->def_send.sinfo_timetolive; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_default_prinfo); + } + break; + } + case SCTP_PEER_ADDR_THLDS: + { + struct sctp_paddrthlds *thlds; + struct sctp_nets *net; + struct sockaddr *addr; +#if defined(INET) && defined(INET6) + struct sockaddr_in sin_store; +#endif + + SCTP_CHECK_AND_CAST(thlds, optval, struct sctp_paddrthlds, *optsize); + SCTP_FIND_STCB(inp, stcb, thlds->spt_assoc_id); + +#if defined(INET) && defined(INET6) + if (thlds->spt_address.ss_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)&thlds->spt_address; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + in6_sin6_2_sin(&sin_store, sin6); + addr = (struct sockaddr *)&sin_store; + } else { + addr = (struct sockaddr *)&thlds->spt_address; + } + } else { + addr = (struct sockaddr *)&thlds->spt_address; + } +#else + addr = (struct sockaddr *)&thlds->spt_address; +#endif + if (stcb != NULL) { + net = sctp_findnet(stcb, addr); + } else { + /* We increment here since sctp_findassociation_ep_addr() wil + * do a decrement if it finds the stcb as long as the locked + * tcb (last argument) is NOT a TCB.. aka NULL. + */ + net = NULL; + SCTP_INP_INCR_REF(inp); + stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_DECR_REF(inp); + } + } + if ((stcb != NULL) && (net == NULL)) { +#ifdef INET + if (addr->sa_family == AF_INET) { + struct sockaddr_in *sin; + + sin = (struct sockaddr_in *)addr; + if (sin->sin_addr.s_addr != INADDR_ANY) { + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } else +#endif +#ifdef INET6 + if (addr->sa_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)addr; + if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } else +#endif +#if defined(__Userspace__) + if (addr->sa_family == AF_CONN) { + struct sockaddr_conn *sconn; + + sconn = (struct sockaddr_conn *)addr; + if (sconn->sconn_addr != NULL) { + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } else +#endif + { + error = EAFNOSUPPORT; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } + + if (stcb != NULL) { + if (net != NULL) { + thlds->spt_pathmaxrxt = net->failure_threshold; + thlds->spt_pathpfthld = net->pf_threshold; + thlds->spt_pathcpthld = 0xffff; + } else { + thlds->spt_pathmaxrxt = stcb->asoc.def_net_failure; + thlds->spt_pathpfthld = stcb->asoc.def_net_pf_threshold; + thlds->spt_pathcpthld = 0xffff; + } + thlds->spt_assoc_id = sctp_get_associd(stcb); + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (thlds->spt_assoc_id == SCTP_FUTURE_ASSOC))) { + /* Use endpoint defaults */ + SCTP_INP_RLOCK(inp); + thlds->spt_pathmaxrxt = inp->sctp_ep.def_net_failure; + thlds->spt_pathpfthld = inp->sctp_ep.def_net_pf_threshold; + thlds->spt_pathcpthld = 0xffff; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_paddrthlds); + } + break; + } + case SCTP_REMOTE_UDP_ENCAPS_PORT: + { + struct sctp_udpencaps *encaps; + struct sctp_nets *net; + struct sockaddr *addr; +#if defined(INET) && defined(INET6) + struct sockaddr_in sin_store; +#endif + + SCTP_CHECK_AND_CAST(encaps, optval, struct sctp_udpencaps, *optsize); + SCTP_FIND_STCB(inp, stcb, encaps->sue_assoc_id); + +#if defined(INET) && defined(INET6) + if (encaps->sue_address.ss_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)&encaps->sue_address; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + in6_sin6_2_sin(&sin_store, sin6); + addr = (struct sockaddr *)&sin_store; + } else { + addr = (struct sockaddr *)&encaps->sue_address; + } + } else { + addr = (struct sockaddr *)&encaps->sue_address; + } +#else + addr = (struct sockaddr *)&encaps->sue_address; +#endif + if (stcb) { + net = sctp_findnet(stcb, addr); + } else { + /* We increment here since sctp_findassociation_ep_addr() wil + * do a decrement if it finds the stcb as long as the locked + * tcb (last argument) is NOT a TCB.. aka NULL. + */ + net = NULL; + SCTP_INP_INCR_REF(inp); + stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_DECR_REF(inp); + } + } + if ((stcb != NULL) && (net == NULL)) { +#ifdef INET + if (addr->sa_family == AF_INET) { + struct sockaddr_in *sin; + + sin = (struct sockaddr_in *)addr; + if (sin->sin_addr.s_addr != INADDR_ANY) { + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } else +#endif +#ifdef INET6 + if (addr->sa_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)addr; + if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } else +#endif +#if defined(__Userspace__) + if (addr->sa_family == AF_CONN) { + struct sockaddr_conn *sconn; + + sconn = (struct sockaddr_conn *)addr; + if (sconn->sconn_addr != NULL) { + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } else +#endif + { + error = EAFNOSUPPORT; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } + + if (stcb != NULL) { + if (net) { + encaps->sue_port = net->port; + } else { + encaps->sue_port = stcb->asoc.port; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (encaps->sue_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + encaps->sue_port = inp->sctp_ep.port; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_udpencaps); + } + break; + } + case SCTP_ECN_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = stcb->asoc.ecn_supported; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->ecn_supported; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_PR_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = stcb->asoc.prsctp_supported; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->prsctp_supported; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_AUTH_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = stcb->asoc.auth_supported; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->auth_supported; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_ASCONF_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = stcb->asoc.asconf_supported; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->asconf_supported; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_RECONFIG_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = stcb->asoc.reconfig_supported; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->reconfig_supported; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_NRSACK_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = stcb->asoc.nrsack_supported; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->nrsack_supported; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_PKTDROP_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = stcb->asoc.pktdrop_supported; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->pktdrop_supported; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_ENABLE_STREAM_RESET: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = (uint32_t)stcb->asoc.local_strreset_support; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = (uint32_t)inp->local_strreset_support; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + case SCTP_PR_STREAM_STATUS: + { + struct sctp_prstatus *sprstat; + uint16_t sid; + uint16_t policy; + + SCTP_CHECK_AND_CAST(sprstat, optval, struct sctp_prstatus, *optsize); + SCTP_FIND_STCB(inp, stcb, sprstat->sprstat_assoc_id); + + sid = sprstat->sprstat_sid; + policy = sprstat->sprstat_policy; +#if defined(SCTP_DETAILED_STR_STATS) + if ((stcb != NULL) && + (sid < stcb->asoc.streamoutcnt) && + (policy != SCTP_PR_SCTP_NONE) && + ((policy <= SCTP_PR_SCTP_MAX) || + (policy == SCTP_PR_SCTP_ALL))) { + if (policy == SCTP_PR_SCTP_ALL) { + sprstat->sprstat_abandoned_unsent = stcb->asoc.strmout[sid].abandoned_unsent[0]; + sprstat->sprstat_abandoned_sent = stcb->asoc.strmout[sid].abandoned_sent[0]; + } else { + sprstat->sprstat_abandoned_unsent = stcb->asoc.strmout[sid].abandoned_unsent[policy]; + sprstat->sprstat_abandoned_sent = stcb->asoc.strmout[sid].abandoned_sent[policy]; + } +#else + if ((stcb != NULL) && + (sid < stcb->asoc.streamoutcnt) && + (policy == SCTP_PR_SCTP_ALL)) { + sprstat->sprstat_abandoned_unsent = stcb->asoc.strmout[sid].abandoned_unsent[0]; + sprstat->sprstat_abandoned_sent = stcb->asoc.strmout[sid].abandoned_sent[0]; +#endif + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + if (error == 0) { + *optsize = sizeof(struct sctp_prstatus); + } + break; + } + case SCTP_PR_ASSOC_STATUS: + { + struct sctp_prstatus *sprstat; + uint16_t policy; + + SCTP_CHECK_AND_CAST(sprstat, optval, struct sctp_prstatus, *optsize); + SCTP_FIND_STCB(inp, stcb, sprstat->sprstat_assoc_id); + + policy = sprstat->sprstat_policy; + if ((stcb != NULL) && + (policy != SCTP_PR_SCTP_NONE) && + ((policy <= SCTP_PR_SCTP_MAX) || + (policy == SCTP_PR_SCTP_ALL))) { + if (policy == SCTP_PR_SCTP_ALL) { + sprstat->sprstat_abandoned_unsent = stcb->asoc.abandoned_unsent[0]; + sprstat->sprstat_abandoned_sent = stcb->asoc.abandoned_sent[0]; + } else { + sprstat->sprstat_abandoned_unsent = stcb->asoc.abandoned_unsent[policy]; + sprstat->sprstat_abandoned_sent = stcb->asoc.abandoned_sent[policy]; + } + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + if (error == 0) { + *optsize = sizeof(struct sctp_prstatus); + } + break; + } + case SCTP_MAX_CWND: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + av->assoc_value = stcb->asoc.max_cwnd; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_RLOCK(inp); + av->assoc_value = inp->max_cwnd; + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + if (error == 0) { + *optsize = sizeof(struct sctp_assoc_value); + } + break; + } + default: + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); + error = ENOPROTOOPT; + break; + } /* end switch (sopt->sopt_name) */ + if (error) { + *optsize = 0; + } + return (error); +} + +#if defined(__Userspace__) +int +#else +static int +#endif +sctp_setopt(struct socket *so, int optname, void *optval, size_t optsize, + void *p) +{ + int error, set_opt; + uint32_t *mopt; + struct sctp_tcb *stcb = NULL; + struct sctp_inpcb *inp = NULL; + uint32_t vrf_id; + + if (optval == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + vrf_id = inp->def_vrf_id; + + error = 0; + switch (optname) { + case SCTP_NODELAY: + case SCTP_AUTOCLOSE: + case SCTP_AUTO_ASCONF: + case SCTP_EXPLICIT_EOR: + case SCTP_DISABLE_FRAGMENTS: + case SCTP_USE_EXT_RCVINFO: + case SCTP_I_WANT_MAPPED_V4_ADDR: + /* copy in the option value */ + SCTP_CHECK_AND_CAST(mopt, optval, uint32_t, optsize); + set_opt = 0; + if (error) + break; + switch (optname) { + case SCTP_DISABLE_FRAGMENTS: + set_opt = SCTP_PCB_FLAGS_NO_FRAGMENT; + break; + case SCTP_AUTO_ASCONF: + /* + * NOTE: we don't really support this flag + */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + /* only valid for bound all sockets */ + if ((SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) && + (*mopt != 0)) { + /* forbidden by admin */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EPERM); + return (EPERM); + } + set_opt = SCTP_PCB_FLAGS_AUTO_ASCONF; + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + break; + case SCTP_EXPLICIT_EOR: + set_opt = SCTP_PCB_FLAGS_EXPLICIT_EOR; + break; + case SCTP_USE_EXT_RCVINFO: + set_opt = SCTP_PCB_FLAGS_EXT_RCVINFO; + break; + case SCTP_I_WANT_MAPPED_V4_ADDR: + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + set_opt = SCTP_PCB_FLAGS_NEEDS_MAPPED_V4; + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + break; + case SCTP_NODELAY: + set_opt = SCTP_PCB_FLAGS_NODELAY; + break; + case SCTP_AUTOCLOSE: + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + set_opt = SCTP_PCB_FLAGS_AUTOCLOSE; + /* + * The value is in ticks. Note this does not effect + * old associations, only new ones. + */ + inp->sctp_ep.auto_close_time = sctp_secs_to_ticks(*mopt); + break; + } + SCTP_INP_WLOCK(inp); + if (*mopt != 0) { + sctp_feature_on(inp, set_opt); + } else { + sctp_feature_off(inp, set_opt); + } + SCTP_INP_WUNLOCK(inp); + break; + case SCTP_REUSE_PORT: + { + SCTP_CHECK_AND_CAST(mopt, optval, uint32_t, optsize); + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) { + /* Can't set it after we are bound */ + error = EINVAL; + break; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE)) { + /* Can't do this for a 1-m socket */ + error = EINVAL; + break; + } + if (optval) + sctp_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE); + else + sctp_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE); + break; + } + case SCTP_PARTIAL_DELIVERY_POINT: + { + uint32_t *value; + + SCTP_CHECK_AND_CAST(value, optval, uint32_t, optsize); + if (*value > SCTP_SB_LIMIT_RCV(so)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + inp->partial_delivery_point = *value; + break; + } + case SCTP_FRAGMENT_INTERLEAVE: + /* not yet until we re-write sctp_recvmsg() */ + { + uint32_t *level; + + SCTP_CHECK_AND_CAST(level, optval, uint32_t, optsize); + if (*level == SCTP_FRAG_LEVEL_2) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); + sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); + } else if (*level == SCTP_FRAG_LEVEL_1) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); + sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); + } else if (*level == SCTP_FRAG_LEVEL_0) { + sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); + sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); + + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + break; + } + case SCTP_INTERLEAVING_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + if (av->assoc_value == 0) { + inp->idata_supported = 0; + } else { + if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE)) && + (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS))) { + inp->idata_supported = 1; + } else { + /* Must have Frag interleave and stream interleave on */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_CMT_ON_OFF: + if (SCTP_BASE_SYSCTL(sctp_cmt_on_off)) { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + if (av->assoc_value > SCTP_CMT_MAX) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + if (stcb) { + stcb->asoc.sctp_cmt_on_off = av->assoc_value; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((av->assoc_id == SCTP_FUTURE_ASSOC) || + (av->assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + inp->sctp_cmt_on_off = av->assoc_value; + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((av->assoc_id == SCTP_CURRENT_ASSOC) || + (av->assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + stcb->asoc.sctp_cmt_on_off = av->assoc_value; + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); + error = ENOPROTOOPT; + } + break; + case SCTP_PLUGGABLE_CC: + { + struct sctp_assoc_value *av; + struct sctp_nets *net; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + if ((av->assoc_value != SCTP_CC_RFC2581) && + (av->assoc_value != SCTP_CC_HSTCP) && + (av->assoc_value != SCTP_CC_HTCP) && + (av->assoc_value != SCTP_CC_RTCC)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + if (stcb) { + stcb->asoc.cc_functions = sctp_cc_functions[av->assoc_value]; + stcb->asoc.congestion_control_module = av->assoc_value; + if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL) { + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net); + } + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((av->assoc_id == SCTP_FUTURE_ASSOC) || + (av->assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + inp->sctp_ep.sctp_default_cc_module = av->assoc_value; + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((av->assoc_id == SCTP_CURRENT_ASSOC) || + (av->assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + stcb->asoc.cc_functions = sctp_cc_functions[av->assoc_value]; + stcb->asoc.congestion_control_module = av->assoc_value; + if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL) { + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net); + } + } + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_CC_OPTION: + { + struct sctp_cc_option *cc_opt; + + SCTP_CHECK_AND_CAST(cc_opt, optval, struct sctp_cc_option, optsize); + SCTP_FIND_STCB(inp, stcb, cc_opt->aid_value.assoc_id); + if (stcb == NULL) { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (cc_opt->aid_value.assoc_id == SCTP_CURRENT_ASSOC)) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + if (stcb->asoc.cc_functions.sctp_cwnd_socket_option) { + (*stcb->asoc.cc_functions.sctp_cwnd_socket_option)(stcb, 1, cc_opt); + } + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } else { + error = EINVAL; + } + } else { + if (stcb->asoc.cc_functions.sctp_cwnd_socket_option == NULL) { + error = ENOTSUP; + } else { + error = (*stcb->asoc.cc_functions.sctp_cwnd_socket_option)(stcb, 1, + cc_opt); + } + SCTP_TCB_UNLOCK(stcb); + } + break; + } + case SCTP_PLUGGABLE_SS: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + if ((av->assoc_value != SCTP_SS_DEFAULT) && + (av->assoc_value != SCTP_SS_ROUND_ROBIN) && + (av->assoc_value != SCTP_SS_ROUND_ROBIN_PACKET) && + (av->assoc_value != SCTP_SS_PRIORITY) && + (av->assoc_value != SCTP_SS_FAIR_BANDWITH) && + (av->assoc_value != SCTP_SS_FIRST_COME)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + if (stcb) { + stcb->asoc.ss_functions.sctp_ss_clear(stcb, &stcb->asoc, true); + stcb->asoc.ss_functions = sctp_ss_functions[av->assoc_value]; + stcb->asoc.stream_scheduling_module = av->assoc_value; + stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc); + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((av->assoc_id == SCTP_FUTURE_ASSOC) || + (av->assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + inp->sctp_ep.sctp_default_ss_module = av->assoc_value; + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((av->assoc_id == SCTP_CURRENT_ASSOC) || + (av->assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + stcb->asoc.ss_functions.sctp_ss_clear(stcb, &stcb->asoc, true); + stcb->asoc.ss_functions = sctp_ss_functions[av->assoc_value]; + stcb->asoc.stream_scheduling_module = av->assoc_value; + stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc); + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_SS_VALUE: + { + struct sctp_stream_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_stream_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + if (stcb) { + if ((av->stream_id >= stcb->asoc.streamoutcnt) || + (stcb->asoc.ss_functions.sctp_ss_set_value(stcb, &stcb->asoc, &stcb->asoc.strmout[av->stream_id], + av->stream_value) < 0)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_CURRENT_ASSOC)) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + if (av->stream_id < stcb->asoc.streamoutcnt) { + stcb->asoc.ss_functions.sctp_ss_set_value(stcb, + &stcb->asoc, + &stcb->asoc.strmout[av->stream_id], + av->stream_value); + } + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } else { + /* Can't set stream value without association */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_CLR_STAT_LOG: + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); + error = EOPNOTSUPP; + break; + case SCTP_CONTEXT: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + stcb->asoc.context = av->assoc_value; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((av->assoc_id == SCTP_FUTURE_ASSOC) || + (av->assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + inp->sctp_context = av->assoc_value; + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((av->assoc_id == SCTP_CURRENT_ASSOC) || + (av->assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + stcb->asoc.context = av->assoc_value; + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_VRF_ID: + { + uint32_t *default_vrfid; +#ifdef SCTP_MVRF + int i; +#endif + SCTP_CHECK_AND_CAST(default_vrfid, optval, uint32_t, optsize); + if (*default_vrfid > SCTP_MAX_VRF_ID) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } +#ifdef SCTP_MVRF + for (i = 0; i < inp->num_vrfs; i++) { + /* The VRF must be in the VRF list */ + if (*default_vrfid == inp->m_vrf_ids[i]) { + SCTP_INP_WLOCK(inp); + inp->def_vrf_id = *default_vrfid; + SCTP_INP_WUNLOCK(inp); + goto sctp_done; + } + } + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; +#else + inp->def_vrf_id = *default_vrfid; +#endif +#ifdef SCTP_MVRF + sctp_done: +#endif + break; + } + case SCTP_DEL_VRF_ID: + { +#ifdef SCTP_MVRF + uint32_t *del_vrfid; + int i, fnd = 0; + + SCTP_CHECK_AND_CAST(del_vrfid, optval, uint32_t, optsize); + if (*del_vrfid > SCTP_MAX_VRF_ID) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + if (inp->num_vrfs == 1) { + /* Can't delete last one */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) { + /* Can't add more once you are bound */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + SCTP_INP_WLOCK(inp); + for (i = 0; i < inp->num_vrfs; i++) { + if (*del_vrfid == inp->m_vrf_ids[i]) { + fnd = 1; + break; + } + } + if (!fnd) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + if (i != (inp->num_vrfs - 1)) { + /* Take bottom one and move to this slot */ + inp->m_vrf_ids[i] = inp->m_vrf_ids[(inp->num_vrfs-1)]; + } + if (*del_vrfid == inp->def_vrf_id) { + /* Take the first one as the new default */ + inp->def_vrf_id = inp->m_vrf_ids[0]; + } + /* Drop the number by one killing last one */ + inp->num_vrfs--; +#else + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); + error = EOPNOTSUPP; +#endif + break; + } + case SCTP_ADD_VRF_ID: + { +#ifdef SCTP_MVRF + uint32_t *add_vrfid; + int i; + + SCTP_CHECK_AND_CAST(add_vrfid, optval, uint32_t, optsize); + if (*add_vrfid > SCTP_MAX_VRF_ID) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) { + /* Can't add more once you are bound */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + SCTP_INP_WLOCK(inp); + /* Verify its not already here */ + for (i = 0; i < inp->num_vrfs; i++) { + if (*add_vrfid == inp->m_vrf_ids[i]) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); + error = EALREADY; + SCTP_INP_WUNLOCK(inp); + break; + } + } + if ((inp->num_vrfs + 1) > inp->vrf_size) { + /* need to grow array */ + uint32_t *tarray; + SCTP_MALLOC(tarray, uint32_t *, + (sizeof(uint32_t) * (inp->vrf_size + SCTP_DEFAULT_VRF_SIZE)), + SCTP_M_MVRF); + if (tarray == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); + error = ENOMEM; + SCTP_INP_WUNLOCK(inp); + break; + } + memcpy(tarray, inp->m_vrf_ids, (sizeof(uint32_t) * inp->vrf_size)); + SCTP_FREE(inp->m_vrf_ids, SCTP_M_MVRF); + inp->m_vrf_ids = tarray; + inp->vrf_size += SCTP_DEFAULT_VRF_SIZE; + } + inp->m_vrf_ids[inp->num_vrfs] = *add_vrfid; + inp->num_vrfs++; + SCTP_INP_WUNLOCK(inp); +#else + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); + error = EOPNOTSUPP; +#endif + break; + } + case SCTP_DELAYED_SACK: + { + struct sctp_sack_info *sack; + + SCTP_CHECK_AND_CAST(sack, optval, struct sctp_sack_info, optsize); + SCTP_FIND_STCB(inp, stcb, sack->sack_assoc_id); + if (sack->sack_delay) { + if (sack->sack_delay > SCTP_MAX_SACK_DELAY) { + error = EINVAL; + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + break; + } + } + if (stcb) { + if (sack->sack_delay) { + stcb->asoc.delayed_ack = sack->sack_delay; + } + if (sack->sack_freq) { + stcb->asoc.sack_freq = sack->sack_freq; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((sack->sack_assoc_id == SCTP_FUTURE_ASSOC) || + (sack->sack_assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + if (sack->sack_delay) { + inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_RECV] = sctp_msecs_to_ticks(sack->sack_delay); + } + if (sack->sack_freq) { + inp->sctp_ep.sctp_sack_freq = sack->sack_freq; + } + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((sack->sack_assoc_id == SCTP_CURRENT_ASSOC) || + (sack->sack_assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + if (sack->sack_delay) { + stcb->asoc.delayed_ack = sack->sack_delay; + } + if (sack->sack_freq) { + stcb->asoc.sack_freq = sack->sack_freq; + } + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_AUTH_CHUNK: + { + struct sctp_authchunk *sauth; + + SCTP_CHECK_AND_CAST(sauth, optval, struct sctp_authchunk, optsize); + + SCTP_INP_WLOCK(inp); + if (sctp_auth_add_chunk(sauth->sauth_chunk, inp->sctp_ep.local_auth_chunks)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } else { + inp->auth_supported = 1; + } + SCTP_INP_WUNLOCK(inp); + break; + } + case SCTP_AUTH_KEY: + { + struct sctp_authkey *sca; + struct sctp_keyhead *shared_keys; + sctp_sharedkey_t *shared_key; + sctp_key_t *key = NULL; + size_t size; + + SCTP_CHECK_AND_CAST(sca, optval, struct sctp_authkey, optsize); + if (sca->sca_keylength == 0) { + size = optsize - sizeof(struct sctp_authkey); + } else { + if (sca->sca_keylength + sizeof(struct sctp_authkey) <= optsize) { + size = sca->sca_keylength; + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + } + SCTP_FIND_STCB(inp, stcb, sca->sca_assoc_id); + + if (stcb) { + shared_keys = &stcb->asoc.shared_keys; + /* clear the cached keys for this key id */ + sctp_clear_cachedkeys(stcb, sca->sca_keynumber); + /* + * create the new shared key and + * insert/replace it + */ + if (size > 0) { + key = sctp_set_key(sca->sca_key, (uint32_t) size); + if (key == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); + error = ENOMEM; + SCTP_TCB_UNLOCK(stcb); + break; + } + } + shared_key = sctp_alloc_sharedkey(); + if (shared_key == NULL) { + sctp_free_key(key); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); + error = ENOMEM; + SCTP_TCB_UNLOCK(stcb); + break; + } + shared_key->key = key; + shared_key->keyid = sca->sca_keynumber; + error = sctp_insert_sharedkey(shared_keys, shared_key); + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((sca->sca_assoc_id == SCTP_FUTURE_ASSOC) || + (sca->sca_assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + shared_keys = &inp->sctp_ep.shared_keys; + /* + * clear the cached keys on all assocs for + * this key id + */ + sctp_clear_cachedkeys_ep(inp, sca->sca_keynumber); + /* + * create the new shared key and + * insert/replace it + */ + if (size > 0) { + key = sctp_set_key(sca->sca_key, (uint32_t) size); + if (key == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); + error = ENOMEM; + SCTP_INP_WUNLOCK(inp); + break; + } + } + shared_key = sctp_alloc_sharedkey(); + if (shared_key == NULL) { + sctp_free_key(key); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); + error = ENOMEM; + SCTP_INP_WUNLOCK(inp); + break; + } + shared_key->key = key; + shared_key->keyid = sca->sca_keynumber; + error = sctp_insert_sharedkey(shared_keys, shared_key); + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((sca->sca_assoc_id == SCTP_CURRENT_ASSOC) || + (sca->sca_assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + shared_keys = &stcb->asoc.shared_keys; + /* clear the cached keys for this key id */ + sctp_clear_cachedkeys(stcb, sca->sca_keynumber); + /* + * create the new shared key and + * insert/replace it + */ + if (size > 0) { + key = sctp_set_key(sca->sca_key, (uint32_t) size); + if (key == NULL) { + SCTP_TCB_UNLOCK(stcb); + continue; + } + } + shared_key = sctp_alloc_sharedkey(); + if (shared_key == NULL) { + sctp_free_key(key); + SCTP_TCB_UNLOCK(stcb); + continue; + } + shared_key->key = key; + shared_key->keyid = sca->sca_keynumber; + error = sctp_insert_sharedkey(shared_keys, shared_key); + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_HMAC_IDENT: + { + struct sctp_hmacalgo *shmac; + sctp_hmaclist_t *hmaclist; + uint16_t hmacid; + uint32_t i; + + SCTP_CHECK_AND_CAST(shmac, optval, struct sctp_hmacalgo, optsize); + if ((optsize < sizeof(struct sctp_hmacalgo) + shmac->shmac_number_of_idents * sizeof(uint16_t)) || + (shmac->shmac_number_of_idents > 0xffff)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + + hmaclist = sctp_alloc_hmaclist((uint16_t)shmac->shmac_number_of_idents); + if (hmaclist == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); + error = ENOMEM; + break; + } + for (i = 0; i < shmac->shmac_number_of_idents; i++) { + hmacid = shmac->shmac_idents[i]; + if (sctp_auth_add_hmacid(hmaclist, hmacid)) { + /* invalid HMACs were found */; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + sctp_free_hmaclist(hmaclist); + goto sctp_set_hmac_done; + } + } + for (i = 0; i < hmaclist->num_algo; i++) { + if (hmaclist->hmac[i] == SCTP_AUTH_HMAC_ID_SHA1) { + /* already in list */ + break; + } + } + if (i == hmaclist->num_algo) { + /* not found in list */ + sctp_free_hmaclist(hmaclist); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + /* set it on the endpoint */ + SCTP_INP_WLOCK(inp); + if (inp->sctp_ep.local_hmacs) + sctp_free_hmaclist(inp->sctp_ep.local_hmacs); + inp->sctp_ep.local_hmacs = hmaclist; + SCTP_INP_WUNLOCK(inp); + sctp_set_hmac_done: + break; + } + case SCTP_AUTH_ACTIVE_KEY: + { + struct sctp_authkeyid *scact; + + SCTP_CHECK_AND_CAST(scact, optval, struct sctp_authkeyid, optsize); + SCTP_FIND_STCB(inp, stcb, scact->scact_assoc_id); + + /* set the active key on the right place */ + if (stcb) { + /* set the active key on the assoc */ + if (sctp_auth_setactivekey(stcb, + scact->scact_keynumber)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, + SCTP_FROM_SCTP_USRREQ, + EINVAL); + error = EINVAL; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((scact->scact_assoc_id == SCTP_FUTURE_ASSOC) || + (scact->scact_assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + if (sctp_auth_setactivekey_ep(inp, scact->scact_keynumber)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((scact->scact_assoc_id == SCTP_CURRENT_ASSOC) || + (scact->scact_assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + sctp_auth_setactivekey(stcb, scact->scact_keynumber); + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_AUTH_DELETE_KEY: + { + struct sctp_authkeyid *scdel; + + SCTP_CHECK_AND_CAST(scdel, optval, struct sctp_authkeyid, optsize); + SCTP_FIND_STCB(inp, stcb, scdel->scact_assoc_id); + + /* delete the key from the right place */ + if (stcb) { + if (sctp_delete_sharedkey(stcb, scdel->scact_keynumber)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((scdel->scact_assoc_id == SCTP_FUTURE_ASSOC) || + (scdel->scact_assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + if (sctp_delete_sharedkey_ep(inp, scdel->scact_keynumber)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((scdel->scact_assoc_id == SCTP_CURRENT_ASSOC) || + (scdel->scact_assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + sctp_delete_sharedkey(stcb, scdel->scact_keynumber); + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_AUTH_DEACTIVATE_KEY: + { + struct sctp_authkeyid *keyid; + + SCTP_CHECK_AND_CAST(keyid, optval, struct sctp_authkeyid, optsize); + SCTP_FIND_STCB(inp, stcb, keyid->scact_assoc_id); + + /* deactivate the key from the right place */ + if (stcb) { + if (sctp_deact_sharedkey(stcb, keyid->scact_keynumber)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((keyid->scact_assoc_id == SCTP_FUTURE_ASSOC) || + (keyid->scact_assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + if (sctp_deact_sharedkey_ep(inp, keyid->scact_keynumber)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((keyid->scact_assoc_id == SCTP_CURRENT_ASSOC) || + (keyid->scact_assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + sctp_deact_sharedkey(stcb, keyid->scact_keynumber); + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_ENABLE_STREAM_RESET: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + if (av->assoc_value & (~SCTP_ENABLE_VALUE_MASK)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + if (stcb) { + stcb->asoc.local_strreset_support = (uint8_t)av->assoc_value; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((av->assoc_id == SCTP_FUTURE_ASSOC) || + (av->assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + inp->local_strreset_support = (uint8_t)av->assoc_value; + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((av->assoc_id == SCTP_CURRENT_ASSOC) || + (av->assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + stcb->asoc.local_strreset_support = (uint8_t)av->assoc_value; + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_RESET_STREAMS: + { + struct sctp_reset_streams *strrst; + int i, send_out = 0; + int send_in = 0; + + SCTP_CHECK_AND_CAST(strrst, optval, struct sctp_reset_streams, optsize); + SCTP_FIND_STCB(inp, stcb, strrst->srs_assoc_id); + if (stcb == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); + error = ENOENT; + break; + } + if (stcb->asoc.reconfig_supported == 0) { + /* + * Peer does not support the chunk type. + */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); + error = EOPNOTSUPP; + SCTP_TCB_UNLOCK(stcb); + break; + } + if (SCTP_GET_STATE(stcb) != SCTP_STATE_OPEN) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + break; + } + if (sizeof(struct sctp_reset_streams) + + strrst->srs_number_streams * sizeof(uint16_t) > optsize) { + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + break; + } + if (strrst->srs_flags & SCTP_STREAM_RESET_INCOMING) { + send_in = 1; + if (stcb->asoc.stream_reset_outstanding) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); + error = EALREADY; + SCTP_TCB_UNLOCK(stcb); + break; + } + } + if (strrst->srs_flags & SCTP_STREAM_RESET_OUTGOING) { + send_out = 1; + } + if ((strrst->srs_number_streams > SCTP_MAX_STREAMS_AT_ONCE_RESET) && send_in) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); + error = ENOMEM; + SCTP_TCB_UNLOCK(stcb); + break; + } + if ((send_in == 0) && (send_out == 0)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + break; + } + for (i = 0; i < strrst->srs_number_streams; i++) { + if ((send_in) && + (strrst->srs_stream_list[i] >= stcb->asoc.streamincnt)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + if ((send_out) && + (strrst->srs_stream_list[i] >= stcb->asoc.streamoutcnt)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + } + if (error) { + SCTP_TCB_UNLOCK(stcb); + break; + } + if (send_out) { + int cnt; + uint16_t strm; + if (strrst->srs_number_streams) { + for (i = 0, cnt = 0; i < strrst->srs_number_streams; i++) { + strm = strrst->srs_stream_list[i]; + if (stcb->asoc.strmout[strm].state == SCTP_STREAM_OPEN) { + stcb->asoc.strmout[strm].state = SCTP_STREAM_RESET_PENDING; + cnt++; + } + } + } else { + /* Its all */ + for (i = 0, cnt = 0; i < stcb->asoc.streamoutcnt; i++) { + if (stcb->asoc.strmout[i].state == SCTP_STREAM_OPEN) { + stcb->asoc.strmout[i].state = SCTP_STREAM_RESET_PENDING; + cnt++; + } + } + } + } + if (send_in) { + error = sctp_send_str_reset_req(stcb, strrst->srs_number_streams, + strrst->srs_stream_list, + send_in, 0, 0, 0, 0, 0); + } else { + error = sctp_send_stream_reset_out_if_possible(stcb, SCTP_SO_LOCKED); + } + if (error == 0) { + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_STRRST_REQ, SCTP_SO_LOCKED); + } else { + /* + * For outgoing streams don't report any problems in + * sending the request to the application. + * XXX: Double check resetting incoming streams. + */ + error = 0; + } + SCTP_TCB_UNLOCK(stcb); + break; + } + case SCTP_ADD_STREAMS: + { + struct sctp_add_streams *stradd; + uint8_t addstream = 0; + uint16_t add_o_strmcnt = 0; + uint16_t add_i_strmcnt = 0; + + SCTP_CHECK_AND_CAST(stradd, optval, struct sctp_add_streams, optsize); + SCTP_FIND_STCB(inp, stcb, stradd->sas_assoc_id); + if (stcb == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); + error = ENOENT; + break; + } + if (stcb->asoc.reconfig_supported == 0) { + /* + * Peer does not support the chunk type. + */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); + error = EOPNOTSUPP; + SCTP_TCB_UNLOCK(stcb); + break; + } + if (SCTP_GET_STATE(stcb) != SCTP_STATE_OPEN) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + break; + } + if (stcb->asoc.stream_reset_outstanding) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); + error = EALREADY; + SCTP_TCB_UNLOCK(stcb); + break; + } + if ((stradd->sas_outstrms == 0) && + (stradd->sas_instrms == 0)) { + error = EINVAL; + goto skip_stuff; + } + if (stradd->sas_outstrms) { + addstream = 1; + /* We allocate here */ + add_o_strmcnt = stradd->sas_outstrms; + if ((((int)add_o_strmcnt) + ((int)stcb->asoc.streamoutcnt)) > 0x0000ffff) { + /* You can't have more than 64k */ + error = EINVAL; + goto skip_stuff; + } + } + if (stradd->sas_instrms) { + int cnt; + + addstream |= 2; + /* We allocate inside sctp_send_str_reset_req() */ + add_i_strmcnt = stradd->sas_instrms; + cnt = add_i_strmcnt; + cnt += stcb->asoc.streamincnt; + if (cnt > 0x0000ffff) { + /* You can't have more than 64k */ + error = EINVAL; + goto skip_stuff; + } + if (cnt > (int)stcb->asoc.max_inbound_streams) { + /* More than you are allowed */ + error = EINVAL; + goto skip_stuff; + } + } + error = sctp_send_str_reset_req(stcb, 0, NULL, 0, 0, addstream, add_o_strmcnt, add_i_strmcnt, 0); + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_STRRST_REQ, SCTP_SO_LOCKED); + skip_stuff: + SCTP_TCB_UNLOCK(stcb); + break; + } + case SCTP_RESET_ASSOC: + { + int i; + uint32_t *value; + + SCTP_CHECK_AND_CAST(value, optval, uint32_t, optsize); + SCTP_FIND_STCB(inp, stcb, (sctp_assoc_t) *value); + if (stcb == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); + error = ENOENT; + break; + } + if (stcb->asoc.reconfig_supported == 0) { + /* + * Peer does not support the chunk type. + */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); + error = EOPNOTSUPP; + SCTP_TCB_UNLOCK(stcb); + break; + } + if (SCTP_GET_STATE(stcb) != SCTP_STATE_OPEN) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + break; + } + if (stcb->asoc.stream_reset_outstanding) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); + error = EALREADY; + SCTP_TCB_UNLOCK(stcb); + break; + } + /* Is there any data pending in the send or sent queues? */ + if (!TAILQ_EMPTY(&stcb->asoc.send_queue) || + !TAILQ_EMPTY(&stcb->asoc.sent_queue)) { + busy_out: + error = EBUSY; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + SCTP_TCB_UNLOCK(stcb); + break; + } + /* Do any streams have data queued? */ + for (i = 0; i < stcb->asoc.streamoutcnt; i++) { + if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) { + goto busy_out; + } + } + error = sctp_send_str_reset_req(stcb, 0, NULL, 0, 1, 0, 0, 0, 0); + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_STRRST_REQ, SCTP_SO_LOCKED); + SCTP_TCB_UNLOCK(stcb); + break; + } + case SCTP_CONNECT_X: + if (optsize < (sizeof(int) + sizeof(struct sockaddr_in))) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + error = sctp_do_connect_x(so, inp, optval, optsize, p, 0); + break; + case SCTP_CONNECT_X_DELAYED: + if (optsize < (sizeof(int) + sizeof(struct sockaddr_in))) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + error = sctp_do_connect_x(so, inp, optval, optsize, p, 1); + break; + case SCTP_CONNECT_X_COMPLETE: + { + struct sockaddr *sa; + + /* FIXME MT: check correct? */ + SCTP_CHECK_AND_CAST(sa, optval, struct sockaddr, optsize); + + /* find tcb */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { + SCTP_INP_RLOCK(inp); + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb) { + SCTP_TCB_LOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } else { + /* We increment here since sctp_findassociation_ep_addr() wil + * do a decrement if it finds the stcb as long as the locked + * tcb (last argument) is NOT a TCB.. aka NULL. + */ + SCTP_INP_INCR_REF(inp); + stcb = sctp_findassociation_ep_addr(&inp, sa, NULL, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_DECR_REF(inp); + } + } + + if (stcb == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); + error = ENOENT; + break; + } + if (stcb->asoc.delayed_connection == 1) { + stcb->asoc.delayed_connection = 0; + (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); + sctp_timer_stop(SCTP_TIMER_TYPE_INIT, inp, stcb, + stcb->asoc.primary_destination, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_8); + sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); + } else { + /* + * already expired or did not use delayed + * connectx + */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); + error = EALREADY; + } + SCTP_TCB_UNLOCK(stcb); + break; + } + case SCTP_MAX_BURST: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + stcb->asoc.max_burst = av->assoc_value; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((av->assoc_id == SCTP_FUTURE_ASSOC) || + (av->assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + inp->sctp_ep.max_burst = av->assoc_value; + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((av->assoc_id == SCTP_CURRENT_ASSOC) || + (av->assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + stcb->asoc.max_burst = av->assoc_value; + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_MAXSEG: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + stcb->asoc.sctp_frag_point = av->assoc_value; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + inp->sctp_frag_point = av->assoc_value; + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_EVENTS: + { + struct sctp_event_subscribe *events; + + SCTP_CHECK_AND_CAST(events, optval, struct sctp_event_subscribe, optsize); + + SCTP_INP_WLOCK(inp); + if (events->sctp_data_io_event) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT); + } + + if (events->sctp_association_event) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVASSOCEVNT); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVASSOCEVNT); + } + + if (events->sctp_address_event) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVPADDREVNT); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVPADDREVNT); + } + + if (events->sctp_send_failure_event) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVSENDFAILEVNT); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVSENDFAILEVNT); + } + + if (events->sctp_peer_error_event) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVPEERERR); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVPEERERR); + } + + if (events->sctp_shutdown_event) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT); + } + + if (events->sctp_partial_delivery_event) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_PDAPIEVNT); + } + + if (events->sctp_adaptation_layer_event) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_ADAPTATIONEVNT); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_ADAPTATIONEVNT); + } + + if (events->sctp_authentication_event) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTHEVNT); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTHEVNT); + } + + if (events->sctp_sender_dry_event) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_DRYEVNT); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_DRYEVNT); + } + + if (events->sctp_stream_reset_event) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_STREAM_RESETEVNT); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_STREAM_RESETEVNT); + } + + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + if (events->sctp_association_event) { + sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_RECVASSOCEVNT); + } else { + sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_RECVASSOCEVNT); + } + if (events->sctp_address_event) { + sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_RECVPADDREVNT); + } else { + sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_RECVPADDREVNT); + } + if (events->sctp_send_failure_event) { + sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_RECVSENDFAILEVNT); + } else { + sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_RECVSENDFAILEVNT); + } + if (events->sctp_peer_error_event) { + sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_RECVPEERERR); + } else { + sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_RECVPEERERR); + } + if (events->sctp_shutdown_event) { + sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT); + } else { + sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT); + } + if (events->sctp_partial_delivery_event) { + sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT); + } else { + sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT); + } + if (events->sctp_adaptation_layer_event) { + sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_ADAPTATIONEVNT); + } else { + sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_ADAPTATIONEVNT); + } + if (events->sctp_authentication_event) { + sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_AUTHEVNT); + } else { + sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_AUTHEVNT); + } + if (events->sctp_sender_dry_event) { + sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_DRYEVNT); + } else { + sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_DRYEVNT); + } + if (events->sctp_stream_reset_event) { + sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_STREAM_RESETEVNT); + } else { + sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_STREAM_RESETEVNT); + } + SCTP_TCB_UNLOCK(stcb); + } + /* Send up the sender dry event only for 1-to-1 style sockets. */ + if (events->sctp_sender_dry_event) { + if (((inp->sctp_flags & (SCTP_PCB_FLAGS_TCPTYPE | SCTP_PCB_FLAGS_IN_TCPPOOL)) != 0) && + !SCTP_IS_LISTENING(inp)) { + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb != NULL) { + SCTP_TCB_LOCK(stcb); + if (TAILQ_EMPTY(&stcb->asoc.send_queue) && + TAILQ_EMPTY(&stcb->asoc.sent_queue) && + (stcb->asoc.stream_queue_cnt == 0)) { + sctp_ulp_notify(SCTP_NOTIFY_SENDER_DRY, stcb, 0, NULL, SCTP_SO_LOCKED); + } + SCTP_TCB_UNLOCK(stcb); + } + } + } + SCTP_INP_WUNLOCK(inp); + break; + } + case SCTP_ADAPTATION_LAYER: + { + struct sctp_setadaptation *adap_bits; + + SCTP_CHECK_AND_CAST(adap_bits, optval, struct sctp_setadaptation, optsize); + SCTP_INP_WLOCK(inp); + inp->sctp_ep.adaptation_layer_indicator = adap_bits->ssb_adaptation_ind; + inp->sctp_ep.adaptation_layer_indicator_provided = 1; + SCTP_INP_WUNLOCK(inp); + break; + } +#ifdef SCTP_DEBUG + case SCTP_SET_INITIAL_DBG_SEQ: + { + uint32_t *vvv; + + SCTP_CHECK_AND_CAST(vvv, optval, uint32_t, optsize); + SCTP_INP_WLOCK(inp); + inp->sctp_ep.initial_sequence_debug = *vvv; + SCTP_INP_WUNLOCK(inp); + break; + } +#endif + case SCTP_DEFAULT_SEND_PARAM: + { + struct sctp_sndrcvinfo *s_info; + + SCTP_CHECK_AND_CAST(s_info, optval, struct sctp_sndrcvinfo, optsize); + SCTP_FIND_STCB(inp, stcb, s_info->sinfo_assoc_id); + + if (stcb) { + if (s_info->sinfo_stream < stcb->asoc.streamoutcnt) { + memcpy(&stcb->asoc.def_send, s_info, min(optsize, sizeof(stcb->asoc.def_send))); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((s_info->sinfo_assoc_id == SCTP_FUTURE_ASSOC) || + (s_info->sinfo_assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + memcpy(&inp->def_send, s_info, min(optsize, sizeof(inp->def_send))); + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((s_info->sinfo_assoc_id == SCTP_CURRENT_ASSOC) || + (s_info->sinfo_assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + if (s_info->sinfo_stream < stcb->asoc.streamoutcnt) { + memcpy(&stcb->asoc.def_send, s_info, min(optsize, sizeof(stcb->asoc.def_send))); + } + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_PEER_ADDR_PARAMS: + { + struct sctp_paddrparams *paddrp; + struct sctp_nets *net; + struct sockaddr *addr; +#if defined(INET) && defined(INET6) + struct sockaddr_in sin_store; +#endif + + SCTP_CHECK_AND_CAST(paddrp, optval, struct sctp_paddrparams, optsize); + SCTP_FIND_STCB(inp, stcb, paddrp->spp_assoc_id); + +#if defined(INET) && defined(INET6) + if (paddrp->spp_address.ss_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)&paddrp->spp_address; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + in6_sin6_2_sin(&sin_store, sin6); + addr = (struct sockaddr *)&sin_store; + } else { + addr = (struct sockaddr *)&paddrp->spp_address; + } + } else { + addr = (struct sockaddr *)&paddrp->spp_address; + } +#else + addr = (struct sockaddr *)&paddrp->spp_address; +#endif + if (stcb != NULL) { + net = sctp_findnet(stcb, addr); + } else { + /* We increment here since sctp_findassociation_ep_addr() wil + * do a decrement if it finds the stcb as long as the locked + * tcb (last argument) is NOT a TCB.. aka NULL. + */ + net = NULL; + SCTP_INP_INCR_REF(inp); + stcb = sctp_findassociation_ep_addr(&inp, addr, + &net, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_DECR_REF(inp); + } + } + if ((stcb != NULL) && (net == NULL)) { +#ifdef INET + if (addr->sa_family == AF_INET) { + struct sockaddr_in *sin; + + sin = (struct sockaddr_in *)addr; + if (sin->sin_addr.s_addr != INADDR_ANY) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + SCTP_TCB_UNLOCK(stcb); + error = EINVAL; + break; + } + } else +#endif +#ifdef INET6 + if (addr->sa_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)addr; + if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + SCTP_TCB_UNLOCK(stcb); + error = EINVAL; + break; + } + } else +#endif +#if defined(__Userspace__) + if (addr->sa_family == AF_CONN) { + struct sockaddr_conn *sconn; + + sconn = (struct sockaddr_conn *)addr; + if (sconn->sconn_addr != NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + SCTP_TCB_UNLOCK(stcb); + error = EINVAL; + break; + } + } else +#endif + { + error = EAFNOSUPPORT; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } + /* sanity checks */ + if ((paddrp->spp_flags & SPP_HB_ENABLE) && (paddrp->spp_flags & SPP_HB_DISABLE)) { + if (stcb) + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + + if ((paddrp->spp_flags & SPP_PMTUD_ENABLE) && (paddrp->spp_flags & SPP_PMTUD_DISABLE)) { + if (stcb) + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + if ((paddrp->spp_flags & SPP_PMTUD_DISABLE) && + (paddrp->spp_pathmtu > 0) && + ((paddrp->spp_pathmtu < SCTP_SMALLEST_PMTU) || + (paddrp->spp_pathmtu > SCTP_LARGEST_PMTU))) { + if (stcb) + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } + + if (stcb != NULL) { + /************************TCB SPECIFIC SET ******************/ + if (net != NULL) { + /************************NET SPECIFIC SET ******************/ + if (paddrp->spp_flags & SPP_HB_DISABLE) { + if (((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) && + ((net->dest_state & SCTP_ADDR_NOHB) == 0)) { + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_9); + } + net->dest_state |= SCTP_ADDR_NOHB; + } + if (paddrp->spp_flags & SPP_HB_ENABLE) { + if (paddrp->spp_hbinterval) { + net->heart_beat_delay = paddrp->spp_hbinterval; + } else if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO) { + net->heart_beat_delay = 0; + } + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_10); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net); + net->dest_state &= ~SCTP_ADDR_NOHB; + } + if (paddrp->spp_flags & SPP_HB_DEMAND) { + if (SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) { + sctp_send_hb(stcb, net, SCTP_SO_LOCKED); + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SOCKOPT, SCTP_SO_LOCKED); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net); + } + } + if (paddrp->spp_flags & SPP_PMTUD_DISABLE) { + if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { + sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_11); + } + net->dest_state |= SCTP_ADDR_NO_PMTUD; + if (paddrp->spp_pathmtu > 0) { + net->mtu = paddrp->spp_pathmtu; + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + net->mtu += SCTP_MIN_V4_OVERHEAD; + break; +#endif +#ifdef INET6 + case AF_INET6: + net->mtu += SCTP_MIN_OVERHEAD; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + net->mtu += sizeof(struct sctphdr); + break; +#endif + default: + break; + } + if (net->mtu < stcb->asoc.smallest_mtu) { + sctp_pathmtu_adjustment(stcb, net->mtu, true); + } + } + } + if (paddrp->spp_flags & SPP_PMTUD_ENABLE) { + if (!SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { + sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net); + } + net->dest_state &= ~SCTP_ADDR_NO_PMTUD; + } + if (paddrp->spp_pathmaxrxt > 0) { + if (net->dest_state & SCTP_ADDR_PF) { + if (net->error_count > paddrp->spp_pathmaxrxt) { + net->dest_state &= ~SCTP_ADDR_PF; + } + } else { + if ((net->error_count <= paddrp->spp_pathmaxrxt) && + (net->error_count > net->pf_threshold)) { + net->dest_state |= SCTP_ADDR_PF; + sctp_send_hb(stcb, net, SCTP_SO_LOCKED); + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, + stcb->sctp_ep, stcb, net, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_12); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); + } + } + if (net->dest_state & SCTP_ADDR_REACHABLE) { + if (net->error_count > paddrp->spp_pathmaxrxt) { + net->dest_state &= ~SCTP_ADDR_REACHABLE; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, 0, net, SCTP_SO_LOCKED); + } + } else { + if (net->error_count <= paddrp->spp_pathmaxrxt) { + net->dest_state |= SCTP_ADDR_REACHABLE; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, 0, net, SCTP_SO_LOCKED); + } + } + net->failure_threshold = paddrp->spp_pathmaxrxt; + } + if (paddrp->spp_flags & SPP_DSCP) { + net->dscp = paddrp->spp_dscp & 0xfc; + net->dscp |= 0x01; + } +#ifdef INET6 + if (paddrp->spp_flags & SPP_IPV6_FLOWLABEL) { + if (net->ro._l_addr.sa.sa_family == AF_INET6) { + net->flowlabel = paddrp->spp_ipv6_flowlabel & 0x000fffff; + net->flowlabel |= 0x80000000; + } + } +#endif + } else { + /************************ASSOC ONLY -- NO NET SPECIFIC SET ******************/ + if (paddrp->spp_pathmaxrxt > 0) { + stcb->asoc.def_net_failure = paddrp->spp_pathmaxrxt; + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if (net->dest_state & SCTP_ADDR_PF) { + if (net->error_count > paddrp->spp_pathmaxrxt) { + net->dest_state &= ~SCTP_ADDR_PF; + } + } else { + if ((net->error_count <= paddrp->spp_pathmaxrxt) && + (net->error_count > net->pf_threshold)) { + net->dest_state |= SCTP_ADDR_PF; + sctp_send_hb(stcb, net, SCTP_SO_LOCKED); + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, + stcb->sctp_ep, stcb, net, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_13); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); + } + } + if (net->dest_state & SCTP_ADDR_REACHABLE) { + if (net->error_count > paddrp->spp_pathmaxrxt) { + net->dest_state &= ~SCTP_ADDR_REACHABLE; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, 0, net, SCTP_SO_LOCKED); + } + } else { + if (net->error_count <= paddrp->spp_pathmaxrxt) { + net->dest_state |= SCTP_ADDR_REACHABLE; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, 0, net, SCTP_SO_LOCKED); + } + } + net->failure_threshold = paddrp->spp_pathmaxrxt; + } + } + if (paddrp->spp_flags & SPP_HB_ENABLE) { + if (paddrp->spp_hbinterval != 0) { + stcb->asoc.heart_beat_delay = paddrp->spp_hbinterval; + } else if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO) { + stcb->asoc.heart_beat_delay = 0; + } + /* Turn back on the timer */ + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if (paddrp->spp_hbinterval != 0) { + net->heart_beat_delay = paddrp->spp_hbinterval; + } else if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO) { + net->heart_beat_delay = 0; + } + if (net->dest_state & SCTP_ADDR_NOHB) { + net->dest_state &= ~SCTP_ADDR_NOHB; + } + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_14); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net); + } + sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT); + } + if (paddrp->spp_flags & SPP_HB_DISABLE) { + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if ((net->dest_state & SCTP_ADDR_NOHB) == 0) { + net->dest_state |= SCTP_ADDR_NOHB; + if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) { + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, + inp, stcb, net, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_15); + } + } + } + sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT); + } + if (paddrp->spp_flags & SPP_PMTUD_DISABLE) { + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { + sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_16); + } + net->dest_state |= SCTP_ADDR_NO_PMTUD; + if (paddrp->spp_pathmtu > 0) { + net->mtu = paddrp->spp_pathmtu; + switch (net->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + net->mtu += SCTP_MIN_V4_OVERHEAD; + break; +#endif +#ifdef INET6 + case AF_INET6: + net->mtu += SCTP_MIN_OVERHEAD; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + net->mtu += sizeof(struct sctphdr); + break; +#endif + default: + break; + } + if (net->mtu < stcb->asoc.smallest_mtu) { + sctp_pathmtu_adjustment(stcb, net->mtu, true); + } + } + } + if (paddrp->spp_pathmtu > 0) { + stcb->asoc.default_mtu = paddrp->spp_pathmtu; + } + sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD); + } + if (paddrp->spp_flags & SPP_PMTUD_ENABLE) { + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if (!SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { + sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net); + } + net->dest_state &= ~SCTP_ADDR_NO_PMTUD; + } + stcb->asoc.default_mtu = 0; + sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD); + } + if (paddrp->spp_flags & SPP_DSCP) { + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + net->dscp = paddrp->spp_dscp & 0xfc; + net->dscp |= 0x01; + } + stcb->asoc.default_dscp = paddrp->spp_dscp & 0xfc; + stcb->asoc.default_dscp |= 0x01; + } +#ifdef INET6 + if (paddrp->spp_flags & SPP_IPV6_FLOWLABEL) { + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if (net->ro._l_addr.sa.sa_family == AF_INET6) { + net->flowlabel = paddrp->spp_ipv6_flowlabel & 0x000fffff; + net->flowlabel |= 0x80000000; + } + } + stcb->asoc.default_flowlabel = paddrp->spp_ipv6_flowlabel & 0x000fffff; + stcb->asoc.default_flowlabel |= 0x80000000; + } +#endif + } + SCTP_TCB_UNLOCK(stcb); + } else { + /************************NO TCB, SET TO default stuff ******************/ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (paddrp->spp_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + /* + * For the TOS/FLOWLABEL stuff you set it + * with the options on the socket + */ + if (paddrp->spp_pathmaxrxt > 0) { + inp->sctp_ep.def_net_failure = paddrp->spp_pathmaxrxt; + } + + if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO) + inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = 0; + else if (paddrp->spp_hbinterval != 0) { + if (paddrp->spp_hbinterval > SCTP_MAX_HB_INTERVAL) + paddrp->spp_hbinterval= SCTP_MAX_HB_INTERVAL; + inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = sctp_msecs_to_ticks(paddrp->spp_hbinterval); + } + + if (paddrp->spp_flags & SPP_HB_ENABLE) { + if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO) { + inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = 0; + } else if (paddrp->spp_hbinterval) { + inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = sctp_msecs_to_ticks(paddrp->spp_hbinterval); + } + sctp_feature_off(inp, SCTP_PCB_FLAGS_DONOT_HEARTBEAT); + } else if (paddrp->spp_flags & SPP_HB_DISABLE) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_DONOT_HEARTBEAT); + } + if (paddrp->spp_flags & SPP_PMTUD_ENABLE) { + inp->sctp_ep.default_mtu = 0; + sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_NOT_PMTUD); + } else if (paddrp->spp_flags & SPP_PMTUD_DISABLE) { + if (paddrp->spp_pathmtu > 0) { + inp->sctp_ep.default_mtu = paddrp->spp_pathmtu; + } + sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_NOT_PMTUD); + } + if (paddrp->spp_flags & SPP_DSCP) { + inp->sctp_ep.default_dscp = paddrp->spp_dscp & 0xfc; + inp->sctp_ep.default_dscp |= 0x01; + } +#ifdef INET6 + if (paddrp->spp_flags & SPP_IPV6_FLOWLABEL) { + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + inp->sctp_ep.default_flowlabel = paddrp->spp_ipv6_flowlabel & 0x000fffff; + inp->sctp_ep.default_flowlabel |= 0x80000000; + } + } +#endif + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_RTOINFO: + { + struct sctp_rtoinfo *srto; + uint32_t new_init, new_min, new_max; + + SCTP_CHECK_AND_CAST(srto, optval, struct sctp_rtoinfo, optsize); + SCTP_FIND_STCB(inp, stcb, srto->srto_assoc_id); + + if (stcb) { + if (srto->srto_initial) + new_init = srto->srto_initial; + else + new_init = stcb->asoc.initial_rto; + if (srto->srto_max) + new_max = srto->srto_max; + else + new_max = stcb->asoc.maxrto; + if (srto->srto_min) + new_min = srto->srto_min; + else + new_min = stcb->asoc.minrto; + if ((new_min <= new_init) && (new_init <= new_max)) { + stcb->asoc.initial_rto = new_init; + stcb->asoc.maxrto = new_max; + stcb->asoc.minrto = new_min; + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (srto->srto_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + if (srto->srto_initial) + new_init = srto->srto_initial; + else + new_init = inp->sctp_ep.initial_rto; + if (srto->srto_max) + new_max = srto->srto_max; + else + new_max = inp->sctp_ep.sctp_maxrto; + if (srto->srto_min) + new_min = srto->srto_min; + else + new_min = inp->sctp_ep.sctp_minrto; + if ((new_min <= new_init) && (new_init <= new_max)) { + inp->sctp_ep.initial_rto = new_init; + inp->sctp_ep.sctp_maxrto = new_max; + inp->sctp_ep.sctp_minrto = new_min; + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_ASSOCINFO: + { + struct sctp_assocparams *sasoc; + + SCTP_CHECK_AND_CAST(sasoc, optval, struct sctp_assocparams, optsize); + SCTP_FIND_STCB(inp, stcb, sasoc->sasoc_assoc_id); + if (sasoc->sasoc_cookie_life > 0) { + /* boundary check the cookie life */ + if (sasoc->sasoc_cookie_life < SCTP_MIN_COOKIE_LIFE) { + sasoc->sasoc_cookie_life = SCTP_MIN_COOKIE_LIFE; + } + if (sasoc->sasoc_cookie_life > SCTP_MAX_COOKIE_LIFE) { + sasoc->sasoc_cookie_life = SCTP_MAX_COOKIE_LIFE; + } + } + if (stcb) { + if (sasoc->sasoc_asocmaxrxt > 0) { + stcb->asoc.max_send_times = sasoc->sasoc_asocmaxrxt; + } + if (sasoc->sasoc_cookie_life > 0) { + stcb->asoc.cookie_life = sctp_msecs_to_ticks(sasoc->sasoc_cookie_life); + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (sasoc->sasoc_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + if (sasoc->sasoc_asocmaxrxt > 0) { + inp->sctp_ep.max_send_times = sasoc->sasoc_asocmaxrxt; + } + if (sasoc->sasoc_cookie_life > 0) { + inp->sctp_ep.def_cookie_life = sctp_msecs_to_ticks(sasoc->sasoc_cookie_life); + } + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_INITMSG: + { + struct sctp_initmsg *sinit; + + SCTP_CHECK_AND_CAST(sinit, optval, struct sctp_initmsg, optsize); + SCTP_INP_WLOCK(inp); + if (sinit->sinit_num_ostreams) + inp->sctp_ep.pre_open_stream_count = sinit->sinit_num_ostreams; + + if (sinit->sinit_max_instreams) + inp->sctp_ep.max_open_streams_intome = sinit->sinit_max_instreams; + + if (sinit->sinit_max_attempts) + inp->sctp_ep.max_init_times = sinit->sinit_max_attempts; + + if (sinit->sinit_max_init_timeo) + inp->sctp_ep.initial_init_rto_max = sinit->sinit_max_init_timeo; + SCTP_INP_WUNLOCK(inp); + break; + } + case SCTP_PRIMARY_ADDR: + { + struct sctp_setprim *spa; + struct sctp_nets *net; + struct sockaddr *addr; +#if defined(INET) && defined(INET6) + struct sockaddr_in sin_store; +#endif + + SCTP_CHECK_AND_CAST(spa, optval, struct sctp_setprim, optsize); + SCTP_FIND_STCB(inp, stcb, spa->ssp_assoc_id); + +#if defined(INET) && defined(INET6) + if (spa->ssp_addr.ss_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)&spa->ssp_addr; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + in6_sin6_2_sin(&sin_store, sin6); + addr = (struct sockaddr *)&sin_store; + } else { + addr = (struct sockaddr *)&spa->ssp_addr; + } + } else { + addr = (struct sockaddr *)&spa->ssp_addr; + } +#else + addr = (struct sockaddr *)&spa->ssp_addr; +#endif + if (stcb != NULL) { + net = sctp_findnet(stcb, addr); + } else { + /* We increment here since sctp_findassociation_ep_addr() wil + * do a decrement if it finds the stcb as long as the locked + * tcb (last argument) is NOT a TCB.. aka NULL. + */ + net = NULL; + SCTP_INP_INCR_REF(inp); + stcb = sctp_findassociation_ep_addr(&inp, addr, + &net, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_DECR_REF(inp); + } + } + + if ((stcb != NULL) && (net != NULL)) { + if (net != stcb->asoc.primary_destination) { + if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) { + /* Ok we need to set it */ + if (sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, net) == 0) { + if ((stcb->asoc.alternate) && + ((net->dest_state & SCTP_ADDR_PF) == 0) && + (net->dest_state & SCTP_ADDR_REACHABLE)) { + sctp_free_remote_addr(stcb->asoc.alternate); + stcb->asoc.alternate = NULL; + } + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + break; + } + case SCTP_SET_DYNAMIC_PRIMARY: + { + union sctp_sockstore *ss; +#ifdef SCTP_MVRF + int i, fnd = 0; +#endif +#if !defined(_WIN32) && !defined(__Userspace__) +#if defined(__APPLE__) + struct proc *proc; +#endif +#if defined(__FreeBSD__) + error = priv_check(curthread, + PRIV_NETINET_RESERVEDPORT); +#elif defined(__APPLE__) + proc = (struct proc *)p; + if (p) { + error = suser(proc->p_ucred, &proc->p_acflag); + } else { + break; + } +#else + error = suser(p, 0); +#endif + if (error) + break; +#endif + + SCTP_CHECK_AND_CAST(ss, optval, union sctp_sockstore, optsize); + /* SUPER USER CHECK? */ +#ifdef SCTP_MVRF + for (i = 0; i < inp->num_vrfs; i++) { + if (vrf_id == inp->m_vrf_ids[i]) { + fnd = 1; + break; + } + } + if (!fnd) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } +#endif + error = sctp_dynamic_set_primary(&ss->sa, vrf_id); + break; + } + case SCTP_SET_PEER_PRIMARY_ADDR: + { + struct sctp_setpeerprim *sspp; + struct sockaddr *addr; +#if defined(INET) && defined(INET6) + struct sockaddr_in sin_store; +#endif + + SCTP_CHECK_AND_CAST(sspp, optval, struct sctp_setpeerprim, optsize); + SCTP_FIND_STCB(inp, stcb, sspp->sspp_assoc_id); + if (stcb != NULL) { + struct sctp_ifa *ifa; + +#if defined(INET) && defined(INET6) + if (sspp->sspp_addr.ss_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)&sspp->sspp_addr; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + in6_sin6_2_sin(&sin_store, sin6); + addr = (struct sockaddr *)&sin_store; + } else { + addr = (struct sockaddr *)&sspp->sspp_addr; + } + } else { + addr = (struct sockaddr *)&sspp->sspp_addr; + } +#else + addr = (struct sockaddr *)&sspp->sspp_addr; +#endif + ifa = sctp_find_ifa_by_addr(addr, stcb->asoc.vrf_id, SCTP_ADDR_NOT_LOCKED); + if (ifa == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_of_it; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { + /* Must validate the ifa found is in our ep */ + struct sctp_laddr *laddr; + int found = 0; + + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa == NULL) { + SCTPDBG(SCTP_DEBUG_OUTPUT1, "%s: NULL ifa\n", + __func__); + continue; + } + if ((sctp_is_addr_restricted(stcb, laddr->ifa)) && + (!sctp_is_addr_pending(stcb, laddr->ifa))) { + continue; + } + if (laddr->ifa == ifa) { + found = 1; + break; + } + } + if (!found) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_of_it; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + } else { + switch (addr->sa_family) { +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin; + + sin = (struct sockaddr_in *)addr; + if (prison_check_ip4(inp->ip_inp.inp.inp_cred, + &sin->sin_addr) != 0) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_of_it; + } + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)addr; + if (prison_check_ip6(inp->ip_inp.inp.inp_cred, + &sin6->sin6_addr) != 0) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_of_it; + } + break; + } +#endif + default: + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_of_it; + } +#endif + } + if (sctp_set_primary_ip_address_sa(stcb, addr) != 0) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SOCKOPT, SCTP_SO_LOCKED); + out_of_it: + SCTP_TCB_UNLOCK(stcb); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + break; + } + case SCTP_BINDX_ADD_ADDR: + { + struct sockaddr *sa; +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct thread *td; + + td = (struct thread *)p; +#endif + SCTP_CHECK_AND_CAST(sa, optval, struct sockaddr, optsize); +#ifdef INET + if (sa->sa_family == AF_INET) { + if (optsize < sizeof(struct sockaddr_in)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (td != NULL && + (error = prison_local_ip4(td->td_ucred, &(((struct sockaddr_in *)sa)->sin_addr)))) { + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } +#endif + } else +#endif +#ifdef INET6 + if (sa->sa_family == AF_INET6) { + if (optsize < sizeof(struct sockaddr_in6)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (td != NULL && + (error = prison_local_ip6(td->td_ucred, + &(((struct sockaddr_in6 *)sa)->sin6_addr), + (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) { + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } +#endif + } else +#endif + { + error = EAFNOSUPPORT; + break; + } + sctp_bindx_add_address(so, inp, sa, vrf_id, &error, p); + break; + } + case SCTP_BINDX_REM_ADDR: + { + struct sockaddr *sa; +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct thread *td; + td = (struct thread *)p; + +#endif + SCTP_CHECK_AND_CAST(sa, optval, struct sockaddr, optsize); +#ifdef INET + if (sa->sa_family == AF_INET) { + if (optsize < sizeof(struct sockaddr_in)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (td != NULL && + (error = prison_local_ip4(td->td_ucred, &(((struct sockaddr_in *)sa)->sin_addr)))) { + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } +#endif + } else +#endif +#ifdef INET6 + if (sa->sa_family == AF_INET6) { + if (optsize < sizeof(struct sockaddr_in6)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (td != NULL && + (error = prison_local_ip6(td->td_ucred, + &(((struct sockaddr_in6 *)sa)->sin6_addr), + (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) { + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } +#endif + } else +#endif + { + error = EAFNOSUPPORT; + break; + } + sctp_bindx_delete_address(inp, sa, vrf_id, &error); + break; + } +#if defined(__APPLE__) && !defined(__Userspace__) + case SCTP_LISTEN_FIX: + /* only applies to one-to-many sockets */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) { + /* make sure the ACCEPTCONN flag is OFF */ + so->so_options &= ~SO_ACCEPTCONN; + } else { + /* otherwise, not allowed */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + break; +#endif + case SCTP_EVENT: + { + struct sctp_event *event; + uint32_t event_type; + + SCTP_CHECK_AND_CAST(event, optval, struct sctp_event, optsize); + SCTP_FIND_STCB(inp, stcb, event->se_assoc_id); + switch (event->se_type) { + case SCTP_ASSOC_CHANGE: + event_type = SCTP_PCB_FLAGS_RECVASSOCEVNT; + break; + case SCTP_PEER_ADDR_CHANGE: + event_type = SCTP_PCB_FLAGS_RECVPADDREVNT; + break; + case SCTP_REMOTE_ERROR: + event_type = SCTP_PCB_FLAGS_RECVPEERERR; + break; + case SCTP_SEND_FAILED: + event_type = SCTP_PCB_FLAGS_RECVSENDFAILEVNT; + break; + case SCTP_SHUTDOWN_EVENT: + event_type = SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT; + break; + case SCTP_ADAPTATION_INDICATION: + event_type = SCTP_PCB_FLAGS_ADAPTATIONEVNT; + break; + case SCTP_PARTIAL_DELIVERY_EVENT: + event_type = SCTP_PCB_FLAGS_PDAPIEVNT; + break; + case SCTP_AUTHENTICATION_EVENT: + event_type = SCTP_PCB_FLAGS_AUTHEVNT; + break; + case SCTP_STREAM_RESET_EVENT: + event_type = SCTP_PCB_FLAGS_STREAM_RESETEVNT; + break; + case SCTP_SENDER_DRY_EVENT: + event_type = SCTP_PCB_FLAGS_DRYEVNT; + break; + case SCTP_NOTIFICATIONS_STOPPED_EVENT: + event_type = 0; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTSUP); + error = ENOTSUP; + break; + case SCTP_ASSOC_RESET_EVENT: + event_type = SCTP_PCB_FLAGS_ASSOC_RESETEVNT; + break; + case SCTP_STREAM_CHANGE_EVENT: + event_type = SCTP_PCB_FLAGS_STREAM_CHANGEEVNT; + break; + case SCTP_SEND_FAILED_EVENT: + event_type = SCTP_PCB_FLAGS_RECVNSENDFAILEVNT; + break; + default: + event_type = 0; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + if (event_type > 0) { + if (stcb) { + if (event->se_on) { + sctp_stcb_feature_on(inp, stcb, event_type); + if (event_type == SCTP_PCB_FLAGS_DRYEVNT) { + if (TAILQ_EMPTY(&stcb->asoc.send_queue) && + TAILQ_EMPTY(&stcb->asoc.sent_queue) && + (stcb->asoc.stream_queue_cnt == 0)) { + sctp_ulp_notify(SCTP_NOTIFY_SENDER_DRY, stcb, 0, NULL, SCTP_SO_LOCKED); + } + } + } else { + sctp_stcb_feature_off(inp, stcb, event_type); + } + SCTP_TCB_UNLOCK(stcb); + } else { + /* + * We don't want to send up a storm of events, + * so return an error for sender dry events + */ + if ((event_type == SCTP_PCB_FLAGS_DRYEVNT) && + (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((event->se_assoc_id == SCTP_ALL_ASSOC) || + (event->se_assoc_id == SCTP_CURRENT_ASSOC))) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTSUP); + error = ENOTSUP; + break; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((event->se_assoc_id == SCTP_FUTURE_ASSOC) || + (event->se_assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + if (event->se_on) { + sctp_feature_on(inp, event_type); + } else { + sctp_feature_off(inp, event_type); + } + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((event->se_assoc_id == SCTP_CURRENT_ASSOC) || + (event->se_assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + if (event->se_on) { + sctp_stcb_feature_on(inp, stcb, event_type); + } else { + sctp_stcb_feature_off(inp, stcb, event_type); + } + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + } else { + if (stcb) { + SCTP_TCB_UNLOCK(stcb); + } + } + break; + } + case SCTP_RECVRCVINFO: + { + int *onoff; + + SCTP_CHECK_AND_CAST(onoff, optval, int, optsize); + SCTP_INP_WLOCK(inp); + if (*onoff != 0) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVRCVINFO); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVRCVINFO); + } + SCTP_INP_WUNLOCK(inp); + break; + } + case SCTP_RECVNXTINFO: + { + int *onoff; + + SCTP_CHECK_AND_CAST(onoff, optval, int, optsize); + SCTP_INP_WLOCK(inp); + if (*onoff != 0) { + sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVNXTINFO); + } else { + sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVNXTINFO); + } + SCTP_INP_WUNLOCK(inp); + break; + } + case SCTP_DEFAULT_SNDINFO: + { + struct sctp_sndinfo *info; + uint16_t policy; + + SCTP_CHECK_AND_CAST(info, optval, struct sctp_sndinfo, optsize); + SCTP_FIND_STCB(inp, stcb, info->snd_assoc_id); + + if (stcb) { + if (info->snd_sid < stcb->asoc.streamoutcnt) { + stcb->asoc.def_send.sinfo_stream = info->snd_sid; + policy = PR_SCTP_POLICY(stcb->asoc.def_send.sinfo_flags); + stcb->asoc.def_send.sinfo_flags = info->snd_flags; + stcb->asoc.def_send.sinfo_flags |= policy; + stcb->asoc.def_send.sinfo_ppid = info->snd_ppid; + stcb->asoc.def_send.sinfo_context = info->snd_context; + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((info->snd_assoc_id == SCTP_FUTURE_ASSOC) || + (info->snd_assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + inp->def_send.sinfo_stream = info->snd_sid; + policy = PR_SCTP_POLICY(inp->def_send.sinfo_flags); + inp->def_send.sinfo_flags = info->snd_flags; + inp->def_send.sinfo_flags |= policy; + inp->def_send.sinfo_ppid = info->snd_ppid; + inp->def_send.sinfo_context = info->snd_context; + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((info->snd_assoc_id == SCTP_CURRENT_ASSOC) || + (info->snd_assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + if (info->snd_sid < stcb->asoc.streamoutcnt) { + stcb->asoc.def_send.sinfo_stream = info->snd_sid; + policy = PR_SCTP_POLICY(stcb->asoc.def_send.sinfo_flags); + stcb->asoc.def_send.sinfo_flags = info->snd_flags; + stcb->asoc.def_send.sinfo_flags |= policy; + stcb->asoc.def_send.sinfo_ppid = info->snd_ppid; + stcb->asoc.def_send.sinfo_context = info->snd_context; + } + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_DEFAULT_PRINFO: + { + struct sctp_default_prinfo *info; + + SCTP_CHECK_AND_CAST(info, optval, struct sctp_default_prinfo, optsize); + SCTP_FIND_STCB(inp, stcb, info->pr_assoc_id); + + if (info->pr_policy > SCTP_PR_SCTP_MAX) { + if (stcb) { + SCTP_TCB_UNLOCK(stcb); + } + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + if (stcb) { + stcb->asoc.def_send.sinfo_flags &= 0xfff0; + stcb->asoc.def_send.sinfo_flags |= info->pr_policy; + stcb->asoc.def_send.sinfo_timetolive = info->pr_value; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((info->pr_assoc_id == SCTP_FUTURE_ASSOC) || + (info->pr_assoc_id == SCTP_ALL_ASSOC)))) { + SCTP_INP_WLOCK(inp); + inp->def_send.sinfo_flags &= 0xfff0; + inp->def_send.sinfo_flags |= info->pr_policy; + inp->def_send.sinfo_timetolive = info->pr_value; + SCTP_INP_WUNLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + ((info->pr_assoc_id == SCTP_CURRENT_ASSOC) || + (info->pr_assoc_id == SCTP_ALL_ASSOC))) { + SCTP_INP_RLOCK(inp); + LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { + SCTP_TCB_LOCK(stcb); + stcb->asoc.def_send.sinfo_flags &= 0xfff0; + stcb->asoc.def_send.sinfo_flags |= info->pr_policy; + stcb->asoc.def_send.sinfo_timetolive = info->pr_value; + SCTP_TCB_UNLOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } + } + break; + } + case SCTP_PEER_ADDR_THLDS: + /* Applies to the specific association */ + { + struct sctp_paddrthlds *thlds; + struct sctp_nets *net; + struct sockaddr *addr; +#if defined(INET) && defined(INET6) + struct sockaddr_in sin_store; +#endif + + SCTP_CHECK_AND_CAST(thlds, optval, struct sctp_paddrthlds, optsize); + SCTP_FIND_STCB(inp, stcb, thlds->spt_assoc_id); + +#if defined(INET) && defined(INET6) + if (thlds->spt_address.ss_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)&thlds->spt_address; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + in6_sin6_2_sin(&sin_store, sin6); + addr = (struct sockaddr *)&sin_store; + } else { + addr = (struct sockaddr *)&thlds->spt_address; + } + } else { + addr = (struct sockaddr *)&thlds->spt_address; + } +#else + addr = (struct sockaddr *)&thlds->spt_address; +#endif + if (stcb != NULL) { + net = sctp_findnet(stcb, addr); + } else { + /* We increment here since sctp_findassociation_ep_addr() wil + * do a decrement if it finds the stcb as long as the locked + * tcb (last argument) is NOT a TCB.. aka NULL. + */ + net = NULL; + SCTP_INP_INCR_REF(inp); + stcb = sctp_findassociation_ep_addr(&inp, addr, + &net, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_DECR_REF(inp); + } + } + if ((stcb != NULL) && (net == NULL)) { +#ifdef INET + if (addr->sa_family == AF_INET) { + struct sockaddr_in *sin; + + sin = (struct sockaddr_in *)addr; + if (sin->sin_addr.s_addr != INADDR_ANY) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + SCTP_TCB_UNLOCK(stcb); + error = EINVAL; + break; + } + } else +#endif +#ifdef INET6 + if (addr->sa_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)addr; + if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + SCTP_TCB_UNLOCK(stcb); + error = EINVAL; + break; + } + } else +#endif +#if defined(__Userspace__) + if (addr->sa_family == AF_CONN) { + struct sockaddr_conn *sconn; + + sconn = (struct sockaddr_conn *)addr; + if (sconn->sconn_addr != NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + SCTP_TCB_UNLOCK(stcb); + error = EINVAL; + break; + } + } else +#endif + { + error = EAFNOSUPPORT; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } + if (thlds->spt_pathcpthld != 0xffff) { + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + if (stcb != NULL) { + if (net != NULL) { + net->failure_threshold = thlds->spt_pathmaxrxt; + net->pf_threshold = thlds->spt_pathpfthld; + if (net->dest_state & SCTP_ADDR_PF) { + if ((net->error_count > net->failure_threshold) || + (net->error_count <= net->pf_threshold)) { + net->dest_state &= ~SCTP_ADDR_PF; + } + } else { + if ((net->error_count > net->pf_threshold) && + (net->error_count <= net->failure_threshold)) { + net->dest_state |= SCTP_ADDR_PF; + sctp_send_hb(stcb, net, SCTP_SO_LOCKED); + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, + stcb->sctp_ep, stcb, net, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_17); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); + } + } + if (net->dest_state & SCTP_ADDR_REACHABLE) { + if (net->error_count > net->failure_threshold) { + net->dest_state &= ~SCTP_ADDR_REACHABLE; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, 0, net, SCTP_SO_LOCKED); + } + } else { + if (net->error_count <= net->failure_threshold) { + net->dest_state |= SCTP_ADDR_REACHABLE; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, 0, net, SCTP_SO_LOCKED); + } + } + } else { + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + net->failure_threshold = thlds->spt_pathmaxrxt; + net->pf_threshold = thlds->spt_pathpfthld; + if (net->dest_state & SCTP_ADDR_PF) { + if ((net->error_count > net->failure_threshold) || + (net->error_count <= net->pf_threshold)) { + net->dest_state &= ~SCTP_ADDR_PF; + } + } else { + if ((net->error_count > net->pf_threshold) && + (net->error_count <= net->failure_threshold)) { + net->dest_state |= SCTP_ADDR_PF; + sctp_send_hb(stcb, net, SCTP_SO_LOCKED); + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, + stcb->sctp_ep, stcb, net, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_18); + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); + } + } + if (net->dest_state & SCTP_ADDR_REACHABLE) { + if (net->error_count > net->failure_threshold) { + net->dest_state &= ~SCTP_ADDR_REACHABLE; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, 0, net, SCTP_SO_LOCKED); + } + } else { + if (net->error_count <= net->failure_threshold) { + net->dest_state |= SCTP_ADDR_REACHABLE; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, 0, net, SCTP_SO_LOCKED); + } + } + } + stcb->asoc.def_net_failure = thlds->spt_pathmaxrxt; + stcb->asoc.def_net_pf_threshold = thlds->spt_pathpfthld; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (thlds->spt_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + inp->sctp_ep.def_net_failure = thlds->spt_pathmaxrxt; + inp->sctp_ep.def_net_pf_threshold = thlds->spt_pathpfthld; + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_REMOTE_UDP_ENCAPS_PORT: + { + struct sctp_udpencaps *encaps; + struct sctp_nets *net; + struct sockaddr *addr; +#if defined(INET) && defined(INET6) + struct sockaddr_in sin_store; +#endif + + SCTP_CHECK_AND_CAST(encaps, optval, struct sctp_udpencaps, optsize); + SCTP_FIND_STCB(inp, stcb, encaps->sue_assoc_id); + +#if defined(INET) && defined(INET6) + if (encaps->sue_address.ss_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)&encaps->sue_address; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + in6_sin6_2_sin(&sin_store, sin6); + addr = (struct sockaddr *)&sin_store; + } else { + addr = (struct sockaddr *)&encaps->sue_address; + } + } else { + addr = (struct sockaddr *)&encaps->sue_address; + } +#else + addr = (struct sockaddr *)&encaps->sue_address; +#endif + if (stcb != NULL) { + net = sctp_findnet(stcb, addr); + } else { + /* We increment here since sctp_findassociation_ep_addr() wil + * do a decrement if it finds the stcb as long as the locked + * tcb (last argument) is NOT a TCB.. aka NULL. + */ + net = NULL; + SCTP_INP_INCR_REF(inp); + stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_DECR_REF(inp); + } + } + if ((stcb != NULL) && (net == NULL)) { +#ifdef INET + if (addr->sa_family == AF_INET) { + struct sockaddr_in *sin; + + sin = (struct sockaddr_in *)addr; + if (sin->sin_addr.s_addr != INADDR_ANY) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + SCTP_TCB_UNLOCK(stcb); + error = EINVAL; + break; + } + } else +#endif +#ifdef INET6 + if (addr->sa_family == AF_INET6) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)addr; + if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + SCTP_TCB_UNLOCK(stcb); + error = EINVAL; + break; + } + } else +#endif +#if defined(__Userspace__) + if (addr->sa_family == AF_CONN) { + struct sockaddr_conn *sconn; + + sconn = (struct sockaddr_conn *)addr; + if (sconn->sconn_addr != NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + SCTP_TCB_UNLOCK(stcb); + error = EINVAL; + break; + } + } else +#endif + { + error = EAFNOSUPPORT; + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + break; + } + } + + if (stcb != NULL) { + if (net != NULL) { + net->port = encaps->sue_port; + } else { + stcb->asoc.port = encaps->sue_port; + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (encaps->sue_assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + inp->sctp_ep.port = encaps->sue_port; + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_ECN_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + if (av->assoc_value == 0) { + inp->ecn_supported = 0; + } else { + inp->ecn_supported = 1; + } + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_PR_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + if (av->assoc_value == 0) { + inp->prsctp_supported = 0; + } else { + inp->prsctp_supported = 1; + } + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_AUTH_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + if ((av->assoc_value == 0) && + (inp->asconf_supported == 1)) { + /* AUTH is required for ASCONF */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } else { + SCTP_INP_WLOCK(inp); + if (av->assoc_value == 0) { + inp->auth_supported = 0; + } else { + inp->auth_supported = 1; + } + SCTP_INP_WUNLOCK(inp); + } + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_ASCONF_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + if ((av->assoc_value != 0) && + (inp->auth_supported == 0)) { + /* AUTH is required for ASCONF */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } else { + SCTP_INP_WLOCK(inp); + if (av->assoc_value == 0) { + inp->asconf_supported = 0; + sctp_auth_delete_chunk(SCTP_ASCONF, + inp->sctp_ep.local_auth_chunks); + sctp_auth_delete_chunk(SCTP_ASCONF_ACK, + inp->sctp_ep.local_auth_chunks); + } else { + inp->asconf_supported = 1; + sctp_auth_add_chunk(SCTP_ASCONF, + inp->sctp_ep.local_auth_chunks); + sctp_auth_add_chunk(SCTP_ASCONF_ACK, + inp->sctp_ep.local_auth_chunks); + } + SCTP_INP_WUNLOCK(inp); + } + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_RECONFIG_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + if (av->assoc_value == 0) { + inp->reconfig_supported = 0; + } else { + inp->reconfig_supported = 1; + } + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_NRSACK_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + if (av->assoc_value == 0) { + inp->nrsack_supported = 0; + } else { + inp->nrsack_supported = 1; + } + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_PKTDROP_SUPPORTED: + { + struct sctp_assoc_value *av; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + if (av->assoc_value == 0) { + inp->pktdrop_supported = 0; + } else { + inp->pktdrop_supported = 1; + } + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_MAX_CWND: + { + struct sctp_assoc_value *av; + struct sctp_nets *net; + + SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); + SCTP_FIND_STCB(inp, stcb, av->assoc_id); + + if (stcb) { + stcb->asoc.max_cwnd = av->assoc_value; + if (stcb->asoc.max_cwnd > 0) { + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + if ((net->cwnd > stcb->asoc.max_cwnd) && + (net->cwnd > (net->mtu - sizeof(struct sctphdr)))) { + net->cwnd = stcb->asoc.max_cwnd; + if (net->cwnd < (net->mtu - sizeof(struct sctphdr))) { + net->cwnd = net->mtu - sizeof(struct sctphdr); + } + } + } + } + SCTP_TCB_UNLOCK(stcb); + } else { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) && + (av->assoc_id == SCTP_FUTURE_ASSOC))) { + SCTP_INP_WLOCK(inp); + inp->max_cwnd = av->assoc_value; + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + } + break; + } + case SCTP_ACCEPT_ZERO_CHECKSUM: + { + uint32_t *value; + + SCTP_CHECK_AND_CAST(value, optval, uint32_t, optsize); + SCTP_INP_WLOCK(inp); + if (*value != 0) { + inp->zero_checksum = 1; + } else { + inp->zero_checksum = 0; + } + SCTP_INP_WUNLOCK(inp); + break; + } + + default: + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); + error = ENOPROTOOPT; + break; + } /* end switch (opt) */ + return (error); +} + +#if !defined(__Userspace__) +int +sctp_ctloutput(struct socket *so, struct sockopt *sopt) +{ +#if defined(__FreeBSD__) + struct epoch_tracker et; + struct sctp_inpcb *inp; +#endif + void *optval = NULL; + void *p; + size_t optsize = 0; + int error = 0; + +#if defined(__FreeBSD__) + if ((sopt->sopt_level == SOL_SOCKET) && + (sopt->sopt_name == SO_SETFIB)) { + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(so->so_pcb, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOBUFS); + return (EINVAL); + } + SCTP_INP_WLOCK(inp); + inp->fibnum = so->so_fibnum; + SCTP_INP_WUNLOCK(inp); + return (0); + } +#endif + if (sopt->sopt_level != IPPROTO_SCTP) { + /* wrong proto level... send back up to IP */ +#ifdef INET6 + if (INP_CHECK_SOCKAF(so, AF_INET6)) + error = ip6_ctloutput(so, sopt); +#endif /* INET6 */ +#if defined(INET) && defined(INET6) + else +#endif +#ifdef INET + error = ip_ctloutput(so, sopt); +#endif + return (error); + } + optsize = sopt->sopt_valsize; + if (optsize > SCTP_SOCKET_OPTION_LIMIT) { + SCTP_LTRACE_ERR_RET(so->so_pcb, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOBUFS); + return (ENOBUFS); + } + if (optsize) { + SCTP_MALLOC(optval, void *, optsize, SCTP_M_SOCKOPT); + if (optval == NULL) { + SCTP_LTRACE_ERR_RET(so->so_pcb, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOBUFS); + return (ENOBUFS); + } + error = sooptcopyin(sopt, optval, optsize, optsize); + if (error) { + SCTP_FREE(optval, SCTP_M_SOCKOPT); + goto out; + } + } +#if defined(__FreeBSD__) || defined(_WIN32) + p = (void *)sopt->sopt_td; +#else + p = (void *)sopt->sopt_p; +#endif + if (sopt->sopt_dir == SOPT_SET) { +#if defined(__FreeBSD__) + NET_EPOCH_ENTER(et); +#endif + error = sctp_setopt(so, sopt->sopt_name, optval, optsize, p); +#if defined(__FreeBSD__) + NET_EPOCH_EXIT(et); +#endif + } else if (sopt->sopt_dir == SOPT_GET) { + error = sctp_getopt(so, sopt->sopt_name, optval, &optsize, p); + } else { + SCTP_LTRACE_ERR_RET(so->so_pcb, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + } + if ((error == 0) && (optval != NULL)) { + error = sooptcopyout(sopt, optval, optsize); + SCTP_FREE(optval, SCTP_M_SOCKOPT); + } else if (optval != NULL) { + SCTP_FREE(optval, SCTP_M_SOCKOPT); + } +out: + return (error); +} +#endif + +#ifdef INET +#if defined(__Userspace__) +int +sctp_connect(struct socket *so, struct sockaddr *addr) +{ + void *p = NULL; +#elif defined(__FreeBSD__) +static int +sctp_connect(struct socket *so, struct sockaddr *addr, struct thread *p) +{ +#elif defined(__APPLE__) +static int +sctp_connect(struct socket *so, struct sockaddr *addr, struct proc *p) +{ +#elif defined(_WIN32) +static int +sctp_connect(struct socket *so, struct sockaddr *addr, PKTHREAD p) +{ +#else +static int +sctp_connect(struct socket *so, struct mbuf *nam, struct proc *p) +{ + struct sockaddr *addr = mtod(nam, struct sockaddr *); + +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif +#ifdef SCTP_MVRF + int i, fnd = 0; +#endif + int error = 0; + int create_lock_on = 0; + uint32_t vrf_id; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb = NULL; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + /* I made the same as TCP since we are not setup? */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (ECONNRESET); + } + if (addr == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return EINVAL; + } + +#if defined(__Userspace__) + /* TODO __Userspace__ falls into this code for IPv6 stuff at the moment... */ +#endif +#if !defined(_WIN32) && !defined(__linux__) && !defined(__EMSCRIPTEN__) + switch (addr->sa_family) { +#ifdef INET6 + case AF_INET6: + { +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct sockaddr_in6 *sin6; + +#endif + if (addr->sa_len != sizeof(struct sockaddr_in6)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + sin6 = (struct sockaddr_in6 *)addr; + if (p != NULL && (error = prison_remote_ip6(p->td_ucred, &sin6->sin6_addr)) != 0) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + return (error); + } +#endif + break; + } +#endif +#ifdef INET + case AF_INET: + { +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct sockaddr_in *sin; + +#endif +#if !defined(_WIN32) + if (addr->sa_len != sizeof(struct sockaddr_in)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + sin = (struct sockaddr_in *)addr; + if (p != NULL && (error = prison_remote_ip4(p->td_ucred, &sin->sin_addr)) != 0) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + return (error); + } +#endif + break; + } +#endif + default: + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EAFNOSUPPORT); + return (EAFNOSUPPORT); + } +#endif + SCTP_INP_INCR_REF(inp); + SCTP_ASOC_CREATE_LOCK(inp); + create_lock_on = 1; +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + + if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { + /* Should I really unlock ? */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EFAULT); + error = EFAULT; + goto out_now; + } +#ifdef INET6 + if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) && + (addr->sa_family == AF_INET6)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_now; + } +#endif +#if defined(__Userspace__) + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_CONN) && + (addr->sa_family != AF_CONN)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_now; + } +#endif + if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { + /* Bind a ephemeral port */ + error = sctp_inpcb_bind(so, NULL, NULL, p); + if (error) { + goto out_now; + } + } + /* Now do we connect? */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) && + (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE))) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_now; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && + (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) { + /* We are already connected AND the TCP model */ + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE); + error = EADDRINUSE; + goto out_now; + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { + SCTP_INP_RLOCK(inp); + stcb = LIST_FIRST(&inp->sctp_asoc_list); + SCTP_INP_RUNLOCK(inp); + } else { + /* We increment here since sctp_findassociation_ep_addr() will + * do a decrement if it finds the stcb as long as the locked + * tcb (last argument) is NOT a TCB.. aka NULL. + */ + SCTP_INP_INCR_REF(inp); + stcb = sctp_findassociation_ep_addr(&inp, addr, NULL, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_DECR_REF(inp); + } else { + SCTP_TCB_UNLOCK(stcb); + } + } + if (stcb != NULL) { + /* Already have or am bring up an association */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); + error = EALREADY; + goto out_now; + } + + vrf_id = inp->def_vrf_id; +#ifdef SCTP_MVRF + for (i = 0; i < inp->num_vrfs; i++) { + if (vrf_id == inp->m_vrf_ids[i]) { + fnd = 1; + break; + } + } + if (!fnd) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_now; + } +#endif + /* We are GOOD to go */ + stcb = sctp_aloc_assoc_connected(inp, addr, &error, 0, 0, vrf_id, + inp->sctp_ep.pre_open_stream_count, + inp->sctp_ep.port, p, + SCTP_INITIALIZE_AUTH_PARAMS); + if (stcb == NULL) { + /* Gak! no memory */ + goto out_now; + } + SCTP_SET_STATE(stcb, SCTP_STATE_COOKIE_WAIT); + (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); + + sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); + SCTP_TCB_UNLOCK(stcb); + out_now: +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + if (create_lock_on) { + SCTP_ASOC_CREATE_UNLOCK(inp); + } + SCTP_INP_DECR_REF(inp); + return (error); +} +#endif + +#if defined(__Userspace__) +int +sctpconn_connect(struct socket *so, struct sockaddr *addr) +{ +#ifdef SCTP_MVRF + int i, fnd = 0; +#endif + void *p = NULL; + int error = 0; + int create_lock_on = 0; + uint32_t vrf_id; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb = NULL; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + /* I made the same as TCP since we are not setup? */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (ECONNRESET); + } + if (addr == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return EINVAL; + } + switch (addr->sa_family) { +#ifdef INET + case AF_INET: +#ifdef HAVE_SA_LEN + if (addr->sa_len != sizeof(struct sockaddr_in)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: +#ifdef HAVE_SA_LEN + if (addr->sa_len != sizeof(struct sockaddr_in6)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } +#endif + break; +#endif + case AF_CONN: +#ifdef HAVE_SA_LEN + if (addr->sa_len != sizeof(struct sockaddr_conn)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (EINVAL); + } +#endif + break; + default: + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EAFNOSUPPORT); + return (EAFNOSUPPORT); + } + SCTP_INP_INCR_REF(inp); + SCTP_ASOC_CREATE_LOCK(inp); + create_lock_on = 1; + + + if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { + /* Should I really unlock ? */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EFAULT); + error = EFAULT; + goto out_now; + } +#ifdef INET6 + if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) && + (addr->sa_family == AF_INET6)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_now; + } +#endif + if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { + /* Bind a ephemeral port */ + error = sctp_inpcb_bind(so, NULL, NULL, p); + if (error) { + goto out_now; + } + } + /* Now do we connect? */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) && + (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE))) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_now; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && + (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) { + /* We are already connected AND the TCP model */ + SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE); + error = EADDRINUSE; + goto out_now; + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { + SCTP_INP_RLOCK(inp); + stcb = LIST_FIRST(&inp->sctp_asoc_list); + SCTP_INP_RUNLOCK(inp); + } else { + /* We increment here since sctp_findassociation_ep_addr() will + * do a decrement if it finds the stcb as long as the locked + * tcb (last argument) is NOT a TCB.. aka NULL. + */ + SCTP_INP_INCR_REF(inp); + stcb = sctp_findassociation_ep_addr(&inp, addr, NULL, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_DECR_REF(inp); + } else { + SCTP_TCB_UNLOCK(stcb); + } + } + if (stcb != NULL) { + /* Already have or am bring up an association */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); + error = EALREADY; + goto out_now; + } + + vrf_id = inp->def_vrf_id; +#ifdef SCTP_MVRF + for (i = 0; i < inp->num_vrfs; i++) { + if (vrf_id == inp->m_vrf_ids[i]) { + fnd = 1; + break; + } + } + if (!fnd) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + goto out_now; + } +#endif + /* We are GOOD to go */ + stcb = sctp_aloc_assoc_connected(inp, addr, &error, 0, 0, vrf_id, + inp->sctp_ep.pre_open_stream_count, + inp->sctp_ep.port, p, + SCTP_INITIALIZE_AUTH_PARAMS); + if (stcb == NULL) { + /* Gak! no memory */ + goto out_now; + } + SCTP_SET_STATE(stcb, SCTP_STATE_COOKIE_WAIT); + (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); + + sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); + SCTP_TCB_UNLOCK(stcb); + out_now: + if (create_lock_on) { + SCTP_ASOC_CREATE_UNLOCK(inp); + } + + SCTP_INP_DECR_REF(inp); + return (error); +} +#endif +int +#if defined(__Userspace__) +sctp_listen(struct socket *so, int backlog, struct proc *p) +#elif defined(__FreeBSD__) +sctp_listen(struct socket *so, int backlog, struct thread *p) +#elif defined(_WIN32) +sctp_listen(struct socket *so, int backlog, PKTHREAD p) +#else +sctp_listen(struct socket *so, struct proc *p) +#endif +{ + /* + * Note this module depends on the protocol processing being called + * AFTER any socket level flags and backlog are applied to the + * socket. The traditional way that the socket flags are applied is + * AFTER protocol processing. We have made a change to the + * sys/kern/uipc_socket.c module to reverse this but this MUST be in + * place if the socket API for SCTP is to work properly. + */ + + int error = 0; + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + /* I made the same as TCP since we are not setup? */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (ECONNRESET); + } + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) { + /* See if we have a listener */ + struct sctp_inpcb *tinp; + union sctp_sockstore store; + + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { + /* not bound all */ + struct sctp_laddr *laddr; + + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + memcpy(&store, &laddr->ifa->address, sizeof(store)); + switch (store.sa.sa_family) { +#ifdef INET + case AF_INET: + store.sin.sin_port = inp->sctp_lport; + break; +#endif +#ifdef INET6 + case AF_INET6: + store.sin6.sin6_port = inp->sctp_lport; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + store.sconn.sconn_port = inp->sctp_lport; + break; +#endif + default: + break; + } + tinp = sctp_pcb_findep(&store.sa, 0, 0, inp->def_vrf_id); + if (tinp && (tinp != inp) && + ((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) == 0) && + ((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (SCTP_IS_LISTENING(tinp))) { + /* we have a listener already and its not this inp. */ + SCTP_INP_DECR_REF(tinp); + return (EADDRINUSE); + } else if (tinp) { + SCTP_INP_DECR_REF(tinp); + } + } + } else { + /* Setup a local addr bound all */ + memset(&store, 0, sizeof(store)); +#ifdef INET6 + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + store.sa.sa_family = AF_INET6; +#ifdef HAVE_SA_LEN + store.sa.sa_len = sizeof(struct sockaddr_in6); +#endif + } +#endif +#if defined(__Userspace__) + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_CONN) { + store.sa.sa_family = AF_CONN; +#ifdef HAVE_SA_LEN + store.sa.sa_len = sizeof(struct sockaddr_conn); +#endif + } +#endif +#ifdef INET +#if defined(__Userspace__) + if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) && + ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_CONN) == 0)) { +#else + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { +#endif + store.sa.sa_family = AF_INET; +#ifdef HAVE_SA_LEN + store.sa.sa_len = sizeof(struct sockaddr_in); +#endif + } +#endif + switch (store.sa.sa_family) { +#ifdef INET + case AF_INET: + store.sin.sin_port = inp->sctp_lport; + break; +#endif +#ifdef INET6 + case AF_INET6: + store.sin6.sin6_port = inp->sctp_lport; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + store.sconn.sconn_port = inp->sctp_lport; + break; +#endif + default: + break; + } + tinp = sctp_pcb_findep(&store.sa, 0, 0, inp->def_vrf_id); + if (tinp && (tinp != inp) && + ((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) == 0) && + ((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (SCTP_IS_LISTENING(tinp))) { + /* we have a listener already and its not this inp. */ + SCTP_INP_DECR_REF(tinp); + return (EADDRINUSE); + } else if (tinp) { + SCTP_INP_DECR_REF(tinp); + } + } + } + SCTP_INP_INFO_WLOCK(); + SCTP_INP_WLOCK(inp); +#ifdef SCTP_LOCK_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) { + sctp_log_lock(inp, (struct sctp_tcb *)NULL, SCTP_LOG_LOCK_SOCK); + } +#endif + if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { + /* The unlucky case + * - We are in the tcp pool with this guy. + * - Someone else is in the main inp slot. + * - We must move this guy (the listener) to the main slot + * - We must then move the guy that was listener to the TCP Pool. + */ + if (sctp_swap_inpcb_for_listen(inp)) { + error = EADDRINUSE; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + goto out; + } + } +#if defined(__FreeBSD__) || defined(__Userspace__) + SOCK_LOCK(so); + error = solisten_proto_check(so); + if (error) { + SOCK_UNLOCK(so); + goto out; + } +#endif + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && + (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) { + SOCK_UNLOCK(so); +#if defined(__FreeBSD__) && !defined(__Userspace__) + solisten_proto_abort(so); +#endif + error = EADDRINUSE; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + goto out; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && + ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) || + (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED))) { + SOCK_UNLOCK(so); +#if defined(__FreeBSD__) && !defined(__Userspace__) + solisten_proto_abort(so); +#endif + error = EINVAL; + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); + goto out; + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { + if ((error = sctp_inpcb_bind_locked(inp, NULL, NULL, p))) { + SOCK_UNLOCK(so); +#if defined(__FreeBSD__) && !defined(__Userspace__) + solisten_proto_abort(so); +#endif + /* bind error, probably perm */ + goto out; + } + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) == 0) { + solisten_proto(so, backlog); + SOCK_UNLOCK(so); + inp->sctp_flags |= SCTP_PCB_FLAGS_ACCEPTING; + } else { + solisten_proto_abort(so); + SOCK_UNLOCK(so); + if (backlog > 0) { + inp->sctp_flags |= SCTP_PCB_FLAGS_ACCEPTING; + } else { + inp->sctp_flags &= ~SCTP_PCB_FLAGS_ACCEPTING; + } + } +#elif defined(_WIN32) || defined(__Userspace__) + solisten_proto(so, backlog); +#endif +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) { + /* remove the ACCEPTCONN flag for one-to-many sockets */ +#if defined(__Userspace__) + so->so_options &= ~SCTP_SO_ACCEPTCONN; +#else + so->so_options &= ~SO_ACCEPTCONN; +#endif + } + SOCK_UNLOCK(so); + if (backlog > 0) { + inp->sctp_flags |= SCTP_PCB_FLAGS_ACCEPTING; + } else { + inp->sctp_flags &= ~SCTP_PCB_FLAGS_ACCEPTING; + } +#endif +out: + SCTP_INP_WUNLOCK(inp); + SCTP_INP_INFO_WUNLOCK(); + return (error); +} + +static int sctp_defered_wakeup_cnt = 0; + +int +sctp_accept(struct socket *so, struct sockaddr **addr) +{ + struct sctp_tcb *stcb; + struct sctp_inpcb *inp; + union sctp_sockstore store; +#ifdef INET6 +#if defined(SCTP_KAME) && defined(SCTP_EMBEDDED_V6_SCOPE) + int error; +#endif +#endif + inp = (struct sctp_inpcb *)so->so_pcb; + + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (ECONNRESET); + } + SCTP_INP_WLOCK(inp); + if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) { + SCTP_INP_WUNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); + return (EOPNOTSUPP); + } + if (so->so_state & SS_ISDISCONNECTED) { + SCTP_INP_WUNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ECONNABORTED); + return (ECONNABORTED); + } + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb == NULL) { + SCTP_INP_WUNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (ECONNRESET); + } + SCTP_TCB_LOCK(stcb); + store = stcb->asoc.primary_destination->ro._l_addr; + SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE); + /* Wake any delayed sleep action */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { + inp->sctp_flags &= ~SCTP_PCB_FLAGS_DONT_WAKE; + if (inp->sctp_flags & SCTP_PCB_FLAGS_WAKEOUTPUT) { + inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT; + SOCKBUF_LOCK(&inp->sctp_socket->so_snd); + if (sowriteable(inp->sctp_socket)) { +#if defined(__Userspace__) + /*__Userspace__ calling sowwakup_locked because of SOCKBUF_LOCK above. */ +#endif +#if defined(__FreeBSD__) || defined(_WIN32) || defined(__Userspace__) + sowwakeup_locked(inp->sctp_socket); +#else +#if defined(__APPLE__) + /* socket is locked */ +#endif + sowwakeup(inp->sctp_socket); +#endif + } else { + SOCKBUF_UNLOCK(&inp->sctp_socket->so_snd); + } + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_WAKEINPUT) { + inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT; + SOCKBUF_LOCK(&inp->sctp_socket->so_rcv); + if (soreadable(inp->sctp_socket)) { + sctp_defered_wakeup_cnt++; +#if defined(__Userspace__) + /*__Userspace__ calling sorwakup_locked because of SOCKBUF_LOCK above */ +#endif +#if defined(__FreeBSD__) || defined(_WIN32) || defined(__Userspace__) + sorwakeup_locked(inp->sctp_socket); +#else +#if defined(__APPLE__) + /* socket is locked */ +#endif + sorwakeup(inp->sctp_socket); +#endif + } else { + SOCKBUF_UNLOCK(&inp->sctp_socket->so_rcv); + } + } + } + SCTP_INP_WUNLOCK(inp); + if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_19); + } else { + SCTP_TCB_UNLOCK(stcb); + } + switch (store.sa.sa_family) { +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin; + + SCTP_MALLOC_SONAME(sin, struct sockaddr_in *, sizeof *sin); + if (sin == NULL) + return (ENOMEM); + sin->sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + sin->sin_len = sizeof(*sin); +#endif + sin->sin_port = store.sin.sin_port; + sin->sin_addr = store.sin.sin_addr; + *addr = (struct sockaddr *)sin; + break; + } +#endif +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6; + + SCTP_MALLOC_SONAME(sin6, struct sockaddr_in6 *, sizeof *sin6); + if (sin6 == NULL) + return (ENOMEM); + sin6->sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6->sin6_len = sizeof(*sin6); +#endif + sin6->sin6_port = store.sin6.sin6_port; + sin6->sin6_addr = store.sin6.sin6_addr; +#if defined(SCTP_EMBEDDED_V6_SCOPE) +#ifdef SCTP_KAME + if ((error = sa6_recoverscope(sin6)) != 0) { + SCTP_FREE_SONAME(sin6); + return (error); + } +#else + if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) + /* + * sin6->sin6_scope_id = + * ntohs(sin6->sin6_addr.s6_addr16[1]); + */ + in6_recoverscope(sin6, &sin6->sin6_addr, NULL); /* skip ifp check */ + else + sin6->sin6_scope_id = 0; /* XXX */ +#endif /* SCTP_KAME */ +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + *addr = (struct sockaddr *)sin6; + break; + } +#endif +#if defined(__Userspace__) + case AF_CONN: + { + struct sockaddr_conn *sconn; + + SCTP_MALLOC_SONAME(sconn, struct sockaddr_conn *, sizeof(struct sockaddr_conn)); + if (sconn == NULL) { + return (ENOMEM); + } + sconn->sconn_family = AF_CONN; +#ifdef HAVE_SCONN_LEN + sconn->sconn_len = sizeof(struct sockaddr_conn); +#endif + sconn->sconn_port = store.sconn.sconn_port; + sconn->sconn_addr = store.sconn.sconn_addr; + *addr = (struct sockaddr *)sconn; + break; + } +#endif + default: + /* TSNH */ + break; + } + return (0); +} + +#ifdef INET +int +#if !defined(__Userspace__) +sctp_ingetaddr(struct socket *so, struct sockaddr **addr) +{ + struct sockaddr_in *sin; +#else +sctp_ingetaddr(struct socket *so, struct mbuf *nam) +{ + struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *); +#endif + uint32_t vrf_id; + struct sctp_inpcb *inp; + struct sctp_ifa *sctp_ifa; + + /* + * Do the malloc first in case it blocks. + */ +#if !defined(__Userspace__) + SCTP_MALLOC_SONAME(sin, struct sockaddr_in *, sizeof *sin); + if (sin == NULL) + return (ENOMEM); +#else + SCTP_BUF_LEN(nam) = sizeof(*sin); + memset(sin, 0, sizeof(*sin)); +#endif + sin->sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + sin->sin_len = sizeof(*sin); +#endif + inp = (struct sctp_inpcb *)so->so_pcb; + if (!inp) { +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin); +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (ECONNRESET); + } + SCTP_INP_RLOCK(inp); + sin->sin_port = inp->sctp_lport; + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { + struct sctp_tcb *stcb; + struct sockaddr_in *sin_a; + struct sctp_nets *net; + int fnd; + + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb == NULL) { + goto notConn; + } + fnd = 0; + sin_a = NULL; + SCTP_TCB_LOCK(stcb); + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + sin_a = (struct sockaddr_in *)&net->ro._l_addr; + if (sin_a == NULL) + /* this will make coverity happy */ + continue; + + if (sin_a->sin_family == AF_INET) { + fnd = 1; + break; + } + } + if ((!fnd) || (sin_a == NULL)) { + /* punt */ + SCTP_TCB_UNLOCK(stcb); + goto notConn; + } + + vrf_id = inp->def_vrf_id; + sctp_ifa = sctp_source_address_selection(inp, + stcb, + (sctp_route_t *)&net->ro, + net, 0, vrf_id); + if (sctp_ifa) { + sin->sin_addr = sctp_ifa->address.sin.sin_addr; + sctp_free_ifa(sctp_ifa); + } + SCTP_TCB_UNLOCK(stcb); + } else { + /* For the bound all case you get back 0 */ + notConn: + sin->sin_addr.s_addr = 0; + } + + } else { + /* Take the first IPv4 address in the list */ + struct sctp_laddr *laddr; + int fnd = 0; + + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa->address.sa.sa_family == AF_INET) { + struct sockaddr_in *sin_a; + + sin_a = &laddr->ifa->address.sin; + sin->sin_addr = sin_a->sin_addr; + fnd = 1; + break; + } + } + if (!fnd) { +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin); +#endif + SCTP_INP_RUNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); + return (ENOENT); + } + } + SCTP_INP_RUNLOCK(inp); +#if !defined(__Userspace__) + (*addr) = (struct sockaddr *)sin; +#endif + return (0); +} + +int +#if !defined(__Userspace__) +sctp_peeraddr(struct socket *so, struct sockaddr **addr) +{ + struct sockaddr_in *sin; +#else +sctp_peeraddr(struct socket *so, struct mbuf *nam) +{ + struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *); + +#endif + int fnd; + struct sockaddr_in *sin_a; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; + struct sctp_nets *net; + + /* Do the malloc first in case it blocks. */ +#if !defined(__Userspace__) + SCTP_MALLOC_SONAME(sin, struct sockaddr_in *, sizeof *sin); + if (sin == NULL) + return (ENOMEM); +#else + SCTP_BUF_LEN(nam) = sizeof(*sin); + memset(sin, 0, sizeof(*sin)); +#endif + sin->sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + sin->sin_len = sizeof(*sin); +#endif + + inp = (struct sctp_inpcb *)so->so_pcb; + if ((inp == NULL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0)) { + /* UDP type and listeners will drop out here */ +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin); +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); + return (ENOTCONN); + } + SCTP_INP_RLOCK(inp); + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb) { + SCTP_TCB_LOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + if (stcb == NULL) { +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin); +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + return (ECONNRESET); + } + fnd = 0; + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + sin_a = (struct sockaddr_in *)&net->ro._l_addr; + if (sin_a->sin_family == AF_INET) { + fnd = 1; + sin->sin_port = stcb->rport; + sin->sin_addr = sin_a->sin_addr; + break; + } + } + SCTP_TCB_UNLOCK(stcb); + if (!fnd) { + /* No IPv4 address */ +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin); +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); + return (ENOENT); + } +#if !defined(__Userspace__) + (*addr) = (struct sockaddr *)sin; +#endif + return (0); +} + +#if !defined(__Userspace__) +#if defined(__FreeBSD__) +#define SCTP_PROTOSW \ + .pr_protocol = IPPROTO_SCTP, \ + .pr_ctloutput = sctp_ctloutput, \ + .pr_abort = sctp_abort, \ + .pr_accept = sctp_accept, \ + .pr_attach = sctp_attach, \ + .pr_bind = sctp_bind, \ + .pr_connect = sctp_connect, \ + .pr_control = in_control, \ + .pr_close = sctp_close, \ + .pr_detach = sctp_close, \ + .pr_sopoll = sopoll_generic, \ + .pr_flush = sctp_flush, \ + .pr_disconnect = sctp_disconnect, \ + .pr_listen = sctp_listen, \ + .pr_peeraddr = sctp_peeraddr, \ + .pr_send = sctp_sendm, \ + .pr_shutdown = sctp_shutdown, \ + .pr_sockaddr = sctp_ingetaddr, \ + .pr_sosend = sctp_sosend, \ + .pr_soreceive = sctp_soreceive \ + +struct protosw sctp_seqpacket_protosw = { + .pr_type = SOCK_SEQPACKET, + .pr_flags = PR_WANTRCVD, + SCTP_PROTOSW +}; + +struct protosw sctp_stream_protosw = { + .pr_type = SOCK_STREAM, + .pr_flags = PR_CONNREQUIRED | PR_WANTRCVD, + SCTP_PROTOSW +}; +#else +struct pr_usrreqs sctp_usrreqs = { +#if defined(__APPLE__) + .pru_abort = sctp_abort, + .pru_accept = sctp_accept, + .pru_attach = sctp_attach, + .pru_bind = sctp_bind, + .pru_connect = sctp_connect, + .pru_connect2 = pru_connect2_notsupp, + .pru_control = in_control, + .pru_detach = sctp_detach, + .pru_disconnect = sctp_disconnect, + .pru_listen = sctp_listen, + .pru_peeraddr = sctp_peeraddr, + .pru_rcvd = NULL, + .pru_rcvoob = pru_rcvoob_notsupp, + .pru_send = sctp_sendm, + .pru_sense = pru_sense_null, + .pru_shutdown = sctp_shutdown, + .pru_sockaddr = sctp_ingetaddr, + .pru_sosend = sctp_sosend, + .pru_soreceive = sctp_soreceive, + .pru_sopoll = sopoll +#elif defined(_WIN32) && !defined(__Userspace__) + sctp_abort, + sctp_accept, + sctp_attach, + sctp_bind, + sctp_connect, + pru_connect2_notsupp, + NULL, + NULL, + sctp_disconnect, + sctp_listen, + sctp_peeraddr, + NULL, + pru_rcvoob_notsupp, + NULL, + pru_sense_null, + sctp_shutdown, + sctp_flush, + sctp_ingetaddr, + sctp_sosend, + sctp_soreceive, + sopoll_generic, + NULL, + sctp_close +#endif +}; +#endif +#endif +#endif + +#if defined(__Userspace__) +int +register_recv_cb(struct socket *so, + int (*receive_cb)(struct socket *sock, union sctp_sockstore addr, void *data, + size_t datalen, struct sctp_rcvinfo, int flags, void *ulp_info)) +{ + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *) so->so_pcb; + if (inp == NULL) { + return (0); + } + SCTP_INP_WLOCK(inp); + inp->recv_callback = receive_cb; + SCTP_INP_WUNLOCK(inp); + return (1); +} + +int +register_send_cb(struct socket *so, uint32_t sb_threshold, int (*send_cb)(struct socket *sock, uint32_t sb_free, void *ulp_info)) +{ + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *) so->so_pcb; + if (inp == NULL) { + return (0); + } + SCTP_INP_WLOCK(inp); + inp->send_callback = send_cb; + inp->send_sb_threshold = sb_threshold; + SCTP_INP_WUNLOCK(inp); + /* FIXME change to current amount free. This will be the full buffer + * the first time this is registered but it could be only a portion + * of the send buffer if this is called a second time e.g. if the + * threshold changes. + */ + return (1); +} + +int +register_ulp_info (struct socket *so, void *ulp_info) +{ + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *) so->so_pcb; + if (inp == NULL) { + return (0); + } + SCTP_INP_WLOCK(inp); + inp->ulp_info = ulp_info; + SCTP_INP_WUNLOCK(inp); + return (1); +} + +int +retrieve_ulp_info (struct socket *so, void **pulp_info) +{ + struct sctp_inpcb *inp; + + if (pulp_info == NULL) { + return (0); + } + + inp = (struct sctp_inpcb *) so->so_pcb; + if (inp == NULL) { + return (0); + } + SCTP_INP_RLOCK(inp); + *pulp_info = inp->ulp_info; + SCTP_INP_RUNLOCK(inp); + return (1); +} +#endif diff --git a/netwerk/sctp/src/netinet/sctp_var.h b/netwerk/sctp/src/netinet/sctp_var.h new file mode 100644 index 0000000000..de26bb1d39 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctp_var.h @@ -0,0 +1,453 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_VAR_H_ +#define _NETINET_SCTP_VAR_H_ + +#include <netinet/sctp_uio.h> + +#if defined(_KERNEL) || defined(__Userspace__) + +#if !defined(__Userspace__) +#if defined(__FreeBSD__) +extern struct protosw sctp_seqpacket_protosw, sctp_stream_protosw; +#else +extern struct pr_usrreqs sctp_usrreqs; +#endif +#endif + +#define sctp_feature_on(inp, feature) (inp->sctp_features |= feature) +#define sctp_feature_off(inp, feature) (inp->sctp_features &= ~feature) +#define sctp_is_feature_on(inp, feature) ((inp->sctp_features & feature) == feature) +#define sctp_is_feature_off(inp, feature) ((inp->sctp_features & feature) == 0) + +#define sctp_stcb_feature_on(inp, stcb, feature) {\ + if (stcb) { \ + stcb->asoc.sctp_features |= feature; \ + } else if (inp) { \ + inp->sctp_features |= feature; \ + } \ +} +#define sctp_stcb_feature_off(inp, stcb, feature) {\ + if (stcb) { \ + stcb->asoc.sctp_features &= ~feature; \ + } else if (inp) { \ + inp->sctp_features &= ~feature; \ + } \ +} +#define sctp_stcb_is_feature_on(inp, stcb, feature) \ + (((stcb != NULL) && \ + ((stcb->asoc.sctp_features & feature) == feature)) || \ + ((stcb == NULL) && (inp != NULL) && \ + ((inp->sctp_features & feature) == feature))) +#define sctp_stcb_is_feature_off(inp, stcb, feature) \ + (((stcb != NULL) && \ + ((stcb->asoc.sctp_features & feature) == 0)) || \ + ((stcb == NULL) && (inp != NULL) && \ + ((inp->sctp_features & feature) == 0)) || \ + ((stcb == NULL) && (inp == NULL))) + +/* managing mobility_feature in inpcb (by micchie) */ +#define sctp_mobility_feature_on(inp, feature) (inp->sctp_mobility_features |= feature) +#define sctp_mobility_feature_off(inp, feature) (inp->sctp_mobility_features &= ~feature) +#define sctp_is_mobility_feature_on(inp, feature) (inp->sctp_mobility_features & feature) +#define sctp_is_mobility_feature_off(inp, feature) ((inp->sctp_mobility_features & feature) == 0) + +#define sctp_maxspace(sb) (max((sb)->sb_hiwat,SCTP_MINIMAL_RWND)) + +#define sctp_sbspace(asoc, sb) ((long) ((sctp_maxspace(sb) > (asoc)->sb_cc) ? (sctp_maxspace(sb) - (asoc)->sb_cc) : 0)) + +#define sctp_sbspace_failedmsgs(sb) ((long) ((sctp_maxspace(sb) > SCTP_SBAVAIL(sb)) ? (sctp_maxspace(sb) - SCTP_SBAVAIL(sb)) : 0)) + +#define sctp_sbspace_sub(a,b) (((a) > (b)) ? ((a) - (b)) : 0) + +/* + * I tried to cache the readq entries at one point. But the reality + * is that it did not add any performance since this meant we had to + * lock the STCB on read. And at that point once you have to do an + * extra lock, it really does not matter if the lock is in the ZONE + * stuff or in our code. Note that this same problem would occur with + * an mbuf cache as well so it is not really worth doing, at least + * right now :-D + */ +#ifdef INVARIANTS +#define sctp_free_a_readq(_stcb, _readq) { \ + if ((_readq)->on_strm_q) \ + panic("On strm q stcb:%p readq:%p", (_stcb), (_readq)); \ + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), (_readq)); \ + SCTP_DECR_READQ_COUNT(); \ +} +#else +#define sctp_free_a_readq(_stcb, _readq) { \ + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), (_readq)); \ + SCTP_DECR_READQ_COUNT(); \ +} +#endif + +#define sctp_alloc_a_readq(_stcb, _readq) { \ + (_readq) = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_readq), struct sctp_queued_to_read); \ + if ((_readq)) { \ + SCTP_INCR_READQ_COUNT(); \ + } \ +} + +#define sctp_free_a_strmoq(_stcb, _strmoq, _so_locked) { \ + if ((_strmoq)->holds_key_ref) { \ + sctp_auth_key_release(stcb, sp->auth_keyid, _so_locked); \ + (_strmoq)->holds_key_ref = 0; \ + } \ + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_strmoq), (_strmoq)); \ + SCTP_DECR_STRMOQ_COUNT(); \ +} + +#define sctp_alloc_a_strmoq(_stcb, _strmoq) { \ + (_strmoq) = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_strmoq), struct sctp_stream_queue_pending); \ + if ((_strmoq)) { \ + memset(_strmoq, 0, sizeof(struct sctp_stream_queue_pending)); \ + SCTP_INCR_STRMOQ_COUNT(); \ + (_strmoq)->holds_key_ref = 0; \ + } \ +} + +#define sctp_free_a_chunk(_stcb, _chk, _so_locked) { \ + if ((_chk)->holds_key_ref) {\ + sctp_auth_key_release((_stcb), (_chk)->auth_keyid, _so_locked); \ + (_chk)->holds_key_ref = 0; \ + } \ + if (_stcb) { \ + SCTP_TCB_LOCK_ASSERT((_stcb)); \ + if ((_chk)->whoTo) { \ + sctp_free_remote_addr((_chk)->whoTo); \ + (_chk)->whoTo = NULL; \ + } \ + if (((_stcb)->asoc.free_chunk_cnt > SCTP_BASE_SYSCTL(sctp_asoc_free_resc_limit)) || \ + (SCTP_BASE_INFO(ipi_free_chunks) > SCTP_BASE_SYSCTL(sctp_system_free_resc_limit))) { \ + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), (_chk)); \ + SCTP_DECR_CHK_COUNT(); \ + } else { \ + TAILQ_INSERT_TAIL(&(_stcb)->asoc.free_chunks, (_chk), sctp_next); \ + (_stcb)->asoc.free_chunk_cnt++; \ + atomic_add_int(&SCTP_BASE_INFO(ipi_free_chunks), 1); \ + } \ + } else { \ + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), (_chk)); \ + SCTP_DECR_CHK_COUNT(); \ + } \ +} + +#define sctp_alloc_a_chunk(_stcb, _chk) { \ + if (TAILQ_EMPTY(&(_stcb)->asoc.free_chunks)) { \ + (_chk) = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_chunk), struct sctp_tmit_chunk); \ + if ((_chk)) { \ + SCTP_INCR_CHK_COUNT(); \ + (_chk)->whoTo = NULL; \ + (_chk)->holds_key_ref = 0; \ + } \ + } else { \ + (_chk) = TAILQ_FIRST(&(_stcb)->asoc.free_chunks); \ + TAILQ_REMOVE(&(_stcb)->asoc.free_chunks, (_chk), sctp_next); \ + atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1); \ + (_chk)->holds_key_ref = 0; \ + SCTP_STAT_INCR(sctps_cached_chk); \ + (_stcb)->asoc.free_chunk_cnt--; \ + } \ +} + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#define sctp_free_remote_addr(__net) { \ + if ((__net)) { \ + if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&(__net)->ref_count)) { \ + RO_NHFREE(&(__net)->ro); \ + if ((__net)->src_addr_selected) { \ + sctp_free_ifa((__net)->ro._s_addr); \ + (__net)->ro._s_addr = NULL; \ + } \ + (__net)->src_addr_selected = 0; \ + (__net)->dest_state &= ~SCTP_ADDR_REACHABLE; \ + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_net), (__net)); \ + SCTP_DECR_RADDR_COUNT(); \ + } \ + } \ +} + +#define sctp_sbfree(ctl, stcb, sb, m) { \ + SCTP_SAVE_ATOMIC_DECREMENT(&(sb)->sb_cc, SCTP_BUF_LEN((m))); \ + SCTP_SAVE_ATOMIC_DECREMENT(&(sb)->sb_mbcnt, MSIZE); \ + if (((ctl)->do_not_ref_stcb == 0) && stcb) {\ + SCTP_SAVE_ATOMIC_DECREMENT(&(stcb)->asoc.sb_cc, SCTP_BUF_LEN((m))); \ + SCTP_SAVE_ATOMIC_DECREMENT(&(stcb)->asoc.my_rwnd_control_len, MSIZE); \ + } \ + if (SCTP_BUF_TYPE(m) != MT_DATA && SCTP_BUF_TYPE(m) != MT_HEADER && \ + SCTP_BUF_TYPE(m) != MT_OOBDATA) \ + atomic_subtract_int(&(sb)->sb_ctl,SCTP_BUF_LEN((m))); \ +} + +#define sctp_sballoc(stcb, sb, m) { \ + atomic_add_int(&(sb)->sb_cc,SCTP_BUF_LEN((m))); \ + atomic_add_int(&(sb)->sb_mbcnt, MSIZE); \ + if (stcb) { \ + atomic_add_int(&(stcb)->asoc.sb_cc, SCTP_BUF_LEN((m))); \ + atomic_add_int(&(stcb)->asoc.my_rwnd_control_len, MSIZE); \ + } \ + if (SCTP_BUF_TYPE(m) != MT_DATA && SCTP_BUF_TYPE(m) != MT_HEADER && \ + SCTP_BUF_TYPE(m) != MT_OOBDATA) \ + atomic_add_int(&(sb)->sb_ctl,SCTP_BUF_LEN((m))); \ +} +#else /* FreeBSD Version <= 500000 or non-FreeBSD */ +#define sctp_free_remote_addr(__net) { \ + if ((__net)) { \ + if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&(__net)->ref_count)) { \ + if ((__net)->ro.ro_rt) { \ + RTFREE((__net)->ro.ro_rt); \ + (__net)->ro.ro_rt = NULL; \ + } \ + if ((__net)->src_addr_selected) { \ + sctp_free_ifa((__net)->ro._s_addr); \ + (__net)->ro._s_addr = NULL; \ + } \ + (__net)->src_addr_selected = 0; \ + (__net)->dest_state &=~SCTP_ADDR_REACHABLE; \ + SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_net), (__net)); \ + SCTP_DECR_RADDR_COUNT(); \ + } \ + } \ +} + +#define sctp_sbfree(ctl, stcb, sb, m) { \ + SCTP_SAVE_ATOMIC_DECREMENT(&(sb)->sb_cc, SCTP_BUF_LEN((m))); \ + SCTP_SAVE_ATOMIC_DECREMENT(&(sb)->sb_mbcnt, MSIZE); \ + if (((ctl)->do_not_ref_stcb == 0) && stcb) { \ + SCTP_SAVE_ATOMIC_DECREMENT(&(stcb)->asoc.sb_cc, SCTP_BUF_LEN((m))); \ + SCTP_SAVE_ATOMIC_DECREMENT(&(stcb)->asoc.my_rwnd_control_len, MSIZE); \ + } \ +} + +#define sctp_sballoc(stcb, sb, m) { \ + atomic_add_int(&(sb)->sb_cc, SCTP_BUF_LEN((m))); \ + atomic_add_int(&(sb)->sb_mbcnt, MSIZE); \ + if (stcb) { \ + atomic_add_int(&(stcb)->asoc.sb_cc, SCTP_BUF_LEN((m))); \ + atomic_add_int(&(stcb)->asoc.my_rwnd_control_len, MSIZE); \ + } \ +} +#endif + +#define sctp_ucount_incr(val) { \ + val++; \ +} + +#define sctp_ucount_decr(val) { \ + if (val > 0) { \ + val--; \ + } else { \ + val = 0; \ + } \ +} + +#define sctp_mbuf_crush(data) do { \ + struct mbuf *_m; \ + _m = (data); \ + while (_m && (SCTP_BUF_LEN(_m) == 0)) { \ + (data) = SCTP_BUF_NEXT(_m); \ + SCTP_BUF_NEXT(_m) = NULL; \ + sctp_m_free(_m); \ + _m = (data); \ + } \ +} while (0) + +#define sctp_flight_size_decrease(tp1) do { \ + if (tp1->whoTo->flight_size >= tp1->book_size) \ + tp1->whoTo->flight_size -= tp1->book_size; \ + else \ + tp1->whoTo->flight_size = 0; \ +} while (0) + +#define sctp_flight_size_increase(tp1) do { \ + (tp1)->whoTo->flight_size += (tp1)->book_size; \ +} while (0) + +#ifdef SCTP_FS_SPEC_LOG +#define sctp_total_flight_decrease(stcb, tp1) do { \ + if (stcb->asoc.fs_index > SCTP_FS_SPEC_LOG_SIZE) \ + stcb->asoc.fs_index = 0;\ + stcb->asoc.fslog[stcb->asoc.fs_index].total_flight = stcb->asoc.total_flight; \ + stcb->asoc.fslog[stcb->asoc.fs_index].tsn = tp1->rec.data.tsn; \ + stcb->asoc.fslog[stcb->asoc.fs_index].book = tp1->book_size; \ + stcb->asoc.fslog[stcb->asoc.fs_index].sent = tp1->sent; \ + stcb->asoc.fslog[stcb->asoc.fs_index].incr = 0; \ + stcb->asoc.fslog[stcb->asoc.fs_index].decr = 1; \ + stcb->asoc.fs_index++; \ + tp1->window_probe = 0; \ + if (stcb->asoc.total_flight >= tp1->book_size) { \ + stcb->asoc.total_flight -= tp1->book_size; \ + if (stcb->asoc.total_flight_count > 0) \ + stcb->asoc.total_flight_count--; \ + } else { \ + stcb->asoc.total_flight = 0; \ + stcb->asoc.total_flight_count = 0; \ + } \ +} while (0) + +#define sctp_total_flight_increase(stcb, tp1) do { \ + if (stcb->asoc.fs_index > SCTP_FS_SPEC_LOG_SIZE) \ + stcb->asoc.fs_index = 0;\ + stcb->asoc.fslog[stcb->asoc.fs_index].total_flight = stcb->asoc.total_flight; \ + stcb->asoc.fslog[stcb->asoc.fs_index].tsn = tp1->rec.data.tsn; \ + stcb->asoc.fslog[stcb->asoc.fs_index].book = tp1->book_size; \ + stcb->asoc.fslog[stcb->asoc.fs_index].sent = tp1->sent; \ + stcb->asoc.fslog[stcb->asoc.fs_index].incr = 1; \ + stcb->asoc.fslog[stcb->asoc.fs_index].decr = 0; \ + stcb->asoc.fs_index++; \ + (stcb)->asoc.total_flight_count++; \ + (stcb)->asoc.total_flight += (tp1)->book_size; \ +} while (0) + +#else + +#define sctp_total_flight_decrease(stcb, tp1) do { \ + tp1->window_probe = 0; \ + if (stcb->asoc.total_flight >= tp1->book_size) { \ + stcb->asoc.total_flight -= tp1->book_size; \ + if (stcb->asoc.total_flight_count > 0) \ + stcb->asoc.total_flight_count--; \ + } else { \ + stcb->asoc.total_flight = 0; \ + stcb->asoc.total_flight_count = 0; \ + } \ +} while (0) + +#define sctp_total_flight_increase(stcb, tp1) do { \ + (stcb)->asoc.total_flight_count++; \ + (stcb)->asoc.total_flight += (tp1)->book_size; \ +} while (0) + +#endif + +#define SCTP_PF_ENABLED(_net) (_net->pf_threshold < _net->failure_threshold) +#define SCTP_NET_IS_PF(_net) (_net->pf_threshold < _net->error_count) + +struct sctp_nets; +struct sctp_inpcb; +struct sctp_tcb; +struct sctphdr; + +#if defined(__FreeBSD__) || defined(_WIN32) || defined(__Userspace__) +void sctp_close(struct socket *so); +#else +int sctp_detach(struct socket *so); +#endif +int sctp_disconnect(struct socket *so); +#if !defined(__Userspace__) +#if defined(__APPLE__) && !defined(APPLE_LEOPARD) && !defined(APPLE_SNOWLEOPARD) && !defined(APPLE_LION) && !defined(APPLE_MOUNTAINLION) && !defined(APPLE_ELCAPITAN) +void sctp_ctlinput(int, struct sockaddr *, void *, struct ifnet * SCTP_UNUSED); +#elif defined(__FreeBSD__) +ipproto_ctlinput_t sctp_ctlinput; +#else +void sctp_ctlinput(int, struct sockaddr *, void *); +#endif +int sctp_ctloutput(struct socket *, struct sockopt *); +#ifdef INET +void sctp_input_with_port(struct mbuf *, int, uint16_t); +#if defined(__FreeBSD__) && !defined(__Userspace__) +int sctp_input(struct mbuf **, int *, int); +#else +void sctp_input(struct mbuf *, int); +#endif +#endif +void sctp_pathmtu_adjustment(struct sctp_tcb *, uint32_t, bool); +#else +#if defined(__Userspace__) +void sctp_pathmtu_adjustment(struct sctp_tcb *, uint32_t, bool); +#else +void sctp_input(struct mbuf *,...); +#endif +void *sctp_ctlinput(int, struct sockaddr *, void *); +int sctp_ctloutput(int, struct socket *, int, int, struct mbuf **); +#endif +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) +void sctp_drain(void); +#endif +#if defined(__Userspace__) +void sctp_init(uint16_t, + int (*)(void *addr, void *buffer, size_t length, uint8_t tos, uint8_t set_df), + void (*)(const char *, ...), int start_threads); +#elif defined(__APPLE__) && (!defined(APPLE_LEOPARD) && !defined(APPLE_SNOWLEOPARD) &&!defined(APPLE_LION) && !defined(APPLE_MOUNTAINLION)) +void sctp_init(struct protosw *pp, struct domain *dp); +#else +#if !defined(__FreeBSD__) +void sctp_init(void); +#endif +void sctp_notify(struct sctp_inpcb *, struct sctp_tcb *, struct sctp_nets *, + uint8_t, uint8_t, uint16_t, uint32_t); +#endif +#if !defined(__FreeBSD__) && !defined(__Userspace__) +void sctp_finish(void); +#endif +#if defined(__FreeBSD__) || defined(_WIN32) || defined(__Userspace__) +int sctp_flush(struct socket *, int); +#endif +int sctp_shutdown(struct socket *); +int sctp_bindx(struct socket *, int, struct sockaddr_storage *, + int, int, struct proc *); +/* can't use sctp_assoc_t here */ +int sctp_peeloff(struct socket *, struct socket *, int, caddr_t, int *); +#if !defined(__Userspace__) +int sctp_ingetaddr(struct socket *, struct sockaddr **); +#else +int sctp_ingetaddr(struct socket *, struct mbuf *); +#endif +#if !defined(__Userspace__) +int sctp_peeraddr(struct socket *, struct sockaddr **); +#else +int sctp_peeraddr(struct socket *, struct mbuf *); +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) +int sctp_listen(struct socket *, int, struct thread *); +#elif defined(_WIN32) && !defined(__Userspace__) +int sctp_listen(struct socket *, int, PKTHREAD); +#elif defined(__Userspace__) +int sctp_listen(struct socket *, int, struct proc *); +#else +int sctp_listen(struct socket *, struct proc *); +#endif +int sctp_accept(struct socket *, struct sockaddr **); + +#endif /* _KERNEL */ + +#endif /* !_NETINET_SCTP_VAR_H_ */ diff --git a/netwerk/sctp/src/netinet/sctputil.c b/netwerk/sctp/src/netinet/sctputil.c new file mode 100644 index 0000000000..7c9e2edfe5 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctputil.c @@ -0,0 +1,8718 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_var.h> +#include <netinet/sctp_sysctl.h> +#ifdef INET6 +#if defined(__Userspace__) || defined(__FreeBSD__) +#include <netinet6/sctp6_var.h> +#endif +#endif +#include <netinet/sctp_header.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_uio.h> +#include <netinet/sctp_timer.h> +#include <netinet/sctp_indata.h> +#include <netinet/sctp_auth.h> +#include <netinet/sctp_asconf.h> +#include <netinet/sctp_bsd_addr.h> +#if defined(__Userspace__) +#include <netinet/sctp_constants.h> +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <netinet/sctp_kdtrace.h> +#if defined(INET6) || defined(INET) +#include <netinet/tcp_var.h> +#endif +#include <netinet/udp.h> +#include <netinet/udp_var.h> +#include <sys/proc.h> +#ifdef INET6 +#include <netinet/icmp6.h> +#endif +#endif + +#if defined(_WIN32) && !defined(__Userspace__) +#if !defined(SCTP_LOCAL_TRACE_BUF) +#include "eventrace_netinet.h" +#include "sctputil.tmh" /* this is the file that will be auto generated */ +#endif +#else +#ifndef KTR_SCTP +#define KTR_SCTP KTR_SUBSYS +#endif +#endif + +extern const struct sctp_cc_functions sctp_cc_functions[]; +extern const struct sctp_ss_functions sctp_ss_functions[]; + +void +sctp_sblog(struct sockbuf *sb, struct sctp_tcb *stcb, int from, int incr) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + sctp_clog.x.sb.stcb = stcb; + sctp_clog.x.sb.so_sbcc = SCTP_SBAVAIL(sb); + if (stcb) + sctp_clog.x.sb.stcb_sbcc = stcb->asoc.sb_cc; + else + sctp_clog.x.sb.stcb_sbcc = 0; + sctp_clog.x.sb.incr = incr; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_SB, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +void +sctp_log_closing(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int16_t loc) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + sctp_clog.x.close.inp = (void *)inp; + sctp_clog.x.close.sctp_flags = inp->sctp_flags; + if (stcb) { + sctp_clog.x.close.stcb = (void *)stcb; + sctp_clog.x.close.state = (uint16_t)stcb->asoc.state; + } else { + sctp_clog.x.close.stcb = 0; + sctp_clog.x.close.state = 0; + } + sctp_clog.x.close.loc = loc; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_CLOSE, + 0, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +void +rto_logging(struct sctp_nets *net, int from) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + memset(&sctp_clog, 0, sizeof(sctp_clog)); + sctp_clog.x.rto.net = (void *) net; + sctp_clog.x.rto.rtt = net->rtt / 1000; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_RTT, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +void +sctp_log_strm_del_alt(struct sctp_tcb *stcb, uint32_t tsn, uint16_t sseq, uint16_t stream, int from) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + sctp_clog.x.strlog.stcb = stcb; + sctp_clog.x.strlog.n_tsn = tsn; + sctp_clog.x.strlog.n_sseq = sseq; + sctp_clog.x.strlog.e_tsn = 0; + sctp_clog.x.strlog.e_sseq = 0; + sctp_clog.x.strlog.strm = stream; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_STRM, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +void +sctp_log_nagle_event(struct sctp_tcb *stcb, int action) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + sctp_clog.x.nagle.stcb = (void *)stcb; + sctp_clog.x.nagle.total_flight = stcb->asoc.total_flight; + sctp_clog.x.nagle.total_in_queue = stcb->asoc.total_output_queue_size; + sctp_clog.x.nagle.count_in_queue = stcb->asoc.chunks_on_out_queue; + sctp_clog.x.nagle.count_in_flight = stcb->asoc.total_flight_count; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_NAGLE, + action, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +void +sctp_log_sack(uint32_t old_cumack, uint32_t cumack, uint32_t tsn, uint16_t gaps, uint16_t dups, int from) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + sctp_clog.x.sack.cumack = cumack; + sctp_clog.x.sack.oldcumack = old_cumack; + sctp_clog.x.sack.tsn = tsn; + sctp_clog.x.sack.numGaps = gaps; + sctp_clog.x.sack.numDups = dups; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_SACK, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +void +sctp_log_map(uint32_t map, uint32_t cum, uint32_t high, int from) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + memset(&sctp_clog, 0, sizeof(sctp_clog)); + sctp_clog.x.map.base = map; + sctp_clog.x.map.cum = cum; + sctp_clog.x.map.high = high; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_MAP, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +void +sctp_log_fr(uint32_t biggest_tsn, uint32_t biggest_new_tsn, uint32_t tsn, int from) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + memset(&sctp_clog, 0, sizeof(sctp_clog)); + sctp_clog.x.fr.largest_tsn = biggest_tsn; + sctp_clog.x.fr.largest_new_tsn = biggest_new_tsn; + sctp_clog.x.fr.tsn = tsn; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_FR, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +#ifdef SCTP_MBUF_LOGGING +void +sctp_log_mb(struct mbuf *m, int from) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + sctp_clog.x.mb.mp = m; + sctp_clog.x.mb.mbuf_flags = (uint8_t)(SCTP_BUF_GET_FLAGS(m)); + sctp_clog.x.mb.size = (uint16_t)(SCTP_BUF_LEN(m)); + sctp_clog.x.mb.data = SCTP_BUF_AT(m, 0); + if (SCTP_BUF_IS_EXTENDED(m)) { + sctp_clog.x.mb.ext = SCTP_BUF_EXTEND_BASE(m); +#if defined(__APPLE__) && !defined(__Userspace__) + /* APPLE does not use a ref_cnt, but a forward/backward ref queue */ +#else + sctp_clog.x.mb.refcnt = (uint8_t)(SCTP_BUF_EXTEND_REFCNT(m)); +#endif + } else { + sctp_clog.x.mb.ext = 0; + sctp_clog.x.mb.refcnt = 0; + } + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_MBUF, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +void +sctp_log_mbc(struct mbuf *m, int from) +{ + struct mbuf *mat; + + for (mat = m; mat; mat = SCTP_BUF_NEXT(mat)) { + sctp_log_mb(mat, from); + } +} +#endif + +void +sctp_log_strm_del(struct sctp_queued_to_read *control, struct sctp_queued_to_read *poschk, int from) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + if (control == NULL) { + SCTP_PRINTF("Gak log of NULL?\n"); + return; + } + sctp_clog.x.strlog.stcb = control->stcb; + sctp_clog.x.strlog.n_tsn = control->sinfo_tsn; + sctp_clog.x.strlog.n_sseq = (uint16_t)control->mid; + sctp_clog.x.strlog.strm = control->sinfo_stream; + if (poschk != NULL) { + sctp_clog.x.strlog.e_tsn = poschk->sinfo_tsn; + sctp_clog.x.strlog.e_sseq = (uint16_t)poschk->mid; + } else { + sctp_clog.x.strlog.e_tsn = 0; + sctp_clog.x.strlog.e_sseq = 0; + } + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_STRM, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +void +sctp_log_cwnd(struct sctp_tcb *stcb, struct sctp_nets *net, int augment, uint8_t from) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + sctp_clog.x.cwnd.net = net; + if (stcb->asoc.send_queue_cnt > 255) + sctp_clog.x.cwnd.cnt_in_send = 255; + else + sctp_clog.x.cwnd.cnt_in_send = stcb->asoc.send_queue_cnt; + if (stcb->asoc.stream_queue_cnt > 255) + sctp_clog.x.cwnd.cnt_in_str = 255; + else + sctp_clog.x.cwnd.cnt_in_str = stcb->asoc.stream_queue_cnt; + + if (net) { + sctp_clog.x.cwnd.cwnd_new_value = net->cwnd; + sctp_clog.x.cwnd.inflight = net->flight_size; + sctp_clog.x.cwnd.pseudo_cumack = net->pseudo_cumack; + sctp_clog.x.cwnd.meets_pseudo_cumack = net->new_pseudo_cumack; + sctp_clog.x.cwnd.need_new_pseudo_cumack = net->find_pseudo_cumack; + } + if (SCTP_CWNDLOG_PRESEND == from) { + sctp_clog.x.cwnd.meets_pseudo_cumack = stcb->asoc.peers_rwnd; + } + sctp_clog.x.cwnd.cwnd_augment = augment; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_CWND, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +#if !defined(__APPLE__) && !defined(__Userspace__) +void +sctp_log_lock(struct sctp_inpcb *inp, struct sctp_tcb *stcb, uint8_t from) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + memset(&sctp_clog, 0, sizeof(sctp_clog)); + if (inp) { + sctp_clog.x.lock.sock = (void *) inp->sctp_socket; + + } else { + sctp_clog.x.lock.sock = (void *) NULL; + } + sctp_clog.x.lock.inp = (void *) inp; +#if defined(__FreeBSD__) + if (stcb) { + sctp_clog.x.lock.tcb_lock = mtx_owned(&stcb->tcb_mtx); + } else { + sctp_clog.x.lock.tcb_lock = SCTP_LOCK_UNKNOWN; + } + if (inp) { + sctp_clog.x.lock.inp_lock = mtx_owned(&inp->inp_mtx); + sctp_clog.x.lock.create_lock = mtx_owned(&inp->inp_create_mtx); + } else { + sctp_clog.x.lock.inp_lock = SCTP_LOCK_UNKNOWN; + sctp_clog.x.lock.create_lock = SCTP_LOCK_UNKNOWN; + } + sctp_clog.x.lock.info_lock = rw_wowned(&SCTP_BASE_INFO(ipi_ep_mtx)); + if (inp && (inp->sctp_socket)) { + sctp_clog.x.lock.sock_lock = mtx_owned(SOCK_MTX(inp->sctp_socket)); + sctp_clog.x.lock.sockrcvbuf_lock = mtx_owned(SOCKBUF_MTX(&inp->sctp_socket->so_rcv)); + sctp_clog.x.lock.socksndbuf_lock = mtx_owned(SOCKBUF_MTX(&inp->sctp_socket->so_snd)); + } else { + sctp_clog.x.lock.sock_lock = SCTP_LOCK_UNKNOWN; + sctp_clog.x.lock.sockrcvbuf_lock = SCTP_LOCK_UNKNOWN; + sctp_clog.x.lock.socksndbuf_lock = SCTP_LOCK_UNKNOWN; + } +#endif + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_LOCK_EVENT, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} +#endif + +void +sctp_log_maxburst(struct sctp_tcb *stcb, struct sctp_nets *net, int error, int burst, uint8_t from) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + memset(&sctp_clog, 0, sizeof(sctp_clog)); + sctp_clog.x.cwnd.net = net; + sctp_clog.x.cwnd.cwnd_new_value = error; + sctp_clog.x.cwnd.inflight = net->flight_size; + sctp_clog.x.cwnd.cwnd_augment = burst; + if (stcb->asoc.send_queue_cnt > 255) + sctp_clog.x.cwnd.cnt_in_send = 255; + else + sctp_clog.x.cwnd.cnt_in_send = stcb->asoc.send_queue_cnt; + if (stcb->asoc.stream_queue_cnt > 255) + sctp_clog.x.cwnd.cnt_in_str = 255; + else + sctp_clog.x.cwnd.cnt_in_str = stcb->asoc.stream_queue_cnt; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_MAXBURST, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +void +sctp_log_rwnd(uint8_t from, uint32_t peers_rwnd, uint32_t snd_size, uint32_t overhead) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + sctp_clog.x.rwnd.rwnd = peers_rwnd; + sctp_clog.x.rwnd.send_size = snd_size; + sctp_clog.x.rwnd.overhead = overhead; + sctp_clog.x.rwnd.new_rwnd = 0; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_RWND, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +void +sctp_log_rwnd_set(uint8_t from, uint32_t peers_rwnd, uint32_t flight_size, uint32_t overhead, uint32_t a_rwndval) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + sctp_clog.x.rwnd.rwnd = peers_rwnd; + sctp_clog.x.rwnd.send_size = flight_size; + sctp_clog.x.rwnd.overhead = overhead; + sctp_clog.x.rwnd.new_rwnd = a_rwndval; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_RWND, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +#ifdef SCTP_MBCNT_LOGGING +static void +sctp_log_mbcnt(uint8_t from, uint32_t total_oq, uint32_t book, uint32_t total_mbcnt_q, uint32_t mbcnt) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + sctp_clog.x.mbcnt.total_queue_size = total_oq; + sctp_clog.x.mbcnt.size_change = book; + sctp_clog.x.mbcnt.total_queue_mb_size = total_mbcnt_q; + sctp_clog.x.mbcnt.mbcnt_change = mbcnt; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_MBCNT, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} +#endif + +void +sctp_misc_ints(uint8_t from, uint32_t a, uint32_t b, uint32_t c, uint32_t d) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_MISC_EVENT, + from, + a, b, c, d); +#endif +} + +void +sctp_wakeup_log(struct sctp_tcb *stcb, uint32_t wake_cnt, int from) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + sctp_clog.x.wake.stcb = (void *)stcb; + sctp_clog.x.wake.wake_cnt = wake_cnt; + sctp_clog.x.wake.flight = stcb->asoc.total_flight_count; + sctp_clog.x.wake.send_q = stcb->asoc.send_queue_cnt; + sctp_clog.x.wake.sent_q = stcb->asoc.sent_queue_cnt; + + if (stcb->asoc.stream_queue_cnt < 0xff) + sctp_clog.x.wake.stream_qcnt = (uint8_t) stcb->asoc.stream_queue_cnt; + else + sctp_clog.x.wake.stream_qcnt = 0xff; + + if (stcb->asoc.chunks_on_out_queue < 0xff) + sctp_clog.x.wake.chunks_on_oque = (uint8_t) stcb->asoc.chunks_on_out_queue; + else + sctp_clog.x.wake.chunks_on_oque = 0xff; + + sctp_clog.x.wake.sctpflags = 0; + /* set in the defered mode stuff */ + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) + sctp_clog.x.wake.sctpflags |= 1; + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_WAKEOUTPUT) + sctp_clog.x.wake.sctpflags |= 2; + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_WAKEINPUT) + sctp_clog.x.wake.sctpflags |= 4; + /* what about the sb */ + if (stcb->sctp_socket) { + struct socket *so = stcb->sctp_socket; + + sctp_clog.x.wake.sbflags = (uint8_t)((so->so_snd.sb_flags & 0x00ff)); + } else { + sctp_clog.x.wake.sbflags = 0xff; + } + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_WAKE, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +void +sctp_log_block(uint8_t from, struct sctp_association *asoc, ssize_t sendlen) +{ +#if defined(SCTP_LOCAL_TRACE_BUF) + struct sctp_cwnd_log sctp_clog; + + sctp_clog.x.blk.onsb = asoc->total_output_queue_size; + sctp_clog.x.blk.send_sent_qcnt = (uint16_t) (asoc->send_queue_cnt + asoc->sent_queue_cnt); + sctp_clog.x.blk.peer_rwnd = asoc->peers_rwnd; + sctp_clog.x.blk.stream_qcnt = (uint16_t) asoc->stream_queue_cnt; + sctp_clog.x.blk.chunks_on_oque = (uint16_t) asoc->chunks_on_out_queue; + sctp_clog.x.blk.flight_size = (uint16_t) (asoc->total_flight/1024); + sctp_clog.x.blk.sndlen = (uint32_t)sendlen; + SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", + SCTP_LOG_EVENT_BLOCK, + from, + sctp_clog.x.misc.log1, + sctp_clog.x.misc.log2, + sctp_clog.x.misc.log3, + sctp_clog.x.misc.log4); +#endif +} + +int +sctp_fill_stat_log(void *optval SCTP_UNUSED, size_t *optsize SCTP_UNUSED) +{ + /* May need to fix this if ktrdump does not work */ + return (0); +} + +#ifdef SCTP_AUDITING_ENABLED +uint8_t sctp_audit_data[SCTP_AUDIT_SIZE][2]; +static int sctp_audit_indx = 0; + +static +void +sctp_print_audit_report(void) +{ + int i; + int cnt; + + cnt = 0; + for (i = sctp_audit_indx; i < SCTP_AUDIT_SIZE; i++) { + if ((sctp_audit_data[i][0] == 0xe0) && + (sctp_audit_data[i][1] == 0x01)) { + cnt = 0; + SCTP_PRINTF("\n"); + } else if (sctp_audit_data[i][0] == 0xf0) { + cnt = 0; + SCTP_PRINTF("\n"); + } else if ((sctp_audit_data[i][0] == 0xc0) && + (sctp_audit_data[i][1] == 0x01)) { + SCTP_PRINTF("\n"); + cnt = 0; + } + SCTP_PRINTF("%2.2x%2.2x ", (uint32_t) sctp_audit_data[i][0], + (uint32_t) sctp_audit_data[i][1]); + cnt++; + if ((cnt % 14) == 0) + SCTP_PRINTF("\n"); + } + for (i = 0; i < sctp_audit_indx; i++) { + if ((sctp_audit_data[i][0] == 0xe0) && + (sctp_audit_data[i][1] == 0x01)) { + cnt = 0; + SCTP_PRINTF("\n"); + } else if (sctp_audit_data[i][0] == 0xf0) { + cnt = 0; + SCTP_PRINTF("\n"); + } else if ((sctp_audit_data[i][0] == 0xc0) && + (sctp_audit_data[i][1] == 0x01)) { + SCTP_PRINTF("\n"); + cnt = 0; + } + SCTP_PRINTF("%2.2x%2.2x ", (uint32_t) sctp_audit_data[i][0], + (uint32_t) sctp_audit_data[i][1]); + cnt++; + if ((cnt % 14) == 0) + SCTP_PRINTF("\n"); + } + SCTP_PRINTF("\n"); +} + +void +sctp_auditing(int from, struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + int resend_cnt, tot_out, rep, tot_book_cnt; + struct sctp_nets *lnet; + struct sctp_tmit_chunk *chk; + + sctp_audit_data[sctp_audit_indx][0] = 0xAA; + sctp_audit_data[sctp_audit_indx][1] = 0x000000ff & from; + sctp_audit_indx++; + if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { + sctp_audit_indx = 0; + } + if (inp == NULL) { + sctp_audit_data[sctp_audit_indx][0] = 0xAF; + sctp_audit_data[sctp_audit_indx][1] = 0x01; + sctp_audit_indx++; + if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { + sctp_audit_indx = 0; + } + return; + } + if (stcb == NULL) { + sctp_audit_data[sctp_audit_indx][0] = 0xAF; + sctp_audit_data[sctp_audit_indx][1] = 0x02; + sctp_audit_indx++; + if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { + sctp_audit_indx = 0; + } + return; + } + sctp_audit_data[sctp_audit_indx][0] = 0xA1; + sctp_audit_data[sctp_audit_indx][1] = + (0x000000ff & stcb->asoc.sent_queue_retran_cnt); + sctp_audit_indx++; + if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { + sctp_audit_indx = 0; + } + rep = 0; + tot_book_cnt = 0; + resend_cnt = tot_out = 0; + TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { + if (chk->sent == SCTP_DATAGRAM_RESEND) { + resend_cnt++; + } else if (chk->sent < SCTP_DATAGRAM_RESEND) { + tot_out += chk->book_size; + tot_book_cnt++; + } + } + if (resend_cnt != stcb->asoc.sent_queue_retran_cnt) { + sctp_audit_data[sctp_audit_indx][0] = 0xAF; + sctp_audit_data[sctp_audit_indx][1] = 0xA1; + sctp_audit_indx++; + if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { + sctp_audit_indx = 0; + } + SCTP_PRINTF("resend_cnt:%d asoc-tot:%d\n", + resend_cnt, stcb->asoc.sent_queue_retran_cnt); + rep = 1; + stcb->asoc.sent_queue_retran_cnt = resend_cnt; + sctp_audit_data[sctp_audit_indx][0] = 0xA2; + sctp_audit_data[sctp_audit_indx][1] = + (0x000000ff & stcb->asoc.sent_queue_retran_cnt); + sctp_audit_indx++; + if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { + sctp_audit_indx = 0; + } + } + if (tot_out != stcb->asoc.total_flight) { + sctp_audit_data[sctp_audit_indx][0] = 0xAF; + sctp_audit_data[sctp_audit_indx][1] = 0xA2; + sctp_audit_indx++; + if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { + sctp_audit_indx = 0; + } + rep = 1; + SCTP_PRINTF("tot_flt:%d asoc_tot:%d\n", tot_out, + (int)stcb->asoc.total_flight); + stcb->asoc.total_flight = tot_out; + } + if (tot_book_cnt != stcb->asoc.total_flight_count) { + sctp_audit_data[sctp_audit_indx][0] = 0xAF; + sctp_audit_data[sctp_audit_indx][1] = 0xA5; + sctp_audit_indx++; + if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { + sctp_audit_indx = 0; + } + rep = 1; + SCTP_PRINTF("tot_flt_book:%d\n", tot_book_cnt); + + stcb->asoc.total_flight_count = tot_book_cnt; + } + tot_out = 0; + TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { + tot_out += lnet->flight_size; + } + if (tot_out != stcb->asoc.total_flight) { + sctp_audit_data[sctp_audit_indx][0] = 0xAF; + sctp_audit_data[sctp_audit_indx][1] = 0xA3; + sctp_audit_indx++; + if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { + sctp_audit_indx = 0; + } + rep = 1; + SCTP_PRINTF("real flight:%d net total was %d\n", + stcb->asoc.total_flight, tot_out); + /* now corrective action */ + TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { + tot_out = 0; + TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { + if ((chk->whoTo == lnet) && + (chk->sent < SCTP_DATAGRAM_RESEND)) { + tot_out += chk->book_size; + } + } + if (lnet->flight_size != tot_out) { + SCTP_PRINTF("net:%p flight was %d corrected to %d\n", + (void *)lnet, lnet->flight_size, + tot_out); + lnet->flight_size = tot_out; + } + } + } + if (rep) { + sctp_print_audit_report(); + } +} + +void +sctp_audit_log(uint8_t ev, uint8_t fd) +{ + + sctp_audit_data[sctp_audit_indx][0] = ev; + sctp_audit_data[sctp_audit_indx][1] = fd; + sctp_audit_indx++; + if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { + sctp_audit_indx = 0; + } +} + +#endif + +/* + * The conversion from time to ticks and vice versa is done by rounding + * upwards. This way we can test in the code the time to be positive and + * know that this corresponds to a positive number of ticks. + */ + +uint32_t +sctp_msecs_to_ticks(uint32_t msecs) +{ + uint64_t temp; + uint32_t ticks; + + if (hz == 1000) { + ticks = msecs; + } else { + temp = (((uint64_t)msecs * hz) + 999) / 1000; + if (temp > UINT32_MAX) { + ticks = UINT32_MAX; + } else { + ticks = (uint32_t)temp; + } + } + return (ticks); +} + +uint32_t +sctp_ticks_to_msecs(uint32_t ticks) +{ + uint64_t temp; + uint32_t msecs; + + if (hz == 1000) { + msecs = ticks; + } else { + temp = (((uint64_t)ticks * 1000) + (hz - 1)) / hz; + if (temp > UINT32_MAX) { + msecs = UINT32_MAX; + } else { + msecs = (uint32_t)temp; + } + } + return (msecs); +} + +uint32_t +sctp_secs_to_ticks(uint32_t secs) +{ + uint64_t temp; + uint32_t ticks; + + temp = (uint64_t)secs * hz; + if (temp > UINT32_MAX) { + ticks = UINT32_MAX; + } else { + ticks = (uint32_t)temp; + } + return (ticks); +} + +uint32_t +sctp_ticks_to_secs(uint32_t ticks) +{ + uint64_t temp; + uint32_t secs; + + temp = ((uint64_t)ticks + (hz - 1)) / hz; + if (temp > UINT32_MAX) { + secs = UINT32_MAX; + } else { + secs = (uint32_t)temp; + } + return (secs); +} + +/* + * sctp_stop_timers_for_shutdown() should be called + * when entering the SHUTDOWN_SENT or SHUTDOWN_ACK_SENT + * state to make sure that all timers are stopped. + */ +void +sctp_stop_timers_for_shutdown(struct sctp_tcb *stcb) +{ + struct sctp_inpcb *inp; + struct sctp_nets *net; + + inp = stcb->sctp_ep; + + sctp_timer_stop(SCTP_TIMER_TYPE_RECV, inp, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_12); + sctp_timer_stop(SCTP_TIMER_TYPE_STRRESET, inp, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_13); + sctp_timer_stop(SCTP_TIMER_TYPE_ASCONF, inp, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_14); + sctp_timer_stop(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_15); + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, + SCTP_FROM_SCTPUTIL + SCTP_LOC_16); + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, + SCTP_FROM_SCTPUTIL + SCTP_LOC_17); + } +} + +void +sctp_stop_association_timers(struct sctp_tcb *stcb, bool stop_assoc_kill_timer) +{ + struct sctp_inpcb *inp; + struct sctp_nets *net; + + inp = stcb->sctp_ep; + sctp_timer_stop(SCTP_TIMER_TYPE_RECV, inp, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_18); + sctp_timer_stop(SCTP_TIMER_TYPE_STRRESET, inp, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_19); + if (stop_assoc_kill_timer) { + sctp_timer_stop(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_20); + } + sctp_timer_stop(SCTP_TIMER_TYPE_ASCONF, inp, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_21); + sctp_timer_stop(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_22); + sctp_timer_stop(SCTP_TIMER_TYPE_SHUTDOWNGUARD, inp, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_23); + /* Mobility adaptation */ + sctp_timer_stop(SCTP_TIMER_TYPE_PRIM_DELETED, inp, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_24); + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + sctp_timer_stop(SCTP_TIMER_TYPE_SEND, inp, stcb, net, + SCTP_FROM_SCTPUTIL + SCTP_LOC_25); + sctp_timer_stop(SCTP_TIMER_TYPE_INIT, inp, stcb, net, + SCTP_FROM_SCTPUTIL + SCTP_LOC_26); + sctp_timer_stop(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, net, + SCTP_FROM_SCTPUTIL + SCTP_LOC_27); + sctp_timer_stop(SCTP_TIMER_TYPE_COOKIE, inp, stcb, net, + SCTP_FROM_SCTPUTIL + SCTP_LOC_28); + sctp_timer_stop(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, net, + SCTP_FROM_SCTPUTIL + SCTP_LOC_29); + sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, + SCTP_FROM_SCTPUTIL + SCTP_LOC_30); + sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, + SCTP_FROM_SCTPUTIL + SCTP_LOC_31); + } +} + +/* + * A list of sizes based on typical mtu's, used only if next hop size not + * returned. These values MUST be multiples of 4 and MUST be ordered. + */ +static uint32_t sctp_mtu_sizes[] = { + 68, + 296, + 508, + 512, + 544, + 576, + 1004, + 1492, + 1500, + 1536, + 2000, + 2048, + 4352, + 4464, + 8168, + 17912, + 32000, + 65532 +}; + +/* + * Return the largest MTU in sctp_mtu_sizes smaller than val. + * If val is smaller than the minimum, just return the largest + * multiple of 4 smaller or equal to val. + * Ensure that the result is a multiple of 4. + */ +uint32_t +sctp_get_prev_mtu(uint32_t val) +{ + uint32_t i; + + val &= 0xfffffffc; + if (val <= sctp_mtu_sizes[0]) { + return (val); + } + for (i = 1; i < (sizeof(sctp_mtu_sizes) / sizeof(uint32_t)); i++) { + if (val <= sctp_mtu_sizes[i]) { + break; + } + } + KASSERT((sctp_mtu_sizes[i - 1] & 0x00000003) == 0, + ("sctp_mtu_sizes[%u] not a multiple of 4", i - 1)); + return (sctp_mtu_sizes[i - 1]); +} + +/* + * Return the smallest MTU in sctp_mtu_sizes larger than val. + * If val is larger than the maximum, just return the largest multiple of 4 smaller + * or equal to val. + * Ensure that the result is a multiple of 4. + */ +uint32_t +sctp_get_next_mtu(uint32_t val) +{ + /* select another MTU that is just bigger than this one */ + uint32_t i; + + val &= 0xfffffffc; + for (i = 0; i < (sizeof(sctp_mtu_sizes) / sizeof(uint32_t)); i++) { + if (val < sctp_mtu_sizes[i]) { + KASSERT((sctp_mtu_sizes[i] & 0x00000003) == 0, + ("sctp_mtu_sizes[%u] not a multiple of 4", i)); + return (sctp_mtu_sizes[i]); + } + } + return (val); +} + +void +sctp_fill_random_store(struct sctp_pcb *m) +{ + /* + * Here we use the MD5/SHA-1 to hash with our good randomNumbers and + * our counter. The result becomes our good random numbers and we + * then setup to give these out. Note that we do no locking to + * protect this. This is ok, since if competing folks call this we + * will get more gobbled gook in the random store which is what we + * want. There is a danger that two guys will use the same random + * numbers, but thats ok too since that is random as well :-> + */ + m->store_at = 0; +#if defined(__Userspace__) && defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) + for (int i = 0; i < (int) (sizeof(m->random_store) / sizeof(m->random_store[0])); i++) { + m->random_store[i] = (uint8_t) rand(); + } +#else + (void)sctp_hmac(SCTP_HMAC, (uint8_t *)m->random_numbers, + sizeof(m->random_numbers), (uint8_t *)&m->random_counter, + sizeof(m->random_counter), (uint8_t *)m->random_store); +#endif + m->random_counter++; +} + +uint32_t +sctp_select_initial_TSN(struct sctp_pcb *inp) +{ + /* + * A true implementation should use random selection process to get + * the initial stream sequence number, using RFC1750 as a good + * guideline + */ + uint32_t x, *xp; + uint8_t *p; + int store_at, new_store; + + if (inp->initial_sequence_debug != 0) { + uint32_t ret; + + ret = inp->initial_sequence_debug; + inp->initial_sequence_debug++; + return (ret); + } + retry: + store_at = inp->store_at; + new_store = store_at + sizeof(uint32_t); + if (new_store >= (SCTP_SIGNATURE_SIZE-3)) { + new_store = 0; + } + if (!atomic_cmpset_int(&inp->store_at, store_at, new_store)) { + goto retry; + } + if (new_store == 0) { + /* Refill the random store */ + sctp_fill_random_store(inp); + } + p = &inp->random_store[store_at]; + xp = (uint32_t *)p; + x = *xp; + return (x); +} + +uint32_t +sctp_select_a_tag(struct sctp_inpcb *inp, uint16_t lport, uint16_t rport, int check) +{ + uint32_t x; + struct timeval now; + + if (check) { + (void)SCTP_GETTIME_TIMEVAL(&now); + } + for (;;) { + x = sctp_select_initial_TSN(&inp->sctp_ep); + if (x == 0) { + /* we never use 0 */ + continue; + } + if (!check || sctp_is_vtag_good(x, lport, rport, &now)) { + break; + } + } + return (x); +} + +int32_t +sctp_map_assoc_state(int kernel_state) +{ + int32_t user_state; + + if (kernel_state & SCTP_STATE_WAS_ABORTED) { + user_state = SCTP_CLOSED; + } else if (kernel_state & SCTP_STATE_SHUTDOWN_PENDING) { + user_state = SCTP_SHUTDOWN_PENDING; + } else { + switch (kernel_state & SCTP_STATE_MASK) { + case SCTP_STATE_EMPTY: + user_state = SCTP_CLOSED; + break; + case SCTP_STATE_INUSE: + user_state = SCTP_CLOSED; + break; + case SCTP_STATE_COOKIE_WAIT: + user_state = SCTP_COOKIE_WAIT; + break; + case SCTP_STATE_COOKIE_ECHOED: + user_state = SCTP_COOKIE_ECHOED; + break; + case SCTP_STATE_OPEN: + user_state = SCTP_ESTABLISHED; + break; + case SCTP_STATE_SHUTDOWN_SENT: + user_state = SCTP_SHUTDOWN_SENT; + break; + case SCTP_STATE_SHUTDOWN_RECEIVED: + user_state = SCTP_SHUTDOWN_RECEIVED; + break; + case SCTP_STATE_SHUTDOWN_ACK_SENT: + user_state = SCTP_SHUTDOWN_ACK_SENT; + break; + default: + user_state = SCTP_CLOSED; + break; + } + } + return (user_state); +} + +int +sctp_init_asoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + uint32_t override_tag, uint32_t initial_tsn, uint32_t vrf_id, + uint16_t o_strms) +{ + struct sctp_association *asoc; + /* + * Anything set to zero is taken care of by the allocation routine's + * bzero + */ + + /* + * Up front select what scoping to apply on addresses I tell my peer + * Not sure what to do with these right now, we will need to come up + * with a way to set them. We may need to pass them through from the + * caller in the sctp_aloc_assoc() function. + */ + int i; +#if defined(SCTP_DETAILED_STR_STATS) + int j; +#endif + + asoc = &stcb->asoc; + /* init all variables to a known value. */ + SCTP_SET_STATE(stcb, SCTP_STATE_INUSE); + asoc->max_burst = inp->sctp_ep.max_burst; + asoc->fr_max_burst = inp->sctp_ep.fr_max_burst; + asoc->heart_beat_delay = sctp_ticks_to_msecs(inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT]); + asoc->cookie_life = inp->sctp_ep.def_cookie_life; + asoc->sctp_cmt_on_off = inp->sctp_cmt_on_off; + asoc->ecn_supported = inp->ecn_supported; + asoc->prsctp_supported = inp->prsctp_supported; + asoc->auth_supported = inp->auth_supported; + asoc->asconf_supported = inp->asconf_supported; + asoc->reconfig_supported = inp->reconfig_supported; + asoc->nrsack_supported = inp->nrsack_supported; + asoc->pktdrop_supported = inp->pktdrop_supported; + asoc->idata_supported = inp->idata_supported; + asoc->zero_checksum = inp->zero_checksum; + asoc->sctp_cmt_pf = (uint8_t)0; + asoc->sctp_frag_point = inp->sctp_frag_point; + asoc->sctp_features = inp->sctp_features; + asoc->default_dscp = inp->sctp_ep.default_dscp; + asoc->max_cwnd = inp->max_cwnd; +#ifdef INET6 + if (inp->sctp_ep.default_flowlabel) { + asoc->default_flowlabel = inp->sctp_ep.default_flowlabel; + } else { + if (inp->ip_inp.inp.inp_flags & IN6P_AUTOFLOWLABEL) { + asoc->default_flowlabel = sctp_select_initial_TSN(&inp->sctp_ep); + asoc->default_flowlabel &= 0x000fffff; + asoc->default_flowlabel |= 0x80000000; + } else { + asoc->default_flowlabel = 0; + } + } +#endif + asoc->sb_send_resv = 0; + if (override_tag) { + asoc->my_vtag = override_tag; + } else { + asoc->my_vtag = sctp_select_a_tag(inp, stcb->sctp_ep->sctp_lport, stcb->rport, 1); + } + /* Get the nonce tags */ + asoc->my_vtag_nonce = sctp_select_a_tag(inp, stcb->sctp_ep->sctp_lport, stcb->rport, 0); + asoc->peer_vtag_nonce = sctp_select_a_tag(inp, stcb->sctp_ep->sctp_lport, stcb->rport, 0); + asoc->vrf_id = vrf_id; + +#ifdef SCTP_ASOCLOG_OF_TSNS + asoc->tsn_in_at = 0; + asoc->tsn_out_at = 0; + asoc->tsn_in_wrapped = 0; + asoc->tsn_out_wrapped = 0; + asoc->cumack_log_at = 0; + asoc->cumack_log_atsnt = 0; +#endif +#ifdef SCTP_FS_SPEC_LOG + asoc->fs_index = 0; +#endif + asoc->refcnt = 0; + asoc->assoc_up_sent = 0; + if (override_tag) { + asoc->init_seq_number = initial_tsn; + } else { + asoc->init_seq_number = sctp_select_initial_TSN(&inp->sctp_ep); + } + asoc->asconf_seq_out = asoc->init_seq_number; + asoc->str_reset_seq_out = asoc->init_seq_number; + asoc->sending_seq = asoc->init_seq_number; + asoc->asconf_seq_out_acked = asoc->init_seq_number - 1; + /* we are optimistic here */ + asoc->peer_supports_nat = 0; + asoc->sent_queue_retran_cnt = 0; + + /* for CMT */ + asoc->last_net_cmt_send_started = NULL; + + asoc->last_acked_seq = asoc->init_seq_number - 1; + asoc->advanced_peer_ack_point = asoc->init_seq_number - 1; + asoc->asconf_seq_in = asoc->init_seq_number - 1; + + /* here we are different, we hold the next one we expect */ + asoc->str_reset_seq_in = asoc->init_seq_number; + + asoc->initial_init_rto_max = inp->sctp_ep.initial_init_rto_max; + asoc->initial_rto = inp->sctp_ep.initial_rto; + + asoc->default_mtu = inp->sctp_ep.default_mtu; + asoc->max_init_times = inp->sctp_ep.max_init_times; + asoc->max_send_times = inp->sctp_ep.max_send_times; + asoc->def_net_failure = inp->sctp_ep.def_net_failure; + asoc->def_net_pf_threshold = inp->sctp_ep.def_net_pf_threshold; + asoc->free_chunk_cnt = 0; + + asoc->iam_blocking = 0; + asoc->context = inp->sctp_context; + asoc->local_strreset_support = inp->local_strreset_support; + asoc->def_send = inp->def_send; + asoc->delayed_ack = sctp_ticks_to_msecs(inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_RECV]); + asoc->sack_freq = inp->sctp_ep.sctp_sack_freq; + asoc->pr_sctp_cnt = 0; + asoc->total_output_queue_size = 0; + + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + asoc->scope.ipv6_addr_legal = 1; + if (SCTP_IPV6_V6ONLY(inp) == 0) { + asoc->scope.ipv4_addr_legal = 1; + } else { + asoc->scope.ipv4_addr_legal = 0; + } +#if defined(__Userspace__) + asoc->scope.conn_addr_legal = 0; +#endif + } else { + asoc->scope.ipv6_addr_legal = 0; +#if defined(__Userspace__) + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_CONN) { + asoc->scope.conn_addr_legal = 1; + asoc->scope.ipv4_addr_legal = 0; + } else { + asoc->scope.conn_addr_legal = 0; + asoc->scope.ipv4_addr_legal = 1; + } +#else + asoc->scope.ipv4_addr_legal = 1; +#endif + } + + asoc->my_rwnd = max(SCTP_SB_LIMIT_RCV(inp->sctp_socket), SCTP_MINIMAL_RWND); + asoc->peers_rwnd = SCTP_SB_LIMIT_RCV(inp->sctp_socket); + + asoc->smallest_mtu = 0; + asoc->minrto = inp->sctp_ep.sctp_minrto; + asoc->maxrto = inp->sctp_ep.sctp_maxrto; + + asoc->stream_locked_on = 0; + asoc->ecn_echo_cnt_onq = 0; + asoc->stream_locked = 0; + + asoc->send_sack = 1; + + LIST_INIT(&asoc->sctp_restricted_addrs); + + TAILQ_INIT(&asoc->nets); + TAILQ_INIT(&asoc->pending_reply_queue); + TAILQ_INIT(&asoc->asconf_ack_sent); + /* Setup to fill the hb random cache at first HB */ + asoc->hb_random_idx = 4; + + asoc->sctp_autoclose_ticks = inp->sctp_ep.auto_close_time; + + stcb->asoc.congestion_control_module = inp->sctp_ep.sctp_default_cc_module; + stcb->asoc.cc_functions = sctp_cc_functions[inp->sctp_ep.sctp_default_cc_module]; + + stcb->asoc.stream_scheduling_module = inp->sctp_ep.sctp_default_ss_module; + stcb->asoc.ss_functions = sctp_ss_functions[inp->sctp_ep.sctp_default_ss_module]; + + /* + * Now the stream parameters, here we allocate space for all streams + * that we request by default. + */ + asoc->strm_realoutsize = asoc->streamoutcnt = asoc->pre_open_streams = + o_strms; + SCTP_MALLOC(asoc->strmout, struct sctp_stream_out *, + asoc->streamoutcnt * sizeof(struct sctp_stream_out), + SCTP_M_STRMO); + if (asoc->strmout == NULL) { + /* big trouble no memory */ + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOMEM); + return (ENOMEM); + } + SCTP_TCB_LOCK(stcb); + for (i = 0; i < asoc->streamoutcnt; i++) { + /* + * inbound side must be set to 0xffff, also NOTE when we get + * the INIT-ACK back (for INIT sender) we MUST reduce the + * count (streamoutcnt) but first check if we sent to any of + * the upper streams that were dropped (if some were). Those + * that were dropped must be notified to the upper layer as + * failed to send. + */ + TAILQ_INIT(&asoc->strmout[i].outqueue); + asoc->ss_functions.sctp_ss_init_stream(stcb, &asoc->strmout[i], NULL); + asoc->strmout[i].chunks_on_queues = 0; +#if defined(SCTP_DETAILED_STR_STATS) + for (j = 0; j < SCTP_PR_SCTP_MAX + 1; j++) { + asoc->strmout[i].abandoned_sent[j] = 0; + asoc->strmout[i].abandoned_unsent[j] = 0; + } +#else + asoc->strmout[i].abandoned_sent[0] = 0; + asoc->strmout[i].abandoned_unsent[0] = 0; +#endif + asoc->strmout[i].next_mid_ordered = 0; + asoc->strmout[i].next_mid_unordered = 0; + asoc->strmout[i].sid = i; + asoc->strmout[i].last_msg_incomplete = 0; + asoc->strmout[i].state = SCTP_STREAM_OPENING; + } + asoc->ss_functions.sctp_ss_init(stcb, asoc); + SCTP_TCB_UNLOCK(stcb); + + /* Now the mapping array */ + asoc->mapping_array_size = SCTP_INITIAL_MAPPING_ARRAY; + SCTP_MALLOC(asoc->mapping_array, uint8_t *, asoc->mapping_array_size, + SCTP_M_MAP); + if (asoc->mapping_array == NULL) { + SCTP_FREE(asoc->strmout, SCTP_M_STRMO); + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOMEM); + return (ENOMEM); + } + memset(asoc->mapping_array, 0, asoc->mapping_array_size); + SCTP_MALLOC(asoc->nr_mapping_array, uint8_t *, asoc->mapping_array_size, + SCTP_M_MAP); + if (asoc->nr_mapping_array == NULL) { + SCTP_FREE(asoc->strmout, SCTP_M_STRMO); + SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOMEM); + return (ENOMEM); + } + memset(asoc->nr_mapping_array, 0, asoc->mapping_array_size); + + /* Now the init of the other outqueues */ + TAILQ_INIT(&asoc->free_chunks); + TAILQ_INIT(&asoc->control_send_queue); + TAILQ_INIT(&asoc->asconf_send_queue); + TAILQ_INIT(&asoc->send_queue); + TAILQ_INIT(&asoc->sent_queue); + TAILQ_INIT(&asoc->resetHead); + asoc->max_inbound_streams = inp->sctp_ep.max_open_streams_intome; + TAILQ_INIT(&asoc->asconf_queue); + /* authentication fields */ + asoc->authinfo.random = NULL; + asoc->authinfo.active_keyid = 0; + asoc->authinfo.assoc_key = NULL; + asoc->authinfo.assoc_keyid = 0; + asoc->authinfo.recv_key = NULL; + asoc->authinfo.recv_keyid = 0; + LIST_INIT(&asoc->shared_keys); + asoc->marked_retrans = 0; + asoc->port = inp->sctp_ep.port; + asoc->timoinit = 0; + asoc->timodata = 0; + asoc->timosack = 0; + asoc->timoshutdown = 0; + asoc->timoheartbeat = 0; + asoc->timocookie = 0; + asoc->timoshutdownack = 0; + (void)SCTP_GETTIME_TIMEVAL(&asoc->start_time); + asoc->discontinuity_time = asoc->start_time; + for (i = 0; i < SCTP_PR_SCTP_MAX + 1; i++) { + asoc->abandoned_unsent[i] = 0; + asoc->abandoned_sent[i] = 0; + } + /* sa_ignore MEMLEAK {memory is put in the assoc mapping array and freed later when + * the association is freed. + */ + return (0); +} + +void +sctp_print_mapping_array(struct sctp_association *asoc) +{ + unsigned int i, limit; + + SCTP_PRINTF("Mapping array size: %d, baseTSN: %8.8x, cumAck: %8.8x, highestTSN: (%8.8x, %8.8x).\n", + asoc->mapping_array_size, + asoc->mapping_array_base_tsn, + asoc->cumulative_tsn, + asoc->highest_tsn_inside_map, + asoc->highest_tsn_inside_nr_map); + for (limit = asoc->mapping_array_size; limit > 1; limit--) { + if (asoc->mapping_array[limit - 1] != 0) { + break; + } + } + SCTP_PRINTF("Renegable mapping array (last %d entries are zero):\n", asoc->mapping_array_size - limit); + for (i = 0; i < limit; i++) { + SCTP_PRINTF("%2.2x%c", asoc->mapping_array[i], ((i + 1) % 16) ? ' ' : '\n'); + } + if (limit % 16) + SCTP_PRINTF("\n"); + for (limit = asoc->mapping_array_size; limit > 1; limit--) { + if (asoc->nr_mapping_array[limit - 1]) { + break; + } + } + SCTP_PRINTF("Non renegable mapping array (last %d entries are zero):\n", asoc->mapping_array_size - limit); + for (i = 0; i < limit; i++) { + SCTP_PRINTF("%2.2x%c", asoc->nr_mapping_array[i], ((i + 1) % 16) ? ' ': '\n'); + } + if (limit % 16) + SCTP_PRINTF("\n"); +} + +int +sctp_expand_mapping_array(struct sctp_association *asoc, uint32_t needed) +{ + /* mapping array needs to grow */ + uint8_t *new_array1, *new_array2; + uint32_t new_size; + + new_size = asoc->mapping_array_size + ((needed+7)/8 + SCTP_MAPPING_ARRAY_INCR); + SCTP_MALLOC(new_array1, uint8_t *, new_size, SCTP_M_MAP); + SCTP_MALLOC(new_array2, uint8_t *, new_size, SCTP_M_MAP); + if ((new_array1 == NULL) || (new_array2 == NULL)) { + /* can't get more, forget it */ + SCTP_PRINTF("No memory for expansion of SCTP mapping array %d\n", new_size); + if (new_array1) { + SCTP_FREE(new_array1, SCTP_M_MAP); + } + if (new_array2) { + SCTP_FREE(new_array2, SCTP_M_MAP); + } + return (-1); + } + memset(new_array1, 0, new_size); + memset(new_array2, 0, new_size); + memcpy(new_array1, asoc->mapping_array, asoc->mapping_array_size); + memcpy(new_array2, asoc->nr_mapping_array, asoc->mapping_array_size); + SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); + SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP); + asoc->mapping_array = new_array1; + asoc->nr_mapping_array = new_array2; + asoc->mapping_array_size = new_size; + return (0); +} + +static void +sctp_iterator_work(struct sctp_iterator *it) +{ +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif + struct sctp_inpcb *tinp; + int iteration_count = 0; + int inp_skip = 0; + int first_in = 1; + +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + SCTP_INP_INFO_RLOCK(); + SCTP_ITERATOR_LOCK(); + sctp_it_ctl.cur_it = it; + if (it->inp) { + SCTP_INP_RLOCK(it->inp); + SCTP_INP_DECR_REF(it->inp); + } + if (it->inp == NULL) { + /* iterator is complete */ +done_with_iterator: + sctp_it_ctl.cur_it = NULL; + SCTP_ITERATOR_UNLOCK(); + SCTP_INP_INFO_RUNLOCK(); + if (it->function_atend != NULL) { + (*it->function_atend) (it->pointer, it->val); + } + SCTP_FREE(it, SCTP_M_ITER); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + return; + } +select_a_new_ep: + if (first_in) { + first_in = 0; + } else { + SCTP_INP_RLOCK(it->inp); + } + while (((it->pcb_flags) && + ((it->inp->sctp_flags & it->pcb_flags) != it->pcb_flags)) || + ((it->pcb_features) && + ((it->inp->sctp_features & it->pcb_features) != it->pcb_features))) { + /* endpoint flags or features don't match, so keep looking */ + if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { + SCTP_INP_RUNLOCK(it->inp); + goto done_with_iterator; + } + tinp = it->inp; + it->inp = LIST_NEXT(it->inp, sctp_list); + it->stcb = NULL; + SCTP_INP_RUNLOCK(tinp); + if (it->inp == NULL) { + goto done_with_iterator; + } + SCTP_INP_RLOCK(it->inp); + } + /* now go through each assoc which is in the desired state */ + if (it->done_current_ep == 0) { + if (it->function_inp != NULL) + inp_skip = (*it->function_inp)(it->inp, it->pointer, it->val); + it->done_current_ep = 1; + } + if (it->stcb == NULL) { + /* run the per instance function */ + it->stcb = LIST_FIRST(&it->inp->sctp_asoc_list); + } + if ((inp_skip) || it->stcb == NULL) { + if (it->function_inp_end != NULL) { + inp_skip = (*it->function_inp_end)(it->inp, + it->pointer, + it->val); + } + SCTP_INP_RUNLOCK(it->inp); + goto no_stcb; + } + while (it->stcb != NULL) { + SCTP_TCB_LOCK(it->stcb); + if (it->asoc_state && ((it->stcb->asoc.state & it->asoc_state) != it->asoc_state)) { + /* not in the right state... keep looking */ + SCTP_TCB_UNLOCK(it->stcb); + goto next_assoc; + } + /* see if we have limited out the iterator loop */ + iteration_count++; + if (iteration_count > SCTP_ITERATOR_MAX_AT_ONCE) { + /* Pause to let others grab the lock */ + atomic_add_int(&it->stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(it->stcb); + SCTP_INP_INCR_REF(it->inp); + SCTP_INP_RUNLOCK(it->inp); + SCTP_ITERATOR_UNLOCK(); + SCTP_INP_INFO_RUNLOCK(); + SCTP_INP_INFO_RLOCK(); + SCTP_ITERATOR_LOCK(); + if (sctp_it_ctl.iterator_flags) { + /* We won't be staying here */ + SCTP_INP_DECR_REF(it->inp); + atomic_subtract_int(&it->stcb->asoc.refcnt, 1); +#if !(defined(__FreeBSD__) && !defined(__Userspace__)) + if (sctp_it_ctl.iterator_flags & + SCTP_ITERATOR_MUST_EXIT) { + goto done_with_iterator; + } +#endif + if (sctp_it_ctl.iterator_flags & + SCTP_ITERATOR_STOP_CUR_IT) { + sctp_it_ctl.iterator_flags &= ~SCTP_ITERATOR_STOP_CUR_IT; + goto done_with_iterator; + } + if (sctp_it_ctl.iterator_flags & + SCTP_ITERATOR_STOP_CUR_INP) { + sctp_it_ctl.iterator_flags &= ~SCTP_ITERATOR_STOP_CUR_INP; + goto no_stcb; + } + /* If we reach here huh? */ + SCTP_PRINTF("Unknown it ctl flag %x\n", + sctp_it_ctl.iterator_flags); + sctp_it_ctl.iterator_flags = 0; + } + SCTP_INP_RLOCK(it->inp); + SCTP_INP_DECR_REF(it->inp); + SCTP_TCB_LOCK(it->stcb); + atomic_subtract_int(&it->stcb->asoc.refcnt, 1); + iteration_count = 0; + } + KASSERT(it->inp == it->stcb->sctp_ep, + ("%s: stcb %p does not belong to inp %p, but inp %p", + __func__, it->stcb, it->inp, it->stcb->sctp_ep)); + SCTP_INP_RLOCK_ASSERT(it->inp); + SCTP_TCB_LOCK_ASSERT(it->stcb); + + /* run function on this one */ + (*it->function_assoc)(it->inp, it->stcb, it->pointer, it->val); + SCTP_INP_RLOCK_ASSERT(it->inp); + SCTP_TCB_LOCK_ASSERT(it->stcb); + + /* + * we lie here, it really needs to have its own type but + * first I must verify that this won't effect things :-0 + */ + if (it->no_chunk_output == 0) { + sctp_chunk_output(it->inp, it->stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); + SCTP_INP_RLOCK_ASSERT(it->inp); + SCTP_TCB_LOCK_ASSERT(it->stcb); + } + + SCTP_TCB_UNLOCK(it->stcb); + next_assoc: + it->stcb = LIST_NEXT(it->stcb, sctp_tcblist); + if (it->stcb == NULL) { + /* Run last function */ + if (it->function_inp_end != NULL) { + inp_skip = (*it->function_inp_end)(it->inp, + it->pointer, + it->val); + } + } + } + SCTP_INP_RUNLOCK(it->inp); + no_stcb: + /* done with all assocs on this endpoint, move on to next endpoint */ + it->done_current_ep = 0; + if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { + it->inp = NULL; + } else { + it->inp = LIST_NEXT(it->inp, sctp_list); + } + it->stcb = NULL; + if (it->inp == NULL) { + goto done_with_iterator; + } + goto select_a_new_ep; +} + +void +sctp_iterator_worker(void) +{ + struct sctp_iterator *it; + + /* This function is called with the WQ lock in place */ + sctp_it_ctl.iterator_running = 1; + while ((it = TAILQ_FIRST(&sctp_it_ctl.iteratorhead)) != NULL) { + /* now lets work on this one */ + TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); + SCTP_IPI_ITERATOR_WQ_UNLOCK(); +#if defined(__FreeBSD__) && !defined(__Userspace__) + CURVNET_SET(it->vn); +#endif + sctp_iterator_work(it); +#if defined(__FreeBSD__) && !defined(__Userspace__) + CURVNET_RESTORE(); +#endif + SCTP_IPI_ITERATOR_WQ_LOCK(); +#if !defined(__FreeBSD__) && !defined(__Userspace__) + if (sctp_it_ctl.iterator_flags & SCTP_ITERATOR_MUST_EXIT) { + break; + } +#endif + /*sa_ignore FREED_MEMORY*/ + } + sctp_it_ctl.iterator_running = 0; + return; +} + +static void +sctp_handle_addr_wq(void) +{ + /* deal with the ADDR wq from the rtsock calls */ + struct sctp_laddr *wi, *nwi; + struct sctp_asconf_iterator *asc; + + SCTP_MALLOC(asc, struct sctp_asconf_iterator *, + sizeof(struct sctp_asconf_iterator), SCTP_M_ASC_IT); + if (asc == NULL) { + /* Try later, no memory */ + sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, + (struct sctp_inpcb *)NULL, + (struct sctp_tcb *)NULL, + (struct sctp_nets *)NULL); + return; + } + LIST_INIT(&asc->list_of_work); + asc->cnt = 0; + + LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) { + LIST_REMOVE(wi, sctp_nxt_addr); + LIST_INSERT_HEAD(&asc->list_of_work, wi, sctp_nxt_addr); + asc->cnt++; + } + + if (asc->cnt == 0) { + SCTP_FREE(asc, SCTP_M_ASC_IT); + } else { + int ret; + + ret = sctp_initiate_iterator(sctp_asconf_iterator_ep, + sctp_asconf_iterator_stcb, + NULL, /* No ep end for boundall */ + SCTP_PCB_FLAGS_BOUNDALL, + SCTP_PCB_ANY_FEATURES, + SCTP_ASOC_ANY_STATE, + (void *)asc, 0, + sctp_asconf_iterator_end, NULL, 0); + if (ret) { + SCTP_PRINTF("Failed to initiate iterator for handle_addr_wq\n"); + /* Freeing if we are stopping or put back on the addr_wq. */ + if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) { + sctp_asconf_iterator_end(asc, 0); + } else { + LIST_FOREACH(wi, &asc->list_of_work, sctp_nxt_addr) { + LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); + } + SCTP_FREE(asc, SCTP_M_ASC_IT); + } + } + } +} + +/*- + * The following table shows which pointers for the inp, stcb, or net are + * stored for each timer after it was started. + * + *|Name |Timer |inp |stcb|net | + *|-----------------------------|-----------------------------|----|----|----| + *|SCTP_TIMER_TYPE_SEND |net->rxt_timer |Yes |Yes |Yes | + *|SCTP_TIMER_TYPE_INIT |net->rxt_timer |Yes |Yes |Yes | + *|SCTP_TIMER_TYPE_RECV |stcb->asoc.dack_timer |Yes |Yes |No | + *|SCTP_TIMER_TYPE_SHUTDOWN |net->rxt_timer |Yes |Yes |Yes | + *|SCTP_TIMER_TYPE_HEARTBEAT |net->hb_timer |Yes |Yes |Yes | + *|SCTP_TIMER_TYPE_COOKIE |net->rxt_timer |Yes |Yes |Yes | + *|SCTP_TIMER_TYPE_NEWCOOKIE |inp->sctp_ep.signature_change|Yes |No |No | + *|SCTP_TIMER_TYPE_PATHMTURAISE |net->pmtu_timer |Yes |Yes |Yes | + *|SCTP_TIMER_TYPE_SHUTDOWNACK |net->rxt_timer |Yes |Yes |Yes | + *|SCTP_TIMER_TYPE_ASCONF |stcb->asoc.asconf_timer |Yes |Yes |Yes | + *|SCTP_TIMER_TYPE_SHUTDOWNGUARD|stcb->asoc.shut_guard_timer |Yes |Yes |No | + *|SCTP_TIMER_TYPE_AUTOCLOSE |stcb->asoc.autoclose_timer |Yes |Yes |No | + *|SCTP_TIMER_TYPE_STRRESET |stcb->asoc.strreset_timer |Yes |Yes |No | + *|SCTP_TIMER_TYPE_INPKILL |inp->sctp_ep.signature_change|Yes |No |No | + *|SCTP_TIMER_TYPE_ASOCKILL |stcb->asoc.strreset_timer |Yes |Yes |No | + *|SCTP_TIMER_TYPE_ADDR_WQ |SCTP_BASE_INFO(addr_wq_timer)|No |No |No | + *|SCTP_TIMER_TYPE_PRIM_DELETED |stcb->asoc.delete_prim_timer |Yes |Yes |No | + */ + +void +sctp_timeout_handler(void *t) +{ +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif + struct timeval tv; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; + struct sctp_nets *net; + struct sctp_timer *tmr; + struct mbuf *op_err; +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif +#if defined(__Userspace__) + struct socket *upcall_socket = NULL; +#endif + int type; + int i, secret; + bool did_output, released_asoc_reference; + + /* + * If inp, stcb or net are not NULL, then references to these were + * added when the timer was started, and must be released before this + * function returns. + */ + tmr = (struct sctp_timer *)t; + inp = (struct sctp_inpcb *)tmr->ep; + stcb = (struct sctp_tcb *)tmr->tcb; + net = (struct sctp_nets *)tmr->net; +#if defined(__FreeBSD__) && !defined(__Userspace__) + CURVNET_SET((struct vnet *)tmr->vnet); + NET_EPOCH_ENTER(et); +#endif + released_asoc_reference = false; + +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xF0, (uint8_t) tmr->type); + sctp_auditing(3, inp, stcb, net); +#endif + + /* sanity checks... */ + KASSERT(tmr->self == NULL || tmr->self == tmr, + ("sctp_timeout_handler: tmr->self corrupted")); + KASSERT(SCTP_IS_TIMER_TYPE_VALID(tmr->type), + ("sctp_timeout_handler: invalid timer type %d", tmr->type)); + type = tmr->type; + KASSERT(stcb == NULL || stcb->sctp_ep == inp, + ("sctp_timeout_handler of type %d: inp = %p, stcb->sctp_ep %p", + type, stcb, stcb->sctp_ep)); + tmr->stopped_from = 0xa001; + if ((stcb != NULL) && (stcb->asoc.state == SCTP_STATE_EMPTY)) { + SCTPDBG(SCTP_DEBUG_TIMER2, + "Timer type %d handler exiting due to CLOSED association.\n", + type); + goto out_decr; + } + tmr->stopped_from = 0xa002; + SCTPDBG(SCTP_DEBUG_TIMER2, "Timer type %d goes off.\n", type); + if (!SCTP_OS_TIMER_ACTIVE(&tmr->timer)) { + SCTPDBG(SCTP_DEBUG_TIMER2, + "Timer type %d handler exiting due to not being active.\n", + type); + goto out_decr; + } + + tmr->stopped_from = 0xa003; + if (stcb) { + SCTP_TCB_LOCK(stcb); + /* + * Release reference so that association can be freed if + * necessary below. + * This is safe now that we have acquired the lock. + */ + atomic_subtract_int(&stcb->asoc.refcnt, 1); + released_asoc_reference = true; + if ((type != SCTP_TIMER_TYPE_ASOCKILL) && + ((stcb->asoc.state == SCTP_STATE_EMPTY) || + (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED))) { + SCTPDBG(SCTP_DEBUG_TIMER2, + "Timer type %d handler exiting due to CLOSED association.\n", + type); + goto out; + } + } else if (inp != NULL) { + SCTP_INP_WLOCK(inp); + } else { + SCTP_WQ_ADDR_LOCK(); + } + + /* Record in stopped_from which timeout occurred. */ + tmr->stopped_from = type; + /* mark as being serviced now */ + if (SCTP_OS_TIMER_PENDING(&tmr->timer)) { + /* + * Callout has been rescheduled. + */ + goto out; + } + if (!SCTP_OS_TIMER_ACTIVE(&tmr->timer)) { + /* + * Not active, so no action. + */ + goto out; + } + SCTP_OS_TIMER_DEACTIVATE(&tmr->timer); + +#if defined(__Userspace__) + if ((stcb != NULL) && + ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (stcb->sctp_socket != NULL)) { + upcall_socket = stcb->sctp_socket; + SOCK_LOCK(upcall_socket); + soref(upcall_socket); + SOCK_UNLOCK(upcall_socket); + } +#endif + /* call the handler for the appropriate timer type */ + switch (type) { + case SCTP_TIMER_TYPE_SEND: + KASSERT(inp != NULL && stcb != NULL && net != NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timodata); + stcb->asoc.timodata++; + stcb->asoc.num_send_timers_up--; + if (stcb->asoc.num_send_timers_up < 0) { + stcb->asoc.num_send_timers_up = 0; + } + SCTP_TCB_LOCK_ASSERT(stcb); + if (sctp_t3rxt_timer(inp, stcb, net)) { + /* no need to unlock on tcb its gone */ + + goto out_decr; + } + SCTP_TCB_LOCK_ASSERT(stcb); +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(4, inp, stcb, net); +#endif + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); + did_output = true; + if ((stcb->asoc.num_send_timers_up == 0) && + (stcb->asoc.sent_queue_cnt > 0)) { + struct sctp_tmit_chunk *chk; + + /* + * Safeguard. If there on some on the sent queue + * somewhere but no timers running something is + * wrong... so we start a timer on the first chunk + * on the send queue on whatever net it is sent to. + */ + TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { + if (chk->whoTo != NULL) { + break; + } + } + if (chk != NULL) { + sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, chk->whoTo); + } + } + break; + case SCTP_TIMER_TYPE_INIT: + KASSERT(inp != NULL && stcb != NULL && net != NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timoinit); + stcb->asoc.timoinit++; + if (sctp_t1init_timer(inp, stcb, net)) { + /* no need to unlock on tcb its gone */ + goto out_decr; + } + did_output = false; + break; + case SCTP_TIMER_TYPE_RECV: + KASSERT(inp != NULL && stcb != NULL && net == NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timosack); + stcb->asoc.timosack++; + sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED); +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(4, inp, stcb, NULL); +#endif + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SACK_TMR, SCTP_SO_NOT_LOCKED); + did_output = true; + break; + case SCTP_TIMER_TYPE_SHUTDOWN: + KASSERT(inp != NULL && stcb != NULL && net != NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timoshutdown); + stcb->asoc.timoshutdown++; + if (sctp_shutdown_timer(inp, stcb, net)) { + /* no need to unlock on tcb its gone */ + goto out_decr; + } +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(4, inp, stcb, net); +#endif + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_NOT_LOCKED); + did_output = true; + break; + case SCTP_TIMER_TYPE_HEARTBEAT: + KASSERT(inp != NULL && stcb != NULL && net != NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timoheartbeat); + stcb->asoc.timoheartbeat++; + if (sctp_heartbeat_timer(inp, stcb, net)) { + /* no need to unlock on tcb its gone */ + goto out_decr; + } +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(4, inp, stcb, net); +#endif + if ((net->dest_state & SCTP_ADDR_NOHB) == 0) { + sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net); + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_HB_TMR, SCTP_SO_NOT_LOCKED); + did_output = true; + } else { + did_output = false; + } + break; + case SCTP_TIMER_TYPE_COOKIE: + KASSERT(inp != NULL && stcb != NULL && net != NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timocookie); + stcb->asoc.timocookie++; + if (sctp_cookie_timer(inp, stcb, net)) { + /* no need to unlock on tcb its gone */ + goto out_decr; + } +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(4, inp, stcb, net); +#endif + /* + * We consider T3 and Cookie timer pretty much the same with + * respect to where from in chunk_output. + */ + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); + did_output = true; + break; + case SCTP_TIMER_TYPE_NEWCOOKIE: + KASSERT(inp != NULL && stcb == NULL && net == NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timosecret); + (void)SCTP_GETTIME_TIMEVAL(&tv); + inp->sctp_ep.time_of_secret_change = (unsigned int)tv.tv_sec; + inp->sctp_ep.last_secret_number = + inp->sctp_ep.current_secret_number; + inp->sctp_ep.current_secret_number++; + if (inp->sctp_ep.current_secret_number >= + SCTP_HOW_MANY_SECRETS) { + inp->sctp_ep.current_secret_number = 0; + } + secret = (int)inp->sctp_ep.current_secret_number; + for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) { + inp->sctp_ep.secret_key[secret][i] = + sctp_select_initial_TSN(&inp->sctp_ep); + } + sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL); + did_output = false; + break; + case SCTP_TIMER_TYPE_PATHMTURAISE: + KASSERT(inp != NULL && stcb != NULL && net != NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timopathmtu); + sctp_pathmtu_timer(inp, stcb, net); + did_output = false; + break; + case SCTP_TIMER_TYPE_SHUTDOWNACK: + KASSERT(inp != NULL && stcb != NULL && net != NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + if (sctp_shutdownack_timer(inp, stcb, net)) { + /* no need to unlock on tcb its gone */ + goto out_decr; + } + SCTP_STAT_INCR(sctps_timoshutdownack); + stcb->asoc.timoshutdownack++; +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(4, inp, stcb, net); +#endif + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SHUT_ACK_TMR, SCTP_SO_NOT_LOCKED); + did_output = true; + break; + case SCTP_TIMER_TYPE_ASCONF: + KASSERT(inp != NULL && stcb != NULL && net != NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timoasconf); + if (sctp_asconf_timer(inp, stcb, net)) { + /* no need to unlock on tcb its gone */ + goto out_decr; + } +#ifdef SCTP_AUDITING_ENABLED + sctp_auditing(4, inp, stcb, net); +#endif + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_ASCONF_TMR, SCTP_SO_NOT_LOCKED); + did_output = true; + break; + case SCTP_TIMER_TYPE_SHUTDOWNGUARD: + KASSERT(inp != NULL && stcb != NULL && net == NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timoshutdownguard); + op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), + "Shutdown guard timer expired"); + sctp_abort_an_association(inp, stcb, op_err, true, SCTP_SO_NOT_LOCKED); + /* no need to unlock on tcb its gone */ + goto out_decr; + case SCTP_TIMER_TYPE_AUTOCLOSE: + KASSERT(inp != NULL && stcb != NULL && net == NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timoautoclose); + sctp_autoclose_timer(inp, stcb); + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED); + did_output = true; + break; + case SCTP_TIMER_TYPE_STRRESET: + KASSERT(inp != NULL && stcb != NULL && net == NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timostrmrst); + if (sctp_strreset_timer(inp, stcb)) { + /* no need to unlock on tcb its gone */ + goto out_decr; + } + sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_STRRST_TMR, SCTP_SO_NOT_LOCKED); + did_output = true; + break; + case SCTP_TIMER_TYPE_INPKILL: + KASSERT(inp != NULL && stcb == NULL && net == NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timoinpkill); + /* + * special case, take away our increment since WE are the + * killer + */ + sctp_timer_stop(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_3); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_LOCK(SCTP_INP_SO(inp), 1); +#endif + SCTP_INP_DECR_REF(inp); + SCTP_INP_WUNLOCK(inp); + sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, + SCTP_CALLED_FROM_INPKILL_TIMER); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(SCTP_INP_SO(inp), 1); +#endif + inp = NULL; + goto out_decr; + case SCTP_TIMER_TYPE_ASOCKILL: + KASSERT(inp != NULL && stcb != NULL && net == NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timoassockill); + /* Can we free it yet? */ + sctp_timer_stop(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_1); +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(inp); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTPUTIL + SCTP_LOC_2); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + /* + * free asoc, always unlocks (or destroy's) so prevent + * duplicate unlock or unlock of a free mtx :-0 + */ + stcb = NULL; + goto out_decr; + case SCTP_TIMER_TYPE_ADDR_WQ: + KASSERT(inp == NULL && stcb == NULL && net == NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + sctp_handle_addr_wq(); + did_output = true; + break; + case SCTP_TIMER_TYPE_PRIM_DELETED: + KASSERT(inp != NULL && stcb != NULL && net == NULL, + ("timeout of type %d: inp = %p, stcb = %p, net = %p", + type, inp, stcb, net)); + SCTP_STAT_INCR(sctps_timodelprim); + sctp_delete_prim_timer(inp, stcb); + did_output = false; + break; + default: +#ifdef INVARIANTS + panic("Unknown timer type %d", type); +#else + goto out; +#endif + } +#ifdef SCTP_AUDITING_ENABLED + sctp_audit_log(0xF1, (uint8_t) type); + if (inp != NULL) + sctp_auditing(5, inp, stcb, net); +#endif + if (did_output && (stcb != NULL)) { + /* + * Now we need to clean up the control chunk chain if an + * ECNE is on it. It must be marked as UNSENT again so next + * call will continue to send it until such time that we get + * a CWR, to remove it. It is, however, less likely that we + * will find a ecn echo on the chain though. + */ + sctp_fix_ecn_echo(&stcb->asoc); + } +out: + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + } else if (inp != NULL) { + SCTP_INP_WUNLOCK(inp); + } else { + SCTP_WQ_ADDR_UNLOCK(); + } + +out_decr: +#if defined(__Userspace__) + if (upcall_socket != NULL) { + if ((upcall_socket->so_upcall != NULL) && + (upcall_socket->so_error != 0)) { + (*upcall_socket->so_upcall)(upcall_socket, upcall_socket->so_upcallarg, M_NOWAIT); + } + ACCEPT_LOCK(); + SOCK_LOCK(upcall_socket); + sorele(upcall_socket); + } +#endif + /* These reference counts were incremented in sctp_timer_start(). */ + if (inp != NULL) { + SCTP_INP_DECR_REF(inp); + } + if ((stcb != NULL) && !released_asoc_reference) { + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } + if (net != NULL) { + sctp_free_remote_addr(net); + } + SCTPDBG(SCTP_DEBUG_TIMER2, "Timer type %d handler finished.\n", type); +#if defined(__FreeBSD__) && !defined(__Userspace__) + CURVNET_RESTORE(); + NET_EPOCH_EXIT(et); +#endif +} + +/*- + * The following table shows which parameters must be provided + * when calling sctp_timer_start(). For parameters not being + * provided, NULL must be used. + * + * |Name |inp |stcb|net | + * |-----------------------------|----|----|----| + * |SCTP_TIMER_TYPE_SEND |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_INIT |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_RECV |Yes |Yes |No | + * |SCTP_TIMER_TYPE_SHUTDOWN |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_HEARTBEAT |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_COOKIE |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_NEWCOOKIE |Yes |No |No | + * |SCTP_TIMER_TYPE_PATHMTURAISE |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_SHUTDOWNACK |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_ASCONF |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_SHUTDOWNGUARD|Yes |Yes |No | + * |SCTP_TIMER_TYPE_AUTOCLOSE |Yes |Yes |No | + * |SCTP_TIMER_TYPE_STRRESET |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_INPKILL |Yes |No |No | + * |SCTP_TIMER_TYPE_ASOCKILL |Yes |Yes |No | + * |SCTP_TIMER_TYPE_ADDR_WQ |No |No |No | + * |SCTP_TIMER_TYPE_PRIM_DELETED |Yes |Yes |No | + * + */ + +void +sctp_timer_start(int t_type, struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct sctp_nets *net) +{ + struct sctp_timer *tmr; + uint32_t to_ticks; + uint32_t rndval, jitter; + + KASSERT(stcb == NULL || stcb->sctp_ep == inp, + ("sctp_timer_start of type %d: inp = %p, stcb->sctp_ep %p", + t_type, stcb, stcb->sctp_ep)); + tmr = NULL; + if (stcb != NULL) { + SCTP_TCB_LOCK_ASSERT(stcb); + } else if (inp != NULL) { + SCTP_INP_WLOCK_ASSERT(inp); + } else { + SCTP_WQ_ADDR_LOCK_ASSERT(); + } + if (stcb != NULL) { + /* Don't restart timer on association that's about to be killed. */ + if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) && + (t_type != SCTP_TIMER_TYPE_ASOCKILL)) { + SCTPDBG(SCTP_DEBUG_TIMER2, + "Timer type %d not started: inp=%p, stcb=%p, net=%p (stcb deleted).\n", + t_type, inp, stcb, net); + return; + } + /* Don't restart timer on net that's been removed. */ + if (net != NULL && (net->dest_state & SCTP_ADDR_BEING_DELETED)) { + SCTPDBG(SCTP_DEBUG_TIMER2, + "Timer type %d not started: inp=%p, stcb=%p, net=%p (net deleted).\n", + t_type, inp, stcb, net); + return; + } + } + switch (t_type) { + case SCTP_TIMER_TYPE_SEND: + /* Here we use the RTO timer. */ + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &net->rxt_timer; + if (net->RTO == 0) { + to_ticks = sctp_msecs_to_ticks(stcb->asoc.initial_rto); + } else { + to_ticks = sctp_msecs_to_ticks(net->RTO); + } + break; + case SCTP_TIMER_TYPE_INIT: + /* + * Here we use the INIT timer default usually about 1 + * second. + */ + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &net->rxt_timer; + if (net->RTO == 0) { + to_ticks = sctp_msecs_to_ticks(stcb->asoc.initial_rto); + } else { + to_ticks = sctp_msecs_to_ticks(net->RTO); + } + break; + case SCTP_TIMER_TYPE_RECV: + /* + * Here we use the Delayed-Ack timer value from the inp, + * usually about 200ms. + */ + if ((inp == NULL) || (stcb == NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.dack_timer; + to_ticks = sctp_msecs_to_ticks(stcb->asoc.delayed_ack); + break; + case SCTP_TIMER_TYPE_SHUTDOWN: + /* Here we use the RTO of the destination. */ + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &net->rxt_timer; + if (net->RTO == 0) { + to_ticks = sctp_msecs_to_ticks(stcb->asoc.initial_rto); + } else { + to_ticks = sctp_msecs_to_ticks(net->RTO); + } + break; + case SCTP_TIMER_TYPE_HEARTBEAT: + /* + * The net is used here so that we can add in the RTO. Even + * though we use a different timer. We also add the HB timer + * PLUS a random jitter. + */ + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + if ((net->dest_state & SCTP_ADDR_NOHB) && + ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) { + SCTPDBG(SCTP_DEBUG_TIMER2, + "Timer type %d not started: inp=%p, stcb=%p, net=%p.\n", + t_type, inp, stcb, net); + return; + } + tmr = &net->hb_timer; + if (net->RTO == 0) { + to_ticks = stcb->asoc.initial_rto; + } else { + to_ticks = net->RTO; + } + rndval = sctp_select_initial_TSN(&inp->sctp_ep); + jitter = rndval % to_ticks; + if (to_ticks > 1) { + to_ticks >>= 1; + } + if (jitter < (UINT32_MAX - to_ticks)) { + to_ticks += jitter; + } else { + to_ticks = UINT32_MAX; + } + if (!((net->dest_state & SCTP_ADDR_UNCONFIRMED) && + (net->dest_state & SCTP_ADDR_REACHABLE)) && + ((net->dest_state & SCTP_ADDR_PF) == 0)) { + if (net->heart_beat_delay < (UINT32_MAX - to_ticks)) { + to_ticks += net->heart_beat_delay; + } else { + to_ticks = UINT32_MAX; + } + } + /* + * Now we must convert the to_ticks that are now in + * ms to ticks. + */ + to_ticks = sctp_msecs_to_ticks(to_ticks); + break; + case SCTP_TIMER_TYPE_COOKIE: + /* + * Here we can use the RTO timer from the network since one + * RTT was complete. If a retransmission happened then we will + * be using the RTO initial value. + */ + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &net->rxt_timer; + if (net->RTO == 0) { + to_ticks = sctp_msecs_to_ticks(stcb->asoc.initial_rto); + } else { + to_ticks = sctp_msecs_to_ticks(net->RTO); + } + break; + case SCTP_TIMER_TYPE_NEWCOOKIE: + /* + * Nothing needed but the endpoint here usually about 60 + * minutes. + */ + if ((inp == NULL) || (stcb != NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &inp->sctp_ep.signature_change; + to_ticks = inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_SIGNATURE]; + break; + case SCTP_TIMER_TYPE_PATHMTURAISE: + /* + * Here we use the value found in the EP for PMTUD, usually + * about 10 minutes. + */ + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + if (net->dest_state & SCTP_ADDR_NO_PMTUD) { + SCTPDBG(SCTP_DEBUG_TIMER2, + "Timer type %d not started: inp=%p, stcb=%p, net=%p.\n", + t_type, inp, stcb, net); + return; + } + tmr = &net->pmtu_timer; + to_ticks = inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_PMTU]; + break; + case SCTP_TIMER_TYPE_SHUTDOWNACK: + /* Here we use the RTO of the destination. */ + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &net->rxt_timer; + if (net->RTO == 0) { + to_ticks = sctp_msecs_to_ticks(stcb->asoc.initial_rto); + } else { + to_ticks = sctp_msecs_to_ticks(net->RTO); + } + break; + case SCTP_TIMER_TYPE_ASCONF: + /* + * Here the timer comes from the stcb but its value is from + * the net's RTO. + */ + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.asconf_timer; + if (net->RTO == 0) { + to_ticks = sctp_msecs_to_ticks(stcb->asoc.initial_rto); + } else { + to_ticks = sctp_msecs_to_ticks(net->RTO); + } + break; + case SCTP_TIMER_TYPE_SHUTDOWNGUARD: + /* + * Here we use the endpoints shutdown guard timer usually + * about 3 minutes. + */ + if ((inp == NULL) || (stcb == NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.shut_guard_timer; + if (inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] == 0) { + if (stcb->asoc.maxrto < UINT32_MAX / 5) { + to_ticks = sctp_msecs_to_ticks(5 * stcb->asoc.maxrto); + } else { + to_ticks = sctp_msecs_to_ticks(UINT32_MAX); + } + } else { + to_ticks = inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN]; + } + break; + case SCTP_TIMER_TYPE_AUTOCLOSE: + if ((inp == NULL) || (stcb == NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.autoclose_timer; + to_ticks = stcb->asoc.sctp_autoclose_ticks; + break; + case SCTP_TIMER_TYPE_STRRESET: + /* + * Here the timer comes from the stcb but its value is from + * the net's RTO. + */ + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.strreset_timer; + if (net->RTO == 0) { + to_ticks = sctp_msecs_to_ticks(stcb->asoc.initial_rto); + } else { + to_ticks = sctp_msecs_to_ticks(net->RTO); + } + break; + case SCTP_TIMER_TYPE_INPKILL: + /* + * The inp is setup to die. We re-use the signature_change + * timer since that has stopped and we are in the GONE + * state. + */ + if ((inp == NULL) || (stcb != NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &inp->sctp_ep.signature_change; + to_ticks = sctp_msecs_to_ticks(SCTP_INP_KILL_TIMEOUT); + break; + case SCTP_TIMER_TYPE_ASOCKILL: + if ((inp == NULL) || (stcb == NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.strreset_timer; + to_ticks = sctp_msecs_to_ticks(SCTP_ASOC_KILL_TIMEOUT); + break; + case SCTP_TIMER_TYPE_ADDR_WQ: + if ((inp != NULL) || (stcb != NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + /* Only 1 tick away :-) */ + tmr = &SCTP_BASE_INFO(addr_wq_timer); + to_ticks = SCTP_ADDRESS_TICK_DELAY; + break; + case SCTP_TIMER_TYPE_PRIM_DELETED: + if ((inp == NULL) || (stcb == NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_start of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.delete_prim_timer; + to_ticks = sctp_msecs_to_ticks(stcb->asoc.initial_rto); + break; + default: +#ifdef INVARIANTS + panic("Unknown timer type %d", t_type); +#else + return; +#endif + } + KASSERT(tmr != NULL, ("tmr is NULL for timer type %d", t_type)); + KASSERT(to_ticks > 0, ("to_ticks == 0 for timer type %d", t_type)); + if (SCTP_OS_TIMER_PENDING(&tmr->timer)) { + /* + * We do NOT allow you to have it already running. If it is, + * we leave the current one up unchanged. + */ + SCTPDBG(SCTP_DEBUG_TIMER2, + "Timer type %d already running: inp=%p, stcb=%p, net=%p.\n", + t_type, inp, stcb, net); + return; + } + /* At this point we can proceed. */ + if (t_type == SCTP_TIMER_TYPE_SEND) { + stcb->asoc.num_send_timers_up++; + } + tmr->stopped_from = 0; + tmr->type = t_type; + tmr->ep = (void *)inp; + tmr->tcb = (void *)stcb; + if (t_type == SCTP_TIMER_TYPE_STRRESET) { + tmr->net = NULL; + } else { + tmr->net = (void *)net; + } + tmr->self = (void *)tmr; +#if defined(__FreeBSD__) && !defined(__Userspace__) + tmr->vnet = (void *)curvnet; +#endif + tmr->ticks = sctp_get_tick_count(); + if (SCTP_OS_TIMER_START(&tmr->timer, to_ticks, sctp_timeout_handler, tmr) == 0) { + SCTPDBG(SCTP_DEBUG_TIMER2, + "Timer type %d started: ticks=%u, inp=%p, stcb=%p, net=%p.\n", + t_type, to_ticks, inp, stcb, net); + /* + * If this is a newly scheduled callout, as opposed to a + * rescheduled one, increment relevant reference counts. + */ + if (tmr->ep != NULL) { + SCTP_INP_INCR_REF(inp); + } + if (tmr->tcb != NULL) { + atomic_add_int(&stcb->asoc.refcnt, 1); + } + if (tmr->net != NULL) { + atomic_add_int(&net->ref_count, 1); + } + } else { + /* + * This should not happen, since we checked for pending + * above. + */ + SCTPDBG(SCTP_DEBUG_TIMER2, + "Timer type %d restarted: ticks=%u, inp=%p, stcb=%p, net=%p.\n", + t_type, to_ticks, inp, stcb, net); + } + return; +} + +/*- + * The following table shows which parameters must be provided + * when calling sctp_timer_stop(). For parameters not being + * provided, NULL must be used. + * + * |Name |inp |stcb|net | + * |-----------------------------|----|----|----| + * |SCTP_TIMER_TYPE_SEND |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_INIT |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_RECV |Yes |Yes |No | + * |SCTP_TIMER_TYPE_SHUTDOWN |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_HEARTBEAT |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_COOKIE |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_NEWCOOKIE |Yes |No |No | + * |SCTP_TIMER_TYPE_PATHMTURAISE |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_SHUTDOWNACK |Yes |Yes |Yes | + * |SCTP_TIMER_TYPE_ASCONF |Yes |Yes |No | + * |SCTP_TIMER_TYPE_SHUTDOWNGUARD|Yes |Yes |No | + * |SCTP_TIMER_TYPE_AUTOCLOSE |Yes |Yes |No | + * |SCTP_TIMER_TYPE_STRRESET |Yes |Yes |No | + * |SCTP_TIMER_TYPE_INPKILL |Yes |No |No | + * |SCTP_TIMER_TYPE_ASOCKILL |Yes |Yes |No | + * |SCTP_TIMER_TYPE_ADDR_WQ |No |No |No | + * |SCTP_TIMER_TYPE_PRIM_DELETED |Yes |Yes |No | + * + */ + +void +sctp_timer_stop(int t_type, struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct sctp_nets *net, uint32_t from) +{ + struct sctp_timer *tmr; + + KASSERT(stcb == NULL || stcb->sctp_ep == inp, + ("sctp_timer_stop of type %d: inp = %p, stcb->sctp_ep %p", + t_type, stcb, stcb->sctp_ep)); + if (stcb != NULL) { + SCTP_TCB_LOCK_ASSERT(stcb); + } else if (inp != NULL) { + SCTP_INP_WLOCK_ASSERT(inp); + } else { + SCTP_WQ_ADDR_LOCK_ASSERT(); + } + tmr = NULL; + switch (t_type) { + case SCTP_TIMER_TYPE_SEND: + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &net->rxt_timer; + break; + case SCTP_TIMER_TYPE_INIT: + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &net->rxt_timer; + break; + case SCTP_TIMER_TYPE_RECV: + if ((inp == NULL) || (stcb == NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.dack_timer; + break; + case SCTP_TIMER_TYPE_SHUTDOWN: + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &net->rxt_timer; + break; + case SCTP_TIMER_TYPE_HEARTBEAT: + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &net->hb_timer; + break; + case SCTP_TIMER_TYPE_COOKIE: + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &net->rxt_timer; + break; + case SCTP_TIMER_TYPE_NEWCOOKIE: + if ((inp == NULL) || (stcb != NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &inp->sctp_ep.signature_change; + break; + case SCTP_TIMER_TYPE_PATHMTURAISE: + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &net->pmtu_timer; + break; + case SCTP_TIMER_TYPE_SHUTDOWNACK: + if ((inp == NULL) || (stcb == NULL) || (net == NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &net->rxt_timer; + break; + case SCTP_TIMER_TYPE_ASCONF: + if ((inp == NULL) || (stcb == NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.asconf_timer; + break; + case SCTP_TIMER_TYPE_SHUTDOWNGUARD: + if ((inp == NULL) || (stcb == NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.shut_guard_timer; + break; + case SCTP_TIMER_TYPE_AUTOCLOSE: + if ((inp == NULL) || (stcb == NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.autoclose_timer; + break; + case SCTP_TIMER_TYPE_STRRESET: + if ((inp == NULL) || (stcb == NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.strreset_timer; + break; + case SCTP_TIMER_TYPE_INPKILL: + /* + * The inp is setup to die. We re-use the signature_change + * timer since that has stopped and we are in the GONE + * state. + */ + if ((inp == NULL) || (stcb != NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &inp->sctp_ep.signature_change; + break; + case SCTP_TIMER_TYPE_ASOCKILL: + if ((inp == NULL) || (stcb == NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.strreset_timer; + break; + case SCTP_TIMER_TYPE_ADDR_WQ: + if ((inp != NULL) || (stcb != NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &SCTP_BASE_INFO(addr_wq_timer); + break; + case SCTP_TIMER_TYPE_PRIM_DELETED: + if ((inp == NULL) || (stcb == NULL) || (net != NULL)) { +#ifdef INVARIANTS + panic("sctp_timer_stop of type %d: inp = %p, stcb = %p, net = %p", + t_type, inp, stcb, net); +#else + return; +#endif + } + tmr = &stcb->asoc.delete_prim_timer; + break; + default: +#ifdef INVARIANTS + panic("Unknown timer type %d", t_type); +#else + return; +#endif + } + KASSERT(tmr != NULL, ("tmr is NULL for timer type %d", t_type)); + if ((tmr->type != SCTP_TIMER_TYPE_NONE) && + (tmr->type != t_type)) { + /* + * Ok we have a timer that is under joint use. Cookie timer + * per chance with the SEND timer. We therefore are NOT + * running the timer that the caller wants stopped. So just + * return. + */ + SCTPDBG(SCTP_DEBUG_TIMER2, + "Shared timer type %d not running: inp=%p, stcb=%p, net=%p.\n", + t_type, inp, stcb, net); + return; + } + if ((t_type == SCTP_TIMER_TYPE_SEND) && (stcb != NULL)) { + stcb->asoc.num_send_timers_up--; + if (stcb->asoc.num_send_timers_up < 0) { + stcb->asoc.num_send_timers_up = 0; + } + } + tmr->self = NULL; + tmr->stopped_from = from; + if (SCTP_OS_TIMER_STOP(&tmr->timer) == 1) { + KASSERT(tmr->ep == inp, + ("sctp_timer_stop of type %d: inp = %p, tmr->inp = %p", + t_type, inp, tmr->ep)); + KASSERT(tmr->tcb == stcb, + ("sctp_timer_stop of type %d: stcb = %p, tmr->stcb = %p", + t_type, stcb, tmr->tcb)); + KASSERT(((t_type == SCTP_TIMER_TYPE_ASCONF) && (tmr->net != NULL)) || + ((t_type != SCTP_TIMER_TYPE_ASCONF) && (tmr->net == net)), + ("sctp_timer_stop of type %d: net = %p, tmr->net = %p", + t_type, net, tmr->net)); + SCTPDBG(SCTP_DEBUG_TIMER2, + "Timer type %d stopped: inp=%p, stcb=%p, net=%p.\n", + t_type, inp, stcb, net); + /* + * If the timer was actually stopped, decrement reference counts + * that were incremented in sctp_timer_start(). + */ + if (tmr->ep != NULL) { + tmr->ep = NULL; + SCTP_INP_DECR_REF(inp); + } + if (tmr->tcb != NULL) { + tmr->tcb = NULL; + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } + if (tmr->net != NULL) { + struct sctp_nets *tmr_net; + + /* + * Can't use net, since it doesn't work for + * SCTP_TIMER_TYPE_ASCONF. + */ + tmr_net = tmr->net; + tmr->net = NULL; + sctp_free_remote_addr(tmr_net); + } + } else { + SCTPDBG(SCTP_DEBUG_TIMER2, + "Timer type %d not stopped: inp=%p, stcb=%p, net=%p.\n", + t_type, inp, stcb, net); + } + return; +} + +uint32_t +sctp_calculate_len(struct mbuf *m) +{ + struct mbuf *at; + uint32_t tlen; + + tlen = 0; + for (at = m; at != NULL; at = SCTP_BUF_NEXT(at)) { + tlen += SCTP_BUF_LEN(at); + } + return (tlen); +} + +/* + * Given an association and starting time of the current RTT period, update + * RTO in number of msecs. net should point to the current network. + * Return 1, if an RTO update was performed, return 0 if no update was + * performed due to invalid starting point. + */ + +int +sctp_calculate_rto(struct sctp_tcb *stcb, + struct sctp_association *asoc, + struct sctp_nets *net, + struct timeval *old, + int rtt_from_sack) +{ + struct timeval now; + uint64_t rtt_us; /* RTT in us */ + int32_t rtt; /* RTT in ms */ + uint32_t new_rto; + int first_measure = 0; + + /************************/ + /* 1. calculate new RTT */ + /************************/ + /* get the current time */ + if (stcb->asoc.use_precise_time) { + (void)SCTP_GETPTIME_TIMEVAL(&now); + } else { + (void)SCTP_GETTIME_TIMEVAL(&now); + } + if ((old->tv_sec > now.tv_sec) || + ((old->tv_sec == now.tv_sec) && (old->tv_usec > now.tv_usec))) { + /* The starting point is in the future. */ + return (0); + } + timevalsub(&now, old); + rtt_us = (uint64_t)1000000 * (uint64_t)now.tv_sec + (uint64_t)now.tv_usec; + if (rtt_us > SCTP_RTO_UPPER_BOUND * 1000) { + /* The RTT is larger than a sane value. */ + return (0); + } + /* store the current RTT in us */ + net->rtt = rtt_us; + /* compute rtt in ms */ + rtt = (int32_t)(net->rtt / 1000); + if ((asoc->cc_functions.sctp_rtt_calculated) && (rtt_from_sack == SCTP_RTT_FROM_DATA)) { + /* Tell the CC module that a new update has just occurred from a sack */ + (*asoc->cc_functions.sctp_rtt_calculated)(stcb, net, &now); + } + /* Do we need to determine the lan? We do this only + * on sacks i.e. RTT being determined from data not + * non-data (HB/INIT->INITACK). + */ + if ((rtt_from_sack == SCTP_RTT_FROM_DATA) && + (net->lan_type == SCTP_LAN_UNKNOWN)) { + if (net->rtt > SCTP_LOCAL_LAN_RTT) { + net->lan_type = SCTP_LAN_INTERNET; + } else { + net->lan_type = SCTP_LAN_LOCAL; + } + } + + /***************************/ + /* 2. update RTTVAR & SRTT */ + /***************************/ + /*- + * Compute the scaled average lastsa and the + * scaled variance lastsv as described in van Jacobson + * Paper "Congestion Avoidance and Control", Annex A. + * + * (net->lastsa >> SCTP_RTT_SHIFT) is the srtt + * (net->lastsv >> SCTP_RTT_VAR_SHIFT) is the rttvar + */ + if (net->RTO_measured) { + rtt -= (net->lastsa >> SCTP_RTT_SHIFT); + net->lastsa += rtt; + if (rtt < 0) { + rtt = -rtt; + } + rtt -= (net->lastsv >> SCTP_RTT_VAR_SHIFT); + net->lastsv += rtt; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RTTVAR_LOGGING_ENABLE) { + rto_logging(net, SCTP_LOG_RTTVAR); + } + } else { + /* First RTO measurement */ + net->RTO_measured = 1; + first_measure = 1; + net->lastsa = rtt << SCTP_RTT_SHIFT; + net->lastsv = (rtt / 2) << SCTP_RTT_VAR_SHIFT; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RTTVAR_LOGGING_ENABLE) { + rto_logging(net, SCTP_LOG_INITIAL_RTT); + } + } + if (net->lastsv == 0) { + net->lastsv = SCTP_CLOCK_GRANULARITY; + } + new_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv; + if ((new_rto > SCTP_SAT_NETWORK_MIN) && + (stcb->asoc.sat_network_lockout == 0)) { + stcb->asoc.sat_network = 1; + } else if ((!first_measure) && stcb->asoc.sat_network) { + stcb->asoc.sat_network = 0; + stcb->asoc.sat_network_lockout = 1; + } + /* bound it, per C6/C7 in Section 5.3.1 */ + if (new_rto < stcb->asoc.minrto) { + new_rto = stcb->asoc.minrto; + } + if (new_rto > stcb->asoc.maxrto) { + new_rto = stcb->asoc.maxrto; + } + net->RTO = new_rto; + return (1); +} + +/* + * return a pointer to a contiguous piece of data from the given mbuf chain + * starting at 'off' for 'len' bytes. If the desired piece spans more than + * one mbuf, a copy is made at 'ptr'. caller must ensure that the buffer size + * is >= 'len' returns NULL if there there isn't 'len' bytes in the chain. + */ +caddr_t +sctp_m_getptr(struct mbuf *m, int off, int len, uint8_t * in_ptr) +{ + uint32_t count; + uint8_t *ptr; + + ptr = in_ptr; + if ((off < 0) || (len <= 0)) + return (NULL); + + /* find the desired start location */ + while ((m != NULL) && (off > 0)) { + if (off < SCTP_BUF_LEN(m)) + break; + off -= SCTP_BUF_LEN(m); + m = SCTP_BUF_NEXT(m); + } + if (m == NULL) + return (NULL); + + /* is the current mbuf large enough (eg. contiguous)? */ + if ((SCTP_BUF_LEN(m) - off) >= len) { + return (mtod(m, caddr_t) + off); + } else { + /* else, it spans more than one mbuf, so save a temp copy... */ + while ((m != NULL) && (len > 0)) { + count = min(SCTP_BUF_LEN(m) - off, len); + memcpy(ptr, mtod(m, caddr_t) + off, count); + len -= count; + ptr += count; + off = 0; + m = SCTP_BUF_NEXT(m); + } + if ((m == NULL) && (len > 0)) + return (NULL); + else + return ((caddr_t)in_ptr); + } +} + +struct sctp_paramhdr * +sctp_get_next_param(struct mbuf *m, + int offset, + struct sctp_paramhdr *pull, + int pull_limit) +{ + /* This just provides a typed signature to Peter's Pull routine */ + return ((struct sctp_paramhdr *)sctp_m_getptr(m, offset, pull_limit, + (uint8_t *) pull)); +} + +struct mbuf * +sctp_add_pad_tombuf(struct mbuf *m, int padlen) +{ + struct mbuf *m_last; + caddr_t dp; + + if (padlen > 3) { + return (NULL); + } + if (padlen <= M_TRAILINGSPACE(m)) { + /* + * The easy way. We hope the majority of the time we hit + * here :) + */ + m_last = m; + } else { + /* Hard way we must grow the mbuf chain */ + m_last = sctp_get_mbuf_for_msg(padlen, 0, M_NOWAIT, 1, MT_DATA); + if (m_last == NULL) { + return (NULL); + } + SCTP_BUF_LEN(m_last) = 0; + SCTP_BUF_NEXT(m_last) = NULL; + SCTP_BUF_NEXT(m) = m_last; + } + dp = mtod(m_last, caddr_t) + SCTP_BUF_LEN(m_last); + SCTP_BUF_LEN(m_last) += padlen; + memset(dp, 0, padlen); + return (m_last); +} + +struct mbuf * +sctp_pad_lastmbuf(struct mbuf *m, int padval, struct mbuf *last_mbuf) +{ + /* find the last mbuf in chain and pad it */ + struct mbuf *m_at; + + if (last_mbuf != NULL) { + return (sctp_add_pad_tombuf(last_mbuf, padval)); + } else { + for (m_at = m; m_at; m_at = SCTP_BUF_NEXT(m_at)) { + if (SCTP_BUF_NEXT(m_at) == NULL) { + return (sctp_add_pad_tombuf(m_at, padval)); + } + } + } + return (NULL); +} + +static void +sctp_notify_assoc_change(uint16_t state, struct sctp_tcb *stcb, + uint16_t error, struct sctp_abort_chunk *abort, + bool from_peer, bool timedout, int so_locked) +{ + struct mbuf *m_notify; + struct sctp_assoc_change *sac; + struct sctp_queued_to_read *control; + unsigned int notif_len; + uint16_t abort_len; + unsigned int i; +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif + + KASSERT(abort == NULL || from_peer, + ("sctp_notify_assoc_change: ABORT chunk provided for local termination")); + KASSERT(!from_peer || !timedout, + ("sctp_notify_assoc_change: timeouts can only be local")); + if (stcb == NULL) { + return; + } + if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_RECVASSOCEVNT)) { + notif_len = (unsigned int)sizeof(struct sctp_assoc_change); + if (abort != NULL) { + abort_len = ntohs(abort->ch.chunk_length); + /* + * Only SCTP_CHUNK_BUFFER_SIZE are guaranteed to be + * contiguous. + */ + if (abort_len > SCTP_CHUNK_BUFFER_SIZE) { + abort_len = SCTP_CHUNK_BUFFER_SIZE; + } + } else { + abort_len = 0; + } + if ((state == SCTP_COMM_UP) || (state == SCTP_RESTART)) { + notif_len += SCTP_ASSOC_SUPPORTS_MAX; + } else if ((state == SCTP_COMM_LOST) || (state == SCTP_CANT_STR_ASSOC)) { + notif_len += abort_len; + } + m_notify = sctp_get_mbuf_for_msg(notif_len, 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) { + /* Retry with smaller value. */ + notif_len = (unsigned int)sizeof(struct sctp_assoc_change); + m_notify = sctp_get_mbuf_for_msg(notif_len, 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) { + goto set_error; + } + } + SCTP_BUF_NEXT(m_notify) = NULL; + sac = mtod(m_notify, struct sctp_assoc_change *); + memset(sac, 0, notif_len); + sac->sac_type = SCTP_ASSOC_CHANGE; + sac->sac_flags = 0; + sac->sac_length = sizeof(struct sctp_assoc_change); + sac->sac_state = state; + sac->sac_error = error; + if (state == SCTP_CANT_STR_ASSOC) { + sac->sac_outbound_streams = 0; + sac->sac_inbound_streams = 0; + } else { + sac->sac_outbound_streams = stcb->asoc.streamoutcnt; + sac->sac_inbound_streams = stcb->asoc.streamincnt; + } + sac->sac_assoc_id = sctp_get_associd(stcb); + if (notif_len > sizeof(struct sctp_assoc_change)) { + if ((state == SCTP_COMM_UP) || (state == SCTP_RESTART)) { + i = 0; + if (stcb->asoc.prsctp_supported == 1) { + sac->sac_info[i++] = SCTP_ASSOC_SUPPORTS_PR; + } + if (stcb->asoc.auth_supported == 1) { + sac->sac_info[i++] = SCTP_ASSOC_SUPPORTS_AUTH; + } + if (stcb->asoc.asconf_supported == 1) { + sac->sac_info[i++] = SCTP_ASSOC_SUPPORTS_ASCONF; + } + if (stcb->asoc.idata_supported == 1) { + sac->sac_info[i++] = SCTP_ASSOC_SUPPORTS_INTERLEAVING; + } + sac->sac_info[i++] = SCTP_ASSOC_SUPPORTS_MULTIBUF; + if (stcb->asoc.reconfig_supported == 1) { + sac->sac_info[i++] = SCTP_ASSOC_SUPPORTS_RE_CONFIG; + } + sac->sac_length += i; + } else if ((state == SCTP_COMM_LOST) || (state == SCTP_CANT_STR_ASSOC)) { + memcpy(sac->sac_info, abort, abort_len); + sac->sac_length += abort_len; + } + } + SCTP_BUF_LEN(m_notify) = sac->sac_length; + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, + m_notify); + if (control != NULL) { + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, + so_locked); + } else { + sctp_m_freem(m_notify); + } + } + /* + * For 1-to-1 style sockets, we send up and error when an ABORT + * comes in. + */ +set_error: + if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) && + ((state == SCTP_COMM_LOST) || (state == SCTP_CANT_STR_ASSOC))) { + SOCK_LOCK(stcb->sctp_socket); + if (from_peer) { + if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) { + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ECONNREFUSED); + stcb->sctp_socket->so_error = ECONNREFUSED; + } else { + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ECONNRESET); + stcb->sctp_socket->so_error = ECONNRESET; + } + } else { + if (timedout) { + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ETIMEDOUT); + stcb->sctp_socket->so_error = ETIMEDOUT; + } else { + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ECONNABORTED); + stcb->sctp_socket->so_error = ECONNABORTED; + } + } + SOCK_UNLOCK(stcb->sctp_socket); + } + /* Wake ANY sleepers */ +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(stcb->sctp_ep); + if (!so_locked) { + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { + SCTP_SOCKET_UNLOCK(so, 1); + return; + } + } +#endif + if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) && + ((state == SCTP_COMM_LOST) || (state == SCTP_CANT_STR_ASSOC))) { + socantrcvmore(stcb->sctp_socket); + } + sorwakeup(stcb->sctp_socket); + sowwakeup(stcb->sctp_socket); +#if defined(__APPLE__) && !defined(__Userspace__) + if (!so_locked) { + SCTP_SOCKET_UNLOCK(so, 1); + } +#endif +} + +static void +sctp_notify_peer_addr_change(struct sctp_tcb *stcb, uint32_t state, + struct sockaddr *sa, uint32_t error, int so_locked) +{ + struct mbuf *m_notify; + struct sctp_paddr_change *spc; + struct sctp_queued_to_read *control; + + if ((stcb == NULL) || + sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_RECVPADDREVNT)) { + /* event not enabled */ + return; + } + m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_paddr_change), 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) + return; + SCTP_BUF_LEN(m_notify) = 0; + spc = mtod(m_notify, struct sctp_paddr_change *); + memset(spc, 0, sizeof(struct sctp_paddr_change)); + spc->spc_type = SCTP_PEER_ADDR_CHANGE; + spc->spc_flags = 0; + spc->spc_length = sizeof(struct sctp_paddr_change); + switch (sa->sa_family) { +#ifdef INET + case AF_INET: +#ifdef INET6 + if (sctp_is_feature_on(stcb->sctp_ep, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) { + in6_sin_2_v4mapsin6((struct sockaddr_in *)sa, + (struct sockaddr_in6 *)&spc->spc_aaddr); + } else { + memcpy(&spc->spc_aaddr, sa, sizeof(struct sockaddr_in)); + } +#else + memcpy(&spc->spc_aaddr, sa, sizeof(struct sockaddr_in)); +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: + { +#ifdef SCTP_EMBEDDED_V6_SCOPE + struct sockaddr_in6 *sin6; +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + memcpy(&spc->spc_aaddr, sa, sizeof(struct sockaddr_in6)); + +#ifdef SCTP_EMBEDDED_V6_SCOPE + sin6 = (struct sockaddr_in6 *)&spc->spc_aaddr; + if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) { + if (sin6->sin6_scope_id == 0) { + /* recover scope_id for user */ +#ifdef SCTP_KAME + (void)sa6_recoverscope(sin6); +#else + (void)in6_recoverscope(sin6, &sin6->sin6_addr, + NULL); +#endif + } else { + /* clear embedded scope_id for user */ + in6_clearscope(&sin6->sin6_addr); + } + } +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + break; + } +#endif +#if defined(__Userspace__) + case AF_CONN: + memcpy(&spc->spc_aaddr, sa, sizeof(struct sockaddr_conn)); + break; +#endif + default: + /* TSNH */ + break; + } + spc->spc_state = state; + spc->spc_error = error; + spc->spc_assoc_id = sctp_get_associd(stcb); + + SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_paddr_change); + SCTP_BUF_NEXT(m_notify) = NULL; + + /* append to socket */ + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, + m_notify); + if (control == NULL) { + /* no memory */ + sctp_m_freem(m_notify); + return; + } + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, + SCTP_READ_LOCK_NOT_HELD, + so_locked); +} + +static void +sctp_notify_send_failed(struct sctp_tcb *stcb, uint8_t sent, uint32_t error, + struct sctp_tmit_chunk *chk, int so_locked) +{ + struct mbuf *m_notify; + struct sctp_send_failed *ssf; + struct sctp_send_failed_event *ssfe; + struct sctp_queued_to_read *control; + struct sctp_chunkhdr *chkhdr; + int notifhdr_len, chk_len, chkhdr_len, padding_len, payload_len; + + if ((stcb == NULL) || + (sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_RECVSENDFAILEVNT) && + sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_RECVNSENDFAILEVNT))) { + /* event not enabled */ + return; + } + + if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_RECVNSENDFAILEVNT)) { + notifhdr_len = sizeof(struct sctp_send_failed_event); + } else { + notifhdr_len = sizeof(struct sctp_send_failed); + } + m_notify = sctp_get_mbuf_for_msg(notifhdr_len, 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) + /* no space left */ + return; + SCTP_BUF_LEN(m_notify) = notifhdr_len; + if (stcb->asoc.idata_supported) { + chkhdr_len = sizeof(struct sctp_idata_chunk); + } else { + chkhdr_len = sizeof(struct sctp_data_chunk); + } + /* Use some defaults in case we can't access the chunk header */ + if (chk->send_size >= chkhdr_len) { + payload_len = chk->send_size - chkhdr_len; + } else { + payload_len = 0; + } + padding_len = 0; + if (chk->data != NULL) { + chkhdr = mtod(chk->data, struct sctp_chunkhdr *); + if (chkhdr != NULL) { + chk_len = ntohs(chkhdr->chunk_length); + if ((chk_len >= chkhdr_len) && + (chk->send_size >= chk_len) && + (chk->send_size - chk_len < 4)) { + padding_len = chk->send_size - chk_len; + payload_len = chk->send_size - chkhdr_len - padding_len; + } + } + } + if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_RECVNSENDFAILEVNT)) { + ssfe = mtod(m_notify, struct sctp_send_failed_event *); + memset(ssfe, 0, notifhdr_len); + ssfe->ssfe_type = SCTP_SEND_FAILED_EVENT; + if (sent) { + ssfe->ssfe_flags = SCTP_DATA_SENT; + } else { + ssfe->ssfe_flags = SCTP_DATA_UNSENT; + } + ssfe->ssfe_length = (uint32_t)(notifhdr_len + payload_len); + ssfe->ssfe_error = error; + /* not exactly what the user sent in, but should be close :) */ + ssfe->ssfe_info.snd_sid = chk->rec.data.sid; + ssfe->ssfe_info.snd_flags = chk->rec.data.rcv_flags; + ssfe->ssfe_info.snd_ppid = chk->rec.data.ppid; + ssfe->ssfe_info.snd_context = chk->rec.data.context; + ssfe->ssfe_info.snd_assoc_id = sctp_get_associd(stcb); + ssfe->ssfe_assoc_id = sctp_get_associd(stcb); + } else { + ssf = mtod(m_notify, struct sctp_send_failed *); + memset(ssf, 0, notifhdr_len); + ssf->ssf_type = SCTP_SEND_FAILED; + if (sent) { + ssf->ssf_flags = SCTP_DATA_SENT; + } else { + ssf->ssf_flags = SCTP_DATA_UNSENT; + } + ssf->ssf_length = (uint32_t)(notifhdr_len + payload_len); + ssf->ssf_error = error; + /* not exactly what the user sent in, but should be close :) */ + ssf->ssf_info.sinfo_stream = chk->rec.data.sid; + ssf->ssf_info.sinfo_ssn = (uint16_t)chk->rec.data.mid; + ssf->ssf_info.sinfo_flags = chk->rec.data.rcv_flags; + ssf->ssf_info.sinfo_ppid = chk->rec.data.ppid; + ssf->ssf_info.sinfo_context = chk->rec.data.context; + ssf->ssf_info.sinfo_assoc_id = sctp_get_associd(stcb); + ssf->ssf_assoc_id = sctp_get_associd(stcb); + } + if (chk->data != NULL) { + /* Trim off the sctp chunk header (it should be there) */ + if (chk->send_size == chkhdr_len + payload_len + padding_len) { + m_adj(chk->data, chkhdr_len); + m_adj(chk->data, -padding_len); + sctp_mbuf_crush(chk->data); + chk->send_size -= (chkhdr_len + padding_len); + } + } + SCTP_BUF_NEXT(m_notify) = chk->data; + /* Steal off the mbuf */ + chk->data = NULL; + /* + * For this case, we check the actual socket buffer, since the assoc + * is going away we don't want to overfill the socket buffer for a + * non-reader + */ + if (sctp_sbspace_failedmsgs(&stcb->sctp_socket->so_rcv) < SCTP_BUF_LEN(m_notify)) { + sctp_m_freem(m_notify); + return; + } + /* append to socket */ + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, + m_notify); + if (control == NULL) { + /* no memory */ + sctp_m_freem(m_notify); + return; + } + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, + SCTP_READ_LOCK_NOT_HELD, + so_locked); +} + +static void +sctp_notify_send_failed2(struct sctp_tcb *stcb, uint32_t error, + struct sctp_stream_queue_pending *sp, int so_locked) +{ + struct mbuf *m_notify; + struct sctp_send_failed *ssf; + struct sctp_send_failed_event *ssfe; + struct sctp_queued_to_read *control; + int notifhdr_len; + + if ((stcb == NULL) || + (sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_RECVSENDFAILEVNT) && + sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_RECVNSENDFAILEVNT))) { + /* event not enabled */ + return; + } + if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_RECVNSENDFAILEVNT)) { + notifhdr_len = sizeof(struct sctp_send_failed_event); + } else { + notifhdr_len = sizeof(struct sctp_send_failed); + } + m_notify = sctp_get_mbuf_for_msg(notifhdr_len, 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) { + /* no space left */ + return; + } + SCTP_BUF_LEN(m_notify) = notifhdr_len; + if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_RECVNSENDFAILEVNT)) { + ssfe = mtod(m_notify, struct sctp_send_failed_event *); + memset(ssfe, 0, notifhdr_len); + ssfe->ssfe_type = SCTP_SEND_FAILED_EVENT; + ssfe->ssfe_flags = SCTP_DATA_UNSENT; + ssfe->ssfe_length = (uint32_t)(notifhdr_len + sp->length); + ssfe->ssfe_error = error; + /* not exactly what the user sent in, but should be close :) */ + ssfe->ssfe_info.snd_sid = sp->sid; + if (sp->some_taken) { + ssfe->ssfe_info.snd_flags = SCTP_DATA_LAST_FRAG; + } else { + ssfe->ssfe_info.snd_flags = SCTP_DATA_NOT_FRAG; + } + ssfe->ssfe_info.snd_ppid = sp->ppid; + ssfe->ssfe_info.snd_context = sp->context; + ssfe->ssfe_info.snd_assoc_id = sctp_get_associd(stcb); + ssfe->ssfe_assoc_id = sctp_get_associd(stcb); + } else { + ssf = mtod(m_notify, struct sctp_send_failed *); + memset(ssf, 0, notifhdr_len); + ssf->ssf_type = SCTP_SEND_FAILED; + ssf->ssf_flags = SCTP_DATA_UNSENT; + ssf->ssf_length = (uint32_t)(notifhdr_len + sp->length); + ssf->ssf_error = error; + /* not exactly what the user sent in, but should be close :) */ + ssf->ssf_info.sinfo_stream = sp->sid; + ssf->ssf_info.sinfo_ssn = 0; + if (sp->some_taken) { + ssf->ssf_info.sinfo_flags = SCTP_DATA_LAST_FRAG; + } else { + ssf->ssf_info.sinfo_flags = SCTP_DATA_NOT_FRAG; + } + ssf->ssf_info.sinfo_ppid = sp->ppid; + ssf->ssf_info.sinfo_context = sp->context; + ssf->ssf_info.sinfo_assoc_id = sctp_get_associd(stcb); + ssf->ssf_assoc_id = sctp_get_associd(stcb); + } + SCTP_BUF_NEXT(m_notify) = sp->data; + + /* Steal off the mbuf */ + sp->data = NULL; + /* + * For this case, we check the actual socket buffer, since the assoc + * is going away we don't want to overfill the socket buffer for a + * non-reader + */ + if (sctp_sbspace_failedmsgs(&stcb->sctp_socket->so_rcv) < SCTP_BUF_LEN(m_notify)) { + sctp_m_freem(m_notify); + return; + } + /* append to socket */ + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, + m_notify); + if (control == NULL) { + /* no memory */ + sctp_m_freem(m_notify); + return; + } + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, so_locked); +} + +static void +sctp_notify_adaptation_layer(struct sctp_tcb *stcb) +{ + struct mbuf *m_notify; + struct sctp_adaptation_event *sai; + struct sctp_queued_to_read *control; + + if ((stcb == NULL) || + sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_ADAPTATIONEVNT)) { + /* event not enabled */ + return; + } + + m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_adaption_event), 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) + /* no space left */ + return; + SCTP_BUF_LEN(m_notify) = 0; + sai = mtod(m_notify, struct sctp_adaptation_event *); + memset(sai, 0, sizeof(struct sctp_adaptation_event)); + sai->sai_type = SCTP_ADAPTATION_INDICATION; + sai->sai_flags = 0; + sai->sai_length = sizeof(struct sctp_adaptation_event); + sai->sai_adaptation_ind = stcb->asoc.peers_adaptation; + sai->sai_assoc_id = sctp_get_associd(stcb); + + SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_adaptation_event); + SCTP_BUF_NEXT(m_notify) = NULL; + + /* append to socket */ + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, + m_notify); + if (control == NULL) { + /* no memory */ + sctp_m_freem(m_notify); + return; + } + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); +} + +/* This always must be called with the read-queue LOCKED in the INP */ +static void +sctp_notify_partial_delivery_indication(struct sctp_tcb *stcb, uint32_t error, + uint32_t val, int so_locked) +{ + struct mbuf *m_notify; + struct sctp_pdapi_event *pdapi; + struct sctp_queued_to_read *control; + struct sockbuf *sb; + + if ((stcb == NULL) || + sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_PDAPIEVNT)) { + /* event not enabled */ + return; + } + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_CANT_READ) { + return; + } + + m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_pdapi_event), 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) + /* no space left */ + return; + SCTP_BUF_LEN(m_notify) = 0; + pdapi = mtod(m_notify, struct sctp_pdapi_event *); + memset(pdapi, 0, sizeof(struct sctp_pdapi_event)); + pdapi->pdapi_type = SCTP_PARTIAL_DELIVERY_EVENT; + pdapi->pdapi_flags = 0; + pdapi->pdapi_length = sizeof(struct sctp_pdapi_event); + pdapi->pdapi_indication = error; + pdapi->pdapi_stream = (val >> 16); + pdapi->pdapi_seq = (val & 0x0000ffff); + pdapi->pdapi_assoc_id = sctp_get_associd(stcb); + + SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_pdapi_event); + SCTP_BUF_NEXT(m_notify) = NULL; + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, + m_notify); + if (control == NULL) { + /* no memory */ + sctp_m_freem(m_notify); + return; + } + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sb = &stcb->sctp_socket->so_rcv; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(sb, control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBALLOC, SCTP_BUF_LEN(m_notify)); + } + sctp_sballoc(stcb, sb, m_notify); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(sb, control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBRESULT, 0); + } + control->end_added = 1; + if (stcb->asoc.control_pdapi) + TAILQ_INSERT_AFTER(&stcb->sctp_ep->read_queue, stcb->asoc.control_pdapi, control, next); + else { + /* we really should not see this case */ + TAILQ_INSERT_TAIL(&stcb->sctp_ep->read_queue, control, next); + } + if (stcb->sctp_ep && stcb->sctp_socket) { + /* This should always be the case */ +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; + + so = SCTP_INP_SO(stcb->sctp_ep); + if (!so_locked) { + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + SCTP_SOCKET_UNLOCK(so, 1); + return; + } + } +#endif + sctp_sorwakeup(stcb->sctp_ep, stcb->sctp_socket); +#if defined(__APPLE__) && !defined(__Userspace__) + if (!so_locked) { + SCTP_SOCKET_UNLOCK(so, 1); + } +#endif + } +} + +static void +sctp_notify_shutdown_event(struct sctp_tcb *stcb) +{ + struct mbuf *m_notify; + struct sctp_shutdown_event *sse; + struct sctp_queued_to_read *control; + + /* + * For TCP model AND UDP connected sockets we will send an error up + * when an SHUTDOWN completes + */ + if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { + /* mark socket closed for read/write and wakeup! */ +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; + + so = SCTP_INP_SO(stcb->sctp_ep); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { + SCTP_SOCKET_UNLOCK(so, 1); + return; + } +#endif + socantsendmore(stcb->sctp_socket); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } + if (sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT)) { + /* event not enabled */ + return; + } + + m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_shutdown_event), 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) + /* no space left */ + return; + sse = mtod(m_notify, struct sctp_shutdown_event *); + memset(sse, 0, sizeof(struct sctp_shutdown_event)); + sse->sse_type = SCTP_SHUTDOWN_EVENT; + sse->sse_flags = 0; + sse->sse_length = sizeof(struct sctp_shutdown_event); + sse->sse_assoc_id = sctp_get_associd(stcb); + + SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_shutdown_event); + SCTP_BUF_NEXT(m_notify) = NULL; + + /* append to socket */ + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, + m_notify); + if (control == NULL) { + /* no memory */ + sctp_m_freem(m_notify); + return; + } + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); +} + +static void +sctp_notify_sender_dry_event(struct sctp_tcb *stcb, + int so_locked) +{ + struct mbuf *m_notify; + struct sctp_sender_dry_event *event; + struct sctp_queued_to_read *control; + + if ((stcb == NULL) || + sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DRYEVNT)) { + /* event not enabled */ + return; + } + + m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_sender_dry_event), 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) { + /* no space left */ + return; + } + SCTP_BUF_LEN(m_notify) = 0; + event = mtod(m_notify, struct sctp_sender_dry_event *); + memset(event, 0, sizeof(struct sctp_sender_dry_event)); + event->sender_dry_type = SCTP_SENDER_DRY_EVENT; + event->sender_dry_flags = 0; + event->sender_dry_length = sizeof(struct sctp_sender_dry_event); + event->sender_dry_assoc_id = sctp_get_associd(stcb); + + SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_sender_dry_event); + SCTP_BUF_NEXT(m_notify) = NULL; + + /* append to socket */ + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, + m_notify); + if (control == NULL) { + /* no memory */ + sctp_m_freem(m_notify); + return; + } + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sctp_add_to_readq(stcb->sctp_ep, stcb, control, + &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, so_locked); +} + +void +sctp_notify_stream_reset_add(struct sctp_tcb *stcb, uint16_t numberin, uint16_t numberout, int flag) +{ + struct mbuf *m_notify; + struct sctp_queued_to_read *control; + struct sctp_stream_change_event *stradd; + + if ((stcb == NULL) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET)) { + /* If the socket is gone we are out of here. */ + return; + } + if (sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_STREAM_CHANGEEVNT)) { + /* event not enabled */ + return; + } + + if ((stcb->asoc.peer_req_out) && flag) { + /* Peer made the request, don't tell the local user */ + stcb->asoc.peer_req_out = 0; + return; + } + stcb->asoc.peer_req_out = 0; + m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_stream_change_event), 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) + /* no space left */ + return; + SCTP_BUF_LEN(m_notify) = 0; + stradd = mtod(m_notify, struct sctp_stream_change_event *); + memset(stradd, 0, sizeof(struct sctp_stream_change_event)); + stradd->strchange_type = SCTP_STREAM_CHANGE_EVENT; + stradd->strchange_flags = flag; + stradd->strchange_length = sizeof(struct sctp_stream_change_event); + stradd->strchange_assoc_id = sctp_get_associd(stcb); + stradd->strchange_instrms = numberin; + stradd->strchange_outstrms = numberout; + SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_stream_change_event); + SCTP_BUF_NEXT(m_notify) = NULL; + if (sctp_sbspace(&stcb->asoc, &stcb->sctp_socket->so_rcv) < SCTP_BUF_LEN(m_notify)) { + /* no space */ + sctp_m_freem(m_notify); + return; + } + /* append to socket */ + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, + m_notify); + if (control == NULL) { + /* no memory */ + sctp_m_freem(m_notify); + return; + } + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); +} + +void +sctp_notify_stream_reset_tsn(struct sctp_tcb *stcb, uint32_t sending_tsn, uint32_t recv_tsn, int flag) +{ + struct mbuf *m_notify; + struct sctp_queued_to_read *control; + struct sctp_assoc_reset_event *strasoc; + + if ((stcb == NULL) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET)) { + /* If the socket is gone we are out of here. */ + return; + } + if (sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_ASSOC_RESETEVNT)) { + /* event not enabled */ + return; + } + + m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_assoc_reset_event), 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) + /* no space left */ + return; + SCTP_BUF_LEN(m_notify) = 0; + strasoc = mtod(m_notify, struct sctp_assoc_reset_event *); + memset(strasoc, 0, sizeof(struct sctp_assoc_reset_event)); + strasoc->assocreset_type = SCTP_ASSOC_RESET_EVENT; + strasoc->assocreset_flags = flag; + strasoc->assocreset_length = sizeof(struct sctp_assoc_reset_event); + strasoc->assocreset_assoc_id= sctp_get_associd(stcb); + strasoc->assocreset_local_tsn = sending_tsn; + strasoc->assocreset_remote_tsn = recv_tsn; + SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_assoc_reset_event); + SCTP_BUF_NEXT(m_notify) = NULL; + if (sctp_sbspace(&stcb->asoc, &stcb->sctp_socket->so_rcv) < SCTP_BUF_LEN(m_notify)) { + /* no space */ + sctp_m_freem(m_notify); + return; + } + /* append to socket */ + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, + m_notify); + if (control == NULL) { + /* no memory */ + sctp_m_freem(m_notify); + return; + } + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); +} + +static void +sctp_notify_stream_reset(struct sctp_tcb *stcb, + int number_entries, uint16_t * list, int flag) +{ + struct mbuf *m_notify; + struct sctp_queued_to_read *control; + struct sctp_stream_reset_event *strreset; + int len; + + if ((stcb == NULL) || + (sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_STREAM_RESETEVNT))) { + /* event not enabled */ + return; + } + + m_notify = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) + /* no space left */ + return; + SCTP_BUF_LEN(m_notify) = 0; + len = sizeof(struct sctp_stream_reset_event) + (number_entries * sizeof(uint16_t)); + if (len > M_TRAILINGSPACE(m_notify)) { + /* never enough room */ + sctp_m_freem(m_notify); + return; + } + strreset = mtod(m_notify, struct sctp_stream_reset_event *); + memset(strreset, 0, len); + strreset->strreset_type = SCTP_STREAM_RESET_EVENT; + strreset->strreset_flags = flag; + strreset->strreset_length = len; + strreset->strreset_assoc_id = sctp_get_associd(stcb); + if (number_entries) { + int i; + + for (i = 0; i < number_entries; i++) { + strreset->strreset_stream_list[i] = ntohs(list[i]); + } + } + SCTP_BUF_LEN(m_notify) = len; + SCTP_BUF_NEXT(m_notify) = NULL; + if (sctp_sbspace(&stcb->asoc, &stcb->sctp_socket->so_rcv) < SCTP_BUF_LEN(m_notify)) { + /* no space */ + sctp_m_freem(m_notify); + return; + } + /* append to socket */ + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, + m_notify); + if (control == NULL) { + /* no memory */ + sctp_m_freem(m_notify); + return; + } + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); +} + +static void +sctp_notify_remote_error(struct sctp_tcb *stcb, uint16_t error, struct sctp_error_chunk *chunk) +{ + struct mbuf *m_notify; + struct sctp_remote_error *sre; + struct sctp_queued_to_read *control; + unsigned int notif_len; + uint16_t chunk_len; + + if ((stcb == NULL) || + sctp_stcb_is_feature_off(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_RECVPEERERR)) { + return; + } + if (chunk != NULL) { + chunk_len = ntohs(chunk->ch.chunk_length); + /* + * Only SCTP_CHUNK_BUFFER_SIZE are guaranteed to be + * contiguous. + */ + if (chunk_len > SCTP_CHUNK_BUFFER_SIZE) { + chunk_len = SCTP_CHUNK_BUFFER_SIZE; + } + } else { + chunk_len = 0; + } + notif_len = (unsigned int)(sizeof(struct sctp_remote_error) + chunk_len); + m_notify = sctp_get_mbuf_for_msg(notif_len, 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) { + /* Retry with smaller value. */ + notif_len = (unsigned int)sizeof(struct sctp_remote_error); + m_notify = sctp_get_mbuf_for_msg(notif_len, 0, M_NOWAIT, 1, MT_DATA); + if (m_notify == NULL) { + return; + } + } + SCTP_BUF_NEXT(m_notify) = NULL; + sre = mtod(m_notify, struct sctp_remote_error *); + memset(sre, 0, notif_len); + sre->sre_type = SCTP_REMOTE_ERROR; + sre->sre_flags = 0; + sre->sre_length = sizeof(struct sctp_remote_error); + sre->sre_error = error; + sre->sre_assoc_id = sctp_get_associd(stcb); + if (notif_len > sizeof(struct sctp_remote_error)) { + memcpy(sre->sre_data, chunk, chunk_len); + sre->sre_length += chunk_len; + } + SCTP_BUF_LEN(m_notify) = sre->sre_length; + control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, + 0, 0, stcb->asoc.context, 0, 0, 0, + m_notify); + if (control != NULL) { + control->length = SCTP_BUF_LEN(m_notify); + control->spec_flags = M_NOTIFICATION; + /* not that we need this */ + control->tail_mbuf = m_notify; + sctp_add_to_readq(stcb->sctp_ep, stcb, + control, + &stcb->sctp_socket->so_rcv, 1, + SCTP_READ_LOCK_NOT_HELD, SCTP_SO_NOT_LOCKED); + } else { + sctp_m_freem(m_notify); + } +} + +void +sctp_ulp_notify(uint32_t notification, struct sctp_tcb *stcb, + uint32_t error, void *data, int so_locked) +{ + if ((stcb == NULL) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET)) { + /* If the socket is gone we are out of here */ + return; + } +#if (defined(__FreeBSD__) || defined(_WIN32)) && !defined(__Userspace__) + if (stcb->sctp_socket->so_rcv.sb_state & SBS_CANTRCVMORE) { +#else + if (stcb->sctp_socket->so_state & SS_CANTRCVMORE) { +#endif + return; + } +#if defined(__APPLE__) && !defined(__Userspace__) + if (so_locked) { + sctp_lock_assert(SCTP_INP_SO(stcb->sctp_ep)); + } else { + sctp_unlock_assert(SCTP_INP_SO(stcb->sctp_ep)); + } +#endif + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) { + if ((notification == SCTP_NOTIFY_INTERFACE_DOWN) || + (notification == SCTP_NOTIFY_INTERFACE_UP) || + (notification == SCTP_NOTIFY_INTERFACE_CONFIRMED)) { + /* Don't report these in front states */ + return; + } + } + switch (notification) { + case SCTP_NOTIFY_ASSOC_UP: + if (stcb->asoc.assoc_up_sent == 0) { + sctp_notify_assoc_change(SCTP_COMM_UP, stcb, error, NULL, false, false, so_locked); + stcb->asoc.assoc_up_sent = 1; + } + if (stcb->asoc.adaptation_needed && (stcb->asoc.adaptation_sent == 0)) { + sctp_notify_adaptation_layer(stcb); + } + if (stcb->asoc.auth_supported == 0) { + sctp_ulp_notify(SCTP_NOTIFY_NO_PEER_AUTH, stcb, 0, + NULL, so_locked); + } + break; + case SCTP_NOTIFY_ASSOC_DOWN: + sctp_notify_assoc_change(SCTP_SHUTDOWN_COMP, stcb, error, NULL, false, false, so_locked); +#if defined(__Userspace__) + if (stcb->sctp_ep->recv_callback) { + if (stcb->sctp_socket) { + union sctp_sockstore addr; + struct sctp_rcvinfo rcv; + + memset(&addr, 0, sizeof(union sctp_sockstore)); + memset(&rcv, 0, sizeof(struct sctp_rcvinfo)); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + stcb->sctp_ep->recv_callback(stcb->sctp_socket, addr, NULL, 0, rcv, 0, stcb->sctp_ep->ulp_info); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } + } +#endif + break; + case SCTP_NOTIFY_INTERFACE_DOWN: + { + struct sctp_nets *net; + + net = (struct sctp_nets *)data; + sctp_notify_peer_addr_change(stcb, SCTP_ADDR_UNREACHABLE, + (struct sockaddr *)&net->ro._l_addr, error, so_locked); + break; + } + case SCTP_NOTIFY_INTERFACE_UP: + { + struct sctp_nets *net; + + net = (struct sctp_nets *)data; + sctp_notify_peer_addr_change(stcb, SCTP_ADDR_AVAILABLE, + (struct sockaddr *)&net->ro._l_addr, error, so_locked); + break; + } + case SCTP_NOTIFY_INTERFACE_CONFIRMED: + { + struct sctp_nets *net; + + net = (struct sctp_nets *)data; + sctp_notify_peer_addr_change(stcb, SCTP_ADDR_CONFIRMED, + (struct sockaddr *)&net->ro._l_addr, error, so_locked); + break; + } + case SCTP_NOTIFY_SPECIAL_SP_FAIL: + sctp_notify_send_failed2(stcb, error, + (struct sctp_stream_queue_pending *)data, so_locked); + break; + case SCTP_NOTIFY_SENT_DG_FAIL: + sctp_notify_send_failed(stcb, 1, error, + (struct sctp_tmit_chunk *)data, so_locked); + break; + case SCTP_NOTIFY_UNSENT_DG_FAIL: + sctp_notify_send_failed(stcb, 0, error, + (struct sctp_tmit_chunk *)data, so_locked); + break; + case SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION: + { + uint32_t val; + val = *((uint32_t *)data); + + sctp_notify_partial_delivery_indication(stcb, error, val, so_locked); + break; + } + case SCTP_NOTIFY_ASSOC_LOC_ABORTED: + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) { + sctp_notify_assoc_change(SCTP_CANT_STR_ASSOC, stcb, error, data, false, false, so_locked); + } else { + sctp_notify_assoc_change(SCTP_COMM_LOST, stcb, error, data, false, false, so_locked); + } + break; + case SCTP_NOTIFY_ASSOC_REM_ABORTED: + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) { + sctp_notify_assoc_change(SCTP_CANT_STR_ASSOC, stcb, error, data, true, false, so_locked); + } else { + sctp_notify_assoc_change(SCTP_COMM_LOST, stcb, error, data, true, false, so_locked); + } + break; + case SCTP_NOTIFY_ASSOC_TIMEDOUT: + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) { + sctp_notify_assoc_change(SCTP_CANT_STR_ASSOC, stcb, error, data, false, true, so_locked); + } else { + sctp_notify_assoc_change(SCTP_COMM_LOST, stcb, error, data, false, true, so_locked); + } + break; + case SCTP_NOTIFY_ASSOC_RESTART: + sctp_notify_assoc_change(SCTP_RESTART, stcb, error, NULL, false, false, so_locked); + if (stcb->asoc.auth_supported == 0) { + sctp_ulp_notify(SCTP_NOTIFY_NO_PEER_AUTH, stcb, 0, + NULL, so_locked); + } + break; + case SCTP_NOTIFY_STR_RESET_SEND: + sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), SCTP_STREAM_RESET_OUTGOING_SSN); + break; + case SCTP_NOTIFY_STR_RESET_RECV: + sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), SCTP_STREAM_RESET_INCOMING); + break; + case SCTP_NOTIFY_STR_RESET_FAILED_OUT: + sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), + (SCTP_STREAM_RESET_OUTGOING_SSN|SCTP_STREAM_RESET_FAILED)); + break; + case SCTP_NOTIFY_STR_RESET_DENIED_OUT: + sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), + (SCTP_STREAM_RESET_OUTGOING_SSN|SCTP_STREAM_RESET_DENIED)); + break; + case SCTP_NOTIFY_STR_RESET_FAILED_IN: + sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), + (SCTP_STREAM_RESET_INCOMING|SCTP_STREAM_RESET_FAILED)); + break; + case SCTP_NOTIFY_STR_RESET_DENIED_IN: + sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), + (SCTP_STREAM_RESET_INCOMING|SCTP_STREAM_RESET_DENIED)); + break; + case SCTP_NOTIFY_ASCONF_ADD_IP: + sctp_notify_peer_addr_change(stcb, SCTP_ADDR_ADDED, data, + error, so_locked); + break; + case SCTP_NOTIFY_ASCONF_DELETE_IP: + sctp_notify_peer_addr_change(stcb, SCTP_ADDR_REMOVED, data, + error, so_locked); + break; + case SCTP_NOTIFY_ASCONF_SET_PRIMARY: + sctp_notify_peer_addr_change(stcb, SCTP_ADDR_MADE_PRIM, data, + error, so_locked); + break; + case SCTP_NOTIFY_PEER_SHUTDOWN: + sctp_notify_shutdown_event(stcb); + break; + case SCTP_NOTIFY_AUTH_NEW_KEY: + sctp_notify_authentication(stcb, SCTP_AUTH_NEW_KEY, error, + (uint16_t)(uintptr_t)data, + so_locked); + break; + case SCTP_NOTIFY_AUTH_FREE_KEY: + sctp_notify_authentication(stcb, SCTP_AUTH_FREE_KEY, error, + (uint16_t)(uintptr_t)data, + so_locked); + break; + case SCTP_NOTIFY_NO_PEER_AUTH: + sctp_notify_authentication(stcb, SCTP_AUTH_NO_AUTH, error, + (uint16_t)(uintptr_t)data, + so_locked); + break; + case SCTP_NOTIFY_SENDER_DRY: + sctp_notify_sender_dry_event(stcb, so_locked); + break; + case SCTP_NOTIFY_REMOTE_ERROR: + sctp_notify_remote_error(stcb, error, data); + break; + default: + SCTPDBG(SCTP_DEBUG_UTIL1, "%s: unknown notification %xh (%u)\n", + __func__, notification, notification); + break; + } /* end switch */ +} + +void +sctp_report_all_outbound(struct sctp_tcb *stcb, uint16_t error, int so_locked) +{ + struct sctp_association *asoc; + struct sctp_stream_out *outs; + struct sctp_tmit_chunk *chk, *nchk; + struct sctp_stream_queue_pending *sp, *nsp; + int i; + + if (stcb == NULL) { + return; + } + asoc = &stcb->asoc; + if (asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) { + /* already being freed */ + return; + } +#if defined(__APPLE__) && !defined(__Userspace__) + if (so_locked) { + sctp_lock_assert(SCTP_INP_SO(stcb->sctp_ep)); + } else { + sctp_unlock_assert(SCTP_INP_SO(stcb->sctp_ep)); + } +#endif + if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (asoc->state & SCTP_STATE_CLOSED_SOCKET)) { + return; + } + /* now through all the gunk freeing chunks */ + /* sent queue SHOULD be empty */ + TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) { + TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next); + asoc->sent_queue_cnt--; + if (chk->sent != SCTP_DATAGRAM_NR_ACKED) { + if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) { + asoc->strmout[chk->rec.data.sid].chunks_on_queues--; +#ifdef INVARIANTS + } else { + panic("No chunks on the queues for sid %u.", chk->rec.data.sid); +#endif + } + } + if (chk->data != NULL) { + sctp_free_bufspace(stcb, asoc, chk, 1); + sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb, + error, chk, so_locked); + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + } + sctp_free_a_chunk(stcb, chk, so_locked); + /*sa_ignore FREED_MEMORY*/ + } + /* pending send queue SHOULD be empty */ + TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) { + TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next); + asoc->send_queue_cnt--; + if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) { + asoc->strmout[chk->rec.data.sid].chunks_on_queues--; +#ifdef INVARIANTS + } else { + panic("No chunks on the queues for sid %u.", chk->rec.data.sid); +#endif + } + if (chk->data != NULL) { + sctp_free_bufspace(stcb, asoc, chk, 1); + sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb, + error, chk, so_locked); + if (chk->data) { + sctp_m_freem(chk->data); + chk->data = NULL; + } + } + sctp_free_a_chunk(stcb, chk, so_locked); + /*sa_ignore FREED_MEMORY*/ + } + for (i = 0; i < asoc->streamoutcnt; i++) { + /* For each stream */ + outs = &asoc->strmout[i]; + /* clean up any sends there */ + TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) { + atomic_subtract_int(&asoc->stream_queue_cnt, 1); + TAILQ_REMOVE(&outs->outqueue, sp, next); + stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, outs, sp); + sctp_free_spbufspace(stcb, asoc, sp); + if (sp->data) { + sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb, + error, (void *)sp, so_locked); + if (sp->data) { + sctp_m_freem(sp->data); + sp->data = NULL; + sp->tail_mbuf = NULL; + sp->length = 0; + } + } + if (sp->net) { + sctp_free_remote_addr(sp->net); + sp->net = NULL; + } + /* Free the chunk */ + sctp_free_a_strmoq(stcb, sp, so_locked); + /*sa_ignore FREED_MEMORY*/ + } + } +} + +void +sctp_abort_notification(struct sctp_tcb *stcb, bool from_peer, bool timeout, + uint16_t error, struct sctp_abort_chunk *abort, + int so_locked) +{ + if (stcb == NULL) { + return; + } +#if defined(__APPLE__) && !defined(__Userspace__) + if (so_locked) { + sctp_lock_assert(SCTP_INP_SO(stcb->sctp_ep)); + } else { + sctp_unlock_assert(SCTP_INP_SO(stcb->sctp_ep)); + } +#endif + SCTP_TCB_LOCK_ASSERT(stcb); + + if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || + ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) { + sctp_pcb_add_flags(stcb->sctp_ep, SCTP_PCB_FLAGS_WAS_ABORTED); + } + if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || + (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET)) { + return; + } + SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_WAS_ABORTED); + /* Tell them we lost the asoc */ + sctp_report_all_outbound(stcb, error, so_locked); + if (from_peer) { + sctp_ulp_notify(SCTP_NOTIFY_ASSOC_REM_ABORTED, stcb, error, abort, so_locked); + } else { + if (timeout) { + sctp_ulp_notify(SCTP_NOTIFY_ASSOC_TIMEDOUT, stcb, error, abort, so_locked); + } else { + sctp_ulp_notify(SCTP_NOTIFY_ASSOC_LOC_ABORTED, stcb, error, abort, so_locked); + } + } +} + +void +sctp_abort_association(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct mbuf *m, int iphlen, + struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, struct mbuf *op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, +#endif + uint32_t vrf_id, uint16_t port) +{ +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif + struct sctp_gen_error_cause* cause; + uint32_t vtag; + uint16_t cause_code; + + if (stcb != NULL) { + vtag = stcb->asoc.peer_vtag; + vrf_id = stcb->asoc.vrf_id; + if (op_err != NULL) { + /* Read the cause code from the error cause. */ + cause = mtod(op_err, struct sctp_gen_error_cause *); + cause_code = ntohs(cause->code); + } else { + cause_code = 0; + } + } else { + vtag = 0; + } + sctp_send_abort(m, iphlen, src, dst, sh, vtag, op_err, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, inp->fibnum, +#endif + vrf_id, port); + if (stcb != NULL) { + /* We have a TCB to abort, send notification too */ + sctp_abort_notification(stcb, false, false, cause_code, NULL, SCTP_SO_NOT_LOCKED); + /* Ok, now lets free it */ +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(inp); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + SCTP_STAT_INCR_COUNTER32(sctps_aborted); + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTPUTIL + SCTP_LOC_4); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } +} +#ifdef SCTP_ASOCLOG_OF_TSNS +void +sctp_print_out_track_log(struct sctp_tcb *stcb) +{ +#ifdef NOSIY_PRINTS + int i; + SCTP_PRINTF("Last ep reason:%x\n", stcb->sctp_ep->last_abort_code); + SCTP_PRINTF("IN bound TSN log-aaa\n"); + if ((stcb->asoc.tsn_in_at == 0) && (stcb->asoc.tsn_in_wrapped == 0)) { + SCTP_PRINTF("None rcvd\n"); + goto none_in; + } + if (stcb->asoc.tsn_in_wrapped) { + for (i = stcb->asoc.tsn_in_at; i < SCTP_TSN_LOG_SIZE; i++) { + SCTP_PRINTF("TSN:%x strm:%d seq:%d flags:%x sz:%d\n", + stcb->asoc.in_tsnlog[i].tsn, + stcb->asoc.in_tsnlog[i].strm, + stcb->asoc.in_tsnlog[i].seq, + stcb->asoc.in_tsnlog[i].flgs, + stcb->asoc.in_tsnlog[i].sz); + } + } + if (stcb->asoc.tsn_in_at) { + for (i = 0; i < stcb->asoc.tsn_in_at; i++) { + SCTP_PRINTF("TSN:%x strm:%d seq:%d flags:%x sz:%d\n", + stcb->asoc.in_tsnlog[i].tsn, + stcb->asoc.in_tsnlog[i].strm, + stcb->asoc.in_tsnlog[i].seq, + stcb->asoc.in_tsnlog[i].flgs, + stcb->asoc.in_tsnlog[i].sz); + } + } + none_in: + SCTP_PRINTF("OUT bound TSN log-aaa\n"); + if ((stcb->asoc.tsn_out_at == 0) && + (stcb->asoc.tsn_out_wrapped == 0)) { + SCTP_PRINTF("None sent\n"); + } + if (stcb->asoc.tsn_out_wrapped) { + for (i = stcb->asoc.tsn_out_at; i < SCTP_TSN_LOG_SIZE; i++) { + SCTP_PRINTF("TSN:%x strm:%d seq:%d flags:%x sz:%d\n", + stcb->asoc.out_tsnlog[i].tsn, + stcb->asoc.out_tsnlog[i].strm, + stcb->asoc.out_tsnlog[i].seq, + stcb->asoc.out_tsnlog[i].flgs, + stcb->asoc.out_tsnlog[i].sz); + } + } + if (stcb->asoc.tsn_out_at) { + for (i = 0; i < stcb->asoc.tsn_out_at; i++) { + SCTP_PRINTF("TSN:%x strm:%d seq:%d flags:%x sz:%d\n", + stcb->asoc.out_tsnlog[i].tsn, + stcb->asoc.out_tsnlog[i].strm, + stcb->asoc.out_tsnlog[i].seq, + stcb->asoc.out_tsnlog[i].flgs, + stcb->asoc.out_tsnlog[i].sz); + } + } +#endif +} +#endif + +void +sctp_abort_an_association(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + struct mbuf *op_err, bool timedout, int so_locked) +{ +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; +#endif + struct sctp_gen_error_cause* cause; + uint16_t cause_code; + +#if defined(__APPLE__) && !defined(__Userspace__) + so = SCTP_INP_SO(inp); +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + if (so_locked) { + sctp_lock_assert(SCTP_INP_SO(inp)); + } else { + sctp_unlock_assert(SCTP_INP_SO(inp)); + } +#endif + if (stcb == NULL) { + /* Got to have a TCB */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + if (LIST_EMPTY(&inp->sctp_asoc_list)) { +#if defined(__APPLE__) && !defined(__Userspace__) + if (!so_locked) { + SCTP_SOCKET_LOCK(so, 1); + } +#endif + sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, + SCTP_CALLED_DIRECTLY_NOCMPSET); +#if defined(__APPLE__) && !defined(__Userspace__) + if (!so_locked) { + SCTP_SOCKET_UNLOCK(so, 1); + } +#endif + } + } + return; + } + if (op_err != NULL) { + /* Read the cause code from the error cause. */ + cause = mtod(op_err, struct sctp_gen_error_cause *); + cause_code = ntohs(cause->code); + } else { + cause_code = 0; + } + /* notify the peer */ + sctp_send_abort_tcb(stcb, op_err, so_locked); + SCTP_STAT_INCR_COUNTER32(sctps_aborted); + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) || + (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) { + SCTP_STAT_DECR_GAUGE32(sctps_currestab); + } + /* notify the ulp */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) { + sctp_abort_notification(stcb, false, timedout, cause_code, NULL, so_locked); + } + /* now free the asoc */ +#ifdef SCTP_ASOCLOG_OF_TSNS + sctp_print_out_track_log(stcb); +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + if (!so_locked) { + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } +#endif + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTPUTIL + SCTP_LOC_5); +#if defined(__APPLE__) && !defined(__Userspace__) + if (!so_locked) { + SCTP_SOCKET_UNLOCK(so, 1); + } +#endif +} + +void +sctp_handle_ootb(struct mbuf *m, int iphlen, int offset, + struct sockaddr *src, struct sockaddr *dst, + struct sctphdr *sh, struct sctp_inpcb *inp, + struct mbuf *cause, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t mflowtype, uint32_t mflowid, uint16_t fibnum, +#endif + uint32_t vrf_id, uint16_t port) +{ + struct sctp_chunkhdr *ch, chunk_buf; + unsigned int chk_length; + int contains_init_chunk; + + SCTP_STAT_INCR_COUNTER32(sctps_outoftheblue); + /* Generate a TO address for future reference */ + if (inp && (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { + if (LIST_EMPTY(&inp->sctp_asoc_list)) { +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_LOCK(SCTP_INP_SO(inp), 1); +#endif + sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, + SCTP_CALLED_DIRECTLY_NOCMPSET); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(SCTP_INP_SO(inp), 1); +#endif + } + } + contains_init_chunk = 0; + ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, + sizeof(*ch), (uint8_t *) & chunk_buf); + while (ch != NULL) { + chk_length = ntohs(ch->chunk_length); + if (chk_length < sizeof(*ch)) { + /* break to abort land */ + break; + } + switch (ch->chunk_type) { + case SCTP_INIT: + contains_init_chunk = 1; + break; + case SCTP_PACKET_DROPPED: + /* we don't respond to pkt-dropped */ + return; + case SCTP_ABORT_ASSOCIATION: + /* we don't respond with an ABORT to an ABORT */ + return; + case SCTP_SHUTDOWN_COMPLETE: + /* + * we ignore it since we are not waiting for it and + * peer is gone + */ + return; + case SCTP_SHUTDOWN_ACK: + sctp_send_shutdown_complete2(src, dst, sh, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, fibnum, +#endif + vrf_id, port); + return; + default: + break; + } + offset += SCTP_SIZE32(chk_length); + ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, + sizeof(*ch), (uint8_t *) & chunk_buf); + } + if ((SCTP_BASE_SYSCTL(sctp_blackhole) == 0) || + ((SCTP_BASE_SYSCTL(sctp_blackhole) == 1) && + (contains_init_chunk == 0))) { + sctp_send_abort(m, iphlen, src, dst, sh, 0, cause, +#if defined(__FreeBSD__) && !defined(__Userspace__) + mflowtype, mflowid, fibnum, +#endif + vrf_id, port); + } +} + +/* + * check the inbound datagram to make sure there is not an abort inside it, + * if there is return 1, else return 0. + */ +int +sctp_is_there_an_abort_here(struct mbuf *m, int iphlen, uint32_t *vtag) +{ + struct sctp_chunkhdr *ch; + struct sctp_init_chunk *init_chk, chunk_buf; + int offset; + unsigned int chk_length; + + offset = iphlen + sizeof(struct sctphdr); + ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), + (uint8_t *) & chunk_buf); + while (ch != NULL) { + chk_length = ntohs(ch->chunk_length); + if (chk_length < sizeof(*ch)) { + /* packet is probably corrupt */ + break; + } + /* we seem to be ok, is it an abort? */ + if (ch->chunk_type == SCTP_ABORT_ASSOCIATION) { + /* yep, tell them */ + return (1); + } + if ((ch->chunk_type == SCTP_INITIATION) || + (ch->chunk_type == SCTP_INITIATION_ACK)) { + /* need to update the Vtag */ + init_chk = (struct sctp_init_chunk *)sctp_m_getptr(m, + offset, sizeof(struct sctp_init_chunk), (uint8_t *) & chunk_buf); + if (init_chk != NULL) { + *vtag = ntohl(init_chk->init.initiate_tag); + } + } + /* Nope, move to the next chunk */ + offset += SCTP_SIZE32(chk_length); + ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, + sizeof(*ch), (uint8_t *) & chunk_buf); + } + return (0); +} + +/* + * currently (2/02), ifa_addr embeds scope_id's and don't have sin6_scope_id + * set (i.e. it's 0) so, create this function to compare link local scopes + */ +#ifdef INET6 +uint32_t +sctp_is_same_scope(struct sockaddr_in6 *addr1, struct sockaddr_in6 *addr2) +{ +#if defined(__Userspace__) + /*__Userspace__ Returning 1 here always */ +#endif +#if defined(SCTP_EMBEDDED_V6_SCOPE) + struct sockaddr_in6 a, b; + + /* save copies */ + a = *addr1; + b = *addr2; + + if (a.sin6_scope_id == 0) +#ifdef SCTP_KAME + if (sa6_recoverscope(&a)) { +#else + if (in6_recoverscope(&a, &a.sin6_addr, NULL)) { +#endif /* SCTP_KAME */ + /* can't get scope, so can't match */ + return (0); + } + if (b.sin6_scope_id == 0) +#ifdef SCTP_KAME + if (sa6_recoverscope(&b)) { +#else + if (in6_recoverscope(&b, &b.sin6_addr, NULL)) { +#endif /* SCTP_KAME */ + /* can't get scope, so can't match */ + return (0); + } + if (a.sin6_scope_id != b.sin6_scope_id) + return (0); +#else + if (addr1->sin6_scope_id != addr2->sin6_scope_id) + return (0); +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + + return (1); +} + +#if defined(SCTP_EMBEDDED_V6_SCOPE) +/* + * returns a sockaddr_in6 with embedded scope recovered and removed + */ +struct sockaddr_in6 * +sctp_recover_scope(struct sockaddr_in6 *addr, struct sockaddr_in6 *store) +{ + /* check and strip embedded scope junk */ + if (addr->sin6_family == AF_INET6) { + if (IN6_IS_SCOPE_LINKLOCAL(&addr->sin6_addr)) { + if (addr->sin6_scope_id == 0) { + *store = *addr; +#ifdef SCTP_KAME + if (!sa6_recoverscope(store)) { +#else + if (!in6_recoverscope(store, &store->sin6_addr, + NULL)) { +#endif /* SCTP_KAME */ + /* use the recovered scope */ + addr = store; + } + } else { + /* else, return the original "to" addr */ + in6_clearscope(&addr->sin6_addr); + } + } + } + return (addr); +} +#endif /* SCTP_EMBEDDED_V6_SCOPE */ +#endif + +/* + * are the two addresses the same? currently a "scopeless" check returns: 1 + * if same, 0 if not + */ +int +sctp_cmpaddr(struct sockaddr *sa1, struct sockaddr *sa2) +{ + + /* must be valid */ + if (sa1 == NULL || sa2 == NULL) + return (0); + + /* must be the same family */ + if (sa1->sa_family != sa2->sa_family) + return (0); + + switch (sa1->sa_family) { +#ifdef INET6 + case AF_INET6: + { + /* IPv6 addresses */ + struct sockaddr_in6 *sin6_1, *sin6_2; + + sin6_1 = (struct sockaddr_in6 *)sa1; + sin6_2 = (struct sockaddr_in6 *)sa2; + return (SCTP6_ARE_ADDR_EQUAL(sin6_1, + sin6_2)); + } +#endif +#ifdef INET + case AF_INET: + { + /* IPv4 addresses */ + struct sockaddr_in *sin_1, *sin_2; + + sin_1 = (struct sockaddr_in *)sa1; + sin_2 = (struct sockaddr_in *)sa2; + return (sin_1->sin_addr.s_addr == sin_2->sin_addr.s_addr); + } +#endif +#if defined(__Userspace__) + case AF_CONN: + { + struct sockaddr_conn *sconn_1, *sconn_2; + + sconn_1 = (struct sockaddr_conn *)sa1; + sconn_2 = (struct sockaddr_conn *)sa2; + return (sconn_1->sconn_addr == sconn_2->sconn_addr); + } +#endif + default: + /* we don't do these... */ + return (0); + } +} + +void +sctp_print_address(struct sockaddr *sa) +{ +#ifdef INET6 +#if defined(__FreeBSD__) && !defined(__Userspace__) + char ip6buf[INET6_ADDRSTRLEN]; +#endif +#endif + + switch (sa->sa_family) { +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)sa; +#if defined(__Userspace__) + SCTP_PRINTF("IPv6 address: %x:%x:%x:%x:%x:%x:%x:%x:port:%d scope:%u\n", + ntohs(sin6->sin6_addr.s6_addr16[0]), + ntohs(sin6->sin6_addr.s6_addr16[1]), + ntohs(sin6->sin6_addr.s6_addr16[2]), + ntohs(sin6->sin6_addr.s6_addr16[3]), + ntohs(sin6->sin6_addr.s6_addr16[4]), + ntohs(sin6->sin6_addr.s6_addr16[5]), + ntohs(sin6->sin6_addr.s6_addr16[6]), + ntohs(sin6->sin6_addr.s6_addr16[7]), + ntohs(sin6->sin6_port), + sin6->sin6_scope_id); +#else +#if defined(__FreeBSD__) && !defined(__Userspace__) + SCTP_PRINTF("IPv6 address: %s:port:%d scope:%u\n", + ip6_sprintf(ip6buf, &sin6->sin6_addr), + ntohs(sin6->sin6_port), + sin6->sin6_scope_id); +#else + SCTP_PRINTF("IPv6 address: %s:port:%d scope:%u\n", + ip6_sprintf(&sin6->sin6_addr), + ntohs(sin6->sin6_port), + sin6->sin6_scope_id); +#endif +#endif + break; + } +#endif +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin; + unsigned char *p; + + sin = (struct sockaddr_in *)sa; + p = (unsigned char *)&sin->sin_addr; + SCTP_PRINTF("IPv4 address: %u.%u.%u.%u:%d\n", + p[0], p[1], p[2], p[3], ntohs(sin->sin_port)); + break; + } +#endif +#if defined(__Userspace__) + case AF_CONN: + { + struct sockaddr_conn *sconn; + + sconn = (struct sockaddr_conn *)sa; + SCTP_PRINTF("AF_CONN address: %p\n", sconn->sconn_addr); + break; + } +#endif + default: + SCTP_PRINTF("?\n"); + break; + } +} + +void +sctp_pull_off_control_to_new_inp(struct sctp_inpcb *old_inp, + struct sctp_inpcb *new_inp, + struct sctp_tcb *stcb, + int waitflags) +{ + /* + * go through our old INP and pull off any control structures that + * belong to stcb and move then to the new inp. + */ + struct socket *old_so, *new_so; + struct sctp_queued_to_read *control, *nctl; + struct sctp_readhead tmp_queue; + struct mbuf *m; +#if (defined(__FreeBSD__) || defined(__APPLE__)) && !defined(__Userspace__) + int error = 0; +#endif + + old_so = old_inp->sctp_socket; + new_so = new_inp->sctp_socket; + TAILQ_INIT(&tmp_queue); +#if (defined(__FreeBSD__) || defined(__APPLE__)) && !defined(__Userspace__) +#if defined(__FreeBSD__) + error = SOCK_IO_RECV_LOCK(old_so, waitflags); +#else + error = sblock(&old_so->so_rcv, waitflags); +#endif + if (error) { + /* Gak, can't get I/O lock, we have a problem. + * data will be left stranded.. and we + * don't dare look at it since the + * other thread may be reading something. + * Oh well, its a screwed up app that does + * a peeloff OR a accept while reading + * from the main socket... actually its + * only the peeloff() case, since I think + * read will fail on a listening socket.. + */ + return; + } +#endif + /* lock the socket buffers */ + SCTP_INP_READ_LOCK(old_inp); + TAILQ_FOREACH_SAFE(control, &old_inp->read_queue, next, nctl) { + /* Pull off all for out target stcb */ + if (control->stcb == stcb) { + /* remove it we want it */ + TAILQ_REMOVE(&old_inp->read_queue, control, next); + TAILQ_INSERT_TAIL(&tmp_queue, control, next); + m = control->data; + while (m) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(&old_so->so_rcv, control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBFREE,SCTP_BUF_LEN(m)); + } + sctp_sbfree(control, stcb, &old_so->so_rcv, m); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(&old_so->so_rcv, control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBRESULT, 0); + } + m = SCTP_BUF_NEXT(m); + } + } + } + SCTP_INP_READ_UNLOCK(old_inp); + /* Remove the recv-lock on the old socket */ +#if defined(__APPLE__) && !defined(__Userspace__) + sbunlock(&old_so->so_rcv, 1); +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + SOCK_IO_RECV_UNLOCK(old_so); +#endif + /* Now we move them over to the new socket buffer */ + SCTP_INP_READ_LOCK(new_inp); + TAILQ_FOREACH_SAFE(control, &tmp_queue, next, nctl) { + TAILQ_INSERT_TAIL(&new_inp->read_queue, control, next); + m = control->data; + while (m) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(&new_so->so_rcv, control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBALLOC, SCTP_BUF_LEN(m)); + } + sctp_sballoc(stcb, &new_so->so_rcv, m); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(&new_so->so_rcv, control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBRESULT, 0); + } + m = SCTP_BUF_NEXT(m); + } + } + SCTP_INP_READ_UNLOCK(new_inp); +} + +void +sctp_wakeup_the_read_socket(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + int so_locked +#if !(defined(__APPLE__) && !defined(__Userspace__)) + SCTP_UNUSED +#endif +) +{ + if ((inp != NULL) && + (inp->sctp_socket != NULL) && + (((inp->sctp_flags & (SCTP_PCB_FLAGS_TCPTYPE | SCTP_PCB_FLAGS_IN_TCPPOOL)) == 0) || + !SCTP_IS_LISTENING(inp))) { +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; + + so = SCTP_INP_SO(inp); + if (!so_locked) { + if (stcb) { + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + } + SCTP_SOCKET_LOCK(so, 1); + if (stcb) { + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + SCTP_SOCKET_UNLOCK(so, 1); + return; + } + } +#endif + sctp_sorwakeup(inp, inp->sctp_socket); +#if defined(__APPLE__) && !defined(__Userspace__) + if (!so_locked) { + SCTP_SOCKET_UNLOCK(so, 1); + } +#endif + } +} +#if defined(__Userspace__) + +void +sctp_invoke_recv_callback(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_queued_to_read *control, + int inp_read_lock_held) +{ + uint32_t pd_point, length; + + if ((inp->recv_callback == NULL) || + (stcb == NULL) || + (stcb->sctp_socket == NULL)) { + return; + } + + length = control->length; + if (stcb != NULL && stcb->sctp_socket != NULL) { + pd_point = min(SCTP_SB_LIMIT_RCV(stcb->sctp_socket) >> SCTP_PARTIAL_DELIVERY_SHIFT, + stcb->sctp_ep->partial_delivery_point); + } else { + pd_point = inp->partial_delivery_point; + } + if ((control->end_added == 1) || (length >= pd_point)) { + struct socket *so; + struct mbuf *m; + char *buffer; + struct sctp_rcvinfo rcv; + union sctp_sockstore addr; + int flags; + + if ((buffer = malloc(length)) == NULL) { + return; + } + if (inp_read_lock_held == 0) { + SCTP_INP_READ_LOCK(inp); + } + so = stcb->sctp_socket; + for (m = control->data; m; m = SCTP_BUF_NEXT(m)) { + sctp_sbfree(control, control->stcb, &so->so_rcv, m); + } + m_copydata(control->data, 0, length, buffer); + memset(&rcv, 0, sizeof(struct sctp_rcvinfo)); + rcv.rcv_sid = control->sinfo_stream; + rcv.rcv_ssn = (uint16_t)control->mid; + rcv.rcv_flags = control->sinfo_flags; + rcv.rcv_ppid = control->sinfo_ppid; + rcv.rcv_tsn = control->sinfo_tsn; + rcv.rcv_cumtsn = control->sinfo_cumtsn; + rcv.rcv_context = control->sinfo_context; + rcv.rcv_assoc_id = control->sinfo_assoc_id; + memset(&addr, 0, sizeof(union sctp_sockstore)); + switch (control->whoFrom->ro._l_addr.sa.sa_family) { +#ifdef INET + case AF_INET: + addr.sin = control->whoFrom->ro._l_addr.sin; + break; +#endif +#ifdef INET6 + case AF_INET6: + addr.sin6 = control->whoFrom->ro._l_addr.sin6; + break; +#endif + case AF_CONN: + addr.sconn = control->whoFrom->ro._l_addr.sconn; + break; + default: + addr.sa = control->whoFrom->ro._l_addr.sa; + break; + } + flags = 0; + if (control->end_added == 1) { + flags |= MSG_EOR; + } + if (control->spec_flags & M_NOTIFICATION) { + flags |= MSG_NOTIFICATION; + } + sctp_m_freem(control->data); + control->data = NULL; + control->tail_mbuf = NULL; + control->length = 0; + if (control->end_added) { + TAILQ_REMOVE(&stcb->sctp_ep->read_queue, control, next); + control->on_read_q = 0; + sctp_free_remote_addr(control->whoFrom); + control->whoFrom = NULL; + sctp_free_a_readq(stcb, control); + } + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + if (inp_read_lock_held == 0) { + SCTP_INP_READ_UNLOCK(inp); + } + inp->recv_callback(so, addr, buffer, length, rcv, flags, inp->ulp_info); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } +} +#endif + +void +sctp_add_to_readq(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_queued_to_read *control, + struct sockbuf *sb, + int end, + int inp_read_lock_held, + int so_locked) +{ + /* + * Here we must place the control on the end of the socket read + * queue AND increment sb_cc so that select will work properly on + * read. + */ + struct mbuf *m, *prev = NULL; + + if (inp == NULL) { + /* Gak, TSNH!! */ +#ifdef INVARIANTS + panic("Gak, inp NULL on add_to_readq"); +#endif + return; + } +#if defined(__APPLE__) && !defined(__Userspace__) + if (so_locked) { + sctp_lock_assert(SCTP_INP_SO(inp)); + } else { + sctp_unlock_assert(SCTP_INP_SO(inp)); + } +#endif + if (inp_read_lock_held == SCTP_READ_LOCK_NOT_HELD) { + SCTP_INP_READ_LOCK(inp); + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_CANT_READ) { + if (!control->on_strm_q) { + sctp_free_remote_addr(control->whoFrom); + if (control->data) { + sctp_m_freem(control->data); + control->data = NULL; + } + sctp_free_a_readq(stcb, control); + } + if (inp_read_lock_held == SCTP_READ_LOCK_NOT_HELD) { + SCTP_INP_READ_UNLOCK(inp); + } + return; + } + if ((control->spec_flags & M_NOTIFICATION) == 0) { + atomic_add_int(&inp->total_recvs, 1); + if (!control->do_not_ref_stcb) { + atomic_add_int(&stcb->total_recvs, 1); + } + } + m = control->data; + control->held_length = 0; + control->length = 0; + while (m != NULL) { + if (SCTP_BUF_LEN(m) == 0) { + /* Skip mbufs with NO length */ + if (prev == NULL) { + /* First one */ + control->data = sctp_m_free(m); + m = control->data; + } else { + SCTP_BUF_NEXT(prev) = sctp_m_free(m); + m = SCTP_BUF_NEXT(prev); + } + if (m == NULL) { + control->tail_mbuf = prev; + } + continue; + } + prev = m; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(sb, control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBALLOC, SCTP_BUF_LEN(m)); + } + sctp_sballoc(stcb, sb, m); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(sb, control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBRESULT, 0); + } + atomic_add_int(&control->length, SCTP_BUF_LEN(m)); + m = SCTP_BUF_NEXT(m); + } + if (prev != NULL) { + control->tail_mbuf = prev; + } else { + /* Everything got collapsed out?? */ + if (!control->on_strm_q) { + sctp_free_remote_addr(control->whoFrom); + sctp_free_a_readq(stcb, control); + } + if (inp_read_lock_held == 0) + SCTP_INP_READ_UNLOCK(inp); + return; + } + if (end) { + control->end_added = 1; + } + TAILQ_INSERT_TAIL(&inp->read_queue, control, next); + control->on_read_q = 1; +#if defined(__Userspace__) + sctp_invoke_recv_callback(inp, stcb, control, SCTP_READ_LOCK_HELD); +#endif + if ((inp != NULL) && (inp->sctp_socket != NULL)) { + sctp_wakeup_the_read_socket(inp, stcb, so_locked); + } + if (inp_read_lock_held == SCTP_READ_LOCK_NOT_HELD) { + SCTP_INP_READ_UNLOCK(inp); + } +} + +/*************HOLD THIS COMMENT FOR PATCH FILE OF + *************ALTERNATE ROUTING CODE + */ + +/*************HOLD THIS COMMENT FOR END OF PATCH FILE OF + *************ALTERNATE ROUTING CODE + */ + +struct mbuf * +sctp_generate_cause(uint16_t code, char *info) +{ + struct mbuf *m; + struct sctp_gen_error_cause *cause; + size_t info_len; + uint16_t len; + + if ((code == 0) || (info == NULL)) { + return (NULL); + } + info_len = strlen(info); + if (info_len > (SCTP_MAX_CAUSE_LENGTH - sizeof(struct sctp_paramhdr))) { + return (NULL); + } + len = (uint16_t)(sizeof(struct sctp_paramhdr) + info_len); + m = sctp_get_mbuf_for_msg(len, 0, M_NOWAIT, 1, MT_DATA); + if (m != NULL) { + SCTP_BUF_LEN(m) = len; + cause = mtod(m, struct sctp_gen_error_cause *); + cause->code = htons(code); + cause->length = htons(len); + memcpy(cause->info, info, info_len); + } + return (m); +} + +struct mbuf * +sctp_generate_no_user_data_cause(uint32_t tsn) +{ + struct mbuf *m; + struct sctp_error_no_user_data *no_user_data_cause; + uint16_t len; + + len = (uint16_t)sizeof(struct sctp_error_no_user_data); + m = sctp_get_mbuf_for_msg(len, 0, M_NOWAIT, 1, MT_DATA); + if (m != NULL) { + SCTP_BUF_LEN(m) = len; + no_user_data_cause = mtod(m, struct sctp_error_no_user_data *); + no_user_data_cause->cause.code = htons(SCTP_CAUSE_NO_USER_DATA); + no_user_data_cause->cause.length = htons(len); + no_user_data_cause->tsn = htonl(tsn); + } + return (m); +} + +void +sctp_free_bufspace(struct sctp_tcb *stcb, struct sctp_association *asoc, + struct sctp_tmit_chunk *tp1, int chk_cnt) +{ + if (tp1->data == NULL) { + return; + } + atomic_subtract_int(&asoc->chunks_on_out_queue, chk_cnt); +#ifdef SCTP_MBCNT_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBCNT_LOGGING_ENABLE) { + sctp_log_mbcnt(SCTP_LOG_MBCNT_DECREASE, + asoc->total_output_queue_size, + tp1->book_size, + 0, + tp1->mbcnt); + } +#endif + if (asoc->total_output_queue_size >= tp1->book_size) { + atomic_subtract_int(&asoc->total_output_queue_size, tp1->book_size); + } else { + asoc->total_output_queue_size = 0; + } + if ((stcb->sctp_socket != NULL) && + (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) || + ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)))) { + if (stcb->sctp_socket->so_snd.sb_cc >= tp1->book_size) { + atomic_subtract_int(&((stcb)->sctp_socket->so_snd.sb_cc), tp1->book_size); + } else { + stcb->sctp_socket->so_snd.sb_cc = 0; + } + } +} + +int +sctp_release_pr_sctp_chunk(struct sctp_tcb *stcb, struct sctp_tmit_chunk *tp1, + uint8_t sent, int so_locked) +{ + struct sctp_stream_out *strq; + struct sctp_tmit_chunk *chk = NULL, *tp2; + struct sctp_stream_queue_pending *sp; + uint32_t mid; + uint16_t sid; + uint8_t foundeom = 0; + int ret_sz = 0; + int notdone; + int do_wakeup_routine = 0; + +#if defined(__APPLE__) && !defined(__Userspace__) + if (so_locked) { + sctp_lock_assert(SCTP_INP_SO(stcb->sctp_ep)); + } else { + sctp_unlock_assert(SCTP_INP_SO(stcb->sctp_ep)); + } +#endif + SCTP_TCB_LOCK_ASSERT(stcb); + + sid = tp1->rec.data.sid; + mid = tp1->rec.data.mid; + if (sent || ((tp1->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) == 0)) { + stcb->asoc.abandoned_sent[0]++; + stcb->asoc.abandoned_sent[PR_SCTP_POLICY(tp1->flags)]++; + stcb->asoc.strmout[sid].abandoned_sent[0]++; +#if defined(SCTP_DETAILED_STR_STATS) + stcb->asoc.strmout[sid].abandoned_sent[PR_SCTP_POLICY(tp1->flags)]++; +#endif + } else { + stcb->asoc.abandoned_unsent[0]++; + stcb->asoc.abandoned_unsent[PR_SCTP_POLICY(tp1->flags)]++; + stcb->asoc.strmout[sid].abandoned_unsent[0]++; +#if defined(SCTP_DETAILED_STR_STATS) + stcb->asoc.strmout[sid].abandoned_unsent[PR_SCTP_POLICY(tp1->flags)]++; +#endif + } + do { + ret_sz += tp1->book_size; + if (tp1->data != NULL) { + if (tp1->sent < SCTP_DATAGRAM_RESEND) { + sctp_flight_size_decrease(tp1); + sctp_total_flight_decrease(stcb, tp1); + } + sctp_free_bufspace(stcb, &stcb->asoc, tp1, 1); + stcb->asoc.peers_rwnd += tp1->send_size; + stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh); + if (sent) { + sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb, 0, tp1, so_locked); + } else { + sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb, 0, tp1, so_locked); + } + if (tp1->data) { + sctp_m_freem(tp1->data); + tp1->data = NULL; + } + do_wakeup_routine = 1; + if (PR_SCTP_BUF_ENABLED(tp1->flags)) { + stcb->asoc.sent_queue_cnt_removeable--; + } + } + tp1->sent = SCTP_FORWARD_TSN_SKIP; + if ((tp1->rec.data.rcv_flags & SCTP_DATA_NOT_FRAG) == + SCTP_DATA_NOT_FRAG) { + /* not frag'ed we ae done */ + notdone = 0; + foundeom = 1; + } else if (tp1->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) { + /* end of frag, we are done */ + notdone = 0; + foundeom = 1; + } else { + /* + * Its a begin or middle piece, we must mark all of + * it + */ + notdone = 1; + tp1 = TAILQ_NEXT(tp1, sctp_next); + } + } while (tp1 && notdone); + if (foundeom == 0) { + /* + * The multi-part message was scattered across the send and + * sent queue. + */ + TAILQ_FOREACH_SAFE(tp1, &stcb->asoc.send_queue, sctp_next, tp2) { + if ((tp1->rec.data.sid != sid) || + (!SCTP_MID_EQ(stcb->asoc.idata_supported, tp1->rec.data.mid, mid))) { + break; + } + /* save to chk in case we have some on stream out + * queue. If so and we have an un-transmitted one + * we don't have to fudge the TSN. + */ + chk = tp1; + ret_sz += tp1->book_size; + sctp_free_bufspace(stcb, &stcb->asoc, tp1, 1); + if (sent) { + sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb, 0, tp1, so_locked); + } else { + sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb, 0, tp1, so_locked); + } + if (tp1->data) { + sctp_m_freem(tp1->data); + tp1->data = NULL; + } + /* No flight involved here book the size to 0 */ + tp1->book_size = 0; + if (tp1->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) { + foundeom = 1; + } + do_wakeup_routine = 1; + tp1->sent = SCTP_FORWARD_TSN_SKIP; + TAILQ_REMOVE(&stcb->asoc.send_queue, tp1, sctp_next); + /* on to the sent queue so we can wait for it to be passed by. */ + TAILQ_INSERT_TAIL(&stcb->asoc.sent_queue, tp1, + sctp_next); + stcb->asoc.send_queue_cnt--; + stcb->asoc.sent_queue_cnt++; + } + } + if (foundeom == 0) { + /* + * Still no eom found. That means there + * is stuff left on the stream out queue.. yuck. + */ + strq = &stcb->asoc.strmout[sid]; + sp = TAILQ_FIRST(&strq->outqueue); + if (sp != NULL) { + sp->discard_rest = 1; + /* + * We may need to put a chunk on the + * queue that holds the TSN that + * would have been sent with the LAST + * bit. + */ + if (chk == NULL) { + /* Yep, we have to */ + sctp_alloc_a_chunk(stcb, chk); + if (chk == NULL) { + /* we are hosed. All we can + * do is nothing.. which will + * cause an abort if the peer is + * paying attention. + */ + goto oh_well; + } + memset(chk, 0, sizeof(*chk)); + chk->rec.data.rcv_flags = 0; + chk->sent = SCTP_FORWARD_TSN_SKIP; + chk->asoc = &stcb->asoc; + if (stcb->asoc.idata_supported == 0) { + if (sp->sinfo_flags & SCTP_UNORDERED) { + chk->rec.data.mid = 0; + } else { + chk->rec.data.mid = strq->next_mid_ordered; + } + } else { + if (sp->sinfo_flags & SCTP_UNORDERED) { + chk->rec.data.mid = strq->next_mid_unordered; + } else { + chk->rec.data.mid = strq->next_mid_ordered; + } + } + chk->rec.data.sid = sp->sid; + chk->rec.data.ppid = sp->ppid; + chk->rec.data.context = sp->context; + chk->flags = sp->act_flags; + chk->whoTo = NULL; +#if defined(__FreeBSD__) && !defined(__Userspace__) + chk->rec.data.tsn = atomic_fetchadd_int(&stcb->asoc.sending_seq, 1); +#else + chk->rec.data.tsn = stcb->asoc.sending_seq++; +#endif + strq->chunks_on_queues++; + TAILQ_INSERT_TAIL(&stcb->asoc.sent_queue, chk, sctp_next); + stcb->asoc.sent_queue_cnt++; + stcb->asoc.pr_sctp_cnt++; + } + chk->rec.data.rcv_flags |= SCTP_DATA_LAST_FRAG; + if (sp->sinfo_flags & SCTP_UNORDERED) { + chk->rec.data.rcv_flags |= SCTP_DATA_UNORDERED; + } + if (stcb->asoc.idata_supported == 0) { + if ((sp->sinfo_flags & SCTP_UNORDERED) == 0) { + strq->next_mid_ordered++; + } + } else { + if (sp->sinfo_flags & SCTP_UNORDERED) { + strq->next_mid_unordered++; + } else { + strq->next_mid_ordered++; + } + } + oh_well: + if (sp->data) { + /* Pull any data to free up the SB and + * allow sender to "add more" while we + * will throw away :-) + */ + sctp_free_spbufspace(stcb, &stcb->asoc, sp); + ret_sz += sp->length; + do_wakeup_routine = 1; + sp->some_taken = 1; + sctp_m_freem(sp->data); + sp->data = NULL; + sp->tail_mbuf = NULL; + sp->length = 0; + } + } + } + if (do_wakeup_routine) { +#if defined(__APPLE__) && !defined(__Userspace__) + struct socket *so; + + so = SCTP_INP_SO(stcb->sctp_ep); + if (!so_locked) { + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); + if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { + /* assoc was freed while we were unlocked */ + SCTP_SOCKET_UNLOCK(so, 1); + return (ret_sz); + } + } +#endif + sctp_sowwakeup(stcb->sctp_ep, stcb->sctp_socket); +#if defined(__APPLE__) && !defined(__Userspace__) + if (!so_locked) { + SCTP_SOCKET_UNLOCK(so, 1); + } +#endif + } + return (ret_sz); +} + +/* + * checks to see if the given address, sa, is one that is currently known by + * the kernel note: can't distinguish the same address on multiple interfaces + * and doesn't handle multiple addresses with different zone/scope id's note: + * ifa_ifwithaddr() compares the entire sockaddr struct + */ +struct sctp_ifa * +sctp_find_ifa_in_ep(struct sctp_inpcb *inp, struct sockaddr *addr, + int holds_lock) +{ + struct sctp_laddr *laddr; + + if (holds_lock == 0) { + SCTP_INP_RLOCK(inp); + } + + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa == NULL) + continue; + if (addr->sa_family != laddr->ifa->address.sa.sa_family) + continue; +#ifdef INET + if (addr->sa_family == AF_INET) { + if (((struct sockaddr_in *)addr)->sin_addr.s_addr == + laddr->ifa->address.sin.sin_addr.s_addr) { + /* found him. */ + break; + } + } +#endif +#ifdef INET6 + if (addr->sa_family == AF_INET6) { + if (SCTP6_ARE_ADDR_EQUAL((struct sockaddr_in6 *)addr, + &laddr->ifa->address.sin6)) { + /* found him. */ + break; + } + } +#endif +#if defined(__Userspace__) + if (addr->sa_family == AF_CONN) { + if (((struct sockaddr_conn *)addr)->sconn_addr == laddr->ifa->address.sconn.sconn_addr) { + /* found him. */ + break; + } + } +#endif + } + if (holds_lock == 0) { + SCTP_INP_RUNLOCK(inp); + } + if (laddr != NULL) { + return (laddr->ifa); + } else { + return (NULL); + } +} + +uint32_t +sctp_get_ifa_hash_val(struct sockaddr *addr) +{ + switch (addr->sa_family) { +#ifdef INET + case AF_INET: + { + struct sockaddr_in *sin; + + sin = (struct sockaddr_in *)addr; + return (sin->sin_addr.s_addr ^ (sin->sin_addr.s_addr >> 16)); + } +#endif +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6; + uint32_t hash_of_addr; + + sin6 = (struct sockaddr_in6 *)addr; +#if !defined(_WIN32) && !(defined(__FreeBSD__) && defined(__Userspace__)) && !defined(__APPLE__) + hash_of_addr = (sin6->sin6_addr.s6_addr32[0] + + sin6->sin6_addr.s6_addr32[1] + + sin6->sin6_addr.s6_addr32[2] + + sin6->sin6_addr.s6_addr32[3]); +#else + hash_of_addr = (((uint32_t *)&sin6->sin6_addr)[0] + + ((uint32_t *)&sin6->sin6_addr)[1] + + ((uint32_t *)&sin6->sin6_addr)[2] + + ((uint32_t *)&sin6->sin6_addr)[3]); +#endif + hash_of_addr = (hash_of_addr ^ (hash_of_addr >> 16)); + return (hash_of_addr); + } +#endif +#if defined(__Userspace__) + case AF_CONN: + { + struct sockaddr_conn *sconn; + uintptr_t temp; + + sconn = (struct sockaddr_conn *)addr; + temp = (uintptr_t)sconn->sconn_addr; + return ((uint32_t)(temp ^ (temp >> 16))); + } +#endif + default: + break; + } + return (0); +} + +struct sctp_ifa * +sctp_find_ifa_by_addr(struct sockaddr *addr, uint32_t vrf_id, int holds_lock) +{ + struct sctp_ifa *sctp_ifap; + struct sctp_vrf *vrf; + struct sctp_ifalist *hash_head; + uint32_t hash_of_addr; + + if (holds_lock == 0) { + SCTP_IPI_ADDR_RLOCK(); + } else { + SCTP_IPI_ADDR_LOCK_ASSERT(); + } + + vrf = sctp_find_vrf(vrf_id); + if (vrf == NULL) { + if (holds_lock == 0) + SCTP_IPI_ADDR_RUNLOCK(); + return (NULL); + } + + hash_of_addr = sctp_get_ifa_hash_val(addr); + + hash_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)]; + if (hash_head == NULL) { + SCTP_PRINTF("hash_of_addr:%x mask:%x table:%x - ", + hash_of_addr, (uint32_t)vrf->vrf_addr_hashmark, + (uint32_t)(hash_of_addr & vrf->vrf_addr_hashmark)); + sctp_print_address(addr); + SCTP_PRINTF("No such bucket for address\n"); + if (holds_lock == 0) + SCTP_IPI_ADDR_RUNLOCK(); + + return (NULL); + } + LIST_FOREACH(sctp_ifap, hash_head, next_bucket) { + if (addr->sa_family != sctp_ifap->address.sa.sa_family) + continue; +#ifdef INET + if (addr->sa_family == AF_INET) { + if (((struct sockaddr_in *)addr)->sin_addr.s_addr == + sctp_ifap->address.sin.sin_addr.s_addr) { + /* found him. */ + break; + } + } +#endif +#ifdef INET6 + if (addr->sa_family == AF_INET6) { + if (SCTP6_ARE_ADDR_EQUAL((struct sockaddr_in6 *)addr, + &sctp_ifap->address.sin6)) { + /* found him. */ + break; + } + } +#endif +#if defined(__Userspace__) + if (addr->sa_family == AF_CONN) { + if (((struct sockaddr_conn *)addr)->sconn_addr == sctp_ifap->address.sconn.sconn_addr) { + /* found him. */ + break; + } + } +#endif + } + if (holds_lock == 0) + SCTP_IPI_ADDR_RUNLOCK(); + return (sctp_ifap); +} + +static void +sctp_user_rcvd(struct sctp_tcb *stcb, uint32_t *freed_so_far, int hold_rlock, + uint32_t rwnd_req) +{ + /* User pulled some data, do we need a rwnd update? */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif + int r_unlocked = 0; + uint32_t dif, rwnd; + struct socket *so = NULL; + + if (stcb == NULL) + return; + + atomic_add_int(&stcb->asoc.refcnt, 1); + + if ((SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_ACK_SENT) || + (stcb->asoc.state & (SCTP_STATE_ABOUT_TO_BE_FREED | SCTP_STATE_SHUTDOWN_RECEIVED))) { + /* Pre-check If we are freeing no update */ + goto no_lock; + } + SCTP_INP_INCR_REF(stcb->sctp_ep); + if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { + goto out; + } + so = stcb->sctp_socket; + if (so == NULL) { + goto out; + } + atomic_add_int(&stcb->freed_by_sorcv_sincelast, *freed_so_far); + /* Have you have freed enough to look */ + *freed_so_far = 0; + /* Yep, its worth a look and the lock overhead */ + + /* Figure out what the rwnd would be */ + rwnd = sctp_calc_rwnd(stcb, &stcb->asoc); + if (rwnd >= stcb->asoc.my_last_reported_rwnd) { + dif = rwnd - stcb->asoc.my_last_reported_rwnd; + } else { + dif = 0; + } + if (dif >= rwnd_req) { + if (hold_rlock) { + SCTP_INP_READ_UNLOCK(stcb->sctp_ep); + r_unlocked = 1; + } + if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + /* + * One last check before we allow the guy possibly + * to get in. There is a race, where the guy has not + * reached the gate. In that case + */ + goto out; + } + SCTP_TCB_LOCK(stcb); + if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + /* No reports here */ + SCTP_TCB_UNLOCK(stcb); + goto out; + } + SCTP_STAT_INCR(sctps_wu_sacks_sent); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + sctp_send_sack(stcb, SCTP_SO_LOCKED); + + sctp_chunk_output(stcb->sctp_ep, stcb, + SCTP_OUTPUT_FROM_USR_RCVD, SCTP_SO_LOCKED); + /* make sure no timer is running */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, + SCTP_FROM_SCTPUTIL + SCTP_LOC_6); + SCTP_TCB_UNLOCK(stcb); + } else { + /* Update how much we have pending */ + stcb->freed_by_sorcv_sincelast = dif; + } + out: + if (so && r_unlocked && hold_rlock) { + SCTP_INP_READ_LOCK(stcb->sctp_ep); + } + + SCTP_INP_DECR_REF(stcb->sctp_ep); + no_lock: + atomic_subtract_int(&stcb->asoc.refcnt, 1); + return; +} + +int +sctp_sorecvmsg(struct socket *so, + struct uio *uio, + struct mbuf **mp, + struct sockaddr *from, + int fromlen, + int *msg_flags, + struct sctp_sndrcvinfo *sinfo, + int filling_sinfo) +{ + /* + * MSG flags we will look at MSG_DONTWAIT - non-blocking IO. + * MSG_PEEK - Look don't touch :-D (only valid with OUT mbuf copy + * mp=NULL thus uio is the copy method to userland) MSG_WAITALL - ?? + * On the way out we may send out any combination of: + * MSG_NOTIFICATION MSG_EOR + * + */ + struct sctp_inpcb *inp = NULL; + ssize_t my_len = 0; + ssize_t cp_len = 0; + int error = 0; + struct sctp_queued_to_read *control = NULL, *ctl = NULL, *nxt = NULL; + struct mbuf *m = NULL; + struct sctp_tcb *stcb = NULL; + int wakeup_read_socket = 0; + int freecnt_applied = 0; + int out_flags = 0, in_flags = 0; + int block_allowed = 1; + uint32_t freed_so_far = 0; + ssize_t copied_so_far = 0; + int in_eeor_mode = 0; + int no_rcv_needed = 0; + uint32_t rwnd_req = 0; + int hold_sblock = 0; + int hold_rlock = 0; + ssize_t slen = 0; + uint32_t held_length = 0; +#if defined(__FreeBSD__) && !defined(__Userspace__) + int sockbuf_lock = 0; +#endif + + if (uio == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + return (EINVAL); + } + + if (msg_flags) { + in_flags = *msg_flags; + if (in_flags & MSG_PEEK) + SCTP_STAT_INCR(sctps_read_peeks); + } else { + in_flags = 0; + } +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + slen = uio->uio_resid; +#else + slen = uio_resid(uio); +#endif +#else + slen = uio->uio_resid; +#endif + + /* Pull in and set up our int flags */ + if (in_flags & MSG_OOB) { + /* Out of band's NOT supported */ + return (EOPNOTSUPP); + } + if ((in_flags & MSG_PEEK) && (mp != NULL)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + return (EINVAL); + } + if ((in_flags & (MSG_DONTWAIT +#if defined(__FreeBSD__) && !defined(__Userspace__) + | MSG_NBIO +#endif + )) || + SCTP_SO_IS_NBIO(so)) { + block_allowed = 0; + } + /* setup the endpoint */ + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, EFAULT); + return (EFAULT); + } + rwnd_req = (SCTP_SB_LIMIT_RCV(so) >> SCTP_RWND_HIWAT_SHIFT); + /* Must be at least a MTU's worth */ + if (rwnd_req < SCTP_MIN_RWND) + rwnd_req = SCTP_MIN_RWND; + in_eeor_mode = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RECV_RWND_LOGGING_ENABLE) { +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + sctp_misc_ints(SCTP_SORECV_ENTER, + rwnd_req, in_eeor_mode, SCTP_SBAVAIL(&so->so_rcv), uio->uio_resid); +#else + sctp_misc_ints(SCTP_SORECV_ENTER, + rwnd_req, in_eeor_mode, SCTP_SBAVAIL(&so->so_rcv), uio_resid(uio)); +#endif +#else + sctp_misc_ints(SCTP_SORECV_ENTER, + rwnd_req, in_eeor_mode, SCTP_SBAVAIL(&so->so_rcv), (uint32_t)uio->uio_resid); +#endif + } +#if defined(__Userspace__) + SOCKBUF_LOCK(&so->so_rcv); + hold_sblock = 1; +#endif + if (SCTP_BASE_SYSCTL(sctp_logging_level) &SCTP_RECV_RWND_LOGGING_ENABLE) { +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + sctp_misc_ints(SCTP_SORECV_ENTERPL, + rwnd_req, block_allowed, SCTP_SBAVAIL(&so->so_rcv), uio->uio_resid); +#else + sctp_misc_ints(SCTP_SORECV_ENTERPL, + rwnd_req, block_allowed, SCTP_SBAVAIL(&so->so_rcv), uio_resid(uio)); +#endif +#else + sctp_misc_ints(SCTP_SORECV_ENTERPL, + rwnd_req, block_allowed, SCTP_SBAVAIL(&so->so_rcv), (uint32_t)uio->uio_resid); +#endif + } + +#if defined(__APPLE__) && !defined(__Userspace__) + error = sblock(&so->so_rcv, SBLOCKWAIT(in_flags)); +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + error = SOCK_IO_RECV_LOCK(so, SBLOCKWAIT(in_flags)); +#endif + if (error) { + goto release_unlocked; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + sockbuf_lock = 1; +#endif + restart: +#if defined(__Userspace__) + if (hold_sblock == 0) { + SOCKBUF_LOCK(&so->so_rcv); + hold_sblock = 1; + } +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + sbunlock(&so->so_rcv, 1); +#endif + + restart_nosblocks: + if (hold_sblock == 0) { + SOCKBUF_LOCK(&so->so_rcv); + hold_sblock = 1; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { + goto out; + } +#if (defined(__FreeBSD__) || defined(_WIN32)) && !defined(__Userspace__) + if ((so->so_rcv.sb_state & SBS_CANTRCVMORE) && SCTP_SBAVAIL(&so->so_rcv) == 0) { +#else + if ((so->so_state & SS_CANTRCVMORE) && SCTP_SBAVAIL(&so->so_rcv) == 0) { +#endif + if (so->so_error) { + error = so->so_error; + if ((in_flags & MSG_PEEK) == 0) + so->so_error = 0; + goto out; + } else { + if (SCTP_SBAVAIL(&so->so_rcv) == 0) { + /* indicate EOF */ + error = 0; + goto out; + } + } + } + if (SCTP_SBAVAIL(&so->so_rcv) <= held_length) { + if (so->so_error) { + error = so->so_error; + if ((in_flags & MSG_PEEK) == 0) { + so->so_error = 0; + } + goto out; + } + if ((SCTP_SBAVAIL(&so->so_rcv) == 0) && + ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL))) { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0) { + /* For active open side clear flags for re-use + * passive open is blocked by connect. + */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED) { + /* You were aborted, passive side always hits here */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, ECONNRESET); + error = ECONNRESET; + } + so->so_state &= ~(SS_ISCONNECTING | + SS_ISDISCONNECTING | + SS_ISCONFIRMING | + SS_ISCONNECTED); + if (error == 0) { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) == 0) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, ENOTCONN); + error = ENOTCONN; + } + } + goto out; + } + } + if (block_allowed) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + error = sbwait(so, SO_RCV); +#else + error = sbwait(&so->so_rcv); +#endif + if (error) { + goto out; + } + held_length = 0; + goto restart_nosblocks; + } else { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EWOULDBLOCK); + error = EWOULDBLOCK; + goto out; + } + } + if (hold_sblock == 1) { + SOCKBUF_UNLOCK(&so->so_rcv); + hold_sblock = 0; + } +#if defined(__APPLE__) && !defined(__Userspace__) + error = sblock(&so->so_rcv, SBLOCKWAIT(in_flags)); +#endif + /* we possibly have data we can read */ + /*sa_ignore FREED_MEMORY*/ + control = TAILQ_FIRST(&inp->read_queue); + if (control == NULL) { + /* This could be happening since + * the appender did the increment but as not + * yet did the tailq insert onto the read_queue + */ + if (hold_rlock == 0) { + SCTP_INP_READ_LOCK(inp); + } + control = TAILQ_FIRST(&inp->read_queue); + if ((control == NULL) && (SCTP_SBAVAIL(&so->so_rcv) > 0)) { +#ifdef INVARIANTS + panic("Huh, its non zero and nothing on control?"); +#endif + SCTP_SB_CLEAR(so->so_rcv); + } + SCTP_INP_READ_UNLOCK(inp); + hold_rlock = 0; + goto restart; + } + + if ((control->length == 0) && + (control->do_not_ref_stcb)) { + /* Clean up code for freeing assoc that left behind a pdapi.. + * maybe a peer in EEOR that just closed after sending and + * never indicated a EOR. + */ + if (hold_rlock == 0) { + hold_rlock = 1; + SCTP_INP_READ_LOCK(inp); + } + control->held_length = 0; + if (control->data) { + /* Hmm there is data here .. fix */ + struct mbuf *m_tmp; + int cnt = 0; + m_tmp = control->data; + while (m_tmp) { + cnt += SCTP_BUF_LEN(m_tmp); + if (SCTP_BUF_NEXT(m_tmp) == NULL) { + control->tail_mbuf = m_tmp; + control->end_added = 1; + } + m_tmp = SCTP_BUF_NEXT(m_tmp); + } + control->length = cnt; + } else { + /* remove it */ + TAILQ_REMOVE(&inp->read_queue, control, next); + /* Add back any hidden data */ + sctp_free_remote_addr(control->whoFrom); + sctp_free_a_readq(stcb, control); + } + if (hold_rlock) { + hold_rlock = 0; + SCTP_INP_READ_UNLOCK(inp); + } + goto restart; + } + if ((control->length == 0) && + (control->end_added == 1)) { + /* Do we also need to check for (control->pdapi_aborted == 1)? */ + if (hold_rlock == 0) { + hold_rlock = 1; + SCTP_INP_READ_LOCK(inp); + } + TAILQ_REMOVE(&inp->read_queue, control, next); + if (control->data) { +#ifdef INVARIANTS + panic("control->data not null but control->length == 0"); +#else + SCTP_PRINTF("Strange, data left in the control buffer. Cleaning up.\n"); + sctp_m_freem(control->data); + control->data = NULL; +#endif + } + if (control->aux_data) { + sctp_m_free (control->aux_data); + control->aux_data = NULL; + } +#ifdef INVARIANTS + if (control->on_strm_q) { + panic("About to free ctl:%p so:%p and its in %d", + control, so, control->on_strm_q); + } +#endif + sctp_free_remote_addr(control->whoFrom); + sctp_free_a_readq(stcb, control); + if (hold_rlock) { + hold_rlock = 0; + SCTP_INP_READ_UNLOCK(inp); + } + goto restart; + } + if (control->length == 0) { + if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE)) && + (filling_sinfo)) { + /* find a more suitable one then this */ + ctl = TAILQ_NEXT(control, next); + while (ctl) { + if ((ctl->stcb != control->stcb) && (ctl->length) && + (ctl->some_taken || + (ctl->spec_flags & M_NOTIFICATION) || + ((ctl->do_not_ref_stcb == 0) && + (ctl->stcb->asoc.strmin[ctl->sinfo_stream].delivery_started == 0))) + ) { + /*- + * If we have a different TCB next, and there is data + * present. If we have already taken some (pdapi), OR we can + * ref the tcb and no delivery as started on this stream, we + * take it. Note we allow a notification on a different + * assoc to be delivered.. + */ + control = ctl; + goto found_one; + } else if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS)) && + (ctl->length) && + ((ctl->some_taken) || + ((ctl->do_not_ref_stcb == 0) && + ((ctl->spec_flags & M_NOTIFICATION) == 0) && + (ctl->stcb->asoc.strmin[ctl->sinfo_stream].delivery_started == 0)))) { + /*- + * If we have the same tcb, and there is data present, and we + * have the strm interleave feature present. Then if we have + * taken some (pdapi) or we can refer to tht tcb AND we have + * not started a delivery for this stream, we can take it. + * Note we do NOT allow a notification on the same assoc to + * be delivered. + */ + control = ctl; + goto found_one; + } + ctl = TAILQ_NEXT(ctl, next); + } + } + /* + * if we reach here, not suitable replacement is available + * <or> fragment interleave is NOT on. So stuff the sb_cc + * into the our held count, and its time to sleep again. + */ + held_length = SCTP_SBAVAIL(&so->so_rcv); + control->held_length = SCTP_SBAVAIL(&so->so_rcv); + goto restart; + } + /* Clear the held length since there is something to read */ + control->held_length = 0; + found_one: + /* + * If we reach here, control has a some data for us to read off. + * Note that stcb COULD be NULL. + */ + if (hold_rlock == 0) { + hold_rlock = 1; + SCTP_INP_READ_LOCK(inp); + } + control->some_taken++; + stcb = control->stcb; + if (stcb) { + if ((control->do_not_ref_stcb == 0) && + (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED)) { + if (freecnt_applied == 0) + stcb = NULL; + } else if (control->do_not_ref_stcb == 0) { + /* you can't free it on me please */ + /* + * The lock on the socket buffer protects us so the + * free code will stop. But since we used the socketbuf + * lock and the sender uses the tcb_lock to increment, + * we need to use the atomic add to the refcnt + */ + if (freecnt_applied) { +#ifdef INVARIANTS + panic("refcnt already incremented"); +#else + SCTP_PRINTF("refcnt already incremented?\n"); +#endif + } else { + atomic_add_int(&stcb->asoc.refcnt, 1); + freecnt_applied = 1; + } + /* + * Setup to remember how much we have not yet told + * the peer our rwnd has opened up. Note we grab + * the value from the tcb from last time. + * Note too that sack sending clears this when a sack + * is sent, which is fine. Once we hit the rwnd_req, + * we then will go to the sctp_user_rcvd() that will + * not lock until it KNOWs it MUST send a WUP-SACK. + */ + freed_so_far = (uint32_t)stcb->freed_by_sorcv_sincelast; + stcb->freed_by_sorcv_sincelast = 0; + } + } + if (stcb && + ((control->spec_flags & M_NOTIFICATION) == 0) && + control->do_not_ref_stcb == 0) { + stcb->asoc.strmin[control->sinfo_stream].delivery_started = 1; + } + + /* First lets get off the sinfo and sockaddr info */ + if ((sinfo != NULL) && (filling_sinfo != 0)) { + sinfo->sinfo_stream = control->sinfo_stream; + sinfo->sinfo_ssn = (uint16_t)control->mid; + sinfo->sinfo_flags = control->sinfo_flags; + sinfo->sinfo_ppid = control->sinfo_ppid; + sinfo->sinfo_context =control->sinfo_context; + sinfo->sinfo_timetolive = control->sinfo_timetolive; + sinfo->sinfo_tsn = control->sinfo_tsn; + sinfo->sinfo_cumtsn = control->sinfo_cumtsn; + sinfo->sinfo_assoc_id = control->sinfo_assoc_id; + nxt = TAILQ_NEXT(control, next); + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO) || + sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVNXTINFO)) { + struct sctp_extrcvinfo *s_extra; + s_extra = (struct sctp_extrcvinfo *)sinfo; + if ((nxt) && + (nxt->length)) { + s_extra->serinfo_next_flags = SCTP_NEXT_MSG_AVAIL; + if (nxt->sinfo_flags & SCTP_UNORDERED) { + s_extra->serinfo_next_flags |= SCTP_NEXT_MSG_IS_UNORDERED; + } + if (nxt->spec_flags & M_NOTIFICATION) { + s_extra->serinfo_next_flags |= SCTP_NEXT_MSG_IS_NOTIFICATION; + } + s_extra->serinfo_next_aid = nxt->sinfo_assoc_id; + s_extra->serinfo_next_length = nxt->length; + s_extra->serinfo_next_ppid = nxt->sinfo_ppid; + s_extra->serinfo_next_stream = nxt->sinfo_stream; + if (nxt->tail_mbuf != NULL) { + if (nxt->end_added) { + s_extra->serinfo_next_flags |= SCTP_NEXT_MSG_ISCOMPLETE; + } + } + } else { + /* we explicitly 0 this, since the memcpy got + * some other things beyond the older sinfo_ + * that is on the control's structure :-D + */ + nxt = NULL; + s_extra->serinfo_next_flags = SCTP_NO_NEXT_MSG; + s_extra->serinfo_next_aid = 0; + s_extra->serinfo_next_length = 0; + s_extra->serinfo_next_ppid = 0; + s_extra->serinfo_next_stream = 0; + } + } + /* + * update off the real current cum-ack, if we have an stcb. + */ + if ((control->do_not_ref_stcb == 0) && stcb) + sinfo->sinfo_cumtsn = stcb->asoc.cumulative_tsn; + /* + * mask off the high bits, we keep the actual chunk bits in + * there. + */ + sinfo->sinfo_flags &= 0x00ff; + if ((control->sinfo_flags >> 8) & SCTP_DATA_UNORDERED) { + sinfo->sinfo_flags |= SCTP_UNORDERED; + } + } +#ifdef SCTP_ASOCLOG_OF_TSNS + { + int index, newindex; + struct sctp_pcbtsn_rlog *entry; + do { + index = inp->readlog_index; + newindex = index + 1; + if (newindex >= SCTP_READ_LOG_SIZE) { + newindex = 0; + } + } while (atomic_cmpset_int(&inp->readlog_index, index, newindex) == 0); + entry = &inp->readlog[index]; + entry->vtag = control->sinfo_assoc_id; + entry->strm = control->sinfo_stream; + entry->seq = (uint16_t)control->mid; + entry->sz = control->length; + entry->flgs = control->sinfo_flags; + } +#endif + if ((fromlen > 0) && (from != NULL)) { + union sctp_sockstore store; + size_t len; + + switch (control->whoFrom->ro._l_addr.sa.sa_family) { +#ifdef INET6 + case AF_INET6: + len = sizeof(struct sockaddr_in6); + store.sin6 = control->whoFrom->ro._l_addr.sin6; + store.sin6.sin6_port = control->port_from; + break; +#endif +#ifdef INET + case AF_INET: +#ifdef INET6 + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) { + len = sizeof(struct sockaddr_in6); + in6_sin_2_v4mapsin6(&control->whoFrom->ro._l_addr.sin, + &store.sin6); + store.sin6.sin6_port = control->port_from; + } else { + len = sizeof(struct sockaddr_in); + store.sin = control->whoFrom->ro._l_addr.sin; + store.sin.sin_port = control->port_from; + } +#else + len = sizeof(struct sockaddr_in); + store.sin = control->whoFrom->ro._l_addr.sin; + store.sin.sin_port = control->port_from; +#endif + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + len = sizeof(struct sockaddr_conn); + store.sconn = control->whoFrom->ro._l_addr.sconn; + store.sconn.sconn_port = control->port_from; + break; +#endif + default: + len = 0; + break; + } + memcpy(from, &store, min((size_t)fromlen, len)); +#if defined(SCTP_EMBEDDED_V6_SCOPE) +#ifdef INET6 + { + struct sockaddr_in6 lsa6, *from6; + + from6 = (struct sockaddr_in6 *)from; + sctp_recover_scope_mac(from6, (&lsa6)); + } +#endif +#endif + } + if (hold_rlock) { + SCTP_INP_READ_UNLOCK(inp); + hold_rlock = 0; + } + if (hold_sblock) { + SOCKBUF_UNLOCK(&so->so_rcv); + hold_sblock = 0; + } + /* now copy out what data we can */ + if (mp == NULL) { + /* copy out each mbuf in the chain up to length */ + get_more_data: + m = control->data; + while (m) { + /* Move out all we can */ +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + cp_len = uio->uio_resid; +#else + cp_len = uio_resid(uio); +#endif +#else + cp_len = uio->uio_resid; +#endif + my_len = SCTP_BUF_LEN(m); + if (cp_len > my_len) { + /* not enough in this buf */ + cp_len = my_len; + } + if (hold_rlock) { + SCTP_INP_READ_UNLOCK(inp); + hold_rlock = 0; + } +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 0); +#endif + if (cp_len > 0) + error = uiomove(mtod(m, char *), (int)cp_len, uio); +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_LOCK(so, 0); +#endif + /* re-read */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + goto release; + } + + if ((control->do_not_ref_stcb == 0) && stcb && + stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { + no_rcv_needed = 1; + } + if (error) { + /* error we are out of here */ + goto release; + } + SCTP_INP_READ_LOCK(inp); + hold_rlock = 1; + if (cp_len == SCTP_BUF_LEN(m)) { + if ((SCTP_BUF_NEXT(m)== NULL) && + (control->end_added)) { + out_flags |= MSG_EOR; + if ((control->do_not_ref_stcb == 0) && + (control->stcb != NULL) && + ((control->spec_flags & M_NOTIFICATION) == 0)) + control->stcb->asoc.strmin[control->sinfo_stream].delivery_started = 0; + } + if (control->spec_flags & M_NOTIFICATION) { + out_flags |= MSG_NOTIFICATION; + } + /* we ate up the mbuf */ + if (in_flags & MSG_PEEK) { + /* just looking */ + m = SCTP_BUF_NEXT(m); + copied_so_far += cp_len; + } else { + /* dispose of the mbuf */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(&so->so_rcv, + control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBFREE, SCTP_BUF_LEN(m)); + } + sctp_sbfree(control, stcb, &so->so_rcv, m); + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(&so->so_rcv, + control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBRESULT, 0); + } + copied_so_far += cp_len; + freed_so_far += (uint32_t)cp_len; + freed_so_far += MSIZE; + atomic_subtract_int(&control->length, (int)cp_len); + control->data = sctp_m_free(m); + m = control->data; + /* been through it all, must hold sb lock ok to null tail */ + if (control->data == NULL) { +#ifdef INVARIANTS +#if defined(__FreeBSD__) && !defined(__Userspace__) + if ((control->end_added == 0) || + (TAILQ_NEXT(control, next) == NULL)) { + /* If the end is not added, OR the + * next is NOT null we MUST have the lock. + */ + if (mtx_owned(&inp->inp_rdata_mtx) == 0) { + panic("Hmm we don't own the lock?"); + } + } +#endif +#endif + control->tail_mbuf = NULL; +#ifdef INVARIANTS + if ((control->end_added) && ((out_flags & MSG_EOR) == 0)) { + panic("end_added, nothing left and no MSG_EOR"); + } +#endif + } + } + } else { + /* Do we need to trim the mbuf? */ + if (control->spec_flags & M_NOTIFICATION) { + out_flags |= MSG_NOTIFICATION; + } + if ((in_flags & MSG_PEEK) == 0) { + SCTP_BUF_RESV_UF(m, cp_len); + SCTP_BUF_LEN(m) -= (int)cp_len; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(&so->so_rcv, control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBFREE, (int)cp_len); + } + atomic_subtract_int(&so->so_rcv.sb_cc, (int)cp_len); + if ((control->do_not_ref_stcb == 0) && + stcb) { + atomic_subtract_int(&stcb->asoc.sb_cc, (int)cp_len); + } + copied_so_far += cp_len; + freed_so_far += (uint32_t)cp_len; + freed_so_far += MSIZE; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(&so->so_rcv, control->do_not_ref_stcb?NULL:stcb, + SCTP_LOG_SBRESULT, 0); + } + atomic_subtract_int(&control->length, (int)cp_len); + } else { + copied_so_far += cp_len; + } + } +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + if ((out_flags & MSG_EOR) || (uio->uio_resid == 0)) { +#else + if ((out_flags & MSG_EOR) || (uio_resid(uio) == 0)) { +#endif +#else + if ((out_flags & MSG_EOR) || (uio->uio_resid == 0)) { +#endif + break; + } + if (((stcb) && (in_flags & MSG_PEEK) == 0) && + (control->do_not_ref_stcb == 0) && + (freed_so_far >= rwnd_req)) { + sctp_user_rcvd(stcb, &freed_so_far, hold_rlock, rwnd_req); + } + } /* end while(m) */ + /* + * At this point we have looked at it all and we either have + * a MSG_EOR/or read all the user wants... <OR> + * control->length == 0. + */ + if ((out_flags & MSG_EOR) && ((in_flags & MSG_PEEK) == 0)) { + /* we are done with this control */ + if (control->length == 0) { + if (control->data) { +#ifdef INVARIANTS + panic("control->data not null at read eor?"); +#else + SCTP_PRINTF("Strange, data left in the control buffer .. invariants would panic?\n"); + sctp_m_freem(control->data); + control->data = NULL; +#endif + } + done_with_control: + if (hold_rlock == 0) { + SCTP_INP_READ_LOCK(inp); + hold_rlock = 1; + } + TAILQ_REMOVE(&inp->read_queue, control, next); + /* Add back any hidden data */ + if (control->held_length) { + held_length = 0; + control->held_length = 0; + wakeup_read_socket = 1; + } + if (control->aux_data) { + sctp_m_free (control->aux_data); + control->aux_data = NULL; + } + no_rcv_needed = control->do_not_ref_stcb; + sctp_free_remote_addr(control->whoFrom); + control->data = NULL; +#ifdef INVARIANTS + if (control->on_strm_q) { + panic("About to free ctl:%p so:%p and its in %d", + control, so, control->on_strm_q); + } +#endif + sctp_free_a_readq(stcb, control); + control = NULL; + if ((freed_so_far >= rwnd_req) && + (no_rcv_needed == 0)) + sctp_user_rcvd(stcb, &freed_so_far, hold_rlock, rwnd_req); + + } else { + /* + * The user did not read all of this + * message, turn off the returned MSG_EOR + * since we are leaving more behind on the + * control to read. + */ +#ifdef INVARIANTS + if (control->end_added && + (control->data == NULL) && + (control->tail_mbuf == NULL)) { + panic("Gak, control->length is corrupt?"); + } +#endif + no_rcv_needed = control->do_not_ref_stcb; + out_flags &= ~MSG_EOR; + } + } + if (out_flags & MSG_EOR) { + goto release; + } +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + if ((uio->uio_resid == 0) || +#else + if ((uio_resid(uio) == 0) || +#endif +#else + if ((uio->uio_resid == 0) || +#endif + ((in_eeor_mode) && + (copied_so_far >= max(so->so_rcv.sb_lowat, 1)))) { + goto release; + } + /* + * If I hit here the receiver wants more and this message is + * NOT done (pd-api). So two questions. Can we block? if not + * we are done. Did the user NOT set MSG_WAITALL? + */ + if (block_allowed == 0) { + goto release; + } + /* + * We need to wait for more data a few things: + * - We don't release the I/O lock so we don't get someone else + * reading. + * - We must be sure to account for the case where what is added + * is NOT to our control when we wakeup. + */ + + /* Do we need to tell the transport a rwnd update might be + * needed before we go to sleep? + */ + if (((stcb) && (in_flags & MSG_PEEK) == 0) && + ((freed_so_far >= rwnd_req) && + (control->do_not_ref_stcb == 0) && + (no_rcv_needed == 0))) { + sctp_user_rcvd(stcb, &freed_so_far, hold_rlock, rwnd_req); + } + wait_some_more: +#if (defined(__FreeBSD__) || defined(_WIN32)) && !defined(__Userspace__) + if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { + goto release; + } +#else + if (so->so_state & SS_CANTRCVMORE) { + goto release; + } +#endif + + if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) + goto release; + + if (hold_rlock == 1) { + SCTP_INP_READ_UNLOCK(inp); + hold_rlock = 0; + } + if (hold_sblock == 0) { + SOCKBUF_LOCK(&so->so_rcv); + hold_sblock = 1; + } + if ((copied_so_far) && (control->length == 0) && + (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE))) { + goto release; + } +#if defined(__APPLE__) && !defined(__Userspace__) + sbunlock(&so->so_rcv, 1); +#endif + if (SCTP_SBAVAIL(&so->so_rcv) <= control->held_length) { +#if defined(__FreeBSD__) && !defined(__Userspace__) + error = sbwait(so, SO_RCV); +#else + error = sbwait(&so->so_rcv); +#endif + if (error) { +#if defined(__APPLE__) && !defined(__Userspace__) + goto release_unlocked; +#else + goto release; +#endif + } + control->held_length = 0; + } +#if defined(__APPLE__) && !defined(__Userspace__) + error = sblock(&so->so_rcv, SBLOCKWAIT(in_flags)); +#endif + if (hold_sblock) { + SOCKBUF_UNLOCK(&so->so_rcv); + hold_sblock = 0; + } + if (control->length == 0) { + /* still nothing here */ + if (control->end_added == 1) { + /* he aborted, or is done i.e.did a shutdown */ + out_flags |= MSG_EOR; + if (control->pdapi_aborted) { + if ((control->do_not_ref_stcb == 0) && ((control->spec_flags & M_NOTIFICATION) == 0)) + control->stcb->asoc.strmin[control->sinfo_stream].delivery_started = 0; + + out_flags |= MSG_TRUNC; + } else { + if ((control->do_not_ref_stcb == 0) && ((control->spec_flags & M_NOTIFICATION) == 0)) + control->stcb->asoc.strmin[control->sinfo_stream].delivery_started = 0; + } + goto done_with_control; + } + if (SCTP_SBAVAIL(&so->so_rcv) > held_length) { + control->held_length = SCTP_SBAVAIL(&so->so_rcv); + held_length = 0; + } + goto wait_some_more; + } else if (control->data == NULL) { + /* we must re-sync since data + * is probably being added + */ + SCTP_INP_READ_LOCK(inp); + if ((control->length > 0) && (control->data == NULL)) { + /* big trouble.. we have the lock and its corrupt? */ +#ifdef INVARIANTS + panic ("Impossible data==NULL length !=0"); +#endif + out_flags |= MSG_EOR; + out_flags |= MSG_TRUNC; + control->length = 0; + SCTP_INP_READ_UNLOCK(inp); + goto done_with_control; + } + SCTP_INP_READ_UNLOCK(inp); + /* We will fall around to get more data */ + } + goto get_more_data; + } else { + /*- + * Give caller back the mbuf chain, + * store in uio_resid the length + */ + wakeup_read_socket = 0; + if ((control->end_added == 0) || + (TAILQ_NEXT(control, next) == NULL)) { + /* Need to get rlock */ + if (hold_rlock == 0) { + SCTP_INP_READ_LOCK(inp); + hold_rlock = 1; + } + } + if (control->end_added) { + out_flags |= MSG_EOR; + if ((control->do_not_ref_stcb == 0) && + (control->stcb != NULL) && + ((control->spec_flags & M_NOTIFICATION) == 0)) + control->stcb->asoc.strmin[control->sinfo_stream].delivery_started = 0; + } + if (control->spec_flags & M_NOTIFICATION) { + out_flags |= MSG_NOTIFICATION; + } +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + uio->uio_resid = control->length; +#else + uio_setresid(uio, control->length); +#endif +#else + uio->uio_resid = control->length; +#endif + *mp = control->data; + m = control->data; + while (m) { + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(&so->so_rcv, + control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBFREE, SCTP_BUF_LEN(m)); + } + sctp_sbfree(control, stcb, &so->so_rcv, m); + freed_so_far += (uint32_t)SCTP_BUF_LEN(m); + freed_so_far += MSIZE; + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { + sctp_sblog(&so->so_rcv, + control->do_not_ref_stcb?NULL:stcb, SCTP_LOG_SBRESULT, 0); + } + m = SCTP_BUF_NEXT(m); + } + control->data = control->tail_mbuf = NULL; + control->length = 0; + if (out_flags & MSG_EOR) { + /* Done with this control */ + goto done_with_control; + } + } + release: + if (hold_rlock == 1) { + SCTP_INP_READ_UNLOCK(inp); + hold_rlock = 0; + } +#if defined(__Userspace__) + if (hold_sblock == 0) { + SOCKBUF_LOCK(&so->so_rcv); + hold_sblock = 1; + } +#else + if (hold_sblock == 1) { + SOCKBUF_UNLOCK(&so->so_rcv); + hold_sblock = 0; + } +#endif +#if defined(__APPLE__) && !defined(__Userspace__) + sbunlock(&so->so_rcv, 1); +#endif + +#if defined(__FreeBSD__) && !defined(__Userspace__) + SOCK_IO_RECV_UNLOCK(so); + sockbuf_lock = 0; +#endif + + release_unlocked: + if (hold_sblock) { + SOCKBUF_UNLOCK(&so->so_rcv); + hold_sblock = 0; + } + if ((stcb) && (in_flags & MSG_PEEK) == 0) { + if ((freed_so_far >= rwnd_req) && + (control && (control->do_not_ref_stcb == 0)) && + (no_rcv_needed == 0)) + sctp_user_rcvd(stcb, &freed_so_far, hold_rlock, rwnd_req); + } + out: + if (msg_flags) { + *msg_flags = out_flags; + } + if (((out_flags & MSG_EOR) == 0) && + ((in_flags & MSG_PEEK) == 0) && + (sinfo) && + (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO) || + sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVNXTINFO))) { + struct sctp_extrcvinfo *s_extra; + s_extra = (struct sctp_extrcvinfo *)sinfo; + s_extra->serinfo_next_flags = SCTP_NO_NEXT_MSG; + } + if (hold_rlock == 1) { + SCTP_INP_READ_UNLOCK(inp); + } + if (hold_sblock) { + SOCKBUF_UNLOCK(&so->so_rcv); + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (sockbuf_lock) { + SOCK_IO_RECV_UNLOCK(so); + } +#endif + + if (freecnt_applied) { + /* + * The lock on the socket buffer protects us so the free + * code will stop. But since we used the socketbuf lock and + * the sender uses the tcb_lock to increment, we need to use + * the atomic add to the refcnt. + */ + if (stcb == NULL) { +#ifdef INVARIANTS + panic("stcb for refcnt has gone NULL?"); + goto stage_left; +#else + goto stage_left; +#endif + } + /* Save the value back for next time */ + stcb->freed_by_sorcv_sincelast = freed_so_far; + atomic_subtract_int(&stcb->asoc.refcnt, 1); + } + if (SCTP_BASE_SYSCTL(sctp_logging_level) &SCTP_RECV_RWND_LOGGING_ENABLE) { + if (stcb) { + sctp_misc_ints(SCTP_SORECV_DONE, + freed_so_far, +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + ((uio) ? (slen - uio->uio_resid) : slen), +#else + ((uio) ? (slen - uio_resid(uio)) : slen), +#endif +#else + (uint32_t)((uio) ? (slen - uio->uio_resid) : slen), +#endif + stcb->asoc.my_rwnd, + SCTP_SBAVAIL(&so->so_rcv)); + } else { + sctp_misc_ints(SCTP_SORECV_DONE, + freed_so_far, +#if defined(__APPLE__) && !defined(__Userspace__) +#if defined(APPLE_LEOPARD) + ((uio) ? (slen - uio->uio_resid) : slen), +#else + ((uio) ? (slen - uio_resid(uio)) : slen), +#endif +#else + (uint32_t)((uio) ? (slen - uio->uio_resid) : slen), +#endif + 0, + SCTP_SBAVAIL(&so->so_rcv)); + } + } + stage_left: + if (wakeup_read_socket) { + sctp_sorwakeup(inp, so); + } + return (error); +} + +#ifdef SCTP_MBUF_LOGGING +struct mbuf * +sctp_m_free(struct mbuf *m) +{ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mb(m, SCTP_MBUF_IFREE); + } + return (m_free(m)); +} + +void +sctp_m_freem(struct mbuf *mb) +{ + while (mb != NULL) + mb = sctp_m_free(mb); +} + +#endif + +int +sctp_dynamic_set_primary(struct sockaddr *sa, uint32_t vrf_id) +{ + /* Given a local address. For all associations + * that holds the address, request a peer-set-primary. + */ + struct sctp_ifa *ifa; + struct sctp_laddr *wi; + + ifa = sctp_find_ifa_by_addr(sa, vrf_id, SCTP_ADDR_NOT_LOCKED); + if (ifa == NULL) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, EADDRNOTAVAIL); + return (EADDRNOTAVAIL); + } + /* Now that we have the ifa we must awaken the + * iterator with this message. + */ + wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); + if (wi == NULL) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, ENOMEM); + return (ENOMEM); + } + /* Now incr the count and int wi structure */ + SCTP_INCR_LADDR_COUNT(); + memset(wi, 0, sizeof(*wi)); + (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); + wi->ifa = ifa; + wi->action = SCTP_SET_PRIM_ADDR; + atomic_add_int(&ifa->refcount, 1); + + /* Now add it to the work queue */ + SCTP_WQ_ADDR_LOCK(); + /* + * Should this really be a tailq? As it is we will process the + * newest first :-0 + */ + LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); + sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, + (struct sctp_inpcb *)NULL, + (struct sctp_tcb *)NULL, + (struct sctp_nets *)NULL); + SCTP_WQ_ADDR_UNLOCK(); + return (0); +} + +#if defined(__Userspace__) +/* no sctp_soreceive for __Userspace__ now */ +#endif +#if !defined(__Userspace__) +int +sctp_soreceive( struct socket *so, + struct sockaddr **psa, + struct uio *uio, + struct mbuf **mp0, + struct mbuf **controlp, + int *flagsp) +{ + int error, fromlen; + uint8_t sockbuf[256]; + struct sockaddr *from; + struct sctp_extrcvinfo sinfo; + int filling_sinfo = 1; + int flags; + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *)so->so_pcb; + /* pickup the assoc we are reading from */ + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + return (EINVAL); + } + if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT) && + sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVRCVINFO) && + sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVNXTINFO)) || + (controlp == NULL)) { + /* user does not want the sndrcv ctl */ + filling_sinfo = 0; + } + if (psa) { + from = (struct sockaddr *)sockbuf; + fromlen = sizeof(sockbuf); +#ifdef HAVE_SA_LEN + from->sa_len = 0; +#endif + } else { + from = NULL; + fromlen = 0; + } + +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_LOCK(so, 1); +#endif + if (filling_sinfo) { + memset(&sinfo, 0, sizeof(struct sctp_extrcvinfo)); + } + if (flagsp != NULL) { + flags = *flagsp; + } else { + flags = 0; + } + error = sctp_sorecvmsg(so, uio, mp0, from, fromlen, &flags, + (struct sctp_sndrcvinfo *)&sinfo, filling_sinfo); + if (flagsp != NULL) { + *flagsp = flags; + } + if (controlp != NULL) { + /* copy back the sinfo in a CMSG format */ + if (filling_sinfo && ((flags & MSG_NOTIFICATION) == 0)) { + *controlp = sctp_build_ctl_nchunk(inp, + (struct sctp_sndrcvinfo *)&sinfo); + } else { + *controlp = NULL; + } + } + if (psa) { + /* copy back the address info */ +#ifdef HAVE_SA_LEN + if (from && from->sa_len) { +#else + if (from) { +#endif +#if (defined(__FreeBSD__) || defined(_WIN32)) && !defined(__Userspace__) + *psa = sodupsockaddr(from, M_NOWAIT); +#else + *psa = dup_sockaddr(from, mp0 == 0); +#endif + } else { + *psa = NULL; + } + } +#if defined(__APPLE__) && !defined(__Userspace__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + return (error); +} + +#if defined(_WIN32) && !defined(__Userspace__) +/* + * General routine to allocate a hash table with control of memory flags. + * is in 7.0 and beyond for sure :-) + */ +void * +sctp_hashinit_flags(int elements, struct malloc_type *type, + u_long *hashmask, int flags) +{ + long hashsize; + LIST_HEAD(generic, generic) *hashtbl; + int i; + + + if (elements <= 0) { +#ifdef INVARIANTS + panic("hashinit: bad elements"); +#else + SCTP_PRINTF("hashinit: bad elements?"); + elements = 1; +#endif + } + for (hashsize = 1; hashsize <= elements; hashsize <<= 1) + continue; + hashsize >>= 1; + if (flags & HASH_WAITOK) + hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl), type, M_WAITOK); + else if (flags & HASH_NOWAIT) + hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl), type, M_NOWAIT); + else { +#ifdef INVARIANTS + panic("flag incorrect in hashinit_flags"); +#else + return (NULL); +#endif + } + + /* no memory? */ + if (hashtbl == NULL) + return (NULL); + + for (i = 0; i < hashsize; i++) + LIST_INIT(&hashtbl[i]); + *hashmask = hashsize - 1; + return (hashtbl); +} +#endif +#else /* __Userspace__ ifdef above sctp_soreceive */ +/* + * __Userspace__ Defining sctp_hashinit_flags() and sctp_hashdestroy() for userland. + * NOTE: We don't want multiple definitions here. So sctp_hashinit_flags() above for + *__FreeBSD__ must be excluded. + * + */ + +void * +sctp_hashinit_flags(int elements, struct malloc_type *type, + u_long *hashmask, int flags) +{ + long hashsize; + LIST_HEAD(generic, generic) *hashtbl; + int i; + + if (elements <= 0) { + SCTP_PRINTF("hashinit: bad elements?"); +#ifdef INVARIANTS + return (NULL); +#else + elements = 1; +#endif + } + for (hashsize = 1; hashsize <= elements; hashsize <<= 1) + continue; + hashsize >>= 1; + /*cannot use MALLOC here because it has to be declared or defined + using MALLOC_DECLARE or MALLOC_DEFINE first. */ + if (flags & HASH_WAITOK) + hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl)); + else if (flags & HASH_NOWAIT) + hashtbl = malloc((u_long)hashsize * sizeof(*hashtbl)); + else { +#ifdef INVARIANTS + SCTP_PRINTF("flag incorrect in hashinit_flags.\n"); +#endif + return (NULL); + } + + /* no memory? */ + if (hashtbl == NULL) + return (NULL); + + for (i = 0; i < hashsize; i++) + LIST_INIT(&hashtbl[i]); + *hashmask = hashsize - 1; + return (hashtbl); +} + +void +sctp_hashdestroy(void *vhashtbl, struct malloc_type *type, u_long hashmask) +{ + LIST_HEAD(generic, generic) *hashtbl, *hp; + + hashtbl = vhashtbl; + for (hp = hashtbl; hp <= &hashtbl[hashmask]; hp++) + if (!LIST_EMPTY(hp)) { + SCTP_PRINTF("hashdestroy: hash not empty.\n"); + return; + } + FREE(hashtbl, type); +} + +void +sctp_hashfreedestroy(void *vhashtbl, struct malloc_type *type, u_long hashmask) +{ + LIST_HEAD(generic, generic) *hashtbl/*, *hp*/; + /* + LIST_ENTRY(type) *start, *temp; + */ + hashtbl = vhashtbl; + /* Apparently temp is not dynamically allocated, so attempts to + free it results in error. + for (hp = hashtbl; hp <= &hashtbl[hashmask]; hp++) + if (!LIST_EMPTY(hp)) { + start = LIST_FIRST(hp); + while (start != NULL) { + temp = start; + start = start->le_next; + SCTP_PRINTF("%s: %p \n", __func__, (void *)temp); + FREE(temp, type); + } + } + */ + FREE(hashtbl, type); +} + +#endif +int +sctp_connectx_helper_add(struct sctp_tcb *stcb, struct sockaddr *addr, + int totaddr, int *error) +{ + int added = 0; + int i; + struct sctp_inpcb *inp; + struct sockaddr *sa; + size_t incr = 0; +#ifdef INET + struct sockaddr_in *sin; +#endif +#ifdef INET6 + struct sockaddr_in6 *sin6; +#endif + + sa = addr; + inp = stcb->sctp_ep; + *error = 0; + for (i = 0; i < totaddr; i++) { + switch (sa->sa_family) { +#ifdef INET + case AF_INET: + incr = sizeof(struct sockaddr_in); + sin = (struct sockaddr_in *)sa; + if ((sin->sin_addr.s_addr == INADDR_ANY) || + (sin->sin_addr.s_addr == INADDR_BROADCAST) || + IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTPUTIL + SCTP_LOC_7); + *error = EINVAL; + goto out_now; + } + if (sctp_add_remote_addr(stcb, sa, NULL, stcb->asoc.port, + SCTP_DONOT_SETSCOPE, + SCTP_ADDR_IS_CONFIRMED)) { + /* assoc gone no un-lock */ + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOBUFS); + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTPUTIL + SCTP_LOC_8); + *error = ENOBUFS; + goto out_now; + } + added++; + break; +#endif +#ifdef INET6 + case AF_INET6: + incr = sizeof(struct sockaddr_in6); + sin6 = (struct sockaddr_in6 *)sa; + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) || + IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTPUTIL + SCTP_LOC_9); + *error = EINVAL; + goto out_now; + } + if (sctp_add_remote_addr(stcb, sa, NULL, stcb->asoc.port, + SCTP_DONOT_SETSCOPE, + SCTP_ADDR_IS_CONFIRMED)) { + /* assoc gone no un-lock */ + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOBUFS); + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTPUTIL + SCTP_LOC_10); + *error = ENOBUFS; + goto out_now; + } + added++; + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + incr = sizeof(struct sockaddr_conn); + if (sctp_add_remote_addr(stcb, sa, NULL, stcb->asoc.port, + SCTP_DONOT_SETSCOPE, + SCTP_ADDR_IS_CONFIRMED)) { + /* assoc gone no un-lock */ + SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOBUFS); + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTPUTIL + SCTP_LOC_11); + *error = ENOBUFS; + goto out_now; + } + added++; + break; +#endif + default: + break; + } + sa = (struct sockaddr *)((caddr_t)sa + incr); + } + out_now: + return (added); +} + +int +sctp_connectx_helper_find(struct sctp_inpcb *inp, struct sockaddr *addr, + unsigned int totaddr, + unsigned int *num_v4, unsigned int *num_v6, + unsigned int limit) +{ + struct sockaddr *sa; + struct sctp_tcb *stcb; + unsigned int incr, at, i; + + at = 0; + sa = addr; + *num_v6 = *num_v4 = 0; + /* account and validate addresses */ + if (totaddr == 0) { + return (EINVAL); + } + for (i = 0; i < totaddr; i++) { + if (at + sizeof(struct sockaddr) > limit) { + return (EINVAL); + } + switch (sa->sa_family) { +#ifdef INET + case AF_INET: + incr = (unsigned int)sizeof(struct sockaddr_in); +#ifdef HAVE_SA_LEN + if (sa->sa_len != incr) { + return (EINVAL); + } +#endif + (*num_v4) += 1; + break; +#endif +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6; + + incr = (unsigned int)sizeof(struct sockaddr_in6); +#ifdef HAVE_SA_LEN + if (sa->sa_len != incr) { + return (EINVAL); + } +#endif + sin6 = (struct sockaddr_in6 *)sa; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + /* Must be non-mapped for connectx */ + return (EINVAL); + } + (*num_v6) += 1; + break; + } +#endif + default: + return (EINVAL); + } + if ((at + incr) > limit) { + return (EINVAL); + } + SCTP_INP_INCR_REF(inp); + stcb = sctp_findassociation_ep_addr(&inp, sa, NULL, NULL, NULL); + if (stcb != NULL) { + SCTP_TCB_UNLOCK(stcb); + return (EALREADY); + } else { + SCTP_INP_DECR_REF(inp); + } + at += incr; + sa = (struct sockaddr *)((caddr_t)sa + incr); + } + return (0); +} + +/* + * sctp_bindx(ADD) for one address. + * assumes all arguments are valid/checked by caller. + */ +void +sctp_bindx_add_address(struct socket *so, struct sctp_inpcb *inp, + struct sockaddr *sa, uint32_t vrf_id, int *error, + void *p) +{ +#if defined(INET) && defined(INET6) + struct sockaddr_in sin; +#endif +#ifdef INET6 + struct sockaddr_in6 *sin6; +#endif +#ifdef INET + struct sockaddr_in *sinp; +#endif + struct sockaddr *addr_to_use; + struct sctp_inpcb *lep; +#ifdef SCTP_MVRF + int i; +#endif + uint16_t port; + + /* see if we're bound all already! */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } +#ifdef SCTP_MVRF + /* Is the VRF one we have */ + for (i = 0; i < inp->num_vrfs; i++) { + if (vrf_id == inp->m_vrf_ids[i]) { + break; + } + } + if (i == inp->num_vrfs) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } +#endif + switch (sa->sa_family) { +#ifdef INET6 + case AF_INET6: +#ifdef HAVE_SA_LEN + if (sa->sa_len != sizeof(struct sockaddr_in6)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } +#endif + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { + /* can only bind v6 on PF_INET6 sockets */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } + sin6 = (struct sockaddr_in6 *)sa; + port = sin6->sin6_port; +#ifdef INET + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(inp)) { + /* can't bind v4-mapped on PF_INET sockets */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } + in6_sin6_2_sin(&sin, sin6); + addr_to_use = (struct sockaddr *)&sin; + } else { + addr_to_use = sa; + } +#else + addr_to_use = sa; +#endif + break; +#endif +#ifdef INET + case AF_INET: +#ifdef HAVE_SA_LEN + if (sa->sa_len != sizeof(struct sockaddr_in)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } +#endif + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(inp)) { + /* can't bind v4 on PF_INET sockets */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } + sinp = (struct sockaddr_in *)sa; + port = sinp->sin_port; + addr_to_use = sa; + break; +#endif + default: + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } + if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { +#if !(defined(_WIN32) || defined(__Userspace__)) + if (p == NULL) { + /* Can't get proc for Net/Open BSD */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } +#endif + *error = sctp_inpcb_bind(so, addr_to_use, NULL, p); + return; + } + /* Validate the incoming port. */ + if ((port != 0) && (port != inp->sctp_lport)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } + lep = sctp_pcb_findep(addr_to_use, 1, 0, vrf_id); + if (lep == NULL) { + /* add the address */ + *error = sctp_addr_mgmt_ep_sa(inp, addr_to_use, + SCTP_ADD_IP_ADDRESS, vrf_id); + } else { + if (lep != inp) { + *error = EADDRINUSE; + } + SCTP_INP_DECR_REF(lep); + } +} + +/* + * sctp_bindx(DELETE) for one address. + * assumes all arguments are valid/checked by caller. + */ +void +sctp_bindx_delete_address(struct sctp_inpcb *inp, + struct sockaddr *sa, uint32_t vrf_id, int *error) +{ + struct sockaddr *addr_to_use; +#if defined(INET) && defined(INET6) + struct sockaddr_in6 *sin6; + struct sockaddr_in sin; +#endif +#ifdef SCTP_MVRF + int i; +#endif + + /* see if we're bound all already! */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } +#ifdef SCTP_MVRF + /* Is the VRF one we have */ + for (i = 0; i < inp->num_vrfs; i++) { + if (vrf_id == inp->m_vrf_ids[i]) { + break; + } + } + if (i == inp->num_vrfs) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } +#endif + switch (sa->sa_family) { +#ifdef INET6 + case AF_INET6: +#ifdef HAVE_SA_LEN + if (sa->sa_len != sizeof(struct sockaddr_in6)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } +#endif + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { + /* can only bind v6 on PF_INET6 sockets */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } +#ifdef INET + sin6 = (struct sockaddr_in6 *)sa; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(inp)) { + /* can't bind mapped-v4 on PF_INET sockets */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } + in6_sin6_2_sin(&sin, sin6); + addr_to_use = (struct sockaddr *)&sin; + } else { + addr_to_use = sa; + } +#else + addr_to_use = sa; +#endif + break; +#endif +#ifdef INET + case AF_INET: +#ifdef HAVE_SA_LEN + if (sa->sa_len != sizeof(struct sockaddr_in)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } +#endif + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(inp)) { + /* can't bind v4 on PF_INET sockets */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } + addr_to_use = sa; + break; +#endif + default: + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); + *error = EINVAL; + return; + } + /* No lock required mgmt_ep_sa does its own locking. */ + *error = sctp_addr_mgmt_ep_sa(inp, addr_to_use, SCTP_DEL_IP_ADDRESS, + vrf_id); +} + +/* + * returns the valid local address count for an assoc, taking into account + * all scoping rules + */ +int +sctp_local_addr_count(struct sctp_tcb *stcb) +{ + int loopback_scope; +#if defined(INET) + int ipv4_local_scope, ipv4_addr_legal; +#endif +#if defined(INET6) + int local_scope, site_scope, ipv6_addr_legal; +#endif +#if defined(__Userspace__) + int conn_addr_legal; +#endif + struct sctp_vrf *vrf; + struct sctp_ifn *sctp_ifn; + struct sctp_ifa *sctp_ifa; + int count = 0; + + /* Turn on all the appropriate scopes */ + loopback_scope = stcb->asoc.scope.loopback_scope; +#if defined(INET) + ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope; + ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal; +#endif +#if defined(INET6) + local_scope = stcb->asoc.scope.local_scope; + site_scope = stcb->asoc.scope.site_scope; + ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal; +#endif +#if defined(__Userspace__) + conn_addr_legal = stcb->asoc.scope.conn_addr_legal; +#endif + SCTP_IPI_ADDR_RLOCK(); + vrf = sctp_find_vrf(stcb->asoc.vrf_id); + if (vrf == NULL) { + /* no vrf, no addresses */ + SCTP_IPI_ADDR_RUNLOCK(); + return (0); + } + + if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + /* + * bound all case: go through all ifns on the vrf + */ + LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { + if ((loopback_scope == 0) && + SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { + continue; + } + LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { + if (sctp_is_addr_restricted(stcb, sctp_ifa)) + continue; + switch (sctp_ifa->address.sa.sa_family) { +#ifdef INET + case AF_INET: + if (ipv4_addr_legal) { + struct sockaddr_in *sin; + + sin = &sctp_ifa->address.sin; + if (sin->sin_addr.s_addr == 0) { + /* skip unspecified addrs */ + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred, + &sin->sin_addr) != 0) { + continue; + } +#endif + if ((ipv4_local_scope == 0) && + (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { + continue; + } + /* count this one */ + count++; + } else { + continue; + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (ipv6_addr_legal) { + struct sockaddr_in6 *sin6; + +#if defined(SCTP_EMBEDDED_V6_SCOPE) && !defined(SCTP_KAME) + struct sockaddr_in6 lsa6; +#endif + sin6 = &sctp_ifa->address.sin6; + if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { + continue; + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred, + &sin6->sin6_addr) != 0) { + continue; + } +#endif + if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { + if (local_scope == 0) + continue; +#if defined(SCTP_EMBEDDED_V6_SCOPE) + if (sin6->sin6_scope_id == 0) { +#ifdef SCTP_KAME + if (sa6_recoverscope(sin6) != 0) + /* + * bad link + * local + * address + */ + continue; +#else + lsa6 = *sin6; + if (in6_recoverscope(&lsa6, + &lsa6.sin6_addr, + NULL)) + /* + * bad link + * local + * address + */ + continue; + sin6 = &lsa6; +#endif /* SCTP_KAME */ + } +#endif /* SCTP_EMBEDDED_V6_SCOPE */ + } + if ((site_scope == 0) && + (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { + continue; + } + /* count this one */ + count++; + } + break; +#endif +#if defined(__Userspace__) + case AF_CONN: + if (conn_addr_legal) { + count++; + } + break; +#endif + default: + /* TSNH */ + break; + } + } + } + } else { + /* + * subset bound case + */ + struct sctp_laddr *laddr; + LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, + sctp_nxt_addr) { + if (sctp_is_addr_restricted(stcb, laddr->ifa)) { + continue; + } + /* count this one */ + count++; + } + } + SCTP_IPI_ADDR_RUNLOCK(); + return (count); +} + +#if defined(SCTP_LOCAL_TRACE_BUF) + +void +sctp_log_trace(uint32_t subsys, const char *str SCTP_UNUSED, uint32_t a, uint32_t b, uint32_t c, uint32_t d, uint32_t e, uint32_t f) +{ + uint32_t saveindex, newindex; + +#if defined(_WIN32) && !defined(__Userspace__) + if (SCTP_BASE_SYSCTL(sctp_log) == NULL) { + return; + } + do { + saveindex = SCTP_BASE_SYSCTL(sctp_log)->index; + if (saveindex >= SCTP_MAX_LOGGING_SIZE) { + newindex = 1; + } else { + newindex = saveindex + 1; + } + } while (atomic_cmpset_int(&SCTP_BASE_SYSCTL(sctp_log)->index, saveindex, newindex) == 0); + if (saveindex >= SCTP_MAX_LOGGING_SIZE) { + saveindex = 0; + } + SCTP_BASE_SYSCTL(sctp_log)->entry[saveindex].timestamp = SCTP_GET_CYCLECOUNT; + SCTP_BASE_SYSCTL(sctp_log)->entry[saveindex].subsys = subsys; + SCTP_BASE_SYSCTL(sctp_log)->entry[saveindex].params[0] = a; + SCTP_BASE_SYSCTL(sctp_log)->entry[saveindex].params[1] = b; + SCTP_BASE_SYSCTL(sctp_log)->entry[saveindex].params[2] = c; + SCTP_BASE_SYSCTL(sctp_log)->entry[saveindex].params[3] = d; + SCTP_BASE_SYSCTL(sctp_log)->entry[saveindex].params[4] = e; + SCTP_BASE_SYSCTL(sctp_log)->entry[saveindex].params[5] = f; +#else + do { + saveindex = SCTP_BASE_SYSCTL(sctp_log).index; + if (saveindex >= SCTP_MAX_LOGGING_SIZE) { + newindex = 1; + } else { + newindex = saveindex + 1; + } + } while (atomic_cmpset_int(&SCTP_BASE_SYSCTL(sctp_log).index, saveindex, newindex) == 0); + if (saveindex >= SCTP_MAX_LOGGING_SIZE) { + saveindex = 0; + } + SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].timestamp = SCTP_GET_CYCLECOUNT; + SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].subsys = subsys; + SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[0] = a; + SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[1] = b; + SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[2] = c; + SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[3] = d; + SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[4] = e; + SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[5] = f; +#endif +} + +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) +static bool +sctp_recv_udp_tunneled_packet(struct mbuf *m, int off, struct inpcb *inp, + const struct sockaddr *sa SCTP_UNUSED, void *ctx SCTP_UNUSED) +{ + struct ip *iph; +#ifdef INET6 + struct ip6_hdr *ip6; +#endif + struct mbuf *sp, *last; + struct udphdr *uhdr; + uint16_t port; + + if ((m->m_flags & M_PKTHDR) == 0) { + /* Can't handle one that is not a pkt hdr */ + goto out; + } + /* Pull the src port */ + iph = mtod(m, struct ip *); + uhdr = (struct udphdr *)((caddr_t)iph + off); + port = uhdr->uh_sport; + /* Split out the mbuf chain. Leave the + * IP header in m, place the + * rest in the sp. + */ + sp = m_split(m, off, M_NOWAIT); + if (sp == NULL) { + /* Gak, drop packet, we can't do a split */ + goto out; + } + if (sp->m_pkthdr.len < sizeof(struct udphdr) + sizeof(struct sctphdr)) { + /* Gak, packet can't have an SCTP header in it - too small */ + m_freem(sp); + goto out; + } + /* Now pull up the UDP header and SCTP header together */ + sp = m_pullup(sp, sizeof(struct udphdr) + sizeof(struct sctphdr)); + if (sp == NULL) { + /* Gak pullup failed */ + goto out; + } + /* Trim out the UDP header */ + m_adj(sp, sizeof(struct udphdr)); + + /* Now reconstruct the mbuf chain */ + for (last = m; last->m_next; last = last->m_next); + last->m_next = sp; + m->m_pkthdr.len += sp->m_pkthdr.len; + /* + * The CSUM_DATA_VALID flags indicates that the HW checked the + * UDP checksum and it was valid. + * Since CSUM_DATA_VALID == CSUM_SCTP_VALID this would imply that + * the HW also verified the SCTP checksum. Therefore, clear the bit. + */ + SCTPDBG(SCTP_DEBUG_CRCOFFLOAD, + "sctp_recv_udp_tunneled_packet(): Packet of length %d received on %s with csum_flags 0x%b.\n", + m->m_pkthdr.len, + if_name(m->m_pkthdr.rcvif), + (int)m->m_pkthdr.csum_flags, CSUM_BITS); + m->m_pkthdr.csum_flags &= ~CSUM_DATA_VALID; + iph = mtod(m, struct ip *); + switch (iph->ip_v) { +#ifdef INET + case IPVERSION: + iph->ip_len = htons(ntohs(iph->ip_len) - sizeof(struct udphdr)); + sctp_input_with_port(m, off, port); + break; +#endif +#ifdef INET6 + case IPV6_VERSION >> 4: + ip6 = mtod(m, struct ip6_hdr *); + ip6->ip6_plen = htons(ntohs(ip6->ip6_plen) - sizeof(struct udphdr)); + sctp6_input_with_port(&m, &off, port); + break; +#endif + default: + goto out; + break; + } + return (true); + out: + m_freem(m); + + return (true); +} + +#ifdef INET +static void +sctp_recv_icmp_tunneled_packet(udp_tun_icmp_param_t param) +{ + struct icmp *icmp = param.icmp; + struct ip *outer_ip, *inner_ip; + struct sctphdr *sh; + struct udphdr *udp; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; + struct sctp_nets *net; + struct sctp_init_chunk *ch; + struct sockaddr_in src, dst; + uint8_t type, code; + + inner_ip = &icmp->icmp_ip; + outer_ip = (struct ip *)((caddr_t)icmp - sizeof(struct ip)); + if (ntohs(outer_ip->ip_len) < + sizeof(struct ip) + 8 + (inner_ip->ip_hl << 2) + sizeof(struct udphdr) + 8) { + return; + } + udp = (struct udphdr *)((caddr_t)inner_ip + (inner_ip->ip_hl << 2)); + sh = (struct sctphdr *)(udp + 1); + memset(&src, 0, sizeof(struct sockaddr_in)); + src.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + src.sin_len = sizeof(struct sockaddr_in); +#endif + src.sin_port = sh->src_port; + src.sin_addr = inner_ip->ip_src; + memset(&dst, 0, sizeof(struct sockaddr_in)); + dst.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + dst.sin_len = sizeof(struct sockaddr_in); +#endif + dst.sin_port = sh->dest_port; + dst.sin_addr = inner_ip->ip_dst; + /* + * 'dst' holds the dest of the packet that failed to be sent. + * 'src' holds our local endpoint address. Thus we reverse + * the dst and the src in the lookup. + */ + inp = NULL; + net = NULL; + stcb = sctp_findassociation_addr_sa((struct sockaddr *)&dst, + (struct sockaddr *)&src, + &inp, &net, 1, + SCTP_DEFAULT_VRFID); + if ((stcb != NULL) && + (net != NULL) && + (inp != NULL)) { + /* Check the UDP port numbers */ + if ((udp->uh_dport != net->port) || + (udp->uh_sport != htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)))) { + SCTP_TCB_UNLOCK(stcb); + return; + } + /* Check the verification tag */ + if (ntohl(sh->v_tag) != 0) { + /* + * This must be the verification tag used + * for sending out packets. We don't + * consider packets reflecting the + * verification tag. + */ + if (ntohl(sh->v_tag) != stcb->asoc.peer_vtag) { + SCTP_TCB_UNLOCK(stcb); + return; + } + } else { + if (ntohs(outer_ip->ip_len) >= + sizeof(struct ip) + + 8 + (inner_ip->ip_hl << 2) + 8 + 20) { + /* + * In this case we can check if we + * got an INIT chunk and if the + * initiate tag matches. + */ + ch = (struct sctp_init_chunk *)(sh + 1); + if ((ch->ch.chunk_type != SCTP_INITIATION) || + (ntohl(ch->init.initiate_tag) != stcb->asoc.my_vtag)) { + SCTP_TCB_UNLOCK(stcb); + return; + } + } else { + SCTP_TCB_UNLOCK(stcb); + return; + } + } + type = icmp->icmp_type; + code = icmp->icmp_code; + if ((type == ICMP_UNREACH) && + (code == ICMP_UNREACH_PORT)) { + code = ICMP_UNREACH_PROTOCOL; + } + sctp_notify(inp, stcb, net, type, code, + ntohs(inner_ip->ip_len), + (uint32_t)ntohs(icmp->icmp_nextmtu)); +#if defined(__Userspace__) + if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (stcb->sctp_socket != NULL)) { + struct socket *upcall_socket; + + upcall_socket = stcb->sctp_socket; + SOCK_LOCK(upcall_socket); + soref(upcall_socket); + SOCK_UNLOCK(upcall_socket); + if ((upcall_socket->so_upcall != NULL) && + (upcall_socket->so_error != 0)) { + (*upcall_socket->so_upcall)(upcall_socket, upcall_socket->so_upcallarg, M_NOWAIT); + } + ACCEPT_LOCK(); + SOCK_LOCK(upcall_socket); + sorele(upcall_socket); + } +#endif + } else { + if ((stcb == NULL) && (inp != NULL)) { + /* reduce ref-count */ + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + SCTP_INP_WUNLOCK(inp); + } + if (stcb) { + SCTP_TCB_UNLOCK(stcb); + } + } + return; +} +#endif + +#ifdef INET6 +static void +sctp_recv_icmp6_tunneled_packet(udp_tun_icmp_param_t param) +{ + struct ip6ctlparam *ip6cp = param.ip6cp; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; + struct sctp_nets *net; + struct sctphdr sh; + struct udphdr udp; + struct sockaddr_in6 src, dst; + uint8_t type, code; + + /* + * XXX: We assume that when IPV6 is non NULL, M and OFF are + * valid. + */ + if (ip6cp->ip6c_m == NULL) { + return; + } + /* Check if we can safely examine the ports and the + * verification tag of the SCTP common header. + */ + if (ip6cp->ip6c_m->m_pkthdr.len < + ip6cp->ip6c_off + sizeof(struct udphdr)+ offsetof(struct sctphdr, checksum)) { + return; + } + /* Copy out the UDP header. */ + memset(&udp, 0, sizeof(struct udphdr)); + m_copydata(ip6cp->ip6c_m, + ip6cp->ip6c_off, + sizeof(struct udphdr), + (caddr_t)&udp); + /* Copy out the port numbers and the verification tag. */ + memset(&sh, 0, sizeof(struct sctphdr)); + m_copydata(ip6cp->ip6c_m, + ip6cp->ip6c_off + sizeof(struct udphdr), + sizeof(uint16_t) + sizeof(uint16_t) + sizeof(uint32_t), + (caddr_t)&sh); + memset(&src, 0, sizeof(struct sockaddr_in6)); + src.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + src.sin6_len = sizeof(struct sockaddr_in6); +#endif + src.sin6_port = sh.src_port; + src.sin6_addr = ip6cp->ip6c_ip6->ip6_src; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (in6_setscope(&src.sin6_addr, ip6cp->ip6c_m->m_pkthdr.rcvif, NULL) != 0) { + return; + } +#endif + memset(&dst, 0, sizeof(struct sockaddr_in6)); + dst.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + dst.sin6_len = sizeof(struct sockaddr_in6); +#endif + dst.sin6_port = sh.dest_port; + dst.sin6_addr = ip6cp->ip6c_ip6->ip6_dst; +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (in6_setscope(&dst.sin6_addr, ip6cp->ip6c_m->m_pkthdr.rcvif, NULL) != 0) { + return; + } +#endif + inp = NULL; + net = NULL; + stcb = sctp_findassociation_addr_sa((struct sockaddr *)&dst, + (struct sockaddr *)&src, + &inp, &net, 1, SCTP_DEFAULT_VRFID); + if ((stcb != NULL) && + (net != NULL) && + (inp != NULL)) { + /* Check the UDP port numbers */ + if ((udp.uh_dport != net->port) || + (udp.uh_sport != htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)))) { + SCTP_TCB_UNLOCK(stcb); + return; + } + /* Check the verification tag */ + if (ntohl(sh.v_tag) != 0) { + /* + * This must be the verification tag used for + * sending out packets. We don't consider + * packets reflecting the verification tag. + */ + if (ntohl(sh.v_tag) != stcb->asoc.peer_vtag) { + SCTP_TCB_UNLOCK(stcb); + return; + } + } else { +#if defined(__FreeBSD__) && !defined(__Userspace__) + if (ip6cp->ip6c_m->m_pkthdr.len >= + ip6cp->ip6c_off + sizeof(struct udphdr) + + sizeof(struct sctphdr) + + sizeof(struct sctp_chunkhdr) + + offsetof(struct sctp_init, a_rwnd)) { + /* + * In this case we can check if we + * got an INIT chunk and if the + * initiate tag matches. + */ + uint32_t initiate_tag; + uint8_t chunk_type; + + m_copydata(ip6cp->ip6c_m, + ip6cp->ip6c_off + + sizeof(struct udphdr) + + sizeof(struct sctphdr), + sizeof(uint8_t), + (caddr_t)&chunk_type); + m_copydata(ip6cp->ip6c_m, + ip6cp->ip6c_off + + sizeof(struct udphdr) + + sizeof(struct sctphdr) + + sizeof(struct sctp_chunkhdr), + sizeof(uint32_t), + (caddr_t)&initiate_tag); + if ((chunk_type != SCTP_INITIATION) || + (ntohl(initiate_tag) != stcb->asoc.my_vtag)) { + SCTP_TCB_UNLOCK(stcb); + return; + } + } else { + SCTP_TCB_UNLOCK(stcb); + return; + } +#else + SCTP_TCB_UNLOCK(stcb); + return; +#endif + } + type = ip6cp->ip6c_icmp6->icmp6_type; + code = ip6cp->ip6c_icmp6->icmp6_code; + if ((type == ICMP6_DST_UNREACH) && + (code == ICMP6_DST_UNREACH_NOPORT)) { + type = ICMP6_PARAM_PROB; + code = ICMP6_PARAMPROB_NEXTHEADER; + } + sctp6_notify(inp, stcb, net, type, code, + ntohl(ip6cp->ip6c_icmp6->icmp6_mtu)); +#if defined(__Userspace__) + if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (stcb->sctp_socket != NULL)) { + struct socket *upcall_socket; + + upcall_socket = stcb->sctp_socket; + SOCK_LOCK(upcall_socket); + soref(upcall_socket); + SOCK_UNLOCK(upcall_socket); + if ((upcall_socket->so_upcall != NULL) && + (upcall_socket->so_error != 0)) { + (*upcall_socket->so_upcall)(upcall_socket, upcall_socket->so_upcallarg, M_NOWAIT); + } + ACCEPT_LOCK(); + SOCK_LOCK(upcall_socket); + sorele(upcall_socket); + } +#endif + } else { + if ((stcb == NULL) && (inp != NULL)) { + /* reduce inp's ref-count */ + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + SCTP_INP_WUNLOCK(inp); + } + if (stcb) { + SCTP_TCB_UNLOCK(stcb); + } + } +} +#endif + +void +sctp_over_udp_stop(void) +{ + /* + * This function assumes sysctl caller holds sctp_sysctl_info_lock() for writing! + */ +#ifdef INET + if (SCTP_BASE_INFO(udp4_tun_socket) != NULL) { + soclose(SCTP_BASE_INFO(udp4_tun_socket)); + SCTP_BASE_INFO(udp4_tun_socket) = NULL; + } +#endif +#ifdef INET6 + if (SCTP_BASE_INFO(udp6_tun_socket) != NULL) { + soclose(SCTP_BASE_INFO(udp6_tun_socket)); + SCTP_BASE_INFO(udp6_tun_socket) = NULL; + } +#endif +} + +int +sctp_over_udp_start(void) +{ + uint16_t port; + int ret; +#ifdef INET + struct sockaddr_in sin; +#endif +#ifdef INET6 + struct sockaddr_in6 sin6; +#endif + /* + * This function assumes sysctl caller holds sctp_sysctl_info_lock() for writing! + */ + port = SCTP_BASE_SYSCTL(sctp_udp_tunneling_port); + if (ntohs(port) == 0) { + /* Must have a port set */ + return (EINVAL); + } +#ifdef INET + if (SCTP_BASE_INFO(udp4_tun_socket) != NULL) { + /* Already running -- must stop first */ + return (EALREADY); + } +#endif +#ifdef INET6 + if (SCTP_BASE_INFO(udp6_tun_socket) != NULL) { + /* Already running -- must stop first */ + return (EALREADY); + } +#endif +#ifdef INET + if ((ret = socreate(PF_INET, &SCTP_BASE_INFO(udp4_tun_socket), + SOCK_DGRAM, IPPROTO_UDP, + curthread->td_ucred, curthread))) { + sctp_over_udp_stop(); + return (ret); + } + /* Call the special UDP hook. */ + if ((ret = udp_set_kernel_tunneling(SCTP_BASE_INFO(udp4_tun_socket), + sctp_recv_udp_tunneled_packet, + sctp_recv_icmp_tunneled_packet, + NULL))) { + sctp_over_udp_stop(); + return (ret); + } + /* Ok, we have a socket, bind it to the port. */ + memset(&sin, 0, sizeof(struct sockaddr_in)); + sin.sin_len = sizeof(struct sockaddr_in); + sin.sin_family = AF_INET; + sin.sin_port = htons(port); + if ((ret = sobind(SCTP_BASE_INFO(udp4_tun_socket), + (struct sockaddr *)&sin, curthread))) { + sctp_over_udp_stop(); + return (ret); + } +#endif +#ifdef INET6 + if ((ret = socreate(PF_INET6, &SCTP_BASE_INFO(udp6_tun_socket), + SOCK_DGRAM, IPPROTO_UDP, + curthread->td_ucred, curthread))) { + sctp_over_udp_stop(); + return (ret); + } + /* Call the special UDP hook. */ + if ((ret = udp_set_kernel_tunneling(SCTP_BASE_INFO(udp6_tun_socket), + sctp_recv_udp_tunneled_packet, + sctp_recv_icmp6_tunneled_packet, + NULL))) { + sctp_over_udp_stop(); + return (ret); + } + /* Ok, we have a socket, bind it to the port. */ + memset(&sin6, 0, sizeof(struct sockaddr_in6)); + sin6.sin6_len = sizeof(struct sockaddr_in6); + sin6.sin6_family = AF_INET6; + sin6.sin6_port = htons(port); + if ((ret = sobind(SCTP_BASE_INFO(udp6_tun_socket), + (struct sockaddr *)&sin6, curthread))) { + sctp_over_udp_stop(); + return (ret); + } +#endif + return (0); +} +#endif + +/* + * sctp_min_mtu ()returns the minimum of all non-zero arguments. + * If all arguments are zero, zero is returned. + */ +uint32_t +sctp_min_mtu(uint32_t mtu1, uint32_t mtu2, uint32_t mtu3) +{ + if (mtu1 > 0) { + if (mtu2 > 0) { + if (mtu3 > 0) { + return (min(mtu1, min(mtu2, mtu3))); + } else { + return (min(mtu1, mtu2)); + } + } else { + if (mtu3 > 0) { + return (min(mtu1, mtu3)); + } else { + return (mtu1); + } + } + } else { + if (mtu2 > 0) { + if (mtu3 > 0) { + return (min(mtu2, mtu3)); + } else { + return (mtu2); + } + } else { + return (mtu3); + } + } +} + +#if defined(__FreeBSD__) && !defined(__Userspace__) +void +sctp_hc_set_mtu(union sctp_sockstore *addr, uint16_t fibnum, uint32_t mtu) +{ + struct in_conninfo inc; + + memset(&inc, 0, sizeof(struct in_conninfo)); + inc.inc_fibnum = fibnum; + switch (addr->sa.sa_family) { +#ifdef INET + case AF_INET: + inc.inc_faddr = addr->sin.sin_addr; + break; +#endif +#ifdef INET6 + case AF_INET6: + inc.inc_flags |= INC_ISIPV6; + inc.inc6_faddr = addr->sin6.sin6_addr; + break; +#endif + default: + return; + } + tcp_hc_updatemtu(&inc, (u_long)mtu); +} + +uint32_t +sctp_hc_get_mtu(union sctp_sockstore *addr, uint16_t fibnum) +{ + struct in_conninfo inc; + + memset(&inc, 0, sizeof(struct in_conninfo)); + inc.inc_fibnum = fibnum; + switch (addr->sa.sa_family) { +#ifdef INET + case AF_INET: + inc.inc_faddr = addr->sin.sin_addr; + break; +#endif +#ifdef INET6 + case AF_INET6: + inc.inc_flags |= INC_ISIPV6; + inc.inc6_faddr = addr->sin6.sin6_addr; + break; +#endif + default: + return (0); + } + return ((uint32_t)tcp_hc_getmtu(&inc)); +} +#endif + +void +sctp_set_state(struct sctp_tcb *stcb, int new_state) +{ +#if defined(KDTRACE_HOOKS) + int old_state = stcb->asoc.state; +#endif + + KASSERT((new_state & ~SCTP_STATE_MASK) == 0, + ("sctp_set_state: Can't set substate (new_state = %x)", + new_state)); + stcb->asoc.state = (stcb->asoc.state & ~SCTP_STATE_MASK) | new_state; + if ((new_state == SCTP_STATE_SHUTDOWN_RECEIVED) || + (new_state == SCTP_STATE_SHUTDOWN_SENT) || + (new_state == SCTP_STATE_SHUTDOWN_ACK_SENT)) { + SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_SHUTDOWN_PENDING); + } +#if defined(KDTRACE_HOOKS) + if (((old_state & SCTP_STATE_MASK) != new_state) && + !(((old_state & SCTP_STATE_MASK) == SCTP_STATE_EMPTY) && + (new_state == SCTP_STATE_INUSE))) { + SCTP_PROBE6(state__change, NULL, stcb, NULL, stcb, NULL, old_state); + } +#endif +} + +void +sctp_add_substate(struct sctp_tcb *stcb, int substate) +{ +#if defined(KDTRACE_HOOKS) + int old_state = stcb->asoc.state; +#endif + + KASSERT((substate & SCTP_STATE_MASK) == 0, + ("sctp_add_substate: Can't set state (substate = %x)", + substate)); + stcb->asoc.state |= substate; +#if defined(KDTRACE_HOOKS) + if (((substate & SCTP_STATE_ABOUT_TO_BE_FREED) && + ((old_state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) || + ((substate & SCTP_STATE_SHUTDOWN_PENDING) && + ((old_state & SCTP_STATE_SHUTDOWN_PENDING) == 0))) { + SCTP_PROBE6(state__change, NULL, stcb, NULL, stcb, NULL, old_state); + } +#endif +} + diff --git a/netwerk/sctp/src/netinet/sctputil.h b/netwerk/sctp/src/netinet/sctputil.h new file mode 100644 index 0000000000..74140b99d2 --- /dev/null +++ b/netwerk/sctp/src/netinet/sctputil.h @@ -0,0 +1,390 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET_SCTP_UTIL_H_ +#define _NETINET_SCTP_UTIL_H_ + +#if defined(_KERNEL) || defined(__Userspace__) + +#define SCTP_READ_LOCK_HELD 1 +#define SCTP_READ_LOCK_NOT_HELD 0 + +#ifdef SCTP_ASOCLOG_OF_TSNS +void sctp_print_out_track_log(struct sctp_tcb *stcb); +#endif + +#ifdef SCTP_MBUF_LOGGING +struct mbuf *sctp_m_free(struct mbuf *m); +void sctp_m_freem(struct mbuf *m); +#else +#define sctp_m_free m_free +#define sctp_m_freem m_freem +#endif + +#if defined(SCTP_LOCAL_TRACE_BUF) +void +sctp_log_trace(uint32_t fr, const char *str SCTP_UNUSED, uint32_t a, uint32_t b, uint32_t c, uint32_t d, uint32_t e, uint32_t f); +#endif + +#define sctp_get_associd(stcb) ((sctp_assoc_t)stcb->asoc.assoc_id) + +/* + * Function prototypes + */ +int32_t +sctp_map_assoc_state(int); + +uint32_t +sctp_get_ifa_hash_val(struct sockaddr *addr); + +struct sctp_ifa * +sctp_find_ifa_in_ep(struct sctp_inpcb *inp, struct sockaddr *addr, int hold_lock); + +struct sctp_ifa * +sctp_find_ifa_by_addr(struct sockaddr *addr, uint32_t vrf_id, int holds_lock); + +uint32_t sctp_select_initial_TSN(struct sctp_pcb *); + +uint32_t sctp_select_a_tag(struct sctp_inpcb *, uint16_t lport, uint16_t rport, int); + +int sctp_init_asoc(struct sctp_inpcb *, struct sctp_tcb *, uint32_t, uint32_t, uint32_t, uint16_t); + +void sctp_fill_random_store(struct sctp_pcb *); + +void +sctp_notify_stream_reset_add(struct sctp_tcb *stcb, uint16_t numberin, + uint16_t numberout, int flag); +void +sctp_notify_stream_reset_tsn(struct sctp_tcb *stcb, uint32_t sending_tsn, uint32_t recv_tsn, int flag); + +/* + * NOTE: sctp_timer_start() will increment the reference count of any relevant + * structure the timer is referencing, in order to prevent a race condition + * between the timer executing and the structure being freed. + * + * When the timer fires or sctp_timer_stop() is called, these references are + * removed. + */ +void +sctp_timer_start(int, struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *); + +void +sctp_timer_stop(int, struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *, uint32_t); + +int +sctp_dynamic_set_primary(struct sockaddr *sa, uint32_t vrf_id); + +void +sctp_wakeup_the_read_socket(struct sctp_inpcb *inp, struct sctp_tcb *stcb, + int so_locked +#if !(defined(__APPLE__) && !defined(__Userspace__)) + SCTP_UNUSED +#endif +); + +#if defined(__Userspace__) +void sctp_invoke_recv_callback(struct sctp_inpcb *, + struct sctp_tcb *, + struct sctp_queued_to_read *, + int); + +#endif +void +sctp_add_to_readq(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_queued_to_read *control, + struct sockbuf *sb, + int end, + int inpread_locked, + int so_locked); + +void sctp_iterator_worker(void); + +uint32_t sctp_get_prev_mtu(uint32_t); +uint32_t sctp_get_next_mtu(uint32_t); + +void +sctp_timeout_handler(void *); + +int +sctp_calculate_rto(struct sctp_tcb *, struct sctp_association *, + struct sctp_nets *, struct timeval *, int); + +uint32_t sctp_calculate_len(struct mbuf *); + +caddr_t sctp_m_getptr(struct mbuf *, int, int, uint8_t *); + +struct sctp_paramhdr * +sctp_get_next_param(struct mbuf *, int, + struct sctp_paramhdr *, int); + +struct mbuf * +sctp_add_pad_tombuf(struct mbuf *, int); + +struct mbuf * +sctp_pad_lastmbuf(struct mbuf *, int, struct mbuf *); + +void sctp_ulp_notify(uint32_t, struct sctp_tcb *, uint32_t, void *, int); + +void +sctp_pull_off_control_to_new_inp(struct sctp_inpcb *old_inp, + struct sctp_inpcb *new_inp, + struct sctp_tcb *stcb, int waitflags); + +void sctp_stop_timers_for_shutdown(struct sctp_tcb *); + +/* Stop all timers for association and remote addresses. */ +void sctp_stop_association_timers(struct sctp_tcb *, bool); + +void sctp_report_all_outbound(struct sctp_tcb *, uint16_t, int); + +int sctp_expand_mapping_array(struct sctp_association *, uint32_t); + +void sctp_abort_notification(struct sctp_tcb *, bool, bool, uint16_t, + struct sctp_abort_chunk *, int); + +/* We abort responding to an IP packet for some reason */ +void +sctp_abort_association(struct sctp_inpcb *, struct sctp_tcb *, struct mbuf *, + int, struct sockaddr *, struct sockaddr *, + struct sctphdr *, struct mbuf *, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t, uint32_t, +#endif + uint32_t, uint16_t); + +/* We choose to abort via user input */ +void +sctp_abort_an_association(struct sctp_inpcb *, struct sctp_tcb *, + struct mbuf *, bool, int); + +void sctp_handle_ootb(struct mbuf *, int, int, + struct sockaddr *, struct sockaddr *, + struct sctphdr *, struct sctp_inpcb *, + struct mbuf *, +#if defined(__FreeBSD__) && !defined(__Userspace__) + uint8_t, uint32_t, uint16_t, +#endif + uint32_t, uint16_t); + +int sctp_connectx_helper_add(struct sctp_tcb *stcb, struct sockaddr *addr, + int totaddr, int *error); + +int +sctp_connectx_helper_find(struct sctp_inpcb *, struct sockaddr *, + unsigned int, unsigned int *, unsigned int *, unsigned int); + +int sctp_is_there_an_abort_here(struct mbuf *, int, uint32_t *); +#ifdef INET6 +uint32_t sctp_is_same_scope(struct sockaddr_in6 *, struct sockaddr_in6 *); + +#if defined(SCTP_EMBEDDED_V6_SCOPE) +struct sockaddr_in6 * +sctp_recover_scope(struct sockaddr_in6 *, struct sockaddr_in6 *); + +#ifdef SCTP_KAME +#define sctp_recover_scope_mac(addr, store) do { \ + if ((addr->sin6_family == AF_INET6) && \ + (IN6_IS_SCOPE_LINKLOCAL(&addr->sin6_addr))) { \ + *store = *addr; \ + if (addr->sin6_scope_id == 0) { \ + if (!sa6_recoverscope(store)) { \ + addr = store; \ + } \ + } else { \ + in6_clearscope(&addr->sin6_addr); \ + addr = store; \ + } \ + } \ +} while (0) +#else +#define sctp_recover_scope_mac(addr, store) do { \ + if ((addr->sin6_family == AF_INET6) && \ + (IN6_IS_SCOPE_LINKLOCAL(&addr->sin6_addr))) { \ + *store = *addr; \ + if (addr->sin6_scope_id == 0) { \ + if (!in6_recoverscope(store, &store->sin6_addr, \ + NULL)) { \ + addr = store; \ + } \ + } else { \ + in6_clearscope(&addr->sin6_addr); \ + addr = store; \ + } \ + } \ +} while (0) +#endif +#endif +#endif + +int sctp_cmpaddr(struct sockaddr *, struct sockaddr *); + +void sctp_print_address(struct sockaddr *); + +int +sctp_release_pr_sctp_chunk(struct sctp_tcb *, struct sctp_tmit_chunk *, + uint8_t, int); + +struct mbuf *sctp_generate_cause(uint16_t, char *); +struct mbuf *sctp_generate_no_user_data_cause(uint32_t); + +void sctp_bindx_add_address(struct socket *so, struct sctp_inpcb *inp, + struct sockaddr *sa, uint32_t vrf_id, int *error, + void *p); +void sctp_bindx_delete_address(struct sctp_inpcb *inp, struct sockaddr *sa, + uint32_t vrf_id, int *error); + +int sctp_local_addr_count(struct sctp_tcb *stcb); + +void +sctp_free_bufspace(struct sctp_tcb *, struct sctp_association *, + struct sctp_tmit_chunk *, int); + +#define sctp_free_spbufspace(stcb, asoc, sp) \ +do { \ + if (sp->data != NULL) { \ + if ((asoc)->total_output_queue_size >= sp->length) { \ + atomic_subtract_int(&(asoc)->total_output_queue_size, sp->length); \ + } else { \ + (asoc)->total_output_queue_size = 0; \ + } \ + if (stcb->sctp_socket && ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || \ + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL))) { \ + if (stcb->sctp_socket->so_snd.sb_cc >= sp->length) { \ + atomic_subtract_int(&stcb->sctp_socket->so_snd.sb_cc,sp->length); \ + } else { \ + stcb->sctp_socket->so_snd.sb_cc = 0; \ + } \ + } \ + } \ +} while (0) + +#define sctp_snd_sb_alloc(stcb, sz) \ +do { \ + atomic_add_int(&stcb->asoc.total_output_queue_size,sz); \ + if ((stcb->sctp_socket != NULL) && \ + ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || \ + (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL))) { \ + atomic_add_int(&stcb->sctp_socket->so_snd.sb_cc,sz); \ + } \ +} while (0) + +/* functions to start/stop udp tunneling */ +#if (defined(__APPLE__) || defined(__FreeBSD__)) && !defined(__Userspace__) +void sctp_over_udp_stop(void); +int sctp_over_udp_start(void); +#endif +#if defined(_WIN32) && !defined(__Userspace__) +void sctp_over_udp_restart(void); +#endif + +int +sctp_soreceive(struct socket *so, struct sockaddr **psa, + struct uio *uio, + struct mbuf **mp0, + struct mbuf **controlp, + int *flagsp); + +void +sctp_misc_ints(uint8_t from, uint32_t a, uint32_t b, uint32_t c, uint32_t d); + +void +sctp_wakeup_log(struct sctp_tcb *stcb, + uint32_t wake_cnt, int from); + +void sctp_log_strm_del_alt(struct sctp_tcb *stcb, uint32_t, uint16_t, uint16_t, int); + +void sctp_log_nagle_event(struct sctp_tcb *stcb, int action); + +#ifdef SCTP_MBUF_LOGGING +void +sctp_log_mb(struct mbuf *m, int from); + +void +sctp_log_mbc(struct mbuf *m, int from); +#endif + +void +sctp_sblog(struct sockbuf *sb, + struct sctp_tcb *stcb, int from, int incr); + +void +sctp_log_strm_del(struct sctp_queued_to_read *control, + struct sctp_queued_to_read *poschk, + int from); +void sctp_log_cwnd(struct sctp_tcb *stcb, struct sctp_nets *, int, uint8_t); +void rto_logging(struct sctp_nets *net, int from); + +void sctp_log_closing(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int16_t loc); + +void sctp_log_lock(struct sctp_inpcb *inp, struct sctp_tcb *stcb, uint8_t from); +void sctp_log_maxburst(struct sctp_tcb *stcb, struct sctp_nets *, int, int, uint8_t); +void sctp_log_block(uint8_t, struct sctp_association *, ssize_t); +void sctp_log_rwnd(uint8_t, uint32_t, uint32_t, uint32_t); +void sctp_log_rwnd_set(uint8_t, uint32_t, uint32_t, uint32_t, uint32_t); +int sctp_fill_stat_log(void *, size_t *); +void sctp_log_fr(uint32_t, uint32_t, uint32_t, int); +void sctp_log_sack(uint32_t, uint32_t, uint32_t, uint16_t, uint16_t, int); +void sctp_log_map(uint32_t, uint32_t, uint32_t, int); +void sctp_print_mapping_array(struct sctp_association *asoc); +void sctp_clr_stat_log(void); + +#ifdef SCTP_AUDITING_ENABLED +void +sctp_auditing(int, struct sctp_inpcb *, struct sctp_tcb *, + struct sctp_nets *); +void sctp_audit_log(uint8_t, uint8_t); + +#endif +uint32_t sctp_min_mtu(uint32_t, uint32_t, uint32_t); +#if defined(__FreeBSD__) && !defined(__Userspace__) +void sctp_hc_set_mtu(union sctp_sockstore *, uint16_t, uint32_t); +uint32_t sctp_hc_get_mtu(union sctp_sockstore *, uint16_t); +#endif +void sctp_set_state(struct sctp_tcb *, int); +void sctp_add_substate(struct sctp_tcb *, int); +uint32_t sctp_ticks_to_msecs(uint32_t); +uint32_t sctp_msecs_to_ticks(uint32_t); +uint32_t sctp_ticks_to_secs(uint32_t); +uint32_t sctp_secs_to_ticks(uint32_t); + +#endif /* _KERNEL */ +#endif diff --git a/netwerk/sctp/src/netinet6/sctp6_usrreq.c b/netwerk/sctp/src/netinet6/sctp6_usrreq.c new file mode 100644 index 0000000000..a4606e51a8 --- /dev/null +++ b/netwerk/sctp/src/netinet6/sctp6_usrreq.c @@ -0,0 +1,1826 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#include <netinet/sctp_os.h> +#ifdef INET6 +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/proc.h> +#endif +#include <netinet/sctp_pcb.h> +#include <netinet/sctp_header.h> +#include <netinet/sctp_var.h> +#include <netinet6/sctp6_var.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_uio.h> +#include <netinet/sctp_asconf.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_indata.h> +#include <netinet/sctp_timer.h> +#include <netinet/sctp_auth.h> +#include <netinet/sctp_input.h> +#include <netinet/sctp_output.h> +#include <netinet/sctp_bsd_addr.h> +#include <netinet/sctp_crc32.h> +#if !defined(_WIN32) +#include <netinet/icmp6.h> +#include <netinet/udp.h> +#endif +#if defined(__Userspace__) +int ip6_v6only=0; +#endif +#if defined(__Userspace__) +#ifdef INET +void +in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6) +{ +#if defined(_WIN32) + uint32_t temp; +#endif + memset(sin, 0, sizeof(*sin)); +#ifdef HAVE_SIN_LEN + sin->sin_len = sizeof(struct sockaddr_in); +#endif + sin->sin_family = AF_INET; + sin->sin_port = sin6->sin6_port; +#if defined(_WIN32) + temp = sin6->sin6_addr.s6_addr16[7]; + temp = temp << 16; + temp = temp | sin6->sin6_addr.s6_addr16[6]; + sin->sin_addr.s_addr = temp; +#else + sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3]; +#endif +} + +void +in6_sin6_2_sin_in_sock(struct sockaddr *nam) +{ + struct sockaddr_in *sin_p; + struct sockaddr_in6 sin6; + + /* save original sockaddr_in6 addr and convert it to sockaddr_in */ + sin6 = *(struct sockaddr_in6 *)nam; + sin_p = (struct sockaddr_in *)nam; + in6_sin6_2_sin(sin_p, &sin6); +} + +void +in6_sin_2_v4mapsin6(const struct sockaddr_in *sin, struct sockaddr_in6 *sin6) +{ + memset(sin6, 0, sizeof(struct sockaddr_in6)); + sin6->sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6->sin6_len = sizeof(struct sockaddr_in6); +#endif + sin6->sin6_port = sin->sin_port; +#if defined(_WIN32) + ((uint32_t *)&sin6->sin6_addr)[0] = 0; + ((uint32_t *)&sin6->sin6_addr)[1] = 0; + ((uint32_t *)&sin6->sin6_addr)[2] = htonl(0xffff); + ((uint32_t *)&sin6->sin6_addr)[3] = sin->sin_addr.s_addr; +#else + sin6->sin6_addr.s6_addr32[0] = 0; + sin6->sin6_addr.s6_addr32[1] = 0; + sin6->sin6_addr.s6_addr32[2] = htonl(0xffff); + sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr; +#endif +} +#endif +#endif + +#if !defined(__Userspace__) +int +#if defined(__APPLE__) || defined(__FreeBSD__) +sctp6_input_with_port(struct mbuf **i_pak, int *offp, uint16_t port) +#else +sctp6_input(struct mbuf **i_pak, int *offp, int proto) +#endif +{ + struct mbuf *m; + int iphlen; + uint32_t vrf_id; + uint8_t ecn_bits; + struct sockaddr_in6 src, dst; + struct ip6_hdr *ip6; + struct sctphdr *sh; + struct sctp_chunkhdr *ch; + int length, offset; + uint8_t compute_crc; +#if defined(__FreeBSD__) + uint32_t mflowid; + uint8_t mflowtype; + uint16_t fibnum; +#endif +#if !(defined(__APPLE__) || defined(__FreeBSD__)) + uint16_t port = 0; +#endif + + iphlen = *offp; + if (SCTP_GET_PKT_VRFID(*i_pak, vrf_id)) { + SCTP_RELEASE_PKT(*i_pak); + return (IPPROTO_DONE); + } + m = SCTP_HEADER_TO_CHAIN(*i_pak); +#ifdef SCTP_MBUF_LOGGING + /* Log in any input mbufs */ + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { + sctp_log_mbc(m, SCTP_MBUF_INPUT); + } +#endif +#ifdef SCTP_PACKET_LOGGING + if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) { + sctp_packet_log(m); + } +#endif +#if defined(__FreeBSD__) + SCTPDBG(SCTP_DEBUG_CRCOFFLOAD, + "sctp6_input(): Packet of length %d received on %s with csum_flags 0x%b.\n", + m->m_pkthdr.len, + if_name(m->m_pkthdr.rcvif), + (int)m->m_pkthdr.csum_flags, CSUM_BITS); +#endif +#if defined(__APPLE__) + SCTPDBG(SCTP_DEBUG_CRCOFFLOAD, + "sctp6_input(): Packet of length %d received on %s%d with csum_flags 0x%x.\n", + m->m_pkthdr.len, + m->m_pkthdr.rcvif->if_name, + m->m_pkthdr.rcvif->if_unit, + m->m_pkthdr.csum_flags); +#endif +#if defined(_WIN32) && !defined(__Userspace__) + SCTPDBG(SCTP_DEBUG_CRCOFFLOAD, + "sctp6_input(): Packet of length %d received on %s with csum_flags 0x%x.\n", + m->m_pkthdr.len, + m->m_pkthdr.rcvif->if_xname, + m->m_pkthdr.csum_flags); +#endif +#if defined(__FreeBSD__) + mflowid = m->m_pkthdr.flowid; + mflowtype = M_HASHTYPE_GET(m); + fibnum = M_GETFIB(m); +#endif + SCTP_STAT_INCR(sctps_recvpackets); + SCTP_STAT_INCR_COUNTER64(sctps_inpackets); + /* Get IP, SCTP, and first chunk header together in the first mbuf. */ + offset = iphlen + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); + if (m->m_len < offset) { + m = m_pullup(m, offset); + if (m == NULL) { + SCTP_STAT_INCR(sctps_hdrops); + return (IPPROTO_DONE); + } + } + ip6 = mtod(m, struct ip6_hdr *); + sh = (struct sctphdr *)(mtod(m, caddr_t) + iphlen); + ch = (struct sctp_chunkhdr *)((caddr_t)sh + sizeof(struct sctphdr)); + offset -= sizeof(struct sctp_chunkhdr); + memset(&src, 0, sizeof(struct sockaddr_in6)); + src.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + src.sin6_len = sizeof(struct sockaddr_in6); +#endif + src.sin6_port = sh->src_port; + src.sin6_addr = ip6->ip6_src; +#if defined(__FreeBSD__) +#if defined(__APPLE__) + /* XXX: This code should also be used on Apple */ +#endif + if (in6_setscope(&src.sin6_addr, m->m_pkthdr.rcvif, NULL) != 0) { + goto out; + } +#endif + memset(&dst, 0, sizeof(struct sockaddr_in6)); + dst.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + dst.sin6_len = sizeof(struct sockaddr_in6); +#endif + dst.sin6_port = sh->dest_port; + dst.sin6_addr = ip6->ip6_dst; +#if defined(__FreeBSD__) +#if defined(__APPLE__) + /* XXX: This code should also be used on Apple */ +#endif + if (in6_setscope(&dst.sin6_addr, m->m_pkthdr.rcvif, NULL) != 0) { + goto out; + } +#endif +#if defined(__APPLE__) +#if defined(NFAITH) && 0 < NFAITH + if (faithprefix(&dst.sin6_addr)) { + goto out; + } +#endif +#endif + length = ntohs(ip6->ip6_plen) + iphlen; + /* Validate mbuf chain length with IP payload length. */ + if (SCTP_HEADER_LEN(m) != length) { + SCTPDBG(SCTP_DEBUG_INPUT1, + "sctp6_input() length:%d reported length:%d\n", length, SCTP_HEADER_LEN(m)); + SCTP_STAT_INCR(sctps_hdrops); + goto out; + } + if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { + goto out; + } +#if defined(__FreeBSD__) + ecn_bits = IPV6_TRAFFIC_CLASS(ip6); + if (m->m_pkthdr.csum_flags & CSUM_SCTP_VALID) { + SCTP_STAT_INCR(sctps_recvhwcrc); + compute_crc = 0; + } else { +#else + ecn_bits = ((ntohl(ip6->ip6_flow) >> 20) & 0x000000ff); + if (SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) && + (IN6_ARE_ADDR_EQUAL(&src.sin6_addr, &dst.sin6_addr))) { + SCTP_STAT_INCR(sctps_recvhwcrc); + compute_crc = 0; + } else { +#endif + SCTP_STAT_INCR(sctps_recvswcrc); + compute_crc = 1; + } + sctp_common_input_processing(&m, iphlen, offset, length, + (struct sockaddr *)&src, + (struct sockaddr *)&dst, + sh, ch, + compute_crc, + ecn_bits, +#if defined(__FreeBSD__) + mflowtype, mflowid, fibnum, +#endif + vrf_id, port); + out: + if (m) { + sctp_m_freem(m); + } + return (IPPROTO_DONE); +} + +#if defined(__APPLE__) +int +sctp6_input(struct mbuf **i_pak, int *offp) +{ + return (sctp6_input_with_port(i_pak, offp, 0)); +} +#endif +#if defined(__FreeBSD__) +int +sctp6_input(struct mbuf **i_pak, int *offp, int proto SCTP_UNUSED) +{ + return (sctp6_input_with_port(i_pak, offp, 0)); +} +#endif + +void +sctp6_notify(struct sctp_inpcb *inp, + struct sctp_tcb *stcb, + struct sctp_nets *net, + uint8_t icmp6_type, + uint8_t icmp6_code, + uint32_t next_mtu) +{ +#if defined(__APPLE__) + struct socket *so; +#endif + int timer_stopped; + + switch (icmp6_type) { + case ICMP6_DST_UNREACH: + if ((icmp6_code == ICMP6_DST_UNREACH_NOROUTE) || + (icmp6_code == ICMP6_DST_UNREACH_ADMIN) || + (icmp6_code == ICMP6_DST_UNREACH_BEYONDSCOPE) || + (icmp6_code == ICMP6_DST_UNREACH_ADDR)) { + /* Mark the net unreachable. */ + if (net->dest_state & SCTP_ADDR_REACHABLE) { + /* Ok that destination is not reachable */ + net->dest_state &= ~SCTP_ADDR_REACHABLE; + net->dest_state &= ~SCTP_ADDR_PF; + sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, + stcb, 0, (void *)net, SCTP_SO_NOT_LOCKED); + } + } + SCTP_TCB_UNLOCK(stcb); + break; + case ICMP6_PARAM_PROB: + /* Treat it like an ABORT. */ + if (icmp6_code == ICMP6_PARAMPROB_NEXTHEADER) { + sctp_abort_notification(stcb, true, false, 0, NULL, SCTP_SO_NOT_LOCKED); +#if defined(__APPLE__) + so = SCTP_INP_SO(inp); + atomic_add_int(&stcb->asoc.refcnt, 1); + SCTP_TCB_UNLOCK(stcb); + SCTP_SOCKET_LOCK(so, 1); + SCTP_TCB_LOCK(stcb); + atomic_subtract_int(&stcb->asoc.refcnt, 1); +#endif + (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_2); +#if defined(__APPLE__) + SCTP_SOCKET_UNLOCK(so, 1); +#endif + } else { + SCTP_TCB_UNLOCK(stcb); + } + break; + case ICMP6_PACKET_TOO_BIG: + if (net->dest_state & SCTP_ADDR_NO_PMTUD) { + SCTP_TCB_UNLOCK(stcb); + break; + } + if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { + timer_stopped = 1; + sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, + SCTP_FROM_SCTP_USRREQ + SCTP_LOC_1); + } else { + timer_stopped = 0; + } + /* Update the path MTU. */ + if (net->port) { + next_mtu -= sizeof(struct udphdr); + } + if (net->mtu > next_mtu) { + net->mtu = next_mtu; +#if defined(__FreeBSD__) + if (net->port) { + sctp_hc_set_mtu(&net->ro._l_addr, inp->fibnum, next_mtu + sizeof(struct udphdr)); + } else { + sctp_hc_set_mtu(&net->ro._l_addr, inp->fibnum, next_mtu); + } +#endif + } + /* Update the association MTU */ + if (stcb->asoc.smallest_mtu > next_mtu) { + sctp_pathmtu_adjustment(stcb, next_mtu, true); + } + /* Finally, start the PMTU timer if it was running before. */ + if (timer_stopped) { + sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net); + } + SCTP_TCB_UNLOCK(stcb); + break; + default: + SCTP_TCB_UNLOCK(stcb); + break; + } +} + +#if defined(__FreeBSD__) && !defined(__Userspace__) +void +sctp6_ctlinput(struct ip6ctlparam *ip6cp) +{ + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; + struct sctp_nets *net; + struct sctphdr sh; + struct sockaddr_in6 src, dst; + + if (icmp6_errmap(ip6cp->ip6c_icmp6) == 0) { + return; + } + + /* + * Check if we can safely examine the ports and the + * verification tag of the SCTP common header. + */ + if (ip6cp->ip6c_m->m_pkthdr.len < + (int32_t)(ip6cp->ip6c_off + offsetof(struct sctphdr, checksum))) { + return; + } + + /* Copy out the port numbers and the verification tag. */ + memset(&sh, 0, sizeof(sh)); + m_copydata(ip6cp->ip6c_m, + ip6cp->ip6c_off, + sizeof(uint16_t) + sizeof(uint16_t) + sizeof(uint32_t), + (caddr_t)&sh); + memset(&src, 0, sizeof(struct sockaddr_in6)); + src.sin6_family = AF_INET6; + src.sin6_len = sizeof(struct sockaddr_in6); + src.sin6_port = sh.src_port; + src.sin6_addr = ip6cp->ip6c_ip6->ip6_src; + if (in6_setscope(&src.sin6_addr, ip6cp->ip6c_m->m_pkthdr.rcvif, NULL) != 0) { + return; + } + memset(&dst, 0, sizeof(struct sockaddr_in6)); + dst.sin6_family = AF_INET6; + dst.sin6_len = sizeof(struct sockaddr_in6); + dst.sin6_port = sh.dest_port; + dst.sin6_addr = ip6cp->ip6c_ip6->ip6_dst; + if (in6_setscope(&dst.sin6_addr, ip6cp->ip6c_m->m_pkthdr.rcvif, NULL) != 0) { + return; + } + inp = NULL; + net = NULL; + stcb = sctp_findassociation_addr_sa((struct sockaddr *)&dst, + (struct sockaddr *)&src, + &inp, &net, 1, SCTP_DEFAULT_VRFID); + if ((stcb != NULL) && + (net != NULL) && + (inp != NULL)) { + /* Check the verification tag */ + if (ntohl(sh.v_tag) != 0) { + /* + * This must be the verification tag used for + * sending out packets. We don't consider + * packets reflecting the verification tag. + */ + if (ntohl(sh.v_tag) != stcb->asoc.peer_vtag) { + SCTP_TCB_UNLOCK(stcb); + return; + } + } else { + if (ip6cp->ip6c_m->m_pkthdr.len >= + ip6cp->ip6c_off + sizeof(struct sctphdr) + + sizeof(struct sctp_chunkhdr) + + offsetof(struct sctp_init, a_rwnd)) { + /* + * In this case we can check if we + * got an INIT chunk and if the + * initiate tag matches. + */ + uint32_t initiate_tag; + uint8_t chunk_type; + + m_copydata(ip6cp->ip6c_m, + ip6cp->ip6c_off + + sizeof(struct sctphdr), + sizeof(uint8_t), + (caddr_t)&chunk_type); + m_copydata(ip6cp->ip6c_m, + ip6cp->ip6c_off + + sizeof(struct sctphdr) + + sizeof(struct sctp_chunkhdr), + sizeof(uint32_t), + (caddr_t)&initiate_tag); + if ((chunk_type != SCTP_INITIATION) || + (ntohl(initiate_tag) != stcb->asoc.my_vtag)) { + SCTP_TCB_UNLOCK(stcb); + return; + } + } else { + SCTP_TCB_UNLOCK(stcb); + return; + } + } + sctp6_notify(inp, stcb, net, + ip6cp->ip6c_icmp6->icmp6_type, + ip6cp->ip6c_icmp6->icmp6_code, + ntohl(ip6cp->ip6c_icmp6->icmp6_mtu)); + } else { + if ((stcb == NULL) && (inp != NULL)) { + /* reduce inp's ref-count */ + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + SCTP_INP_WUNLOCK(inp); + } + if (stcb) { + SCTP_TCB_UNLOCK(stcb); + } + } +} +#else +void +#if defined(__APPLE__) && !defined(APPLE_LEOPARD) && !defined(APPLE_SNOWLEOPARD) && !defined(APPLE_LION) && !defined(APPLE_MOUNTAINLION) && !defined(APPLE_ELCAPITAN) +sctp6_ctlinput(int cmd, struct sockaddr *pktdst, void *d, struct ifnet *ifp SCTP_UNUSED) +#else +sctp6_ctlinput(int cmd, struct sockaddr *pktdst, void *d) +#endif +{ + struct ip6ctlparam *ip6cp; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; + struct sctp_nets *net; + struct sctphdr sh; + struct sockaddr_in6 src, dst; + +#ifdef HAVE_SA_LEN + if (pktdst->sa_family != AF_INET6 || + pktdst->sa_len != sizeof(struct sockaddr_in6)) { +#else + if (pktdst->sa_family != AF_INET6) { +#endif + return; + } + + if ((unsigned)cmd >= PRC_NCMDS) { + return; + } + if (PRC_IS_REDIRECT(cmd)) { + d = NULL; + } else if (inet6ctlerrmap[cmd] == 0) { + return; + } + /* If the parameter is from icmp6, decode it. */ + if (d != NULL) { + ip6cp = (struct ip6ctlparam *)d; + } else { + ip6cp = (struct ip6ctlparam *)NULL; + } + + if (ip6cp != NULL) { + /* + * XXX: We assume that when IPV6 is non NULL, M and OFF are + * valid. + */ + if (ip6cp->ip6c_m == NULL) { + return; + } + + /* Check if we can safely examine the ports and the + * verification tag of the SCTP common header. + */ + if (ip6cp->ip6c_m->m_pkthdr.len < + (int32_t)(ip6cp->ip6c_off + offsetof(struct sctphdr, checksum))) { + return; + } + + /* Copy out the port numbers and the verification tag. */ + memset(&sh, 0, sizeof(sh)); + m_copydata(ip6cp->ip6c_m, + ip6cp->ip6c_off, + sizeof(uint16_t) + sizeof(uint16_t) + sizeof(uint32_t), + (caddr_t)&sh); + memset(&src, 0, sizeof(struct sockaddr_in6)); + src.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + src.sin6_len = sizeof(struct sockaddr_in6); +#endif + src.sin6_port = sh.src_port; + src.sin6_addr = ip6cp->ip6c_ip6->ip6_src; + memset(&dst, 0, sizeof(struct sockaddr_in6)); + dst.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + dst.sin6_len = sizeof(struct sockaddr_in6); +#endif + dst.sin6_port = sh.dest_port; + dst.sin6_addr = ip6cp->ip6c_ip6->ip6_dst; + inp = NULL; + net = NULL; + stcb = sctp_findassociation_addr_sa((struct sockaddr *)&dst, + (struct sockaddr *)&src, + &inp, &net, 1, SCTP_DEFAULT_VRFID); + if ((stcb != NULL) && + (net != NULL) && + (inp != NULL)) { + /* Check the verification tag */ + if (ntohl(sh.v_tag) != 0) { + /* + * This must be the verification tag used for + * sending out packets. We don't consider + * packets reflecting the verification tag. + */ + if (ntohl(sh.v_tag) != stcb->asoc.peer_vtag) { + SCTP_TCB_UNLOCK(stcb); + return; + } + } else { + SCTP_TCB_UNLOCK(stcb); + return; + } + sctp6_notify(inp, stcb, net, + ip6cp->ip6c_icmp6->icmp6_type, + ip6cp->ip6c_icmp6->icmp6_code, + ntohl(ip6cp->ip6c_icmp6->icmp6_mtu)); +#if defined(__Userspace__) + if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (stcb->sctp_socket != NULL)) { + struct socket *upcall_socket; + + upcall_socket = stcb->sctp_socket; + SOCK_LOCK(upcall_socket); + soref(upcall_socket); + SOCK_UNLOCK(upcall_socket); + if ((upcall_socket->so_upcall != NULL) && + (upcall_socket->so_error != 0)) { + (*upcall_socket->so_upcall)(upcall_socket, upcall_socket->so_upcallarg, M_NOWAIT); + } + ACCEPT_LOCK(); + SOCK_LOCK(upcall_socket); + sorele(upcall_socket); + } +#endif + } else { + if ((stcb == NULL) && (inp != NULL)) { + /* reduce inp's ref-count */ + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + SCTP_INP_WUNLOCK(inp); + } + if (stcb) { + SCTP_TCB_UNLOCK(stcb); + } + } + } +} +#endif +#endif + +/* + * this routine can probably be collasped into the one in sctp_userreq.c + * since they do the same thing and now we lookup with a sockaddr + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) +static int +sctp6_getcred(SYSCTL_HANDLER_ARGS) +{ + struct xucred xuc; + struct sockaddr_in6 addrs[2]; + struct sctp_inpcb *inp; + struct sctp_nets *net; + struct sctp_tcb *stcb; + int error; + uint32_t vrf_id; + + vrf_id = SCTP_DEFAULT_VRFID; + +#if defined(__FreeBSD__) && !defined(__Userspace__) + error = priv_check(req->td, PRIV_NETINET_GETCRED); +#else + error = suser(req->p); +#endif + if (error) + return (error); + + if (req->newlen != sizeof(addrs)) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } + if (req->oldlen != sizeof(struct ucred)) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } + error = SYSCTL_IN(req, addrs, sizeof(addrs)); + if (error) + return (error); + + stcb = sctp_findassociation_addr_sa(sin6tosa(&addrs[1]), + sin6tosa(&addrs[0]), + &inp, &net, 1, vrf_id); + if (stcb == NULL || inp == NULL || inp->sctp_socket == NULL) { + if ((inp != NULL) && (stcb == NULL)) { + /* reduce ref-count */ + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + goto cred_can_cont; + } + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ENOENT); + error = ENOENT; + goto out; + } + SCTP_TCB_UNLOCK(stcb); + /* We use the write lock here, only + * since in the error leg we need it. + * If we used RLOCK, then we would have + * to wlock/decr/unlock/rlock. Which + * in theory could create a hole. Better + * to use higher wlock. + */ + SCTP_INP_WLOCK(inp); + cred_can_cont: + error = cr_canseesocket(req->td->td_ucred, inp->sctp_socket); + if (error) { + SCTP_INP_WUNLOCK(inp); + goto out; + } + cru2x(inp->sctp_socket->so_cred, &xuc); + SCTP_INP_WUNLOCK(inp); + error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); +out: + return (error); +} + +SYSCTL_PROC(_net_inet6_sctp6, OID_AUTO, getcred, + CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_NEEDGIANT, + 0, 0, sctp6_getcred, "S,ucred", + "Get the ucred of a SCTP6 connection"); +#endif + +/* This is the same as the sctp_abort() could be made common */ +#if defined(__Userspace__) +int +#elif defined(__FreeBSD__) || defined(_WIN32) +static void +#else +static int +#endif +sctp6_abort(struct socket *so) +{ +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif + struct sctp_inpcb *inp; + uint32_t flags; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); +#if (defined(__FreeBSD__) || defined(_WIN32)) && !defined(__Userspace__) + return; +#else + return (EINVAL); +#endif + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + sctp_must_try_again: + flags = inp->sctp_flags; +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 17); +#endif + if (((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (atomic_cmpset_int(&inp->sctp_flags, flags, (flags | SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP)))) { +#ifdef SCTP_LOG_CLOSING + sctp_log_closing(inp, NULL, 16); +#endif + sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, + SCTP_CALLED_AFTER_CMPSET_OFCLOSE); + SOCK_LOCK(so); + SCTP_SB_CLEAR(so->so_snd); + /* same for the rcv ones, they are only + * here for the accounting/select. + */ + SCTP_SB_CLEAR(so->so_rcv); +#if defined(__APPLE__) && !defined(__Userspace__) + so->so_usecount--; +#else + /* Now null out the reference, we are completely detached. */ + so->so_pcb = NULL; +#endif + SOCK_UNLOCK(so); + } else { + flags = inp->sctp_flags; + if ((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) { + goto sctp_must_try_again; + } + } +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); + return; +#else + return (0); +#endif +} + +#if defined(__Userspace__) +int +sctp6_attach(struct socket *so, int proto SCTP_UNUSED, uint32_t vrf_id) +#elif defined(__FreeBSD__) +static int +sctp6_attach(struct socket *so, int proto SCTP_UNUSED, struct thread *p SCTP_UNUSED) +#elif defined(_WIN32) +static int +sctp6_attach(struct socket *so, int proto SCTP_UNUSED, PKTHREAD p SCTP_UNUSED) +#else +static int +sctp6_attach(struct socket *so, int proto SCTP_UNUSED, struct proc *p SCTP_UNUSED) +#endif +{ + int error; + struct sctp_inpcb *inp; +#if !defined(__Userspace__) + uint32_t vrf_id = SCTP_DEFAULT_VRFID; +#endif + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp != NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } + + if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { + error = SCTP_SORESERVE(so, SCTP_BASE_SYSCTL(sctp_sendspace), SCTP_BASE_SYSCTL(sctp_recvspace)); + if (error) + return (error); + } + error = sctp_inpcb_alloc(so, vrf_id); + if (error) + return (error); + inp = (struct sctp_inpcb *)so->so_pcb; + SCTP_INP_WLOCK(inp); + inp->sctp_flags |= SCTP_PCB_FLAGS_BOUND_V6; /* I'm v6! */ + + inp->ip_inp.inp.inp_vflag |= INP_IPV6; + inp->ip_inp.inp.in6p_hops = -1; /* use kernel default */ + inp->ip_inp.inp.in6p_cksum = -1; /* just to be sure */ +#ifdef INET + /* + * XXX: ugly!! IPv4 TTL initialization is necessary for an IPv6 + * socket as well, because the socket may be bound to an IPv6 + * wildcard address, which may match an IPv4-mapped IPv6 address. + */ + inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl); +#endif + SCTP_INP_WUNLOCK(inp); + return (0); +} + +#if defined(__Userspace__) +int +sctp6_bind(struct socket *so, struct sockaddr *addr, void * p) +{ +#elif defined(__FreeBSD__) +static int +sctp6_bind(struct socket *so, struct sockaddr *addr, struct thread *p) +{ +#elif defined(__APPLE__) +static int +sctp6_bind(struct socket *so, struct sockaddr *addr, struct proc *p) +{ +#elif defined(_WIN32) +static int +sctp6_bind(struct socket *so, struct sockaddr *addr, PKTHREAD p) +{ +#else +static int +sctp6_bind(struct socket *so, struct mbuf *nam, struct proc *p) +{ + struct sockaddr *addr = nam ? mtod(nam, struct sockaddr *): NULL; + +#endif + struct sctp_inpcb *inp; + int error; + u_char vflagsav; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } + +#if !(defined(_WIN32) && !defined(__Userspace__)) + if (addr) { + switch (addr->sa_family) { +#ifdef INET + case AF_INET: +#ifdef HAVE_SA_LEN + if (addr->sa_len != sizeof(struct sockaddr_in)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: +#ifdef HAVE_SA_LEN + if (addr->sa_len != sizeof(struct sockaddr_in6)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } +#endif + break; +#endif + default: + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } + } +#endif + vflagsav = inp->ip_inp.inp.inp_vflag; + inp->ip_inp.inp.inp_vflag &= ~INP_IPV4; + inp->ip_inp.inp.inp_vflag |= INP_IPV6; + if ((addr != NULL) && (SCTP_IPV6_V6ONLY(inp) == 0)) { + switch (addr->sa_family) { +#ifdef INET + case AF_INET: + /* binding v4 addr to v6 socket, so reset flags */ + inp->ip_inp.inp.inp_vflag |= INP_IPV4; + inp->ip_inp.inp.inp_vflag &= ~INP_IPV6; + break; +#endif +#ifdef INET6 + case AF_INET6: + { + struct sockaddr_in6 *sin6_p; + + sin6_p = (struct sockaddr_in6 *)addr; + + if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr)) { + inp->ip_inp.inp.inp_vflag |= INP_IPV4; + } +#ifdef INET + if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) { + struct sockaddr_in sin; + + in6_sin6_2_sin(&sin, sin6_p); + inp->ip_inp.inp.inp_vflag |= INP_IPV4; + inp->ip_inp.inp.inp_vflag &= ~INP_IPV6; + error = sctp_inpcb_bind(so, (struct sockaddr *)&sin, NULL, p); + goto out; + } +#endif + break; + } +#endif + default: + break; + } + } else if (addr != NULL) { + struct sockaddr_in6 *sin6_p; + + /* IPV6_V6ONLY socket */ +#ifdef INET + if (addr->sa_family == AF_INET) { + /* can't bind v4 addr to v6 only socket! */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + error = EINVAL; + goto out; + } +#endif + sin6_p = (struct sockaddr_in6 *)addr; + + if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) { + /* can't bind v4-mapped addrs either! */ + /* NOTE: we don't support SIIT */ + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + error = EINVAL; + goto out; + } + } + error = sctp_inpcb_bind(so, addr, NULL, p); +out: + if (error != 0) + inp->ip_inp.inp.inp_vflag = vflagsav; + return (error); +} + +#if defined(__FreeBSD__) || defined(_WIN32) || defined(__Userspace__) +#if !defined(__Userspace__) +static void +#else +void +#endif +sctp6_close(struct socket *so) +{ + sctp_close(so); +} + +/* This could be made common with sctp_detach() since they are identical */ +#else + +static +int +sctp6_detach(struct socket *so) +{ +#if defined(__Userspace__) + sctp_close(so); + return (0); +#else + return (sctp_detach(so)); +#endif +} + +#endif + +#if !defined(__Userspace__) +static +#endif +int +sctp6_disconnect(struct socket *so) +{ + return (sctp_disconnect(so)); +} + +int +#if defined(__FreeBSD__) && !defined(__Userspace__) +sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, + struct mbuf *control, struct thread *p); +#else +sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, + struct mbuf *control, struct proc *p); +#endif + +#if !defined(_WIN32) && !defined(__Userspace__) +#if defined(__FreeBSD__) +static int +sctp6_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, + struct mbuf *control, struct thread *p) +{ +#elif defined(__APPLE__) +static int +sctp6_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, + struct mbuf *control, struct proc *p) +{ +#else +static int +sctp6_send(struct socket *so, int flags, struct mbuf *m, struct mbuf *nam, + struct mbuf *control, struct proc *p) +{ + struct sockaddr *addr = nam ? mtod(nam, struct sockaddr *): NULL; +#endif + struct sctp_inpcb *inp; + +#ifdef INET + struct sockaddr_in6 *sin6; +#endif /* INET */ + /* No SPL needed since sctp_output does this */ + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + if (control) { + SCTP_RELEASE_PKT(control); + control = NULL; + } + SCTP_RELEASE_PKT(m); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } + /* + * For the TCP model we may get a NULL addr, if we are a connected + * socket thats ok. + */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) && + (addr == NULL)) { + goto connected_type; + } + if (addr == NULL) { + SCTP_RELEASE_PKT(m); + if (control) { + SCTP_RELEASE_PKT(control); + control = NULL; + } + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EDESTADDRREQ); + return (EDESTADDRREQ); + } + switch (addr->sa_family) { +#ifdef INET + case AF_INET: +#if defined(HAVE_SA_LEN) + if (addr->sa_len != sizeof(struct sockaddr_in)) { + if (control) { + SCTP_RELEASE_PKT(control); + control = NULL; + } + SCTP_RELEASE_PKT(m); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: +#if defined(HAVE_SA_LEN) + if (addr->sa_len != sizeof(struct sockaddr_in6)) { + if (control) { + SCTP_RELEASE_PKT(control); + control = NULL; + } + SCTP_RELEASE_PKT(m); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } +#endif + break; +#endif + default: + if (control) { + SCTP_RELEASE_PKT(control); + control = NULL; + } + SCTP_RELEASE_PKT(m); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } +#ifdef INET + sin6 = (struct sockaddr_in6 *)addr; + if (SCTP_IPV6_V6ONLY(inp)) { + /* + * if IPV6_V6ONLY flag, we discard datagrams destined to a + * v4 addr or v4-mapped addr + */ + if (addr->sa_family == AF_INET) { + if (control) { + SCTP_RELEASE_PKT(control); + control = NULL; + } + SCTP_RELEASE_PKT(m); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + if (control) { + SCTP_RELEASE_PKT(control); + control = NULL; + } + SCTP_RELEASE_PKT(m); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } + } + if ((addr->sa_family == AF_INET6) && + IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + struct sockaddr_in sin; + + /* convert v4-mapped into v4 addr and send */ + in6_sin6_2_sin(&sin, sin6); + return (sctp_sendm(so, flags, m, (struct sockaddr *)&sin, control, p)); + } +#endif /* INET */ +connected_type: + /* now what about control */ + if (control) { + if (inp->control) { + SCTP_PRINTF("huh? control set?\n"); + SCTP_RELEASE_PKT(inp->control); + inp->control = NULL; + } + inp->control = control; + } + /* Place the data */ + if (inp->pkt) { + SCTP_BUF_NEXT(inp->pkt_last) = m; + inp->pkt_last = m; + } else { + inp->pkt_last = inp->pkt = m; + } + if ( +#if (defined(__FreeBSD__) || defined(__APPLE__)) && !defined(__Userspace__) + /* FreeBSD and MacOSX uses a flag passed */ + ((flags & PRUS_MORETOCOME) == 0) +#else + 1 /* Open BSD does not have any "more to come" + * indication */ +#endif + ) { + /* + * note with the current version this code will only be used + * by OpenBSD, NetBSD and FreeBSD have methods for + * re-defining sosend() to use sctp_sosend(). One can + * optionaly switch back to this code (by changing back the + * defininitions but this is not advisable. + */ +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif + int ret; + +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + ret = sctp_output(inp, inp->pkt, addr, inp->control, p, flags); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + inp->pkt = NULL; + inp->control = NULL; + return (ret); + } else { + return (0); + } +} +#endif + +#if defined(__Userspace__) +int +sctp6_connect(struct socket *so, struct sockaddr *addr) +{ + void *p = NULL; +#elif defined(__FreeBSD__) +static int +sctp6_connect(struct socket *so, struct sockaddr *addr, struct thread *p) +{ +#elif defined(__APPLE__) +static int +sctp6_connect(struct socket *so, struct sockaddr *addr, struct proc *p) +{ +#elif defined(_WIN32) +static int +sctp6_connect(struct socket *so, struct sockaddr *addr, PKTHREAD p) +{ +#else +static int +sctp6_connect(struct socket *so, struct mbuf *nam, struct proc *p) +{ + struct sockaddr *addr = mtod(nam, struct sockaddr *); +#endif +#if defined(__FreeBSD__) && !defined(__Userspace__) + struct epoch_tracker et; +#endif + uint32_t vrf_id; + int error = 0; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; +#ifdef INET + struct sockaddr_in6 *sin6; + union sctp_sockstore store; +#endif + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ECONNRESET); + return (ECONNRESET); /* I made the same as TCP since we are + * not setup? */ + } + if (addr == NULL) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } +#if !(defined(_WIN32) && !defined(__Userspace__)) + switch (addr->sa_family) { +#ifdef INET + case AF_INET: +#ifdef HAVE_SA_LEN + if (addr->sa_len != sizeof(struct sockaddr_in)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: +#ifdef HAVE_SA_LEN + if (addr->sa_len != sizeof(struct sockaddr_in6)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } +#endif + break; +#endif + default: + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } +#endif + + vrf_id = inp->def_vrf_id; + SCTP_ASOC_CREATE_LOCK(inp); + SCTP_INP_RLOCK(inp); + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == + SCTP_PCB_FLAGS_UNBOUND) { + /* Bind a ephemeral port */ + SCTP_INP_RUNLOCK(inp); + error = sctp6_bind(so, NULL, p); + if (error) { + SCTP_ASOC_CREATE_UNLOCK(inp); + + return (error); + } + SCTP_INP_RLOCK(inp); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && + (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) { + /* We are already connected AND the TCP model */ + SCTP_INP_RUNLOCK(inp); + SCTP_ASOC_CREATE_UNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EADDRINUSE); + return (EADDRINUSE); + } +#ifdef INET + sin6 = (struct sockaddr_in6 *)addr; + if (SCTP_IPV6_V6ONLY(inp)) { + /* + * if IPV6_V6ONLY flag, ignore connections destined to a v4 + * addr or v4-mapped addr + */ + if (addr->sa_family == AF_INET) { + SCTP_INP_RUNLOCK(inp); + SCTP_ASOC_CREATE_UNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + SCTP_INP_RUNLOCK(inp); + SCTP_ASOC_CREATE_UNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } + } + if ((addr->sa_family == AF_INET6) && + IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + /* convert v4-mapped into v4 addr */ + in6_sin6_2_sin(&store.sin, sin6); + addr = &store.sa; + } +#endif /* INET */ + /* Now do we connect? */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb) { + SCTP_TCB_LOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + } else { + SCTP_INP_RUNLOCK(inp); + SCTP_INP_WLOCK(inp); + SCTP_INP_INCR_REF(inp); + SCTP_INP_WUNLOCK(inp); + stcb = sctp_findassociation_ep_addr(&inp, addr, NULL, NULL, NULL); + if (stcb == NULL) { + SCTP_INP_WLOCK(inp); + SCTP_INP_DECR_REF(inp); + SCTP_INP_WUNLOCK(inp); + } + } + + if (stcb != NULL) { + /* Already have or am bring up an association */ + SCTP_ASOC_CREATE_UNLOCK(inp); + SCTP_TCB_UNLOCK(stcb); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EALREADY); + return (EALREADY); + } + /* We are GOOD to go */ + stcb = sctp_aloc_assoc_connected(inp, addr, &error, 0, 0, vrf_id, + inp->sctp_ep.pre_open_stream_count, + inp->sctp_ep.port, p, + SCTP_INITIALIZE_AUTH_PARAMS); + SCTP_ASOC_CREATE_UNLOCK(inp); + if (stcb == NULL) { + /* Gak! no memory */ + return (error); + } + SCTP_SET_STATE(stcb, SCTP_STATE_COOKIE_WAIT); + (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_ENTER(et); +#endif + sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); + SCTP_TCB_UNLOCK(stcb); +#if defined(__FreeBSD__) && !defined(__Userspace__) + NET_EPOCH_EXIT(et); +#endif + return (error); +} + +static int +#if !defined(__Userspace__) +sctp6_getaddr(struct socket *so, struct sockaddr **addr) +{ + struct sockaddr_in6 *sin6; +#else +sctp6_getaddr(struct socket *so, struct mbuf *nam) +{ + struct sockaddr_in6 *sin6 = mtod(nam, struct sockaddr_in6 *); +#endif + struct sctp_inpcb *inp; + uint32_t vrf_id; + struct sctp_ifa *sctp_ifa; + +#if defined(SCTP_KAME) && defined(SCTP_EMBEDDED_V6_SCOPE) + int error; +#endif + + /* + * Do the malloc first in case it blocks. + */ +#if !defined(__Userspace__) + SCTP_MALLOC_SONAME(sin6, struct sockaddr_in6 *, sizeof(*sin6)); + if (sin6 == NULL) + return (ENOMEM); +#else + SCTP_BUF_LEN(nam) = sizeof(*sin6); + memset(sin6, 0, sizeof(*sin6)); +#endif + sin6->sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6->sin6_len = sizeof(*sin6); +#endif + + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp == NULL) { +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin6); +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ECONNRESET); + return (ECONNRESET); + } + SCTP_INP_RLOCK(inp); + sin6->sin6_port = inp->sctp_lport; + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { + /* For the bound all case you get back 0 */ + if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { + struct sctp_tcb *stcb; + struct sockaddr_in6 *sin_a6; + struct sctp_nets *net; + int fnd; + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb == NULL) { + SCTP_INP_RUNLOCK(inp); +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin6); +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ENOENT); + return (ENOENT); + } + fnd = 0; + sin_a6 = NULL; + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + sin_a6 = (struct sockaddr_in6 *)&net->ro._l_addr; + if (sin_a6 == NULL) + /* this will make coverity happy */ + continue; + + if (sin_a6->sin6_family == AF_INET6) { + fnd = 1; + break; + } + } + if ((!fnd) || (sin_a6 == NULL)) { + /* punt */ + SCTP_INP_RUNLOCK(inp); +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin6); +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ENOENT); + return (ENOENT); + } + vrf_id = inp->def_vrf_id; + sctp_ifa = sctp_source_address_selection(inp, stcb, (sctp_route_t *)&net->ro, net, 0, vrf_id); + if (sctp_ifa) { + sin6->sin6_addr = sctp_ifa->address.sin6.sin6_addr; + } + } else { + /* For the bound all case you get back 0 */ + memset(&sin6->sin6_addr, 0, sizeof(sin6->sin6_addr)); + } + } else { + /* Take the first IPv6 address in the list */ + struct sctp_laddr *laddr; + int fnd = 0; + + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + if (laddr->ifa->address.sa.sa_family == AF_INET6) { + struct sockaddr_in6 *sin_a; + + sin_a = &laddr->ifa->address.sin6; + sin6->sin6_addr = sin_a->sin6_addr; + fnd = 1; + break; + } + } + if (!fnd) { +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin6); +#endif + SCTP_INP_RUNLOCK(inp); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ENOENT); + return (ENOENT); + } + } + SCTP_INP_RUNLOCK(inp); + /* Scoping things for v6 */ +#ifdef SCTP_EMBEDDED_V6_SCOPE +#ifdef SCTP_KAME + if ((error = sa6_recoverscope(sin6)) != 0) { + SCTP_FREE_SONAME(sin6); + return (error); + } +#else + if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) + /* skip ifp check below */ + in6_recoverscope(sin6, &sin6->sin6_addr, NULL); + else + sin6->sin6_scope_id = 0; /* XXX */ +#endif /* SCTP_KAME */ +#endif /* SCTP_EMBEDDED_V6_SCOPE */ +#if !defined(__Userspace__) + (*addr) = (struct sockaddr *)sin6; +#endif + return (0); +} + +static int +#if !defined(__Userspace__) +sctp6_peeraddr(struct socket *so, struct sockaddr **addr) +{ + struct sockaddr_in6 *sin6; +#else +sctp6_peeraddr(struct socket *so, struct mbuf *nam) +{ + struct sockaddr_in6 *sin6 = mtod(nam, struct sockaddr_in6 *); +#endif + int fnd; + struct sockaddr_in6 *sin_a6; + struct sctp_inpcb *inp; + struct sctp_tcb *stcb; + struct sctp_nets *net; +#ifdef SCTP_KAME + int error; +#endif + + /* Do the malloc first in case it blocks. */ +#if !defined(__Userspace__) + SCTP_MALLOC_SONAME(sin6, struct sockaddr_in6 *, sizeof *sin6); + if (sin6 == NULL) + return (ENOMEM); +#else + SCTP_BUF_LEN(nam) = sizeof(*sin6); + memset(sin6, 0, sizeof(*sin6)); +#endif + sin6->sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + sin6->sin6_len = sizeof(*sin6); +#endif + + inp = (struct sctp_inpcb *)so->so_pcb; + if ((inp == NULL) || + ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0)) { + /* UDP type and listeners will drop out here */ +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin6); +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ENOTCONN); + return (ENOTCONN); + } + SCTP_INP_RLOCK(inp); + stcb = LIST_FIRST(&inp->sctp_asoc_list); + if (stcb) { + SCTP_TCB_LOCK(stcb); + } + SCTP_INP_RUNLOCK(inp); + if (stcb == NULL) { +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin6); +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ECONNRESET); + return (ECONNRESET); + } + fnd = 0; + TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { + sin_a6 = (struct sockaddr_in6 *)&net->ro._l_addr; + if (sin_a6->sin6_family == AF_INET6) { + fnd = 1; + sin6->sin6_port = stcb->rport; + sin6->sin6_addr = sin_a6->sin6_addr; + break; + } + } + SCTP_TCB_UNLOCK(stcb); + if (!fnd) { + /* No IPv4 address */ +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin6); +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ENOENT); + return (ENOENT); + } +#ifdef SCTP_EMBEDDED_V6_SCOPE +#ifdef SCTP_KAME + if ((error = sa6_recoverscope(sin6)) != 0) { +#if !defined(__Userspace__) + SCTP_FREE_SONAME(sin6); +#endif + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, error); + return (error); + } +#else + in6_recoverscope(sin6, &sin6->sin6_addr, NULL); +#endif /* SCTP_KAME */ +#endif /* SCTP_EMBEDDED_V6_SCOPE */ +#if !defined(__Userspace__) + *addr = (struct sockaddr *)sin6; +#endif + return (0); +} + +#if !defined(__Userspace__) +static int +sctp6_in6getaddr(struct socket *so, struct sockaddr **nam) +{ +#elif defined(__Userspace__) +int +sctp6_in6getaddr(struct socket *so, struct mbuf *nam) +{ +#ifdef INET + struct sockaddr *addr = mtod(nam, struct sockaddr *); +#endif +#else +static int +sctp6_in6getaddr(struct socket *so, struct mbuf *nam) +{ +#ifdef INET + struct sockaddr *addr = mtod(nam, struct sockaddr *); +#endif +#endif + struct inpcb *inp = sotoinpcb(so); + int error; + + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } + + /* allow v6 addresses precedence */ + error = sctp6_getaddr(so, nam); +#ifdef INET + if (error) { +#if !defined(__Userspace__) + struct sockaddr_in6 *sin6; +#else + struct sockaddr_in6 sin6; +#endif + + /* try v4 next if v6 failed */ + error = sctp_ingetaddr(so, nam); + if (error) { + return (error); + } +#if !defined(__Userspace__) + SCTP_MALLOC_SONAME(sin6, struct sockaddr_in6 *, sizeof *sin6); + if (sin6 == NULL) { + SCTP_FREE_SONAME(*nam); + return (ENOMEM); + } + in6_sin_2_v4mapsin6((struct sockaddr_in *)*nam, sin6); + SCTP_FREE_SONAME(*nam); + *nam = (struct sockaddr *)sin6; +#else + in6_sin_2_v4mapsin6((struct sockaddr_in *)addr, &sin6); + SCTP_BUF_LEN(nam) = sizeof(struct sockaddr_in6); + memcpy(addr, &sin6, sizeof(struct sockaddr_in6)); +#endif + } +#endif + return (error); +} + +#if !defined(__Userspace__) +static int +sctp6_getpeeraddr(struct socket *so, struct sockaddr **nam) +{ +#elif defined(__Userspace__) +int +sctp6_getpeeraddr(struct socket *so, struct mbuf *nam) +{ +#ifdef INET + struct sockaddr *addr = mtod(nam, struct sockaddr *); +#endif +#else +static +int +sctp6_getpeeraddr(struct socket *so, struct mbuf *nam) +{ +#ifdef INET + struct sockaddr *addr = mtod(nam, struct sockaddr *); +#endif + +#endif + struct inpcb *inp = sotoinpcb(so); + int error; + + if (inp == NULL) { + SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL); + return (EINVAL); + } + + /* allow v6 addresses precedence */ + error = sctp6_peeraddr(so, nam); +#ifdef INET + if (error) { +#if !defined(__Userspace__) + struct sockaddr_in6 *sin6; +#else + struct sockaddr_in6 sin6; +#endif + + /* try v4 next if v6 failed */ + error = sctp_peeraddr(so, nam); + if (error) { + return (error); + } +#if !defined(__Userspace__) + SCTP_MALLOC_SONAME(sin6, struct sockaddr_in6 *, sizeof *sin6); + if (sin6 == NULL) { + SCTP_FREE_SONAME(*nam); + return (ENOMEM); + } + in6_sin_2_v4mapsin6((struct sockaddr_in *)*nam, sin6); + SCTP_FREE_SONAME(*nam); + *nam = (struct sockaddr *)sin6; +#else + in6_sin_2_v4mapsin6((struct sockaddr_in *)addr, &sin6); + SCTP_BUF_LEN(nam) = sizeof(struct sockaddr_in6); + memcpy(addr, &sin6, sizeof(struct sockaddr_in6)); +#endif + } +#endif + return (error); +} + +#if !defined(__Userspace__) +#if defined(__FreeBSD__) +#define SCTP6_PROTOSW \ + .pr_protocol = IPPROTO_SCTP, \ + .pr_ctloutput = sctp_ctloutput, \ + .pr_abort = sctp6_abort, \ + .pr_accept = sctp_accept, \ + .pr_attach = sctp6_attach, \ + .pr_bind = sctp6_bind, \ + .pr_connect = sctp6_connect, \ + .pr_control = in6_control, \ + .pr_close = sctp6_close, \ + .pr_detach = sctp6_close, \ + .pr_sopoll = sopoll_generic, \ + .pr_flush = sctp_flush, \ + .pr_disconnect = sctp6_disconnect, \ + .pr_listen = sctp_listen, \ + .pr_peeraddr = sctp6_getpeeraddr, \ + .pr_send = sctp6_send, \ + .pr_shutdown = sctp_shutdown, \ + .pr_sockaddr = sctp6_in6getaddr, \ + .pr_sosend = sctp_sosend, \ + .pr_soreceive = sctp_soreceive + +struct protosw sctp6_seqpacket_protosw = { + .pr_type = SOCK_SEQPACKET, + .pr_flags = PR_WANTRCVD, + SCTP6_PROTOSW +}; + +struct protosw sctp6_stream_protosw = { + .pr_type = SOCK_STREAM, + .pr_flags = PR_CONNREQUIRED | PR_WANTRCVD, + SCTP6_PROTOSW +}; +#else +struct pr_usrreqs sctp6_usrreqs = { +#if defined(__APPLE__) && !defined(__Userspace__) + .pru_abort = sctp6_abort, + .pru_accept = sctp_accept, + .pru_attach = sctp6_attach, + .pru_bind = sctp6_bind, + .pru_connect = sctp6_connect, + .pru_connect2 = pru_connect2_notsupp, + .pru_control = in6_control, + .pru_detach = sctp6_detach, + .pru_disconnect = sctp6_disconnect, + .pru_listen = sctp_listen, + .pru_peeraddr = sctp6_getpeeraddr, + .pru_rcvd = NULL, + .pru_rcvoob = pru_rcvoob_notsupp, + .pru_send = sctp6_send, + .pru_sense = pru_sense_null, + .pru_shutdown = sctp_shutdown, + .pru_sockaddr = sctp6_in6getaddr, + .pru_sosend = sctp_sosend, + .pru_soreceive = sctp_soreceive, + .pru_sopoll = sopoll +#elif defined(_WIN32) && !defined(__Userspace__) + sctp6_abort, + sctp_accept, + sctp6_attach, + sctp6_bind, + sctp6_connect, + pru_connect2_notsupp, + NULL, + NULL, + sctp6_disconnect, + sctp_listen, + sctp6_getpeeraddr, + NULL, + pru_rcvoob_notsupp, + NULL, + pru_sense_null, + sctp_shutdown, + sctp_flush, + sctp6_in6getaddr, + sctp_sosend, + sctp_soreceive, + sopoll_generic, + NULL, + sctp6_close +}; +#endif +#endif +#endif +#endif diff --git a/netwerk/sctp/src/netinet6/sctp6_var.h b/netwerk/sctp/src/netinet6/sctp6_var.h new file mode 100644 index 0000000000..e4d737b211 --- /dev/null +++ b/netwerk/sctp/src/netinet6/sctp6_var.h @@ -0,0 +1,87 @@ +/*- + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. + * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. + * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * a) Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * b) Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the distribution. + * + * c) Neither the name of Cisco Systems, Inc. nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, + * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF + * THE POSSIBILITY OF SUCH DAMAGE. + */ + +#if defined(__FreeBSD__) && !defined(__Userspace__) +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); +#endif + +#ifndef _NETINET6_SCTP6_VAR_H_ +#define _NETINET6_SCTP6_VAR_H_ + +#if defined(__Userspace__) +#ifdef INET +extern void in6_sin6_2_sin(struct sockaddr_in *, struct sockaddr_in6 *); +extern void in6_sin6_2_sin_in_sock(struct sockaddr *); +extern void in6_sin_2_v4mapsin6(const struct sockaddr_in *, struct sockaddr_in6 *); +#endif +#endif +#if defined(_KERNEL) + +#if !defined(__Userspace__) +SYSCTL_DECL(_net_inet6_sctp6); +#if defined(__FreeBSD__) +extern struct protosw sctp6_seqpacket_protosw, sctp6_stream_protosw; +#else +extern struct pr_usrreqs sctp6_usrreqs; +#endif +#else +int sctp6_usrreq(struct socket *, int, struct mbuf *, struct mbuf *, struct mbuf *); +#endif + +#if defined(__APPLE__) && !defined(__Userspace__) +int sctp6_input(struct mbuf **, int *); +int sctp6_input_with_port(struct mbuf **, int *, uint16_t); +#else +int sctp6_input(struct mbuf **, int *, int); +int sctp6_input_with_port(struct mbuf **, int *, uint16_t); +#endif +int sctp6_output(struct sctp_inpcb *, struct mbuf *, struct sockaddr *, + struct mbuf *, struct proc *); +#if defined(__APPLE__) && !defined(__Userspace__) && !defined(APPLE_LEOPARD) && !defined(APPLE_SNOWLEOPARD) && !defined(APPLE_LION) && !defined(APPLE_MOUNTAINLION) && !defined(APPLE_ELCAPITAN) +void sctp6_ctlinput(int, struct sockaddr *, void *, struct ifnet * SCTP_UNUSED); +#elif defined(__FreeBSD__) && !defined(__Userspace__) +ip6proto_ctlinput_t sctp6_ctlinput; +#else +void sctp6_ctlinput(int, struct sockaddr_in6 *, ip6ctlparam *); +#endif +#if !((defined(__FreeBSD__) || defined(__APPLE__)) && !defined(__Userspace__)) +extern void in6_sin_2_v4mapsin6(struct sockaddr_in *, struct sockaddr_in6 *); +extern void in6_sin6_2_sin(struct sockaddr_in *, struct sockaddr_in6 *); +extern void in6_sin6_2_sin_in_sock(struct sockaddr *); +#endif +void sctp6_notify(struct sctp_inpcb *, struct sctp_tcb *, struct sctp_nets *, + uint8_t, uint8_t, uint32_t); +#endif +#endif diff --git a/netwerk/sctp/src/restore_mod.sh b/netwerk/sctp/src/restore_mod.sh new file mode 100755 index 0000000000..8305e478ed --- /dev/null +++ b/netwerk/sctp/src/restore_mod.sh @@ -0,0 +1,2 @@ +#!/bin/sh +find -name '*.[ch]' -exec chmod 644 {} \; diff --git a/netwerk/sctp/src/user_atomic.h b/netwerk/sctp/src/user_atomic.h new file mode 100644 index 0000000000..6a59587efc --- /dev/null +++ b/netwerk/sctp/src/user_atomic.h @@ -0,0 +1,315 @@ +/*- + * Copyright (c) 2009-2010 Brad Penoff + * Copyright (c) 2009-2010 Humaira Kamal + * Copyright (c) 2011-2012 Irene Ruengeler + * Copyright (c) 2011-2012 Michael Tuexen + * + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +#ifndef _USER_ATOMIC_H_ +#define _USER_ATOMIC_H_ + +/* __Userspace__ version of sys/i386/include/atomic.h goes here */ + +/* TODO In the future, might want to not use i386 specific assembly. + * The options include: + * - implement them generically (but maybe not truly atomic?) in userspace + * - have ifdef's for __Userspace_arch_ perhaps (OS isn't enough...) + */ + +#include <stdio.h> +#include <sys/types.h> + +#if defined(__APPLE__) || defined(_WIN32) +#if defined(_WIN32) +#define atomic_add_int(addr, val) InterlockedExchangeAdd((LPLONG)addr, (LONG)val) +#define atomic_fetchadd_int(addr, val) InterlockedExchangeAdd((LPLONG)addr, (LONG)val) +#define atomic_subtract_int(addr, val) InterlockedExchangeAdd((LPLONG)addr,-((LONG)val)) +#define atomic_cmpset_int(dst, exp, src) InterlockedCompareExchange((LPLONG)dst, src, exp) +#define SCTP_DECREMENT_AND_CHECK_REFCOUNT(addr) (InterlockedExchangeAdd((LPLONG)addr, (-1L)) == 1) +#else +#include <libkern/OSAtomic.h> +#define atomic_add_int(addr, val) OSAtomicAdd32Barrier(val, (int32_t *)addr) +#define atomic_fetchadd_int(addr, val) OSAtomicAdd32Barrier(val, (int32_t *)addr) +#define atomic_subtract_int(addr, val) OSAtomicAdd32Barrier(-val, (int32_t *)addr) +#define atomic_cmpset_int(dst, exp, src) OSAtomicCompareAndSwapIntBarrier(exp, src, (int *)dst) +#define SCTP_DECREMENT_AND_CHECK_REFCOUNT(addr) (atomic_fetchadd_int(addr, -1) == 0) +#endif + +#if defined(INVARIANTS) +#define SCTP_SAVE_ATOMIC_DECREMENT(addr, val) \ +{ \ + int32_t newval; \ + newval = atomic_fetchadd_int(addr, -val); \ + if (newval < 0) { \ + panic("Counter goes negative"); \ + } \ +} +#else +#define SCTP_SAVE_ATOMIC_DECREMENT(addr, val) \ +{ \ + int32_t newval; \ + newval = atomic_fetchadd_int(addr, -val); \ + if (newval < 0) { \ + *addr = 0; \ + } \ +} +#endif +#if defined(_WIN32) +static void atomic_init(void) {} /* empty when we are not using atomic_mtx */ +#else +static inline void atomic_init(void) {} /* empty when we are not using atomic_mtx */ +#endif + +#else +/* Using gcc built-in functions for atomic memory operations + Reference: http://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html + Requires gcc version 4.1.0 + compile with -march=i486 + */ + +/*Atomically add V to *P.*/ +#define atomic_add_int(P, V) (void) __sync_fetch_and_add(P, V) + +/*Atomically subtrace V from *P.*/ +#define atomic_subtract_int(P, V) (void) __sync_fetch_and_sub(P, V) + +/* + * Atomically add the value of v to the integer pointed to by p and return + * the previous value of *p. + */ +#define atomic_fetchadd_int(p, v) __sync_fetch_and_add(p, v) + +/* Following explanation from src/sys/i386/include/atomic.h, + * for atomic compare and set + * + * if (*dst == exp) *dst = src (all 32 bit words) + * + * Returns 0 on failure, non-zero on success + */ + +#define atomic_cmpset_int(dst, exp, src) __sync_bool_compare_and_swap(dst, exp, src) + +#define SCTP_DECREMENT_AND_CHECK_REFCOUNT(addr) (atomic_fetchadd_int(addr, -1) == 1) +#if defined(INVARIANTS) +#define SCTP_SAVE_ATOMIC_DECREMENT(addr, val) \ +{ \ + int32_t oldval; \ + oldval = atomic_fetchadd_int(addr, -val); \ + if (oldval < val) { \ + panic("Counter goes negative"); \ + } \ +} +#else +#define SCTP_SAVE_ATOMIC_DECREMENT(addr, val) \ +{ \ + int32_t oldval; \ + oldval = atomic_fetchadd_int(addr, -val); \ + if (oldval < val) { \ + *addr = 0; \ + } \ +} +#endif +static inline void atomic_init(void) {} /* empty when we are not using atomic_mtx */ +#endif + +#if 0 /* using libatomic_ops */ +#include "user_include/atomic_ops.h" + +/*Atomically add incr to *P, and return the original value of *P.*/ +#define atomic_add_int(P, V) AO_fetch_and_add((AO_t*)P, V) + +#define atomic_subtract_int(P, V) AO_fetch_and_add((AO_t*)P, -(V)) + +/* + * Atomically add the value of v to the integer pointed to by p and return + * the previous value of *p. + */ +#define atomic_fetchadd_int(p, v) AO_fetch_and_add((AO_t*)p, v) + +/* Atomically compare *addr to old_val, and replace *addr by new_val + if the first comparison succeeds. Returns nonzero if the comparison + succeeded and *addr was updated. +*/ +/* Following Explanation from src/sys/i386/include/atomic.h, which + matches that of AO_compare_and_swap above. + * Atomic compare and set, used by the mutex functions + * + * if (*dst == exp) *dst = src (all 32 bit words) + * + * Returns 0 on failure, non-zero on success + */ + +#define atomic_cmpset_int(dst, exp, src) AO_compare_and_swap((AO_t*)dst, exp, src) + +static inline void atomic_init() {} /* empty when we are not using atomic_mtx */ +#endif /* closing #if for libatomic */ + +#if 0 /* using atomic_mtx */ + +#include <pthread.h> + +extern userland_mutex_t atomic_mtx; + +#if defined(_WIN32) +static inline void atomic_init() { + InitializeCriticalSection(&atomic_mtx); +} +static inline void atomic_destroy() { + DeleteCriticalSection(&atomic_mtx); +} +static inline void atomic_lock() { + EnterCriticalSection(&atomic_mtx); +} +static inline void atomic_unlock() { + LeaveCriticalSection(&atomic_mtx); +} +#else +static inline void atomic_init() { + pthread_mutexattr_t mutex_attr; + + pthread_mutexattr_init(&mutex_attr); +#ifdef INVARIANTS + pthread_mutexattr_settype(&mutex_attr, PTHREAD_MUTEX_ERRORCHECK); +#endif + pthread_mutex_init(&accept_mtx, &mutex_attr); + pthread_mutexattr_destroy(&mutex_attr); +} +static inline void atomic_destroy() { + (void)pthread_mutex_destroy(&atomic_mtx); +} +static inline void atomic_lock() { +#ifdef INVARIANTS + KASSERT(pthread_mutex_lock(&atomic_mtx) == 0, ("atomic_lock: atomic_mtx already locked")) +#else + (void)pthread_mutex_lock(&atomic_mtx); +#endif +} +static inline void atomic_unlock() { +#ifdef INVARIANTS + KASSERT(pthread_mutex_unlock(&atomic_mtx) == 0, ("atomic_unlock: atomic_mtx not locked")) +#else + (void)pthread_mutex_unlock(&atomic_mtx); +#endif +} +#endif +/* + * For userland, always use lock prefixes so that the binaries will run + * on both SMP and !SMP systems. + */ + +#define MPLOCKED "lock ; " + +/* + * Atomically add the value of v to the integer pointed to by p and return + * the previous value of *p. + */ +static __inline u_int +atomic_fetchadd_int(volatile void *n, u_int v) +{ + int *p = (int *) n; + atomic_lock(); + __asm __volatile( + " " MPLOCKED " " + " xaddl %0, %1 ; " + "# atomic_fetchadd_int" + : "+r" (v), /* 0 (result) */ + "=m" (*p) /* 1 */ + : "m" (*p)); /* 2 */ + atomic_unlock(); + + return (v); +} + + +#ifdef CPU_DISABLE_CMPXCHG + +static __inline int +atomic_cmpset_int(volatile u_int *dst, u_int exp, u_int src) +{ + u_char res; + + atomic_lock(); + __asm __volatile( + " pushfl ; " + " cli ; " + " cmpl %3,%4 ; " + " jne 1f ; " + " movl %2,%1 ; " + "1: " + " sete %0 ; " + " popfl ; " + "# atomic_cmpset_int" + : "=q" (res), /* 0 */ + "=m" (*dst) /* 1 */ + : "r" (src), /* 2 */ + "r" (exp), /* 3 */ + "m" (*dst) /* 4 */ + : "memory"); + atomic_unlock(); + + return (res); +} + +#else /* !CPU_DISABLE_CMPXCHG */ + +static __inline int +atomic_cmpset_int(volatile u_int *dst, u_int exp, u_int src) +{ + atomic_lock(); + u_char res; + + __asm __volatile( + " " MPLOCKED " " + " cmpxchgl %2,%1 ; " + " sete %0 ; " + "1: " + "# atomic_cmpset_int" + : "=a" (res), /* 0 */ + "=m" (*dst) /* 1 */ + : "r" (src), /* 2 */ + "a" (exp), /* 3 */ + "m" (*dst) /* 4 */ + : "memory"); + atomic_unlock(); + + return (res); +} + +#endif /* CPU_DISABLE_CMPXCHG */ + +#define atomic_add_int(P, V) do { \ + atomic_lock(); \ + (*(u_int *)(P) += (V)); \ + atomic_unlock(); \ +} while(0) +#define atomic_subtract_int(P, V) do { \ + atomic_lock(); \ + (*(u_int *)(P) -= (V)); \ + atomic_unlock(); \ +} while(0) + +#endif +#endif diff --git a/netwerk/sctp/src/user_environment.c b/netwerk/sctp/src/user_environment.c new file mode 100644 index 0000000000..3deb3ef0d4 --- /dev/null +++ b/netwerk/sctp/src/user_environment.c @@ -0,0 +1,384 @@ +/*- + * Copyright (c) 2009-2010 Brad Penoff + * Copyright (c) 2009-2010 Humaira Kamal + * Copyright (c) 2011-2012 Irene Ruengeler + * Copyright (c) 2011-2012 Michael Tuexen + * + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +/* __Userspace__ */ + +#if defined(_WIN32) +#if !defined(_CRT_RAND_S) && !defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) +#define _CRT_RAND_S +#endif +#else +#include <stdint.h> +#include <netinet/sctp_os_userspace.h> +#endif +#ifdef INVARIANTS +#include <netinet/sctp_pcb.h> +#endif +#include <user_environment.h> +#include <sys/types.h> +/* #include <sys/param.h> defines MIN */ +#if !defined(MIN) +#define MIN(arg1,arg2) ((arg1) < (arg2) ? (arg1) : (arg2)) +#endif + +#define uHZ 1000 + +/* See user_include/user_environment.h for comments about these variables */ +int maxsockets = 25600; +int hz = uHZ; +int ip_defttl = 64; +int ipport_firstauto = 49152, ipport_lastauto = 65535; +int nmbclusters = 65536; + +/* Source ip_output.c. extern'd in ip_var.h */ +u_short ip_id = 0; /*__Userspace__ TODO Should it be initialized to zero? */ + +/* used in user_include/user_atomic.h in order to make the operations + * defined there truly atomic + */ +userland_mutex_t atomic_mtx; + +/* If the entropy device is not loaded, make a token effort to + * provide _some_ kind of randomness. This should only be used + * inside other RNG's, like arc4random(9). + */ +#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) +#include <string.h> + +void +init_random(void) +{ + return; +} + +void +read_random(void *buf, size_t size) +{ + memset(buf, 'A', size); + return; +} + +void +finish_random(void) +{ + return; +} +/* This define can be used to optionally use OpenSSL's random number utility, + * which is capable of bypassing the chromium sandbox which normally would + * prevent opening files, including /dev/urandom. + */ +#elif defined(SCTP_USE_OPENSSL_RAND) +#include <openssl/rand.h> + +/* Requiring BoringSSL because it guarantees that RAND_bytes will succeed. */ +#ifndef OPENSSL_IS_BORINGSSL +#error Only BoringSSL is supported with SCTP_USE_OPENSSL_RAND. +#endif + +void +init_random(void) +{ + return; +} + +void +read_random(void *buf, size_t size) +{ + RAND_bytes((uint8_t *)buf, size); + return; +} + +void +finish_random(void) +{ + return; +} +#elif defined(__FreeBSD__) || defined(__DragonFly__) || defined(__OpenBSD__) || defined(__NetBSD__) || defined(__APPLE__) || defined(__Bitrig__) +#include <stdlib.h> + +void +init_random(void) +{ + return; +} + +void +read_random(void *buf, size_t size) +{ + arc4random_buf(buf, size); + return; +} + +void +finish_random(void) +{ + return; +} +#elif defined(_WIN32) +#include <stdlib.h> + +void +init_random(void) +{ + return; +} + +void +read_random(void *buf, size_t size) +{ + unsigned int randval; + size_t position, remaining; + + position = 0; + while (position < size) { + if (rand_s(&randval) == 0) { + remaining = MIN(size - position, sizeof(unsigned int)); + memcpy((char *)buf + position, &randval, remaining); + position += sizeof(unsigned int); + } + } + return; +} + +void +finish_random(void) +{ + return; +} +#elif (defined(__ANDROID__) && (__ANDROID_API__ < 28)) || defined(__QNX__) || defined(__EMSCRIPTEN__) +#include <fcntl.h> + +static int fd = -1; + +void +init_random(void) +{ + fd = open("/dev/urandom", O_RDONLY); + return; +} + +void +read_random(void *buf, size_t size) +{ + size_t position; + ssize_t n; + + position = 0; + while (position < size) { + n = read(fd, (char *)buf + position, size - position); + if (n > 0) { + position += n; + } + } + return; +} + +void +finish_random(void) +{ + close(fd); + return; +} +#elif defined(__ANDROID__) && (__ANDROID_API__ >= 28) +#include <sys/random.h> + +void +init_random(void) +{ + return; +} + +void +read_random(void *buf, size_t size) +{ + size_t position; + ssize_t n; + + position = 0; + while (position < size) { + n = getrandom((char *)buf + position, size - position, 0); + if (n > 0) { + position += n; + } + } + return; +} + +void +finish_random(void) +{ + return; +} +#elif defined(__linux__) +#include <fcntl.h> +#include <unistd.h> +#include <sys/syscall.h> + +#if defined(__has_feature) +#if __has_feature(memory_sanitizer) +void __msan_unpoison(void *, size_t); +#endif +#endif + +#ifdef __NR_getrandom +#if !defined(GRND_NONBLOCK) +#define GRND_NONBLOCK 1 +#endif +static int getrandom_available = 0; +#endif +static int fd = -1; + +void +init_random(void) +{ +#ifdef __NR_getrandom + char dummy; + ssize_t n = syscall(__NR_getrandom, &dummy, sizeof(dummy), GRND_NONBLOCK); + if (n > 0 || errno == EINTR || errno == EAGAIN) { + /* Either getrandom succeeded, was interrupted or is waiting for entropy; + * all of which mean the syscall is available. + */ + getrandom_available = 1; + } else { +#ifdef INVARIANTS + if (errno != ENOSYS) { + panic("getrandom syscall returned unexpected error: %d", errno); + } +#endif + /* If the syscall isn't available, fall back to /dev/urandom. */ +#endif + fd = open("/dev/urandom", O_RDONLY); +#ifdef __NR_getrandom + } +#endif + return; +} + +void +read_random(void *buf, size_t size) +{ + size_t position; + ssize_t n; + + position = 0; + while (position < size) { +#ifdef __NR_getrandom + if (getrandom_available) { + /* Using syscall directly because getrandom isn't present in glibc < 2.25. + */ + n = syscall(__NR_getrandom, (char *)buf + position, size - position, 0); + if (n > 0) { +#if defined(__has_feature) +#if __has_feature(memory_sanitizer) + /* Need to do this because MSan doesn't realize that syscall has + * initialized the output buffer. + */ + __msan_unpoison(buf + position, n); +#endif +#endif + position += n; + } else if (errno != EINTR && errno != EAGAIN) { +#ifdef INVARIANTS + panic("getrandom syscall returned unexpected error: %d", errno); +#endif + } + } else +#endif /* __NR_getrandom */ + { + n = read(fd, (char *)buf + position, size - position); + if (n > 0) { + position += n; + } + } + } + return; +} + +void +finish_random(void) +{ + if (fd != -1) { + close(fd); + } + return; +} +#elif defined(__Fuchsia__) +#include <zircon/syscalls.h> + +void +init_random(void) +{ + return; +} + +void +read_random(void *buf, size_t size) +{ + zx_cprng_draw(buf, size); + return; +} + +void +finish_random(void) +{ + return; +} +#elif defined(__native_client__) +#include <nacl/nacl_random.h> + +void +init_random(void) +{ + return; +} + +void +read_random(void *buf, size_t size) +{ + size_t position; + size_t n; + + position = 0; + while (position < size) { + if (nacl_secure_random((char *)buf + position, size - position, &n) == 0) + position += n; + } + } + return; +} + +void +finish_random(void) +{ + return; +} +#else +#error "Unknown platform. Please provide platform specific RNG." +#endif diff --git a/netwerk/sctp/src/user_environment.h b/netwerk/sctp/src/user_environment.h new file mode 100644 index 0000000000..930d085637 --- /dev/null +++ b/netwerk/sctp/src/user_environment.h @@ -0,0 +1,132 @@ +/*- + * Copyright (c) 2009-2010 Brad Penoff + * Copyright (c) 2009-2010 Humaira Kamal + * Copyright (c) 2011-2012 Irene Ruengeler + * Copyright (c) 2011-2012 Michael Tuexen + * + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +#ifndef _USER_ENVIRONMENT_H_ +#define _USER_ENVIRONMENT_H_ + +#if defined(_WIN32) +// Needed for unified build so that rand_s is available to all unified +// sources. +#if !defined(_CRT_RAND_S) && !defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) +#define _CRT_RAND_S +#endif +#endif + +/* __Userspace__ */ +#include <sys/types.h> + +#ifdef __FreeBSD__ +#ifndef _SYS_MUTEX_H_ +#include <sys/mutex.h> +#endif +#endif +#if defined(_WIN32) +#include "netinet/sctp_os_userspace.h" +#endif + +/* maxsockets is used in SCTP_ZONE_INIT call. It refers to + * kern.ipc.maxsockets kernel environment variable. + */ +extern int maxsockets; + +/* int hz; is declared in sys/kern/subr_param.c and refers to kernel timer frequency. + * See http://ivoras.sharanet.org/freebsd/vmware.html for additional info about kern.hz + * hz is initialized in void init_param1(void) in that file. + */ +extern int hz; + + +/* The following two ints define a range of available ephemeral ports. */ +extern int ipport_firstauto, ipport_lastauto; + +/* nmbclusters is used in sctp_usrreq.c (e.g., sctp_init). In the FreeBSD kernel, + * this is 1024 + maxusers * 64. + */ +extern int nmbclusters; + +#if !defined(_MSC_VER) && !defined(__MINGW32__) +#define min(a,b) (((a)>(b))?(b):(a)) +#define max(a,b) (((a)>(b))?(a):(b)) +#endif + +void init_random(void); +void read_random(void *, size_t); +void finish_random(void); + +/* errno's may differ per OS. errno.h now included in sctp_os_userspace.h */ +/* Source: /usr/src/sys/sys/errno.h */ +/* #define ENOSPC 28 */ /* No space left on device */ +/* #define ENOBUFS 55 */ /* No buffer space available */ +/* #define ENOMEM 12 */ /* Cannot allocate memory */ +/* #define EACCES 13 */ /* Permission denied */ +/* #define EFAULT 14 */ /* Bad address */ +/* #define EHOSTDOWN 64 */ /* Host is down */ +/* #define EHOSTUNREACH 65 */ /* No route to host */ + +/* Source ip_output.c. extern'd in ip_var.h */ +extern u_short ip_id; + +#if defined(__linux__) +#define IPV6_VERSION 0x60 +#endif + +#if defined(INVARIANTS) +#include <stdlib.h> + +#if defined(_WIN32) +static inline void __declspec(noreturn) +#else +static inline void __attribute__((__noreturn__)) +#endif +terminate_non_graceful(void) { + abort(); +} + +#define panic(...) \ + do { \ + SCTP_PRINTF("%s(): ", __func__); \ + SCTP_PRINTF(__VA_ARGS__); \ + SCTP_PRINTF("\n"); \ + terminate_non_graceful(); \ +} while (0) + +#define KASSERT(cond, args) \ + do { \ + if (!(cond)) { \ + panic args ; \ + } \ + } while (0) +#else +#define KASSERT(cond, args) +#endif + +/* necessary for sctp_pcb.c */ +extern int ip_defttl; +#endif diff --git a/netwerk/sctp/src/user_inpcb.h b/netwerk/sctp/src/user_inpcb.h new file mode 100644 index 0000000000..2e6e9334ea --- /dev/null +++ b/netwerk/sctp/src/user_inpcb.h @@ -0,0 +1,373 @@ +/*- + * Copyright (c) 1982, 1986, 1990, 1993 + * The Regents of the University of California. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + * @(#)in_pcb.h 8.1 (Berkeley) 6/10/93 + * $FreeBSD: src/sys/netinet/in_pcb.h,v 1.100.2.1 2007/12/07 05:46:08 kmacy Exp $ + */ + +#ifndef _USER_INPCB_H_ +#define _USER_INPCB_H_ + +#include <user_route.h> /* was <net/route.h> */ + +struct inpcbpolicy; + +/* + * Struct inpcb is the ommon structure pcb for the Internet Protocol + * implementation. + * + * Pointers to local and foreign host table entries, local and foreign socket + * numbers, and pointers up (to a socket structure) and down (to a + * protocol-specific control block) are stored here. + */ +LIST_HEAD(inpcbhead, inpcb); +LIST_HEAD(inpcbporthead, inpcbport); + +/* + * PCB with AF_INET6 null bind'ed laddr can receive AF_INET input packet. + * So, AF_INET6 null laddr is also used as AF_INET null laddr, by utilizing + * the following structure. + */ +struct in_addr_4in6 { + uint32_t ia46_pad32[3]; + struct in_addr ia46_addr4; +}; + +/* + * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553. in_conninfo has + * some extra padding to accomplish this. + */ +struct in_endpoints { + uint16_t ie_fport; /* foreign port */ + uint16_t ie_lport; /* local port */ + /* protocol dependent part, local and foreign addr */ + union { + /* foreign host table entry */ + struct in_addr_4in6 ie46_foreign; + struct in6_addr ie6_foreign; + } ie_dependfaddr; + union { + /* local host table entry */ + struct in_addr_4in6 ie46_local; + struct in6_addr ie6_local; + } ie_dependladdr; +#define ie_faddr ie_dependfaddr.ie46_foreign.ia46_addr4 +#define ie_laddr ie_dependladdr.ie46_local.ia46_addr4 +#define ie6_faddr ie_dependfaddr.ie6_foreign +#define ie6_laddr ie_dependladdr.ie6_local +}; + +/* + * XXX The defines for inc_* are hacks and should be changed to direct + * references. + */ +struct in_conninfo { + uint8_t inc_flags; + uint8_t inc_len; + uint16_t inc_pad; /* XXX alignment for in_endpoints */ + /* protocol dependent part */ + struct in_endpoints inc_ie; +}; +#define inc_isipv6 inc_flags /* temp compatibility */ +#define inc_fport inc_ie.ie_fport +#define inc_lport inc_ie.ie_lport +#define inc_faddr inc_ie.ie_faddr +#define inc_laddr inc_ie.ie_laddr +#define inc6_faddr inc_ie.ie6_faddr +#define inc6_laddr inc_ie.ie6_laddr + +struct icmp6_filter; + +struct inpcb { + LIST_ENTRY(inpcb) inp_hash; /* hash list */ + LIST_ENTRY(inpcb) inp_list; /* list for all PCBs of this proto */ + void *inp_ppcb; /* pointer to per-protocol pcb */ + struct inpcbinfo *inp_pcbinfo; /* PCB list info */ + struct socket *inp_socket; /* back pointer to socket */ + + uint32_t inp_flow; + int inp_flags; /* generic IP/datagram flags */ + + u_char inp_vflag; /* IP version flag (v4/v6) */ +#define INP_IPV4 0x1 +#define INP_IPV6 0x2 +#define INP_IPV6PROTO 0x4 /* opened under IPv6 protocol */ +#define INP_TIMEWAIT 0x8 /* .. probably doesn't go here */ +#define INP_ONESBCAST 0x10 /* send all-ones broadcast */ +#define INP_DROPPED 0x20 /* protocol drop flag */ +#define INP_SOCKREF 0x40 /* strong socket reference */ +#define INP_CONN 0x80 + u_char inp_ip_ttl; /* time to live proto */ + u_char inp_ip_p; /* protocol proto */ + u_char inp_ip_minttl; /* minimum TTL or drop */ + uint32_t inp_ispare1; /* connection id / queue id */ + void *inp_pspare[2]; /* rtentry / general use */ + + /* Local and foreign ports, local and foreign addr. */ + struct in_conninfo inp_inc; + + /* list for this PCB's local port */ + struct label *inp_label; /* MAC label */ + struct inpcbpolicy *inp_sp; /* for IPSEC */ + + /* Protocol-dependent part; options. */ + struct { + u_char inp4_ip_tos; /* type of service proto */ + struct mbuf *inp4_options; /* IP options */ + struct ip_moptions *inp4_moptions; /* IP multicast options */ + } inp_depend4; +#define inp_fport inp_inc.inc_fport +#define inp_lport inp_inc.inc_lport +#define inp_faddr inp_inc.inc_faddr +#define inp_laddr inp_inc.inc_laddr +#define inp_ip_tos inp_depend4.inp4_ip_tos +#define inp_options inp_depend4.inp4_options +#define inp_moptions inp_depend4.inp4_moptions + struct { + /* IP options */ + struct mbuf *inp6_options; + /* IP6 options for outgoing packets */ + struct ip6_pktopts *inp6_outputopts; + /* IP multicast options */ +#if 0 + struct ip6_moptions *inp6_moptions; +#endif + /* ICMPv6 code type filter */ + struct icmp6_filter *inp6_icmp6filt; + /* IPV6_CHECKSUM setsockopt */ + int inp6_cksum; + short inp6_hops; + } inp_depend6; + LIST_ENTRY(inpcb) inp_portlist; + struct inpcbport *inp_phd; /* head of this list */ +#define inp_zero_size offsetof(struct inpcb, inp_gencnt) + struct mtx inp_mtx; + +#define in6p_faddr inp_inc.inc6_faddr +#define in6p_laddr inp_inc.inc6_laddr +#define in6p_hops inp_depend6.inp6_hops /* default hop limit */ +#define in6p_ip6_nxt inp_ip_p +#define in6p_flowinfo inp_flow +#define in6p_vflag inp_vflag +#define in6p_options inp_depend6.inp6_options +#define in6p_outputopts inp_depend6.inp6_outputopts +#if 0 +#define in6p_moptions inp_depend6.inp6_moptions +#endif +#define in6p_icmp6filt inp_depend6.inp6_icmp6filt +#define in6p_cksum inp_depend6.inp6_cksum +#define in6p_flags inp_flags /* for KAME src sync over BSD*'s */ +#define in6p_socket inp_socket /* for KAME src sync over BSD*'s */ +#define in6p_lport inp_lport /* for KAME src sync over BSD*'s */ +#define in6p_fport inp_fport /* for KAME src sync over BSD*'s */ +#define in6p_ppcb inp_ppcb /* for KAME src sync over BSD*'s */ +}; +/* + * The range of the generation count, as used in this implementation, is 9e19. + * We would have to create 300 billion connections per second for this number + * to roll over in a year. This seems sufficiently unlikely that we simply + * don't concern ourselves with that possibility. + */ + +struct inpcbport { + LIST_ENTRY(inpcbport) phd_hash; + struct inpcbhead phd_pcblist; + u_short phd_port; +}; + +/* + * Global data structure for each high-level protocol (UDP, TCP, ...) in both + * IPv4 and IPv6. Holds inpcb lists and information for managing them. + */ +struct inpcbinfo { + /* + * Global list of inpcbs on the protocol. + */ + struct inpcbhead *ipi_listhead; + u_int ipi_count; + + /* + * Global hash of inpcbs, hashed by local and foreign addresses and + * port numbers. + */ + struct inpcbhead *ipi_hashbase; + u_long ipi_hashmask; + + /* + * Global hash of inpcbs, hashed by only local port number. + */ + struct inpcbporthead *ipi_porthashbase; + u_long ipi_porthashmask; + + /* + * Fields associated with port lookup and allocation. + */ + u_short ipi_lastport; + u_short ipi_lastlow; + u_short ipi_lasthi; + + /* + * UMA zone from which inpcbs are allocated for this protocol. + */ + struct uma_zone *ipi_zone; + + /* + * Generation count--incremented each time a connection is allocated + * or freed. + */ + struct mtx ipi_mtx; + + /* + * vimage 1 + * general use 1 + */ + void *ipi_pspare[2]; +}; + +#define INP_LOCK_INIT(inp, d, t) \ + mtx_init(&(inp)->inp_mtx, (d), (t), MTX_DEF | MTX_RECURSE | MTX_DUPOK) +#define INP_LOCK_DESTROY(inp) mtx_destroy(&(inp)->inp_mtx) +#define INP_LOCK(inp) mtx_lock(&(inp)->inp_mtx) +#define INP_UNLOCK(inp) mtx_unlock(&(inp)->inp_mtx) +#define INP_LOCK_ASSERT(inp) mtx_assert(&(inp)->inp_mtx, MA_OWNED) +#define INP_UNLOCK_ASSERT(inp) mtx_assert(&(inp)->inp_mtx, MA_NOTOWNED) + +#define INP_INFO_LOCK_INIT(ipi, d) \ + mtx_init(&(ipi)->ipi_mtx, (d), NULL, MTX_DEF | MTX_RECURSE) +#define INP_INFO_LOCK_DESTROY(ipi) mtx_destroy(&(ipi)->ipi_mtx) +#define INP_INFO_RLOCK(ipi) mtx_lock(&(ipi)->ipi_mtx) +#define INP_INFO_WLOCK(ipi) mtx_lock(&(ipi)->ipi_mtx) +#define INP_INFO_RUNLOCK(ipi) mtx_unlock(&(ipi)->ipi_mtx) +#define INP_INFO_WUNLOCK(ipi) mtx_unlock(&(ipi)->ipi_mtx) +#define INP_INFO_RLOCK_ASSERT(ipi) mtx_assert(&(ipi)->ipi_mtx, MA_OWNED) +#define INP_INFO_WLOCK_ASSERT(ipi) mtx_assert(&(ipi)->ipi_mtx, MA_OWNED) +#define INP_INFO_UNLOCK_ASSERT(ipi) mtx_assert(&(ipi)->ipi_mtx, MA_NOTOWNED) + +#define INP_PCBHASH(faddr, lport, fport, mask) \ + (((faddr) ^ ((faddr) >> 16) ^ ntohs((lport) ^ (fport))) & (mask)) +#define INP_PCBPORTHASH(lport, mask) \ + (ntohs((lport)) & (mask)) + +/* flags in inp_flags: */ +#define INP_RECVOPTS 0x01 /* receive incoming IP options */ +#define INP_RECVRETOPTS 0x02 /* receive IP options for reply */ +#define INP_RECVDSTADDR 0x04 /* receive IP dst address */ +#define INP_HDRINCL 0x08 /* user supplies entire IP header */ +#define INP_HIGHPORT 0x10 /* user wants "high" port binding */ +#define INP_LOWPORT 0x20 /* user wants "low" port binding */ +#define INP_ANONPORT 0x40 /* port chosen for user */ +#define INP_RECVIF 0x80 /* receive incoming interface */ +#define INP_MTUDISC 0x100 /* user can do MTU discovery */ +#define INP_FAITH 0x200 /* accept FAITH'ed connections */ +#define INP_RECVTTL 0x400 /* receive incoming IP TTL */ +#define INP_DONTFRAG 0x800 /* don't fragment packet */ + +#define IN6P_IPV6_V6ONLY 0x008000 /* restrict AF_INET6 socket for v6 */ + +#define IN6P_PKTINFO 0x010000 /* receive IP6 dst and I/F */ +#define IN6P_HOPLIMIT 0x020000 /* receive hoplimit */ +#define IN6P_HOPOPTS 0x040000 /* receive hop-by-hop options */ +#define IN6P_DSTOPTS 0x080000 /* receive dst options after rthdr */ +#define IN6P_RTHDR 0x100000 /* receive routing header */ +#define IN6P_RTHDRDSTOPTS 0x200000 /* receive dstoptions before rthdr */ +#define IN6P_TCLASS 0x400000 /* receive traffic class value */ +#define IN6P_AUTOFLOWLABEL 0x800000 /* attach flowlabel automatically */ +#define IN6P_RFC2292 0x40000000 /* used RFC2292 API on the socket */ +#define IN6P_MTU 0x80000000 /* receive path MTU */ + +#define INP_CONTROLOPTS (INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\ + INP_RECVIF|INP_RECVTTL|\ + IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\ + IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\ + IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\ + IN6P_MTU) +#define INP_UNMAPPABLEOPTS (IN6P_HOPOPTS|IN6P_DSTOPTS|IN6P_RTHDR|\ + IN6P_TCLASS|IN6P_AUTOFLOWLABEL) + + /* for KAME src sync over BSD*'s */ +#define IN6P_HIGHPORT INP_HIGHPORT +#define IN6P_LOWPORT INP_LOWPORT +#define IN6P_ANONPORT INP_ANONPORT +#define IN6P_RECVIF INP_RECVIF +#define IN6P_MTUDISC INP_MTUDISC +#define IN6P_FAITH INP_FAITH +#define IN6P_CONTROLOPTS INP_CONTROLOPTS + /* + * socket AF version is {newer than,or include} + * actual datagram AF version + */ + +#define INPLOOKUP_WILDCARD 1 +#define sotoinpcb(so) ((struct inpcb *)(so)->so_pcb) + +#define INP_SOCKAF(so) so->so_proto->pr_domain->dom_family + +#define INP_CHECK_SOCKAF(so, af) (INP_SOCKAF(so) == af) + +extern int ipport_reservedhigh; +extern int ipport_reservedlow; +extern int ipport_lowfirstauto; +extern int ipport_lowlastauto; +extern int ipport_firstauto; +extern int ipport_lastauto; +extern int ipport_hifirstauto; +extern int ipport_hilastauto; +extern struct callout ipport_tick_callout; + +void in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *); +int in_pcballoc(struct socket *, struct inpcbinfo *); +int in_pcbbind(struct inpcb *, struct sockaddr *, struct ucred *); +int in_pcbconnect(struct inpcb *, struct sockaddr *, struct ucred *); +void in_pcbdetach(struct inpcb *); +void in_pcbdisconnect(struct inpcb *); +void in_pcbdrop(struct inpcb *); +void in_pcbfree(struct inpcb *); +int in_pcbinshash(struct inpcb *); +struct inpcb * + in_pcblookup_local(struct inpcbinfo *, + struct in_addr, u_int, int); +struct inpcb * + in_pcblookup_hash(struct inpcbinfo *, struct in_addr, u_int, + struct in_addr, u_int, int, struct ifnet *); +void in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr, + int, struct inpcb *(*)(struct inpcb *, int)); +void in_pcbrehash(struct inpcb *); +void in_pcbsetsolabel(struct socket *so); +int in_getpeeraddr(struct socket *so, struct sockaddr **nam); +int in_getsockaddr(struct socket *so, struct sockaddr **nam); +void in_pcbsosetlabel(struct socket *so); +void in_pcbremlists(struct inpcb *inp); +void ipport_tick(void *xtp); + +/* + * Debugging routines compiled in when DDB is present. + */ +void db_print_inpcb(struct inpcb *inp, const char *name, int indent); + + +#endif /* !_NETINET_IN_PCB_H_ */ diff --git a/netwerk/sctp/src/user_ip6_var.h b/netwerk/sctp/src/user_ip6_var.h new file mode 100644 index 0000000000..b970fb8fa5 --- /dev/null +++ b/netwerk/sctp/src/user_ip6_var.h @@ -0,0 +1,124 @@ +/*- + * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the project nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + */ +/*- + * Copyright (c) 1982, 1986, 1993 + * The Regents of the University of California. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + */ + +#ifndef _USER_IP6_VAR_H_ +#define _USER_IP6_VAR_H_ + +#if defined(_WIN32) +struct ip6_hdr { + union { + struct ip6_hdrctl { + uint32_t ip6_un1_flow; /* 20 bits of flow-ID */ + uint16_t ip6_un1_plen; /* payload length */ + uint8_t ip6_un1_nxt; /* next header */ + uint8_t ip6_un1_hlim; /* hop limit */ + } ip6_un1; + uint8_t ip6_un2_vfc; /* 4 bits version, top 4 bits class */ + } ip6_ctlun; + struct in6_addr ip6_src; /* source address */ + struct in6_addr ip6_dst; /* destination address */ +}; +#define ip6_vfc ip6_ctlun.ip6_un2_vfc +#define ip6_flow ip6_ctlun.ip6_un1.ip6_un1_flow +#define ip6_plen ip6_ctlun.ip6_un1.ip6_un1_plen +#define ip6_nxt ip6_ctlun.ip6_un1.ip6_un1_nxt +#define ip6_hlim ip6_ctlun.ip6_un1.ip6_un1_hlim +#define ip6_hops ip6_ctlun.ip6_un1.ip6_un1_hlim + +#define IPV6_VERSION 0x60 +#endif + +#if defined(_WIN32) +#define s6_addr16 u.Word +#endif +#if !defined(_WIN32) && !defined(__linux__) && !defined(__EMSCRIPTEN__) +#define s6_addr8 __u6_addr.__u6_addr8 +#define s6_addr16 __u6_addr.__u6_addr16 +#define s6_addr32 __u6_addr.__u6_addr32 +#endif + +#if !defined(__FreeBSD__) && !defined(__OpenBSD__) && !defined(__DragonFly__) +struct route_in6 { + struct rtentry *ro_rt; + struct llentry *ro_lle; + struct in6_addr *ro_ia6; + int ro_flags; + struct sockaddr_in6 ro_dst; +}; +#endif +#define IP6_EXTHDR_GET(val, typ, m, off, len) \ +do { \ + struct mbuf *t; \ + int tmp; \ + if ((m)->m_len >= (off) + (len)) \ + (val) = (typ)(mtod((m), caddr_t) + (off)); \ + else { \ + t = m_pulldown((m), (off), (len), &tmp); \ + if (t) { \ + KASSERT(t->m_len >= tmp + (len), \ + ("m_pulldown malfunction")); \ + (val) = (typ)(mtod(t, caddr_t) + tmp); \ + } else { \ + (val) = (typ)NULL; \ + (m) = NULL; \ + } \ + } \ +} while (0) + +#endif /* !_USER_IP6_VAR_H_ */ diff --git a/netwerk/sctp/src/user_ip_icmp.h b/netwerk/sctp/src/user_ip_icmp.h new file mode 100644 index 0000000000..a993411a17 --- /dev/null +++ b/netwerk/sctp/src/user_ip_icmp.h @@ -0,0 +1,225 @@ +/*- + * Copyright (c) 1982, 1986, 1993 + * The Regents of the University of California. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + */ + +#ifndef _USER_IP_ICMP_H_ +#define _USER_IP_ICMP_H_ + +/* + * Interface Control Message Protocol Definitions. + * Per RFC 792, September 1981. + */ + +/* + * Internal of an ICMP Router Advertisement + */ +struct icmp_ra_addr { + uint32_t ira_addr; + uint32_t ira_preference; +}; + +/* + * Structure of an icmp header. + */ +struct icmphdr { + u_char icmp_type; /* type of message, see below */ + u_char icmp_code; /* type sub code */ + u_short icmp_cksum; /* ones complement cksum of struct */ +}; + +#if defined(_WIN32) +#pragma pack (push, 1) +struct icmp6_hdr { + uint8_t icmp6_type; + uint8_t icmp6_code; + uint16_t icmp6_cksum; + union { + uint32_t icmp6_un_data32[1]; + uint16_t icmp6_un_data16[2]; + uint8_t icmp6_un_data8[4]; + } icmp6_dataun; +}; +#pragma pack(pop) + +#define icmp6_data32 icmp6_dataun.icmp6_un_data32 +#define icmp6_mtu icmp6_data32[0] +#endif + +/* + * Structure of an icmp packet. + * + * XXX: should start with a struct icmphdr. + */ +struct icmp { + u_char icmp_type; /* type of message, see below */ + u_char icmp_code; /* type sub code */ + u_short icmp_cksum; /* ones complement cksum of struct */ + union { + u_char ih_pptr; /* ICMP_PARAMPROB */ + struct in_addr ih_gwaddr; /* ICMP_REDIRECT */ + struct ih_idseq { + uint16_t icd_id; /* network format */ + uint16_t icd_seq; /* network format */ + } ih_idseq; + int ih_void; + + /* ICMP_UNREACH_NEEDFRAG -- Path MTU Discovery (RFC1191) */ + struct ih_pmtu { + uint16_t ipm_void; /* network format */ + uint16_t ipm_nextmtu; /* network format */ + } ih_pmtu; + + struct ih_rtradv { + u_char irt_num_addrs; + u_char irt_wpa; + uint16_t irt_lifetime; + } ih_rtradv; + } icmp_hun; +#define icmp_pptr icmp_hun.ih_pptr +#define icmp_gwaddr icmp_hun.ih_gwaddr +#define icmp_id icmp_hun.ih_idseq.icd_id +#define icmp_seq icmp_hun.ih_idseq.icd_seq +#define icmp_void icmp_hun.ih_void +#define icmp_pmvoid icmp_hun.ih_pmtu.ipm_void +#define icmp_nextmtu icmp_hun.ih_pmtu.ipm_nextmtu +#define icmp_num_addrs icmp_hun.ih_rtradv.irt_num_addrs +#define icmp_wpa icmp_hun.ih_rtradv.irt_wpa +#define icmp_lifetime icmp_hun.ih_rtradv.irt_lifetime + union { + struct id_ts { /* ICMP Timestamp */ + /* + * The next 3 fields are in network format, + * milliseconds since 00:00 GMT + */ + uint32_t its_otime; /* Originate */ + uint32_t its_rtime; /* Receive */ + uint32_t its_ttime; /* Transmit */ + } id_ts; + struct id_ip { + struct ip idi_ip; + /* options and then 64 bits of data */ + } id_ip; + struct icmp_ra_addr id_radv; + uint32_t id_mask; + char id_data[1]; + } icmp_dun; +#define icmp_otime icmp_dun.id_ts.its_otime +#define icmp_rtime icmp_dun.id_ts.its_rtime +#define icmp_ttime icmp_dun.id_ts.its_ttime +#define icmp_ip icmp_dun.id_ip.idi_ip +#define icmp_radv icmp_dun.id_radv +#define icmp_mask icmp_dun.id_mask +#define icmp_data icmp_dun.id_data +}; + +/* + * Lower bounds on packet lengths for various types. + * For the error advice packets must first insure that the + * packet is large enough to contain the returned ip header. + * Only then can we do the check to see if 64 bits of packet + * data have been returned, since we need to check the returned + * ip header length. + */ +#define ICMP_MINLEN 8 /* abs minimum */ +#define ICMP_TSLEN (8 + 3 * sizeof (uint32_t)) /* timestamp */ +#define ICMP_MASKLEN 12 /* address mask */ +#define ICMP_ADVLENMIN (8 + sizeof (struct ip) + 8) /* min */ +#define ICMP_ADVLEN(p) (8 + ((p)->icmp_ip.ip_hl << 2) + 8) + /* N.B.: must separately check that ip_hl >= 5 */ + +/* + * Definition of type and code field values. + */ +#define ICMP_ECHOREPLY 0 /* echo reply */ +#define ICMP_UNREACH 3 /* dest unreachable, codes: */ +#define ICMP_UNREACH_NET 0 /* bad net */ +#define ICMP_UNREACH_HOST 1 /* bad host */ +#define ICMP_UNREACH_PROTOCOL 2 /* bad protocol */ +#define ICMP_UNREACH_PORT 3 /* bad port */ +#define ICMP_UNREACH_NEEDFRAG 4 /* IP_DF caused drop */ +#define ICMP_UNREACH_SRCFAIL 5 /* src route failed */ +#define ICMP_UNREACH_NET_UNKNOWN 6 /* unknown net */ +#define ICMP_UNREACH_HOST_UNKNOWN 7 /* unknown host */ +#define ICMP_UNREACH_ISOLATED 8 /* src host isolated */ +#define ICMP_UNREACH_NET_PROHIB 9 /* prohibited access */ +#define ICMP_UNREACH_HOST_PROHIB 10 /* ditto */ +#define ICMP_UNREACH_TOSNET 11 /* bad tos for net */ +#define ICMP_UNREACH_TOSHOST 12 /* bad tos for host */ +#define ICMP_UNREACH_FILTER_PROHIB 13 /* admin prohib */ +#define ICMP_UNREACH_HOST_PRECEDENCE 14 /* host prec vio. */ +#define ICMP_UNREACH_PRECEDENCE_CUTOFF 15 /* prec cutoff */ +#define ICMP_SOURCEQUENCH 4 /* packet lost, slow down */ +#define ICMP_REDIRECT 5 /* shorter route, codes: */ +#define ICMP_REDIRECT_NET 0 /* for network */ +#define ICMP_REDIRECT_HOST 1 /* for host */ +#define ICMP_REDIRECT_TOSNET 2 /* for tos and net */ +#define ICMP_REDIRECT_TOSHOST 3 /* for tos and host */ +#define ICMP_ALTHOSTADDR 6 /* alternate host address */ +#define ICMP_ECHO 8 /* echo service */ +#define ICMP_ROUTERADVERT 9 /* router advertisement */ +#define ICMP_ROUTERADVERT_NORMAL 0 /* normal advertisement */ +#define ICMP_ROUTERADVERT_NOROUTE_COMMON 16 /* selective routing */ +#define ICMP_ROUTERSOLICIT 10 /* router solicitation */ +#define ICMP_TIMXCEED 11 /* time exceeded, code: */ +#define ICMP_TIMXCEED_INTRANS 0 /* ttl==0 in transit */ +#define ICMP_TIMXCEED_REASS 1 /* ttl==0 in reass */ +#define ICMP_PARAMPROB 12 /* ip header bad */ +#define ICMP_PARAMPROB_ERRATPTR 0 /* error at param ptr */ +#define ICMP_PARAMPROB_OPTABSENT 1 /* req. opt. absent */ +#define ICMP_PARAMPROB_LENGTH 2 /* bad length */ +#define ICMP_TSTAMP 13 /* timestamp request */ +#define ICMP_TSTAMPREPLY 14 /* timestamp reply */ +#define ICMP_IREQ 15 /* information request */ +#define ICMP_IREQREPLY 16 /* information reply */ +#define ICMP_MASKREQ 17 /* address mask request */ +#define ICMP_MASKREPLY 18 /* address mask reply */ +#define ICMP_TRACEROUTE 30 /* traceroute */ +#define ICMP_DATACONVERR 31 /* data conversion error */ +#define ICMP_MOBILE_REDIRECT 32 /* mobile host redirect */ +#define ICMP_IPV6_WHEREAREYOU 33 /* IPv6 where-are-you */ +#define ICMP_IPV6_IAMHERE 34 /* IPv6 i-am-here */ +#define ICMP_MOBILE_REGREQUEST 35 /* mobile registration req */ +#define ICMP_MOBILE_REGREPLY 36 /* mobile registration reply */ +#define ICMP_SKIP 39 /* SKIP */ +#define ICMP_PHOTURIS 40 /* Photuris */ +#define ICMP_PHOTURIS_UNKNOWN_INDEX 1 /* unknown sec index */ +#define ICMP_PHOTURIS_AUTH_FAILED 2 /* auth failed */ +#define ICMP_PHOTURIS_DECRYPT_FAILED 3 /* decrypt failed */ + +#define ICMP_MAXTYPE 40 + +#define ICMP_INFOTYPE(type) \ + ((type) == ICMP_ECHOREPLY || (type) == ICMP_ECHO || \ + (type) == ICMP_ROUTERADVERT || (type) == ICMP_ROUTERSOLICIT || \ + (type) == ICMP_TSTAMP || (type) == ICMP_TSTAMPREPLY || \ + (type) == ICMP_IREQ || (type) == ICMP_IREQREPLY || \ + (type) == ICMP_MASKREQ || (type) == ICMP_MASKREPLY) + + +#endif diff --git a/netwerk/sctp/src/user_malloc.h b/netwerk/sctp/src/user_malloc.h new file mode 100644 index 0000000000..c588e094af --- /dev/null +++ b/netwerk/sctp/src/user_malloc.h @@ -0,0 +1,203 @@ +/*- + * Copyright (c) 1987, 1993 + * The Regents of the University of California. + * Copyright (c) 2005 Robert N. M. Watson + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + */ + +/* This file has been renamed user_malloc.h for Userspace */ +#ifndef _USER_MALLOC_H_ +#define _USER_MALLOC_H_ + +/*__Userspace__*/ +#include <stdlib.h> +#include <sys/types.h> +#if !defined(_WIN32) +#include <strings.h> +#include <stdint.h> +#else +#if (defined(_MSC_VER) && _MSC_VER >= 1600) || (defined(__MSVCRT_VERSION__) && __MSVCRT_VERSION__ >= 1400) +#include <stdint.h> +#elif defined(SCTP_STDINT_INCLUDE) +#include SCTP_STDINT_INCLUDE +#else +#define uint32_t unsigned __int32 +#define uint64_t unsigned __int64 +#endif +#include <winsock2.h> +#endif + +#define MINALLOCSIZE UMA_SMALLEST_UNIT + +/* + * flags to malloc. + */ +#define M_NOWAIT 0x0001 /* do not block */ +#define M_WAITOK 0x0002 /* ok to block */ +#define M_ZERO 0x0100 /* bzero the allocation */ +#define M_NOVM 0x0200 /* don't ask VM for pages */ +#define M_USE_RESERVE 0x0400 /* can alloc out of reserve memory */ + +#define M_MAGIC 877983977 /* time when first defined :-) */ + +/* + * Two malloc type structures are present: malloc_type, which is used by a + * type owner to declare the type, and malloc_type_internal, which holds + * malloc-owned statistics and other ABI-sensitive fields, such as the set of + * malloc statistics indexed by the compile-time MAXCPU constant. + * Applications should avoid introducing dependence on the allocator private + * data layout and size. + * + * The malloc_type ks_next field is protected by malloc_mtx. Other fields in + * malloc_type are static after initialization so unsynchronized. + * + * Statistics in malloc_type_stats are written only when holding a critical + * section and running on the CPU associated with the index into the stat + * array, but read lock-free resulting in possible (minor) races, which the + * monitoring app should take into account. + */ +struct malloc_type_stats { + uint64_t mts_memalloced; /* Bytes allocated on CPU. */ + uint64_t mts_memfreed; /* Bytes freed on CPU. */ + uint64_t mts_numallocs; /* Number of allocates on CPU. */ + uint64_t mts_numfrees; /* number of frees on CPU. */ + uint64_t mts_size; /* Bitmask of sizes allocated on CPU. */ + uint64_t _mts_reserved1; /* Reserved field. */ + uint64_t _mts_reserved2; /* Reserved field. */ + uint64_t _mts_reserved3; /* Reserved field. */ +}; + +#ifndef MAXCPU /* necessary on Linux */ +#define MAXCPU 4 /* arbitrary? */ +#endif + +struct malloc_type_internal { + struct malloc_type_stats mti_stats[MAXCPU]; +}; + +/* + * ABI-compatible version of the old 'struct malloc_type', only all stats are + * now malloc-managed in malloc-owned memory rather than in caller memory, so + * as to avoid ABI issues. The ks_next pointer is reused as a pointer to the + * internal data handle. + */ +struct malloc_type { + struct malloc_type *ks_next; /* Next in global chain. */ + u_long _ks_memuse; /* No longer used. */ + u_long _ks_size; /* No longer used. */ + u_long _ks_inuse; /* No longer used. */ + uint64_t _ks_calls; /* No longer used. */ + u_long _ks_maxused; /* No longer used. */ + u_long ks_magic; /* Detect programmer error. */ + const char *ks_shortdesc; /* Printable type name. */ + + /* + * struct malloc_type was terminated with a struct mtx, which is no + * longer required. For ABI reasons, continue to flesh out the full + * size of the old structure, but reuse the _lo_class field for our + * internal data handle. + */ + void *ks_handle; /* Priv. data, was lo_class. */ + const char *_lo_name; + const char *_lo_type; + u_int _lo_flags; + void *_lo_list_next; + struct witness *_lo_witness; + uintptr_t _mtx_lock; + u_int _mtx_recurse; +}; + +/* + * Statistics structure headers for user space. The kern.malloc sysctl + * exposes a structure stream consisting of a stream header, then a series of + * malloc type headers and statistics structures (quantity maxcpus). For + * convenience, the kernel will provide the current value of maxcpus at the + * head of the stream. + */ +#define MALLOC_TYPE_STREAM_VERSION 0x00000001 +struct malloc_type_stream_header { + uint32_t mtsh_version; /* Stream format version. */ + uint32_t mtsh_maxcpus; /* Value of MAXCPU for stream. */ + uint32_t mtsh_count; /* Number of records. */ + uint32_t _mtsh_pad; /* Pad/reserved field. */ +}; + +#define MALLOC_MAX_NAME 32 +struct malloc_type_header { + char mth_name[MALLOC_MAX_NAME]; +}; + +/* __Userspace__ +Notice that at places it uses ifdef _KERNEL. That line cannot be +removed because it causes conflicts with malloc definition in +/usr/include/malloc.h, which essentially says that malloc.h has +been overridden by stdlib.h. We will need to use names like +user_malloc.h for isolating kernel interface headers. using +original names like malloc.h in a user_include header can be +confusing, All userspace header files are being placed in ./user_include +Better still to remove from user_include.h all irrelevant code such +as that in the block starting with #ifdef _KERNEL. I am only leaving +it in for the time being to see what functionality is in this file +that kernel uses. + +Start copy: Copied code for __Userspace__ */ +#define MALLOC_DEFINE(type, shortdesc, longdesc) \ + struct malloc_type type[1] = { \ + { NULL, 0, 0, 0, 0, 0, M_MAGIC, shortdesc, NULL, NULL, \ + NULL, 0, NULL, NULL, 0, 0 } \ + } + +/* Removed "extern" in __Userspace__ code */ +/* If we need to use MALLOC_DECLARE before using MALLOC then + we have to remove extern. + In /usr/include/sys/malloc.h there is this definition: + #define MALLOC_DECLARE(type) \ + extern struct malloc_type type[1] + and loader is unable to find the extern malloc_type because + it may be defined in one of kernel object files. + It seems that MALLOC_DECLARE and MALLOC_DEFINE cannot be used at + the same time for same "type" variable. Also, in Randall's architecture + document, where it specifies O/S specific macros and functions, it says + that the name in SCTP_MALLOC does not have to be used. +*/ +#define MALLOC_DECLARE(type) \ + extern struct malloc_type type[1] + +#define FREE(addr, type) free((addr)) + +/* changed definitions of MALLOC and FREE */ +/* Using memset if flag M_ZERO is specified. Todo: M_WAITOK and M_NOWAIT */ +#define MALLOC(space, cast, size, type, flags) \ + ((space) = (cast)malloc((u_long)(size))); \ + do { \ + if (flags & M_ZERO) { \ + memset(space,0,size); \ + } \ + } while (0); + +#endif /* !_SYS_MALLOC_H_ */ diff --git a/netwerk/sctp/src/user_mbuf.c b/netwerk/sctp/src/user_mbuf.c new file mode 100644 index 0000000000..85badc0fa5 --- /dev/null +++ b/netwerk/sctp/src/user_mbuf.c @@ -0,0 +1,1589 @@ +/*- + * Copyright (c) 1982, 1986, 1988, 1993 + * The Regents of the University of California. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + */ + +/* + * __Userspace__ version of /usr/src/sys/kern/kern_mbuf.c + * We are initializing two zones for Mbufs and Clusters. + * + */ + +#include <stdio.h> +#include <string.h> +/* #include <sys/param.h> This defines MSIZE 256 */ +#if !defined(SCTP_SIMPLE_ALLOCATOR) +#include "umem.h" +#endif +#include "user_mbuf.h" +#include "user_environment.h" +#include "user_atomic.h" +#include "netinet/sctp_pcb.h" + +#define KIPC_MAX_LINKHDR 4 /* int: max length of link header (see sys/sysclt.h) */ +#define KIPC_MAX_PROTOHDR 5 /* int: max length of network header (see sys/sysclt.h)*/ +int max_linkhdr = KIPC_MAX_LINKHDR; +int max_protohdr = KIPC_MAX_PROTOHDR; /* Size of largest protocol layer header. */ + +/* + * Zones from which we allocate. + */ +sctp_zone_t zone_mbuf; +sctp_zone_t zone_clust; +sctp_zone_t zone_ext_refcnt; + +/* __Userspace__ clust_mb_args will be passed as callback data to mb_ctor_clust + * and mb_dtor_clust. + * Note: I had to use struct clust_args as an encapsulation for an mbuf pointer. + * struct mbuf * clust_mb_args; does not work. + */ +struct clust_args clust_mb_args; + + +/* __Userspace__ + * Local prototypes. + */ +static int mb_ctor_mbuf(void *, void *, int); +static int mb_ctor_clust(void *, void *, int); +static void mb_dtor_mbuf(void *, void *); +static void mb_dtor_clust(void *, void *); + + +/***************** Functions taken from user_mbuf.h *************/ + +static int mbuf_constructor_dup(struct mbuf *m, int pkthdr, short type) +{ + int flags = pkthdr; + + m->m_next = NULL; + m->m_nextpkt = NULL; + m->m_len = 0; + m->m_flags = flags; + m->m_type = type; + if (flags & M_PKTHDR) { + m->m_data = m->m_pktdat; + m->m_pkthdr.rcvif = NULL; + m->m_pkthdr.len = 0; + m->m_pkthdr.header = NULL; + m->m_pkthdr.csum_flags = 0; + m->m_pkthdr.csum_data = 0; + m->m_pkthdr.tso_segsz = 0; + m->m_pkthdr.ether_vtag = 0; + SLIST_INIT(&m->m_pkthdr.tags); + } else + m->m_data = m->m_dat; + + return (0); +} + +/* __Userspace__ */ +struct mbuf * +m_get(int how, short type) +{ + struct mbuf *mret; +#if defined(SCTP_SIMPLE_ALLOCATOR) + struct mb_args mbuf_mb_args; + + /* The following setter function is not yet being enclosed within + * #if USING_MBUF_CONSTRUCTOR - #endif, until I have thoroughly tested + * mb_dtor_mbuf. See comment there + */ + mbuf_mb_args.flags = 0; + mbuf_mb_args.type = type; +#endif + /* Mbuf master zone, zone_mbuf, has already been + * created in mbuf_initialize() */ + mret = SCTP_ZONE_GET(zone_mbuf, struct mbuf); +#if defined(SCTP_SIMPLE_ALLOCATOR) + mb_ctor_mbuf(mret, &mbuf_mb_args, 0); +#endif + /*mret = ((struct mbuf *)umem_cache_alloc(zone_mbuf, UMEM_DEFAULT));*/ + + /* There are cases when an object available in the current CPU's + * loaded magazine and in those cases the object's constructor is not applied. + * If that is the case, then we are duplicating constructor initialization here, + * so that the mbuf is properly constructed before returning it. + */ + if (mret) { +#if USING_MBUF_CONSTRUCTOR + if (! (mret->m_type == type) ) { + mbuf_constructor_dup(mret, 0, type); + } +#else + mbuf_constructor_dup(mret, 0, type); +#endif + + } + return mret; +} + + +/* __Userspace__ */ +struct mbuf * +m_gethdr(int how, short type) +{ + struct mbuf *mret; +#if defined(SCTP_SIMPLE_ALLOCATOR) + struct mb_args mbuf_mb_args; + + /* The following setter function is not yet being enclosed within + * #if USING_MBUF_CONSTRUCTOR - #endif, until I have thoroughly tested + * mb_dtor_mbuf. See comment there + */ + mbuf_mb_args.flags = M_PKTHDR; + mbuf_mb_args.type = type; +#endif + mret = SCTP_ZONE_GET(zone_mbuf, struct mbuf); +#if defined(SCTP_SIMPLE_ALLOCATOR) + mb_ctor_mbuf(mret, &mbuf_mb_args, 0); +#endif + /*mret = ((struct mbuf *)umem_cache_alloc(zone_mbuf, UMEM_DEFAULT));*/ + /* There are cases when an object available in the current CPU's + * loaded magazine and in those cases the object's constructor is not applied. + * If that is the case, then we are duplicating constructor initialization here, + * so that the mbuf is properly constructed before returning it. + */ + if (mret) { +#if USING_MBUF_CONSTRUCTOR + if (! ((mret->m_flags & M_PKTHDR) && (mret->m_type == type)) ) { + mbuf_constructor_dup(mret, M_PKTHDR, type); + } +#else + mbuf_constructor_dup(mret, M_PKTHDR, type); +#endif + } + return mret; +} + +/* __Userspace__ */ +struct mbuf * +m_free(struct mbuf *m) +{ + + struct mbuf *n = m->m_next; + + if (m->m_flags & M_EXT) + mb_free_ext(m); + else if ((m->m_flags & M_NOFREE) == 0) { +#if defined(SCTP_SIMPLE_ALLOCATOR) + mb_dtor_mbuf(m, NULL); +#endif + SCTP_ZONE_FREE(zone_mbuf, m); + } + /*umem_cache_free(zone_mbuf, m);*/ + return (n); +} + + +static void +clust_constructor_dup(caddr_t m_clust, struct mbuf* m) +{ + u_int *refcnt; + int type, size; + + if (m == NULL) { + return; + } + /* Assigning cluster of MCLBYTES. TODO: Add jumbo frame functionality */ + type = EXT_CLUSTER; + size = MCLBYTES; + + refcnt = SCTP_ZONE_GET(zone_ext_refcnt, u_int); + /*refcnt = (u_int *)umem_cache_alloc(zone_ext_refcnt, UMEM_DEFAULT);*/ +#if !defined(SCTP_SIMPLE_ALLOCATOR) + if (refcnt == NULL) { + umem_reap(); + refcnt = SCTP_ZONE_GET(zone_ext_refcnt, u_int); + /*refcnt = (u_int *)umem_cache_alloc(zone_ext_refcnt, UMEM_DEFAULT);*/ + } +#endif + *refcnt = 1; + m->m_ext.ext_buf = (caddr_t)m_clust; + m->m_data = m->m_ext.ext_buf; + m->m_flags |= M_EXT; + m->m_ext.ext_free = NULL; + m->m_ext.ext_args = NULL; + m->m_ext.ext_size = size; + m->m_ext.ext_type = type; + m->m_ext.ref_cnt = refcnt; + return; +} + + +/* __Userspace__ */ +void +m_clget(struct mbuf *m, int how) +{ + caddr_t mclust_ret; +#if defined(SCTP_SIMPLE_ALLOCATOR) + struct clust_args clust_mb_args_l; +#endif + if (m->m_flags & M_EXT) { + SCTPDBG(SCTP_DEBUG_USR, "%s: %p mbuf already has cluster\n", __func__, (void *)m); + } + m->m_ext.ext_buf = (char *)NULL; +#if defined(SCTP_SIMPLE_ALLOCATOR) + clust_mb_args_l.parent_mbuf = m; +#endif + mclust_ret = SCTP_ZONE_GET(zone_clust, char); +#if defined(SCTP_SIMPLE_ALLOCATOR) + mb_ctor_clust(mclust_ret, &clust_mb_args_l, 0); +#endif + /*mclust_ret = umem_cache_alloc(zone_clust, UMEM_DEFAULT);*/ + /* + On a cluster allocation failure, call umem_reap() and retry. + */ + + if (mclust_ret == NULL) { +#if !defined(SCTP_SIMPLE_ALLOCATOR) + /* mclust_ret = SCTP_ZONE_GET(zone_clust, char); + mb_ctor_clust(mclust_ret, &clust_mb_args, 0); +#else*/ + umem_reap(); + mclust_ret = SCTP_ZONE_GET(zone_clust, char); +#endif + /*mclust_ret = umem_cache_alloc(zone_clust, UMEM_DEFAULT);*/ + /* if (NULL == mclust_ret) { */ + SCTPDBG(SCTP_DEBUG_USR, "Memory allocation failure in %s\n", __func__); + /* } */ + } + +#if USING_MBUF_CONSTRUCTOR + if ((m->m_ext.ext_buf == NULL)) { + clust_constructor_dup(mclust_ret, m); + } +#else + clust_constructor_dup(mclust_ret, m); +#endif +} + +struct mbuf * +m_getm2(struct mbuf *m, int len, int how, short type, int flags, int allonebuf) +{ + struct mbuf *mb, *nm = NULL, *mtail = NULL; + int size, mbuf_threshold, space_needed = len; + + KASSERT(len >= 0, ("%s: len is < 0", __func__)); + + /* Validate flags. */ + flags &= (M_PKTHDR | M_EOR); + + /* Packet header mbuf must be first in chain. */ + if ((flags & M_PKTHDR) && m != NULL) { + flags &= ~M_PKTHDR; + } + + if (allonebuf == 0) + mbuf_threshold = SCTP_BASE_SYSCTL(sctp_mbuf_threshold_count); + else + mbuf_threshold = 1; + + /* Loop and append maximum sized mbufs to the chain tail. */ + while (len > 0) { + if ((!allonebuf && len >= MCLBYTES) || (len > (int)(((mbuf_threshold - 1) * MLEN) + MHLEN))) { + mb = m_gethdr(how, type); + MCLGET(mb, how); + size = MCLBYTES; + /* SCTP_BUF_LEN(mb) = MCLBYTES; */ + } else if (flags & M_PKTHDR) { + mb = m_gethdr(how, type); + if (len < MHLEN) { + size = len; + } else { + size = MHLEN; + } + } else { + mb = m_get(how, type); + if (len < MLEN) { + size = len; + } else { + size = MLEN; + } + } + + /* Fail the whole operation if one mbuf can't be allocated. */ + if (mb == NULL) { + if (nm != NULL) + m_freem(nm); + return (NULL); + } + + if (allonebuf != 0 && size < space_needed) { + m_freem(mb); + return (NULL); + } + + /* Book keeping. */ + len -= size; + if (mtail != NULL) + mtail->m_next = mb; + else + nm = mb; + mtail = mb; + flags &= ~M_PKTHDR; /* Only valid on the first mbuf. */ + } + if (flags & M_EOR) { + mtail->m_flags |= M_EOR; /* Only valid on the last mbuf. */ + } + + /* If mbuf was supplied, append new chain to the end of it. */ + if (m != NULL) { + for (mtail = m; mtail->m_next != NULL; mtail = mtail->m_next); + mtail->m_next = nm; + mtail->m_flags &= ~M_EOR; + } else { + m = nm; + } + + return (m); +} + +/* + * Copy the contents of uio into a properly sized mbuf chain. + */ +struct mbuf * +m_uiotombuf(struct uio *uio, int how, int len, int align, int flags) +{ + struct mbuf *m, *mb; + int error, length; + ssize_t total; + int progress = 0; + + /* + * len can be zero or an arbitrary large value bound by + * the total data supplied by the uio. + */ + if (len > 0) + total = min(uio->uio_resid, len); + else + total = uio->uio_resid; + /* + * The smallest unit returned by m_getm2() is a single mbuf + * with pkthdr. We can't align past it. + */ + if (align >= MHLEN) + return (NULL); + /* + * Give us the full allocation or nothing. + * If len is zero return the smallest empty mbuf. + */ + m = m_getm2(NULL, (int)max(total + align, 1), how, MT_DATA, flags, 0); + if (m == NULL) + return (NULL); + m->m_data += align; + + /* Fill all mbufs with uio data and update header information. */ + for (mb = m; mb != NULL; mb = mb->m_next) { + length = (int)min(M_TRAILINGSPACE(mb), total - progress); + error = uiomove(mtod(mb, void *), length, uio); + if (error) { + m_freem(m); + return (NULL); + } + + mb->m_len = length; + progress += length; + if (flags & M_PKTHDR) + m->m_pkthdr.len += length; + } + KASSERT(progress == total, ("%s: progress != total", __func__)); + + return (m); +} + +u_int +m_length(struct mbuf *m0, struct mbuf **last) +{ + struct mbuf *m; + u_int len; + + len = 0; + for (m = m0; m != NULL; m = m->m_next) { + len += m->m_len; + if (m->m_next == NULL) + break; + } + if (last != NULL) + *last = m; + return (len); +} + +struct mbuf * +m_last(struct mbuf *m) +{ + while (m->m_next) { + m = m->m_next; + } + return (m); +} + +/* + * Unlink a tag from the list of tags associated with an mbuf. + */ +static __inline void +m_tag_unlink(struct mbuf *m, struct m_tag *t) +{ + + SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link); +} + +/* + * Reclaim resources associated with a tag. + */ +static __inline void +m_tag_free(struct m_tag *t) +{ + + (*t->m_tag_free)(t); +} + +/* + * Set up the contents of a tag. Note that this does not fill in the free + * method; the caller is expected to do that. + * + * XXX probably should be called m_tag_init, but that was already taken. + */ +static __inline void +m_tag_setup(struct m_tag *t, uint32_t cookie, int type, int len) +{ + + t->m_tag_id = type; + t->m_tag_len = len; + t->m_tag_cookie = cookie; +} + +/************ End functions from user_mbuf.h ******************/ + + + +/************ End functions to substitute umem_cache_alloc and umem_cache_free **************/ + +void +mbuf_initialize(void *dummy) +{ + + /* + * __Userspace__Configure UMA zones for Mbufs and Clusters. + * (TODO: m_getcl() - using packet secondary zone). + * There is no provision for trash_init and trash_fini in umem. + * + */ + /* zone_mbuf = umem_cache_create(MBUF_MEM_NAME, MSIZE, 0, + mb_ctor_mbuf, mb_dtor_mbuf, NULL, + &mbuf_mb_args, + NULL, 0); + zone_mbuf = umem_cache_create(MBUF_MEM_NAME, MSIZE, 0, NULL, NULL, NULL, NULL, NULL, 0);*/ +#if defined(SCTP_SIMPLE_ALLOCATOR) + SCTP_ZONE_INIT(zone_mbuf, MBUF_MEM_NAME, MSIZE, 0); +#else + zone_mbuf = umem_cache_create(MBUF_MEM_NAME, MSIZE, 0, + mb_ctor_mbuf, mb_dtor_mbuf, NULL, + NULL, + NULL, 0); +#endif + /*zone_ext_refcnt = umem_cache_create(MBUF_EXTREFCNT_MEM_NAME, sizeof(u_int), 0, + NULL, NULL, NULL, + NULL, + NULL, 0);*/ + SCTP_ZONE_INIT(zone_ext_refcnt, MBUF_EXTREFCNT_MEM_NAME, sizeof(u_int), 0); + + /*zone_clust = umem_cache_create(MBUF_CLUSTER_MEM_NAME, MCLBYTES, 0, + mb_ctor_clust, mb_dtor_clust, NULL, + &clust_mb_args, + NULL, 0); + zone_clust = umem_cache_create(MBUF_CLUSTER_MEM_NAME, MCLBYTES, 0, NULL, NULL, NULL, NULL, NULL,0);*/ +#if defined(SCTP_SIMPLE_ALLOCATOR) + SCTP_ZONE_INIT(zone_clust, MBUF_CLUSTER_MEM_NAME, MCLBYTES, 0); +#else + zone_clust = umem_cache_create(MBUF_CLUSTER_MEM_NAME, MCLBYTES, 0, + mb_ctor_clust, mb_dtor_clust, NULL, + &clust_mb_args, + NULL, 0); +#endif + + /* uma_prealloc() goes here... */ + + /* __Userspace__ Add umem_reap here for low memory situation? + * + */ + +} + + + +/* + * __Userspace__ + * + * Constructor for Mbuf master zone. We have a different constructor + * for allocating the cluster. + * + * The 'arg' pointer points to a mb_args structure which + * contains call-specific information required to support the + * mbuf allocation API. See user_mbuf.h. + * + * The flgs parameter below can be UMEM_DEFAULT or UMEM_NOFAIL depending on what + * was passed when umem_cache_alloc was called. + * TODO: Use UMEM_NOFAIL in umem_cache_alloc and also define a failure handler + * and call umem_nofail_callback(my_failure_handler) in the stack initialization routines + * The advantage of using UMEM_NOFAIL is that we don't have to check if umem_cache_alloc + * was successful or not. The failure handler would take care of it, if we use the UMEM_NOFAIL + * flag. + * + * NOTE Ref: http://docs.sun.com/app/docs/doc/819-2243/6n4i099p2?l=en&a=view&q=umem_zalloc) + * The umem_nofail_callback() function sets the **process-wide** UMEM_NOFAIL callback. + * It also mentions that umem_nofail_callback is Evolving. + * + */ +static int +mb_ctor_mbuf(void *mem, void *arg, int flgs) +{ +#if USING_MBUF_CONSTRUCTOR + struct mbuf *m; + struct mb_args *args; + + int flags; + short type; + + m = (struct mbuf *)mem; + args = (struct mb_args *)arg; + flags = args->flags; + type = args->type; + + m->m_next = NULL; + m->m_nextpkt = NULL; + m->m_len = 0; + m->m_flags = flags; + m->m_type = type; + if (flags & M_PKTHDR) { + m->m_data = m->m_pktdat; + m->m_pkthdr.rcvif = NULL; + m->m_pkthdr.len = 0; + m->m_pkthdr.header = NULL; + m->m_pkthdr.csum_flags = 0; + m->m_pkthdr.csum_data = 0; + m->m_pkthdr.tso_segsz = 0; + m->m_pkthdr.ether_vtag = 0; + SLIST_INIT(&m->m_pkthdr.tags); + } else + m->m_data = m->m_dat; +#endif + return (0); +} + + +/* + * __Userspace__ + * The Mbuf master zone destructor. + * This would be called in response to umem_cache_destroy + * TODO: Recheck if this is what we want to do in this destructor. + * (Note: the number of times mb_dtor_mbuf is called is equal to the + * number of individual mbufs allocated from zone_mbuf. + */ +static void +mb_dtor_mbuf(void *mem, void *arg) +{ + struct mbuf *m; + + m = (struct mbuf *)mem; + if ((m->m_flags & M_PKTHDR) != 0) { + m_tag_delete_chain(m, NULL); + } +} + + +/* __Userspace__ + * The Cluster zone constructor. + * + * Here the 'arg' pointer points to the Mbuf which we + * are configuring cluster storage for. If 'arg' is + * empty we allocate just the cluster without setting + * the mbuf to it. See mbuf.h. + */ +static int +mb_ctor_clust(void *mem, void *arg, int flgs) +{ + +#if USING_MBUF_CONSTRUCTOR + struct mbuf *m; + struct clust_args * cla; + u_int *refcnt; + int type, size; + sctp_zone_t zone; + + /* Assigning cluster of MCLBYTES. TODO: Add jumbo frame functionality */ + type = EXT_CLUSTER; + zone = zone_clust; + size = MCLBYTES; + + cla = (struct clust_args *)arg; + m = cla->parent_mbuf; + + refcnt = SCTP_ZONE_GET(zone_ext_refcnt, u_int); + /*refcnt = (u_int *)umem_cache_alloc(zone_ext_refcnt, UMEM_DEFAULT);*/ + *refcnt = 1; + + if (m != NULL) { + m->m_ext.ext_buf = (caddr_t)mem; + m->m_data = m->m_ext.ext_buf; + m->m_flags |= M_EXT; + m->m_ext.ext_free = NULL; + m->m_ext.ext_args = NULL; + m->m_ext.ext_size = size; + m->m_ext.ext_type = type; + m->m_ext.ref_cnt = refcnt; + } +#endif + return (0); +} + +/* __Userspace__ */ +static void +mb_dtor_clust(void *mem, void *arg) +{ + + /* mem is of type caddr_t. In sys/types.h we have typedef char * caddr_t; */ + /* mb_dtor_clust is called at time of umem_cache_destroy() (the number of times + * mb_dtor_clust is called is equal to the number of individual mbufs allocated + * from zone_clust. Similarly for mb_dtor_mbuf). + * At this point the following: + * struct mbuf *m; + * m = (struct mbuf *)arg; + * assert (*(m->m_ext.ref_cnt) == 0); is not meaningful since m->m_ext.ref_cnt = NULL; + * has been done in mb_free_ext(). + */ + +} + + + + +/* Unlink and free a packet tag. */ +void +m_tag_delete(struct mbuf *m, struct m_tag *t) +{ + KASSERT(m && t, ("m_tag_delete: null argument, m %p t %p", (void *)m, (void *)t)); + m_tag_unlink(m, t); + m_tag_free(t); +} + + +/* Unlink and free a packet tag chain, starting from given tag. */ +void +m_tag_delete_chain(struct mbuf *m, struct m_tag *t) +{ + + struct m_tag *p, *q; + + KASSERT(m, ("m_tag_delete_chain: null mbuf")); + if (t != NULL) + p = t; + else + p = SLIST_FIRST(&m->m_pkthdr.tags); + if (p == NULL) + return; + while ((q = SLIST_NEXT(p, m_tag_link)) != NULL) + m_tag_delete(m, q); + m_tag_delete(m, p); +} + +#if 0 +static void +sctp_print_mbuf_chain(struct mbuf *m) +{ + SCTP_DEBUG_USR(SCTP_DEBUG_USR, "Printing mbuf chain %p.\n", (void *)m); + for(; m; m=m->m_next) { + SCTP_DEBUG_USR(SCTP_DEBUG_USR, "%p: m_len = %ld, m_type = %x, m_next = %p.\n", (void *)m, m->m_len, m->m_type, (void *)m->m_next); + if (m->m_flags & M_EXT) + SCTP_DEBUG_USR(SCTP_DEBUG_USR, "%p: extend_size = %d, extend_buffer = %p, ref_cnt = %d.\n", (void *)m, m->m_ext.ext_size, (void *)m->m_ext.ext_buf, *(m->m_ext.ref_cnt)); + } +} +#endif + +/* + * Free an entire chain of mbufs and associated external buffers, if + * applicable. + */ +void +m_freem(struct mbuf *mb) +{ + while (mb != NULL) + mb = m_free(mb); +} + +/* + * __Userspace__ + * clean mbufs with M_EXT storage attached to them + * if the reference count hits 1. + */ +void +mb_free_ext(struct mbuf *m) +{ + + int skipmbuf; + + KASSERT((m->m_flags & M_EXT) == M_EXT, ("%s: M_EXT not set", __func__)); + KASSERT(m->m_ext.ref_cnt != NULL, ("%s: ref_cnt not set", __func__)); + + /* + * check if the header is embedded in the cluster + */ + skipmbuf = (m->m_flags & M_NOFREE); + + /* Free the external attached storage if this + * mbuf is the only reference to it. + *__Userspace__ TODO: jumbo frames + * + */ + /* NOTE: We had the same code that SCTP_DECREMENT_AND_CHECK_REFCOUNT + reduces to here before but the IPHONE malloc commit had changed + this to compare to 0 instead of 1 (see next line). Why? + . .. this caused a huge memory leak in Linux. + */ +#ifdef IPHONE + if (atomic_fetchadd_int(m->m_ext.ref_cnt, -1) == 0) +#else + if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(m->m_ext.ref_cnt)) +#endif + { + if (m->m_ext.ext_type == EXT_CLUSTER){ +#if defined(SCTP_SIMPLE_ALLOCATOR) + mb_dtor_clust(m->m_ext.ext_buf, &clust_mb_args); +#endif + SCTP_ZONE_FREE(zone_clust, m->m_ext.ext_buf); + SCTP_ZONE_FREE(zone_ext_refcnt, (u_int*)m->m_ext.ref_cnt); + m->m_ext.ref_cnt = NULL; + } + } + + if (skipmbuf) + return; + + + /* __Userspace__ Also freeing the storage for ref_cnt + * Free this mbuf back to the mbuf zone with all m_ext + * information purged. + */ + m->m_ext.ext_buf = NULL; + m->m_ext.ext_free = NULL; + m->m_ext.ext_args = NULL; + m->m_ext.ref_cnt = NULL; + m->m_ext.ext_size = 0; + m->m_ext.ext_type = 0; + m->m_flags &= ~M_EXT; +#if defined(SCTP_SIMPLE_ALLOCATOR) + mb_dtor_mbuf(m, NULL); +#endif + SCTP_ZONE_FREE(zone_mbuf, m); + + /*umem_cache_free(zone_mbuf, m);*/ +} + +/* + * "Move" mbuf pkthdr from "from" to "to". + * "from" must have M_PKTHDR set, and "to" must be empty. + */ +void +m_move_pkthdr(struct mbuf *to, struct mbuf *from) +{ + + to->m_flags = (from->m_flags & M_COPYFLAGS) | (to->m_flags & M_EXT); + if ((to->m_flags & M_EXT) == 0) + to->m_data = to->m_pktdat; + to->m_pkthdr = from->m_pkthdr; /* especially tags */ + SLIST_INIT(&from->m_pkthdr.tags); /* purge tags from src */ + from->m_flags &= ~M_PKTHDR; +} + + +/* + * Rearange an mbuf chain so that len bytes are contiguous + * and in the data area of an mbuf (so that mtod and dtom + * will work for a structure of size len). Returns the resulting + * mbuf chain on success, frees it and returns null on failure. + * If there is room, it will add up to max_protohdr-len extra bytes to the + * contiguous region in an attempt to avoid being called next time. + */ +struct mbuf * +m_pullup(struct mbuf *n, int len) +{ + struct mbuf *m; + int count; + int space; + + /* + * If first mbuf has no cluster, and has room for len bytes + * without shifting current data, pullup into it, + * otherwise allocate a new mbuf to prepend to the chain. + */ + if ((n->m_flags & M_EXT) == 0 && + n->m_data + len < &n->m_dat[MLEN] && n->m_next) { + if (n->m_len >= len) + return (n); + m = n; + n = n->m_next; + len -= m->m_len; + } else { + if (len > MHLEN) + goto bad; + MGET(m, M_NOWAIT, n->m_type); + if (m == NULL) + goto bad; + m->m_len = 0; + if (n->m_flags & M_PKTHDR) + M_MOVE_PKTHDR(m, n); + } + space = (int)(&m->m_dat[MLEN] - (m->m_data + m->m_len)); + do { + count = min(min(max(len, max_protohdr), space), n->m_len); + memcpy(mtod(m, caddr_t) + m->m_len,mtod(n, caddr_t), (u_int)count); + len -= count; + m->m_len += count; + n->m_len -= count; + space -= count; + if (n->m_len) + n->m_data += count; + else + n = m_free(n); + } while (len > 0 && n); + if (len > 0) { + (void) m_free(m); + goto bad; + } + m->m_next = n; + return (m); +bad: + m_freem(n); + return (NULL); +} + + +static struct mbuf * +m_dup1(struct mbuf *m, int off, int len, int wait) +{ + struct mbuf *n = NULL; + int copyhdr; + + if (len > MCLBYTES) + return NULL; + if (off == 0 && (m->m_flags & M_PKTHDR) != 0) + copyhdr = 1; + else + copyhdr = 0; + if (len >= MINCLSIZE) { + if (copyhdr == 1) { + m_clget(n, wait); /* TODO: include code for copying the header */ + m_dup_pkthdr(n, m, wait); + } else + m_clget(n, wait); + } else { + if (copyhdr == 1) + n = m_gethdr(wait, m->m_type); + else + n = m_get(wait, m->m_type); + } + if (!n) + return NULL; /* ENOBUFS */ + + if (copyhdr && !m_dup_pkthdr(n, m, wait)) { + m_free(n); + return NULL; + } + m_copydata(m, off, len, mtod(n, caddr_t)); + n->m_len = len; + return n; +} + + +/* Taken from sys/kern/uipc_mbuf2.c */ +struct mbuf * +m_pulldown(struct mbuf *m, int off, int len, int *offp) +{ + struct mbuf *n, *o; + int hlen, tlen, olen; + int writable; + + /* check invalid arguments. */ + KASSERT(m, ("m == NULL in m_pulldown()")); + if (len > MCLBYTES) { + m_freem(m); + return NULL; /* impossible */ + } + +#ifdef PULLDOWN_DEBUG + { + struct mbuf *t; + SCTP_DEBUG_USR(SCTP_DEBUG_USR, "before:"); + for (t = m; t; t = t->m_next) + SCTP_DEBUG_USR(SCTP_DEBUG_USR, " %d", t->m_len); + SCTP_DEBUG_USR(SCTP_DEBUG_USR, "\n"); + } +#endif + n = m; + while (n != NULL && off > 0) { + if (n->m_len > off) + break; + off -= n->m_len; + n = n->m_next; + } + /* be sure to point non-empty mbuf */ + while (n != NULL && n->m_len == 0) + n = n->m_next; + if (!n) { + m_freem(m); + return NULL; /* mbuf chain too short */ + } + + writable = 0; + if ((n->m_flags & M_EXT) == 0 || + (n->m_ext.ext_type == EXT_CLUSTER && M_WRITABLE(n))) + writable = 1; + + /* + * the target data is on <n, off>. + * if we got enough data on the mbuf "n", we're done. + */ + if ((off == 0 || offp) && len <= n->m_len - off && writable) + goto ok; + + /* + * when len <= n->m_len - off and off != 0, it is a special case. + * len bytes from <n, off> sits in single mbuf, but the caller does + * not like the starting position (off). + * chop the current mbuf into two pieces, set off to 0. + */ + if (len <= n->m_len - off) { + o = m_dup1(n, off, n->m_len - off, M_NOWAIT); + if (o == NULL) { + m_freem(m); + return NULL; /* ENOBUFS */ + } + n->m_len = off; + o->m_next = n->m_next; + n->m_next = o; + n = n->m_next; + off = 0; + goto ok; + } + /* + * we need to take hlen from <n, off> and tlen from <n->m_next, 0>, + * and construct contiguous mbuf with m_len == len. + * note that hlen + tlen == len, and tlen > 0. + */ + hlen = n->m_len - off; + tlen = len - hlen; + + /* + * ensure that we have enough trailing data on mbuf chain. + * if not, we can do nothing about the chain. + */ + olen = 0; + for (o = n->m_next; o != NULL; o = o->m_next) + olen += o->m_len; + if (hlen + olen < len) { + m_freem(m); + return NULL; /* mbuf chain too short */ + } + + /* + * easy cases first. + * we need to use m_copydata() to get data from <n->m_next, 0>. + */ + if ((off == 0 || offp) && (M_TRAILINGSPACE(n) >= tlen) && writable) { + m_copydata(n->m_next, 0, tlen, mtod(n, caddr_t) + n->m_len); + n->m_len += tlen; + m_adj(n->m_next, tlen); + goto ok; + } + + if ((off == 0 || offp) && (M_LEADINGSPACE(n->m_next) >= hlen) && writable) { + n->m_next->m_data -= hlen; + n->m_next->m_len += hlen; + memcpy( mtod(n->m_next, caddr_t), mtod(n, caddr_t) + off,hlen); + n->m_len -= hlen; + n = n->m_next; + off = 0; + goto ok; + } + + /* + * now, we need to do the hard way. don't m_copy as there's no room + * on both end. + */ + if (len > MLEN) + m_clget(o, M_NOWAIT); + /* o = m_getcl(M_NOWAIT, m->m_type, 0);*/ + else + o = m_get(M_NOWAIT, m->m_type); + if (!o) { + m_freem(m); + return NULL; /* ENOBUFS */ + } + /* get hlen from <n, off> into <o, 0> */ + o->m_len = hlen; + memcpy(mtod(o, caddr_t), mtod(n, caddr_t) + off, hlen); + n->m_len -= hlen; + /* get tlen from <n->m_next, 0> into <o, hlen> */ + m_copydata(n->m_next, 0, tlen, mtod(o, caddr_t) + o->m_len); + o->m_len += tlen; + m_adj(n->m_next, tlen); + o->m_next = n->m_next; + n->m_next = o; + n = o; + off = 0; +ok: +#ifdef PULLDOWN_DEBUG + { + struct mbuf *t; + SCTP_DEBUG_USR(SCTP_DEBUG_USR, "after:"); + for (t = m; t; t = t->m_next) + SCTP_DEBUG_USR(SCTP_DEBUG_USR, "%c%d", t == n ? '*' : ' ', t->m_len); + SCTP_DEBUG_USR(SCTP_DEBUG_USR, " (off=%d)\n", off); + } +#endif + if (offp) + *offp = off; + return n; +} + +/* + * Attach the the cluster from *m to *n, set up m_ext in *n + * and bump the refcount of the cluster. + */ +static void +mb_dupcl(struct mbuf *n, struct mbuf *m) +{ + KASSERT((m->m_flags & M_EXT) == M_EXT, ("%s: M_EXT not set", __func__)); + KASSERT(m->m_ext.ref_cnt != NULL, ("%s: ref_cnt not set", __func__)); + KASSERT((n->m_flags & M_EXT) == 0, ("%s: M_EXT set", __func__)); + + if (*(m->m_ext.ref_cnt) == 1) + *(m->m_ext.ref_cnt) += 1; + else + atomic_add_int(m->m_ext.ref_cnt, 1); + n->m_ext.ext_buf = m->m_ext.ext_buf; + n->m_ext.ext_free = m->m_ext.ext_free; + n->m_ext.ext_args = m->m_ext.ext_args; + n->m_ext.ext_size = m->m_ext.ext_size; + n->m_ext.ref_cnt = m->m_ext.ref_cnt; + n->m_ext.ext_type = m->m_ext.ext_type; + n->m_flags |= M_EXT; +} + + +/* + * Make a copy of an mbuf chain starting "off0" bytes from the beginning, + * continuing for "len" bytes. If len is M_COPYALL, copy to end of mbuf. + * The wait parameter is a choice of M_TRYWAIT/M_NOWAIT from caller. + * Note that the copy is read-only, because clusters are not copied, + * only their reference counts are incremented. + */ + +struct mbuf * +m_copym(struct mbuf *m, int off0, int len, int wait) +{ + struct mbuf *n, **np; + int off = off0; + struct mbuf *top; + int copyhdr = 0; + + KASSERT(off >= 0, ("m_copym, negative off %d", off)); + KASSERT(len >= 0, ("m_copym, negative len %d", len)); + KASSERT(m != NULL, ("m_copym, m is NULL")); + +#if !defined(INVARIANTS) + if (m == NULL) { + return (NULL); + } +#endif + if (off == 0 && m->m_flags & M_PKTHDR) + copyhdr = 1; + while (off > 0) { + KASSERT(m != NULL, ("m_copym, offset > size of mbuf chain")); + if (off < m->m_len) + break; + off -= m->m_len; + m = m->m_next; + } + np = ⊤ + top = 0; + while (len > 0) { + if (m == NULL) { + KASSERT(len == M_COPYALL, ("m_copym, length > size of mbuf chain")); + break; + } + if (copyhdr) + MGETHDR(n, wait, m->m_type); + else + MGET(n, wait, m->m_type); + *np = n; + if (n == NULL) + goto nospace; + if (copyhdr) { + if (!m_dup_pkthdr(n, m, wait)) + goto nospace; + if (len == M_COPYALL) + n->m_pkthdr.len -= off0; + else + n->m_pkthdr.len = len; + copyhdr = 0; + } + n->m_len = min(len, m->m_len - off); + if (m->m_flags & M_EXT) { + n->m_data = m->m_data + off; + mb_dupcl(n, m); + } else + memcpy(mtod(n, caddr_t), mtod(m, caddr_t) + off, (u_int)n->m_len); + if (len != M_COPYALL) + len -= n->m_len; + off = 0; + m = m->m_next; + np = &n->m_next; + } + + return (top); +nospace: + m_freem(top); + return (NULL); +} + + +int +m_tag_copy_chain(struct mbuf *to, struct mbuf *from, int how) +{ + struct m_tag *p, *t, *tprev = NULL; + + KASSERT(to && from, ("m_tag_copy_chain: null argument, to %p from %p", (void *)to, (void *)from)); + m_tag_delete_chain(to, NULL); + SLIST_FOREACH(p, &from->m_pkthdr.tags, m_tag_link) { + t = m_tag_copy(p, how); + if (t == NULL) { + m_tag_delete_chain(to, NULL); + return 0; + } + if (tprev == NULL) + SLIST_INSERT_HEAD(&to->m_pkthdr.tags, t, m_tag_link); + else + SLIST_INSERT_AFTER(tprev, t, m_tag_link); + tprev = t; + } + return 1; +} + +/* + * Duplicate "from"'s mbuf pkthdr in "to". + * "from" must have M_PKTHDR set, and "to" must be empty. + * In particular, this does a deep copy of the packet tags. + */ +int +m_dup_pkthdr(struct mbuf *to, struct mbuf *from, int how) +{ + + KASSERT(to, ("m_dup_pkthdr: to is NULL")); + KASSERT(from, ("m_dup_pkthdr: from is NULL")); + to->m_flags = (from->m_flags & M_COPYFLAGS) | (to->m_flags & M_EXT); + if ((to->m_flags & M_EXT) == 0) + to->m_data = to->m_pktdat; + to->m_pkthdr = from->m_pkthdr; + SLIST_INIT(&to->m_pkthdr.tags); + return (m_tag_copy_chain(to, from, MBTOM(how))); +} + +/* Copy a single tag. */ +struct m_tag * +m_tag_copy(struct m_tag *t, int how) +{ + struct m_tag *p; + + KASSERT(t, ("m_tag_copy: null tag")); + p = m_tag_alloc(t->m_tag_cookie, t->m_tag_id, t->m_tag_len, how); + if (p == NULL) + return (NULL); + memcpy(p + 1, t + 1, t->m_tag_len); /* Copy the data */ + return p; +} + +/* Get a packet tag structure along with specified data following. */ +struct m_tag * +m_tag_alloc(uint32_t cookie, int type, int len, int wait) +{ + struct m_tag *t; + + if (len < 0) + return NULL; + t = malloc(len + sizeof(struct m_tag)); + if (t == NULL) + return NULL; + m_tag_setup(t, cookie, type, len); + t->m_tag_free = m_tag_free_default; + return t; +} + +/* Free a packet tag. */ +void +m_tag_free_default(struct m_tag *t) +{ + free(t); +} + +/* + * Copy data from a buffer back into the indicated mbuf chain, + * starting "off" bytes from the beginning, extending the mbuf + * chain if necessary. + */ +void +m_copyback(struct mbuf *m0, int off, int len, caddr_t cp) +{ + int mlen; + struct mbuf *m = m0, *n; + int totlen = 0; + + if (m0 == NULL) + return; + while (off > (mlen = m->m_len)) { + off -= mlen; + totlen += mlen; + if (m->m_next == NULL) { + n = m_get(M_NOWAIT, m->m_type); + if (n == NULL) + goto out; + memset(mtod(n, caddr_t), 0, MLEN); + n->m_len = min(MLEN, len + off); + m->m_next = n; + } + m = m->m_next; + } + while (len > 0) { + mlen = min (m->m_len - off, len); + memcpy(off + mtod(m, caddr_t), cp, (u_int)mlen); + cp += mlen; + len -= mlen; + mlen += off; + off = 0; + totlen += mlen; + if (len == 0) + break; + if (m->m_next == NULL) { + n = m_get(M_NOWAIT, m->m_type); + if (n == NULL) + break; + n->m_len = min(MLEN, len); + m->m_next = n; + } + m = m->m_next; + } +out: if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen)) + m->m_pkthdr.len = totlen; +} + +/* + * Apply function f to the data in an mbuf chain starting "off" bytes from + * the beginning, continuing for "len" bytes. + */ +int +m_apply(struct mbuf *m, int off, int len, + int (*f)(void *, void *, u_int), void *arg) +{ + u_int count; + int rval; + + KASSERT(off >= 0, ("m_apply, negative off %d", off)); + KASSERT(len >= 0, ("m_apply, negative len %d", len)); + while (off > 0) { + KASSERT(m != NULL, ("m_apply, offset > size of mbuf chain")); + if (off < m->m_len) + break; + off -= m->m_len; + m = m->m_next; + } + while (len > 0) { + KASSERT(m != NULL, ("m_apply, offset > size of mbuf chain")); + count = min(m->m_len - off, len); + rval = (*f)(arg, mtod(m, caddr_t) + off, count); + if (rval) + return (rval); + len -= count; + off = 0; + m = m->m_next; + } + return (0); +} + +/* + * Lesser-used path for M_PREPEND: + * allocate new mbuf to prepend to chain, + * copy junk along. + */ +struct mbuf * +m_prepend(struct mbuf *m, int len, int how) +{ + struct mbuf *mn; + + if (m->m_flags & M_PKTHDR) + MGETHDR(mn, how, m->m_type); + else + MGET(mn, how, m->m_type); + if (mn == NULL) { + m_freem(m); + return (NULL); + } + if (m->m_flags & M_PKTHDR) + M_MOVE_PKTHDR(mn, m); + mn->m_next = m; + m = mn; + if (m->m_flags & M_PKTHDR) { + if (len < MHLEN) + MH_ALIGN(m, len); + } else { + if (len < MLEN) + M_ALIGN(m, len); + } + m->m_len = len; + return (m); +} + +/* + * Copy data from an mbuf chain starting "off" bytes from the beginning, + * continuing for "len" bytes, into the indicated buffer. + */ +void +m_copydata(const struct mbuf *m, int off, int len, caddr_t cp) +{ + u_int count; + + KASSERT(off >= 0, ("m_copydata, negative off %d", off)); + KASSERT(len >= 0, ("m_copydata, negative len %d", len)); + while (off > 0) { + KASSERT(m != NULL, ("m_copydata, offset > size of mbuf chain")); + if (off < m->m_len) + break; + off -= m->m_len; + m = m->m_next; + } + while (len > 0) { + KASSERT(m != NULL, ("m_copydata, length > size of mbuf chain")); + count = min(m->m_len - off, len); + memcpy(cp, mtod(m, caddr_t) + off, count); + len -= count; + cp += count; + off = 0; + m = m->m_next; + } +} + + +/* + * Concatenate mbuf chain n to m. + * Both chains must be of the same type (e.g. MT_DATA). + * Any m_pkthdr is not updated. + */ +void +m_cat(struct mbuf *m, struct mbuf *n) +{ + while (m->m_next) + m = m->m_next; + while (n) { + if (m->m_flags & M_EXT || + m->m_data + m->m_len + n->m_len >= &m->m_dat[MLEN]) { + /* just join the two chains */ + m->m_next = n; + return; + } + /* splat the data from one into the other */ + memcpy(mtod(m, caddr_t) + m->m_len, mtod(n, caddr_t), (u_int)n->m_len); + m->m_len += n->m_len; + n = m_free(n); + } +} + + +void +m_adj(struct mbuf *mp, int req_len) +{ + int len = req_len; + struct mbuf *m; + int count; + + if ((m = mp) == NULL) + return; + if (len >= 0) { + /* + * Trim from head. + */ + while (m != NULL && len > 0) { + if (m->m_len <= len) { + len -= m->m_len; + m->m_len = 0; + m = m->m_next; + } else { + m->m_len -= len; + m->m_data += len; + len = 0; + } + } + m = mp; + if (mp->m_flags & M_PKTHDR) + m->m_pkthdr.len -= (req_len - len); + } else { + /* + * Trim from tail. Scan the mbuf chain, + * calculating its length and finding the last mbuf. + * If the adjustment only affects this mbuf, then just + * adjust and return. Otherwise, rescan and truncate + * after the remaining size. + */ + len = -len; + count = 0; + for (;;) { + count += m->m_len; + if (m->m_next == (struct mbuf *)0) + break; + m = m->m_next; + } + if (m->m_len >= len) { + m->m_len -= len; + if (mp->m_flags & M_PKTHDR) + mp->m_pkthdr.len -= len; + return; + } + count -= len; + if (count < 0) + count = 0; + /* + * Correct length for chain is "count". + * Find the mbuf with last data, adjust its length, + * and toss data from remaining mbufs on chain. + */ + m = mp; + if (m->m_flags & M_PKTHDR) + m->m_pkthdr.len = count; + for (; m; m = m->m_next) { + if (m->m_len >= count) { + m->m_len = count; + if (m->m_next != NULL) { + m_freem(m->m_next); + m->m_next = NULL; + } + break; + } + count -= m->m_len; + } + } +} + + +/* m_split is used within sctp_handle_cookie_echo. */ + +/* + * Partition an mbuf chain in two pieces, returning the tail -- + * all but the first len0 bytes. In case of failure, it returns NULL and + * attempts to restore the chain to its original state. + * + * Note that the resulting mbufs might be read-only, because the new + * mbuf can end up sharing an mbuf cluster with the original mbuf if + * the "breaking point" happens to lie within a cluster mbuf. Use the + * M_WRITABLE() macro to check for this case. + */ +struct mbuf * +m_split(struct mbuf *m0, int len0, int wait) +{ + struct mbuf *m, *n; + u_int len = len0, remain; + + /* MBUF_CHECKSLEEP(wait); */ + for (m = m0; m && (int)len > m->m_len; m = m->m_next) + len -= m->m_len; + if (m == NULL) + return (NULL); + remain = m->m_len - len; + if (m0->m_flags & M_PKTHDR) { + MGETHDR(n, wait, m0->m_type); + if (n == NULL) + return (NULL); + n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif; + n->m_pkthdr.len = m0->m_pkthdr.len - len0; + m0->m_pkthdr.len = len0; + if (m->m_flags & M_EXT) + goto extpacket; + if (remain > MHLEN) { + /* m can't be the lead packet */ + MH_ALIGN(n, 0); + n->m_next = m_split(m, len, wait); + if (n->m_next == NULL) { + (void) m_free(n); + return (NULL); + } else { + n->m_len = 0; + return (n); + } + } else + MH_ALIGN(n, remain); + } else if (remain == 0) { + n = m->m_next; + m->m_next = NULL; + return (n); + } else { + MGET(n, wait, m->m_type); + if (n == NULL) + return (NULL); + M_ALIGN(n, remain); + } +extpacket: + if (m->m_flags & M_EXT) { + n->m_data = m->m_data + len; + mb_dupcl(n, m); + } else { + memcpy(mtod(n, caddr_t), mtod(m, caddr_t) + len, remain); + } + n->m_len = remain; + m->m_len = len; + n->m_next = m->m_next; + m->m_next = NULL; + return (n); +} + + + + +int +pack_send_buffer(caddr_t buffer, struct mbuf* mb){ + + int count_to_copy; + int total_count_copied = 0; + int offset = 0; + + do { + count_to_copy = mb->m_len; + memcpy(buffer+offset, mtod(mb, caddr_t), count_to_copy); + offset += count_to_copy; + total_count_copied += count_to_copy; + mb = mb->m_next; + } while(mb); + + return (total_count_copied); +} diff --git a/netwerk/sctp/src/user_mbuf.h b/netwerk/sctp/src/user_mbuf.h new file mode 100644 index 0000000000..268988f912 --- /dev/null +++ b/netwerk/sctp/src/user_mbuf.h @@ -0,0 +1,412 @@ +/*- + * Copyright (c) 1982, 1986, 1988, 1993 + * The Regents of the University of California. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + */ + +#ifndef _USER_MBUF_H_ +#define _USER_MBUF_H_ + +/* __Userspace__ header file for mbufs */ +#include <stdio.h> +#if !defined(SCTP_SIMPLE_ALLOCATOR) +#include "umem.h" +#endif +#include "user_malloc.h" +#include "netinet/sctp_os_userspace.h" + +#define USING_MBUF_CONSTRUCTOR 0 + +/* For Linux */ +#ifndef MSIZE +#define MSIZE 256 +/* #define MSIZE 1024 */ +#endif +#ifndef MCLBYTES +#define MCLBYTES 2048 +#endif + +struct mbuf * m_gethdr(int how, short type); +struct mbuf * m_get(int how, short type); +struct mbuf * m_free(struct mbuf *m); +void m_clget(struct mbuf *m, int how); +struct mbuf * m_getm2(struct mbuf *m, int len, int how, short type, int flags, int allonebuf); +struct mbuf *m_uiotombuf(struct uio *uio, int how, int len, int align, int flags); +u_int m_length(struct mbuf *m0, struct mbuf **last); +struct mbuf *m_last(struct mbuf *m); + +/* mbuf initialization function */ +void mbuf_initialize(void *); + +#define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from)) +#define MGET(m, how, type) ((m) = m_get((how), (type))) +#define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type))) +#define MCLGET(m, how) m_clget((m), (how)) + + +#define M_HDR_PAD ((sizeof(intptr_t)==4) ? 2 : 6) /* modified for __Userspace__ */ + +/* Length to m_copy to copy all. */ +#define M_COPYALL 1000000000 + +/* umem_cache_t is defined in user_include/umem.h as + * typedef struct umem_cache umem_cache_t; + * Note:umem_zone_t is a pointer. + */ +#if defined(SCTP_SIMPLE_ALLOCATOR) +typedef size_t sctp_zone_t; +#else +typedef umem_cache_t *sctp_zone_t; +#endif + +extern sctp_zone_t zone_mbuf; +extern sctp_zone_t zone_clust; +extern sctp_zone_t zone_ext_refcnt; + +/*- + * Macros for type conversion: + * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type. + * dtom(x) -- Convert data pointer within mbuf to mbuf pointer (XXX). + */ +#define mtod(m, t) ((t)((m)->m_data)) +#define dtom(x) ((struct mbuf *)((intptr_t)(x) & ~(MSIZE-1))) + +struct mb_args { + int flags; /* Flags for mbuf being allocated */ + short type; /* Type of mbuf being allocated */ +}; + +struct clust_args { + struct mbuf * parent_mbuf; +}; + +struct mbuf * m_split(struct mbuf *, int, int); +void m_cat(struct mbuf *m, struct mbuf *n); +void m_adj(struct mbuf *, int); +void mb_free_ext(struct mbuf *); +void m_freem(struct mbuf *); +struct m_tag *m_tag_alloc(uint32_t, int, int, int); +struct mbuf *m_copym(struct mbuf *, int, int, int); +void m_copyback(struct mbuf *, int, int, caddr_t); +int m_apply(struct mbuf *, int, int, int (*)(void *, void *, u_int), void *arg); +struct mbuf *m_pullup(struct mbuf *, int); +struct mbuf *m_pulldown(struct mbuf *, int off, int len, int *offp); +int m_dup_pkthdr(struct mbuf *, struct mbuf *, int); +struct m_tag *m_tag_copy(struct m_tag *, int); +int m_tag_copy_chain(struct mbuf *, struct mbuf *, int); +struct mbuf *m_prepend(struct mbuf *, int, int); +void m_copydata(const struct mbuf *, int, int, caddr_t); + +#define MBUF_MEM_NAME "mbuf" +#define MBUF_CLUSTER_MEM_NAME "mbuf_cluster" +#define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt" + +/* + * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead. + * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in + * sys/param.h), which has no additional overhead and is used instead of the + * internal data area; this is done when at least MINCLSIZE of data must be + * stored. Additionally, it is possible to allocate a separate buffer + * externally and attach it to the mbuf in a way similar to that of mbuf + * clusters. + */ +#define MLEN ((int)(MSIZE - sizeof(struct m_hdr))) /* normal data len */ +#define MHLEN ((int)(MLEN - sizeof(struct pkthdr))) /* data len w/pkthdr */ +#define MINCLSIZE ((int)(MHLEN + 1)) /* smallest amount to put in cluster */ +#define M_MAXCOMPRESS (MHLEN / 2) /* max amount to copy for compression */ + + +/* + * Header present at the beginning of every mbuf. + */ +struct m_hdr { + struct mbuf *mh_next; /* next buffer in chain */ + struct mbuf *mh_nextpkt; /* next chain in queue/record */ + caddr_t mh_data; /* location of data */ + int mh_len; /* amount of data in this mbuf */ + int mh_flags; /* flags; see below */ + short mh_type; /* type of data in this mbuf */ + uint8_t pad[M_HDR_PAD];/* word align */ +}; + +/* + * Packet tag structure (see below for details). + */ +struct m_tag { + SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */ + uint16_t m_tag_id; /* Tag ID */ + uint16_t m_tag_len; /* Length of data */ + uint32_t m_tag_cookie; /* ABI/Module ID */ + void (*m_tag_free)(struct m_tag *); +}; + +/* + * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set. + */ +struct pkthdr { + struct ifnet *rcvif; /* rcv interface */ + /* variables for ip and tcp reassembly */ + void *header; /* pointer to packet header */ + int len; /* total packet length */ + /* variables for hardware checksum */ + int csum_flags; /* flags regarding checksum */ + int csum_data; /* data field used by csum routines */ + uint16_t tso_segsz; /* TSO segment size */ + uint16_t ether_vtag; /* Ethernet 802.1p+q vlan tag */ + SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */ +}; + +/* + * Description of external storage mapped into mbuf; valid only if M_EXT is + * set. + */ +struct m_ext { + caddr_t ext_buf; /* start of buffer */ + void (*ext_free) /* free routine if not the usual */ + (void *, void *); + void *ext_args; /* optional argument pointer */ + u_int ext_size; /* size of buffer, for ext_free */ + volatile u_int *ref_cnt; /* pointer to ref count info */ + int ext_type; /* type of external storage */ +}; + + +/* + * The core of the mbuf object along with some shortcut defined for practical + * purposes. + */ +struct mbuf { + struct m_hdr m_hdr; + union { + struct { + struct pkthdr MH_pkthdr; /* M_PKTHDR set */ + union { + struct m_ext MH_ext; /* M_EXT set */ + char MH_databuf[MHLEN]; + } MH_dat; + } MH; + char M_databuf[MLEN]; /* !M_PKTHDR, !M_EXT */ + } M_dat; +}; + +#define m_next m_hdr.mh_next +#define m_len m_hdr.mh_len +#define m_data m_hdr.mh_data +#define m_type m_hdr.mh_type +#define m_flags m_hdr.mh_flags +#define m_nextpkt m_hdr.mh_nextpkt +#define m_act m_nextpkt +#define m_pkthdr M_dat.MH.MH_pkthdr +#define m_ext M_dat.MH.MH_dat.MH_ext +#define m_pktdat M_dat.MH.MH_dat.MH_databuf +#define m_dat M_dat.M_databuf + + +/* + * mbuf flags. + */ +#define M_EXT 0x0001 /* has associated external storage */ +#define M_PKTHDR 0x0002 /* start of record */ +#define M_EOR 0x0004 /* end of record */ +#define M_RDONLY 0x0008 /* associated data is marked read-only */ +#define M_PROTO1 0x0010 /* protocol-specific */ +#define M_PROTO2 0x0020 /* protocol-specific */ +#define M_PROTO3 0x0040 /* protocol-specific */ +#define M_PROTO4 0x0080 /* protocol-specific */ +#define M_PROTO5 0x0100 /* protocol-specific */ +#define M_FREELIST 0x8000 /* mbuf is on the free list */ + + +/* + * Flags copied when copying m_pkthdr. + */ +#define M_COPYFLAGS (M_PKTHDR|M_EOR|M_RDONLY|M_PROTO1|M_PROTO1|M_PROTO2|\ + M_PROTO3|M_PROTO4|M_PROTO5|\ + M_BCAST|M_MCAST|M_FRAG|M_FIRSTFRAG|M_LASTFRAG|\ + M_VLANTAG|M_PROMISC) + + +/* + * mbuf pkthdr flags (also stored in m_flags). + */ +#define M_BCAST 0x0200 /* send/received as link-level broadcast */ +#define M_MCAST 0x0400 /* send/received as link-level multicast */ +#define M_FRAG 0x0800 /* packet is a fragment of a larger packet */ +#define M_FIRSTFRAG 0x1000 /* packet is first fragment */ +#define M_LASTFRAG 0x2000 /* packet is last fragment */ +#define M_VLANTAG 0x10000 /* ether_vtag is valid */ +#define M_PROMISC 0x20000 /* packet was not for us */ +#define M_NOFREE 0x40000 /* do not free mbuf - it is embedded in the cluster */ + + +/* + * External buffer types: identify ext_buf type. + */ +#define EXT_CLUSTER 1 /* mbuf cluster */ +#define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */ +#define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */ +#define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ +#define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ +#define EXT_PACKET 6 /* mbuf+cluster from packet zone */ +#define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */ +#define EXT_NET_DRV 100 /* custom ext_buf provided by net driver(s) */ +#define EXT_MOD_TYPE 200 /* custom module's ext_buf type */ +#define EXT_DISPOSABLE 300 /* can throw this buffer away w/page flipping */ +#define EXT_EXTREF 400 /* has externally maintained ref_cnt ptr */ + + +/* + * mbuf types. + */ +#define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */ +#define MT_DATA 1 /* dynamic (data) allocation */ +#define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */ +#define MT_SONAME 8 /* socket name */ +#define MT_CONTROL 14 /* extra-data protocol message */ +#define MT_OOBDATA 15 /* expedited data */ +#define MT_NTYPES 16 /* number of mbuf types for mbtypes[] */ + +/* + * __Userspace__ flags like M_NOWAIT are defined in malloc.h + * Flags like these are used in functions like uma_zalloc() + * but don't have an equivalent in userland umem + * Flags specifying how an allocation should be made. + * + * The flag to use is as follows: + * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation. + * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block. + * + * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and + * if we cannot allocate immediately we may return NULL, whereas + * M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate resources we + * will block until they are available, and thus never return NULL. + * + * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT. + */ +#define MBTOM(how) (how) + +void m_tag_delete(struct mbuf *, struct m_tag *); +void m_tag_delete_chain(struct mbuf *, struct m_tag *); +void m_move_pkthdr(struct mbuf *, struct mbuf *); +void m_tag_free_default(struct m_tag *); + +extern int max_linkhdr; /* Largest link-level header */ +extern int max_protohdr; /* Size of largest protocol layer header. See user_mbuf.c */ + +/* + * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can + * be both the local data payload, or an external buffer area, depending on + * whether M_EXT is set). + */ +#define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \ + (!(((m)->m_flags & M_EXT)) || \ + (*((m)->m_ext.ref_cnt) == 1)) ) \ + +/* Check if the supplied mbuf has a packet header, or else panic. */ +#define M_ASSERTPKTHDR(m) \ + KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \ + ("%s: no mbuf packet header!", __func__)) + +/* + * Compute the amount of space available before the current start of data in + * an mbuf. + * + * The M_WRITABLE() is a temporary, conservative safety measure: the burden + * of checking writability of the mbuf data area rests solely with the caller. + */ +#define M_LEADINGSPACE(m) \ + (((m)->m_flags & M_EXT) ? \ + (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0): \ + ((m)->m_flags & M_PKTHDR)? (m)->m_data - (m)->m_pktdat : \ + (m)->m_data - (m)->m_dat) + +/* + * Compute the amount of space available after the end of data in an mbuf. + * + * The M_WRITABLE() is a temporary, conservative safety measure: the burden + * of checking writability of the mbuf data area rests solely with the caller. + */ +#define M_TRAILINGSPACE(m) \ + (((m)->m_flags & M_EXT) ? \ + (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size \ + - ((m)->m_data + (m)->m_len) : 0) : \ + &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len)) + + + +/* + * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be + * allocated, how specifies whether to wait. If the allocation fails, the + * original mbuf chain is freed and m is set to NULL. + */ +#define M_PREPEND(m, plen, how) do { \ + struct mbuf **_mmp = &(m); \ + struct mbuf *_mm = *_mmp; \ + int _mplen = (plen); \ + int __mhow = (how); \ + \ + if (M_LEADINGSPACE(_mm) >= _mplen) { \ + _mm->m_data -= _mplen; \ + _mm->m_len += _mplen; \ + } else \ + _mm = m_prepend(_mm, _mplen, __mhow); \ + if (_mm != NULL && _mm->m_flags & M_PKTHDR) \ + _mm->m_pkthdr.len += _mplen; \ + *_mmp = _mm; \ +} while (0) + +/* + * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an + * object of the specified size at the end of the mbuf, longword aligned. + */ +#define M_ALIGN(m, len) do { \ + KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)), \ + ("%s: M_ALIGN not normal mbuf", __func__)); \ + KASSERT((m)->m_data == (m)->m_dat, \ + ("%s: M_ALIGN not a virgin mbuf", __func__)); \ + (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1); \ +} while (0) + +/* + * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by + * M_DUP/MOVE_PKTHDR. + */ +#define MH_ALIGN(m, len) do { \ + KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT), \ + ("%s: MH_ALIGN not PKTHDR mbuf", __func__)); \ + KASSERT((m)->m_data == (m)->m_pktdat, \ + ("%s: MH_ALIGN not a virgin mbuf", __func__)); \ + (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1); \ +} while (0) + +#define M_SIZE(m) \ + (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \ + ((m)->m_flags & M_PKTHDR) ? MHLEN : \ + MLEN) + +#endif diff --git a/netwerk/sctp/src/user_queue.h b/netwerk/sctp/src/user_queue.h new file mode 100644 index 0000000000..fcd368bdd2 --- /dev/null +++ b/netwerk/sctp/src/user_queue.h @@ -0,0 +1,639 @@ +/*- + * Copyright (c) 1991, 1993 + * The Regents of the University of California. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + */ + +#ifndef _USER_QUEUE_H_ +#define _USER_QUEUE_H_ + +/* + * This file defines four types of data structures: singly-linked lists, + * singly-linked tail queues, lists and tail queues. + * + * A singly-linked list is headed by a single forward pointer. The elements + * are singly linked for minimum space and pointer manipulation overhead at + * the expense of O(n) removal for arbitrary elements. New elements can be + * added to the list after an existing element or at the head of the list. + * Elements being removed from the head of the list should use the explicit + * macro for this purpose for optimum efficiency. A singly-linked list may + * only be traversed in the forward direction. Singly-linked lists are ideal + * for applications with large datasets and few or no removals or for + * implementing a LIFO queue. + * + * A singly-linked tail queue is headed by a pair of pointers, one to the + * head of the list and the other to the tail of the list. The elements are + * singly linked for minimum space and pointer manipulation overhead at the + * expense of O(n) removal for arbitrary elements. New elements can be added + * to the list after an existing element, at the head of the list, or at the + * end of the list. Elements being removed from the head of the tail queue + * should use the explicit macro for this purpose for optimum efficiency. + * A singly-linked tail queue may only be traversed in the forward direction. + * Singly-linked tail queues are ideal for applications with large datasets + * and few or no removals or for implementing a FIFO queue. + * + * A list is headed by a single forward pointer (or an array of forward + * pointers for a hash table header). The elements are doubly linked + * so that an arbitrary element can be removed without a need to + * traverse the list. New elements can be added to the list before + * or after an existing element or at the head of the list. A list + * may only be traversed in the forward direction. + * + * A tail queue is headed by a pair of pointers, one to the head of the + * list and the other to the tail of the list. The elements are doubly + * linked so that an arbitrary element can be removed without a need to + * traverse the list. New elements can be added to the list before or + * after an existing element, at the head of the list, or at the end of + * the list. A tail queue may be traversed in either direction. + * + * For details on the use of these macros, see the queue(3) manual page. + * + * + * SLIST LIST STAILQ TAILQ + * _HEAD + + + + + * _HEAD_INITIALIZER + + + + + * _ENTRY + + + + + * _INIT + + + + + * _EMPTY + + + + + * _FIRST + + + + + * _NEXT + + + + + * _PREV - - - + + * _LAST - - + + + * _FOREACH + + + + + * _FOREACH_SAFE + + + + + * _FOREACH_REVERSE - - - + + * _FOREACH_REVERSE_SAFE - - - + + * _INSERT_HEAD + + + + + * _INSERT_BEFORE - + - + + * _INSERT_AFTER + + + + + * _INSERT_TAIL - - + + + * _CONCAT - - + + + * _REMOVE_AFTER + - + - + * _REMOVE_HEAD + - + - + * _REMOVE + + + + + * _SWAP + + + + + * + */ +#ifdef QUEUE_MACRO_DEBUG +/* Store the last 2 places the queue element or head was altered */ +struct qm_trace { + char * lastfile; + int lastline; + char * prevfile; + int prevline; +}; + +#define TRACEBUF struct qm_trace trace; +#define TRASHIT(x) do {(x) = (void *)-1;} while (0) +#define QMD_SAVELINK(name, link) void **name = (void *)&(link) + +#define QMD_TRACE_HEAD(head) do { \ + (head)->trace.prevline = (head)->trace.lastline; \ + (head)->trace.prevfile = (head)->trace.lastfile; \ + (head)->trace.lastline = __LINE__; \ + (head)->trace.lastfile = __FILE__; \ +} while (0) + +#define QMD_TRACE_ELEM(elem) do { \ + (elem)->trace.prevline = (elem)->trace.lastline; \ + (elem)->trace.prevfile = (elem)->trace.lastfile; \ + (elem)->trace.lastline = __LINE__; \ + (elem)->trace.lastfile = __FILE__; \ +} while (0) + +#else +#define QMD_TRACE_ELEM(elem) +#define QMD_TRACE_HEAD(head) +#define QMD_SAVELINK(name, link) +#define TRACEBUF +#define TRASHIT(x) +#endif /* QUEUE_MACRO_DEBUG */ + +/* + * Singly-linked List declarations. + */ +#define SLIST_HEAD(name, type) \ +struct name { \ + struct type *slh_first; /* first element */ \ +} + +#define SLIST_HEAD_INITIALIZER(head) \ + { NULL } + +#if defined(_WIN32) +#if defined(SLIST_ENTRY) +#undef SLIST_ENTRY +#endif +#endif + +#define SLIST_ENTRY(type) \ +struct { \ + struct type *sle_next; /* next element */ \ +} + +/* + * Singly-linked List functions. + */ +#define SLIST_EMPTY(head) ((head)->slh_first == NULL) + +#define SLIST_FIRST(head) ((head)->slh_first) + +#define SLIST_FOREACH(var, head, field) \ + for ((var) = SLIST_FIRST((head)); \ + (var); \ + (var) = SLIST_NEXT((var), field)) + +#define SLIST_FOREACH_SAFE(var, head, field, tvar) \ + for ((var) = SLIST_FIRST((head)); \ + (var) && ((tvar) = SLIST_NEXT((var), field), 1); \ + (var) = (tvar)) + +#define SLIST_FOREACH_PREVPTR(var, varp, head, field) \ + for ((varp) = &SLIST_FIRST((head)); \ + ((var) = *(varp)) != NULL; \ + (varp) = &SLIST_NEXT((var), field)) + +#define SLIST_INIT(head) do { \ + SLIST_FIRST((head)) = NULL; \ +} while (0) + +#define SLIST_INSERT_AFTER(slistelm, elm, field) do { \ + SLIST_NEXT((elm), field) = SLIST_NEXT((slistelm), field); \ + SLIST_NEXT((slistelm), field) = (elm); \ +} while (0) + +#define SLIST_INSERT_HEAD(head, elm, field) do { \ + SLIST_NEXT((elm), field) = SLIST_FIRST((head)); \ + SLIST_FIRST((head)) = (elm); \ +} while (0) + +#define SLIST_NEXT(elm, field) ((elm)->field.sle_next) + +#define SLIST_REMOVE(head, elm, type, field) do { \ + QMD_SAVELINK(oldnext, (elm)->field.sle_next); \ + if (SLIST_FIRST((head)) == (elm)) { \ + SLIST_REMOVE_HEAD((head), field); \ + } \ + else { \ + struct type *curelm = SLIST_FIRST((head)); \ + while (SLIST_NEXT(curelm, field) != (elm)) \ + curelm = SLIST_NEXT(curelm, field); \ + SLIST_REMOVE_AFTER(curelm, field); \ + } \ + TRASHIT(*oldnext); \ +} while (0) + +#define SLIST_REMOVE_AFTER(elm, field) do { \ + SLIST_NEXT(elm, field) = \ + SLIST_NEXT(SLIST_NEXT(elm, field), field); \ +} while (0) + +#define SLIST_REMOVE_HEAD(head, field) do { \ + SLIST_FIRST((head)) = SLIST_NEXT(SLIST_FIRST((head)), field); \ +} while (0) + +#define SLIST_SWAP(head1, head2, type) do { \ + struct type *swap_first = SLIST_FIRST(head1); \ + SLIST_FIRST(head1) = SLIST_FIRST(head2); \ + SLIST_FIRST(head2) = swap_first; \ +} while (0) + +/* + * Singly-linked Tail queue declarations. + */ +#define STAILQ_HEAD(name, type) \ +struct name { \ + struct type *stqh_first;/* first element */ \ + struct type **stqh_last;/* addr of last next element */ \ +} + +#define STAILQ_HEAD_INITIALIZER(head) \ + { NULL, &(head).stqh_first } + +#define STAILQ_ENTRY(type) \ +struct { \ + struct type *stqe_next; /* next element */ \ +} + +/* + * Singly-linked Tail queue functions. + */ +#define STAILQ_CONCAT(head1, head2) do { \ + if (!STAILQ_EMPTY((head2))) { \ + *(head1)->stqh_last = (head2)->stqh_first; \ + (head1)->stqh_last = (head2)->stqh_last; \ + STAILQ_INIT((head2)); \ + } \ +} while (0) + +#define STAILQ_EMPTY(head) ((head)->stqh_first == NULL) + +#define STAILQ_FIRST(head) ((head)->stqh_first) + +#define STAILQ_FOREACH(var, head, field) \ + for((var) = STAILQ_FIRST((head)); \ + (var); \ + (var) = STAILQ_NEXT((var), field)) + + +#define STAILQ_FOREACH_SAFE(var, head, field, tvar) \ + for ((var) = STAILQ_FIRST((head)); \ + (var) && ((tvar) = STAILQ_NEXT((var), field), 1); \ + (var) = (tvar)) + +#define STAILQ_INIT(head) do { \ + STAILQ_FIRST((head)) = NULL; \ + (head)->stqh_last = &STAILQ_FIRST((head)); \ +} while (0) + +#define STAILQ_INSERT_AFTER(head, tqelm, elm, field) do { \ + if ((STAILQ_NEXT((elm), field) = STAILQ_NEXT((tqelm), field)) == NULL)\ + (head)->stqh_last = &STAILQ_NEXT((elm), field); \ + STAILQ_NEXT((tqelm), field) = (elm); \ +} while (0) + +#define STAILQ_INSERT_HEAD(head, elm, field) do { \ + if ((STAILQ_NEXT((elm), field) = STAILQ_FIRST((head))) == NULL) \ + (head)->stqh_last = &STAILQ_NEXT((elm), field); \ + STAILQ_FIRST((head)) = (elm); \ +} while (0) + +#define STAILQ_INSERT_TAIL(head, elm, field) do { \ + STAILQ_NEXT((elm), field) = NULL; \ + *(head)->stqh_last = (elm); \ + (head)->stqh_last = &STAILQ_NEXT((elm), field); \ +} while (0) + +#define STAILQ_LAST(head, type, field) \ + (STAILQ_EMPTY((head)) ? \ + NULL : \ + ((struct type *)(void *) \ + ((char *)((head)->stqh_last) - __offsetof(struct type, field)))) + +#define STAILQ_NEXT(elm, field) ((elm)->field.stqe_next) + +#define STAILQ_REMOVE(head, elm, type, field) do { \ + QMD_SAVELINK(oldnext, (elm)->field.stqe_next); \ + if (STAILQ_FIRST((head)) == (elm)) { \ + STAILQ_REMOVE_HEAD((head), field); \ + } \ + else { \ + struct type *curelm = STAILQ_FIRST((head)); \ + while (STAILQ_NEXT(curelm, field) != (elm)) \ + curelm = STAILQ_NEXT(curelm, field); \ + STAILQ_REMOVE_AFTER(head, curelm, field); \ + } \ + TRASHIT(*oldnext); \ +} while (0) + +#define STAILQ_REMOVE_AFTER(head, elm, field) do { \ + if ((STAILQ_NEXT(elm, field) = \ + STAILQ_NEXT(STAILQ_NEXT(elm, field), field)) == NULL) \ + (head)->stqh_last = &STAILQ_NEXT((elm), field); \ +} while (0) + +#define STAILQ_REMOVE_HEAD(head, field) do { \ + if ((STAILQ_FIRST((head)) = \ + STAILQ_NEXT(STAILQ_FIRST((head)), field)) == NULL) \ + (head)->stqh_last = &STAILQ_FIRST((head)); \ +} while (0) + +#define STAILQ_SWAP(head1, head2, type) do { \ + struct type *swap_first = STAILQ_FIRST(head1); \ + struct type **swap_last = (head1)->stqh_last; \ + STAILQ_FIRST(head1) = STAILQ_FIRST(head2); \ + (head1)->stqh_last = (head2)->stqh_last; \ + STAILQ_FIRST(head2) = swap_first; \ + (head2)->stqh_last = swap_last; \ + if (STAILQ_EMPTY(head1)) \ + (head1)->stqh_last = &STAILQ_FIRST(head1); \ + if (STAILQ_EMPTY(head2)) \ + (head2)->stqh_last = &STAILQ_FIRST(head2); \ +} while (0) + + +/* + * List declarations. + */ +#define LIST_HEAD(name, type) \ +struct name { \ + struct type *lh_first; /* first element */ \ +} + +#define LIST_HEAD_INITIALIZER(head) \ + { NULL } + +#define LIST_ENTRY(type) \ +struct { \ + struct type *le_next; /* next element */ \ + struct type **le_prev; /* address of previous next element */ \ +} + +/* + * List functions. + */ + +#if defined(INVARIANTS) +#define QMD_LIST_CHECK_HEAD(head, field) do { \ + if (LIST_FIRST((head)) != NULL && \ + LIST_FIRST((head))->field.le_prev != \ + &LIST_FIRST((head))) \ + panic("Bad list head %p first->prev != head", (void *)(head)); \ +} while (0) + +#define QMD_LIST_CHECK_NEXT(elm, field) do { \ + if (LIST_NEXT((elm), field) != NULL && \ + LIST_NEXT((elm), field)->field.le_prev != \ + &((elm)->field.le_next)) \ + panic("Bad link elm %p next->prev != elm", (void *)(elm)); \ +} while (0) + +#define QMD_LIST_CHECK_PREV(elm, field) do { \ + if (*(elm)->field.le_prev != (elm)) \ + panic("Bad link elm %p prev->next != elm", (void *)(elm)); \ +} while (0) +#else +#define QMD_LIST_CHECK_HEAD(head, field) +#define QMD_LIST_CHECK_NEXT(elm, field) +#define QMD_LIST_CHECK_PREV(elm, field) +#endif /* (INVARIANTS) */ + +#define LIST_EMPTY(head) ((head)->lh_first == NULL) + +#define LIST_FIRST(head) ((head)->lh_first) + +#define LIST_FOREACH(var, head, field) \ + for ((var) = LIST_FIRST((head)); \ + (var); \ + (var) = LIST_NEXT((var), field)) + +#define LIST_FOREACH_SAFE(var, head, field, tvar) \ + for ((var) = LIST_FIRST((head)); \ + (var) && ((tvar) = LIST_NEXT((var), field), 1); \ + (var) = (tvar)) + +#define LIST_INIT(head) do { \ + LIST_FIRST((head)) = NULL; \ +} while (0) + +#define LIST_INSERT_AFTER(listelm, elm, field) do { \ + QMD_LIST_CHECK_NEXT(listelm, field); \ + if ((LIST_NEXT((elm), field) = LIST_NEXT((listelm), field)) != NULL)\ + LIST_NEXT((listelm), field)->field.le_prev = \ + &LIST_NEXT((elm), field); \ + LIST_NEXT((listelm), field) = (elm); \ + (elm)->field.le_prev = &LIST_NEXT((listelm), field); \ +} while (0) + +#define LIST_INSERT_BEFORE(listelm, elm, field) do { \ + QMD_LIST_CHECK_PREV(listelm, field); \ + (elm)->field.le_prev = (listelm)->field.le_prev; \ + LIST_NEXT((elm), field) = (listelm); \ + *(listelm)->field.le_prev = (elm); \ + (listelm)->field.le_prev = &LIST_NEXT((elm), field); \ +} while (0) + +#define LIST_INSERT_HEAD(head, elm, field) do { \ + QMD_LIST_CHECK_HEAD((head), field); \ + if ((LIST_NEXT((elm), field) = LIST_FIRST((head))) != NULL) \ + LIST_FIRST((head))->field.le_prev = &LIST_NEXT((elm), field);\ + LIST_FIRST((head)) = (elm); \ + (elm)->field.le_prev = &LIST_FIRST((head)); \ +} while (0) + +#define LIST_NEXT(elm, field) ((elm)->field.le_next) + +#define LIST_REMOVE(elm, field) do { \ + QMD_SAVELINK(oldnext, (elm)->field.le_next); \ + QMD_SAVELINK(oldprev, (elm)->field.le_prev); \ + QMD_LIST_CHECK_NEXT(elm, field); \ + QMD_LIST_CHECK_PREV(elm, field); \ + if (LIST_NEXT((elm), field) != NULL) \ + LIST_NEXT((elm), field)->field.le_prev = \ + (elm)->field.le_prev; \ + *(elm)->field.le_prev = LIST_NEXT((elm), field); \ + TRASHIT(*oldnext); \ + TRASHIT(*oldprev); \ +} while (0) + +#define LIST_SWAP(head1, head2, type, field) do { \ + struct type *swap_tmp = LIST_FIRST((head1)); \ + LIST_FIRST((head1)) = LIST_FIRST((head2)); \ + LIST_FIRST((head2)) = swap_tmp; \ + if ((swap_tmp = LIST_FIRST((head1))) != NULL) \ + swap_tmp->field.le_prev = &LIST_FIRST((head1)); \ + if ((swap_tmp = LIST_FIRST((head2))) != NULL) \ + swap_tmp->field.le_prev = &LIST_FIRST((head2)); \ +} while (0) + +/* + * Tail queue declarations. + */ +#define TAILQ_HEAD(name, type) \ +struct name { \ + struct type *tqh_first; /* first element */ \ + struct type **tqh_last; /* addr of last next element */ \ + TRACEBUF \ +} + +#define TAILQ_HEAD_INITIALIZER(head) \ + { NULL, &(head).tqh_first } + +#define TAILQ_ENTRY(type) \ +struct { \ + struct type *tqe_next; /* next element */ \ + struct type **tqe_prev; /* address of previous next element */ \ + TRACEBUF \ +} + +/* + * Tail queue functions. + */ +#if defined(INVARIANTS) +#define QMD_TAILQ_CHECK_HEAD(head, field) do { \ + if (!TAILQ_EMPTY(head) && \ + TAILQ_FIRST((head))->field.tqe_prev != \ + &TAILQ_FIRST((head))) \ + panic("Bad tailq head %p first->prev != head", (void *)(head)); \ +} while (0) + +#define QMD_TAILQ_CHECK_TAIL(head, field) do { \ + if (*(head)->tqh_last != NULL) \ + panic("Bad tailq NEXT(%p->tqh_last) != NULL", (void *)(head)); \ +} while (0) + +#define QMD_TAILQ_CHECK_NEXT(elm, field) do { \ + if (TAILQ_NEXT((elm), field) != NULL && \ + TAILQ_NEXT((elm), field)->field.tqe_prev != \ + &((elm)->field.tqe_next)) \ + panic("Bad link elm %p next->prev != elm", (void *)(elm)); \ +} while (0) + +#define QMD_TAILQ_CHECK_PREV(elm, field) do { \ + if (*(elm)->field.tqe_prev != (elm)) \ + panic("Bad link elm %p prev->next != elm", (void *)(elm)); \ +} while (0) +#else +#define QMD_TAILQ_CHECK_HEAD(head, field) +#define QMD_TAILQ_CHECK_TAIL(head, headname) +#define QMD_TAILQ_CHECK_NEXT(elm, field) +#define QMD_TAILQ_CHECK_PREV(elm, field) +#endif /* (INVARIANTS) */ + +#define TAILQ_CONCAT(head1, head2, field) do { \ + if (!TAILQ_EMPTY(head2)) { \ + *(head1)->tqh_last = (head2)->tqh_first; \ + (head2)->tqh_first->field.tqe_prev = (head1)->tqh_last; \ + (head1)->tqh_last = (head2)->tqh_last; \ + TAILQ_INIT((head2)); \ + QMD_TRACE_HEAD(head1); \ + QMD_TRACE_HEAD(head2); \ + } \ +} while (0) + +#define TAILQ_EMPTY(head) ((head)->tqh_first == NULL) + +#define TAILQ_FIRST(head) ((head)->tqh_first) + +#define TAILQ_FOREACH(var, head, field) \ + for ((var) = TAILQ_FIRST((head)); \ + (var); \ + (var) = TAILQ_NEXT((var), field)) + +#define TAILQ_FOREACH_SAFE(var, head, field, tvar) \ + for ((var) = TAILQ_FIRST((head)); \ + (var) && ((tvar) = TAILQ_NEXT((var), field), 1); \ + (var) = (tvar)) + +#define TAILQ_FOREACH_REVERSE(var, head, headname, field) \ + for ((var) = TAILQ_LAST((head), headname); \ + (var); \ + (var) = TAILQ_PREV((var), headname, field)) + +#define TAILQ_FOREACH_REVERSE_SAFE(var, head, headname, field, tvar) \ + for ((var) = TAILQ_LAST((head), headname); \ + (var) && ((tvar) = TAILQ_PREV((var), headname, field), 1); \ + (var) = (tvar)) + +#define TAILQ_INIT(head) do { \ + TAILQ_FIRST((head)) = NULL; \ + (head)->tqh_last = &TAILQ_FIRST((head)); \ + QMD_TRACE_HEAD(head); \ +} while (0) + +#define TAILQ_INSERT_AFTER(head, listelm, elm, field) do { \ + QMD_TAILQ_CHECK_NEXT(listelm, field); \ + if ((TAILQ_NEXT((elm), field) = TAILQ_NEXT((listelm), field)) != NULL)\ + TAILQ_NEXT((elm), field)->field.tqe_prev = \ + &TAILQ_NEXT((elm), field); \ + else { \ + (head)->tqh_last = &TAILQ_NEXT((elm), field); \ + QMD_TRACE_HEAD(head); \ + } \ + TAILQ_NEXT((listelm), field) = (elm); \ + (elm)->field.tqe_prev = &TAILQ_NEXT((listelm), field); \ + QMD_TRACE_ELEM(&(elm)->field); \ + QMD_TRACE_ELEM(&listelm->field); \ +} while (0) + +#define TAILQ_INSERT_BEFORE(listelm, elm, field) do { \ + QMD_TAILQ_CHECK_PREV(listelm, field); \ + (elm)->field.tqe_prev = (listelm)->field.tqe_prev; \ + TAILQ_NEXT((elm), field) = (listelm); \ + *(listelm)->field.tqe_prev = (elm); \ + (listelm)->field.tqe_prev = &TAILQ_NEXT((elm), field); \ + QMD_TRACE_ELEM(&(elm)->field); \ + QMD_TRACE_ELEM(&listelm->field); \ +} while (0) + +#define TAILQ_INSERT_HEAD(head, elm, field) do { \ + QMD_TAILQ_CHECK_HEAD(head, field); \ + if ((TAILQ_NEXT((elm), field) = TAILQ_FIRST((head))) != NULL) \ + TAILQ_FIRST((head))->field.tqe_prev = \ + &TAILQ_NEXT((elm), field); \ + else \ + (head)->tqh_last = &TAILQ_NEXT((elm), field); \ + TAILQ_FIRST((head)) = (elm); \ + (elm)->field.tqe_prev = &TAILQ_FIRST((head)); \ + QMD_TRACE_HEAD(head); \ + QMD_TRACE_ELEM(&(elm)->field); \ +} while (0) + +#define TAILQ_INSERT_TAIL(head, elm, field) do { \ + QMD_TAILQ_CHECK_TAIL(head, field); \ + TAILQ_NEXT((elm), field) = NULL; \ + (elm)->field.tqe_prev = (head)->tqh_last; \ + *(head)->tqh_last = (elm); \ + (head)->tqh_last = &TAILQ_NEXT((elm), field); \ + QMD_TRACE_HEAD(head); \ + QMD_TRACE_ELEM(&(elm)->field); \ +} while (0) + +#define TAILQ_LAST(head, headname) \ + (*(((struct headname *)((head)->tqh_last))->tqh_last)) + +#define TAILQ_NEXT(elm, field) ((elm)->field.tqe_next) + +#define TAILQ_PREV(elm, headname, field) \ + (*(((struct headname *)((elm)->field.tqe_prev))->tqh_last)) + +#define TAILQ_REMOVE(head, elm, field) do { \ + QMD_SAVELINK(oldnext, (elm)->field.tqe_next); \ + QMD_SAVELINK(oldprev, (elm)->field.tqe_prev); \ + QMD_TAILQ_CHECK_NEXT(elm, field); \ + QMD_TAILQ_CHECK_PREV(elm, field); \ + if ((TAILQ_NEXT((elm), field)) != NULL) \ + TAILQ_NEXT((elm), field)->field.tqe_prev = \ + (elm)->field.tqe_prev; \ + else { \ + (head)->tqh_last = (elm)->field.tqe_prev; \ + QMD_TRACE_HEAD(head); \ + } \ + *(elm)->field.tqe_prev = TAILQ_NEXT((elm), field); \ + TRASHIT(*oldnext); \ + TRASHIT(*oldprev); \ + QMD_TRACE_ELEM(&(elm)->field); \ +} while (0) + +#define TAILQ_SWAP(head1, head2, type, field) do { \ + struct type *swap_first = (head1)->tqh_first; \ + struct type **swap_last = (head1)->tqh_last; \ + (head1)->tqh_first = (head2)->tqh_first; \ + (head1)->tqh_last = (head2)->tqh_last; \ + (head2)->tqh_first = swap_first; \ + (head2)->tqh_last = swap_last; \ + if ((swap_first = (head1)->tqh_first) != NULL) \ + swap_first->field.tqe_prev = &(head1)->tqh_first; \ + else \ + (head1)->tqh_last = &(head1)->tqh_first; \ + if ((swap_first = (head2)->tqh_first) != NULL) \ + swap_first->field.tqe_prev = &(head2)->tqh_first; \ + else \ + (head2)->tqh_last = &(head2)->tqh_first; \ +} while (0) + +#endif /* !_SYS_QUEUE_H_ */ diff --git a/netwerk/sctp/src/user_recv_thread.c b/netwerk/sctp/src/user_recv_thread.c new file mode 100644 index 0000000000..66be13d749 --- /dev/null +++ b/netwerk/sctp/src/user_recv_thread.c @@ -0,0 +1,1566 @@ +/*- + * Copyright (c) 2009-2010 Brad Penoff + * Copyright (c) 2009-2010 Humaira Kamal + * Copyright (c) 2011-2012 Irene Ruengeler + * Copyright (c) 2011-2012 Michael Tuexen + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + */ + +#if defined(INET) || defined(INET6) +#include <sys/types.h> +#if !defined(_WIN32) +#include <sys/socket.h> +#include <netinet/in.h> +#include <unistd.h> +#include <pthread.h> +#if !defined(__DragonFly__) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#include <sys/uio.h> +#else +#include <user_ip6_var.h> +#endif +#endif +#include <netinet/sctp_os.h> +#include <netinet/sctp_var.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctp_input.h> +#if 0 +#if defined(__linux__) +#include <linux/netlink.h> +#ifdef HAVE_LINUX_IF_ADDR_H +#include <linux/if_addr.h> +#endif +#ifdef HAVE_LINUX_RTNETLINK_H +#include <linux/rtnetlink.h> +#endif +#endif +#endif +#if defined(HAVE_NET_ROUTE_H) +# include <net/route.h> +#elif defined(__APPLE__) +/* Apple SDKs for iOS, tvOS, watchOS, etc. don't ship this header */ +# define RTM_NEWADDR 0xc +# define RTM_DELADDR 0xd +# define RTAX_IFA 5 +# define RTAX_MAX 8 +#endif +/* local macros and datatypes used to get IP addresses system independently */ +#if !defined(IP_PKTINFO) && !defined(IP_RECVDSTADDR) +# error "Can't determine socket option to use to get UDP IP" +#endif + +void recv_thread_destroy(void); + +#define MAXLEN_MBUF_CHAIN 128 + +#define ROUNDUP(a, size) (((a) & ((size)-1)) ? (1 + ((a) | ((size)-1))) : (a)) + +#if defined(__APPLE__) || defined(__DragonFly__) || defined(__FreeBSD__) +#define NEXT_SA(ap) ap = (struct sockaddr *) \ + ((caddr_t) ap + (ap->sa_len ? ROUNDUP(ap->sa_len, sizeof (uint32_t)) : sizeof(uint32_t))) +#endif + +#if defined(__APPLE__) || defined(__DragonFly__) || defined(__FreeBSD__) +static void +sctp_get_rtaddrs(int addrs, struct sockaddr *sa, struct sockaddr **rti_info) +{ + int i; + + for (i = 0; i < RTAX_MAX; i++) { + if (addrs & (1 << i)) { + rti_info[i] = sa; + NEXT_SA(sa); + } else { + rti_info[i] = NULL; + } + } +} + +static void +sctp_handle_ifamsg(unsigned char type, unsigned short index, struct sockaddr *sa) +{ + int rc; + struct ifaddrs *ifa, *ifas; + + /* handle only the types we want */ + if ((type != RTM_NEWADDR) && (type != RTM_DELADDR)) { + return; + } + + rc = getifaddrs(&ifas); + if (rc != 0) { + return; + } + for (ifa = ifas; ifa; ifa = ifa->ifa_next) { + if (index == if_nametoindex(ifa->ifa_name)) { + break; + } + } + if (ifa == NULL) { + freeifaddrs(ifas); + return; + } + + /* relay the appropriate address change to the base code */ + if (type == RTM_NEWADDR) { + (void)sctp_add_addr_to_vrf(SCTP_DEFAULT_VRFID, + NULL, + if_nametoindex(ifa->ifa_name), + 0, + ifa->ifa_name, + NULL, + sa, + 0, + 1); + } else { + sctp_del_addr_from_vrf(SCTP_DEFAULT_VRFID, ifa->ifa_addr, + if_nametoindex(ifa->ifa_name), + ifa->ifa_name); + } + freeifaddrs(ifas); +} + +static void * +recv_function_route(void *arg) +{ + ssize_t ret; + struct ifa_msghdr *ifa; + char rt_buffer[1024]; + struct sockaddr *sa, *rti_info[RTAX_MAX]; + + sctp_userspace_set_threadname("SCTP addr mon"); + + while (1) { + memset(rt_buffer, 0, sizeof(rt_buffer)); + ret = recv(SCTP_BASE_VAR(userspace_route), rt_buffer, sizeof(rt_buffer), 0); + + if (ret > 0) { + ifa = (struct ifa_msghdr *) rt_buffer; + if (ifa->ifam_type != RTM_DELADDR && ifa->ifam_type != RTM_NEWADDR) { + continue; + } + sa = (struct sockaddr *) (ifa + 1); + sctp_get_rtaddrs(ifa->ifam_addrs, sa, rti_info); + switch (ifa->ifam_type) { + case RTM_DELADDR: + case RTM_NEWADDR: + sctp_handle_ifamsg(ifa->ifam_type, ifa->ifam_index, rti_info[RTAX_IFA]); + break; + default: + /* ignore this routing event */ + break; + } + } + if (ret < 0) { + if (errno == EAGAIN || errno == EINTR) { + continue; + } else { + break; + } + } + } + return (NULL); +} +#endif + +#if 0 +/* This does not yet work on Linux */ +static void * +recv_function_route(void *arg) +{ + int len; + char buf[4096]; + struct iovec iov = { buf, sizeof(buf) }; + struct msghdr msg; + struct nlmsghdr *nh; + struct ifaddrmsg *rtmsg; + struct rtattr *rtatp; + struct in_addr *inp; + struct sockaddr_nl sanl; +#ifdef INET + struct sockaddr_in *sa; +#endif +#ifdef INET6 + struct sockaddr_in6 *sa6; +#endif + + for (;;) { + memset(&sanl, 0, sizeof(sanl)); + sanl.nl_family = AF_NETLINK; + sanl.nl_groups = RTMGRP_IPV6_IFADDR | RTMGRP_IPV4_IFADDR; + memset(&msg, 0, sizeof(struct msghdr)); + msg.msg_name = (void *)&sanl; + msg.msg_namelen = sizeof(sanl); + msg.msg_iov = &iov; + msg.msg_iovlen = 1; + msg.msg_control = NULL; + msg.msg_controllen = 0; + + len = recvmsg(SCTP_BASE_VAR(userspace_route), &msg, 0); + + if (len < 0) { + if (errno == EAGAIN || errno == EINTR) { + continue; + } else { + break; + } + } + for (nh = (struct nlmsghdr *) buf; NLMSG_OK (nh, len); + nh = NLMSG_NEXT (nh, len)) { + if (nh->nlmsg_type == NLMSG_DONE) + break; + + if (nh->nlmsg_type == RTM_NEWADDR || nh->nlmsg_type == RTM_DELADDR) { + rtmsg = (struct ifaddrmsg *)NLMSG_DATA(nh); + rtatp = (struct rtattr *)IFA_RTA(rtmsg); + if (rtatp->rta_type == IFA_ADDRESS) { + inp = (struct in_addr *)RTA_DATA(rtatp); + switch (rtmsg->ifa_family) { +#ifdef INET + case AF_INET: + sa = (struct sockaddr_in *)malloc(sizeof(struct sockaddr_in)); + sa->sin_family = rtmsg->ifa_family; + sa->sin_port = 0; + memcpy(&sa->sin_addr, inp, sizeof(struct in_addr)); + sctp_handle_ifamsg(nh->nlmsg_type, rtmsg->ifa_index, (struct sockaddr *)sa); + break; +#endif +#ifdef INET6 + case AF_INET6: + sa6 = (struct sockaddr_in6 *)malloc(sizeof(struct sockaddr_in6)); + sa6->sin6_family = rtmsg->ifa_family; + sa6->sin6_port = 0; + memcpy(&sa6->sin6_addr, inp, sizeof(struct in6_addr)); + sctp_handle_ifamsg(nh->nlmsg_type, rtmsg->ifa_index, (struct sockaddr *)sa6); + break; +#endif + default: + SCTPDBG(SCTP_DEBUG_USR, "Address family %d not supported.\n", rtmsg->ifa_family); + break; + } + } + } + } + } + return (NULL); +} +#endif + +#ifdef INET +static void * +recv_function_raw(void *arg) +{ + struct mbuf **recvmbuf; + struct ip *iphdr; + struct sctphdr *sh; + uint16_t port; + int offset, ecn = 0; + int compute_crc = 1; + struct sctp_chunkhdr *ch; + struct sockaddr_in src, dst; +#if !defined(_WIN32) + ssize_t res; + unsigned int ncounter; + struct msghdr msg; + struct iovec recv_iovec[MAXLEN_MBUF_CHAIN]; +#else + WSABUF recv_iovec[MAXLEN_MBUF_CHAIN]; + int nResult, m_ErrorCode; + DWORD flags; + DWORD ncounter; + struct sockaddr_in from; + int fromlen; +#endif + /*Initially the entire set of mbufs is to be allocated. + to_fill indicates this amount. */ + int to_fill = MAXLEN_MBUF_CHAIN; + /* iovlen is the size of each mbuf in the chain */ + int i, n; + unsigned int iovlen = MCLBYTES; + int want_ext = (iovlen > MLEN)? 1 : 0; + int want_header = 0; + + sctp_userspace_set_threadname("SCTP/IP4 rcv"); + + memset(&src, 0, sizeof(struct sockaddr_in)); + memset(&dst, 0, sizeof(struct sockaddr_in)); + + recvmbuf = malloc(sizeof(struct mbuf *) * MAXLEN_MBUF_CHAIN); + + while (1) { + for (i = 0; i < to_fill; i++) { + /* Not getting the packet header. Tests with chain of one run + as usual without having the packet header. + Have tried both sending and receiving + */ + recvmbuf[i] = sctp_get_mbuf_for_msg(iovlen, want_header, M_NOWAIT, want_ext, MT_DATA); +#if !defined(_WIN32) + recv_iovec[i].iov_base = (caddr_t)recvmbuf[i]->m_data; + recv_iovec[i].iov_len = iovlen; +#else + recv_iovec[i].buf = (caddr_t)recvmbuf[i]->m_data; + recv_iovec[i].len = iovlen; +#endif + } + to_fill = 0; +#if defined(_WIN32) + flags = 0; + ncounter = 0; + fromlen = sizeof(struct sockaddr_in); + memset(&from, 0, sizeof(struct sockaddr_in)); + + nResult = WSARecvFrom(SCTP_BASE_VAR(userspace_rawsctp), recv_iovec, MAXLEN_MBUF_CHAIN, &ncounter, &flags, (struct sockaddr *)&from, &fromlen, NULL, NULL); + if (nResult != 0) { + m_ErrorCode = WSAGetLastError(); + if ((m_ErrorCode == WSAENOTSOCK) || (m_ErrorCode == WSAEINTR)) { + break; + } + continue; + } + n = ncounter; +#else + memset(&msg, 0, sizeof(struct msghdr)); + msg.msg_name = NULL; + msg.msg_namelen = 0; + msg.msg_iov = recv_iovec; + msg.msg_iovlen = MAXLEN_MBUF_CHAIN; + msg.msg_control = NULL; + msg.msg_controllen = 0; + res = recvmsg(SCTP_BASE_VAR(userspace_rawsctp), &msg, 0); + if (res < 0) { + if (errno == EAGAIN || errno == EINTR) { + continue; + } else { + break; + } + } + ncounter = (unsigned int)res; + n = (int)res; +#endif + SCTP_HEADER_LEN(recvmbuf[0]) = n; /* length of total packet */ + SCTP_STAT_INCR(sctps_recvpackets); + SCTP_STAT_INCR_COUNTER64(sctps_inpackets); + + if ((unsigned int)n <= iovlen) { + SCTP_BUF_LEN(recvmbuf[0]) = n; + (to_fill)++; + } else { + i = 0; + SCTP_BUF_LEN(recvmbuf[0]) = iovlen; + + ncounter -= min(ncounter, iovlen); + (to_fill)++; + do { + recvmbuf[i]->m_next = recvmbuf[i+1]; + SCTP_BUF_LEN(recvmbuf[i]->m_next) = min(ncounter, iovlen); + i++; + ncounter -= min(ncounter, iovlen); + (to_fill)++; + } while (ncounter > 0); + } + + offset = sizeof(struct ip) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); + if (SCTP_BUF_LEN(recvmbuf[0]) < offset) { + if ((recvmbuf[0] = m_pullup(recvmbuf[0], offset)) == NULL) { + SCTP_STAT_INCR(sctps_hdrops); + continue; + } + } + iphdr = mtod(recvmbuf[0], struct ip *); + sh = (struct sctphdr *)((caddr_t)iphdr + sizeof(struct ip)); + ch = (struct sctp_chunkhdr *)((caddr_t)sh + sizeof(struct sctphdr)); + offset -= sizeof(struct sctp_chunkhdr); + + if (iphdr->ip_tos != 0) { + ecn = iphdr->ip_tos & 0x03; + } + + dst.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + dst.sin_len = sizeof(struct sockaddr_in); +#endif + dst.sin_addr = iphdr->ip_dst; + dst.sin_port = sh->dest_port; + + src.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + src.sin_len = sizeof(struct sockaddr_in); +#endif + src.sin_addr = iphdr->ip_src; + src.sin_port = sh->src_port; + + /* SCTP does not allow broadcasts or multicasts */ + if (IN_MULTICAST(ntohl(dst.sin_addr.s_addr))) { + m_freem(recvmbuf[0]); + continue; + } + if (SCTP_IS_IT_BROADCAST(dst.sin_addr, recvmbuf[0])) { + m_freem(recvmbuf[0]); + continue; + } + + port = 0; + + if (SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) && + ((IN4_ISLOOPBACK_ADDRESS(&src.sin_addr) && + IN4_ISLOOPBACK_ADDRESS(&dst.sin_addr)) || + (src.sin_addr.s_addr == dst.sin_addr.s_addr))) { + compute_crc = 0; + SCTP_STAT_INCR(sctps_recvhwcrc); + } else { + SCTP_STAT_INCR(sctps_recvswcrc); + } + SCTPDBG(SCTP_DEBUG_USR, "%s: Received %d bytes.", __func__, n); + SCTPDBG(SCTP_DEBUG_USR, " - calling sctp_common_input_processing with off=%d\n", offset); + sctp_common_input_processing(&recvmbuf[0], sizeof(struct ip), offset, n, + (struct sockaddr *)&src, + (struct sockaddr *)&dst, + sh, ch, + compute_crc, + ecn, + SCTP_DEFAULT_VRFID, port); + if (recvmbuf[0]) { + m_freem(recvmbuf[0]); + } + } + for (i = 0; i < MAXLEN_MBUF_CHAIN; i++) { + m_free(recvmbuf[i]); + } + /* free the array itself */ + free(recvmbuf); + SCTPDBG(SCTP_DEBUG_USR, "%s: Exiting SCTP/IP4 rcv\n", __func__); + return (NULL); +} +#endif + +#if defined(INET6) +static void * +recv_function_raw6(void *arg) +{ + struct mbuf **recvmbuf6; +#if !defined(_WIN32) + ssize_t res; + unsigned int ncounter; + struct iovec recv_iovec[MAXLEN_MBUF_CHAIN]; + struct msghdr msg; + struct cmsghdr *cmsgptr; + char cmsgbuf[CMSG_SPACE(sizeof (struct in6_pktinfo))]; +#else + WSABUF recv_iovec[MAXLEN_MBUF_CHAIN]; + int nResult, m_ErrorCode; + DWORD ncounter = 0; + struct sockaddr_in6 from; + GUID WSARecvMsg_GUID = WSAID_WSARECVMSG; + LPFN_WSARECVMSG WSARecvMsg; + WSACMSGHDR *cmsgptr; + WSAMSG msg; + char ControlBuffer[1024]; +#endif + struct sockaddr_in6 src, dst; + struct sctphdr *sh; + int offset; + struct sctp_chunkhdr *ch; + /*Initially the entire set of mbufs is to be allocated. + to_fill indicates this amount. */ + int to_fill = MAXLEN_MBUF_CHAIN; + /* iovlen is the size of each mbuf in the chain */ + int i, n; + int compute_crc = 1; + unsigned int iovlen = MCLBYTES; + int want_ext = (iovlen > MLEN)? 1 : 0; + int want_header = 0; + + sctp_userspace_set_threadname("SCTP/IP6 rcv"); + + recvmbuf6 = malloc(sizeof(struct mbuf *) * MAXLEN_MBUF_CHAIN); + + for (;;) { + for (i = 0; i < to_fill; i++) { + /* Not getting the packet header. Tests with chain of one run + as usual without having the packet header. + Have tried both sending and receiving + */ + recvmbuf6[i] = sctp_get_mbuf_for_msg(iovlen, want_header, M_NOWAIT, want_ext, MT_DATA); +#if !defined(_WIN32) + recv_iovec[i].iov_base = (caddr_t)recvmbuf6[i]->m_data; + recv_iovec[i].iov_len = iovlen; +#else + recv_iovec[i].buf = (caddr_t)recvmbuf6[i]->m_data; + recv_iovec[i].len = iovlen; +#endif + } + to_fill = 0; +#if defined(_WIN32) + ncounter = 0; + memset(&from, 0, sizeof(struct sockaddr_in6)); + nResult = WSAIoctl(SCTP_BASE_VAR(userspace_rawsctp6), SIO_GET_EXTENSION_FUNCTION_POINTER, + &WSARecvMsg_GUID, sizeof WSARecvMsg_GUID, + &WSARecvMsg, sizeof WSARecvMsg, + &ncounter, NULL, NULL); + if (nResult == 0) { + msg.name = (void *)&src; + msg.namelen = sizeof(struct sockaddr_in6); + msg.lpBuffers = recv_iovec; + msg.dwBufferCount = MAXLEN_MBUF_CHAIN; + msg.Control.len = sizeof ControlBuffer; + msg.Control.buf = ControlBuffer; + msg.dwFlags = 0; + nResult = WSARecvMsg(SCTP_BASE_VAR(userspace_rawsctp6), &msg, &ncounter, NULL, NULL); + } + if (nResult != 0) { + m_ErrorCode = WSAGetLastError(); + if ((m_ErrorCode == WSAENOTSOCK) || (m_ErrorCode == WSAEINTR)) { + break; + } + continue; + } + n = ncounter; +#else + memset(&msg, 0, sizeof(struct msghdr)); + memset(&src, 0, sizeof(struct sockaddr_in6)); + memset(&dst, 0, sizeof(struct sockaddr_in6)); + memset(cmsgbuf, 0, CMSG_SPACE(sizeof (struct in6_pktinfo))); + msg.msg_name = (void *)&src; + msg.msg_namelen = sizeof(struct sockaddr_in6); + msg.msg_iov = recv_iovec; + msg.msg_iovlen = MAXLEN_MBUF_CHAIN; + msg.msg_control = (void *)cmsgbuf; + msg.msg_controllen = (socklen_t)CMSG_SPACE(sizeof (struct in6_pktinfo)); + msg.msg_flags = 0; + res = recvmsg(SCTP_BASE_VAR(userspace_rawsctp6), &msg, 0); + if (res < 0) { + if (errno == EAGAIN || errno == EINTR) { + continue; + } else { + break; + } + } + ncounter = (unsigned int)res; + n = (int)res; +#endif + SCTP_HEADER_LEN(recvmbuf6[0]) = n; /* length of total packet */ + SCTP_STAT_INCR(sctps_recvpackets); + SCTP_STAT_INCR_COUNTER64(sctps_inpackets); + + if ((unsigned int)n <= iovlen) { + SCTP_BUF_LEN(recvmbuf6[0]) = n; + (to_fill)++; + } else { + i = 0; + SCTP_BUF_LEN(recvmbuf6[0]) = iovlen; + + ncounter -= min(ncounter, iovlen); + (to_fill)++; + do { + recvmbuf6[i]->m_next = recvmbuf6[i+1]; + SCTP_BUF_LEN(recvmbuf6[i]->m_next) = min(ncounter, iovlen); + i++; + ncounter -= min(ncounter, iovlen); + (to_fill)++; + } while (ncounter > 0); + } + + for (cmsgptr = CMSG_FIRSTHDR(&msg); cmsgptr != NULL; cmsgptr = CMSG_NXTHDR(&msg, cmsgptr)) { + if ((cmsgptr->cmsg_level == IPPROTO_IPV6) && (cmsgptr->cmsg_type == IPV6_PKTINFO)) { + struct in6_pktinfo * info; + + info = (struct in6_pktinfo *)CMSG_DATA(cmsgptr); + memcpy((void *)&dst.sin6_addr, (const void *) &(info->ipi6_addr), sizeof(struct in6_addr)); + break; + } + } + + /* SCTP does not allow broadcasts or multicasts */ + if (IN6_IS_ADDR_MULTICAST(&dst.sin6_addr)) { + m_freem(recvmbuf6[0]); + continue; + } + + offset = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); + if (SCTP_BUF_LEN(recvmbuf6[0]) < offset) { + if ((recvmbuf6[0] = m_pullup(recvmbuf6[0], offset)) == NULL) { + SCTP_STAT_INCR(sctps_hdrops); + continue; + } + } + sh = mtod(recvmbuf6[0], struct sctphdr *); + ch = (struct sctp_chunkhdr *)((caddr_t)sh + sizeof(struct sctphdr)); + offset -= sizeof(struct sctp_chunkhdr); + + dst.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + dst.sin6_len = sizeof(struct sockaddr_in6); +#endif + dst.sin6_port = sh->dest_port; + + src.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + src.sin6_len = sizeof(struct sockaddr_in6); +#endif + src.sin6_port = sh->src_port; + if (SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) && + (memcmp(&src.sin6_addr, &dst.sin6_addr, sizeof(struct in6_addr)) == 0)) { + compute_crc = 0; + SCTP_STAT_INCR(sctps_recvhwcrc); + } else { + SCTP_STAT_INCR(sctps_recvswcrc); + } + SCTPDBG(SCTP_DEBUG_USR, "%s: Received %d bytes.", __func__, n); + SCTPDBG(SCTP_DEBUG_USR, " - calling sctp_common_input_processing with off=%d\n", offset); + sctp_common_input_processing(&recvmbuf6[0], 0, offset, n, + (struct sockaddr *)&src, + (struct sockaddr *)&dst, + sh, ch, + compute_crc, + 0, + SCTP_DEFAULT_VRFID, 0); + if (recvmbuf6[0]) { + m_freem(recvmbuf6[0]); + } + } + for (i = 0; i < MAXLEN_MBUF_CHAIN; i++) { + m_free(recvmbuf6[i]); + } + /* free the array itself */ + free(recvmbuf6); + SCTPDBG(SCTP_DEBUG_USR, "%s: Exiting SCTP/IP6 rcv\n", __func__); + return (NULL); +} +#endif + +#ifdef INET +static void * +recv_function_udp(void *arg) +{ + struct mbuf **udprecvmbuf; + /*Initially the entire set of mbufs is to be allocated. + to_fill indicates this amount. */ + int to_fill = MAXLEN_MBUF_CHAIN; + /* iovlen is the size of each mbuf in the chain */ + int i, n, offset; + unsigned int iovlen = MCLBYTES; + int want_ext = (iovlen > MLEN)? 1 : 0; + int want_header = 0; + struct sctphdr *sh; + uint16_t port; + struct sctp_chunkhdr *ch; + struct sockaddr_in src, dst; +#if defined(IP_PKTINFO) + char cmsgbuf[CMSG_SPACE(sizeof(struct in_pktinfo))]; +#else + char cmsgbuf[CMSG_SPACE(sizeof(struct in_addr))]; +#endif + int compute_crc = 1; +#if !defined(_WIN32) + ssize_t res; + unsigned int ncounter; + struct iovec iov[MAXLEN_MBUF_CHAIN]; + struct msghdr msg; + struct cmsghdr *cmsgptr; +#else + GUID WSARecvMsg_GUID = WSAID_WSARECVMSG; + LPFN_WSARECVMSG WSARecvMsg; + char ControlBuffer[1024]; + WSABUF iov[MAXLEN_MBUF_CHAIN]; + WSAMSG msg; + int nResult, m_ErrorCode; + WSACMSGHDR *cmsgptr; + DWORD ncounter; +#endif + + sctp_userspace_set_threadname("SCTP/UDP/IP4 rcv"); + + udprecvmbuf = malloc(sizeof(struct mbuf *) * MAXLEN_MBUF_CHAIN); + + while (1) { + for (i = 0; i < to_fill; i++) { + /* Not getting the packet header. Tests with chain of one run + as usual without having the packet header. + Have tried both sending and receiving + */ + udprecvmbuf[i] = sctp_get_mbuf_for_msg(iovlen, want_header, M_NOWAIT, want_ext, MT_DATA); +#if !defined(_WIN32) + iov[i].iov_base = (caddr_t)udprecvmbuf[i]->m_data; + iov[i].iov_len = iovlen; +#else + iov[i].buf = (caddr_t)udprecvmbuf[i]->m_data; + iov[i].len = iovlen; +#endif + } + to_fill = 0; +#if !defined(_WIN32) + memset(&msg, 0, sizeof(struct msghdr)); +#else + memset(&msg, 0, sizeof(WSAMSG)); +#endif + memset(&src, 0, sizeof(struct sockaddr_in)); + memset(&dst, 0, sizeof(struct sockaddr_in)); + memset(cmsgbuf, 0, sizeof(cmsgbuf)); + +#if !defined(_WIN32) + msg.msg_name = (void *)&src; + msg.msg_namelen = sizeof(struct sockaddr_in); + msg.msg_iov = iov; + msg.msg_iovlen = MAXLEN_MBUF_CHAIN; + msg.msg_control = (void *)cmsgbuf; + msg.msg_controllen = sizeof(cmsgbuf); + msg.msg_flags = 0; + + res = recvmsg(SCTP_BASE_VAR(userspace_udpsctp), &msg, 0); + if (res < 0) { + if (errno == EAGAIN || errno == EINTR) { + continue; + } else { + break; + } + } + ncounter = (unsigned int)res; + n = (int)res; +#else + nResult = WSAIoctl(SCTP_BASE_VAR(userspace_udpsctp), SIO_GET_EXTENSION_FUNCTION_POINTER, + &WSARecvMsg_GUID, sizeof WSARecvMsg_GUID, + &WSARecvMsg, sizeof WSARecvMsg, + &ncounter, NULL, NULL); + if (nResult == 0) { + msg.name = (void *)&src; + msg.namelen = sizeof(struct sockaddr_in); + msg.lpBuffers = iov; + msg.dwBufferCount = MAXLEN_MBUF_CHAIN; + msg.Control.len = sizeof ControlBuffer; + msg.Control.buf = ControlBuffer; + msg.dwFlags = 0; + nResult = WSARecvMsg(SCTP_BASE_VAR(userspace_udpsctp), &msg, &ncounter, NULL, NULL); + } + if (nResult != 0) { + m_ErrorCode = WSAGetLastError(); + if ((m_ErrorCode == WSAENOTSOCK) || (m_ErrorCode == WSAEINTR)) { + break; + } + continue; + } + n = ncounter; +#endif + SCTP_HEADER_LEN(udprecvmbuf[0]) = n; /* length of total packet */ + SCTP_STAT_INCR(sctps_recvpackets); + SCTP_STAT_INCR_COUNTER64(sctps_inpackets); + + if ((unsigned int)n <= iovlen) { + SCTP_BUF_LEN(udprecvmbuf[0]) = n; + (to_fill)++; + } else { + i = 0; + SCTP_BUF_LEN(udprecvmbuf[0]) = iovlen; + + ncounter -= min(ncounter, iovlen); + (to_fill)++; + do { + udprecvmbuf[i]->m_next = udprecvmbuf[i+1]; + SCTP_BUF_LEN(udprecvmbuf[i]->m_next) = min(ncounter, iovlen); + i++; + ncounter -= min(ncounter, iovlen); + (to_fill)++; + } while (ncounter > 0); + } + + for (cmsgptr = CMSG_FIRSTHDR(&msg); cmsgptr != NULL; cmsgptr = CMSG_NXTHDR(&msg, cmsgptr)) { +#if defined(IP_PKTINFO) + if ((cmsgptr->cmsg_level == IPPROTO_IP) && (cmsgptr->cmsg_type == IP_PKTINFO)) { + struct in_pktinfo *info; + + dst.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + dst.sin_len = sizeof(struct sockaddr_in); +#endif + info = (struct in_pktinfo *)CMSG_DATA(cmsgptr); + memcpy((void *)&dst.sin_addr, (const void *)&(info->ipi_addr), sizeof(struct in_addr)); + break; + } +#else + if ((cmsgptr->cmsg_level == IPPROTO_IP) && (cmsgptr->cmsg_type == IP_RECVDSTADDR)) { + struct in_addr *addr; + + dst.sin_family = AF_INET; +#ifdef HAVE_SIN_LEN + dst.sin_len = sizeof(struct sockaddr_in); +#endif + addr = (struct in_addr *)CMSG_DATA(cmsgptr); + memcpy((void *)&dst.sin_addr, (const void *)addr, sizeof(struct in_addr)); + break; + } +#endif + } + + /* SCTP does not allow broadcasts or multicasts */ + if (IN_MULTICAST(ntohl(dst.sin_addr.s_addr))) { + m_freem(udprecvmbuf[0]); + continue; + } + if (SCTP_IS_IT_BROADCAST(dst.sin_addr, udprecvmbuf[0])) { + m_freem(udprecvmbuf[0]); + continue; + } + + offset = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); + if (SCTP_BUF_LEN(udprecvmbuf[0]) < offset) { + if ((udprecvmbuf[0] = m_pullup(udprecvmbuf[0], offset)) == NULL) { + SCTP_STAT_INCR(sctps_hdrops); + continue; + } + } + sh = mtod(udprecvmbuf[0], struct sctphdr *); + ch = (struct sctp_chunkhdr *)((caddr_t)sh + sizeof(struct sctphdr)); + offset -= sizeof(struct sctp_chunkhdr); + + port = src.sin_port; + src.sin_port = sh->src_port; + dst.sin_port = sh->dest_port; + if (SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) && + (src.sin_addr.s_addr == dst.sin_addr.s_addr)) { + compute_crc = 0; + SCTP_STAT_INCR(sctps_recvhwcrc); + } else { + SCTP_STAT_INCR(sctps_recvswcrc); + } + SCTPDBG(SCTP_DEBUG_USR, "%s: Received %d bytes.", __func__, n); + SCTPDBG(SCTP_DEBUG_USR, " - calling sctp_common_input_processing with off=%d\n", offset); + sctp_common_input_processing(&udprecvmbuf[0], 0, offset, n, + (struct sockaddr *)&src, + (struct sockaddr *)&dst, + sh, ch, + compute_crc, + 0, + SCTP_DEFAULT_VRFID, port); + if (udprecvmbuf[0]) { + m_freem(udprecvmbuf[0]); + } + } + for (i = 0; i < MAXLEN_MBUF_CHAIN; i++) { + m_free(udprecvmbuf[i]); + } + /* free the array itself */ + free(udprecvmbuf); + SCTPDBG(SCTP_DEBUG_USR, "%s: Exiting SCTP/UDP/IP4 rcv\n", __func__); + return (NULL); +} +#endif + +#if defined(INET6) +static void * +recv_function_udp6(void *arg) +{ + struct mbuf **udprecvmbuf6; + /*Initially the entire set of mbufs is to be allocated. + to_fill indicates this amount. */ + int to_fill = MAXLEN_MBUF_CHAIN; + /* iovlen is the size of each mbuf in the chain */ + int i, n, offset; + unsigned int iovlen = MCLBYTES; + int want_ext = (iovlen > MLEN)? 1 : 0; + int want_header = 0; + struct sockaddr_in6 src, dst; + struct sctphdr *sh; + uint16_t port; + struct sctp_chunkhdr *ch; + char cmsgbuf[CMSG_SPACE(sizeof (struct in6_pktinfo))]; + int compute_crc = 1; +#if !defined(_WIN32) + struct iovec iov[MAXLEN_MBUF_CHAIN]; + struct msghdr msg; + struct cmsghdr *cmsgptr; + ssize_t res; + unsigned int ncounter; +#else + GUID WSARecvMsg_GUID = WSAID_WSARECVMSG; + LPFN_WSARECVMSG WSARecvMsg; + char ControlBuffer[1024]; + WSABUF iov[MAXLEN_MBUF_CHAIN]; + WSAMSG msg; + int nResult, m_ErrorCode; + WSACMSGHDR *cmsgptr; + DWORD ncounter; +#endif + + sctp_userspace_set_threadname("SCTP/UDP/IP6 rcv"); + + udprecvmbuf6 = malloc(sizeof(struct mbuf *) * MAXLEN_MBUF_CHAIN); + while (1) { + for (i = 0; i < to_fill; i++) { + /* Not getting the packet header. Tests with chain of one run + as usual without having the packet header. + Have tried both sending and receiving + */ + udprecvmbuf6[i] = sctp_get_mbuf_for_msg(iovlen, want_header, M_NOWAIT, want_ext, MT_DATA); +#if !defined(_WIN32) + iov[i].iov_base = (caddr_t)udprecvmbuf6[i]->m_data; + iov[i].iov_len = iovlen; +#else + iov[i].buf = (caddr_t)udprecvmbuf6[i]->m_data; + iov[i].len = iovlen; +#endif + } + to_fill = 0; + +#if !defined(_WIN32) + memset(&msg, 0, sizeof(struct msghdr)); +#else + memset(&msg, 0, sizeof(WSAMSG)); +#endif + memset(&src, 0, sizeof(struct sockaddr_in6)); + memset(&dst, 0, sizeof(struct sockaddr_in6)); + memset(cmsgbuf, 0, CMSG_SPACE(sizeof (struct in6_pktinfo))); + +#if !defined(_WIN32) + msg.msg_name = (void *)&src; + msg.msg_namelen = sizeof(struct sockaddr_in6); + msg.msg_iov = iov; + msg.msg_iovlen = MAXLEN_MBUF_CHAIN; + msg.msg_control = (void *)cmsgbuf; + msg.msg_controllen = (socklen_t)CMSG_SPACE(sizeof (struct in6_pktinfo)); + msg.msg_flags = 0; + + res = recvmsg(SCTP_BASE_VAR(userspace_udpsctp6), &msg, 0); + if (res < 0) { + if (errno == EAGAIN || errno == EINTR) { + continue; + } else { + break; + } + } + ncounter = (unsigned int)res; + n = (int)res; +#else + nResult = WSAIoctl(SCTP_BASE_VAR(userspace_udpsctp6), SIO_GET_EXTENSION_FUNCTION_POINTER, + &WSARecvMsg_GUID, sizeof WSARecvMsg_GUID, + &WSARecvMsg, sizeof WSARecvMsg, + &ncounter, NULL, NULL); + if (nResult == SOCKET_ERROR) { + m_ErrorCode = WSAGetLastError(); + WSARecvMsg = NULL; + } + if (nResult == 0) { + msg.name = (void *)&src; + msg.namelen = sizeof(struct sockaddr_in6); + msg.lpBuffers = iov; + msg.dwBufferCount = MAXLEN_MBUF_CHAIN; + msg.Control.len = sizeof ControlBuffer; + msg.Control.buf = ControlBuffer; + msg.dwFlags = 0; + nResult = WSARecvMsg(SCTP_BASE_VAR(userspace_udpsctp6), &msg, &ncounter, NULL, NULL); + } + if (nResult != 0) { + m_ErrorCode = WSAGetLastError(); + if ((m_ErrorCode == WSAENOTSOCK) || (m_ErrorCode == WSAEINTR)) { + break; + } + continue; + } + n = ncounter; +#endif + SCTP_HEADER_LEN(udprecvmbuf6[0]) = n; /* length of total packet */ + SCTP_STAT_INCR(sctps_recvpackets); + SCTP_STAT_INCR_COUNTER64(sctps_inpackets); + + if ((unsigned int)n <= iovlen) { + SCTP_BUF_LEN(udprecvmbuf6[0]) = n; + (to_fill)++; + } else { + i = 0; + SCTP_BUF_LEN(udprecvmbuf6[0]) = iovlen; + + ncounter -= min(ncounter, iovlen); + (to_fill)++; + do { + udprecvmbuf6[i]->m_next = udprecvmbuf6[i+1]; + SCTP_BUF_LEN(udprecvmbuf6[i]->m_next) = min(ncounter, iovlen); + i++; + ncounter -= min(ncounter, iovlen); + (to_fill)++; + } while (ncounter > 0); + } + + for (cmsgptr = CMSG_FIRSTHDR(&msg); cmsgptr != NULL; cmsgptr = CMSG_NXTHDR(&msg, cmsgptr)) { + if ((cmsgptr->cmsg_level == IPPROTO_IPV6) && (cmsgptr->cmsg_type == IPV6_PKTINFO)) { + struct in6_pktinfo *info; + + dst.sin6_family = AF_INET6; +#ifdef HAVE_SIN6_LEN + dst.sin6_len = sizeof(struct sockaddr_in6); +#endif + info = (struct in6_pktinfo *)CMSG_DATA(cmsgptr); + /*dst.sin6_port = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port));*/ + memcpy((void *)&dst.sin6_addr, (const void *)&(info->ipi6_addr), sizeof(struct in6_addr)); + } + } + + /* SCTP does not allow broadcasts or multicasts */ + if (IN6_IS_ADDR_MULTICAST(&dst.sin6_addr)) { + m_freem(udprecvmbuf6[0]); + continue; + } + + offset = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); + if (SCTP_BUF_LEN(udprecvmbuf6[0]) < offset) { + if ((udprecvmbuf6[0] = m_pullup(udprecvmbuf6[0], offset)) == NULL) { + SCTP_STAT_INCR(sctps_hdrops); + continue; + } + } + sh = mtod(udprecvmbuf6[0], struct sctphdr *); + ch = (struct sctp_chunkhdr *)((caddr_t)sh + sizeof(struct sctphdr)); + offset -= sizeof(struct sctp_chunkhdr); + + port = src.sin6_port; + src.sin6_port = sh->src_port; + dst.sin6_port = sh->dest_port; + if (SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) && + (memcmp(&src.sin6_addr, &dst.sin6_addr, sizeof(struct in6_addr)) == 0)) { + compute_crc = 0; + SCTP_STAT_INCR(sctps_recvhwcrc); + } else { + SCTP_STAT_INCR(sctps_recvswcrc); + } + SCTPDBG(SCTP_DEBUG_USR, "%s: Received %d bytes.", __func__, n); + SCTPDBG(SCTP_DEBUG_USR, " - calling sctp_common_input_processing with off=%d\n", (int)sizeof(struct sctphdr)); + sctp_common_input_processing(&udprecvmbuf6[0], 0, offset, n, + (struct sockaddr *)&src, + (struct sockaddr *)&dst, + sh, ch, + compute_crc, + 0, + SCTP_DEFAULT_VRFID, port); + if (udprecvmbuf6[0]) { + m_freem(udprecvmbuf6[0]); + } + } + for (i = 0; i < MAXLEN_MBUF_CHAIN; i++) { + m_free(udprecvmbuf6[i]); + } + /* free the array itself */ + free(udprecvmbuf6); + SCTPDBG(SCTP_DEBUG_USR, "%s: Exiting SCTP/UDP/IP6 rcv\n", __func__); + return (NULL); +} +#endif + +#if defined(_WIN32) +static void +setReceiveBufferSize(SOCKET sfd, int new_size) +#else +static void +setReceiveBufferSize(int sfd, int new_size) +#endif +{ + int ch = new_size; + + if (setsockopt (sfd, SOL_SOCKET, SO_RCVBUF, (void*)&ch, sizeof(ch)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set recv-buffers size (errno = %d).\n", WSAGetLastError()); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set recv-buffers size (errno = %d).\n", errno); +#endif + } + return; +} + +#if defined(_WIN32) +static void +setSendBufferSize(SOCKET sfd, int new_size) +#else +static void +setSendBufferSize(int sfd, int new_size) +#endif +{ + int ch = new_size; + + if (setsockopt (sfd, SOL_SOCKET, SO_SNDBUF, (void*)&ch, sizeof(ch)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set send-buffers size (errno = %d).\n", WSAGetLastError()); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set send-buffers size (errno = %d).\n", errno); +#endif + } + return; +} + +#define SOCKET_TIMEOUT 100 /* in ms */ +void +recv_thread_init(void) +{ +#if defined(INET) + struct sockaddr_in addr_ipv4; + const int hdrincl = 1; +#endif +#if defined(INET6) + struct sockaddr_in6 addr_ipv6; +#endif +#if defined(INET) || defined(INET6) + const int on = 1; +#endif +#if !defined(_WIN32) + struct timeval timeout; + + memset(&timeout, 0, sizeof(struct timeval)); + timeout.tv_sec = (SOCKET_TIMEOUT / 1000); + timeout.tv_usec = (SOCKET_TIMEOUT % 1000) * 1000; +#else + unsigned int timeout = SOCKET_TIMEOUT; /* Timeout in milliseconds */ +#endif +#if defined(__APPLE__) || defined(__DragonFly__) || defined(__FreeBSD__) + if (SCTP_BASE_VAR(userspace_route) == -1) { + if ((SCTP_BASE_VAR(userspace_route) = socket(AF_ROUTE, SOCK_RAW, 0)) == -1) { + SCTPDBG(SCTP_DEBUG_USR, "Can't create routing socket (errno = %d).\n", errno); + } +#if 0 + struct sockaddr_nl sanl; + + if ((SCTP_BASE_VAR(userspace_route) = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE)) < 0) { + SCTPDBG(SCTP_DEBUG_USR, "Can't create routing socket (errno = %d.\n", errno); + } + memset(&sanl, 0, sizeof(sanl)); + sanl.nl_family = AF_NETLINK; + sanl.nl_groups = 0; +#ifdef INET + sanl.nl_groups |= RTMGRP_IPV4_IFADDR; +#endif +#ifdef INET6 + sanl.nl_groups |= RTMGRP_IPV6_IFADDR; +#endif + if (bind(SCTP_BASE_VAR(userspace_route), (struct sockaddr *) &sanl, sizeof(sanl)) < 0) { + SCTPDBG(SCTP_DEBUG_USR, "Can't bind routing socket (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_route)); + SCTP_BASE_VAR(userspace_route) = -1; + } +#endif + if (SCTP_BASE_VAR(userspace_route) != -1) { + if (setsockopt(SCTP_BASE_VAR(userspace_route), SOL_SOCKET, SO_RCVTIMEO,(const void*)&timeout, sizeof(struct timeval)) < 0) { + SCTPDBG(SCTP_DEBUG_USR, "Can't set timeout on routing socket (errno = %d).\n", errno); +#if defined(_WIN32) + closesocket(SCTP_BASE_VAR(userspace_route)); +#else + close(SCTP_BASE_VAR(userspace_route)); +#endif + SCTP_BASE_VAR(userspace_route) = -1; + } + } + } +#endif +#if defined(INET) + if (SCTP_BASE_VAR(userspace_rawsctp) == -1) { + if ((SCTP_BASE_VAR(userspace_rawsctp) = socket(AF_INET, SOCK_RAW, IPPROTO_SCTP)) == -1) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't create raw socket for IPv4 (errno = %d).\n", WSAGetLastError()); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't create raw socket for IPv4 (errno = %d).\n", errno); +#endif + } else { + /* complete setting up the raw SCTP socket */ + if (setsockopt(SCTP_BASE_VAR(userspace_rawsctp), IPPROTO_IP, IP_HDRINCL,(const void*)&hdrincl, sizeof(int)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set IP_HDRINCL (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_rawsctp)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set IP_HDRINCL (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_rawsctp)); +#endif + SCTP_BASE_VAR(userspace_rawsctp) = -1; + } else if (setsockopt(SCTP_BASE_VAR(userspace_rawsctp), SOL_SOCKET, SO_RCVTIMEO, (const void *)&timeout, sizeof(timeout)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set timeout on socket for SCTP/IPv4 (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_rawsctp)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set timeout on socket for SCTP/IPv4 (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_rawsctp)); +#endif + SCTP_BASE_VAR(userspace_rawsctp) = -1; + } else { + memset((void *)&addr_ipv4, 0, sizeof(struct sockaddr_in)); +#ifdef HAVE_SIN_LEN + addr_ipv4.sin_len = sizeof(struct sockaddr_in); +#endif + addr_ipv4.sin_family = AF_INET; + addr_ipv4.sin_port = htons(0); + addr_ipv4.sin_addr.s_addr = htonl(INADDR_ANY); + if (bind(SCTP_BASE_VAR(userspace_rawsctp), (const struct sockaddr *)&addr_ipv4, sizeof(struct sockaddr_in)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't bind socket for SCTP/IPv4 (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_rawsctp)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't bind socket for SCTP/IPv4 (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_rawsctp)); +#endif + SCTP_BASE_VAR(userspace_rawsctp) = -1; + } else { + setReceiveBufferSize(SCTP_BASE_VAR(userspace_rawsctp), SB_RAW); /* 128K */ + setSendBufferSize(SCTP_BASE_VAR(userspace_rawsctp), SB_RAW); /* 128K Is this setting net.inet.raw.maxdgram value? Should it be set to 64K? */ + } + } + } + } + if ((SCTP_BASE_VAR(userspace_udpsctp) == -1) && (SCTP_BASE_SYSCTL(sctp_udp_tunneling_port) != 0)) { + if ((SCTP_BASE_VAR(userspace_udpsctp) = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP)) == -1) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't create socket for SCTP/UDP/IPv4 (errno = %d).\n", WSAGetLastError()); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't create socket for SCTP/UDP/IPv4 (errno = %d).\n", errno); +#endif + } else { +#if defined(IP_PKTINFO) + if (setsockopt(SCTP_BASE_VAR(userspace_udpsctp), IPPROTO_IP, IP_PKTINFO, (const void *)&on, (int)sizeof(int)) < 0) { +#else + if (setsockopt(SCTP_BASE_VAR(userspace_udpsctp), IPPROTO_IP, IP_RECVDSTADDR, (const void *)&on, (int)sizeof(int)) < 0) { +#endif +#if defined(_WIN32) +#if defined(IP_PKTINFO) + SCTPDBG(SCTP_DEBUG_USR, "Can't set IP_PKTINFO on socket for SCTP/UDP/IPv4 (errno = %d).\n", WSAGetLastError()); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set IP_RECVDSTADDR on socket for SCTP/UDP/IPv4 (errno = %d).\n", WSAGetLastError()); +#endif + closesocket(SCTP_BASE_VAR(userspace_udpsctp)); +#else +#if defined(IP_PKTINFO) + SCTPDBG(SCTP_DEBUG_USR, "Can't set IP_PKTINFO on socket for SCTP/UDP/IPv4 (errno = %d).\n", errno); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set IP_RECVDSTADDR on socket for SCTP/UDP/IPv4 (errno = %d).\n", errno); +#endif + close(SCTP_BASE_VAR(userspace_udpsctp)); +#endif + SCTP_BASE_VAR(userspace_udpsctp) = -1; + } else if (setsockopt(SCTP_BASE_VAR(userspace_udpsctp), SOL_SOCKET, SO_RCVTIMEO, (const void *)&timeout, sizeof(timeout)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set timeout on socket for SCTP/UDP/IPv4 (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_udpsctp)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set timeout on socket for SCTP/UDP/IPv4 (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_udpsctp)); +#endif + SCTP_BASE_VAR(userspace_udpsctp) = -1; + } else { + memset((void *)&addr_ipv4, 0, sizeof(struct sockaddr_in)); +#ifdef HAVE_SIN_LEN + addr_ipv4.sin_len = sizeof(struct sockaddr_in); +#endif + addr_ipv4.sin_family = AF_INET; + addr_ipv4.sin_port = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); + addr_ipv4.sin_addr.s_addr = htonl(INADDR_ANY); + if (bind(SCTP_BASE_VAR(userspace_udpsctp), (const struct sockaddr *)&addr_ipv4, sizeof(struct sockaddr_in)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't bind socket for SCTP/UDP/IPv4 (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_udpsctp)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't bind socket for SCTP/UDP/IPv4 (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_udpsctp)); +#endif + SCTP_BASE_VAR(userspace_udpsctp) = -1; + } else { + setReceiveBufferSize(SCTP_BASE_VAR(userspace_udpsctp), SB_RAW); /* 128K */ + setSendBufferSize(SCTP_BASE_VAR(userspace_udpsctp), SB_RAW); /* 128K Is this setting net.inet.raw.maxdgram value? Should it be set to 64K? */ + } + } + } + } +#endif +#if defined(INET6) + if (SCTP_BASE_VAR(userspace_rawsctp6) == -1) { + if ((SCTP_BASE_VAR(userspace_rawsctp6) = socket(AF_INET6, SOCK_RAW, IPPROTO_SCTP)) == -1) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't create socket for SCTP/IPv6 (errno = %d).\n", WSAGetLastError()); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't create socket for SCTP/IPv6 (errno = %d).\n", errno); +#endif + } else { + /* complete setting up the raw SCTP socket */ +#if defined(IPV6_RECVPKTINFO) + if (setsockopt(SCTP_BASE_VAR(userspace_rawsctp6), IPPROTO_IPV6, IPV6_RECVPKTINFO, (const void *)&on, sizeof(on)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set IPV6_RECVPKTINFO on socket for SCTP/IPv6 (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_rawsctp6)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set IPV6_RECVPKTINFO on socket for SCTP/IPv6 (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_rawsctp6)); +#endif + SCTP_BASE_VAR(userspace_rawsctp6) = -1; + } else { +#else + if (setsockopt(SCTP_BASE_VAR(userspace_rawsctp6), IPPROTO_IPV6, IPV6_PKTINFO,(const void*)&on, sizeof(on)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set IPV6_PKTINFO on socket for SCTP/IPv6 (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_rawsctp6)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set IPV6_PKTINFO on socket for SCTP/IPv6 (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_rawsctp6)); +#endif + SCTP_BASE_VAR(userspace_rawsctp6) = -1; + } else { +#endif + if (setsockopt(SCTP_BASE_VAR(userspace_rawsctp6), IPPROTO_IPV6, IPV6_V6ONLY, (const void*)&on, (socklen_t)sizeof(on)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set IPV6_V6ONLY on socket for SCTP/IPv6 (errno = %d).\n", WSAGetLastError()); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set IPV6_V6ONLY on socket for SCTP/IPv6 (errno = %d).\n", errno); +#endif + } + if (setsockopt(SCTP_BASE_VAR(userspace_rawsctp6), SOL_SOCKET, SO_RCVTIMEO, (const void *)&timeout, sizeof(timeout)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set timeout on socket for SCTP/IPv6 (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_rawsctp6)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set timeout on socket for SCTP/IPv6 (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_rawsctp6)); +#endif + SCTP_BASE_VAR(userspace_rawsctp6) = -1; + } else { + memset((void *)&addr_ipv6, 0, sizeof(struct sockaddr_in6)); +#ifdef HAVE_SIN6_LEN + addr_ipv6.sin6_len = sizeof(struct sockaddr_in6); +#endif + addr_ipv6.sin6_family = AF_INET6; + addr_ipv6.sin6_port = htons(0); + addr_ipv6.sin6_addr = in6addr_any; + if (bind(SCTP_BASE_VAR(userspace_rawsctp6), (const struct sockaddr *)&addr_ipv6, sizeof(struct sockaddr_in6)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't bind socket for SCTP/IPv6 (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_rawsctp6)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't bind socket for SCTP/IPv6 (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_rawsctp6)); +#endif + SCTP_BASE_VAR(userspace_rawsctp6) = -1; + } else { + setReceiveBufferSize(SCTP_BASE_VAR(userspace_rawsctp6), SB_RAW); /* 128K */ + setSendBufferSize(SCTP_BASE_VAR(userspace_rawsctp6), SB_RAW); /* 128K Is this setting net.inet.raw.maxdgram value? Should it be set to 64K? */ + } + } + } + } + } + if ((SCTP_BASE_VAR(userspace_udpsctp6) == -1) && (SCTP_BASE_SYSCTL(sctp_udp_tunneling_port) != 0)) { + if ((SCTP_BASE_VAR(userspace_udpsctp6) = socket(AF_INET6, SOCK_DGRAM, IPPROTO_UDP)) == -1) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't create socket for SCTP/UDP/IPv6 (errno = %d).\n", WSAGetLastError()); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't create socket for SCTP/UDP/IPv6 (errno = %d).\n", errno); +#endif + } +#if defined(IPV6_RECVPKTINFO) + if (setsockopt(SCTP_BASE_VAR(userspace_udpsctp6), IPPROTO_IPV6, IPV6_RECVPKTINFO, (const void *)&on, (int)sizeof(int)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set IPV6_RECVPKTINFO on socket for SCTP/UDP/IPv6 (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_udpsctp6)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set IPV6_RECVPKTINFO on socket for SCTP/UDP/IPv6 (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_udpsctp6)); +#endif + SCTP_BASE_VAR(userspace_udpsctp6) = -1; + } else { +#else + if (setsockopt(SCTP_BASE_VAR(userspace_udpsctp6), IPPROTO_IPV6, IPV6_PKTINFO, (const void *)&on, (int)sizeof(int)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set IPV6_PKTINFO on socket for SCTP/UDP/IPv6 (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_udpsctp6)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set IPV6_PKTINFO on socket for SCTP/UDP/IPv6 (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_udpsctp6)); +#endif + SCTP_BASE_VAR(userspace_udpsctp6) = -1; + } else { +#endif + if (setsockopt(SCTP_BASE_VAR(userspace_udpsctp6), IPPROTO_IPV6, IPV6_V6ONLY, (const void *)&on, (socklen_t)sizeof(on)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set IPV6_V6ONLY on socket for SCTP/UDP/IPv6 (errno = %d).\n", WSAGetLastError()); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set IPV6_V6ONLY on socket for SCTP/UDP/IPv6 (errno = %d).\n", errno); +#endif + } + if (setsockopt(SCTP_BASE_VAR(userspace_udpsctp6), SOL_SOCKET, SO_RCVTIMEO, (const void *)&timeout, sizeof(timeout)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't set timeout on socket for SCTP/UDP/IPv6 (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_udpsctp6)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't set timeout on socket for SCTP/UDP/IPv6 (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_udpsctp6)); +#endif + SCTP_BASE_VAR(userspace_udpsctp6) = -1; + } else { + memset((void *)&addr_ipv6, 0, sizeof(struct sockaddr_in6)); +#ifdef HAVE_SIN6_LEN + addr_ipv6.sin6_len = sizeof(struct sockaddr_in6); +#endif + addr_ipv6.sin6_family = AF_INET6; + addr_ipv6.sin6_port = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); + addr_ipv6.sin6_addr = in6addr_any; + if (bind(SCTP_BASE_VAR(userspace_udpsctp6), (const struct sockaddr *)&addr_ipv6, sizeof(struct sockaddr_in6)) < 0) { +#if defined(_WIN32) + SCTPDBG(SCTP_DEBUG_USR, "Can't bind socket for SCTP/UDP/IPv6 (errno = %d).\n", WSAGetLastError()); + closesocket(SCTP_BASE_VAR(userspace_udpsctp6)); +#else + SCTPDBG(SCTP_DEBUG_USR, "Can't bind socket for SCTP/UDP/IPv6 (errno = %d).\n", errno); + close(SCTP_BASE_VAR(userspace_udpsctp6)); +#endif + SCTP_BASE_VAR(userspace_udpsctp6) = -1; + } else { + setReceiveBufferSize(SCTP_BASE_VAR(userspace_udpsctp6), SB_RAW); /* 128K */ + setSendBufferSize(SCTP_BASE_VAR(userspace_udpsctp6), SB_RAW); /* 128K Is this setting net.inet.raw.maxdgram value? Should it be set to 64K? */ + } + } + } + } +#endif +#if defined(__APPLE__) || defined(__DragonFly__) || defined(__FreeBSD__) +#if defined(INET) || defined(INET6) + if (SCTP_BASE_VAR(userspace_route) != -1) { + int rc; + + if ((rc = sctp_userspace_thread_create(&SCTP_BASE_VAR(recvthreadroute), &recv_function_route))) { + SCTPDBG(SCTP_DEBUG_USR, "Can't start routing thread (%d).\n", rc); + close(SCTP_BASE_VAR(userspace_route)); + SCTP_BASE_VAR(userspace_route) = -1; + } + } +#endif +#endif +#if defined(INET) + if (SCTP_BASE_VAR(userspace_rawsctp) != -1) { + int rc; + + if ((rc = sctp_userspace_thread_create(&SCTP_BASE_VAR(recvthreadraw), &recv_function_raw))) { + SCTPDBG(SCTP_DEBUG_USR, "Can't start SCTP/IPv4 recv thread (%d).\n", rc); +#if defined(_WIN32) + closesocket(SCTP_BASE_VAR(userspace_rawsctp)); +#else + close(SCTP_BASE_VAR(userspace_rawsctp)); +#endif + SCTP_BASE_VAR(userspace_rawsctp) = -1; + } + } + if (SCTP_BASE_VAR(userspace_udpsctp) != -1) { + int rc; + + if ((rc = sctp_userspace_thread_create(&SCTP_BASE_VAR(recvthreadudp), &recv_function_udp))) { + SCTPDBG(SCTP_DEBUG_USR, "Can't start SCTP/UDP/IPv4 recv thread (%d).\n", rc); +#if defined(_WIN32) + closesocket(SCTP_BASE_VAR(userspace_udpsctp)); +#else + close(SCTP_BASE_VAR(userspace_udpsctp)); +#endif + SCTP_BASE_VAR(userspace_udpsctp) = -1; + } + } +#endif +#if defined(INET6) + if (SCTP_BASE_VAR(userspace_rawsctp6) != -1) { + int rc; + + if ((rc = sctp_userspace_thread_create(&SCTP_BASE_VAR(recvthreadraw6), &recv_function_raw6))) { + SCTPDBG(SCTP_DEBUG_USR, "Can't start SCTP/IPv6 recv thread (%d).\n", rc); +#if defined(_WIN32) + closesocket(SCTP_BASE_VAR(userspace_rawsctp6)); +#else + close(SCTP_BASE_VAR(userspace_rawsctp6)); +#endif + SCTP_BASE_VAR(userspace_rawsctp6) = -1; + } + } + if (SCTP_BASE_VAR(userspace_udpsctp6) != -1) { + int rc; + + if ((rc = sctp_userspace_thread_create(&SCTP_BASE_VAR(recvthreadudp6), &recv_function_udp6))) { + SCTPDBG(SCTP_DEBUG_USR, "Can't start SCTP/UDP/IPv6 recv thread (%d).\n", rc); +#if defined(_WIN32) + closesocket(SCTP_BASE_VAR(userspace_udpsctp6)); +#else + close(SCTP_BASE_VAR(userspace_udpsctp6)); +#endif + SCTP_BASE_VAR(userspace_udpsctp6) = -1; + } + } +#endif +} + +void +recv_thread_destroy(void) +{ +#if defined(__APPLE__) || defined(__DragonFly__) || defined(__FreeBSD__) +#if defined(INET) || defined(INET6) + if (SCTP_BASE_VAR(userspace_route) != -1) { + close(SCTP_BASE_VAR(userspace_route)); + pthread_join(SCTP_BASE_VAR(recvthreadroute), NULL); + } +#endif +#endif +#if defined(INET) + if (SCTP_BASE_VAR(userspace_rawsctp) != -1) { +#if defined(_WIN32) + closesocket(SCTP_BASE_VAR(userspace_rawsctp)); + SCTP_BASE_VAR(userspace_rawsctp) = -1; + WaitForSingleObject(SCTP_BASE_VAR(recvthreadraw), INFINITE); + CloseHandle(SCTP_BASE_VAR(recvthreadraw)); +#else + close(SCTP_BASE_VAR(userspace_rawsctp)); + SCTP_BASE_VAR(userspace_rawsctp) = -1; + pthread_join(SCTP_BASE_VAR(recvthreadraw), NULL); +#endif + } + if (SCTP_BASE_VAR(userspace_udpsctp) != -1) { +#if defined(_WIN32) + closesocket(SCTP_BASE_VAR(userspace_udpsctp)); + SCTP_BASE_VAR(userspace_udpsctp) = -1; + WaitForSingleObject(SCTP_BASE_VAR(recvthreadudp), INFINITE); + CloseHandle(SCTP_BASE_VAR(recvthreadudp)); +#else + close(SCTP_BASE_VAR(userspace_udpsctp)); + SCTP_BASE_VAR(userspace_udpsctp) = -1; + pthread_join(SCTP_BASE_VAR(recvthreadudp), NULL); +#endif + } +#endif +#if defined(INET6) + if (SCTP_BASE_VAR(userspace_rawsctp6) != -1) { +#if defined(_WIN32) + closesocket(SCTP_BASE_VAR(userspace_rawsctp6)); + SCTP_BASE_VAR(userspace_rawsctp6) = -1; + WaitForSingleObject(SCTP_BASE_VAR(recvthreadraw6), INFINITE); + CloseHandle(SCTP_BASE_VAR(recvthreadraw6)); +#else + close(SCTP_BASE_VAR(userspace_rawsctp6)); + SCTP_BASE_VAR(userspace_rawsctp6) = -1; + pthread_join(SCTP_BASE_VAR(recvthreadraw6), NULL); +#endif + } + if (SCTP_BASE_VAR(userspace_udpsctp6) != -1) { +#if defined(_WIN32) + SCTP_BASE_VAR(userspace_udpsctp6) = -1; + closesocket(SCTP_BASE_VAR(userspace_udpsctp6)); + WaitForSingleObject(SCTP_BASE_VAR(recvthreadudp6), INFINITE); + CloseHandle(SCTP_BASE_VAR(recvthreadudp6)); +#else + close(SCTP_BASE_VAR(userspace_udpsctp6)); + SCTP_BASE_VAR(userspace_udpsctp6) = -1; + pthread_join(SCTP_BASE_VAR(recvthreadudp6), NULL); +#endif + } +#endif +} +#else +int foo; +#endif diff --git a/netwerk/sctp/src/user_recv_thread.h b/netwerk/sctp/src/user_recv_thread.h new file mode 100644 index 0000000000..61cdf17211 --- /dev/null +++ b/netwerk/sctp/src/user_recv_thread.h @@ -0,0 +1,34 @@ +/*- + * Copyright (c) 2012 Michael Tuexen + * Copyright (c) 2012 Irene Ruengeler + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +#ifndef _USER_RECV_THREAD_H_ +#define _USER_RECV_THREAD_H_ + +void recv_thread_init(void); +void recv_thread_destroy(void); + +#endif diff --git a/netwerk/sctp/src/user_route.h b/netwerk/sctp/src/user_route.h new file mode 100644 index 0000000000..4abf2eac98 --- /dev/null +++ b/netwerk/sctp/src/user_route.h @@ -0,0 +1,130 @@ +/*- + * Copyright (c) 1980, 1986, 1993 + * The Regents of the University of California. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + */ + +#ifndef _USER_ROUTE_H_ +#define _USER_ROUTE_H_ + +/* + * Kernel resident routing tables. + * + * The routing tables are initialized when interface addresses + * are set by making entries for all directly connected interfaces. + */ + +/* + * A route consists of a destination address and a reference + * to a routing entry. These are often held by protocols + * in their control blocks, e.g. inpcb. + */ + +struct sctp_route { + struct sctp_rtentry *ro_rt; + struct sockaddr ro_dst; +}; + +/* + * These numbers are used by reliable protocols for determining + * retransmission behavior and are included in the routing structure. + */ +struct sctp_rt_metrics_lite { + uint32_t rmx_mtu; /* MTU for this path */ +#if 0 + u_long rmx_expire; /* lifetime for route, e.g. redirect */ + u_long rmx_pksent; /* packets sent using this route */ +#endif +}; + +/* + * We distinguish between routes to hosts and routes to networks, + * preferring the former if available. For each route we infer + * the interface to use from the gateway address supplied when + * the route was entered. Routes that forward packets through + * gateways are marked so that the output routines know to address the + * gateway rather than the ultimate destination. + */ +struct sctp_rtentry { +#if 0 + struct radix_node rt_nodes[2]; /* tree glue, and other values */ + /* + * XXX struct rtentry must begin with a struct radix_node (or two!) + * because the code does some casts of a 'struct radix_node *' + * to a 'struct rtentry *' + */ +#define rt_key(r) (*((struct sockaddr **)(&(r)->rt_nodes->rn_key))) +#define rt_mask(r) (*((struct sockaddr **)(&(r)->rt_nodes->rn_mask))) + struct sockaddr *rt_gateway; /* value */ + u_long rt_flags; /* up/down?, host/net */ +#endif + struct ifnet *rt_ifp; /* the answer: interface to use */ + struct ifaddr *rt_ifa; /* the answer: interface address to use */ + struct sctp_rt_metrics_lite rt_rmx; /* metrics used by rx'ing protocols */ + long rt_refcnt; /* # held references */ +#if 0 + struct sockaddr *rt_genmask; /* for generation of cloned routes */ + caddr_t rt_llinfo; /* pointer to link level info cache */ + struct rtentry *rt_gwroute; /* implied entry for gatewayed routes */ + struct rtentry *rt_parent; /* cloning parent of this route */ +#endif + struct mtx rt_mtx; /* mutex for routing entry */ +}; + +#define RT_LOCK_INIT(_rt) mtx_init(&(_rt)->rt_mtx, "rtentry", NULL, MTX_DEF | MTX_DUPOK) +#define RT_LOCK(_rt) mtx_lock(&(_rt)->rt_mtx) +#define RT_UNLOCK(_rt) mtx_unlock(&(_rt)->rt_mtx) +#define RT_LOCK_DESTROY(_rt) mtx_destroy(&(_rt)->rt_mtx) +#define RT_LOCK_ASSERT(_rt) mtx_assert(&(_rt)->rt_mtx, MA_OWNED) + +#define RT_ADDREF(_rt) do { \ + RT_LOCK_ASSERT(_rt); \ + KASSERT((_rt)->rt_refcnt >= 0, \ + ("negative refcnt %ld", (_rt)->rt_refcnt)); \ + (_rt)->rt_refcnt++; \ +} while (0) +#define RT_REMREF(_rt) do { \ + RT_LOCK_ASSERT(_rt); \ + KASSERT((_rt)->rt_refcnt > 0, \ + ("bogus refcnt %ld", (_rt)->rt_refcnt)); \ + (_rt)->rt_refcnt--; \ +} while (0) +#define RTFREE_LOCKED(_rt) do { \ + if ((_rt)->rt_refcnt <= 1) { \ + rtfree(_rt); \ + } else { \ + RT_REMREF(_rt); \ + RT_UNLOCK(_rt); \ + } \ + /* guard against invalid refs */ \ + _rt = NULL; \ + } while (0) +#define RTFREE(_rt) do { \ + RT_LOCK(_rt); \ + RTFREE_LOCKED(_rt); \ +} while (0) +#endif diff --git a/netwerk/sctp/src/user_socket.c b/netwerk/sctp/src/user_socket.c new file mode 100644 index 0000000000..ce9daed2aa --- /dev/null +++ b/netwerk/sctp/src/user_socket.c @@ -0,0 +1,3586 @@ +/*- + * Copyright (c) 1982, 1986, 1988, 1990, 1993 + * The Regents of the University of California. + * Copyright (c) 2004 The FreeBSD Foundation + * Copyright (c) 2004-2008 Robert N. M. Watson + * Copyright (c) 2009-2010 Brad Penoff + * Copyright (c) 2009-2010 Humaira Kamal + * Copyright (c) 2011-2012 Irene Ruengeler + * Copyright (c) 2011-2012 Michael Tuexen + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + */ + +#include <netinet/sctp_os.h> +#include <netinet/sctp_pcb.h> +#include <netinet/sctputil.h> +#include <netinet/sctp_var.h> +#include <netinet/sctp_sysctl.h> +#include <netinet/sctp_input.h> +#include <netinet/sctp_peeloff.h> +#include <netinet/sctp_callout.h> +#include <netinet/sctp_crc32.h> +#ifdef INET6 +#include <netinet6/sctp6_var.h> +#endif +#if defined(__FreeBSD__) +#include <sys/param.h> +#endif +#if defined(__linux__) +#define __FAVOR_BSD /* (on Ubuntu at least) enables UDP header field names like BSD in RFC 768 */ +#endif +#if !defined(_WIN32) +#if defined INET || defined INET6 +#include <netinet/udp.h> +#endif +#include <arpa/inet.h> +#else +#include <user_socketvar.h> +#endif +userland_mutex_t accept_mtx; +userland_cond_t accept_cond; +#ifdef _WIN32 +#include <time.h> +#include <sys/timeb.h> +#endif + +MALLOC_DEFINE(M_PCB, "sctp_pcb", "sctp pcb"); +MALLOC_DEFINE(M_SONAME, "sctp_soname", "sctp soname"); +#define MAXLEN_MBUF_CHAIN 32 + +/* Prototypes */ +extern int sctp_sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, + struct mbuf *top, struct mbuf *control, int flags, + /* proc is a dummy in __Userspace__ and will not be passed to sctp_lower_sosend */ + struct proc *p); + +extern int sctp_attach(struct socket *so, int proto, uint32_t vrf_id); +extern int sctpconn_attach(struct socket *so, int proto, uint32_t vrf_id); + +static void init_sync(void) { +#if defined(_WIN32) +#if defined(INET) || defined(INET6) + WSADATA wsaData; + + if (WSAStartup(MAKEWORD(2,2), &wsaData) != 0) { + SCTP_PRINTF("WSAStartup failed\n"); + exit (-1); + } +#endif + InitializeConditionVariable(&accept_cond); + InitializeCriticalSection(&accept_mtx); +#else + pthread_mutexattr_t mutex_attr; + + pthread_mutexattr_init(&mutex_attr); +#ifdef INVARIANTS + pthread_mutexattr_settype(&mutex_attr, PTHREAD_MUTEX_ERRORCHECK); +#endif + pthread_mutex_init(&accept_mtx, &mutex_attr); + pthread_mutexattr_destroy(&mutex_attr); + pthread_cond_init(&accept_cond, NULL); +#endif +} + +void +usrsctp_init(uint16_t port, + int (*conn_output)(void *addr, void *buffer, size_t length, uint8_t tos, uint8_t set_df), + void (*debug_printf)(const char *format, ...)) +{ + init_sync(); + sctp_init(port, conn_output, debug_printf, 1); +} + + +void +usrsctp_init_nothreads(uint16_t port, + int (*conn_output)(void *addr, void *buffer, size_t length, uint8_t tos, uint8_t set_df), + void (*debug_printf)(const char *format, ...)) +{ + init_sync(); + sctp_init(port, conn_output, debug_printf, 0); +} + + +/* Taken from usr/src/sys/kern/uipc_sockbuf.c and modified for __Userspace__*/ +/* + * Socantsendmore indicates that no more data will be sent on the socket; it + * would normally be applied to a socket when the user informs the system + * that no more data is to be sent, by the protocol code (in case + * PRU_SHUTDOWN). Socantrcvmore indicates that no more data will be + * received, and will normally be applied to the socket by a protocol when it + * detects that the peer will send no more data. Data queued for reading in + * the socket may yet be read. + */ + +void socantrcvmore_locked(struct socket *so) +{ + SOCKBUF_LOCK_ASSERT(&so->so_rcv); + so->so_rcv.sb_state |= SBS_CANTRCVMORE; + sorwakeup_locked(so); +} + +void socantrcvmore(struct socket *so) +{ + SOCKBUF_LOCK(&so->so_rcv); + socantrcvmore_locked(so); +} + +void +socantsendmore_locked(struct socket *so) +{ + SOCKBUF_LOCK_ASSERT(&so->so_snd); + so->so_snd.sb_state |= SBS_CANTSENDMORE; + sowwakeup_locked(so); +} + +void +socantsendmore(struct socket *so) +{ + SOCKBUF_LOCK(&so->so_snd); + socantsendmore_locked(so); +} + + + +/* Taken from usr/src/sys/kern/uipc_sockbuf.c and called within sctp_lower_sosend. + */ +int +sbwait(struct sockbuf *sb) +{ + SOCKBUF_LOCK_ASSERT(sb); + + sb->sb_flags |= SB_WAIT; +#if defined(_WIN32) + if (SleepConditionVariableCS(&(sb->sb_cond), &(sb->sb_mtx), INFINITE)) + return (0); + else + return (-1); +#else + return (pthread_cond_wait(&(sb->sb_cond), &(sb->sb_mtx))); +#endif +} + + + + +/* Taken from /src/sys/kern/uipc_socket.c + * and modified for __Userspace__ + */ +static struct socket * +soalloc(void) +{ + struct socket *so; + + /* + * soalloc() sets of socket layer state for a socket, + * called only by socreate() and sonewconn(). + * + * sodealloc() tears down socket layer state for a socket, + * called only by sofree() and sonewconn(). + * __Userspace__ TODO : Make sure so is properly deallocated + * when tearing down the connection. + */ + + so = (struct socket *)malloc(sizeof(struct socket)); + + if (so == NULL) { + return (NULL); + } + memset(so, 0, sizeof(struct socket)); + + /* __Userspace__ Initializing the socket locks here */ + SOCKBUF_LOCK_INIT(&so->so_snd, "so_snd"); + SOCKBUF_LOCK_INIT(&so->so_rcv, "so_rcv"); + SOCKBUF_COND_INIT(&so->so_snd); + SOCKBUF_COND_INIT(&so->so_rcv); + SOCK_COND_INIT(so); /* timeo_cond */ + + /* __Userspace__ Any ref counting required here? Will we have any use for aiojobq? + What about gencnt and numopensockets?*/ + TAILQ_INIT(&so->so_aiojobq); + return (so); +} + +static void +sodealloc(struct socket *so) +{ + + KASSERT(so->so_count == 0, ("sodealloc(): so_count %d", so->so_count)); + KASSERT(so->so_pcb == NULL, ("sodealloc(): so_pcb != NULL")); + + SOCKBUF_COND_DESTROY(&so->so_snd); + SOCKBUF_COND_DESTROY(&so->so_rcv); + + SOCK_COND_DESTROY(so); + + SOCKBUF_LOCK_DESTROY(&so->so_snd); + SOCKBUF_LOCK_DESTROY(&so->so_rcv); + + free(so); +} + +/* Taken from /src/sys/kern/uipc_socket.c + * and modified for __Userspace__ + */ +void +sofree(struct socket *so) +{ + struct socket *head; + + ACCEPT_LOCK_ASSERT(); + SOCK_LOCK_ASSERT(so); + /* SS_NOFDREF unset in accept call. this condition seems irrelevant + * for __Userspace__... + */ + if (so->so_count != 0 || + (so->so_state & SS_PROTOREF) || (so->so_qstate & SQ_COMP)) { + SOCK_UNLOCK(so); + ACCEPT_UNLOCK(); + return; + } + head = so->so_head; + if (head != NULL) { + KASSERT((so->so_qstate & SQ_COMP) != 0 || + (so->so_qstate & SQ_INCOMP) != 0, + ("sofree: so_head != NULL, but neither SQ_COMP nor " + "SQ_INCOMP")); + KASSERT((so->so_qstate & SQ_COMP) == 0 || + (so->so_qstate & SQ_INCOMP) == 0, + ("sofree: so->so_qstate is SQ_COMP and also SQ_INCOMP")); + TAILQ_REMOVE(&head->so_incomp, so, so_list); + head->so_incqlen--; + so->so_qstate &= ~SQ_INCOMP; + so->so_head = NULL; + } + KASSERT((so->so_qstate & SQ_COMP) == 0 && + (so->so_qstate & SQ_INCOMP) == 0, + ("sofree: so_head == NULL, but still SQ_COMP(%d) or SQ_INCOMP(%d)", + so->so_qstate & SQ_COMP, so->so_qstate & SQ_INCOMP)); + if (so->so_options & SCTP_SO_ACCEPTCONN) { + KASSERT((TAILQ_EMPTY(&so->so_comp)), ("sofree: so_comp populated")); + KASSERT((TAILQ_EMPTY(&so->so_incomp)), ("sofree: so_comp populated")); + } + SOCK_UNLOCK(so); + ACCEPT_UNLOCK(); + sctp_close(so); /* was... sctp_detach(so); */ + /* + * From this point on, we assume that no other references to this + * socket exist anywhere else in the stack. Therefore, no locks need + * to be acquired or held. + * + * We used to do a lot of socket buffer and socket locking here, as + * well as invoke sorflush() and perform wakeups. The direct call to + * dom_dispose() and sbrelease_internal() are an inlining of what was + * necessary from sorflush(). + * + * Notice that the socket buffer and kqueue state are torn down + * before calling pru_detach. This means that protocols should not + * assume they can perform socket wakeups, etc, in their detach code. + */ + sodealloc(so); +} + + + +/* Taken from /src/sys/kern/uipc_socket.c */ +void +soabort(struct socket *so) +{ +#if defined(INET6) + struct sctp_inpcb *inp; +#endif + +#if defined(INET6) + inp = (struct sctp_inpcb *)so->so_pcb; + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + sctp6_abort(so); + } else { +#if defined(INET) + sctp_abort(so); +#endif + } +#elif defined(INET) + sctp_abort(so); +#endif + ACCEPT_LOCK(); + SOCK_LOCK(so); + sofree(so); +} + + +/* Taken from usr/src/sys/kern/uipc_socket.c and called within sctp_connect (sctp_usrreq.c). + * We use sctp_connect for send_one_init_real in ms1. + */ +void +soisconnecting(struct socket *so) +{ + + SOCK_LOCK(so); + so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING); + so->so_state |= SS_ISCONNECTING; + SOCK_UNLOCK(so); +} + +/* Taken from usr/src/sys/kern/uipc_socket.c and called within sctp_disconnect (sctp_usrreq.c). + * TODO Do we use sctp_disconnect? + */ +void +soisdisconnecting(struct socket *so) +{ + + /* + * Note: This code assumes that SOCK_LOCK(so) and + * SOCKBUF_LOCK(&so->so_rcv) are the same. + */ + SOCKBUF_LOCK(&so->so_rcv); + so->so_state &= ~SS_ISCONNECTING; + so->so_state |= SS_ISDISCONNECTING; + so->so_rcv.sb_state |= SBS_CANTRCVMORE; + sorwakeup_locked(so); + SOCKBUF_LOCK(&so->so_snd); + so->so_snd.sb_state |= SBS_CANTSENDMORE; + sowwakeup_locked(so); + wakeup("dummy",so); + /* requires 2 args but this was in orig */ + /* wakeup(&so->so_timeo); */ +} + + +/* Taken from sys/kern/kern_synch.c and + modified for __Userspace__ +*/ + +/* + * Make all threads sleeping on the specified identifier runnable. + * Associating wakeup with so_timeo identifier and timeo_cond + * condition variable. TODO. If we use iterator thread then we need to + * modify wakeup so it can distinguish between iterator identifier and + * timeo identifier. + */ +void +wakeup(void *ident, struct socket *so) +{ + SOCK_LOCK(so); +#if defined(_WIN32) + WakeAllConditionVariable(&(so)->timeo_cond); +#else + pthread_cond_broadcast(&(so)->timeo_cond); +#endif + SOCK_UNLOCK(so); +} + + +/* + * Make a thread sleeping on the specified identifier runnable. + * May wake more than one thread if a target thread is currently + * swapped out. + */ +void +wakeup_one(void *ident) +{ + /* __Userspace__ Check: We are using accept_cond for wakeup_one. + It seems that wakeup_one is only called within + soisconnected() and sonewconn() with ident &head->so_timeo + head is so->so_head, which is back pointer to listen socket + This seems to indicate that the use of accept_cond is correct + since socket where accepts occur is so_head in all + subsidiary sockets. + */ + ACCEPT_LOCK(); +#if defined(_WIN32) + WakeAllConditionVariable(&accept_cond); +#else + pthread_cond_broadcast(&accept_cond); +#endif + ACCEPT_UNLOCK(); +} + + +/* Called within sctp_process_cookie_[existing/new] */ +void +soisconnected(struct socket *so) +{ + struct socket *head; + + ACCEPT_LOCK(); + SOCK_LOCK(so); + so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING); + so->so_state |= SS_ISCONNECTED; + head = so->so_head; + if (head != NULL && (so->so_qstate & SQ_INCOMP)) { + SOCK_UNLOCK(so); + TAILQ_REMOVE(&head->so_incomp, so, so_list); + head->so_incqlen--; + so->so_qstate &= ~SQ_INCOMP; + TAILQ_INSERT_TAIL(&head->so_comp, so, so_list); + head->so_qlen++; + so->so_qstate |= SQ_COMP; + ACCEPT_UNLOCK(); + sorwakeup(head); + wakeup_one(&head->so_timeo); + return; + } + SOCK_UNLOCK(so); + ACCEPT_UNLOCK(); + wakeup(&so->so_timeo, so); + sorwakeup(so); + sowwakeup(so); + +} + +/* called within sctp_handle_cookie_echo */ + +struct socket * +sonewconn(struct socket *head, int connstatus) +{ + struct socket *so; + int over; + + ACCEPT_LOCK(); + over = (head->so_qlen > 3 * head->so_qlimit / 2); + ACCEPT_UNLOCK(); +#ifdef REGRESSION + if (regression_sonewconn_earlytest && over) +#else + if (over) +#endif + return (NULL); + so = soalloc(); + if (so == NULL) + return (NULL); + so->so_head = head; + so->so_type = head->so_type; + so->so_options = head->so_options &~ SCTP_SO_ACCEPTCONN; + so->so_linger = head->so_linger; + so->so_state = head->so_state | SS_NOFDREF; + so->so_dom = head->so_dom; +#ifdef MAC + SOCK_LOCK(head); + mac_create_socket_from_socket(head, so); + SOCK_UNLOCK(head); +#endif + if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat)) { + sodealloc(so); + return (NULL); + } + switch (head->so_dom) { +#ifdef INET + case AF_INET: + if (sctp_attach(so, IPPROTO_SCTP, SCTP_DEFAULT_VRFID)) { + sodealloc(so); + return (NULL); + } + break; +#endif +#ifdef INET6 + case AF_INET6: + if (sctp6_attach(so, IPPROTO_SCTP, SCTP_DEFAULT_VRFID)) { + sodealloc(so); + return (NULL); + } + break; +#endif + case AF_CONN: + if (sctpconn_attach(so, IPPROTO_SCTP, SCTP_DEFAULT_VRFID)) { + sodealloc(so); + return (NULL); + } + break; + default: + sodealloc(so); + return (NULL); + break; + } + so->so_rcv.sb_lowat = head->so_rcv.sb_lowat; + so->so_snd.sb_lowat = head->so_snd.sb_lowat; + so->so_rcv.sb_timeo = head->so_rcv.sb_timeo; + so->so_snd.sb_timeo = head->so_snd.sb_timeo; + so->so_rcv.sb_flags |= head->so_rcv.sb_flags & SB_AUTOSIZE; + so->so_snd.sb_flags |= head->so_snd.sb_flags & SB_AUTOSIZE; + so->so_state |= connstatus; + ACCEPT_LOCK(); + if (connstatus) { + TAILQ_INSERT_TAIL(&head->so_comp, so, so_list); + so->so_qstate |= SQ_COMP; + head->so_qlen++; + } else { + /* + * Keep removing sockets from the head until there's room for + * us to insert on the tail. In pre-locking revisions, this + * was a simple if (), but as we could be racing with other + * threads and soabort() requires dropping locks, we must + * loop waiting for the condition to be true. + */ + while (head->so_incqlen > head->so_qlimit) { + struct socket *sp; + sp = TAILQ_FIRST(&head->so_incomp); + TAILQ_REMOVE(&head->so_incomp, sp, so_list); + head->so_incqlen--; + sp->so_qstate &= ~SQ_INCOMP; + sp->so_head = NULL; + ACCEPT_UNLOCK(); + soabort(sp); + ACCEPT_LOCK(); + } + TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list); + so->so_qstate |= SQ_INCOMP; + head->so_incqlen++; + } + ACCEPT_UNLOCK(); + if (connstatus) { + sorwakeup(head); + wakeup_one(&head->so_timeo); + } + return (so); + +} + + /* + Source: /src/sys/gnu/fs/xfs/FreeBSD/xfs_ioctl.c + */ +static __inline__ int +copy_to_user(void *dst, void *src, size_t len) { + memcpy(dst, src, len); + return 0; +} + +static __inline__ int +copy_from_user(void *dst, void *src, size_t len) { + memcpy(dst, src, len); + return 0; +} + +/* + References: + src/sys/dev/lmc/if_lmc.h: + src/sys/powerpc/powerpc/copyinout.c + src/sys/sys/systm.h +*/ +# define copyin(u, k, len) copy_from_user(k, u, len) + +/* References: + src/sys/powerpc/powerpc/copyinout.c + src/sys/sys/systm.h +*/ +# define copyout(k, u, len) copy_to_user(u, k, len) + + +/* copyiniov definition copied/modified from src/sys/kern/kern_subr.c */ +int +copyiniov(struct iovec *iovp, u_int iovcnt, struct iovec **iov, int error) +{ + u_int iovlen; + + *iov = NULL; + if (iovcnt > UIO_MAXIOV) + return (error); + iovlen = iovcnt * sizeof (struct iovec); + *iov = malloc(iovlen); /*, M_IOV, M_WAITOK); */ + error = copyin(iovp, *iov, iovlen); + if (error) { + free(*iov); /*, M_IOV); */ + *iov = NULL; + } + return (error); +} + +/* (__Userspace__) version of uiomove */ +int +uiomove(void *cp, int n, struct uio *uio) +{ + struct iovec *iov; + size_t cnt; + int error = 0; + + if ((uio->uio_rw != UIO_READ) && + (uio->uio_rw != UIO_WRITE)) { + return (EINVAL); + } + + while (n > 0 && uio->uio_resid) { + iov = uio->uio_iov; + cnt = iov->iov_len; + if (cnt == 0) { + uio->uio_iov++; + uio->uio_iovcnt--; + continue; + } + if (cnt > (size_t)n) + cnt = n; + + switch (uio->uio_segflg) { + + case UIO_USERSPACE: + if (uio->uio_rw == UIO_READ) + error = copyout(cp, iov->iov_base, cnt); + else + error = copyin(iov->iov_base, cp, cnt); + if (error) + goto out; + break; + + case UIO_SYSSPACE: + if (uio->uio_rw == UIO_READ) + memcpy(iov->iov_base, cp, cnt); + else + memcpy(cp, iov->iov_base, cnt); + break; + } + iov->iov_base = (char *)iov->iov_base + cnt; + iov->iov_len -= cnt; + uio->uio_resid -= cnt; + uio->uio_offset += (off_t)cnt; + cp = (char *)cp + cnt; + n -= (int)cnt; + } +out: + return (error); +} + + +/* Source: src/sys/kern/uipc_syscalls.c */ +int +getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len) +{ + struct sockaddr *sa; + int error; + + if (len > SOCK_MAXADDRLEN) + return (ENAMETOOLONG); + if (len < offsetof(struct sockaddr, sa_data)) + return (EINVAL); + MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK); + error = copyin(uaddr, sa, len); + if (error) { + FREE(sa, M_SONAME); + } else { +#ifdef HAVE_SA_LEN + sa->sa_len = len; +#endif + *namp = sa; + } + return (error); +} + +int +usrsctp_getsockopt(struct socket *so, int level, int option_name, + void *option_value, socklen_t *option_len); + +sctp_assoc_t +usrsctp_getassocid(struct socket *sock, struct sockaddr *sa) +{ + struct sctp_paddrinfo sp; + socklen_t siz; +#ifndef HAVE_SA_LEN + size_t sa_len; +#endif + + /* First get the assoc id */ + siz = sizeof(sp); + memset(&sp, 0, sizeof(sp)); +#ifdef HAVE_SA_LEN + memcpy((caddr_t)&sp.spinfo_address, sa, sa->sa_len); +#else + switch (sa->sa_family) { +#ifdef INET + case AF_INET: + sa_len = sizeof(struct sockaddr_in); + break; +#endif +#ifdef INET6 + case AF_INET6: + sa_len = sizeof(struct sockaddr_in6); + break; +#endif + case AF_CONN: + sa_len = sizeof(struct sockaddr_conn); + break; + default: + sa_len = 0; + break; + } + memcpy((caddr_t)&sp.spinfo_address, sa, sa_len); +#endif + if (usrsctp_getsockopt(sock, IPPROTO_SCTP, SCTP_GET_PEER_ADDR_INFO, &sp, &siz) != 0) { + /* We depend on the fact that 0 can never be returned */ + return ((sctp_assoc_t) 0); + } + return (sp.spinfo_assoc_id); +} + + +/* Taken from /src/lib/libc/net/sctp_sys_calls.c + * and modified for __Userspace__ + * calling sctp_generic_sendmsg from this function + */ +ssize_t +userspace_sctp_sendmsg(struct socket *so, + const void *data, + size_t len, + struct sockaddr *to, + socklen_t tolen, + uint32_t ppid, + uint32_t flags, + uint16_t stream_no, + uint32_t timetolive, + uint32_t context) +{ + struct sctp_sndrcvinfo sndrcvinfo, *sinfo = &sndrcvinfo; + struct uio auio; + struct iovec iov[1]; + + memset(sinfo, 0, sizeof(struct sctp_sndrcvinfo)); + sinfo->sinfo_ppid = ppid; + sinfo->sinfo_flags = flags; + sinfo->sinfo_stream = stream_no; + sinfo->sinfo_timetolive = timetolive; + sinfo->sinfo_context = context; + sinfo->sinfo_assoc_id = 0; + + + /* Perform error checks on destination (to) */ + if (tolen > SOCK_MAXADDRLEN) { + errno = ENAMETOOLONG; + return (-1); + } + if ((tolen > 0) && + ((to == NULL) || (tolen < (socklen_t)sizeof(struct sockaddr)))) { + errno = EINVAL; + return (-1); + } + if (data == NULL) { + errno = EFAULT; + return (-1); + } + /* Adding the following as part of defensive programming, in case the application + does not do it when preparing the destination address.*/ +#ifdef HAVE_SA_LEN + if (to != NULL) { + to->sa_len = tolen; + } +#endif + + iov[0].iov_base = (caddr_t)data; + iov[0].iov_len = len; + + auio.uio_iov = iov; + auio.uio_iovcnt = 1; + auio.uio_segflg = UIO_USERSPACE; + auio.uio_rw = UIO_WRITE; + auio.uio_offset = 0; /* XXX */ + auio.uio_resid = len; + errno = sctp_lower_sosend(so, to, &auio, NULL, NULL, 0, sinfo); + if (errno == 0) { + return (len - auio.uio_resid); + } else { + return (-1); + } +} + + +ssize_t +usrsctp_sendv(struct socket *so, + const void *data, + size_t len, + struct sockaddr *to, + int addrcnt, + void *info, + socklen_t infolen, + unsigned int infotype, + int flags) +{ + struct sctp_sndrcvinfo sinfo; + struct uio auio; + struct iovec iov[1]; + int use_sinfo; + sctp_assoc_t *assoc_id; + + if (so == NULL) { + errno = EBADF; + return (-1); + } + if (data == NULL) { + errno = EFAULT; + return (-1); + } + memset(&sinfo, 0, sizeof(struct sctp_sndrcvinfo)); + assoc_id = NULL; + use_sinfo = 0; + switch (infotype) { + case SCTP_SENDV_NOINFO: + if ((infolen != 0) || (info != NULL)) { + errno = EINVAL; + return (-1); + } + break; + case SCTP_SENDV_SNDINFO: + if ((info == NULL) || (infolen != sizeof(struct sctp_sndinfo))) { + errno = EINVAL; + return (-1); + } + sinfo.sinfo_stream = ((struct sctp_sndinfo *)info)->snd_sid; + sinfo.sinfo_flags = ((struct sctp_sndinfo *)info)->snd_flags; + sinfo.sinfo_ppid = ((struct sctp_sndinfo *)info)->snd_ppid; + sinfo.sinfo_context = ((struct sctp_sndinfo *)info)->snd_context; + sinfo.sinfo_assoc_id = ((struct sctp_sndinfo *)info)->snd_assoc_id; + assoc_id = &(((struct sctp_sndinfo *)info)->snd_assoc_id); + use_sinfo = 1; + break; + case SCTP_SENDV_PRINFO: + if ((info == NULL) || (infolen != sizeof(struct sctp_prinfo))) { + errno = EINVAL; + return (-1); + } + sinfo.sinfo_stream = 0; + sinfo.sinfo_flags = PR_SCTP_POLICY(((struct sctp_prinfo *)info)->pr_policy); + sinfo.sinfo_timetolive = ((struct sctp_prinfo *)info)->pr_value; + use_sinfo = 1; + break; + case SCTP_SENDV_AUTHINFO: + errno = EINVAL; + return (-1); + case SCTP_SENDV_SPA: + if ((info == NULL) || (infolen != sizeof(struct sctp_sendv_spa))) { + errno = EINVAL; + return (-1); + } + if (((struct sctp_sendv_spa *)info)->sendv_flags & SCTP_SEND_SNDINFO_VALID) { + sinfo.sinfo_stream = ((struct sctp_sendv_spa *)info)->sendv_sndinfo.snd_sid; + sinfo.sinfo_flags = ((struct sctp_sendv_spa *)info)->sendv_sndinfo.snd_flags; + sinfo.sinfo_ppid = ((struct sctp_sendv_spa *)info)->sendv_sndinfo.snd_ppid; + sinfo.sinfo_context = ((struct sctp_sendv_spa *)info)->sendv_sndinfo.snd_context; + sinfo.sinfo_assoc_id = ((struct sctp_sendv_spa *)info)->sendv_sndinfo.snd_assoc_id; + assoc_id = &(((struct sctp_sendv_spa *)info)->sendv_sndinfo.snd_assoc_id); + } else { + sinfo.sinfo_flags = 0; + sinfo.sinfo_stream = 0; + } + if (((struct sctp_sendv_spa *)info)->sendv_flags & SCTP_SEND_PRINFO_VALID) { + sinfo.sinfo_flags |= PR_SCTP_POLICY(((struct sctp_sendv_spa *)info)->sendv_prinfo.pr_policy); + sinfo.sinfo_timetolive = ((struct sctp_sendv_spa *)info)->sendv_prinfo.pr_value; + } + if (((struct sctp_sendv_spa *)info)->sendv_flags & SCTP_SEND_AUTHINFO_VALID) { + errno = EINVAL; + return (-1); + } + use_sinfo = 1; + break; + default: + errno = EINVAL; + return (-1); + } + + /* Perform error checks on destination (to) */ + if (addrcnt > 1) { + errno = EINVAL; + return (-1); + } + + iov[0].iov_base = (caddr_t)data; + iov[0].iov_len = len; + + auio.uio_iov = iov; + auio.uio_iovcnt = 1; + auio.uio_segflg = UIO_USERSPACE; + auio.uio_rw = UIO_WRITE; + auio.uio_offset = 0; /* XXX */ + auio.uio_resid = len; + errno = sctp_lower_sosend(so, to, &auio, NULL, NULL, flags, use_sinfo ? &sinfo : NULL); + if (errno == 0) { + if ((to != NULL) && (assoc_id != NULL)) { + *assoc_id = usrsctp_getassocid(so, to); + } + return (len - auio.uio_resid); + } else { + return (-1); + } +} + + +ssize_t +userspace_sctp_sendmbuf(struct socket *so, + struct mbuf* mbufdata, + size_t len, + struct sockaddr *to, + socklen_t tolen, + uint32_t ppid, + uint32_t flags, + uint16_t stream_no, + uint32_t timetolive, + uint32_t context) +{ + + struct sctp_sndrcvinfo sndrcvinfo, *sinfo = &sndrcvinfo; + /* struct uio auio; + struct iovec iov[1]; */ + int error = 0; + int uflags = 0; + ssize_t retval; + + sinfo->sinfo_ppid = ppid; + sinfo->sinfo_flags = flags; + sinfo->sinfo_stream = stream_no; + sinfo->sinfo_timetolive = timetolive; + sinfo->sinfo_context = context; + sinfo->sinfo_assoc_id = 0; + + /* Perform error checks on destination (to) */ + if (tolen > SOCK_MAXADDRLEN){ + error = (ENAMETOOLONG); + goto sendmsg_return; + } + if (tolen < (socklen_t)offsetof(struct sockaddr, sa_data)){ + error = (EINVAL); + goto sendmsg_return; + } + /* Adding the following as part of defensive programming, in case the application + does not do it when preparing the destination address.*/ +#ifdef HAVE_SA_LEN + to->sa_len = tolen; +#endif + + error = sctp_lower_sosend(so, to, NULL/*uio*/, + (struct mbuf *)mbufdata, (struct mbuf *)NULL, + uflags, sinfo); +sendmsg_return: + /* TODO: Needs a condition for non-blocking when error is EWOULDBLOCK */ + if (0 == error) + retval = len; + else if (error == EWOULDBLOCK) { + errno = EWOULDBLOCK; + retval = -1; + } else { + SCTP_PRINTF("%s: error = %d\n", __func__, error); + errno = error; + retval = -1; + } + return (retval); +} + + +/* taken from usr.lib/sctp_sys_calls.c and needed here */ +#define SCTP_SMALL_IOVEC_SIZE 2 + +/* Taken from /src/lib/libc/net/sctp_sys_calls.c + * and modified for __Userspace__ + * calling sctp_generic_recvmsg from this function + */ +ssize_t +userspace_sctp_recvmsg(struct socket *so, + void *dbuf, + size_t len, + struct sockaddr *from, + socklen_t *fromlenp, + struct sctp_sndrcvinfo *sinfo, + int *msg_flags) +{ + struct uio auio; + struct iovec iov[SCTP_SMALL_IOVEC_SIZE]; + struct iovec *tiov; + int iovlen = 1; + int error = 0; + ssize_t ulen; + int i; + socklen_t fromlen; + + iov[0].iov_base = dbuf; + iov[0].iov_len = len; + + auio.uio_iov = iov; + auio.uio_iovcnt = iovlen; + auio.uio_segflg = UIO_USERSPACE; + auio.uio_rw = UIO_READ; + auio.uio_offset = 0; /* XXX */ + auio.uio_resid = 0; + tiov = iov; + for (i = 0; i <iovlen; i++, tiov++) { + if ((auio.uio_resid += tiov->iov_len) < 0) { + error = EINVAL; + SCTP_PRINTF("%s: error = %d\n", __func__, error); + return (-1); + } + } + ulen = auio.uio_resid; + if (fromlenp != NULL) { + fromlen = *fromlenp; + } else { + fromlen = 0; + } + error = sctp_sorecvmsg(so, &auio, (struct mbuf **)NULL, + from, fromlen, msg_flags, + (struct sctp_sndrcvinfo *)sinfo, 1); + + if (error) { + if ((auio.uio_resid != ulen) && + (error == EINTR || +#if !defined(__NetBSD__) + error == ERESTART || +#endif + error == EWOULDBLOCK)) { + error = 0; + } + } + if ((fromlenp != NULL) && (fromlen > 0) && (from != NULL)) { + switch (from->sa_family) { +#if defined(INET) + case AF_INET: + *fromlenp = sizeof(struct sockaddr_in); + break; +#endif +#if defined(INET6) + case AF_INET6: + *fromlenp = sizeof(struct sockaddr_in6); + break; +#endif + case AF_CONN: + *fromlenp = sizeof(struct sockaddr_conn); + break; + default: + *fromlenp = 0; + break; + } + if (*fromlenp > fromlen) { + *fromlenp = fromlen; + } + } + if (error == 0) { + /* ready return value */ + return (ulen - auio.uio_resid); + } else { + SCTP_PRINTF("%s: error = %d\n", __func__, error); + return (-1); + } +} + +ssize_t +usrsctp_recvv(struct socket *so, + void *dbuf, + size_t len, + struct sockaddr *from, + socklen_t *fromlenp, + void *info, + socklen_t *infolen, + unsigned int *infotype, + int *msg_flags) +{ + struct uio auio; + struct iovec iov[SCTP_SMALL_IOVEC_SIZE]; + struct iovec *tiov; + int iovlen = 1; + ssize_t ulen; + int i; + socklen_t fromlen; + struct sctp_rcvinfo *rcv; + struct sctp_recvv_rn *rn; + struct sctp_extrcvinfo seinfo; + + if (so == NULL) { + errno = EBADF; + return (-1); + } + iov[0].iov_base = dbuf; + iov[0].iov_len = len; + + auio.uio_iov = iov; + auio.uio_iovcnt = iovlen; + auio.uio_segflg = UIO_USERSPACE; + auio.uio_rw = UIO_READ; + auio.uio_offset = 0; /* XXX */ + auio.uio_resid = 0; + tiov = iov; + for (i = 0; i <iovlen; i++, tiov++) { + if ((auio.uio_resid += tiov->iov_len) < 0) { + errno = EINVAL; + return (-1); + } + } + ulen = auio.uio_resid; + if (fromlenp != NULL) { + fromlen = *fromlenp; + } else { + fromlen = 0; + } + errno = sctp_sorecvmsg(so, &auio, (struct mbuf **)NULL, + from, fromlen, msg_flags, + (struct sctp_sndrcvinfo *)&seinfo, 1); + if (errno) { + if ((auio.uio_resid != ulen) && + (errno == EINTR || +#if !defined(__NetBSD__) + errno == ERESTART || +#endif + errno == EWOULDBLOCK)) { + errno = 0; + } + } + if (errno != 0) { + goto out; + } + if ((*msg_flags & MSG_NOTIFICATION) == 0) { + struct sctp_inpcb *inp; + + inp = (struct sctp_inpcb *)so->so_pcb; + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVNXTINFO) && + sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVRCVINFO) && + *infolen >= (socklen_t)sizeof(struct sctp_recvv_rn) && + seinfo.sreinfo_next_flags & SCTP_NEXT_MSG_AVAIL) { + rn = (struct sctp_recvv_rn *)info; + rn->recvv_rcvinfo.rcv_sid = seinfo.sinfo_stream; + rn->recvv_rcvinfo.rcv_ssn = seinfo.sinfo_ssn; + rn->recvv_rcvinfo.rcv_flags = seinfo.sinfo_flags; + rn->recvv_rcvinfo.rcv_ppid = seinfo.sinfo_ppid; + rn->recvv_rcvinfo.rcv_context = seinfo.sinfo_context; + rn->recvv_rcvinfo.rcv_tsn = seinfo.sinfo_tsn; + rn->recvv_rcvinfo.rcv_cumtsn = seinfo.sinfo_cumtsn; + rn->recvv_rcvinfo.rcv_assoc_id = seinfo.sinfo_assoc_id; + rn->recvv_nxtinfo.nxt_sid = seinfo.sreinfo_next_stream; + rn->recvv_nxtinfo.nxt_flags = 0; + if (seinfo.sreinfo_next_flags & SCTP_NEXT_MSG_IS_UNORDERED) { + rn->recvv_nxtinfo.nxt_flags |= SCTP_UNORDERED; + } + if (seinfo.sreinfo_next_flags & SCTP_NEXT_MSG_IS_NOTIFICATION) { + rn->recvv_nxtinfo.nxt_flags |= SCTP_NOTIFICATION; + } + if (seinfo.sreinfo_next_flags & SCTP_NEXT_MSG_ISCOMPLETE) { + rn->recvv_nxtinfo.nxt_flags |= SCTP_COMPLETE; + } + rn->recvv_nxtinfo.nxt_ppid = seinfo.sreinfo_next_ppid; + rn->recvv_nxtinfo.nxt_length = seinfo.sreinfo_next_length; + rn->recvv_nxtinfo.nxt_assoc_id = seinfo.sreinfo_next_aid; + *infolen = (socklen_t)sizeof(struct sctp_recvv_rn); + *infotype = SCTP_RECVV_RN; + } else if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVRCVINFO) && + *infolen >= (socklen_t)sizeof(struct sctp_rcvinfo)) { + rcv = (struct sctp_rcvinfo *)info; + rcv->rcv_sid = seinfo.sinfo_stream; + rcv->rcv_ssn = seinfo.sinfo_ssn; + rcv->rcv_flags = seinfo.sinfo_flags; + rcv->rcv_ppid = seinfo.sinfo_ppid; + rcv->rcv_context = seinfo.sinfo_context; + rcv->rcv_tsn = seinfo.sinfo_tsn; + rcv->rcv_cumtsn = seinfo.sinfo_cumtsn; + rcv->rcv_assoc_id = seinfo.sinfo_assoc_id; + *infolen = (socklen_t)sizeof(struct sctp_rcvinfo); + *infotype = SCTP_RECVV_RCVINFO; + } else { + *infotype = SCTP_RECVV_NOINFO; + *infolen = 0; + } + } + if ((fromlenp != NULL) && + (fromlen > 0) && + (from != NULL) && + (ulen > auio.uio_resid)) { + switch (from->sa_family) { +#if defined(INET) + case AF_INET: + *fromlenp = sizeof(struct sockaddr_in); + break; +#endif +#if defined(INET6) + case AF_INET6: + *fromlenp = sizeof(struct sockaddr_in6); + break; +#endif + case AF_CONN: + *fromlenp = sizeof(struct sockaddr_conn); + break; + default: + *fromlenp = 0; + break; + } + if (*fromlenp > fromlen) { + *fromlenp = fromlen; + } + } +out: + if (errno == 0) { + /* ready return value */ + return (ulen - auio.uio_resid); + } else { + return (-1); + } +} + + + + +/* Taken from /src/sys/kern/uipc_socket.c + * and modified for __Userspace__ + * socreate returns a socket. The socket should be + * closed with soclose(). + */ +int +socreate(int dom, struct socket **aso, int type, int proto) +{ + struct socket *so; + int error; + + if ((dom != AF_CONN) && (dom != AF_INET) && (dom != AF_INET6)) { + return (EINVAL); + } + if ((type != SOCK_STREAM) && (type != SOCK_SEQPACKET)) { + return (EINVAL); + } + if (proto != IPPROTO_SCTP) { + return (EINVAL); + } + + so = soalloc(); + if (so == NULL) { + return (ENOBUFS); + } + + /* + * so_incomp represents a queue of connections that + * must be completed at protocol level before being + * returned. so_comp field heads a list of sockets + * that are ready to be returned to the listening process + *__Userspace__ These queues are being used at a number of places like accept etc. + */ + TAILQ_INIT(&so->so_incomp); + TAILQ_INIT(&so->so_comp); + so->so_type = type; + so->so_count = 1; + so->so_dom = dom; + /* + * Auto-sizing of socket buffers is managed by the protocols and + * the appropriate flags must be set in the pru_attach function. + * For __Userspace__ The pru_attach function in this case is sctp_attach. + */ + switch (dom) { +#if defined(INET) + case AF_INET: + error = sctp_attach(so, proto, SCTP_DEFAULT_VRFID); + break; +#endif +#if defined(INET6) + case AF_INET6: + error = sctp6_attach(so, proto, SCTP_DEFAULT_VRFID); + break; +#endif + case AF_CONN: + error = sctpconn_attach(so, proto, SCTP_DEFAULT_VRFID); + break; + default: + error = EAFNOSUPPORT; + break; + } + if (error) { + KASSERT(so->so_count == 1, ("socreate: so_count %d", so->so_count)); + so->so_count = 0; + sodealloc(so); + return (error); + } + *aso = so; + return (0); +} + + +/* Taken from /src/sys/kern/uipc_syscalls.c + * and modified for __Userspace__ + * Removing struct thread td. + */ +struct socket * +userspace_socket(int domain, int type, int protocol) +{ + struct socket *so = NULL; + + errno = socreate(domain, &so, type, protocol); + if (errno) { + return (NULL); + } + /* + * The original socket call returns the file descriptor fd. + * td->td_retval[0] = fd. + * We are returning struct socket *so. + */ + return (so); +} + +struct socket * +usrsctp_socket(int domain, int type, int protocol, + int (*receive_cb)(struct socket *sock, union sctp_sockstore addr, void *data, + size_t datalen, struct sctp_rcvinfo, int flags, void *ulp_info), + int (*send_cb)(struct socket *sock, uint32_t sb_free, void *ulp_info), + uint32_t sb_threshold, + void *ulp_info) +{ + struct socket *so = NULL; + + if ((protocol == IPPROTO_SCTP) && (SCTP_BASE_VAR(sctp_pcb_initialized) == 0)) { + errno = EPROTONOSUPPORT; + return (NULL); + } + if ((receive_cb == NULL) && + ((send_cb != NULL) || (sb_threshold != 0) || (ulp_info != NULL))) { + errno = EINVAL; + return (NULL); + } + if ((domain == AF_CONN) && (SCTP_BASE_VAR(conn_output) == NULL)) { + errno = EAFNOSUPPORT; + return (NULL); + } + errno = socreate(domain, &so, type, protocol); + if (errno) { + return (NULL); + } + /* + * The original socket call returns the file descriptor fd. + * td->td_retval[0] = fd. + * We are returning struct socket *so. + */ + register_recv_cb(so, receive_cb); + register_send_cb(so, sb_threshold, send_cb); + register_ulp_info(so, ulp_info); + return (so); +} + + +u_long sb_max = SB_MAX; +u_long sb_max_adj = + SB_MAX * MCLBYTES / (MSIZE + MCLBYTES); /* adjusted sb_max */ + +static u_long sb_efficiency = 8; /* parameter for sbreserve() */ + +/* + * Allot mbufs to a sockbuf. Attempt to scale mbmax so that mbcnt doesn't + * become limiting if buffering efficiency is near the normal case. + */ +int +sbreserve_locked(struct sockbuf *sb, u_long cc, struct socket *so) +{ + SOCKBUF_LOCK_ASSERT(sb); + sb->sb_mbmax = (u_int)min(cc * sb_efficiency, sb_max); + sb->sb_hiwat = (u_int)cc; + if (sb->sb_lowat > (int)sb->sb_hiwat) + sb->sb_lowat = (int)sb->sb_hiwat; + return (1); +} + +static int +sbreserve(struct sockbuf *sb, u_long cc, struct socket *so) +{ + int error; + + SOCKBUF_LOCK(sb); + error = sbreserve_locked(sb, cc, so); + SOCKBUF_UNLOCK(sb); + return (error); +} + +int +soreserve(struct socket *so, u_long sndcc, u_long rcvcc) +{ + SOCKBUF_LOCK(&so->so_snd); + SOCKBUF_LOCK(&so->so_rcv); + so->so_snd.sb_hiwat = (uint32_t)sndcc; + so->so_rcv.sb_hiwat = (uint32_t)rcvcc; + + if (sbreserve_locked(&so->so_snd, sndcc, so) == 0) { + goto bad; + } + if (sbreserve_locked(&so->so_rcv, rcvcc, so) == 0) { + goto bad; + } + if (so->so_rcv.sb_lowat == 0) + so->so_rcv.sb_lowat = 1; + if (so->so_snd.sb_lowat == 0) + so->so_snd.sb_lowat = MCLBYTES; + if (so->so_snd.sb_lowat > (int)so->so_snd.sb_hiwat) + so->so_snd.sb_lowat = (int)so->so_snd.sb_hiwat; + SOCKBUF_UNLOCK(&so->so_rcv); + SOCKBUF_UNLOCK(&so->so_snd); + return (0); + + bad: + SOCKBUF_UNLOCK(&so->so_rcv); + SOCKBUF_UNLOCK(&so->so_snd); + return (ENOBUFS); +} + + +/* Taken from /src/sys/kern/uipc_sockbuf.c + * and modified for __Userspace__ + */ + +void +sowakeup(struct socket *so, struct sockbuf *sb) +{ + + SOCKBUF_LOCK_ASSERT(sb); + + sb->sb_flags &= ~SB_SEL; + if (sb->sb_flags & SB_WAIT) { + sb->sb_flags &= ~SB_WAIT; +#if defined(_WIN32) + WakeAllConditionVariable(&(sb)->sb_cond); +#else + pthread_cond_broadcast(&(sb)->sb_cond); +#endif + } + SOCKBUF_UNLOCK(sb); +} + + +/* Taken from /src/sys/kern/uipc_socket.c + * and modified for __Userspace__ + */ + +int +sobind(struct socket *so, struct sockaddr *nam) +{ + switch (nam->sa_family) { +#if defined(INET) + case AF_INET: + return (sctp_bind(so, nam)); +#endif +#if defined(INET6) + case AF_INET6: + return (sctp6_bind(so, nam, NULL)); +#endif + case AF_CONN: + return (sctpconn_bind(so, nam)); + default: + return EAFNOSUPPORT; + } +} + +/* Taken from /src/sys/kern/uipc_syscalls.c + * and modified for __Userspace__ + */ + +int +usrsctp_bind(struct socket *so, struct sockaddr *name, int namelen) +{ + struct sockaddr *sa; + + if (so == NULL) { + errno = EBADF; + return (-1); + } + if ((errno = getsockaddr(&sa, (caddr_t)name, namelen)) != 0) + return (-1); + + errno = sobind(so, sa); + FREE(sa, M_SONAME); + if (errno) { + return (-1); + } else { + return (0); + } +} + +int +userspace_bind(struct socket *so, struct sockaddr *name, int namelen) +{ + return (usrsctp_bind(so, name, namelen)); +} + +/* Taken from /src/sys/kern/uipc_socket.c + * and modified for __Userspace__ + */ + +int +solisten(struct socket *so, int backlog) +{ + if (so == NULL) { + return (EBADF); + } else { + return (sctp_listen(so, backlog, NULL)); + } +} + + +int +solisten_proto_check(struct socket *so) +{ + + SOCK_LOCK_ASSERT(so); + + if (so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | + SS_ISDISCONNECTING)) + return (EINVAL); + return (0); +} + +static int somaxconn = SOMAXCONN; + +void +solisten_proto(struct socket *so, int backlog) +{ + + SOCK_LOCK_ASSERT(so); + + if (backlog < 0 || backlog > somaxconn) + backlog = somaxconn; + so->so_qlimit = backlog; + so->so_options |= SCTP_SO_ACCEPTCONN; +} + + + + +/* Taken from /src/sys/kern/uipc_syscalls.c + * and modified for __Userspace__ + */ + +int +usrsctp_listen(struct socket *so, int backlog) +{ + errno = solisten(so, backlog); + if (errno) { + return (-1); + } else { + return (0); + } +} + +int +userspace_listen(struct socket *so, int backlog) +{ + return (usrsctp_listen(so, backlog)); +} + +/* Taken from /src/sys/kern/uipc_socket.c + * and modified for __Userspace__ + */ + +int +soaccept(struct socket *so, struct sockaddr **nam) +{ + int error; + + SOCK_LOCK(so); + KASSERT((so->so_state & SS_NOFDREF) != 0, ("soaccept: !NOFDREF")); + so->so_state &= ~SS_NOFDREF; + SOCK_UNLOCK(so); + error = sctp_accept(so, nam); + return (error); +} + + + +/* Taken from /src/sys/kern/uipc_syscalls.c + * kern_accept modified for __Userspace__ + */ +int +user_accept(struct socket *head, struct sockaddr **name, socklen_t *namelen, struct socket **ptr_accept_ret_sock) +{ + struct sockaddr *sa = NULL; + int error; + struct socket *so = NULL; + + + if (name) { + *name = NULL; + } + + if ((head->so_options & SCTP_SO_ACCEPTCONN) == 0) { + error = EINVAL; + goto done; + } + + ACCEPT_LOCK(); + if ((head->so_state & SS_NBIO) && TAILQ_EMPTY(&head->so_comp)) { + ACCEPT_UNLOCK(); + error = EWOULDBLOCK; + goto noconnection; + } + while (TAILQ_EMPTY(&head->so_comp) && head->so_error == 0) { + if (head->so_rcv.sb_state & SBS_CANTRCVMORE) { + head->so_error = ECONNABORTED; + break; + } +#if defined(_WIN32) + if (SleepConditionVariableCS(&accept_cond, &accept_mtx, INFINITE)) + error = 0; + else + error = GetLastError(); +#else + error = pthread_cond_wait(&accept_cond, &accept_mtx); +#endif + if (error) { + ACCEPT_UNLOCK(); + goto noconnection; + } + } + if (head->so_error) { + error = head->so_error; + head->so_error = 0; + ACCEPT_UNLOCK(); + goto noconnection; + } + so = TAILQ_FIRST(&head->so_comp); + KASSERT(!(so->so_qstate & SQ_INCOMP), ("accept1: so SQ_INCOMP")); + KASSERT(so->so_qstate & SQ_COMP, ("accept1: so not SQ_COMP")); + + /* + * Before changing the flags on the socket, we have to bump the + * reference count. Otherwise, if the protocol calls sofree(), + * the socket will be released due to a zero refcount. + */ + SOCK_LOCK(so); /* soref() and so_state update */ + soref(so); /* file descriptor reference */ + + TAILQ_REMOVE(&head->so_comp, so, so_list); + head->so_qlen--; + so->so_state |= (head->so_state & SS_NBIO); + so->so_qstate &= ~SQ_COMP; + so->so_head = NULL; + SOCK_UNLOCK(so); + ACCEPT_UNLOCK(); + + + /* + * The original accept returns fd value via td->td_retval[0] = fd; + * we will return the socket for accepted connection. + */ + + error = soaccept(so, &sa); + if (error) { + /* + * return a namelen of zero for older code which might + * ignore the return value from accept. + */ + if (name) + *namelen = 0; + goto noconnection; + } + if (sa == NULL) { + if (name) + *namelen = 0; + goto done; + } + if (name) { +#ifdef HAVE_SA_LEN + /* check sa_len before it is destroyed */ + if (*namelen > sa->sa_len) { + *namelen = sa->sa_len; + } +#else + socklen_t sa_len; + + switch (sa->sa_family) { +#ifdef INET + case AF_INET: + sa_len = sizeof(struct sockaddr_in); + break; +#endif +#ifdef INET6 + case AF_INET6: + sa_len = sizeof(struct sockaddr_in6); + break; +#endif + case AF_CONN: + sa_len = sizeof(struct sockaddr_conn); + break; + default: + sa_len = 0; + break; + } + if (*namelen > sa_len) { + *namelen = sa_len; + } +#endif + *name = sa; + sa = NULL; + } +noconnection: + if (sa) { + FREE(sa, M_SONAME); + } + +done: + *ptr_accept_ret_sock = so; + return (error); +} + + + +/* Taken from /src/sys/kern/uipc_syscalls.c + * and modified for __Userspace__ + */ +/* + * accept1() + */ +static int +accept1(struct socket *so, struct sockaddr *aname, socklen_t *anamelen, struct socket **ptr_accept_ret_sock) +{ + struct sockaddr *name; + socklen_t namelen; + int error; + + if (so == NULL) { + return (EBADF); + } + if (aname == NULL) { + return (user_accept(so, NULL, NULL, ptr_accept_ret_sock)); + } + + error = copyin(anamelen, &namelen, sizeof (namelen)); + if (error) + return (error); + + error = user_accept(so, &name, &namelen, ptr_accept_ret_sock); + + /* + * return a namelen of zero for older code which might + * ignore the return value from accept. + */ + if (error) { + (void) copyout(&namelen, + anamelen, sizeof(*anamelen)); + return (error); + } + + if (error == 0 && name != NULL) { + error = copyout(name, aname, namelen); + } + if (error == 0) { + error = copyout(&namelen, anamelen, sizeof(namelen)); + } + + if (name) { + FREE(name, M_SONAME); + } + return (error); +} + +struct socket * +usrsctp_accept(struct socket *so, struct sockaddr *aname, socklen_t *anamelen) +{ + struct socket *accept_return_sock = NULL; + + errno = accept1(so, aname, anamelen, &accept_return_sock); + if (errno) { + return (NULL); + } else { + return (accept_return_sock); + } +} + +struct socket * +userspace_accept(struct socket *so, struct sockaddr *aname, socklen_t *anamelen) +{ + return (usrsctp_accept(so, aname, anamelen)); +} + +struct socket * +usrsctp_peeloff(struct socket *head, sctp_assoc_t id) +{ + struct socket *so; + + if ((errno = sctp_can_peel_off(head, id)) != 0) { + return (NULL); + } + if ((so = sonewconn(head, SS_ISCONNECTED)) == NULL) { + return (NULL); + } + ACCEPT_LOCK(); + SOCK_LOCK(so); + soref(so); + TAILQ_REMOVE(&head->so_comp, so, so_list); + head->so_qlen--; + so->so_state |= (head->so_state & SS_NBIO); + so->so_qstate &= ~SQ_COMP; + so->so_head = NULL; + SOCK_UNLOCK(so); + ACCEPT_UNLOCK(); + if ((errno = sctp_do_peeloff(head, so, id)) != 0) { + so->so_count = 0; + sodealloc(so); + return (NULL); + } + return (so); +} + +int +sodisconnect(struct socket *so) +{ + int error; + + if ((so->so_state & SS_ISCONNECTED) == 0) + return (ENOTCONN); + if (so->so_state & SS_ISDISCONNECTING) + return (EALREADY); + error = sctp_disconnect(so); + return (error); +} + +int +usrsctp_set_non_blocking(struct socket *so, int onoff) +{ + if (so == NULL) { + errno = EBADF; + return (-1); + } + SOCK_LOCK(so); + if (onoff != 0) { + so->so_state |= SS_NBIO; + } else { + so->so_state &= ~SS_NBIO; + } + SOCK_UNLOCK(so); + return (0); +} + +int +usrsctp_get_non_blocking(struct socket *so) +{ + int result; + + if (so == NULL) { + errno = EBADF; + return (-1); + } + SOCK_LOCK(so); + if (so->so_state & SS_NBIO) { + result = 1; + } else { + result = 0; + } + SOCK_UNLOCK(so); + return (result); +} + +int +soconnect(struct socket *so, struct sockaddr *nam) +{ + int error; + + if (so->so_options & SCTP_SO_ACCEPTCONN) + return (EOPNOTSUPP); + /* + * If protocol is connection-based, can only connect once. + * Otherwise, if connected, try to disconnect first. This allows + * user to disconnect by connecting to, e.g., a null address. + */ + if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && (sodisconnect(so) != 0)) { + error = EISCONN; + } else { + /* + * Prevent accumulated error from previous connection from + * biting us. + */ + so->so_error = 0; + switch (nam->sa_family) { +#if defined(INET) + case AF_INET: + error = sctp_connect(so, nam); + break; +#endif +#if defined(INET6) + case AF_INET6: + error = sctp6_connect(so, nam); + break; +#endif + case AF_CONN: + error = sctpconn_connect(so, nam); + break; + default: + error = EAFNOSUPPORT; + } + } + + return (error); +} + + + +int user_connect(struct socket *so, struct sockaddr *sa) +{ + int error; + int interrupted = 0; + + if (so == NULL) { + error = EBADF; + goto done1; + } + if (so->so_state & SS_ISCONNECTING) { + error = EALREADY; + goto done1; + } + + error = soconnect(so, sa); + if (error) { + goto bad; + } + if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) { + error = EINPROGRESS; + goto done1; + } + + SOCK_LOCK(so); + while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) { +#if defined(_WIN32) + if (SleepConditionVariableCS(SOCK_COND(so), SOCK_MTX(so), INFINITE)) + error = 0; + else + error = -1; +#else + error = pthread_cond_wait(SOCK_COND(so), SOCK_MTX(so)); +#endif + if (error) { +#if defined(__NetBSD__) + if (error == EINTR) { +#else + if (error == EINTR || error == ERESTART) { +#endif + interrupted = 1; + } + break; + } + } + if (error == 0) { + error = so->so_error; + so->so_error = 0; + } + SOCK_UNLOCK(so); + +bad: + if (!interrupted) { + so->so_state &= ~SS_ISCONNECTING; + } +#if !defined(__NetBSD__) + if (error == ERESTART) { + error = EINTR; + } +#endif +done1: + return (error); +} + +int usrsctp_connect(struct socket *so, struct sockaddr *name, int namelen) +{ + struct sockaddr *sa = NULL; + + errno = getsockaddr(&sa, (caddr_t)name, namelen); + if (errno) + return (-1); + + errno = user_connect(so, sa); + FREE(sa, M_SONAME); + if (errno) { + return (-1); + } else { + return (0); + } +} + +int userspace_connect(struct socket *so, struct sockaddr *name, int namelen) +{ + return (usrsctp_connect(so, name, namelen)); +} + +#define SCTP_STACK_BUF_SIZE 2048 + +void +usrsctp_close(struct socket *so) { + if (so != NULL) { + if (so->so_options & SCTP_SO_ACCEPTCONN) { + struct socket *sp; + + ACCEPT_LOCK(); + while ((sp = TAILQ_FIRST(&so->so_comp)) != NULL) { + TAILQ_REMOVE(&so->so_comp, sp, so_list); + so->so_qlen--; + sp->so_qstate &= ~SQ_COMP; + sp->so_head = NULL; + ACCEPT_UNLOCK(); + soabort(sp); + ACCEPT_LOCK(); + } + ACCEPT_UNLOCK(); + } + ACCEPT_LOCK(); + SOCK_LOCK(so); + sorele(so); + } +} + +void +userspace_close(struct socket *so) +{ + usrsctp_close(so); +} + +int +usrsctp_shutdown(struct socket *so, int how) +{ + if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) { + errno = EINVAL; + return (-1); + } + if (so == NULL) { + errno = EBADF; + return (-1); + } + sctp_flush(so, how); + if (how != SHUT_WR) + socantrcvmore(so); + if (how != SHUT_RD) { + errno = sctp_shutdown(so); + if (errno) { + return (-1); + } else { + return (0); + } + } + return (0); +} + +int +userspace_shutdown(struct socket *so, int how) +{ + return (usrsctp_shutdown(so, how)); +} + +int +usrsctp_finish(void) +{ + if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) { + return (0); + } + if (SCTP_INP_INFO_TRYLOCK()) { + if (!LIST_EMPTY(&SCTP_BASE_INFO(listhead))) { + SCTP_INP_INFO_RUNLOCK(); + return (-1); + } + SCTP_INP_INFO_RUNLOCK(); + } else { + return (-1); + } + sctp_finish(); +#if defined(_WIN32) + DeleteConditionVariable(&accept_cond); + DeleteCriticalSection(&accept_mtx); +#if defined(INET) || defined(INET6) + WSACleanup(); +#endif +#else + pthread_cond_destroy(&accept_cond); + pthread_mutex_destroy(&accept_mtx); +#endif + return (0); +} + +int +userspace_finish(void) +{ + return (usrsctp_finish()); +} + +/* needed from sctp_usrreq.c */ +int +sctp_setopt(struct socket *so, int optname, void *optval, size_t optsize, void *p); + +int +usrsctp_setsockopt(struct socket *so, int level, int option_name, + const void *option_value, socklen_t option_len) +{ + if (so == NULL) { + errno = EBADF; + return (-1); + } + switch (level) { + case SOL_SOCKET: + { + switch (option_name) { + case SO_RCVBUF: + if (option_len < (socklen_t)sizeof(int)) { + errno = EINVAL; + return (-1); + } else { + int *buf_size; + + buf_size = (int *)option_value; + if (*buf_size < 1) { + errno = EINVAL; + return (-1); + } + sbreserve(&so->so_rcv, (u_long)*buf_size, so); + return (0); + } + break; + case SO_SNDBUF: + if (option_len < (socklen_t)sizeof(int)) { + errno = EINVAL; + return (-1); + } else { + int *buf_size; + + buf_size = (int *)option_value; + if (*buf_size < 1) { + errno = EINVAL; + return (-1); + } + sbreserve(&so->so_snd, (u_long)*buf_size, so); + return (0); + } + break; + case SO_LINGER: + if (option_len < (socklen_t)sizeof(struct linger)) { + errno = EINVAL; + return (-1); + } else { + struct linger *l; + + l = (struct linger *)option_value; + so->so_linger = l->l_linger; + if (l->l_onoff) { + so->so_options |= SCTP_SO_LINGER; + } else { + so->so_options &= ~SCTP_SO_LINGER; + } + return (0); + } + default: + errno = EINVAL; + return (-1); + } + } + case IPPROTO_SCTP: + errno = sctp_setopt(so, option_name, (void *) option_value, (size_t)option_len, NULL); + if (errno) { + return (-1); + } else { + return (0); + } + default: + errno = ENOPROTOOPT; + return (-1); + } +} + +int +userspace_setsockopt(struct socket *so, int level, int option_name, + const void *option_value, socklen_t option_len) +{ + return (usrsctp_setsockopt(so, level, option_name, option_value, option_len)); +} + +/* needed from sctp_usrreq.c */ +int +sctp_getopt(struct socket *so, int optname, void *optval, size_t *optsize, + void *p); + +int +usrsctp_getsockopt(struct socket *so, int level, int option_name, + void *option_value, socklen_t *option_len) +{ + if (so == NULL) { + errno = EBADF; + return (-1); + } + if (option_len == NULL) { + errno = EFAULT; + return (-1); + } + switch (level) { + case SOL_SOCKET: + switch (option_name) { + case SO_RCVBUF: + if (*option_len < (socklen_t)sizeof(int)) { + errno = EINVAL; + return (-1); + } else { + int *buf_size; + + buf_size = (int *)option_value; + *buf_size = so->so_rcv.sb_hiwat; + *option_len = (socklen_t)sizeof(int); + return (0); + } + break; + case SO_SNDBUF: + if (*option_len < (socklen_t)sizeof(int)) { + errno = EINVAL; + return (-1); + } else { + int *buf_size; + + buf_size = (int *)option_value; + *buf_size = so->so_snd.sb_hiwat; + *option_len = (socklen_t)sizeof(int); + return (0); + } + break; + case SO_LINGER: + if (*option_len < (socklen_t)sizeof(struct linger)) { + errno = EINVAL; + return (-1); + } else { + struct linger *l; + + l = (struct linger *)option_value; + l->l_linger = so->so_linger; + if (so->so_options & SCTP_SO_LINGER) { + l->l_onoff = 1; + } else { + l->l_onoff = 0; + } + *option_len = (socklen_t)sizeof(struct linger); + return (0); + } + break; + case SO_ERROR: + if (*option_len < (socklen_t)sizeof(int)) { + errno = EINVAL; + return (-1); + } else { + int *intval; + + intval = (int *)option_value; + *intval = so->so_error; + *option_len = (socklen_t)sizeof(int); + return (0); + } + break; + default: + errno = EINVAL; + return (-1); + } + case IPPROTO_SCTP: + { + size_t len; + + len = (size_t)*option_len; + errno = sctp_getopt(so, option_name, option_value, &len, NULL); + *option_len = (socklen_t)len; + if (errno) { + return (-1); + } else { + return (0); + } + } + default: + errno = ENOPROTOOPT; + return (-1); + } +} + +int +userspace_getsockopt(struct socket *so, int level, int option_name, + void *option_value, socklen_t *option_len) +{ + return (usrsctp_getsockopt(so, level, option_name, option_value, option_len)); +} + +int +usrsctp_opt_info(struct socket *so, sctp_assoc_t id, int opt, void *arg, socklen_t *size) +{ + if (arg == NULL) { + errno = EINVAL; + return (-1); + } + if ((id == SCTP_CURRENT_ASSOC) || + (id == SCTP_ALL_ASSOC)) { + errno = EINVAL; + return (-1); + } + switch (opt) { + case SCTP_RTOINFO: + ((struct sctp_rtoinfo *)arg)->srto_assoc_id = id; + break; + case SCTP_ASSOCINFO: + ((struct sctp_assocparams *)arg)->sasoc_assoc_id = id; + break; + case SCTP_DEFAULT_SEND_PARAM: + ((struct sctp_assocparams *)arg)->sasoc_assoc_id = id; + break; + case SCTP_PRIMARY_ADDR: + ((struct sctp_setprim *)arg)->ssp_assoc_id = id; + break; + case SCTP_PEER_ADDR_PARAMS: + ((struct sctp_paddrparams *)arg)->spp_assoc_id = id; + break; + case SCTP_MAXSEG: + ((struct sctp_assoc_value *)arg)->assoc_id = id; + break; + case SCTP_AUTH_KEY: + ((struct sctp_authkey *)arg)->sca_assoc_id = id; + break; + case SCTP_AUTH_ACTIVE_KEY: + ((struct sctp_authkeyid *)arg)->scact_assoc_id = id; + break; + case SCTP_DELAYED_SACK: + ((struct sctp_sack_info *)arg)->sack_assoc_id = id; + break; + case SCTP_CONTEXT: + ((struct sctp_assoc_value *)arg)->assoc_id = id; + break; + case SCTP_STATUS: + ((struct sctp_status *)arg)->sstat_assoc_id = id; + break; + case SCTP_GET_PEER_ADDR_INFO: + ((struct sctp_paddrinfo *)arg)->spinfo_assoc_id = id; + break; + case SCTP_PEER_AUTH_CHUNKS: + ((struct sctp_authchunks *)arg)->gauth_assoc_id = id; + break; + case SCTP_LOCAL_AUTH_CHUNKS: + ((struct sctp_authchunks *)arg)->gauth_assoc_id = id; + break; + case SCTP_TIMEOUTS: + ((struct sctp_timeouts *)arg)->stimo_assoc_id = id; + break; + case SCTP_EVENT: + ((struct sctp_event *)arg)->se_assoc_id = id; + break; + case SCTP_DEFAULT_SNDINFO: + ((struct sctp_sndinfo *)arg)->snd_assoc_id = id; + break; + case SCTP_DEFAULT_PRINFO: + ((struct sctp_default_prinfo *)arg)->pr_assoc_id = id; + break; + case SCTP_PEER_ADDR_THLDS: + ((struct sctp_paddrthlds *)arg)->spt_assoc_id = id; + break; + case SCTP_REMOTE_UDP_ENCAPS_PORT: + ((struct sctp_udpencaps *)arg)->sue_assoc_id = id; + break; + case SCTP_ECN_SUPPORTED: + ((struct sctp_assoc_value *)arg)->assoc_id = id; + break; + case SCTP_PR_SUPPORTED: + ((struct sctp_assoc_value *)arg)->assoc_id = id; + break; + case SCTP_AUTH_SUPPORTED: + ((struct sctp_assoc_value *)arg)->assoc_id = id; + break; + case SCTP_ASCONF_SUPPORTED: + ((struct sctp_assoc_value *)arg)->assoc_id = id; + break; + case SCTP_RECONFIG_SUPPORTED: + ((struct sctp_assoc_value *)arg)->assoc_id = id; + break; + case SCTP_NRSACK_SUPPORTED: + ((struct sctp_assoc_value *)arg)->assoc_id = id; + break; + case SCTP_PKTDROP_SUPPORTED: + ((struct sctp_assoc_value *)arg)->assoc_id = id; + break; + case SCTP_MAX_BURST: + ((struct sctp_assoc_value *)arg)->assoc_id = id; + break; + case SCTP_ENABLE_STREAM_RESET: + ((struct sctp_assoc_value *)arg)->assoc_id = id; + break; + case SCTP_PR_STREAM_STATUS: + ((struct sctp_prstatus *)arg)->sprstat_assoc_id = id; + break; + case SCTP_PR_ASSOC_STATUS: + ((struct sctp_prstatus *)arg)->sprstat_assoc_id = id; + break; + case SCTP_MAX_CWND: + ((struct sctp_assoc_value *)arg)->assoc_id = id; + break; + default: + break; + } + return (usrsctp_getsockopt(so, IPPROTO_SCTP, opt, arg, size)); +} + +int +usrsctp_set_ulpinfo(struct socket *so, void *ulp_info) +{ + return (register_ulp_info(so, ulp_info)); +} + + +int +usrsctp_get_ulpinfo(struct socket *so, void **pulp_info) +{ + return (retrieve_ulp_info(so, pulp_info)); +} + +int +usrsctp_bindx(struct socket *so, struct sockaddr *addrs, int addrcnt, int flags) +{ + struct sockaddr *sa; +#ifdef INET + struct sockaddr_in *sin; +#endif +#ifdef INET6 + struct sockaddr_in6 *sin6; +#endif + int i; +#if defined(INET) || defined(INET6) + uint16_t sport; + bool fix_port; +#endif + + /* validate the flags */ + if ((flags != SCTP_BINDX_ADD_ADDR) && + (flags != SCTP_BINDX_REM_ADDR)) { + errno = EFAULT; + return (-1); + } + /* validate the address count and list */ + if ((addrcnt <= 0) || (addrs == NULL)) { + errno = EINVAL; + return (-1); + } +#if defined(INET) || defined(INET6) + sport = 0; + fix_port = false; +#endif + /* First pre-screen the addresses */ + sa = addrs; + for (i = 0; i < addrcnt; i++) { + switch (sa->sa_family) { +#ifdef INET + case AF_INET: +#ifdef HAVE_SA_LEN + if (sa->sa_len != sizeof(struct sockaddr_in)) { + errno = EINVAL; + return (-1); + } +#endif + sin = (struct sockaddr_in *)sa; + if (sin->sin_port) { + /* non-zero port, check or save */ + if (sport) { + /* Check against our port */ + if (sport != sin->sin_port) { + errno = EINVAL; + return (-1); + } + } else { + /* save off the port */ + sport = sin->sin_port; + fix_port = (i > 0); + } + } +#ifndef HAVE_SA_LEN + sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in)); +#endif + break; +#endif +#ifdef INET6 + case AF_INET6: +#ifdef HAVE_SA_LEN + if (sa->sa_len != sizeof(struct sockaddr_in6)) { + errno = EINVAL; + return (-1); + } +#endif + sin6 = (struct sockaddr_in6 *)sa; + if (sin6->sin6_port) { + /* non-zero port, check or save */ + if (sport) { + /* Check against our port */ + if (sport != sin6->sin6_port) { + errno = EINVAL; + return (-1); + } + } else { + /* save off the port */ + sport = sin6->sin6_port; + fix_port = (i > 0); + } + } +#ifndef HAVE_SA_LEN + sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in6)); +#endif + break; +#endif + default: + /* Invalid address family specified. */ + errno = EAFNOSUPPORT; + return (-1); + } +#ifdef HAVE_SA_LEN + sa = (struct sockaddr *)((caddr_t)sa + sa->sa_len); +#endif + } + sa = addrs; + for (i = 0; i < addrcnt; i++) { +#ifndef HAVE_SA_LEN + size_t sa_len; + +#endif +#ifdef HAVE_SA_LEN +#if defined(INET) || defined(INET6) + if (fix_port) { + switch (sa->sa_family) { +#ifdef INET + case AF_INET: + ((struct sockaddr_in *)sa)->sin_port = sport; + break; +#endif +#ifdef INET6 + case AF_INET6: + ((struct sockaddr_in6 *)sa)->sin6_port = sport; + break; +#endif + } + } +#endif + if (usrsctp_setsockopt(so, IPPROTO_SCTP, flags, sa, sa->sa_len) != 0) { + return (-1); + } + sa = (struct sockaddr *)((caddr_t)sa + sa->sa_len); +#else + switch (sa->sa_family) { +#ifdef INET + case AF_INET: + sa_len = sizeof(struct sockaddr_in); + break; +#endif +#ifdef INET6 + case AF_INET6: + sa_len = sizeof(struct sockaddr_in6); + break; +#endif + default: + sa_len = 0; + break; + } + /* + * Now, if there was a port mentioned, assure that the + * first address has that port to make sure it fails or + * succeeds correctly. + */ +#if defined(INET) || defined(INET6) + if (fix_port) { + switch (sa->sa_family) { +#ifdef INET + case AF_INET: + ((struct sockaddr_in *)sa)->sin_port = sport; + break; +#endif +#ifdef INET6 + case AF_INET6: + ((struct sockaddr_in6 *)sa)->sin6_port = sport; + break; +#endif + } + } +#endif + if (usrsctp_setsockopt(so, IPPROTO_SCTP, flags, sa, (socklen_t)sa_len) != 0) { + return (-1); + } + sa = (struct sockaddr *)((caddr_t)sa + sa_len); +#endif + } + return (0); +} + +int +usrsctp_connectx(struct socket *so, + const struct sockaddr *addrs, int addrcnt, + sctp_assoc_t *id) +{ +#if defined(INET) || defined(INET6) + char buf[SCTP_STACK_BUF_SIZE]; + int i, ret, cnt, *aa; + char *cpto; + const struct sockaddr *at; + sctp_assoc_t *p_id; + size_t len = sizeof(int); + + /* validate the address count and list */ + if ((addrs == NULL) || (addrcnt <= 0)) { + errno = EINVAL; + return (-1); + } + at = addrs; + cnt = 0; + cpto = ((caddr_t)buf + sizeof(int)); + /* validate all the addresses and get the size */ + for (i = 0; i < addrcnt; i++) { + switch (at->sa_family) { +#ifdef INET + case AF_INET: +#ifdef HAVE_SA_LEN + if (at->sa_len != sizeof(struct sockaddr_in)) { + errno = EINVAL; + return (-1); + } +#endif + len += sizeof(struct sockaddr_in); + if (len > SCTP_STACK_BUF_SIZE) { + errno = ENOMEM; + return (-1); + } + memcpy(cpto, at, sizeof(struct sockaddr_in)); + cpto = ((caddr_t)cpto + sizeof(struct sockaddr_in)); + at = (struct sockaddr *)((caddr_t)at + sizeof(struct sockaddr_in)); + break; +#endif +#ifdef INET6 + case AF_INET6: +#ifdef HAVE_SA_LEN + if (at->sa_len != sizeof(struct sockaddr_in6)) { + errno = EINVAL; + return (-1); + } +#endif +#ifdef INET + if (IN6_IS_ADDR_V4MAPPED(&((struct sockaddr_in6 *)at)->sin6_addr)) { + len += sizeof(struct sockaddr_in); + if (len > SCTP_STACK_BUF_SIZE) { + errno = ENOMEM; + return (-1); + } + in6_sin6_2_sin((struct sockaddr_in *)cpto, (struct sockaddr_in6 *)at); + cpto = ((caddr_t)cpto + sizeof(struct sockaddr_in)); + } else { + len += sizeof(struct sockaddr_in6); + if (len > SCTP_STACK_BUF_SIZE) { + errno = ENOMEM; + return (-1); + } + memcpy(cpto, at, sizeof(struct sockaddr_in6)); + cpto = ((caddr_t)cpto + sizeof(struct sockaddr_in6)); + } +#else + len += sizeof(struct sockaddr_in6); + if (len > SCTP_STACK_BUF_SIZE) { + errno = ENOMEM; + return (-1); + } + memcpy(cpto, at, sizeof(struct sockaddr_in6)); + cpto = ((caddr_t)cpto + sizeof(struct sockaddr_in6)); +#endif + at = (struct sockaddr *)((caddr_t)at + sizeof(struct sockaddr_in6)); + break; +#endif + default: + errno = EINVAL; + return (-1); + } + cnt++; + } + aa = (int *)buf; + *aa = cnt; + ret = usrsctp_setsockopt(so, IPPROTO_SCTP, SCTP_CONNECT_X, (void *)buf, (socklen_t)len); + if ((ret == 0) && id) { + p_id = (sctp_assoc_t *)buf; + *id = *p_id; + } + return (ret); +#else + errno = EINVAL; + return (-1); +#endif +} + +int +usrsctp_getpaddrs(struct socket *so, sctp_assoc_t id, struct sockaddr **raddrs) +{ + struct sctp_getaddresses *addrs; + struct sockaddr *sa; + caddr_t lim; + socklen_t opt_len; + uint32_t size_of_addresses; + int cnt; + + if (raddrs == NULL) { + errno = EFAULT; + return (-1); + } + /* When calling getsockopt(), the value contains the assoc_id. */ + size_of_addresses = (uint32_t)id; + opt_len = (socklen_t)sizeof(uint32_t); + if (usrsctp_getsockopt(so, IPPROTO_SCTP, SCTP_GET_REMOTE_ADDR_SIZE, &size_of_addresses, &opt_len) != 0) { + if (errno == ENOENT) { + return (0); + } else { + return (-1); + } + } + opt_len = (socklen_t)((size_t)size_of_addresses + sizeof(struct sctp_getaddresses)); + addrs = calloc(1, (size_t)opt_len); + if (addrs == NULL) { + errno = ENOMEM; + return (-1); + } + addrs->sget_assoc_id = id; + /* Now lets get the array of addresses */ + if (usrsctp_getsockopt(so, IPPROTO_SCTP, SCTP_GET_PEER_ADDRESSES, addrs, &opt_len) != 0) { + free(addrs); + return (-1); + } + *raddrs = &addrs->addr[0].sa; + cnt = 0; + sa = &addrs->addr[0].sa; + lim = (caddr_t)addrs + opt_len; +#ifdef HAVE_SA_LEN + while (((caddr_t)sa < lim) && (sa->sa_len > 0)) { + sa = (struct sockaddr *)((caddr_t)sa + sa->sa_len); +#else + while ((caddr_t)sa < lim) { + switch (sa->sa_family) { +#ifdef INET + case AF_INET: + sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in)); + break; +#endif +#ifdef INET6 + case AF_INET6: + sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in6)); + break; +#endif + case AF_CONN: + sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_conn)); + break; + default: + return (cnt); + break; + } +#endif + cnt++; + } + return (cnt); +} + +void +usrsctp_freepaddrs(struct sockaddr *addrs) +{ + /* Take away the hidden association id */ + void *fr_addr; + + fr_addr = (void *)((caddr_t)addrs - offsetof(struct sctp_getaddresses, addr)); + /* Now free it */ + free(fr_addr); +} + +int +usrsctp_getladdrs(struct socket *so, sctp_assoc_t id, struct sockaddr **raddrs) +{ + struct sctp_getaddresses *addrs; + struct sockaddr *sa; + caddr_t lim; + socklen_t opt_len; + uint32_t size_of_addresses; + int cnt; + + if (raddrs == NULL) { + errno = EFAULT; + return (-1); + } + size_of_addresses = 0; + opt_len = (socklen_t)sizeof(uint32_t); + if (usrsctp_getsockopt(so, IPPROTO_SCTP, SCTP_GET_LOCAL_ADDR_SIZE, &size_of_addresses, &opt_len) != 0) { + return (-1); + } + opt_len = (socklen_t)(size_of_addresses + sizeof(struct sctp_getaddresses)); + addrs = calloc(1, (size_t)opt_len); + if (addrs == NULL) { + errno = ENOMEM; + return (-1); + } + addrs->sget_assoc_id = id; + /* Now lets get the array of addresses */ + if (usrsctp_getsockopt(so, IPPROTO_SCTP, SCTP_GET_LOCAL_ADDRESSES, addrs, &opt_len) != 0) { + free(addrs); + return (-1); + } + if (size_of_addresses == 0) { + free(addrs); + return (0); + } + *raddrs = &addrs->addr[0].sa; + cnt = 0; + sa = &addrs->addr[0].sa; + lim = (caddr_t)addrs + opt_len; +#ifdef HAVE_SA_LEN + while (((caddr_t)sa < lim) && (sa->sa_len > 0)) { + sa = (struct sockaddr *)((caddr_t)sa + sa->sa_len); +#else + while ((caddr_t)sa < lim) { + switch (sa->sa_family) { +#ifdef INET + case AF_INET: + sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in)); + break; +#endif +#ifdef INET6 + case AF_INET6: + sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in6)); + break; +#endif + case AF_CONN: + sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_conn)); + break; + default: + return (cnt); + break; + } +#endif + cnt++; + } + return (cnt); +} + +void +usrsctp_freeladdrs(struct sockaddr *addrs) +{ + /* Take away the hidden association id */ + void *fr_addr; + + fr_addr = (void *)((caddr_t)addrs - offsetof(struct sctp_getaddresses, addr)); + /* Now free it */ + free(fr_addr); +} + +#ifdef INET +void +sctp_userspace_ip_output(int *result, struct mbuf *o_pak, + sctp_route_t *ro, void *inp, + uint32_t vrf_id) +{ + struct mbuf *m; + struct mbuf *m_orig; + int iovcnt; + int len; + struct ip *ip; + struct udphdr *udp; + struct sockaddr_in dst; +#if defined(_WIN32) + WSAMSG win_msg_hdr; + DWORD win_sent_len; + WSABUF send_iovec[MAXLEN_MBUF_CHAIN]; + WSABUF winbuf; +#else + struct iovec send_iovec[MAXLEN_MBUF_CHAIN]; + struct msghdr msg_hdr; +#endif + int use_udp_tunneling; + + *result = 0; + + m = SCTP_HEADER_TO_CHAIN(o_pak); + m_orig = m; + + len = sizeof(struct ip); + if (SCTP_BUF_LEN(m) < len) { + if ((m = m_pullup(m, len)) == 0) { + SCTP_PRINTF("Can not get the IP header in the first mbuf.\n"); + return; + } + } + ip = mtod(m, struct ip *); + use_udp_tunneling = (ip->ip_p == IPPROTO_UDP); + + if (use_udp_tunneling) { + len = sizeof(struct ip) + sizeof(struct udphdr); + if (SCTP_BUF_LEN(m) < len) { + if ((m = m_pullup(m, len)) == 0) { + SCTP_PRINTF("Can not get the UDP/IP header in the first mbuf.\n"); + return; + } + ip = mtod(m, struct ip *); + } + udp = (struct udphdr *)(ip + 1); + } else { + udp = NULL; + } + + if (!use_udp_tunneling) { + if (ip->ip_src.s_addr == INADDR_ANY) { + /* TODO get addr of outgoing interface */ + SCTP_PRINTF("Why did the SCTP implementation did not choose a source address?\n"); + } + /* TODO need to worry about ro->ro_dst as in ip_output? */ +#if defined(__linux__) || defined(_WIN32) || (defined(__FreeBSD__) && (__FreeBSD_version >= 1100030)) + /* need to put certain fields into network order for Linux */ + ip->ip_len = htons(ip->ip_len); +#endif + } + + memset((void *)&dst, 0, sizeof(struct sockaddr_in)); + dst.sin_family = AF_INET; + dst.sin_addr.s_addr = ip->ip_dst.s_addr; +#ifdef HAVE_SIN_LEN + dst.sin_len = sizeof(struct sockaddr_in); +#endif + if (use_udp_tunneling) { + dst.sin_port = udp->uh_dport; + } else { + dst.sin_port = 0; + } + + /* tweak the mbuf chain */ + if (use_udp_tunneling) { + m_adj(m, sizeof(struct ip) + sizeof(struct udphdr)); + } + + for (iovcnt = 0; m != NULL && iovcnt < MAXLEN_MBUF_CHAIN; m = m->m_next, iovcnt++) { +#if !defined(_WIN32) + send_iovec[iovcnt].iov_base = (caddr_t)m->m_data; + send_iovec[iovcnt].iov_len = SCTP_BUF_LEN(m); +#else + send_iovec[iovcnt].buf = (caddr_t)m->m_data; + send_iovec[iovcnt].len = SCTP_BUF_LEN(m); +#endif + } + + if (m != NULL) { + SCTP_PRINTF("mbuf chain couldn't be copied completely\n"); + goto free_mbuf; + } + +#if !defined(_WIN32) + msg_hdr.msg_name = (struct sockaddr *) &dst; + msg_hdr.msg_namelen = sizeof(struct sockaddr_in); + msg_hdr.msg_iov = send_iovec; + msg_hdr.msg_iovlen = iovcnt; + msg_hdr.msg_control = NULL; + msg_hdr.msg_controllen = 0; + msg_hdr.msg_flags = 0; + + if ((!use_udp_tunneling) && (SCTP_BASE_VAR(userspace_rawsctp) != -1)) { + if (sendmsg(SCTP_BASE_VAR(userspace_rawsctp), &msg_hdr, MSG_DONTWAIT) < 0) { + *result = errno; + } + } + if ((use_udp_tunneling) && (SCTP_BASE_VAR(userspace_udpsctp) != -1)) { + if (sendmsg(SCTP_BASE_VAR(userspace_udpsctp), &msg_hdr, MSG_DONTWAIT) < 0) { + *result = errno; + } + } +#else + win_msg_hdr.name = (struct sockaddr *) &dst; + win_msg_hdr.namelen = sizeof(struct sockaddr_in); + win_msg_hdr.lpBuffers = (LPWSABUF)send_iovec; + win_msg_hdr.dwBufferCount = iovcnt; + winbuf.len = 0; + winbuf.buf = NULL; + win_msg_hdr.Control = winbuf; + win_msg_hdr.dwFlags = 0; + + if ((!use_udp_tunneling) && (SCTP_BASE_VAR(userspace_rawsctp) != -1)) { + if (WSASendTo(SCTP_BASE_VAR(userspace_rawsctp), (LPWSABUF) send_iovec, iovcnt, &win_sent_len, win_msg_hdr.dwFlags, win_msg_hdr.name, (int) win_msg_hdr.namelen, NULL, NULL) != 0) { + *result = WSAGetLastError(); + } + } + if ((use_udp_tunneling) && (SCTP_BASE_VAR(userspace_udpsctp) != -1)) { + if (WSASendTo(SCTP_BASE_VAR(userspace_udpsctp), (LPWSABUF) send_iovec, iovcnt, &win_sent_len, win_msg_hdr.dwFlags, win_msg_hdr.name, (int) win_msg_hdr.namelen, NULL, NULL) != 0) { + *result = WSAGetLastError(); + } + } +#endif +free_mbuf: + sctp_m_freem(m_orig); +} +#endif + +#if defined(INET6) +void sctp_userspace_ip6_output(int *result, struct mbuf *o_pak, + struct route_in6 *ro, void *inp, + uint32_t vrf_id) +{ + struct mbuf *m; + struct mbuf *m_orig; + int iovcnt; + int len; + struct ip6_hdr *ip6; + struct udphdr *udp; + struct sockaddr_in6 dst; +#if defined(_WIN32) + WSAMSG win_msg_hdr; + DWORD win_sent_len; + WSABUF send_iovec[MAXLEN_MBUF_CHAIN]; + WSABUF winbuf; +#else + struct iovec send_iovec[MAXLEN_MBUF_CHAIN]; + struct msghdr msg_hdr; +#endif + int use_udp_tunneling; + + *result = 0; + + m = SCTP_HEADER_TO_CHAIN(o_pak); + m_orig = m; + + len = sizeof(struct ip6_hdr); + + if (SCTP_BUF_LEN(m) < len) { + if ((m = m_pullup(m, len)) == 0) { + SCTP_PRINTF("Can not get the IP header in the first mbuf.\n"); + return; + } + } + + ip6 = mtod(m, struct ip6_hdr *); + use_udp_tunneling = (ip6->ip6_nxt == IPPROTO_UDP); + + if (use_udp_tunneling) { + len = sizeof(struct ip6_hdr) + sizeof(struct udphdr); + if (SCTP_BUF_LEN(m) < len) { + if ((m = m_pullup(m, len)) == 0) { + SCTP_PRINTF("Can not get the UDP/IP header in the first mbuf.\n"); + return; + } + ip6 = mtod(m, struct ip6_hdr *); + } + udp = (struct udphdr *)(ip6 + 1); + } else { + udp = NULL; + } + + if (!use_udp_tunneling) { + if (ip6->ip6_src.s6_addr == in6addr_any.s6_addr) { + /* TODO get addr of outgoing interface */ + SCTP_PRINTF("Why did the SCTP implementation did not choose a source address?\n"); + } + /* TODO need to worry about ro->ro_dst as in ip_output? */ + } + + memset((void *)&dst, 0, sizeof(struct sockaddr_in6)); + dst.sin6_family = AF_INET6; + dst.sin6_addr = ip6->ip6_dst; +#ifdef HAVE_SIN6_LEN + dst.sin6_len = sizeof(struct sockaddr_in6); +#endif + + if (use_udp_tunneling) { + dst.sin6_port = udp->uh_dport; + } else { + dst.sin6_port = 0; + } + + /* tweak the mbuf chain */ + if (use_udp_tunneling) { + m_adj(m, sizeof(struct ip6_hdr) + sizeof(struct udphdr)); + } else { + m_adj(m, sizeof(struct ip6_hdr)); + } + + for (iovcnt = 0; m != NULL && iovcnt < MAXLEN_MBUF_CHAIN; m = m->m_next, iovcnt++) { +#if !defined(_WIN32) + send_iovec[iovcnt].iov_base = (caddr_t)m->m_data; + send_iovec[iovcnt].iov_len = SCTP_BUF_LEN(m); +#else + send_iovec[iovcnt].buf = (caddr_t)m->m_data; + send_iovec[iovcnt].len = SCTP_BUF_LEN(m); +#endif + } + if (m != NULL) { + SCTP_PRINTF("mbuf chain couldn't be copied completely\n"); + goto free_mbuf; + } + +#if !defined(_WIN32) + msg_hdr.msg_name = (struct sockaddr *) &dst; + msg_hdr.msg_namelen = sizeof(struct sockaddr_in6); + msg_hdr.msg_iov = send_iovec; + msg_hdr.msg_iovlen = iovcnt; + msg_hdr.msg_control = NULL; + msg_hdr.msg_controllen = 0; + msg_hdr.msg_flags = 0; + + if ((!use_udp_tunneling) && (SCTP_BASE_VAR(userspace_rawsctp6) != -1)) { + if (sendmsg(SCTP_BASE_VAR(userspace_rawsctp6), &msg_hdr, MSG_DONTWAIT)< 0) { + *result = errno; + } + } + if ((use_udp_tunneling) && (SCTP_BASE_VAR(userspace_udpsctp6) != -1)) { + if (sendmsg(SCTP_BASE_VAR(userspace_udpsctp6), &msg_hdr, MSG_DONTWAIT) < 0) { + *result = errno; + } + } +#else + win_msg_hdr.name = (struct sockaddr *) &dst; + win_msg_hdr.namelen = sizeof(struct sockaddr_in6); + win_msg_hdr.lpBuffers = (LPWSABUF)send_iovec; + win_msg_hdr.dwBufferCount = iovcnt; + winbuf.len = 0; + winbuf.buf = NULL; + win_msg_hdr.Control = winbuf; + win_msg_hdr.dwFlags = 0; + + if ((!use_udp_tunneling) && (SCTP_BASE_VAR(userspace_rawsctp6) != -1)) { + if (WSASendTo(SCTP_BASE_VAR(userspace_rawsctp6), (LPWSABUF) send_iovec, iovcnt, &win_sent_len, win_msg_hdr.dwFlags, win_msg_hdr.name, (int) win_msg_hdr.namelen, NULL, NULL) != 0) { + *result = WSAGetLastError(); + } + } + if ((use_udp_tunneling) && (SCTP_BASE_VAR(userspace_udpsctp6) != -1)) { + if (WSASendTo(SCTP_BASE_VAR(userspace_udpsctp6), (LPWSABUF) send_iovec, iovcnt, &win_sent_len, win_msg_hdr.dwFlags, win_msg_hdr.name, (int) win_msg_hdr.namelen, NULL, NULL) != 0) { + *result = WSAGetLastError(); + } + } +#endif +free_mbuf: + sctp_m_freem(m_orig); +} +#endif + +void +usrsctp_register_address(void *addr) +{ + struct sockaddr_conn sconn; + + memset(&sconn, 0, sizeof(struct sockaddr_conn)); + sconn.sconn_family = AF_CONN; +#ifdef HAVE_SCONN_LEN + sconn.sconn_len = sizeof(struct sockaddr_conn); +#endif + sconn.sconn_port = 0; + sconn.sconn_addr = addr; + sctp_add_addr_to_vrf(SCTP_DEFAULT_VRFID, + NULL, + 0xffffffff, + 0, + "conn", + NULL, + (struct sockaddr *)&sconn, + 0, + 0); +} + +void +usrsctp_deregister_address(void *addr) +{ + struct sockaddr_conn sconn; + + memset(&sconn, 0, sizeof(struct sockaddr_conn)); + sconn.sconn_family = AF_CONN; +#ifdef HAVE_SCONN_LEN + sconn.sconn_len = sizeof(struct sockaddr_conn); +#endif + sconn.sconn_port = 0; + sconn.sconn_addr = addr; + sctp_del_addr_from_vrf(SCTP_DEFAULT_VRFID, + (struct sockaddr *)&sconn, + 0xffffffff, + "conn"); +} + +#define PREAMBLE_FORMAT "\n%c %02d:%02d:%02d.%06ld " +#define PREAMBLE_LENGTH 19 +#define HEADER "0000 " +#define TRAILER "# SCTP_PACKET\n" + +char * +usrsctp_dumppacket(const void *buf, size_t len, int outbound) +{ + size_t i, pos; + char *dump_buf, *packet; + struct tm t; +#ifdef _WIN32 + struct timeb tb; +#else + struct timeval tv; + time_t sec; +#endif + + if ((len == 0) || (buf == NULL)) { + return (NULL); + } + if ((dump_buf = malloc(PREAMBLE_LENGTH + strlen(HEADER) + 3 * len + strlen(TRAILER) + 1)) == NULL) { + return (NULL); + } + pos = 0; +#ifdef _WIN32 + ftime(&tb); + localtime_s(&t, &tb.time); +#if defined(__MINGW32__) + if (snprintf(dump_buf, PREAMBLE_LENGTH + 1, PREAMBLE_FORMAT, + outbound ? 'O' : 'I', + t.tm_hour, t.tm_min, t.tm_sec, (long)(1000 * tb.millitm)) < 0) { + free(dump_buf); + return (NULL); + } +#else + if (_snprintf_s(dump_buf, PREAMBLE_LENGTH + 1, PREAMBLE_LENGTH, PREAMBLE_FORMAT, + outbound ? 'O' : 'I', + t.tm_hour, t.tm_min, t.tm_sec, (long)(1000 * tb.millitm)) < 0) { + free(dump_buf); + return (NULL); + } +#endif +#else + gettimeofday(&tv, NULL); + sec = (time_t)tv.tv_sec; + localtime_r((const time_t *)&sec, &t); + if (snprintf(dump_buf, PREAMBLE_LENGTH + 1, PREAMBLE_FORMAT, + outbound ? 'O' : 'I', + t.tm_hour, t.tm_min, t.tm_sec, (long)tv.tv_usec) < 0) { + free(dump_buf); + return (NULL); + } +#endif + pos += PREAMBLE_LENGTH; +#if defined(_WIN32) && !defined(__MINGW32__) + strncpy_s(dump_buf + pos, strlen(HEADER) + 1, HEADER, strlen(HEADER)); +#else + strcpy(dump_buf + pos, HEADER); +#endif + pos += strlen(HEADER); + packet = (char *)buf; + for (i = 0; i < len; i++) { + uint8_t byte, low, high; + + byte = (uint8_t)packet[i]; + high = byte / 16; + low = byte % 16; + dump_buf[pos++] = high < 10 ? '0' + high : 'a' + (high - 10); + dump_buf[pos++] = low < 10 ? '0' + low : 'a' + (low - 10); + dump_buf[pos++] = ' '; + } +#if defined(_WIN32) && !defined(__MINGW32__) + strncpy_s(dump_buf + pos, strlen(TRAILER) + 1, TRAILER, strlen(TRAILER)); +#else + strcpy(dump_buf + pos, TRAILER); +#endif + pos += strlen(TRAILER); + dump_buf[pos++] = '\0'; + return (dump_buf); +} + +void +usrsctp_freedumpbuffer(char *buf) +{ + free(buf); +} + +void +usrsctp_enable_crc32c_offload(void) +{ + SCTP_BASE_VAR(crc32c_offloaded) = 1; +} + +void +usrsctp_disable_crc32c_offload(void) +{ + SCTP_BASE_VAR(crc32c_offloaded) = 0; +} + +/* Compute the CRC32C in network byte order */ +uint32_t +usrsctp_crc32c(void *buffer, size_t length) +{ + uint32_t base = 0xffffffff; + + base = calculate_crc32c(0xffffffff, (unsigned char *)buffer, (unsigned int) length); + base = sctp_finalize_crc32c(base); + return (base); +} + +void +usrsctp_conninput(void *addr, const void *buffer, size_t length, uint8_t ecn_bits) +{ + struct sockaddr_conn src, dst; + struct mbuf *m, *mm; + struct sctphdr *sh; + struct sctp_chunkhdr *ch; + int remaining, offset; + + SCTP_STAT_INCR(sctps_recvpackets); + SCTP_STAT_INCR_COUNTER64(sctps_inpackets); + memset(&src, 0, sizeof(struct sockaddr_conn)); + src.sconn_family = AF_CONN; +#ifdef HAVE_SCONN_LEN + src.sconn_len = sizeof(struct sockaddr_conn); +#endif + src.sconn_addr = addr; + memset(&dst, 0, sizeof(struct sockaddr_conn)); + dst.sconn_family = AF_CONN; +#ifdef HAVE_SCONN_LEN + dst.sconn_len = sizeof(struct sockaddr_conn); +#endif + dst.sconn_addr = addr; + if ((m = sctp_get_mbuf_for_msg((unsigned int)length, 1, M_NOWAIT, 0, MT_DATA)) == NULL) { + return; + } + /* Set the lengths fields of the mbuf chain. + * This is expected by m_copyback(). + */ + remaining = (int)length; + for (mm = m; mm != NULL; mm = mm->m_next) { + mm->m_len = min((int)M_SIZE(mm), remaining); + m->m_pkthdr.len += mm->m_len; + remaining -= mm->m_len; + } + KASSERT(remaining == 0, ("usrsctp_conninput: %zu bytes left", remaining)); + m_copyback(m, 0, (int)length, (caddr_t)buffer); + offset = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); + if (SCTP_BUF_LEN(m) < offset) { + if ((m = m_pullup(m, offset)) == NULL) { + SCTP_STAT_INCR(sctps_hdrops); + return; + } + } + sh = mtod(m, struct sctphdr *); + ch = (struct sctp_chunkhdr *)((caddr_t)sh + sizeof(struct sctphdr)); + offset -= sizeof(struct sctp_chunkhdr); + src.sconn_port = sh->src_port; + dst.sconn_port = sh->dest_port; + sctp_common_input_processing(&m, 0, offset, (int)length, + (struct sockaddr *)&src, + (struct sockaddr *)&dst, + sh, ch, + SCTP_BASE_VAR(crc32c_offloaded) == 1 ? 0 : 1, + ecn_bits, + SCTP_DEFAULT_VRFID, 0); + if (m) { + sctp_m_freem(m); + } + return; +} + +void usrsctp_handle_timers(uint32_t elapsed_milliseconds) +{ + sctp_handle_tick(sctp_msecs_to_ticks(elapsed_milliseconds)); +} + +int +usrsctp_get_events(struct socket *so) +{ + int events = 0; + + if (so == NULL) { + errno = EBADF; + return -1; + } + + SOCK_LOCK(so); + if (soreadable(so)) { + events |= SCTP_EVENT_READ; + } + if (sowriteable(so)) { + events |= SCTP_EVENT_WRITE; + } + if (so->so_error) { + events |= SCTP_EVENT_ERROR; + } + SOCK_UNLOCK(so); + + return events; +} + +int +usrsctp_set_upcall(struct socket *so, void (*upcall)(struct socket *, void *, int), void *arg) +{ + if (so == NULL) { + errno = EBADF; + return (-1); + } + + SOCK_LOCK(so); + so->so_upcall = upcall; + so->so_upcallarg = arg; + so->so_snd.sb_flags |= SB_UPCALL; + so->so_rcv.sb_flags |= SB_UPCALL; + SOCK_UNLOCK(so); + + return (0); +} + +#define USRSCTP_TUNABLE_SET_DEF(__field, __prefix) \ +int usrsctp_tunable_set_ ## __field(uint32_t value) \ +{ \ + if ((value < __prefix##_MIN) || \ + (value > __prefix##_MAX)) { \ + errno = EINVAL; \ + return (-1); \ + } else { \ + SCTP_BASE_SYSCTL(__field) = value; \ + return (0); \ + } \ +} + +USRSCTP_TUNABLE_SET_DEF(sctp_hashtblsize, SCTPCTL_TCBHASHSIZE) +USRSCTP_TUNABLE_SET_DEF(sctp_pcbtblsize, SCTPCTL_PCBHASHSIZE) +USRSCTP_TUNABLE_SET_DEF(sctp_chunkscale, SCTPCTL_CHUNKSCALE) + +#define USRSCTP_SYSCTL_SET_DEF(__field, __prefix) \ +int usrsctp_sysctl_set_ ## __field(uint32_t value) \ +{ \ + if ((value < __prefix##_MIN) || \ + (value > __prefix##_MAX)) { \ + errno = EINVAL; \ + return (-1); \ + } else { \ + SCTP_BASE_SYSCTL(__field) = value; \ + return (0); \ + } \ +} + +#if __GNUC__ >= 5 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6) || defined(__clang__) +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wtype-limits" +#endif +USRSCTP_SYSCTL_SET_DEF(sctp_sendspace, SCTPCTL_MAXDGRAM) +USRSCTP_SYSCTL_SET_DEF(sctp_recvspace, SCTPCTL_RECVSPACE) +USRSCTP_SYSCTL_SET_DEF(sctp_auto_asconf, SCTPCTL_AUTOASCONF) +USRSCTP_SYSCTL_SET_DEF(sctp_ecn_enable, SCTPCTL_ECN_ENABLE) +USRSCTP_SYSCTL_SET_DEF(sctp_pr_enable, SCTPCTL_PR_ENABLE) +USRSCTP_SYSCTL_SET_DEF(sctp_auth_enable, SCTPCTL_AUTH_ENABLE) +USRSCTP_SYSCTL_SET_DEF(sctp_asconf_enable, SCTPCTL_ASCONF_ENABLE) +USRSCTP_SYSCTL_SET_DEF(sctp_reconfig_enable, SCTPCTL_RECONFIG_ENABLE) +USRSCTP_SYSCTL_SET_DEF(sctp_nrsack_enable, SCTPCTL_NRSACK_ENABLE) +USRSCTP_SYSCTL_SET_DEF(sctp_pktdrop_enable, SCTPCTL_PKTDROP_ENABLE) +USRSCTP_SYSCTL_SET_DEF(sctp_no_csum_on_loopback, SCTPCTL_LOOPBACK_NOCSUM) +USRSCTP_SYSCTL_SET_DEF(sctp_peer_chunk_oh, SCTPCTL_PEER_CHKOH) +USRSCTP_SYSCTL_SET_DEF(sctp_max_burst_default, SCTPCTL_MAXBURST) +USRSCTP_SYSCTL_SET_DEF(sctp_max_chunks_on_queue, SCTPCTL_MAXCHUNKS) +USRSCTP_SYSCTL_SET_DEF(sctp_min_split_point, SCTPCTL_MIN_SPLIT_POINT) +USRSCTP_SYSCTL_SET_DEF(sctp_delayed_sack_time_default, SCTPCTL_DELAYED_SACK_TIME) +USRSCTP_SYSCTL_SET_DEF(sctp_sack_freq_default, SCTPCTL_SACK_FREQ) +USRSCTP_SYSCTL_SET_DEF(sctp_system_free_resc_limit, SCTPCTL_SYS_RESOURCE) +USRSCTP_SYSCTL_SET_DEF(sctp_asoc_free_resc_limit, SCTPCTL_ASOC_RESOURCE) +USRSCTP_SYSCTL_SET_DEF(sctp_heartbeat_interval_default, SCTPCTL_HEARTBEAT_INTERVAL) +USRSCTP_SYSCTL_SET_DEF(sctp_pmtu_raise_time_default, SCTPCTL_PMTU_RAISE_TIME) +USRSCTP_SYSCTL_SET_DEF(sctp_shutdown_guard_time_default, SCTPCTL_SHUTDOWN_GUARD_TIME) +USRSCTP_SYSCTL_SET_DEF(sctp_secret_lifetime_default, SCTPCTL_SECRET_LIFETIME) +USRSCTP_SYSCTL_SET_DEF(sctp_rto_max_default, SCTPCTL_RTO_MAX) +USRSCTP_SYSCTL_SET_DEF(sctp_rto_min_default, SCTPCTL_RTO_MIN) +USRSCTP_SYSCTL_SET_DEF(sctp_rto_initial_default, SCTPCTL_RTO_INITIAL) +USRSCTP_SYSCTL_SET_DEF(sctp_init_rto_max_default, SCTPCTL_INIT_RTO_MAX) +USRSCTP_SYSCTL_SET_DEF(sctp_valid_cookie_life_default, SCTPCTL_VALID_COOKIE_LIFE) +USRSCTP_SYSCTL_SET_DEF(sctp_init_rtx_max_default, SCTPCTL_INIT_RTX_MAX) +USRSCTP_SYSCTL_SET_DEF(sctp_assoc_rtx_max_default, SCTPCTL_ASSOC_RTX_MAX) +USRSCTP_SYSCTL_SET_DEF(sctp_path_rtx_max_default, SCTPCTL_PATH_RTX_MAX) +USRSCTP_SYSCTL_SET_DEF(sctp_add_more_threshold, SCTPCTL_ADD_MORE_ON_OUTPUT) +USRSCTP_SYSCTL_SET_DEF(sctp_nr_incoming_streams_default, SCTPCTL_INCOMING_STREAMS) +USRSCTP_SYSCTL_SET_DEF(sctp_nr_outgoing_streams_default, SCTPCTL_OUTGOING_STREAMS) +USRSCTP_SYSCTL_SET_DEF(sctp_cmt_on_off, SCTPCTL_CMT_ON_OFF) +USRSCTP_SYSCTL_SET_DEF(sctp_cmt_use_dac, SCTPCTL_CMT_USE_DAC) +USRSCTP_SYSCTL_SET_DEF(sctp_use_cwnd_based_maxburst, SCTPCTL_CWND_MAXBURST) +USRSCTP_SYSCTL_SET_DEF(sctp_nat_friendly, SCTPCTL_NAT_FRIENDLY) +USRSCTP_SYSCTL_SET_DEF(sctp_L2_abc_variable, SCTPCTL_ABC_L_VAR) +USRSCTP_SYSCTL_SET_DEF(sctp_mbuf_threshold_count, SCTPCTL_MAX_CHAINED_MBUFS) +USRSCTP_SYSCTL_SET_DEF(sctp_do_drain, SCTPCTL_DO_SCTP_DRAIN) +USRSCTP_SYSCTL_SET_DEF(sctp_hb_maxburst, SCTPCTL_HB_MAX_BURST) +USRSCTP_SYSCTL_SET_DEF(sctp_abort_if_one_2_one_hits_limit, SCTPCTL_ABORT_AT_LIMIT) +USRSCTP_SYSCTL_SET_DEF(sctp_min_residual, SCTPCTL_MIN_RESIDUAL) +USRSCTP_SYSCTL_SET_DEF(sctp_max_retran_chunk, SCTPCTL_MAX_RETRAN_CHUNK) +USRSCTP_SYSCTL_SET_DEF(sctp_logging_level, SCTPCTL_LOGGING_LEVEL) +USRSCTP_SYSCTL_SET_DEF(sctp_default_cc_module, SCTPCTL_DEFAULT_CC_MODULE) +USRSCTP_SYSCTL_SET_DEF(sctp_default_frag_interleave, SCTPCTL_DEFAULT_FRAG_INTERLEAVE) +USRSCTP_SYSCTL_SET_DEF(sctp_mobility_base, SCTPCTL_MOBILITY_BASE) +USRSCTP_SYSCTL_SET_DEF(sctp_mobility_fasthandoff, SCTPCTL_MOBILITY_FASTHANDOFF) +USRSCTP_SYSCTL_SET_DEF(sctp_inits_include_nat_friendly, SCTPCTL_NAT_FRIENDLY_INITS) +USRSCTP_SYSCTL_SET_DEF(sctp_udp_tunneling_port, SCTPCTL_UDP_TUNNELING_PORT) +USRSCTP_SYSCTL_SET_DEF(sctp_enable_sack_immediately, SCTPCTL_SACK_IMMEDIATELY_ENABLE) +USRSCTP_SYSCTL_SET_DEF(sctp_vtag_time_wait, SCTPCTL_TIME_WAIT) +USRSCTP_SYSCTL_SET_DEF(sctp_blackhole, SCTPCTL_BLACKHOLE) +USRSCTP_SYSCTL_SET_DEF(sctp_diag_info_code, SCTPCTL_DIAG_INFO_CODE) +USRSCTP_SYSCTL_SET_DEF(sctp_fr_max_burst_default, SCTPCTL_FRMAXBURST) +USRSCTP_SYSCTL_SET_DEF(sctp_path_pf_threshold, SCTPCTL_PATH_PF_THRESHOLD) +USRSCTP_SYSCTL_SET_DEF(sctp_default_ss_module, SCTPCTL_DEFAULT_SS_MODULE) +USRSCTP_SYSCTL_SET_DEF(sctp_rttvar_bw, SCTPCTL_RTTVAR_BW) +USRSCTP_SYSCTL_SET_DEF(sctp_rttvar_rtt, SCTPCTL_RTTVAR_RTT) +USRSCTP_SYSCTL_SET_DEF(sctp_rttvar_eqret, SCTPCTL_RTTVAR_EQRET) +USRSCTP_SYSCTL_SET_DEF(sctp_steady_step, SCTPCTL_RTTVAR_STEADYS) +USRSCTP_SYSCTL_SET_DEF(sctp_use_dccc_ecn, SCTPCTL_RTTVAR_DCCCECN) +USRSCTP_SYSCTL_SET_DEF(sctp_buffer_splitting, SCTPCTL_BUFFER_SPLITTING) +USRSCTP_SYSCTL_SET_DEF(sctp_initial_cwnd, SCTPCTL_INITIAL_CWND) +#ifdef SCTP_DEBUG +USRSCTP_SYSCTL_SET_DEF(sctp_debug_on, SCTPCTL_DEBUG) +#endif +#if __GNUC__ >= 5 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6) || defined(__clang__) +#pragma GCC diagnostic pop +#endif + +#define USRSCTP_SYSCTL_GET_DEF(__field) \ +uint32_t usrsctp_sysctl_get_ ## __field(void) { \ + return SCTP_BASE_SYSCTL(__field); \ +} + +USRSCTP_SYSCTL_GET_DEF(sctp_sendspace) +USRSCTP_SYSCTL_GET_DEF(sctp_recvspace) +USRSCTP_SYSCTL_GET_DEF(sctp_auto_asconf) +USRSCTP_SYSCTL_GET_DEF(sctp_multiple_asconfs) +USRSCTP_SYSCTL_GET_DEF(sctp_ecn_enable) +USRSCTP_SYSCTL_GET_DEF(sctp_pr_enable) +USRSCTP_SYSCTL_GET_DEF(sctp_auth_enable) +USRSCTP_SYSCTL_GET_DEF(sctp_asconf_enable) +USRSCTP_SYSCTL_GET_DEF(sctp_reconfig_enable) +USRSCTP_SYSCTL_GET_DEF(sctp_nrsack_enable) +USRSCTP_SYSCTL_GET_DEF(sctp_pktdrop_enable) +USRSCTP_SYSCTL_GET_DEF(sctp_no_csum_on_loopback) +USRSCTP_SYSCTL_GET_DEF(sctp_peer_chunk_oh) +USRSCTP_SYSCTL_GET_DEF(sctp_max_burst_default) +USRSCTP_SYSCTL_GET_DEF(sctp_max_chunks_on_queue) +USRSCTP_SYSCTL_GET_DEF(sctp_hashtblsize) +USRSCTP_SYSCTL_GET_DEF(sctp_pcbtblsize) +USRSCTP_SYSCTL_GET_DEF(sctp_min_split_point) +USRSCTP_SYSCTL_GET_DEF(sctp_chunkscale) +USRSCTP_SYSCTL_GET_DEF(sctp_delayed_sack_time_default) +USRSCTP_SYSCTL_GET_DEF(sctp_sack_freq_default) +USRSCTP_SYSCTL_GET_DEF(sctp_system_free_resc_limit) +USRSCTP_SYSCTL_GET_DEF(sctp_asoc_free_resc_limit) +USRSCTP_SYSCTL_GET_DEF(sctp_heartbeat_interval_default) +USRSCTP_SYSCTL_GET_DEF(sctp_pmtu_raise_time_default) +USRSCTP_SYSCTL_GET_DEF(sctp_shutdown_guard_time_default) +USRSCTP_SYSCTL_GET_DEF(sctp_secret_lifetime_default) +USRSCTP_SYSCTL_GET_DEF(sctp_rto_max_default) +USRSCTP_SYSCTL_GET_DEF(sctp_rto_min_default) +USRSCTP_SYSCTL_GET_DEF(sctp_rto_initial_default) +USRSCTP_SYSCTL_GET_DEF(sctp_init_rto_max_default) +USRSCTP_SYSCTL_GET_DEF(sctp_valid_cookie_life_default) +USRSCTP_SYSCTL_GET_DEF(sctp_init_rtx_max_default) +USRSCTP_SYSCTL_GET_DEF(sctp_assoc_rtx_max_default) +USRSCTP_SYSCTL_GET_DEF(sctp_path_rtx_max_default) +USRSCTP_SYSCTL_GET_DEF(sctp_add_more_threshold) +USRSCTP_SYSCTL_GET_DEF(sctp_nr_incoming_streams_default) +USRSCTP_SYSCTL_GET_DEF(sctp_nr_outgoing_streams_default) +USRSCTP_SYSCTL_GET_DEF(sctp_cmt_on_off) +USRSCTP_SYSCTL_GET_DEF(sctp_cmt_use_dac) +USRSCTP_SYSCTL_GET_DEF(sctp_use_cwnd_based_maxburst) +USRSCTP_SYSCTL_GET_DEF(sctp_nat_friendly) +USRSCTP_SYSCTL_GET_DEF(sctp_L2_abc_variable) +USRSCTP_SYSCTL_GET_DEF(sctp_mbuf_threshold_count) +USRSCTP_SYSCTL_GET_DEF(sctp_do_drain) +USRSCTP_SYSCTL_GET_DEF(sctp_hb_maxburst) +USRSCTP_SYSCTL_GET_DEF(sctp_abort_if_one_2_one_hits_limit) +USRSCTP_SYSCTL_GET_DEF(sctp_min_residual) +USRSCTP_SYSCTL_GET_DEF(sctp_max_retran_chunk) +USRSCTP_SYSCTL_GET_DEF(sctp_logging_level) +USRSCTP_SYSCTL_GET_DEF(sctp_default_cc_module) +USRSCTP_SYSCTL_GET_DEF(sctp_default_frag_interleave) +USRSCTP_SYSCTL_GET_DEF(sctp_mobility_base) +USRSCTP_SYSCTL_GET_DEF(sctp_mobility_fasthandoff) +USRSCTP_SYSCTL_GET_DEF(sctp_inits_include_nat_friendly) +USRSCTP_SYSCTL_GET_DEF(sctp_udp_tunneling_port) +USRSCTP_SYSCTL_GET_DEF(sctp_enable_sack_immediately) +USRSCTP_SYSCTL_GET_DEF(sctp_vtag_time_wait) +USRSCTP_SYSCTL_GET_DEF(sctp_blackhole) +USRSCTP_SYSCTL_GET_DEF(sctp_diag_info_code) +USRSCTP_SYSCTL_GET_DEF(sctp_fr_max_burst_default) +USRSCTP_SYSCTL_GET_DEF(sctp_path_pf_threshold) +USRSCTP_SYSCTL_GET_DEF(sctp_default_ss_module) +USRSCTP_SYSCTL_GET_DEF(sctp_rttvar_bw) +USRSCTP_SYSCTL_GET_DEF(sctp_rttvar_rtt) +USRSCTP_SYSCTL_GET_DEF(sctp_rttvar_eqret) +USRSCTP_SYSCTL_GET_DEF(sctp_steady_step) +USRSCTP_SYSCTL_GET_DEF(sctp_use_dccc_ecn) +USRSCTP_SYSCTL_GET_DEF(sctp_buffer_splitting) +USRSCTP_SYSCTL_GET_DEF(sctp_initial_cwnd) +#ifdef SCTP_DEBUG +USRSCTP_SYSCTL_GET_DEF(sctp_debug_on) +#endif + +void usrsctp_get_stat(struct sctpstat *stat) +{ + *stat = SCTP_BASE_STATS; +} diff --git a/netwerk/sctp/src/user_socketvar.h b/netwerk/sctp/src/user_socketvar.h new file mode 100644 index 0000000000..e8eccc7fb1 --- /dev/null +++ b/netwerk/sctp/src/user_socketvar.h @@ -0,0 +1,527 @@ +/*- + * Copyright (c) 1982, 1986, 1990, 1993 + * The Regents of the University of California. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + */ + +/* __Userspace__ version of <sys/socketvar.h> goes here.*/ + +#ifndef _USER_SOCKETVAR_H_ +#define _USER_SOCKETVAR_H_ + +#if defined(__APPLE__) +#include <sys/types.h> +#include <unistd.h> +#endif + +/* #include <sys/selinfo.h> */ /*__Userspace__ alternative?*/ /* for struct selinfo */ +/* #include <sys/_lock.h> was 0 byte file */ +/* #include <sys/_mutex.h> was 0 byte file */ +/* #include <sys/_sx.h> */ /*__Userspace__ alternative?*/ +#if !defined(__DragonFly__) && !defined(__FreeBSD__) && !defined(__NetBSD__) && !defined(_WIN32) && !defined(__native_client__) +#include <sys/uio.h> +#endif +#define SOCK_MAXADDRLEN 255 +#if !defined(MSG_NOTIFICATION) +#define MSG_NOTIFICATION 0x2000 /* SCTP notification */ +#endif +#define SCTP_SO_LINGER 0x0001 +#define SCTP_SO_ACCEPTCONN 0x0002 +#define SS_CANTRCVMORE 0x020 +#define SS_CANTSENDMORE 0x010 + +#if defined(__APPLE__) || defined(__DragonFly__) || defined(__FreeBSD__) || defined(__OpenBSD__) || defined(_WIN32) || defined(__native_client__) +#define UIO_MAXIOV 1024 +#define ERESTART (-1) +#endif + +#if !defined(__APPLE__) && !defined(__NetBSD__) && !defined(__OpenBSD__) +enum uio_rw { UIO_READ, UIO_WRITE }; +#endif + +#if !defined(__NetBSD__) && !defined(__OpenBSD__) +/* Segment flag values. */ +enum uio_seg { + UIO_USERSPACE, /* from user data space */ + UIO_SYSSPACE /* from system space */ +}; +#endif + +struct proc { + int stub; /* struct proc is a dummy for __Userspace__ */ +}; + +MALLOC_DECLARE(M_ACCF); +MALLOC_DECLARE(M_PCB); +MALLOC_DECLARE(M_SONAME); + +/* __Userspace__ Are these all the fields we need? + * Removing struct thread *uio_td; owner field +*/ +struct uio { + struct iovec *uio_iov; /* scatter/gather list */ + int uio_iovcnt; /* length of scatter/gather list */ + off_t uio_offset; /* offset in target object */ + ssize_t uio_resid; /* remaining bytes to process */ + enum uio_seg uio_segflg; /* address space */ + enum uio_rw uio_rw; /* operation */ +}; + + +/* __Userspace__ */ + +/* + * Kernel structure per socket. + * Contains send and receive buffer queues, + * handle on protocol and pointer to protocol + * private data and error information. + */ +#if defined(_WIN32) +#define AF_ROUTE 17 +#if !defined(__MINGW32__) +typedef __int32 pid_t; +#endif +typedef unsigned __int32 uid_t; +enum sigType { + SIGNAL = 0, + BROADCAST = 1, + MAX_EVENTS = 2 +}; +#endif + +/*- + * Locking key to struct socket: + * (a) constant after allocation, no locking required. + * (b) locked by SOCK_LOCK(so). + * (c) locked by SOCKBUF_LOCK(&so->so_rcv). + * (d) locked by SOCKBUF_LOCK(&so->so_snd). + * (e) locked by ACCEPT_LOCK(). + * (f) not locked since integer reads/writes are atomic. + * (g) used only as a sleep/wakeup address, no value. + * (h) locked by global mutex so_global_mtx. + */ +struct socket { + int so_count; /* (b) reference count */ + short so_type; /* (a) generic type, see socket.h */ + short so_options; /* from socket call, see socket.h */ + short so_linger; /* time to linger while closing */ + short so_state; /* (b) internal state flags SS_* */ + int so_qstate; /* (e) internal state flags SQ_* */ + void *so_pcb; /* protocol control block */ + int so_dom; +/* + * Variables for connection queuing. + * Socket where accepts occur is so_head in all subsidiary sockets. + * If so_head is 0, socket is not related to an accept. + * For head socket so_incomp queues partially completed connections, + * while so_comp is a queue of connections ready to be accepted. + * If a connection is aborted and it has so_head set, then + * it has to be pulled out of either so_incomp or so_comp. + * We allow connections to queue up based on current queue lengths + * and limit on number of queued connections for this socket. + */ + struct socket *so_head; /* (e) back pointer to listen socket */ + TAILQ_HEAD(, socket) so_incomp; /* (e) queue of partial unaccepted connections */ + TAILQ_HEAD(, socket) so_comp; /* (e) queue of complete unaccepted connections */ + TAILQ_ENTRY(socket) so_list; /* (e) list of unaccepted connections */ + u_short so_qlen; /* (e) number of unaccepted connections */ + u_short so_incqlen; /* (e) number of unaccepted incomplete + connections */ + u_short so_qlimit; /* (e) max number queued connections */ + short so_timeo; /* (g) connection timeout */ + userland_cond_t timeo_cond; /* timeo_cond condition variable being used in wakeup */ + + u_short so_error; /* (f) error affecting connection */ + struct sigio *so_sigio; /* [sg] information for async I/O or + out of band data (SIGURG) */ + u_long so_oobmark; /* (c) chars to oob mark */ + TAILQ_HEAD(, aiocblist) so_aiojobq; /* AIO ops waiting on socket */ +/* + * Variables for socket buffering. + */ + struct sockbuf { + /* __Userspace__ Many of these fields may + * not be required for the sctp stack. + * Commenting out the following. + * Including pthread mutex and condition variable to be + * used by sbwait, sorwakeup and sowwakeup. + */ + /* struct selinfo sb_sel;*/ /* process selecting read/write */ + /* struct mtx sb_mtx;*/ /* sockbuf lock */ + /* struct sx sb_sx;*/ /* prevent I/O interlacing */ + userland_cond_t sb_cond; /* sockbuf condition variable */ + userland_mutex_t sb_mtx; /* sockbuf lock associated with sb_cond */ + short sb_state; /* (c/d) socket state on sockbuf */ +#define sb_startzero sb_mb + struct mbuf *sb_mb; /* (c/d) the mbuf chain */ + struct mbuf *sb_mbtail; /* (c/d) the last mbuf in the chain */ + struct mbuf *sb_lastrecord; /* (c/d) first mbuf of last + * record in socket buffer */ + struct mbuf *sb_sndptr; /* (c/d) pointer into mbuf chain */ + u_int sb_sndptroff; /* (c/d) byte offset of ptr into chain */ + u_int sb_cc; /* (c/d) actual chars in buffer */ + u_int sb_hiwat; /* (c/d) max actual char count */ + u_int sb_mbcnt; /* (c/d) chars of mbufs used */ + u_int sb_mbmax; /* (c/d) max chars of mbufs to use */ + u_int sb_ctl; /* (c/d) non-data chars in buffer */ + int sb_lowat; /* (c/d) low water mark */ + int sb_timeo; /* (c/d) timeout for read/write */ + short sb_flags; /* (c/d) flags, see below */ + } so_rcv, so_snd; +/* + * Constants for sb_flags field of struct sockbuf. + */ +#define SB_MAX (256*1024) /* default for max chars in sockbuf */ +#define SB_RAW (64*1024*2) /*Aligning so->so_rcv.sb_hiwat with the receive buffer size of raw socket*/ +/* + * Constants for sb_flags field of struct sockbuf. + */ +#define SB_WAIT 0x04 /* someone is waiting for data/space */ +#define SB_SEL 0x08 /* someone is selecting */ +#define SB_ASYNC 0x10 /* ASYNC I/O, need signals */ +#define SB_UPCALL 0x20 /* someone wants an upcall */ +#define SB_NOINTR 0x40 /* operations not interruptible */ +#define SB_AIO 0x80 /* AIO operations queued */ +#define SB_KNOTE 0x100 /* kernel note attached */ +#define SB_AUTOSIZE 0x800 /* automatically size socket buffer */ + + void (*so_upcall)(struct socket *, void *, int); + void *so_upcallarg; + struct ucred *so_cred; /* (a) user credentials */ + struct label *so_label; /* (b) MAC label for socket */ + struct label *so_peerlabel; /* (b) cached MAC label for peer */ + /* NB: generation count must not be first. */ + uint32_t so_gencnt; /* (h) generation count */ + void *so_emuldata; /* (b) private data for emulators */ + struct so_accf { + struct accept_filter *so_accept_filter; + void *so_accept_filter_arg; /* saved filter args */ + char *so_accept_filter_str; /* saved user args */ + } *so_accf; +}; + +#define SB_EMPTY_FIXUP(sb) do { \ + if ((sb)->sb_mb == NULL) { \ + (sb)->sb_mbtail = NULL; \ + (sb)->sb_lastrecord = NULL; \ + } \ +} while (/*CONSTCOND*/0) + +/* + * Global accept mutex to serialize access to accept queues and + * fields associated with multiple sockets. This allows us to + * avoid defining a lock order between listen and accept sockets + * until such time as it proves to be a good idea. + */ +#if defined(_WIN32) +extern userland_mutex_t accept_mtx; +extern userland_cond_t accept_cond; +#define ACCEPT_LOCK_ASSERT() +#define ACCEPT_LOCK() do { \ + EnterCriticalSection(&accept_mtx); \ +} while (0) +#define ACCEPT_UNLOCK() do { \ + LeaveCriticalSection(&accept_mtx); \ +} while (0) +#define ACCEPT_UNLOCK_ASSERT() +#else +extern userland_mutex_t accept_mtx; + +extern userland_cond_t accept_cond; +#ifdef INVARIANTS +#define ACCEPT_LOCK() KASSERT(pthread_mutex_lock(&accept_mtx) == 0, ("%s: accept_mtx already locked", __func__)) +#define ACCEPT_UNLOCK() KASSERT(pthread_mutex_unlock(&accept_mtx) == 0, ("%s: accept_mtx not locked", __func__)) +#else +#define ACCEPT_LOCK() (void)pthread_mutex_lock(&accept_mtx) +#define ACCEPT_UNLOCK() (void)pthread_mutex_unlock(&accept_mtx) +#endif +#define ACCEPT_LOCK_ASSERT() \ + KASSERT(pthread_mutex_trylock(&accept_mtx) == EBUSY, ("%s: accept_mtx not locked", __func__)) +#define ACCEPT_UNLOCK_ASSERT() do { \ + KASSERT(pthread_mutex_trylock(&accept_mtx) == 0, ("%s: accept_mtx locked", __func__)); \ + (void)pthread_mutex_unlock(&accept_mtx); \ + } while (0) +#endif + +/* + * Per-socket buffer mutex used to protect most fields in the socket + * buffer. + */ +#define SOCKBUF_MTX(_sb) (&(_sb)->sb_mtx) +#if defined(_WIN32) +#define SOCKBUF_LOCK_INIT(_sb, _name) \ + InitializeCriticalSection(SOCKBUF_MTX(_sb)) +#define SOCKBUF_LOCK_DESTROY(_sb) DeleteCriticalSection(SOCKBUF_MTX(_sb)) +#define SOCKBUF_COND_INIT(_sb) InitializeConditionVariable((&(_sb)->sb_cond)) +#define SOCKBUF_COND_DESTROY(_sb) DeleteConditionVariable((&(_sb)->sb_cond)) +#define SOCK_COND_INIT(_so) InitializeConditionVariable((&(_so)->timeo_cond)) +#define SOCK_COND_DESTROY(_so) DeleteConditionVariable((&(_so)->timeo_cond)) +#define SOCK_COND(_so) (&(_so)->timeo_cond) +#else +#ifdef INVARIANTS +#define SOCKBUF_LOCK_INIT(_sb, _name) do { \ + pthread_mutexattr_t mutex_attr; \ + \ + pthread_mutexattr_init(&mutex_attr); \ + pthread_mutexattr_settype(&mutex_attr, PTHREAD_MUTEX_ERRORCHECK); \ + pthread_mutex_init(SOCKBUF_MTX(_sb), &mutex_attr); \ + pthread_mutexattr_destroy(&mutex_attr); \ +} while (0) +#else +#define SOCKBUF_LOCK_INIT(_sb, _name) \ + pthread_mutex_init(SOCKBUF_MTX(_sb), NULL) +#endif +#define SOCKBUF_LOCK_DESTROY(_sb) pthread_mutex_destroy(SOCKBUF_MTX(_sb)) +#define SOCKBUF_COND_INIT(_sb) pthread_cond_init((&(_sb)->sb_cond), NULL) +#define SOCKBUF_COND_DESTROY(_sb) pthread_cond_destroy((&(_sb)->sb_cond)) +#define SOCK_COND_INIT(_so) pthread_cond_init((&(_so)->timeo_cond), NULL) +#define SOCK_COND_DESTROY(_so) pthread_cond_destroy((&(_so)->timeo_cond)) +#define SOCK_COND(_so) (&(_so)->timeo_cond) +#endif +/*__Userspace__ SOCKBUF_LOCK(_sb) is now defined in netinet/sctp_process_lock.h */ + +/* #define SOCKBUF_OWNED(_sb) mtx_owned(SOCKBUF_MTX(_sb)) unused */ +/*__Userspace__ SOCKBUF_UNLOCK(_sb) is now defined in netinet/sctp_process_lock.h */ + +/*__Userspace__ SOCKBUF_LOCK_ASSERT(_sb) is now defined in netinet/sctp_process_lock.h */ + +/* #define SOCKBUF_UNLOCK_ASSERT(_sb) mtx_assert(SOCKBUF_MTX(_sb), MA_NOTOWNED) unused */ + +/* + * Per-socket mutex: we reuse the receive socket buffer mutex for space + * efficiency. This decision should probably be revisited as we optimize + * locking for the socket code. + */ +#define SOCK_MTX(_so) SOCKBUF_MTX(&(_so)->so_rcv) +/*__Userspace__ SOCK_LOCK(_so) is now defined in netinet/sctp_process_lock.h */ + +/* #define SOCK_OWNED(_so) SOCKBUF_OWNED(&(_so)->so_rcv) unused */ +/*__Userspace__ SOCK_UNLOCK(_so) is now defined in netinet/sctp_process_lock.h */ + +#define SOCK_LOCK_ASSERT(_so) SOCKBUF_LOCK_ASSERT(&(_so)->so_rcv) + +/* + * Socket state bits. + * + * Historically, this bits were all kept in the so_state field. For + * locking reasons, they are now in multiple fields, as they are + * locked differently. so_state maintains basic socket state protected + * by the socket lock. so_qstate holds information about the socket + * accept queues. Each socket buffer also has a state field holding + * information relevant to that socket buffer (can't send, rcv). Many + * fields will be read without locks to improve performance and avoid + * lock order issues. However, this approach must be used with caution. + */ +#define SS_NOFDREF 0x0001 /* no file table ref any more */ +#define SS_ISCONNECTED 0x0002 /* socket connected to a peer */ +#define SS_ISCONNECTING 0x0004 /* in process of connecting to peer */ +#define SS_ISDISCONNECTING 0x0008 /* in process of disconnecting */ +#define SS_NBIO 0x0100 /* non-blocking ops */ +#define SS_ASYNC 0x0200 /* async i/o notify */ +#define SS_ISCONFIRMING 0x0400 /* deciding to accept connection req */ +#define SS_ISDISCONNECTED 0x2000 /* socket disconnected from peer */ +/* + * Protocols can mark a socket as SS_PROTOREF to indicate that, following + * pru_detach, they still want the socket to persist, and will free it + * themselves when they are done. Protocols should only ever call sofree() + * following setting this flag in pru_detach(), and never otherwise, as + * sofree() bypasses socket reference counting. + */ +#define SS_PROTOREF 0x4000 /* strong protocol reference */ + +/* + * Socket state bits now stored in the socket buffer state field. + */ +#define SBS_CANTSENDMORE 0x0010 /* can't send more data to peer */ +#define SBS_CANTRCVMORE 0x0020 /* can't receive more data from peer */ +#define SBS_RCVATMARK 0x0040 /* at mark on input */ + +/* + * Socket state bits stored in so_qstate. + */ +#define SQ_INCOMP 0x0800 /* unaccepted, incomplete connection */ +#define SQ_COMP 0x1000 /* unaccepted, complete connection */ + +/* + * Socket event flags + */ +#define SCTP_EVENT_READ 0x0001 /* socket is readable */ +#define SCTP_EVENT_WRITE 0x0002 /* socket is writeable */ +#define SCTP_EVENT_ERROR 0x0004 /* socket has an error state */ + + +/*-------------------------------------------------------------*/ +/*-------------------------------------------------------------*/ +/* __Userspace__ */ +/*-------------------------------------------------------------*/ +/*-------------------------------------------------------------*/ +/* this new __Userspace__ section is to copy portions of the _KERNEL block + * above into, avoiding having to port the entire thing at once... + * For function prototypes, the full bodies are in user_socket.c . + */ + +/* ---------------------------------------------------------- */ +/* --- function prototypes (implemented in user_socket.c) --- */ +/* ---------------------------------------------------------- */ +void soisconnecting(struct socket *so); +void soisdisconnecting(struct socket *so); +void soisconnected(struct socket *so); +struct socket * sonewconn(struct socket *head, int connstatus); +void socantrcvmore(struct socket *so); +void socantsendmore(struct socket *so); +void sofree(struct socket *so); + + + +/* -------------- */ +/* --- macros --- */ +/* -------------- */ + +#define soref(so) do { \ + SOCK_LOCK_ASSERT(so); \ + ++(so)->so_count; \ +} while (0) + +#define sorele(so) do { \ + ACCEPT_LOCK_ASSERT(); \ + SOCK_LOCK_ASSERT(so); \ + KASSERT((so)->so_count > 0, ("sorele")); \ + if (--(so)->so_count == 0) \ + sofree(so); \ + else { \ + SOCK_UNLOCK(so); \ + ACCEPT_UNLOCK(); \ + } \ +} while (0) + + +/* replacing imin with min (user_environment.h) */ +#define sbspace(sb) \ + ((long) min((int)((sb)->sb_hiwat - (sb)->sb_cc), \ + (int)((sb)->sb_mbmax - (sb)->sb_mbcnt))) + +/* do we have to send all at once on a socket? */ +#define sosendallatonce(so) \ + ((so)->so_proto->pr_flags & PR_ATOMIC) + +/* can we read something from so? */ +#define soreadable(so) \ + ((int)((so)->so_rcv.sb_cc) >= (so)->so_rcv.sb_lowat || \ + ((so)->so_rcv.sb_state & SBS_CANTRCVMORE) || \ + !TAILQ_EMPTY(&(so)->so_comp) || (so)->so_error) + +#if 0 /* original */ +#define PR_CONNREQUIRED 0x04 /* from sys/protosw.h "needed" for sowriteable */ +#define sowriteable(so) \ + ((sbspace(&(so)->so_snd) >= (so)->so_snd.sb_lowat && \ + (((so)->so_state&SS_ISCONNECTED) || \ + ((so)->so_proto->pr_flags&PR_CONNREQUIRED)==0)) || \ + ((so)->so_snd.sb_state & SBS_CANTSENDMORE) || \ + (so)->so_error) +#else /* line with PR_CONNREQUIRED removed */ +/* can we write something to so? */ +#define sowriteable(so) \ + ((sbspace(&(so)->so_snd) >= (so)->so_snd.sb_lowat && \ + (((so)->so_state&SS_ISCONNECTED))) || \ + ((so)->so_snd.sb_state & SBS_CANTSENDMORE) || \ + (so)->so_error) +#endif + +extern void solisten_proto(struct socket *so, int backlog); +extern int solisten_proto_check(struct socket *so); +extern int sctp_listen(struct socket *so, int backlog, struct proc *p); +extern void socantrcvmore_locked(struct socket *so); +extern int sctp_bind(struct socket *so, struct sockaddr *addr); +extern int sctp6_bind(struct socket *so, struct sockaddr *addr, void *proc); +extern int sctpconn_bind(struct socket *so, struct sockaddr *addr); +extern int sctp_accept(struct socket *so, struct sockaddr **addr); +extern int sctp_attach(struct socket *so, int proto, uint32_t vrf_id); +extern int sctp6_attach(struct socket *so, int proto, uint32_t vrf_id); +extern int sctp_abort(struct socket *so); +extern int sctp6_abort(struct socket *so); +extern void sctp_close(struct socket *so); +extern int soaccept(struct socket *so, struct sockaddr **nam); +extern int solisten(struct socket *so, int backlog); +extern int soreserve(struct socket *so, u_long sndcc, u_long rcvcc); +extern void sowakeup(struct socket *so, struct sockbuf *sb); +extern void wakeup(void *ident, struct socket *so); /*__Userspace__ */ +extern int uiomove(void *cp, int n, struct uio *uio); +extern int sbwait(struct sockbuf *sb); +extern int sodisconnect(struct socket *so); +extern int soconnect(struct socket *so, struct sockaddr *nam); +extern int sctp_disconnect(struct socket *so); +extern int sctp_connect(struct socket *so, struct sockaddr *addr); +extern int sctp6_connect(struct socket *so, struct sockaddr *addr); +extern int sctpconn_connect(struct socket *so, struct sockaddr *addr); +extern void sctp_finish(void); + +/* ------------------------------------------------ */ +/* ----- macros copied from above ---- */ +/* ------------------------------------------------ */ + +/* + * Do we need to notify the other side when I/O is possible? + */ +#define sb_notify(sb) (((sb)->sb_flags & (SB_WAIT | SB_SEL | SB_ASYNC | \ + SB_UPCALL | SB_AIO | SB_KNOTE)) != 0) + + +/* + * In sorwakeup() and sowwakeup(), acquire the socket buffer lock to + * avoid a non-atomic test-and-wakeup. However, sowakeup is + * responsible for releasing the lock if it is called. We unlock only + * if we don't call into sowakeup. If any code is introduced that + * directly invokes the underlying sowakeup() primitives, it must + * maintain the same semantics. + */ +#define sorwakeup_locked(so) do { \ + SOCKBUF_LOCK_ASSERT(&(so)->so_rcv); \ + if (sb_notify(&(so)->so_rcv)) \ + sowakeup((so), &(so)->so_rcv); \ + else \ + SOCKBUF_UNLOCK(&(so)->so_rcv); \ +} while (0) + +#define sorwakeup(so) do { \ + SOCKBUF_LOCK(&(so)->so_rcv); \ + sorwakeup_locked(so); \ +} while (0) + +#define sowwakeup_locked(so) do { \ + SOCKBUF_LOCK_ASSERT(&(so)->so_snd); \ + if (sb_notify(&(so)->so_snd)) \ + sowakeup((so), &(so)->so_snd); \ + else \ + SOCKBUF_UNLOCK(&(so)->so_snd); \ +} while (0) + +#define sowwakeup(so) do { \ + SOCKBUF_LOCK(&(so)->so_snd); \ + sowwakeup_locked(so); \ +} while (0) + +#endif /* !_SYS_SOCKETVAR_H_ */ diff --git a/netwerk/sctp/src/user_uma.h b/netwerk/sctp/src/user_uma.h new file mode 100644 index 0000000000..59d71fff42 --- /dev/null +++ b/netwerk/sctp/src/user_uma.h @@ -0,0 +1,96 @@ +/*- + * Copyright (c) 2009-2010 Brad Penoff + * Copyright (c) 2009-2010 Humaira Kamal + * Copyright (c) 2011-2012 Irene Ruengeler + * Copyright (c) 2011-2012 Michael Tuexen + * + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +#ifndef _USER_UMA_H_ +#define _USER_UMA_H_ + +#define UMA_ZFLAG_FULL 0x40000000 /* Reached uz_maxpages */ +#define UMA_ALIGN_PTR (sizeof(void *) - 1) /* Alignment fit for ptr */ + +/* __Userspace__ All these definitions will change for +userspace Universal Memory Allocator (UMA). These are included +for reference purposes and to avoid compile errors for the time being. +*/ +typedef int (*uma_ctor)(void *mem, int size, void *arg, int flags); +typedef void (*uma_dtor)(void *mem, int size, void *arg); +typedef int (*uma_init)(void *mem, int size, int flags); +typedef void (*uma_fini)(void *mem, int size); +typedef struct uma_zone * uma_zone_t; +typedef struct uma_keg * uma_keg_t; + +struct uma_cache { + int stub; /* TODO __Userspace__ */ +}; + +struct uma_keg { + int stub; /* TODO __Userspace__ */ +}; + +struct uma_zone { + char *uz_name; /* Text name of the zone */ + struct mtx *uz_lock; /* Lock for the zone (keg's lock) */ + uma_keg_t uz_keg; /* Our underlying Keg */ + + LIST_ENTRY(uma_zone) uz_link; /* List of all zones in keg */ + LIST_HEAD(,uma_bucket) uz_full_bucket; /* full buckets */ + LIST_HEAD(,uma_bucket) uz_free_bucket; /* Buckets for frees */ + + uma_ctor uz_ctor; /* Constructor for each allocation */ + uma_dtor uz_dtor; /* Destructor */ + uma_init uz_init; /* Initializer for each item */ + uma_fini uz_fini; /* Discards memory */ + + u_int64_t uz_allocs; /* Total number of allocations */ + u_int64_t uz_frees; /* Total number of frees */ + u_int64_t uz_fails; /* Total number of alloc failures */ + uint16_t uz_fills; /* Outstanding bucket fills */ + uint16_t uz_count; /* Highest value ub_ptr can have */ + + /* + * This HAS to be the last item because we adjust the zone size + * based on NCPU and then allocate the space for the zones. + */ + struct uma_cache uz_cpu[1]; /* Per cpu caches */ +}; + +/* Prototype */ +uma_zone_t +uma_zcreate(char *name, size_t size, uma_ctor ctor, uma_dtor dtor, + uma_init uminit, uma_fini fini, int align, uint32_t flags); + + +#define uma_zone_set_max(zone, number) /* stub TODO __Userspace__ */ + +uma_zone_t +uma_zcreate(char *name, size_t size, uma_ctor ctor, uma_dtor dtor, + uma_init uminit, uma_fini fini, int align, uint32_t flags) { + return NULL; /* stub TODO __Userspace__. Also place implementation in a separate .c file */ +} +#endif diff --git a/netwerk/sctp/src/usrsctp.h b/netwerk/sctp/src/usrsctp.h new file mode 100644 index 0000000000..4c135a171d --- /dev/null +++ b/netwerk/sctp/src/usrsctp.h @@ -0,0 +1,1324 @@ +/*- + * Copyright (c) 2009-2010 Brad Penoff + * Copyright (c) 2009-2010 Humaira Kamal + * Copyright (c) 2011-2012 Irene Ruengeler + * Copyright (c) 2011-2012 Michael Tuexen + * + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +#ifndef __USRSCTP_H__ +#define __USRSCTP_H__ + +#ifdef __cplusplus +extern "C" { +#endif + +#include <errno.h> +#include <sys/types.h> +#ifdef _WIN32 +#ifdef _MSC_VER +#pragma warning(disable: 4200) +#endif +#include <winsock2.h> +#include <ws2tcpip.h> +#else +#include <sys/socket.h> +#include <netinet/in.h> +#endif + +#ifndef MSG_NOTIFICATION +/* This definition MUST be in sync with usrsctplib/user_socketvar.h */ +#define MSG_NOTIFICATION 0x2000 +#endif + +#ifndef IPPROTO_SCTP +/* This is the IANA assigned protocol number of SCTP. */ +#define IPPROTO_SCTP 132 +#endif + +#ifdef _WIN32 +#if defined(_MSC_VER) && _MSC_VER >= 1600 +#include <stdint.h> +#elif defined(SCTP_STDINT_INCLUDE) +#include SCTP_STDINT_INCLUDE +#else +typedef unsigned __int8 uint8_t; +typedef unsigned __int16 uint16_t; +typedef unsigned __int32 uint32_t; +typedef unsigned __int64 uint64_t; +typedef __int16 int16_t; +typedef __int32 int32_t; +#endif + +#ifndef ssize_t +#ifdef _WIN64 +typedef __int64 ssize_t; +#elif defined _WIN32 +typedef int ssize_t; +#else +#error "Unknown platform!" +#endif +#endif + +#define MSG_EOR 0x8 +#ifndef EWOULDBLOCK +#define EWOULDBLOCK WSAEWOULDBLOCK +#endif +#ifndef EINPROGRESS +#define EINPROGRESS WSAEINPROGRESS +#endif +#define SHUT_RD 1 +#define SHUT_WR 2 +#define SHUT_RDWR 3 +#endif + +typedef uint32_t sctp_assoc_t; + +#if defined(_WIN32) && defined(_MSC_VER) +#pragma pack (push, 1) +#define SCTP_PACKED +#else +#define SCTP_PACKED __attribute__((packed)) +#endif + +struct sctp_common_header { + uint16_t source_port; + uint16_t destination_port; + uint32_t verification_tag; + uint32_t crc32c; +} SCTP_PACKED; + +#if defined(_WIN32) && defined(_MSC_VER) +#pragma pack(pop) +#endif +#undef SCTP_PACKED + +#define AF_CONN 123 +/* The definition of struct sockaddr_conn MUST be in + * tune with other sockaddr_* structures. + */ +#if defined(__APPLE__) || defined(__Bitrig__) || defined(__DragonFly__) || \ + defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__NetBSD__) +struct sockaddr_conn { + uint8_t sconn_len; + uint8_t sconn_family; + uint16_t sconn_port; + void *sconn_addr; +}; +#else +struct sockaddr_conn { + uint16_t sconn_family; + uint16_t sconn_port; + void *sconn_addr; +}; +#endif + +union sctp_sockstore { + struct sockaddr_in sin; + struct sockaddr_in6 sin6; + struct sockaddr_conn sconn; + struct sockaddr sa; +}; + +#define SCTP_FUTURE_ASSOC 0 +#define SCTP_CURRENT_ASSOC 1 +#define SCTP_ALL_ASSOC 2 + +#define SCTP_EVENT_READ 0x0001 +#define SCTP_EVENT_WRITE 0x0002 +#define SCTP_EVENT_ERROR 0x0004 + +/*** Structures and definitions to use the socket API ***/ + +#define SCTP_ALIGN_RESV_PAD 92 +#define SCTP_ALIGN_RESV_PAD_SHORT 76 + +struct sctp_rcvinfo { + uint16_t rcv_sid; + uint16_t rcv_ssn; + uint16_t rcv_flags; + uint32_t rcv_ppid; + uint32_t rcv_tsn; + uint32_t rcv_cumtsn; + uint32_t rcv_context; + sctp_assoc_t rcv_assoc_id; +}; + +struct sctp_nxtinfo { + uint16_t nxt_sid; + uint16_t nxt_flags; + uint32_t nxt_ppid; + uint32_t nxt_length; + sctp_assoc_t nxt_assoc_id; +}; + +#define SCTP_NO_NEXT_MSG 0x0000 +#define SCTP_NEXT_MSG_AVAIL 0x0001 +#define SCTP_NEXT_MSG_ISCOMPLETE 0x0002 +#define SCTP_NEXT_MSG_IS_UNORDERED 0x0004 +#define SCTP_NEXT_MSG_IS_NOTIFICATION 0x0008 + +struct sctp_recvv_rn { + struct sctp_rcvinfo recvv_rcvinfo; + struct sctp_nxtinfo recvv_nxtinfo; +}; + +#define SCTP_RECVV_NOINFO 0 +#define SCTP_RECVV_RCVINFO 1 +#define SCTP_RECVV_NXTINFO 2 +#define SCTP_RECVV_RN 3 + +#define SCTP_SENDV_NOINFO 0 +#define SCTP_SENDV_SNDINFO 1 +#define SCTP_SENDV_PRINFO 2 +#define SCTP_SENDV_AUTHINFO 3 +#define SCTP_SENDV_SPA 4 + +#define SCTP_SEND_SNDINFO_VALID 0x00000001 +#define SCTP_SEND_PRINFO_VALID 0x00000002 +#define SCTP_SEND_AUTHINFO_VALID 0x00000004 + +struct sctp_snd_all_completes { + uint16_t sall_stream; + uint16_t sall_flags; + uint32_t sall_ppid; + uint32_t sall_context; + uint32_t sall_num_sent; + uint32_t sall_num_failed; +}; + +struct sctp_sndinfo { + uint16_t snd_sid; + uint16_t snd_flags; + uint32_t snd_ppid; + uint32_t snd_context; + sctp_assoc_t snd_assoc_id; +}; + +struct sctp_prinfo { + uint16_t pr_policy; + uint32_t pr_value; +}; + +struct sctp_authinfo { + uint16_t auth_keynumber; +}; + +struct sctp_sendv_spa { + uint32_t sendv_flags; + struct sctp_sndinfo sendv_sndinfo; + struct sctp_prinfo sendv_prinfo; + struct sctp_authinfo sendv_authinfo; +}; + +struct sctp_udpencaps { + struct sockaddr_storage sue_address; + uint32_t sue_assoc_id; + uint16_t sue_port; +}; + +/******** Notifications **************/ + +/* notification types */ +#define SCTP_ASSOC_CHANGE 0x0001 +#define SCTP_PEER_ADDR_CHANGE 0x0002 +#define SCTP_REMOTE_ERROR 0x0003 +#define SCTP_SEND_FAILED 0x0004 +#define SCTP_SHUTDOWN_EVENT 0x0005 +#define SCTP_ADAPTATION_INDICATION 0x0006 +#define SCTP_PARTIAL_DELIVERY_EVENT 0x0007 +#define SCTP_AUTHENTICATION_EVENT 0x0008 +#define SCTP_STREAM_RESET_EVENT 0x0009 +#define SCTP_SENDER_DRY_EVENT 0x000a +#define SCTP_NOTIFICATIONS_STOPPED_EVENT 0x000b +#define SCTP_ASSOC_RESET_EVENT 0x000c +#define SCTP_STREAM_CHANGE_EVENT 0x000d +#define SCTP_SEND_FAILED_EVENT 0x000e + +/* notification event structures */ + + +/* association change event */ +struct sctp_assoc_change { + uint16_t sac_type; + uint16_t sac_flags; + uint32_t sac_length; + uint16_t sac_state; + uint16_t sac_error; + uint16_t sac_outbound_streams; + uint16_t sac_inbound_streams; + sctp_assoc_t sac_assoc_id; + uint8_t sac_info[]; /* not available yet */ +}; + +/* sac_state values */ +#define SCTP_COMM_UP 0x0001 +#define SCTP_COMM_LOST 0x0002 +#define SCTP_RESTART 0x0003 +#define SCTP_SHUTDOWN_COMP 0x0004 +#define SCTP_CANT_STR_ASSOC 0x0005 + +/* sac_info values */ +#define SCTP_ASSOC_SUPPORTS_PR 0x01 +#define SCTP_ASSOC_SUPPORTS_AUTH 0x02 +#define SCTP_ASSOC_SUPPORTS_ASCONF 0x03 +#define SCTP_ASSOC_SUPPORTS_MULTIBUF 0x04 +#define SCTP_ASSOC_SUPPORTS_RE_CONFIG 0x05 +#define SCTP_ASSOC_SUPPORTS_INTERLEAVING 0x06 +#define SCTP_ASSOC_SUPPORTS_MAX 0x06 + +/* Address event */ +struct sctp_paddr_change { + uint16_t spc_type; + uint16_t spc_flags; + uint32_t spc_length; + struct sockaddr_storage spc_aaddr; + uint32_t spc_state; + uint32_t spc_error; + sctp_assoc_t spc_assoc_id; + uint8_t spc_padding[4]; +}; + +/* paddr state values */ +#define SCTP_ADDR_AVAILABLE 0x0001 +#define SCTP_ADDR_UNREACHABLE 0x0002 +#define SCTP_ADDR_REMOVED 0x0003 +#define SCTP_ADDR_ADDED 0x0004 +#define SCTP_ADDR_MADE_PRIM 0x0005 +#define SCTP_ADDR_CONFIRMED 0x0006 + +/* remote error events */ +struct sctp_remote_error { + uint16_t sre_type; + uint16_t sre_flags; + uint32_t sre_length; + uint16_t sre_error; + sctp_assoc_t sre_assoc_id; + uint8_t sre_data[]; +}; + +/* shutdown event */ +struct sctp_shutdown_event { + uint16_t sse_type; + uint16_t sse_flags; + uint32_t sse_length; + sctp_assoc_t sse_assoc_id; +}; + +/* Adaptation layer indication */ +struct sctp_adaptation_event { + uint16_t sai_type; + uint16_t sai_flags; + uint32_t sai_length; + uint32_t sai_adaptation_ind; + sctp_assoc_t sai_assoc_id; +}; + +/* Partial delivery event */ +struct sctp_pdapi_event { + uint16_t pdapi_type; + uint16_t pdapi_flags; + uint32_t pdapi_length; + uint32_t pdapi_indication; + uint32_t pdapi_stream; + uint32_t pdapi_seq; + sctp_assoc_t pdapi_assoc_id; +}; + +/* indication values */ +#define SCTP_PARTIAL_DELIVERY_ABORTED 0x0001 + +/* SCTP authentication event */ +struct sctp_authkey_event { + uint16_t auth_type; + uint16_t auth_flags; + uint32_t auth_length; + uint16_t auth_keynumber; + uint32_t auth_indication; + sctp_assoc_t auth_assoc_id; +}; + +/* indication values */ +#define SCTP_AUTH_NEW_KEY 0x0001 +#define SCTP_AUTH_NO_AUTH 0x0002 +#define SCTP_AUTH_FREE_KEY 0x0003 + +/* SCTP sender dry event */ +struct sctp_sender_dry_event { + uint16_t sender_dry_type; + uint16_t sender_dry_flags; + uint32_t sender_dry_length; + sctp_assoc_t sender_dry_assoc_id; +}; + + +/* Stream reset event - subscribe to SCTP_STREAM_RESET_EVENT */ +struct sctp_stream_reset_event { + uint16_t strreset_type; + uint16_t strreset_flags; + uint32_t strreset_length; + sctp_assoc_t strreset_assoc_id; + uint16_t strreset_stream_list[]; +}; + +/* flags in stream_reset_event (strreset_flags) */ +#define SCTP_STREAM_RESET_INCOMING_SSN 0x0001 +#define SCTP_STREAM_RESET_OUTGOING_SSN 0x0002 +#define SCTP_STREAM_RESET_DENIED 0x0004 /* SCTP_STRRESET_FAILED */ +#define SCTP_STREAM_RESET_FAILED 0x0008 /* SCTP_STRRESET_FAILED */ +#define SCTP_STREAM_CHANGED_DENIED 0x0010 + +#define SCTP_STREAM_RESET_INCOMING 0x00000001 +#define SCTP_STREAM_RESET_OUTGOING 0x00000002 + + +/* Assoc reset event - subscribe to SCTP_ASSOC_RESET_EVENT */ +struct sctp_assoc_reset_event { + uint16_t assocreset_type; + uint16_t assocreset_flags; + uint32_t assocreset_length; + sctp_assoc_t assocreset_assoc_id; + uint32_t assocreset_local_tsn; + uint32_t assocreset_remote_tsn; +}; + +#define SCTP_ASSOC_RESET_DENIED 0x0004 +#define SCTP_ASSOC_RESET_FAILED 0x0008 + + +/* Stream change event - subscribe to SCTP_STREAM_CHANGE_EVENT */ +struct sctp_stream_change_event { + uint16_t strchange_type; + uint16_t strchange_flags; + uint32_t strchange_length; + sctp_assoc_t strchange_assoc_id; + uint16_t strchange_instrms; + uint16_t strchange_outstrms; +}; + +#define SCTP_STREAM_CHANGE_DENIED 0x0004 +#define SCTP_STREAM_CHANGE_FAILED 0x0008 + + +/* SCTP send failed event */ +struct sctp_send_failed_event { + uint16_t ssfe_type; + uint16_t ssfe_flags; + uint32_t ssfe_length; + uint32_t ssfe_error; + struct sctp_sndinfo ssfe_info; + sctp_assoc_t ssfe_assoc_id; + uint8_t ssfe_data[]; +}; + +/* flag that indicates state of data */ +#define SCTP_DATA_UNSENT 0x0001 /* inqueue never on wire */ +#define SCTP_DATA_SENT 0x0002 /* on wire at failure */ + +/* SCTP event option */ +struct sctp_event { + sctp_assoc_t se_assoc_id; + uint16_t se_type; + uint8_t se_on; +}; + +union sctp_notification { + struct sctp_tlv { + uint16_t sn_type; + uint16_t sn_flags; + uint32_t sn_length; + } sn_header; + struct sctp_assoc_change sn_assoc_change; + struct sctp_paddr_change sn_paddr_change; + struct sctp_remote_error sn_remote_error; + struct sctp_shutdown_event sn_shutdown_event; + struct sctp_adaptation_event sn_adaptation_event; + struct sctp_pdapi_event sn_pdapi_event; + struct sctp_authkey_event sn_auth_event; + struct sctp_sender_dry_event sn_sender_dry_event; + struct sctp_send_failed_event sn_send_failed_event; + struct sctp_stream_reset_event sn_strreset_event; + struct sctp_assoc_reset_event sn_assocreset_event; + struct sctp_stream_change_event sn_strchange_event; +}; + +struct sctp_event_subscribe { + uint8_t sctp_data_io_event; + uint8_t sctp_association_event; + uint8_t sctp_address_event; + uint8_t sctp_send_failure_event; + uint8_t sctp_peer_error_event; + uint8_t sctp_shutdown_event; + uint8_t sctp_partial_delivery_event; + uint8_t sctp_adaptation_layer_event; + uint8_t sctp_authentication_event; + uint8_t sctp_sender_dry_event; + uint8_t sctp_stream_reset_event; +}; + + + +/* Flags that go into the sinfo->sinfo_flags field */ +#define SCTP_DATA_LAST_FRAG 0x0001 /* tail part of the message could not be sent */ +#define SCTP_DATA_NOT_FRAG 0x0003 /* complete message could not be sent */ +#define SCTP_NOTIFICATION 0x0010 /* next message is a notification */ +#define SCTP_COMPLETE 0x0020 /* next message is complete */ +#define SCTP_EOF 0x0100 /* Start shutdown procedures */ +#define SCTP_ABORT 0x0200 /* Send an ABORT to peer */ +#define SCTP_UNORDERED 0x0400 /* Message is un-ordered */ +#define SCTP_ADDR_OVER 0x0800 /* Override the primary-address */ +#define SCTP_SENDALL 0x1000 /* Send this on all associations */ +#define SCTP_EOR 0x2000 /* end of message signal */ +#define SCTP_SACK_IMMEDIATELY 0x4000 /* Set I-Bit */ + +#define INVALID_SINFO_FLAG(x) (((x) & 0xfffffff0 \ + & ~(SCTP_EOF | SCTP_ABORT | SCTP_UNORDERED |\ + SCTP_ADDR_OVER | SCTP_SENDALL | SCTP_EOR |\ + SCTP_SACK_IMMEDIATELY)) != 0) +/* for the endpoint */ + +/* The lower byte is an enumeration of PR-SCTP policies */ +#define SCTP_PR_SCTP_NONE 0x0000 /* Reliable transfer */ +#define SCTP_PR_SCTP_TTL 0x0001 /* Time based PR-SCTP */ +#define SCTP_PR_SCTP_BUF 0x0002 /* Buffer based PR-SCTP */ +#define SCTP_PR_SCTP_RTX 0x0003 /* Number of retransmissions based PR-SCTP */ + +#define PR_SCTP_POLICY(x) ((x) & 0x0f) +#define PR_SCTP_ENABLED(x) (PR_SCTP_POLICY(x) != SCTP_PR_SCTP_NONE) +#define PR_SCTP_TTL_ENABLED(x) (PR_SCTP_POLICY(x) == SCTP_PR_SCTP_TTL) +#define PR_SCTP_BUF_ENABLED(x) (PR_SCTP_POLICY(x) == SCTP_PR_SCTP_BUF) +#define PR_SCTP_RTX_ENABLED(x) (PR_SCTP_POLICY(x) == SCTP_PR_SCTP_RTX) +#define PR_SCTP_INVALID_POLICY(x) (PR_SCTP_POLICY(x) > SCTP_PR_SCTP_RTX) + + +/* + * user socket options: socket API defined + */ +/* + * read-write options + */ +#define SCTP_RTOINFO 0x00000001 +#define SCTP_ASSOCINFO 0x00000002 +#define SCTP_INITMSG 0x00000003 +#define SCTP_NODELAY 0x00000004 +#define SCTP_AUTOCLOSE 0x00000005 +#define SCTP_PRIMARY_ADDR 0x00000007 +#define SCTP_ADAPTATION_LAYER 0x00000008 +#define SCTP_DISABLE_FRAGMENTS 0x00000009 +#define SCTP_PEER_ADDR_PARAMS 0x0000000a +/* ancillary data/notification interest options */ +/* Without this applied we will give V4 and V6 addresses on a V6 socket */ +#define SCTP_I_WANT_MAPPED_V4_ADDR 0x0000000d +#define SCTP_MAXSEG 0x0000000e +#define SCTP_DELAYED_SACK 0x0000000f +#define SCTP_FRAGMENT_INTERLEAVE 0x00000010 +#define SCTP_PARTIAL_DELIVERY_POINT 0x00000011 +/* authentication support */ +#define SCTP_HMAC_IDENT 0x00000014 +#define SCTP_AUTH_ACTIVE_KEY 0x00000015 +#define SCTP_AUTO_ASCONF 0x00000018 +#define SCTP_MAX_BURST 0x00000019 +/* assoc level context */ +#define SCTP_CONTEXT 0x0000001a +/* explicit EOR signalling */ +#define SCTP_EXPLICIT_EOR 0x0000001b +#define SCTP_REUSE_PORT 0x0000001c + +#define SCTP_EVENT 0x0000001e +#define SCTP_RECVRCVINFO 0x0000001f +#define SCTP_RECVNXTINFO 0x00000020 +#define SCTP_DEFAULT_SNDINFO 0x00000021 +#define SCTP_DEFAULT_PRINFO 0x00000022 +#define SCTP_REMOTE_UDP_ENCAPS_PORT 0x00000024 +#define SCTP_ECN_SUPPORTED 0x00000025 +#define SCTP_PR_SUPPORTED 0x00000026 +#define SCTP_AUTH_SUPPORTED 0x00000027 +#define SCTP_ASCONF_SUPPORTED 0x00000028 +#define SCTP_RECONFIG_SUPPORTED 0x00000029 +#define SCTP_NRSACK_SUPPORTED 0x00000030 +#define SCTP_PKTDROP_SUPPORTED 0x00000031 +#define SCTP_MAX_CWND 0x00000032 + +#define SCTP_ENABLE_STREAM_RESET 0x00000900 /* struct sctp_assoc_value */ + +/* Pluggable Stream Scheduling Socket option */ +#define SCTP_PLUGGABLE_SS 0x00001203 +#define SCTP_SS_VALUE 0x00001204 + +/* + * read-only options + */ +#define SCTP_STATUS 0x00000100 +#define SCTP_GET_PEER_ADDR_INFO 0x00000101 +/* authentication support */ +#define SCTP_PEER_AUTH_CHUNKS 0x00000102 +#define SCTP_LOCAL_AUTH_CHUNKS 0x00000103 +#define SCTP_GET_ASSOC_NUMBER 0x00000104 +#define SCTP_GET_ASSOC_ID_LIST 0x00000105 +#define SCTP_TIMEOUTS 0x00000106 +#define SCTP_PR_STREAM_STATUS 0x00000107 +#define SCTP_PR_ASSOC_STATUS 0x00000108 + +/* + * write-only options + */ +#define SCTP_SET_PEER_PRIMARY_ADDR 0x00000006 +#define SCTP_AUTH_CHUNK 0x00000012 +#define SCTP_AUTH_KEY 0x00000013 +#define SCTP_AUTH_DEACTIVATE_KEY 0x0000001d +#define SCTP_AUTH_DELETE_KEY 0x00000016 +#define SCTP_RESET_STREAMS 0x00000901 /* struct sctp_reset_streams */ +#define SCTP_RESET_ASSOC 0x00000902 /* sctp_assoc_t */ +#define SCTP_ADD_STREAMS 0x00000903 /* struct sctp_add_streams */ + +struct sctp_initmsg { + uint16_t sinit_num_ostreams; + uint16_t sinit_max_instreams; + uint16_t sinit_max_attempts; + uint16_t sinit_max_init_timeo; +}; + +struct sctp_rtoinfo { + sctp_assoc_t srto_assoc_id; + uint32_t srto_initial; + uint32_t srto_max; + uint32_t srto_min; +}; + +struct sctp_assocparams { + sctp_assoc_t sasoc_assoc_id; + uint32_t sasoc_peer_rwnd; + uint32_t sasoc_local_rwnd; + uint32_t sasoc_cookie_life; + uint16_t sasoc_asocmaxrxt; + uint16_t sasoc_number_peer_destinations; +}; + +struct sctp_setprim { + struct sockaddr_storage ssp_addr; + sctp_assoc_t ssp_assoc_id; + uint8_t ssp_padding[4]; +}; + +struct sctp_setadaptation { + uint32_t ssb_adaptation_ind; +}; + +struct sctp_paddrparams { + struct sockaddr_storage spp_address; + sctp_assoc_t spp_assoc_id; + uint32_t spp_hbinterval; + uint32_t spp_pathmtu; + uint32_t spp_flags; + uint32_t spp_ipv6_flowlabel; + uint16_t spp_pathmaxrxt; + uint8_t spp_dscp; +}; + +#define SPP_HB_ENABLE 0x00000001 +#define SPP_HB_DISABLE 0x00000002 +#define SPP_HB_DEMAND 0x00000004 +#define SPP_PMTUD_ENABLE 0x00000008 +#define SPP_PMTUD_DISABLE 0x00000010 +#define SPP_HB_TIME_IS_ZERO 0x00000080 +#define SPP_IPV6_FLOWLABEL 0x00000100 +#define SPP_DSCP 0x00000200 + +/* Used for SCTP_MAXSEG, SCTP_MAX_BURST, SCTP_ENABLE_STREAM_RESET, and SCTP_CONTEXT */ +struct sctp_assoc_value { + sctp_assoc_t assoc_id; + uint32_t assoc_value; +}; + +/* To enable stream reset */ +#define SCTP_ENABLE_RESET_STREAM_REQ 0x00000001 +#define SCTP_ENABLE_RESET_ASSOC_REQ 0x00000002 +#define SCTP_ENABLE_CHANGE_ASSOC_REQ 0x00000004 +#define SCTP_ENABLE_VALUE_MASK 0x00000007 + +struct sctp_reset_streams { + sctp_assoc_t srs_assoc_id; + uint16_t srs_flags; + uint16_t srs_number_streams; /* 0 == ALL */ + uint16_t srs_stream_list[]; /* list if strrst_num_streams is not 0 */ +}; + +struct sctp_add_streams { + sctp_assoc_t sas_assoc_id; + uint16_t sas_instrms; + uint16_t sas_outstrms; +}; + +struct sctp_hmacalgo { + uint32_t shmac_number_of_idents; + uint16_t shmac_idents[]; +}; + +/* AUTH hmac_id */ +#define SCTP_AUTH_HMAC_ID_RSVD 0x0000 +#define SCTP_AUTH_HMAC_ID_SHA1 0x0001 /* default, mandatory */ +#define SCTP_AUTH_HMAC_ID_SHA256 0x0003 +#define SCTP_AUTH_HMAC_ID_SHA224 0x0004 +#define SCTP_AUTH_HMAC_ID_SHA384 0x0005 +#define SCTP_AUTH_HMAC_ID_SHA512 0x0006 + + +struct sctp_sack_info { + sctp_assoc_t sack_assoc_id; + uint32_t sack_delay; + uint32_t sack_freq; +}; + +struct sctp_default_prinfo { + uint16_t pr_policy; + uint32_t pr_value; + sctp_assoc_t pr_assoc_id; +}; + +struct sctp_paddrinfo { + struct sockaddr_storage spinfo_address; + sctp_assoc_t spinfo_assoc_id; + int32_t spinfo_state; + uint32_t spinfo_cwnd; + uint32_t spinfo_srtt; + uint32_t spinfo_rto; + uint32_t spinfo_mtu; +}; + +struct sctp_status { + sctp_assoc_t sstat_assoc_id; + int32_t sstat_state; + uint32_t sstat_rwnd; + uint16_t sstat_unackdata; + uint16_t sstat_penddata; + uint16_t sstat_instrms; + uint16_t sstat_outstrms; + uint32_t sstat_fragmentation_point; + struct sctp_paddrinfo sstat_primary; +}; + +/* + * user state values + */ +#define SCTP_CLOSED 0x0000 +#define SCTP_BOUND 0x1000 +#define SCTP_LISTEN 0x2000 +#define SCTP_COOKIE_WAIT 0x0002 +#define SCTP_COOKIE_ECHOED 0x0004 +#define SCTP_ESTABLISHED 0x0008 +#define SCTP_SHUTDOWN_SENT 0x0010 +#define SCTP_SHUTDOWN_RECEIVED 0x0020 +#define SCTP_SHUTDOWN_ACK_SENT 0x0040 +#define SCTP_SHUTDOWN_PENDING 0x0080 + + +#define SCTP_ACTIVE 0x0001 /* SCTP_ADDR_REACHABLE */ +#define SCTP_INACTIVE 0x0002 /* neither SCTP_ADDR_REACHABLE + nor SCTP_ADDR_UNCONFIRMED */ +#define SCTP_UNCONFIRMED 0x0200 /* SCTP_ADDR_UNCONFIRMED */ + +struct sctp_authchunks { + sctp_assoc_t gauth_assoc_id; +/* uint32_t gauth_number_of_chunks; not available */ + uint8_t gauth_chunks[]; +}; + +struct sctp_assoc_ids { + uint32_t gaids_number_of_ids; + sctp_assoc_t gaids_assoc_id[]; +}; + +struct sctp_setpeerprim { + struct sockaddr_storage sspp_addr; + sctp_assoc_t sspp_assoc_id; + uint8_t sspp_padding[4]; +}; + +struct sctp_authchunk { + uint8_t sauth_chunk; +}; + + +struct sctp_get_nonce_values { + sctp_assoc_t gn_assoc_id; + uint32_t gn_peers_tag; + uint32_t gn_local_tag; +}; + + +/* + * Main SCTP chunk types + */ +/************0x00 series ***********/ +#define SCTP_DATA 0x00 +#define SCTP_INITIATION 0x01 +#define SCTP_INITIATION_ACK 0x02 +#define SCTP_SELECTIVE_ACK 0x03 +#define SCTP_HEARTBEAT_REQUEST 0x04 +#define SCTP_HEARTBEAT_ACK 0x05 +#define SCTP_ABORT_ASSOCIATION 0x06 +#define SCTP_SHUTDOWN 0x07 +#define SCTP_SHUTDOWN_ACK 0x08 +#define SCTP_OPERATION_ERROR 0x09 +#define SCTP_COOKIE_ECHO 0x0a +#define SCTP_COOKIE_ACK 0x0b +#define SCTP_ECN_ECHO 0x0c +#define SCTP_ECN_CWR 0x0d +#define SCTP_SHUTDOWN_COMPLETE 0x0e +/* RFC4895 */ +#define SCTP_AUTHENTICATION 0x0f +/* EY nr_sack chunk id*/ +#define SCTP_NR_SELECTIVE_ACK 0x10 +/************0x40 series ***********/ +/************0x80 series ***********/ +/* RFC5061 */ +#define SCTP_ASCONF_ACK 0x80 +/* draft-ietf-stewart-pktdrpsctp */ +#define SCTP_PACKET_DROPPED 0x81 +/* draft-ietf-stewart-strreset-xxx */ +#define SCTP_STREAM_RESET 0x82 + +/* RFC4820 */ +#define SCTP_PAD_CHUNK 0x84 +/************0xc0 series ***********/ +/* RFC3758 */ +#define SCTP_FORWARD_CUM_TSN 0xc0 +/* RFC5061 */ +#define SCTP_ASCONF 0xc1 + +struct sctp_authkey { + sctp_assoc_t sca_assoc_id; + uint16_t sca_keynumber; + uint16_t sca_keylength; + uint8_t sca_key[]; +}; + +struct sctp_authkeyid { + sctp_assoc_t scact_assoc_id; + uint16_t scact_keynumber; +}; + +struct sctp_cc_option { + int option; + struct sctp_assoc_value aid_value; +}; + +struct sctp_stream_value { + sctp_assoc_t assoc_id; + uint16_t stream_id; + uint16_t stream_value; +}; + +struct sctp_timeouts { + sctp_assoc_t stimo_assoc_id; + uint32_t stimo_init; + uint32_t stimo_data; + uint32_t stimo_sack; + uint32_t stimo_shutdown; + uint32_t stimo_heartbeat; + uint32_t stimo_cookie; + uint32_t stimo_shutdownack; +}; + +struct sctp_prstatus { + sctp_assoc_t sprstat_assoc_id; + uint16_t sprstat_sid; + uint16_t sprstat_policy; + uint64_t sprstat_abandoned_unsent; + uint64_t sprstat_abandoned_sent; +}; + +/* Standard TCP Congestion Control */ +#define SCTP_CC_RFC2581 0x00000000 +/* High Speed TCP Congestion Control (Floyd) */ +#define SCTP_CC_HSTCP 0x00000001 +/* HTCP Congestion Control */ +#define SCTP_CC_HTCP 0x00000002 +/* RTCC Congestion Control - RFC2581 plus */ +#define SCTP_CC_RTCC 0x00000003 + +#define SCTP_CC_OPT_RTCC_SETMODE 0x00002000 +#define SCTP_CC_OPT_USE_DCCC_EC 0x00002001 +#define SCTP_CC_OPT_STEADY_STEP 0x00002002 + +#define SCTP_CMT_OFF 0 +#define SCTP_CMT_BASE 1 +#define SCTP_CMT_RPV1 2 +#define SCTP_CMT_RPV2 3 +#define SCTP_CMT_MPTCP 4 +#define SCTP_CMT_MAX SCTP_CMT_MPTCP + +/* RS - Supported stream scheduling modules for pluggable + * stream scheduling + */ +/* Default simple round-robin */ +#define SCTP_SS_DEFAULT 0x00000000 +/* Real round-robin */ +#define SCTP_SS_ROUND_ROBIN 0x00000001 +/* Real round-robin per packet */ +#define SCTP_SS_ROUND_ROBIN_PACKET 0x00000002 +/* Priority */ +#define SCTP_SS_PRIORITY 0x00000003 +/* Fair Bandwidth */ +#define SCTP_SS_FAIR_BANDWITH 0x00000004 +/* First-come, first-serve */ +#define SCTP_SS_FIRST_COME 0x00000005 + +/******************** System calls *************/ + +struct socket; + +void +usrsctp_init(uint16_t, + int (*)(void *addr, void *buffer, size_t length, uint8_t tos, uint8_t set_df), + void (*)(const char *format, ...)); + +void +usrsctp_init_nothreads(uint16_t, + int (*)(void *addr, void *buffer, size_t length, uint8_t tos, uint8_t set_df), + void (*)(const char *format, ...)); + +struct socket * +usrsctp_socket(int domain, int type, int protocol, + int (*receive_cb)(struct socket *sock, union sctp_sockstore addr, void *data, + size_t datalen, struct sctp_rcvinfo, int flags, void *ulp_info), + int (*send_cb)(struct socket *sock, uint32_t sb_free, void *ulp_info), + uint32_t sb_threshold, + void *ulp_info); + +int +usrsctp_setsockopt(struct socket *so, + int level, + int option_name, + const void *option_value, + socklen_t option_len); + +int +usrsctp_getsockopt(struct socket *so, + int level, + int option_name, + void *option_value, + socklen_t *option_len); + +int +usrsctp_opt_info(struct socket *so, + sctp_assoc_t id, + int opt, + void *arg, + socklen_t *size); + +int +usrsctp_getpaddrs(struct socket *so, + sctp_assoc_t id, + struct sockaddr **raddrs); + +void +usrsctp_freepaddrs(struct sockaddr *addrs); + +int +usrsctp_getladdrs(struct socket *so, + sctp_assoc_t id, + struct sockaddr **raddrs); + +void +usrsctp_freeladdrs(struct sockaddr *addrs); + +ssize_t +usrsctp_sendv(struct socket *so, + const void *data, + size_t len, + struct sockaddr *to, + int addrcnt, + void *info, + socklen_t infolen, + unsigned int infotype, + int flags); + +ssize_t +usrsctp_recvv(struct socket *so, + void *dbuf, + size_t len, + struct sockaddr *from, + socklen_t * fromlen, + void *info, + socklen_t *infolen, + unsigned int *infotype, + int *msg_flags); + +int +usrsctp_bind(struct socket *so, + struct sockaddr *name, + socklen_t namelen); + +#define SCTP_BINDX_ADD_ADDR 0x00008001 +#define SCTP_BINDX_REM_ADDR 0x00008002 + +int +usrsctp_bindx(struct socket *so, + struct sockaddr *addrs, + int addrcnt, + int flags); + +int +usrsctp_listen(struct socket *so, + int backlog); + +struct socket * +usrsctp_accept(struct socket *so, + struct sockaddr * aname, + socklen_t * anamelen); + +struct socket * +usrsctp_peeloff(struct socket *, sctp_assoc_t); + +int +usrsctp_connect(struct socket *so, + struct sockaddr *name, + socklen_t namelen); + +int +usrsctp_connectx(struct socket *so, + const struct sockaddr *addrs, int addrcnt, + sctp_assoc_t *id); + +void +usrsctp_close(struct socket *so); + +sctp_assoc_t +usrsctp_getassocid(struct socket *, struct sockaddr *); + +int +usrsctp_finish(void); + +int +usrsctp_shutdown(struct socket *so, int how); + +void +usrsctp_conninput(void *, const void *, size_t, uint8_t); + +int +usrsctp_set_non_blocking(struct socket *, int); + +int +usrsctp_get_non_blocking(struct socket *); + +void +usrsctp_register_address(void *); + +void +usrsctp_deregister_address(void *); + +int +usrsctp_set_ulpinfo(struct socket *, void *); + +int +usrsctp_get_ulpinfo(struct socket *, void **); + +int +usrsctp_set_upcall(struct socket *so, + void (*upcall)(struct socket *, void *, int), + void *arg); + +int +usrsctp_get_events(struct socket *so); + + +void +usrsctp_handle_timers(uint32_t elapsed_milliseconds); + +#define SCTP_DUMP_OUTBOUND 1 +#define SCTP_DUMP_INBOUND 0 + +char * +usrsctp_dumppacket(const void *, size_t, int); + +void +usrsctp_freedumpbuffer(char *); + +void +usrsctp_enable_crc32c_offload(void); + +void +usrsctp_disable_crc32c_offload(void); + +uint32_t +usrsctp_crc32c(void *, size_t); + +#define USRSCTP_TUNABLE_DECL(__field) \ +int usrsctp_tunable_set_ ## __field(uint32_t value);\ +uint32_t usrsctp_sysctl_get_ ## __field(void); + +USRSCTP_TUNABLE_DECL(sctp_hashtblsize) +USRSCTP_TUNABLE_DECL(sctp_pcbtblsize) +USRSCTP_TUNABLE_DECL(sctp_chunkscale) + +#define USRSCTP_SYSCTL_DECL(__field) \ +int usrsctp_sysctl_set_ ## __field(uint32_t value);\ +uint32_t usrsctp_sysctl_get_ ## __field(void); + +USRSCTP_SYSCTL_DECL(sctp_sendspace) +USRSCTP_SYSCTL_DECL(sctp_recvspace) +USRSCTP_SYSCTL_DECL(sctp_auto_asconf) +USRSCTP_SYSCTL_DECL(sctp_multiple_asconfs) +USRSCTP_SYSCTL_DECL(sctp_ecn_enable) +USRSCTP_SYSCTL_DECL(sctp_pr_enable) +USRSCTP_SYSCTL_DECL(sctp_auth_enable) +USRSCTP_SYSCTL_DECL(sctp_asconf_enable) +USRSCTP_SYSCTL_DECL(sctp_reconfig_enable) +USRSCTP_SYSCTL_DECL(sctp_nrsack_enable) +USRSCTP_SYSCTL_DECL(sctp_pktdrop_enable) +USRSCTP_SYSCTL_DECL(sctp_no_csum_on_loopback) +USRSCTP_SYSCTL_DECL(sctp_peer_chunk_oh) +USRSCTP_SYSCTL_DECL(sctp_max_burst_default) +USRSCTP_SYSCTL_DECL(sctp_max_chunks_on_queue) +USRSCTP_SYSCTL_DECL(sctp_min_split_point) +USRSCTP_SYSCTL_DECL(sctp_delayed_sack_time_default) +USRSCTP_SYSCTL_DECL(sctp_sack_freq_default) +USRSCTP_SYSCTL_DECL(sctp_system_free_resc_limit) +USRSCTP_SYSCTL_DECL(sctp_asoc_free_resc_limit) +USRSCTP_SYSCTL_DECL(sctp_heartbeat_interval_default) +USRSCTP_SYSCTL_DECL(sctp_pmtu_raise_time_default) +USRSCTP_SYSCTL_DECL(sctp_shutdown_guard_time_default) +USRSCTP_SYSCTL_DECL(sctp_secret_lifetime_default) +USRSCTP_SYSCTL_DECL(sctp_rto_max_default) +USRSCTP_SYSCTL_DECL(sctp_rto_min_default) +USRSCTP_SYSCTL_DECL(sctp_rto_initial_default) +USRSCTP_SYSCTL_DECL(sctp_init_rto_max_default) +USRSCTP_SYSCTL_DECL(sctp_valid_cookie_life_default) +USRSCTP_SYSCTL_DECL(sctp_init_rtx_max_default) +USRSCTP_SYSCTL_DECL(sctp_assoc_rtx_max_default) +USRSCTP_SYSCTL_DECL(sctp_path_rtx_max_default) +USRSCTP_SYSCTL_DECL(sctp_add_more_threshold) +USRSCTP_SYSCTL_DECL(sctp_nr_incoming_streams_default) +USRSCTP_SYSCTL_DECL(sctp_nr_outgoing_streams_default) +USRSCTP_SYSCTL_DECL(sctp_cmt_on_off) +USRSCTP_SYSCTL_DECL(sctp_cmt_use_dac) +USRSCTP_SYSCTL_DECL(sctp_use_cwnd_based_maxburst) +USRSCTP_SYSCTL_DECL(sctp_nat_friendly) +USRSCTP_SYSCTL_DECL(sctp_L2_abc_variable) +USRSCTP_SYSCTL_DECL(sctp_mbuf_threshold_count) +USRSCTP_SYSCTL_DECL(sctp_do_drain) +USRSCTP_SYSCTL_DECL(sctp_hb_maxburst) +USRSCTP_SYSCTL_DECL(sctp_abort_if_one_2_one_hits_limit) +USRSCTP_SYSCTL_DECL(sctp_min_residual) +USRSCTP_SYSCTL_DECL(sctp_max_retran_chunk) +USRSCTP_SYSCTL_DECL(sctp_logging_level) +USRSCTP_SYSCTL_DECL(sctp_default_cc_module) +USRSCTP_SYSCTL_DECL(sctp_default_frag_interleave) +USRSCTP_SYSCTL_DECL(sctp_mobility_base) +USRSCTP_SYSCTL_DECL(sctp_mobility_fasthandoff) +USRSCTP_SYSCTL_DECL(sctp_inits_include_nat_friendly) +USRSCTP_SYSCTL_DECL(sctp_udp_tunneling_port) +USRSCTP_SYSCTL_DECL(sctp_enable_sack_immediately) +USRSCTP_SYSCTL_DECL(sctp_vtag_time_wait) +USRSCTP_SYSCTL_DECL(sctp_blackhole) +USRSCTP_SYSCTL_DECL(sctp_sendall_limit) +USRSCTP_SYSCTL_DECL(sctp_diag_info_code) +USRSCTP_SYSCTL_DECL(sctp_fr_max_burst_default) +USRSCTP_SYSCTL_DECL(sctp_path_pf_threshold) +USRSCTP_SYSCTL_DECL(sctp_default_ss_module) +USRSCTP_SYSCTL_DECL(sctp_rttvar_bw) +USRSCTP_SYSCTL_DECL(sctp_rttvar_rtt) +USRSCTP_SYSCTL_DECL(sctp_rttvar_eqret) +USRSCTP_SYSCTL_DECL(sctp_steady_step) +USRSCTP_SYSCTL_DECL(sctp_use_dccc_ecn) +USRSCTP_SYSCTL_DECL(sctp_buffer_splitting) +USRSCTP_SYSCTL_DECL(sctp_initial_cwnd) +USRSCTP_SYSCTL_DECL(sctp_ootb_with_zero_cksum) +#ifdef SCTP_DEBUG +USRSCTP_SYSCTL_DECL(sctp_debug_on) +/* More specific values can be found in sctp_constants, but + * are not considered to be part of the API. + */ +#define SCTP_DEBUG_NONE 0x00000000 +#define SCTP_DEBUG_ALL 0xffffffff +#endif +#undef USRSCTP_SYSCTL_DECL +struct sctp_timeval { + uint32_t tv_sec; + uint32_t tv_usec; +}; + +struct sctpstat { + struct sctp_timeval sctps_discontinuitytime; /* sctpStats 18 (TimeStamp) */ + /* MIB according to RFC 3873 */ + uint32_t sctps_currestab; /* sctpStats 1 (Gauge32) */ + uint32_t sctps_activeestab; /* sctpStats 2 (Counter32) */ + uint32_t sctps_restartestab; + uint32_t sctps_collisionestab; + uint32_t sctps_passiveestab; /* sctpStats 3 (Counter32) */ + uint32_t sctps_aborted; /* sctpStats 4 (Counter32) */ + uint32_t sctps_shutdown; /* sctpStats 5 (Counter32) */ + uint32_t sctps_outoftheblue; /* sctpStats 6 (Counter32) */ + uint32_t sctps_checksumerrors; /* sctpStats 7 (Counter32) */ + uint32_t sctps_outcontrolchunks; /* sctpStats 8 (Counter64) */ + uint32_t sctps_outorderchunks; /* sctpStats 9 (Counter64) */ + uint32_t sctps_outunorderchunks; /* sctpStats 10 (Counter64) */ + uint32_t sctps_incontrolchunks; /* sctpStats 11 (Counter64) */ + uint32_t sctps_inorderchunks; /* sctpStats 12 (Counter64) */ + uint32_t sctps_inunorderchunks; /* sctpStats 13 (Counter64) */ + uint32_t sctps_fragusrmsgs; /* sctpStats 14 (Counter64) */ + uint32_t sctps_reasmusrmsgs; /* sctpStats 15 (Counter64) */ + uint32_t sctps_outpackets; /* sctpStats 16 (Counter64) */ + uint32_t sctps_inpackets; /* sctpStats 17 (Counter64) */ + + /* input statistics: */ + uint32_t sctps_recvpackets; /* total input packets */ + uint32_t sctps_recvdatagrams; /* total input datagrams */ + uint32_t sctps_recvpktwithdata; /* total packets that had data */ + uint32_t sctps_recvsacks; /* total input SACK chunks */ + uint32_t sctps_recvdata; /* total input DATA chunks */ + uint32_t sctps_recvdupdata; /* total input duplicate DATA chunks */ + uint32_t sctps_recvheartbeat; /* total input HB chunks */ + uint32_t sctps_recvheartbeatack; /* total input HB-ACK chunks */ + uint32_t sctps_recvecne; /* total input ECNE chunks */ + uint32_t sctps_recvauth; /* total input AUTH chunks */ + uint32_t sctps_recvauthmissing; /* total input chunks missing AUTH */ + uint32_t sctps_recvivalhmacid; /* total number of invalid HMAC ids received */ + uint32_t sctps_recvivalkeyid; /* total number of invalid secret ids received */ + uint32_t sctps_recvauthfailed; /* total number of auth failed */ + uint32_t sctps_recvexpress; /* total fast path receives all one chunk */ + uint32_t sctps_recvexpressm; /* total fast path multi-part data */ + uint32_t sctps_recv_spare; /* formerly sctps_recvnocrc */ + uint32_t sctps_recvswcrc; + uint32_t sctps_recvhwcrc; + + /* output statistics: */ + uint32_t sctps_sendpackets; /* total output packets */ + uint32_t sctps_sendsacks; /* total output SACKs */ + uint32_t sctps_senddata; /* total output DATA chunks */ + uint32_t sctps_sendretransdata; /* total output retransmitted DATA chunks */ + uint32_t sctps_sendfastretrans; /* total output fast retransmitted DATA chunks */ + uint32_t sctps_sendmultfastretrans; /* total FR's that happened more than once + * to same chunk (u-del multi-fr algo). + */ + uint32_t sctps_sendheartbeat; /* total output HB chunks */ + uint32_t sctps_sendecne; /* total output ECNE chunks */ + uint32_t sctps_sendauth; /* total output AUTH chunks FIXME */ + uint32_t sctps_senderrors; /* ip_output error counter */ + uint32_t sctps_send_spare; /* formerly sctps_sendnocrc */ + uint32_t sctps_sendswcrc; + uint32_t sctps_sendhwcrc; + /* PCKDROPREP statistics: */ + uint32_t sctps_pdrpfmbox; /* Packet drop from middle box */ + uint32_t sctps_pdrpfehos; /* P-drop from end host */ + uint32_t sctps_pdrpmbda; /* P-drops with data */ + uint32_t sctps_pdrpmbct; /* P-drops, non-data, non-endhost */ + uint32_t sctps_pdrpbwrpt; /* P-drop, non-endhost, bandwidth rep only */ + uint32_t sctps_pdrpcrupt; /* P-drop, not enough for chunk header */ + uint32_t sctps_pdrpnedat; /* P-drop, not enough data to confirm */ + uint32_t sctps_pdrppdbrk; /* P-drop, where process_chunk_drop said break */ + uint32_t sctps_pdrptsnnf; /* P-drop, could not find TSN */ + uint32_t sctps_pdrpdnfnd; /* P-drop, attempt reverse TSN lookup */ + uint32_t sctps_pdrpdiwnp; /* P-drop, e-host confirms zero-rwnd */ + uint32_t sctps_pdrpdizrw; /* P-drop, midbox confirms no space */ + uint32_t sctps_pdrpbadd; /* P-drop, data did not match TSN */ + uint32_t sctps_pdrpmark; /* P-drop, TSN's marked for Fast Retran */ + /* timeouts */ + uint32_t sctps_timoiterator; /* Number of iterator timers that fired */ + uint32_t sctps_timodata; /* Number of T3 data time outs */ + uint32_t sctps_timowindowprobe; /* Number of window probe (T3) timers that fired */ + uint32_t sctps_timoinit; /* Number of INIT timers that fired */ + uint32_t sctps_timosack; /* Number of sack timers that fired */ + uint32_t sctps_timoshutdown; /* Number of shutdown timers that fired */ + uint32_t sctps_timoheartbeat; /* Number of heartbeat timers that fired */ + uint32_t sctps_timocookie; /* Number of times a cookie timeout fired */ + uint32_t sctps_timosecret; /* Number of times an endpoint changed its cookie secret*/ + uint32_t sctps_timopathmtu; /* Number of PMTU timers that fired */ + uint32_t sctps_timoshutdownack; /* Number of shutdown ack timers that fired */ + uint32_t sctps_timoshutdownguard; /* Number of shutdown guard timers that fired */ + uint32_t sctps_timostrmrst; /* Number of stream reset timers that fired */ + uint32_t sctps_timoearlyfr; /* Number of early FR timers that fired */ + uint32_t sctps_timoasconf; /* Number of times an asconf timer fired */ + uint32_t sctps_timodelprim; /* Number of times a prim_deleted timer fired */ + uint32_t sctps_timoautoclose; /* Number of times auto close timer fired */ + uint32_t sctps_timoassockill; /* Number of asoc free timers expired */ + uint32_t sctps_timoinpkill; /* Number of inp free timers expired */ + /* former early FR counters */ + uint32_t sctps_spare[11]; + /* others */ + uint32_t sctps_hdrops; /* packet shorter than header */ + uint32_t sctps_badsum; /* checksum error */ + uint32_t sctps_noport; /* no endpoint for port */ + uint32_t sctps_badvtag; /* bad v-tag */ + uint32_t sctps_badsid; /* bad SID */ + uint32_t sctps_nomem; /* no memory */ + uint32_t sctps_fastretransinrtt; /* number of multiple FR in a RTT window */ + uint32_t sctps_markedretrans; + uint32_t sctps_naglesent; /* nagle allowed sending */ + uint32_t sctps_naglequeued; /* nagle doesn't allow sending */ + uint32_t sctps_maxburstqueued; /* max burst doesn't allow sending */ + uint32_t sctps_ifnomemqueued; /* look ahead tells us no memory in + * interface ring buffer OR we had a + * send error and are queuing one send. + */ + uint32_t sctps_windowprobed; /* total number of window probes sent */ + uint32_t sctps_lowlevelerr; /* total times an output error causes us + * to clamp down on next user send. + */ + uint32_t sctps_lowlevelerrusr; /* total times sctp_senderrors were caused from + * a user send from a user invoked send not + * a sack response + */ + uint32_t sctps_datadropchklmt; /* Number of in data drops due to chunk limit reached */ + uint32_t sctps_datadroprwnd; /* Number of in data drops due to rwnd limit reached */ + uint32_t sctps_ecnereducedcwnd; /* Number of times a ECN reduced the cwnd */ + uint32_t sctps_vtagexpress; /* Used express lookup via vtag */ + uint32_t sctps_vtagbogus; /* Collision in express lookup. */ + uint32_t sctps_primary_randry; /* Number of times the sender ran dry of user data on primary */ + uint32_t sctps_cmt_randry; /* Same for above */ + uint32_t sctps_slowpath_sack; /* Sacks the slow way */ + uint32_t sctps_wu_sacks_sent; /* Window Update only sacks sent */ + uint32_t sctps_sends_with_flags; /* number of sends with sinfo_flags !=0 */ + uint32_t sctps_sends_with_unord; /* number of unordered sends */ + uint32_t sctps_sends_with_eof; /* number of sends with EOF flag set */ + uint32_t sctps_sends_with_abort; /* number of sends with ABORT flag set */ + uint32_t sctps_protocol_drain_calls;/* number of times protocol drain called */ + uint32_t sctps_protocol_drains_done;/* number of times we did a protocol drain */ + uint32_t sctps_read_peeks; /* Number of times recv was called with peek */ + uint32_t sctps_cached_chk; /* Number of cached chunks used */ + uint32_t sctps_cached_strmoq; /* Number of cached stream oq's used */ + uint32_t sctps_left_abandon; /* Number of unread messages abandoned by close */ + uint32_t sctps_send_burst_avoid; /* Unused */ + uint32_t sctps_send_cwnd_avoid; /* Send cwnd full avoidance, already max burst inflight to net */ + uint32_t sctps_fwdtsn_map_over; /* number of map array over-runs via fwd-tsn's */ + uint32_t sctps_queue_upd_ecne; /* Number of times we queued or updated an ECN chunk on send queue */ + uint32_t sctps_recvzerocrc; /* Number of accepted packets with zero CRC */ + uint32_t sctps_sendzerocrc; /* Number of packets sent with zero CRC */ + uint32_t sctps_reserved[29]; /* Future ABI compat - remove int's from here when adding new */ +}; + +void +usrsctp_get_stat(struct sctpstat *); + +#ifdef _WIN32 +#ifdef _MSC_VER +#pragma warning(default: 4200) +#endif +#endif +#ifdef __cplusplus +} +#endif +#endif diff --git a/netwerk/sctp/src/win32-free.patch b/netwerk/sctp/src/win32-free.patch new file mode 100644 index 0000000000..566b774b46 --- /dev/null +++ b/netwerk/sctp/src/win32-free.patch @@ -0,0 +1,55 @@ +diff --git a/netinet/sctp_bsd_addr.c b/netinet/sctp_bsd_addr.c +--- a/netinet/sctp_bsd_addr.c ++++ b/netinet/sctp_bsd_addr.c +@@ -304,14 +304,7 @@ sctp_is_vmware_interface(struct ifnet *ifn) + #endif + + #if defined(_WIN32) && defined(__Userspace__) +-#ifdef MALLOC +-#undef MALLOC +-#define MALLOC(x) HeapAlloc(GetProcessHeap(), 0, (x)) +-#endif +-#ifdef FREE +-#undef FREE +-#define FREE(x) HeapFree(GetProcessHeap(), 0, (x)) +-#endif ++#define SCTP_BSD_FREE(x) HeapFree(GetProcessHeap(), 0, (x)) + static void + sctp_init_ifns_for_vrf(int vrfid) + { +@@ -341,7 +334,7 @@ sctp_init_ifns_for_vrf(int vrfid) + /* Get actual adapter information */ + if ((Err = GetAdaptersAddresses(AF_INET, 0, NULL, pAdapterAddrs, &AdapterAddrsSize)) != ERROR_SUCCESS) { + SCTP_PRINTF("GetAdaptersV4Addresses() failed with error code %d\n", Err); +- FREE(pAdapterAddrs); ++ SCTP_BSD_FREE(pAdapterAddrs); + return; + } + /* Enumerate through each returned adapter and save its information */ +@@ -366,7 +359,7 @@ sctp_init_ifns_for_vrf(int vrfid) + } + } + } +- FREE(pAdapterAddrs); ++ SCTP_BSD_FREE(pAdapterAddrs); + #endif + #ifdef INET6 + AdapterAddrsSize = 0; +@@ -386,7 +379,7 @@ sctp_init_ifns_for_vrf(int vrfid) + /* Get actual adapter information */ + if ((Err = GetAdaptersAddresses(AF_INET6, 0, NULL, pAdapterAddrs, &AdapterAddrsSize)) != ERROR_SUCCESS) { + SCTP_PRINTF("GetAdaptersV6Addresses() failed with error code %d\n", Err); +- FREE(pAdapterAddrs); ++ SCTP_BSD_FREE(pAdapterAddrs); + return; + } + /* Enumerate through each returned adapter and save its information */ +@@ -408,7 +401,7 @@ sctp_init_ifns_for_vrf(int vrfid) + } + } + } +- FREE(pAdapterAddrs); ++ SCTP_BSD_FREE(pAdapterAddrs); + #endif + } + #elif defined(__Userspace__) diff --git a/netwerk/sctp/src/win32-rands.patch b/netwerk/sctp/src/win32-rands.patch new file mode 100644 index 0000000000..a7addb8dec --- /dev/null +++ b/netwerk/sctp/src/win32-rands.patch @@ -0,0 +1,36 @@ +diff --git a/netinet/sctp_asconf.c b/netinet/sctp_asconf.c +--- a//netinet/sctp_asconf.c ++++ b//netinet/sctp_asconf.c +@@ -32,6 +32,14 @@ + * THE POSSIBILITY OF SUCH DAMAGE. + */ + ++#if defined(_WIN32) ++// Needed for unified build so that rand_s is available to all unified ++// sources. ++#if !defined(_CRT_RAND_S) && !defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) ++#define _CRT_RAND_S ++#endif ++#endif ++ + #if defined(__FreeBSD__) && !defined(__Userspace__) + #include <sys/cdefs.h> + __FBSDID("$FreeBSD$"); +diff --git a/user_environment.h b/user_environment.h +--- a/user_environment.h ++++ b/user_environment.h +@@ -30,6 +30,15 @@ + + #ifndef _USER_ENVIRONMENT_H_ + #define _USER_ENVIRONMENT_H_ ++ ++#if defined(_WIN32) ++// Needed for unified build so that rand_s is available to all unified ++// sources. ++#if !defined(_CRT_RAND_S) && !defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) ++#define _CRT_RAND_S ++#endif ++#endif ++ + /* __Userspace__ */ + #include <sys/types.h> |