/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 2 -*- * vim: sw=2 ts=4 et : */ /* 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 "mozilla/ipc/MessageChannel.h" #include #include #include "CrashAnnotations.h" #include "mozilla/Assertions.h" #include "mozilla/CycleCollectedJSContext.h" #include "mozilla/DebugOnly.h" #include "mozilla/Logging.h" #include "mozilla/Mutex.h" #include "mozilla/ScopeExit.h" #include "mozilla/Sprintf.h" #include "mozilla/StaticMutex.h" #include "mozilla/Telemetry.h" #include "mozilla/TimeStamp.h" #include "mozilla/UniquePtrExtensions.h" #include "mozilla/dom/ScriptSettings.h" #include "mozilla/ipc/ProcessChild.h" #include "mozilla/ipc/ProtocolUtils.h" #include "nsAppRunner.h" #include "nsContentUtils.h" #include "nsDataHashtable.h" #include "nsDebug.h" #include "nsExceptionHandler.h" #include "nsIMemoryReporter.h" #include "nsISupportsImpl.h" #include "nsPrintfCString.h" #ifdef OS_WIN # include "mozilla/gfx/Logging.h" #endif #ifdef MOZ_TASK_TRACER # include "GeckoTaskTracer.h" using namespace mozilla::tasktracer; #endif // Undo the damage done by mozzconf.h #undef compress static mozilla::LazyLogModule sLogModule("ipc"); #define IPC_LOG(...) MOZ_LOG(sLogModule, LogLevel::Debug, (__VA_ARGS__)) /* * IPC design: * * There are three kinds of messages: async, sync, and intr. Sync and intr * messages are blocking. * * Terminology: To dispatch a message Foo is to run the RecvFoo code for * it. This is also called "handling" the message. * * Sync and async messages can sometimes "nest" inside other sync messages * (i.e., while waiting for the sync reply, we can dispatch the inner * message). Intr messages cannot nest. The three possible nesting levels are * NOT_NESTED, NESTED_INSIDE_SYNC, and NESTED_INSIDE_CPOW. The intended uses * are: * NOT_NESTED - most messages. * NESTED_INSIDE_SYNC - CPOW-related messages, which are always sync * and can go in either direction. * NESTED_INSIDE_CPOW - messages where we don't want to dispatch * incoming CPOWs while waiting for the response. * These nesting levels are ordered: NOT_NESTED, NESTED_INSIDE_SYNC, * NESTED_INSIDE_CPOW. Async messages cannot be NESTED_INSIDE_SYNC but they can * be NESTED_INSIDE_CPOW. * * To avoid jank, the parent process is not allowed to send NOT_NESTED sync * messages. When a process is waiting for a response to a sync message M0, it * will dispatch an incoming message M if: * 1. M has a higher nesting level than M0, or * 2. if M has the same nesting level as M0 and we're in the child, or * 3. if M has the same nesting level as M0 and it was sent by the other side * while dispatching M0. * The idea is that messages with higher nesting should take precendence. The * purpose of rule 2 is to handle a race where both processes send to each other * simultaneously. In this case, we resolve the race in favor of the parent (so * the child dispatches first). * * Messages satisfy the following properties: * A. When waiting for a response to a sync message, we won't dispatch any * messages of nesting level. * B. Messages of the same nesting level will be dispatched roughly in the * order they were sent. The exception is when the parent and child send * sync messages to each other simulataneously. In this case, the parent's * message is dispatched first. While it is dispatched, the child may send * further nested messages, and these messages may be dispatched before the * child's original message. We can consider ordering to be preserved here * because we pretend that the child's original message wasn't sent until * after the parent's message is finished being dispatched. * * When waiting for a sync message reply, we dispatch an async message only if * it is NESTED_INSIDE_CPOW. Normally NESTED_INSIDE_CPOW async * messages are sent only from the child. However, the parent can send * NESTED_INSIDE_CPOW async messages when it is creating a bridged protocol. * * Intr messages are blocking and can nest, but they don't participate in the * nesting levels. While waiting for an intr response, all incoming messages are * dispatched until a response is received. When two intr messages race with * each other, a similar scheme is used to ensure that one side wins. The * winning side is chosen based on the message type. * * Intr messages differ from sync messages in that, while sending an intr * message, we may dispatch an async message. This causes some additional * complexity. One issue is that replies can be received out of order. It's also * more difficult to determine whether one message is nested inside * another. Consequently, intr handling uses mOutOfTurnReplies and * mRemoteStackDepthGuess, which are not needed for sync messages. */ using namespace mozilla; using namespace mozilla::ipc; using mozilla::MonitorAutoLock; using mozilla::MonitorAutoUnlock; using mozilla::dom::AutoNoJSAPI; #define IPC_ASSERT(_cond, ...) \ do { \ if (!(_cond)) DebugAbort(__FILE__, __LINE__, #_cond, ##__VA_ARGS__); \ } while (0) static MessageChannel* gParentProcessBlocker; namespace mozilla { namespace ipc { static const uint32_t kMinTelemetryMessageSize = 4096; // Note: we round the time we spend to the nearest millisecond. So a min value // of 1 ms actually captures from 500us and above. static const uint32_t kMinTelemetryIPCWriteLatencyMs = 1; // Note: we round the time we spend waiting for a response to the nearest // millisecond. So a min value of 1 ms actually captures from 500us and above. // This is used for both the sending and receiving side telemetry for sync IPC, // (IPC_SYNC_MAIN_LATENCY_MS and IPC_SYNC_RECEIVE_MS). static const uint32_t kMinTelemetrySyncIPCLatencyMs = 1; const int32_t MessageChannel::kNoTimeout = INT32_MIN; // static bool MessageChannel::sIsPumpingMessages = false; enum Direction { IN_MESSAGE, OUT_MESSAGE }; class MessageChannel::InterruptFrame { private: enum Semantics { INTR_SEMS, SYNC_SEMS, ASYNC_SEMS }; public: InterruptFrame(Direction direction, const Message* msg) : mMessageName(msg->name()), mMessageRoutingId(msg->routing_id()), mMesageSemantics(msg->is_interrupt() ? INTR_SEMS : msg->is_sync() ? SYNC_SEMS : ASYNC_SEMS), mDirection(direction), mMoved(false) { MOZ_RELEASE_ASSERT(mMessageName); } InterruptFrame(InterruptFrame&& aOther) { MOZ_RELEASE_ASSERT(aOther.mMessageName); mMessageName = aOther.mMessageName; aOther.mMessageName = nullptr; mMoved = aOther.mMoved; aOther.mMoved = true; mMessageRoutingId = aOther.mMessageRoutingId; mMesageSemantics = aOther.mMesageSemantics; mDirection = aOther.mDirection; } ~InterruptFrame() { MOZ_RELEASE_ASSERT(mMessageName || mMoved); } InterruptFrame& operator=(InterruptFrame&& aOther) { MOZ_RELEASE_ASSERT(&aOther != this); this->~InterruptFrame(); new (this) InterruptFrame(std::move(aOther)); return *this; } bool IsInterruptIncall() const { return INTR_SEMS == mMesageSemantics && IN_MESSAGE == mDirection; } bool IsInterruptOutcall() const { return INTR_SEMS == mMesageSemantics && OUT_MESSAGE == mDirection; } bool IsOutgoingSync() const { return (mMesageSemantics == INTR_SEMS || mMesageSemantics == SYNC_SEMS) && mDirection == OUT_MESSAGE; } void Describe(int32_t* id, const char** dir, const char** sems, const char** name) const { *id = mMessageRoutingId; *dir = (IN_MESSAGE == mDirection) ? "in" : "out"; *sems = (INTR_SEMS == mMesageSemantics) ? "intr" : (SYNC_SEMS == mMesageSemantics) ? "sync" : "async"; *name = mMessageName; } int32_t GetRoutingId() const { return mMessageRoutingId; } private: const char* mMessageName; int32_t mMessageRoutingId; Semantics mMesageSemantics; Direction mDirection; bool mMoved; // Disable harmful methods. InterruptFrame(const InterruptFrame& aOther) = delete; InterruptFrame& operator=(const InterruptFrame&) = delete; }; class MOZ_STACK_CLASS MessageChannel::CxxStackFrame { public: CxxStackFrame(MessageChannel& that, Direction direction, const Message* msg) : mThat(that) { mThat.AssertWorkerThread(); if (mThat.mCxxStackFrames.empty()) mThat.EnteredCxxStack(); if (!mThat.mCxxStackFrames.append(InterruptFrame(direction, msg))) MOZ_CRASH(); const InterruptFrame& frame = mThat.mCxxStackFrames.back(); if (frame.IsInterruptIncall()) mThat.EnteredCall(); if (frame.IsOutgoingSync()) mThat.EnteredSyncSend(); mThat.mSawInterruptOutMsg |= frame.IsInterruptOutcall(); } ~CxxStackFrame() { mThat.AssertWorkerThread(); MOZ_RELEASE_ASSERT(!mThat.mCxxStackFrames.empty()); const InterruptFrame& frame = mThat.mCxxStackFrames.back(); bool exitingSync = frame.IsOutgoingSync(); bool exitingCall = frame.IsInterruptIncall(); mThat.mCxxStackFrames.shrinkBy(1); bool exitingStack = mThat.mCxxStackFrames.empty(); // According how lifetime is declared, mListener on MessageChannel // lives longer than MessageChannel itself. Hence is expected to // be alive. There is nothing to even assert here, there is no place // we would be nullifying mListener on MessageChannel. if (exitingCall) mThat.ExitedCall(); if (exitingSync) mThat.ExitedSyncSend(); if (exitingStack) mThat.ExitedCxxStack(); } private: MessageChannel& mThat; // Disable harmful methods. CxxStackFrame() = delete; CxxStackFrame(const CxxStackFrame&) = delete; CxxStackFrame& operator=(const CxxStackFrame&) = delete; }; class AutoEnterTransaction { public: explicit AutoEnterTransaction(MessageChannel* aChan, int32_t aMsgSeqno, int32_t aTransactionID, int aNestedLevel) : mChan(aChan), mActive(true), mOutgoing(true), mNestedLevel(aNestedLevel), mSeqno(aMsgSeqno), mTransaction(aTransactionID), mNext(mChan->mTransactionStack) { mChan->mMonitor->AssertCurrentThreadOwns(); mChan->mTransactionStack = this; } explicit AutoEnterTransaction(MessageChannel* aChan, const IPC::Message& aMessage) : mChan(aChan), mActive(true), mOutgoing(false), mNestedLevel(aMessage.nested_level()), mSeqno(aMessage.seqno()), mTransaction(aMessage.transaction_id()), mNext(mChan->mTransactionStack) { mChan->mMonitor->AssertCurrentThreadOwns(); if (!aMessage.is_sync()) { mActive = false; return; } mChan->mTransactionStack = this; } ~AutoEnterTransaction() { mChan->mMonitor->AssertCurrentThreadOwns(); if (mActive) { mChan->mTransactionStack = mNext; } } void Cancel() { AutoEnterTransaction* cur = mChan->mTransactionStack; MOZ_RELEASE_ASSERT(cur == this); while (cur && cur->mNestedLevel != IPC::Message::NOT_NESTED) { // Note that, in the following situation, we will cancel multiple // transactions: // 1. Parent sends NESTED_INSIDE_SYNC message P1 to child. // 2. Child sends NESTED_INSIDE_SYNC message C1 to child. // 3. Child dispatches P1, parent blocks. // 4. Child cancels. // In this case, both P1 and C1 are cancelled. The parent will // remove C1 from its queue when it gets the cancellation message. MOZ_RELEASE_ASSERT(cur->mActive); cur->mActive = false; cur = cur->mNext; } mChan->mTransactionStack = cur; MOZ_RELEASE_ASSERT(IsComplete()); } bool AwaitingSyncReply() const { MOZ_RELEASE_ASSERT(mActive); if (mOutgoing) { return true; } return mNext ? mNext->AwaitingSyncReply() : false; } int AwaitingSyncReplyNestedLevel() const { MOZ_RELEASE_ASSERT(mActive); if (mOutgoing) { return mNestedLevel; } return mNext ? mNext->AwaitingSyncReplyNestedLevel() : 0; } bool DispatchingSyncMessage() const { MOZ_RELEASE_ASSERT(mActive); if (!mOutgoing) { return true; } return mNext ? mNext->DispatchingSyncMessage() : false; } int DispatchingSyncMessageNestedLevel() const { MOZ_RELEASE_ASSERT(mActive); if (!mOutgoing) { return mNestedLevel; } return mNext ? mNext->DispatchingSyncMessageNestedLevel() : 0; } int NestedLevel() const { MOZ_RELEASE_ASSERT(mActive); return mNestedLevel; } int32_t SequenceNumber() const { MOZ_RELEASE_ASSERT(mActive); return mSeqno; } int32_t TransactionID() const { MOZ_RELEASE_ASSERT(mActive); return mTransaction; } void ReceivedReply(IPC::Message&& aMessage) { MOZ_RELEASE_ASSERT(aMessage.seqno() == mSeqno); MOZ_RELEASE_ASSERT(aMessage.transaction_id() == mTransaction); MOZ_RELEASE_ASSERT(!mReply); IPC_LOG("Reply received on worker thread: seqno=%d", mSeqno); mReply = MakeUnique(std::move(aMessage)); MOZ_RELEASE_ASSERT(IsComplete()); } void HandleReply(IPC::Message&& aMessage) { AutoEnterTransaction* cur = mChan->mTransactionStack; MOZ_RELEASE_ASSERT(cur == this); while (cur) { MOZ_RELEASE_ASSERT(cur->mActive); if (aMessage.seqno() == cur->mSeqno) { cur->ReceivedReply(std::move(aMessage)); break; } cur = cur->mNext; MOZ_RELEASE_ASSERT(cur); } } bool IsComplete() { return !mActive || mReply; } bool IsOutgoing() { return mOutgoing; } bool IsCanceled() { return !mActive; } bool IsBottom() const { return !mNext; } bool IsError() { MOZ_RELEASE_ASSERT(mReply); return mReply->is_reply_error(); } UniquePtr GetReply() { return std::move(mReply); } private: MessageChannel* mChan; // Active is true if this transaction is on the mChan->mTransactionStack // stack. Generally we're not on the stack if the transaction was canceled // or if it was for a message that doesn't require transactions (an async // message). bool mActive; // Is this stack frame for an outgoing message? bool mOutgoing; // Properties of the message being sent/received. int mNestedLevel; int32_t mSeqno; int32_t mTransaction; // Next item in mChan->mTransactionStack. AutoEnterTransaction* mNext; // Pointer the a reply received for this message, if one was received. UniquePtr mReply; }; class PendingResponseReporter final : public nsIMemoryReporter { ~PendingResponseReporter() = default; public: NS_DECL_THREADSAFE_ISUPPORTS NS_IMETHOD CollectReports(nsIHandleReportCallback* aHandleReport, nsISupports* aData, bool aAnonymize) override { MOZ_COLLECT_REPORT( "unresolved-ipc-responses", KIND_OTHER, UNITS_COUNT, MessageChannel::gUnresolvedResponses, "Outstanding IPC async message responses that are still not resolved."); return NS_OK; } }; NS_IMPL_ISUPPORTS(PendingResponseReporter, nsIMemoryReporter) class ChannelCountReporter final : public nsIMemoryReporter { ~ChannelCountReporter() = default; struct ChannelCounts { size_t mNow; size_t mMax; ChannelCounts() : mNow(0), mMax(0) {} void Inc() { ++mNow; if (mMax < mNow) { mMax = mNow; } } void Dec() { MOZ_ASSERT(mNow > 0); --mNow; } }; using CountTable = nsDataHashtable; static StaticMutex sChannelCountMutex; static CountTable* sChannelCounts; public: NS_DECL_THREADSAFE_ISUPPORTS NS_IMETHOD CollectReports(nsIHandleReportCallback* aHandleReport, nsISupports* aData, bool aAnonymize) override { StaticMutexAutoLock countLock(sChannelCountMutex); if (!sChannelCounts) { return NS_OK; } for (auto iter = sChannelCounts->Iter(); !iter.Done(); iter.Next()) { nsPrintfCString pathNow("ipc-channels/%s", iter.Key()); nsPrintfCString pathMax("ipc-channels-peak/%s", iter.Key()); nsPrintfCString descNow( "Number of IPC channels for" " top-level actor type %s", iter.Key()); nsPrintfCString descMax( "Peak number of IPC channels for" " top-level actor type %s", iter.Key()); aHandleReport->Callback(""_ns, pathNow, KIND_OTHER, UNITS_COUNT, iter.Data().mNow, descNow, aData); aHandleReport->Callback(""_ns, pathMax, KIND_OTHER, UNITS_COUNT, iter.Data().mMax, descMax, aData); } return NS_OK; } static void Increment(const char* aName) { StaticMutexAutoLock countLock(sChannelCountMutex); if (!sChannelCounts) { sChannelCounts = new CountTable; } sChannelCounts->GetOrInsert(aName).Inc(); } static void Decrement(const char* aName) { StaticMutexAutoLock countLock(sChannelCountMutex); MOZ_ASSERT(sChannelCounts); sChannelCounts->GetOrInsert(aName).Dec(); } }; StaticMutex ChannelCountReporter::sChannelCountMutex; ChannelCountReporter::CountTable* ChannelCountReporter::sChannelCounts; NS_IMPL_ISUPPORTS(ChannelCountReporter, nsIMemoryReporter) // In child processes, the first MessageChannel is created before // XPCOM is initialized enough to construct the memory reporter // manager. This retries every time a MessageChannel is constructed, // which is good enough in practice. template static void TryRegisterStrongMemoryReporter() { static Atomic registered; if (registered.compareExchange(false, true)) { RefPtr reporter = new Reporter(); if (NS_FAILED(RegisterStrongMemoryReporter(reporter))) { registered = false; } } } Atomic MessageChannel::gUnresolvedResponses; MessageChannel::MessageChannel(const char* aName, IToplevelProtocol* aListener) : mName(aName), mListener(aListener), mChannelState(ChannelClosed), mSide(UnknownSide), mIsCrossProcess(false), mChannelErrorTask(nullptr), mTimeoutMs(kNoTimeout), mInTimeoutSecondHalf(false), mNextSeqno(0), mLastSendError(SyncSendError::SendSuccess), mDispatchingAsyncMessage(false), mDispatchingAsyncMessageNestedLevel(0), mTransactionStack(nullptr), mTimedOutMessageSeqno(0), mTimedOutMessageNestedLevel(0), mMaybeDeferredPendingCount(0), mRemoteStackDepthGuess(0), mSawInterruptOutMsg(false), mIsWaitingForIncoming(false), mAbortOnError(false), mNotifiedChannelDone(false), mFlags(REQUIRE_DEFAULT), mPeerPidSet(false), mPeerPid(-1), mIsPostponingSends(false), mBuildIDsConfirmedMatch(false), mIsSameThreadChannel(false) { MOZ_COUNT_CTOR(ipc::MessageChannel); #ifdef OS_WIN mTopFrame = nullptr; mIsSyncWaitingOnNonMainThread = false; #endif mOnChannelConnectedTask = NewNonOwningCancelableRunnableMethod( "ipc::MessageChannel::DispatchOnChannelConnected", this, &MessageChannel::DispatchOnChannelConnected); #ifdef OS_WIN mEvent = CreateEventW(nullptr, TRUE, FALSE, nullptr); MOZ_RELEASE_ASSERT(mEvent, "CreateEvent failed! Nothing is going to work!"); #endif TryRegisterStrongMemoryReporter(); TryRegisterStrongMemoryReporter(); } MessageChannel::~MessageChannel() { MOZ_COUNT_DTOR(ipc::MessageChannel); IPC_ASSERT(mCxxStackFrames.empty(), "mismatched CxxStackFrame ctor/dtors"); #ifdef OS_WIN if (mEvent) { BOOL ok = CloseHandle(mEvent); mEvent = nullptr; if (!ok) { gfxDevCrash(mozilla::gfx::LogReason::MessageChannelCloseFailure) << "MessageChannel failed to close. GetLastError: " << GetLastError(); } MOZ_RELEASE_ASSERT(ok); } else { gfxDevCrash(mozilla::gfx::LogReason::MessageChannelCloseFailure) << "MessageChannel destructor ran without an mEvent Handle"; } #endif Clear(); } #ifdef DEBUG void MessageChannel::AssertMaybeDeferredCountCorrect() { size_t count = 0; for (MessageTask* task : mPending) { if (!IsAlwaysDeferred(task->Msg())) { count++; } } MOZ_ASSERT(count == mMaybeDeferredPendingCount); } #endif // This function returns the current transaction ID. Since the notion of a // "current transaction" can be hard to define when messages race with each // other and one gets canceled and the other doesn't, we require that this // function is only called when the current transaction is known to be for a // NESTED_INSIDE_SYNC message. In that case, we know for sure what the caller is // looking for. int32_t MessageChannel::CurrentNestedInsideSyncTransaction() const { mMonitor->AssertCurrentThreadOwns(); if (!mTransactionStack) { return 0; } MOZ_RELEASE_ASSERT(mTransactionStack->NestedLevel() == IPC::Message::NESTED_INSIDE_SYNC); return mTransactionStack->TransactionID(); } bool MessageChannel::AwaitingSyncReply() const { mMonitor->AssertCurrentThreadOwns(); return mTransactionStack ? mTransactionStack->AwaitingSyncReply() : false; } int MessageChannel::AwaitingSyncReplyNestedLevel() const { mMonitor->AssertCurrentThreadOwns(); return mTransactionStack ? mTransactionStack->AwaitingSyncReplyNestedLevel() : 0; } bool MessageChannel::DispatchingSyncMessage() const { mMonitor->AssertCurrentThreadOwns(); return mTransactionStack ? mTransactionStack->DispatchingSyncMessage() : false; } int MessageChannel::DispatchingSyncMessageNestedLevel() const { mMonitor->AssertCurrentThreadOwns(); return mTransactionStack ? mTransactionStack->DispatchingSyncMessageNestedLevel() : 0; } static void PrintErrorMessage(Side side, const char* channelName, const char* msg) { const char* from = (side == ChildSide) ? "Child" : ((side == ParentSide) ? "Parent" : "Unknown"); printf_stderr("\n###!!! [%s][%s] Error: %s\n\n", from, channelName, msg); } bool MessageChannel::Connected() const { mMonitor->AssertCurrentThreadOwns(); // The transport layer allows us to send messages before // receiving the "connected" ack from the remote side. return (ChannelOpening == mChannelState || ChannelConnected == mChannelState); } bool MessageChannel::CanSend() const { if (!mMonitor) { return false; } MonitorAutoLock lock(*mMonitor); return Connected(); } void MessageChannel::Clear() { // Don't clear mWorkerThread; we use it in AssertLinkThread() and // AssertWorkerThread(). // // Also don't clear mListener. If we clear it, then sending a message // through this channel after it's Clear()'ed can cause this process to // crash. // // In practice, mListener owns the channel, so the channel gets deleted // before mListener. But just to be safe, mListener is a weak pointer. #if !defined(ANDROID) if (!Unsound_IsClosed()) { CrashReporter::AnnotateCrashReport( CrashReporter::Annotation::IPCFatalErrorProtocol, nsDependentCString(mName)); switch (mChannelState) { case ChannelOpening: MOZ_CRASH( "MessageChannel destroyed without being closed " "(mChannelState == ChannelOpening)."); break; case ChannelConnected: MOZ_CRASH( "MessageChannel destroyed without being closed " "(mChannelState == ChannelConnected)."); break; case ChannelTimeout: MOZ_CRASH( "MessageChannel destroyed without being closed " "(mChannelState == ChannelTimeout)."); break; case ChannelClosing: MOZ_CRASH( "MessageChannel destroyed without being closed " "(mChannelState == ChannelClosing)."); break; case ChannelError: MOZ_CRASH( "MessageChannel destroyed without being closed " "(mChannelState == ChannelError)."); break; default: MOZ_CRASH("MessageChannel destroyed without being closed."); } } #endif if (gParentProcessBlocker == this) { gParentProcessBlocker = nullptr; } gUnresolvedResponses -= mPendingResponses.size(); for (auto& pair : mPendingResponses) { pair.second.get()->Reject(ResponseRejectReason::ChannelClosed); } mPendingResponses.clear(); if (mLink != nullptr && mIsCrossProcess) { ChannelCountReporter::Decrement(mName); } if (mLink) { mLink->PrepareToDestroy(); mLink = nullptr; } mOnChannelConnectedTask->Cancel(); if (mChannelErrorTask) { mChannelErrorTask->Cancel(); mChannelErrorTask = nullptr; } // Free up any memory used by pending messages. for (MessageTask* task : mPending) { task->Clear(); } mPending.clear(); mMaybeDeferredPendingCount = 0; mOutOfTurnReplies.clear(); while (!mDeferred.empty()) { mDeferred.pop(); } } bool MessageChannel::Open(mozilla::UniquePtr aTransport, MessageLoop* aIOLoop, Side aSide) { MOZ_ASSERT(!mLink, "Open() called > once"); mMonitor = new RefCountedMonitor(); mWorkerThread = GetCurrentSerialEventTarget(); MOZ_ASSERT(mWorkerThread, "We should always be on a nsISerialEventTarget"); mListener->OnIPCChannelOpened(); auto link = MakeUnique(this); link->Open(std::move(aTransport), aIOLoop, aSide); // :TODO: n.b.: sets mChild mLink = std::move(link); mIsCrossProcess = true; ChannelCountReporter::Increment(mName); return true; } bool MessageChannel::Open(MessageChannel* aTargetChan, nsISerialEventTarget* aEventTarget, Side aSide) { // Opens a connection to another thread in the same process. // This handshake proceeds as follows: // - Let A be the thread initiating the process (either child or parent) // and B be the other thread. // - A spawns thread for B, obtaining B's message loop // - A creates ProtocolChild and ProtocolParent instances. // Let PA be the one appropriate to A and PB the side for B. // - A invokes PA->Open(PB, ...): // - set state to mChannelOpening // - this will place a work item in B's worker loop (see next bullet) // and then spins until PB->mChannelState becomes mChannelConnected // - meanwhile, on PB's worker loop, the work item is removed and: // - invokes PB->OpenAsOtherThread(PA, ...): // - sets its state and that of PA to Connected MOZ_ASSERT(aTargetChan, "Need a target channel"); MOZ_ASSERT(ChannelClosed == mChannelState, "Not currently closed"); CommonThreadOpenInit(aTargetChan, GetCurrentSerialEventTarget(), aSide); Side oppSide = UnknownSide; switch (aSide) { case ChildSide: oppSide = ParentSide; break; case ParentSide: oppSide = ChildSide; break; case UnknownSide: break; } mMonitor = new RefCountedMonitor(); MonitorAutoLock lock(*mMonitor); mChannelState = ChannelOpening; MOZ_ALWAYS_SUCCEEDS(aEventTarget->Dispatch( NewNonOwningRunnableMethod( "ipc::MessageChannel::OpenAsOtherThread", aTargetChan, &MessageChannel::OpenAsOtherThread, this, aEventTarget, oppSide))); while (ChannelOpening == mChannelState) mMonitor->Wait(); MOZ_RELEASE_ASSERT(ChannelConnected == mChannelState, "not connected when awoken"); return (ChannelConnected == mChannelState); } void MessageChannel::OpenAsOtherThread(MessageChannel* aTargetChan, nsISerialEventTarget* aThread, Side aSide) { // Invoked when the other side has begun the open. MOZ_ASSERT(ChannelClosed == mChannelState, "Not currently closed"); MOZ_ASSERT(ChannelOpening == aTargetChan->mChannelState, "Target channel not in the process of opening"); CommonThreadOpenInit(aTargetChan, aThread, aSide); mMonitor = aTargetChan->mMonitor; MonitorAutoLock lock(*mMonitor); MOZ_RELEASE_ASSERT(ChannelOpening == aTargetChan->mChannelState, "Target channel not in the process of opening"); mChannelState = ChannelConnected; aTargetChan->mChannelState = ChannelConnected; aTargetChan->mMonitor->Notify(); } void MessageChannel::CommonThreadOpenInit(MessageChannel* aTargetChan, nsISerialEventTarget* aThread, Side aSide) { MOZ_ASSERT(aThread); mWorkerThread = aThread; mListener->OnIPCChannelOpened(); mLink = MakeUnique(this, aTargetChan); mSide = aSide; } bool MessageChannel::OpenOnSameThread(MessageChannel* aTargetChan, mozilla::ipc::Side aSide) { nsCOMPtr currentThread = GetCurrentSerialEventTarget(); CommonThreadOpenInit(aTargetChan, currentThread, aSide); Side oppSide = UnknownSide; switch (aSide) { case ChildSide: oppSide = ParentSide; break; case ParentSide: oppSide = ChildSide; break; case UnknownSide: break; } mIsSameThreadChannel = true; // XXX(nika): Avoid setting up a monitor for same thread channels? We // shouldn't need it. mMonitor = new RefCountedMonitor(); mChannelState = ChannelOpening; aTargetChan->CommonThreadOpenInit(this, currentThread, oppSide); aTargetChan->mIsSameThreadChannel = true; aTargetChan->mMonitor = mMonitor; mChannelState = ChannelConnected; aTargetChan->mChannelState = ChannelConnected; return true; } bool MessageChannel::Send(UniquePtr aMsg) { if (aMsg->size() >= kMinTelemetryMessageSize) { Telemetry::Accumulate(Telemetry::IPC_MESSAGE_SIZE2, aMsg->size()); } // If the message was created by the IPC bindings, the create time will be // recorded. Use this information to report the // IPC_WRITE_MAIN_THREAD_LATENCY_MS (time from message creation to it being // sent). if (NS_IsMainThread() && aMsg->create_time()) { uint32_t latencyMs = round( (mozilla::TimeStamp::Now() - aMsg->create_time()).ToMilliseconds()); if (latencyMs >= kMinTelemetryIPCWriteLatencyMs) { mozilla::Telemetry::Accumulate( mozilla::Telemetry::IPC_WRITE_MAIN_THREAD_LATENCY_MS, nsDependentCString(aMsg->name()), latencyMs); } } MOZ_RELEASE_ASSERT(!aMsg->is_sync()); MOZ_RELEASE_ASSERT(aMsg->nested_level() != IPC::Message::NESTED_INSIDE_SYNC); CxxStackFrame frame(*this, OUT_MESSAGE, aMsg.get()); AssertWorkerThread(); mMonitor->AssertNotCurrentThreadOwns(); if (MSG_ROUTING_NONE == aMsg->routing_id()) { ReportMessageRouteError("MessageChannel::Send"); return false; } if (aMsg->seqno() == 0) { aMsg->set_seqno(NextSeqno()); } MonitorAutoLock lock(*mMonitor); if (!Connected()) { ReportConnectionError("MessageChannel", aMsg.get()); return false; } AddProfilerMarker(*aMsg, MessageDirection::eSending); SendMessageToLink(std::move(aMsg)); return true; } void MessageChannel::SendMessageToLink(UniquePtr aMsg) { if (mIsPostponingSends) { mPostponedSends.push_back(std::move(aMsg)); return; } mLink->SendMessage(std::move(aMsg)); } void MessageChannel::BeginPostponingSends() { AssertWorkerThread(); mMonitor->AssertNotCurrentThreadOwns(); MonitorAutoLock lock(*mMonitor); { MOZ_ASSERT(!mIsPostponingSends); mIsPostponingSends = true; } } void MessageChannel::StopPostponingSends() { // Note: this can be called from any thread. MonitorAutoLock lock(*mMonitor); MOZ_ASSERT(mIsPostponingSends); for (UniquePtr& iter : mPostponedSends) { mLink->SendMessage(std::move(iter)); } // We unset this after SendMessage so we can make correct thread // assertions in MessageLink. mIsPostponingSends = false; mPostponedSends.clear(); } UniquePtr MessageChannel::PopCallback( const Message& aMsg) { auto iter = mPendingResponses.find(aMsg.seqno()); if (iter != mPendingResponses.end()) { UniquePtr ret = std::move(iter->second); mPendingResponses.erase(iter); gUnresolvedResponses--; return ret; } return nullptr; } void MessageChannel::RejectPendingResponsesForActor(ActorIdType aActorId) { auto itr = mPendingResponses.begin(); while (itr != mPendingResponses.end()) { if (itr->second.get()->mActorId != aActorId) { ++itr; continue; } itr->second.get()->Reject(ResponseRejectReason::ActorDestroyed); // Take special care of advancing the iterator since we are // removing it while iterating. itr = mPendingResponses.erase(itr); gUnresolvedResponses--; } } class BuildIDsMatchMessage : public IPC::Message { public: BuildIDsMatchMessage() : IPC::Message(MSG_ROUTING_NONE, BUILD_IDS_MATCH_MESSAGE_TYPE) {} void Log(const std::string& aPrefix, FILE* aOutf) const { fputs("(special `Build IDs match' message)", aOutf); } }; // Send the parent a special async message to confirm when the parent and child // are of the same buildID. Skips sending the message and returns false if the // buildIDs don't match. This is a minor variation on // MessageChannel::Send(Message* aMsg). bool MessageChannel::SendBuildIDsMatchMessage(const char* aParentBuildID) { MOZ_ASSERT(!XRE_IsParentProcess()); nsCString parentBuildID(aParentBuildID); nsCString childBuildID(mozilla::PlatformBuildID()); if (parentBuildID != childBuildID) { // The build IDs didn't match, usually because an update occurred in the // background. return false; } auto msg = MakeUnique(); MOZ_RELEASE_ASSERT(!msg->is_sync()); MOZ_RELEASE_ASSERT(msg->nested_level() != IPC::Message::NESTED_INSIDE_SYNC); AssertWorkerThread(); mMonitor->AssertNotCurrentThreadOwns(); // Don't check for MSG_ROUTING_NONE. MonitorAutoLock lock(*mMonitor); if (!Connected()) { ReportConnectionError("MessageChannel", msg.get()); return false; } mLink->SendMessage(std::move(msg)); return true; } class CancelMessage : public IPC::Message { public: explicit CancelMessage(int transaction) : IPC::Message(MSG_ROUTING_NONE, CANCEL_MESSAGE_TYPE) { set_transaction_id(transaction); } static bool Read(const Message* msg) { return true; } void Log(const std::string& aPrefix, FILE* aOutf) const { fputs("(special `Cancel' message)", aOutf); } }; bool MessageChannel::MaybeInterceptSpecialIOMessage(const Message& aMsg) { AssertLinkThread(); mMonitor->AssertCurrentThreadOwns(); if (MSG_ROUTING_NONE == aMsg.routing_id()) { if (GOODBYE_MESSAGE_TYPE == aMsg.type()) { // :TODO: Sort out Close() on this side racing with Close() on the // other side mChannelState = ChannelClosing; if (LoggingEnabled()) { printf("NOTE: %s process received `Goodbye', closing down\n", (mSide == ChildSide) ? "child" : "parent"); } return true; } else if (CANCEL_MESSAGE_TYPE == aMsg.type()) { IPC_LOG("Cancel from message"); CancelTransaction(aMsg.transaction_id()); NotifyWorkerThread(); return true; } else if (BUILD_IDS_MATCH_MESSAGE_TYPE == aMsg.type()) { IPC_LOG("Build IDs match message"); mBuildIDsConfirmedMatch = true; return true; } else if (IMPENDING_SHUTDOWN_MESSAGE_TYPE == aMsg.type()) { IPC_LOG("Impending Shutdown received"); ProcessChild::NotifyImpendingShutdown(); return true; } } return false; } /* static */ bool MessageChannel::IsAlwaysDeferred(const Message& aMsg) { // If a message is not NESTED_INSIDE_CPOW and not sync, then we always defer // it. return aMsg.nested_level() != IPC::Message::NESTED_INSIDE_CPOW && !aMsg.is_sync(); } bool MessageChannel::ShouldDeferMessage(const Message& aMsg) { // Never defer messages that have the highest nested level, even async // ones. This is safe because only the child can send these messages, so // they can never nest. if (aMsg.nested_level() == IPC::Message::NESTED_INSIDE_CPOW) { MOZ_ASSERT(!IsAlwaysDeferred(aMsg)); return false; } // Unless they're NESTED_INSIDE_CPOW, we always defer async messages. // Note that we never send an async NESTED_INSIDE_SYNC message. if (!aMsg.is_sync()) { MOZ_RELEASE_ASSERT(aMsg.nested_level() == IPC::Message::NOT_NESTED); MOZ_ASSERT(IsAlwaysDeferred(aMsg)); return true; } MOZ_ASSERT(!IsAlwaysDeferred(aMsg)); int msgNestedLevel = aMsg.nested_level(); int waitingNestedLevel = AwaitingSyncReplyNestedLevel(); // Always defer if the nested level of the incoming message is less than the // nested level of the message we're awaiting. if (msgNestedLevel < waitingNestedLevel) return true; // Never defer if the message has strictly greater nested level. if (msgNestedLevel > waitingNestedLevel) return false; // When both sides send sync messages of the same nested level, we resolve the // race by dispatching in the child and deferring the incoming message in // the parent. However, the parent still needs to dispatch nested sync // messages. // // Deferring in the parent only sort of breaks message ordering. When the // child's message comes in, we can pretend the child hasn't quite // finished sending it yet. Since the message is sync, we know that the // child hasn't moved on yet. return mSide == ParentSide && aMsg.transaction_id() != CurrentNestedInsideSyncTransaction(); } void MessageChannel::OnMessageReceivedFromLink(Message&& aMsg) { AssertLinkThread(); mMonitor->AssertCurrentThreadOwns(); if (MaybeInterceptSpecialIOMessage(aMsg)) return; mListener->OnChannelReceivedMessage(aMsg); // Regardless of the Interrupt stack, if we're awaiting a sync reply, // we know that it needs to be immediately handled to unblock us. if (aMsg.is_sync() && aMsg.is_reply()) { IPC_LOG("Received reply seqno=%d xid=%d", aMsg.seqno(), aMsg.transaction_id()); if (aMsg.seqno() == mTimedOutMessageSeqno) { // Drop the message, but allow future sync messages to be sent. IPC_LOG("Received reply to timedout message; igoring; xid=%d", mTimedOutMessageSeqno); EndTimeout(); return; } MOZ_RELEASE_ASSERT(AwaitingSyncReply()); MOZ_RELEASE_ASSERT(!mTimedOutMessageSeqno); mTransactionStack->HandleReply(std::move(aMsg)); NotifyWorkerThread(); return; } // Nested messages cannot be compressed. MOZ_RELEASE_ASSERT(aMsg.compress_type() == IPC::Message::COMPRESSION_NONE || aMsg.nested_level() == IPC::Message::NOT_NESTED); bool reuseTask = false; if (aMsg.compress_type() == IPC::Message::COMPRESSION_ENABLED) { bool compress = (!mPending.isEmpty() && mPending.getLast()->Msg().type() == aMsg.type() && mPending.getLast()->Msg().routing_id() == aMsg.routing_id()); if (compress) { // This message type has compression enabled, and the back of the // queue was the same message type and routed to the same destination. // Replace it with the newer message. MOZ_RELEASE_ASSERT(mPending.getLast()->Msg().compress_type() == IPC::Message::COMPRESSION_ENABLED); mPending.getLast()->Msg() = std::move(aMsg); reuseTask = true; } } else if (aMsg.compress_type() == IPC::Message::COMPRESSION_ALL && !mPending.isEmpty()) { for (MessageTask* p = mPending.getLast(); p; p = p->getPrevious()) { if (p->Msg().type() == aMsg.type() && p->Msg().routing_id() == aMsg.routing_id()) { // This message type has compression enabled, and the queue // holds a message with the same message type and routed to the // same destination. Erase it. Note that, since we always // compress these redundancies, There Can Be Only One. MOZ_RELEASE_ASSERT(p->Msg().compress_type() == IPC::Message::COMPRESSION_ALL); MOZ_RELEASE_ASSERT(IsAlwaysDeferred(p->Msg())); p->remove(); break; } } } bool alwaysDeferred = IsAlwaysDeferred(aMsg); bool wakeUpSyncSend = AwaitingSyncReply() && !ShouldDeferMessage(aMsg); bool shouldWakeUp = AwaitingInterruptReply() || wakeUpSyncSend || AwaitingIncomingMessage(); // Although we usually don't need to post a message task if // shouldWakeUp is true, it's easier to post anyway than to have to // guarantee that every Send call processes everything it's supposed to // before returning. bool shouldPostTask = !shouldWakeUp || wakeUpSyncSend; IPC_LOG("Receive on link thread; seqno=%d, xid=%d, shouldWakeUp=%d", aMsg.seqno(), aMsg.transaction_id(), shouldWakeUp); if (reuseTask) { return; } // There are three cases we're concerned about, relating to the state of the // main thread: // // (1) We are waiting on a sync reply - main thread is blocked on the // IPC monitor. // - If the message is NESTED_INSIDE_SYNC, we wake up the main thread to // deliver the message depending on ShouldDeferMessage. Otherwise, we // leave it in the mPending queue, posting a task to the main event // loop, where it will be processed once the synchronous reply has been // received. // // (2) We are waiting on an Interrupt reply - main thread is blocked on the // IPC monitor. // - Always notify and wake up the main thread. // // (3) We are not waiting on a reply. // - We post a task to the main event loop. // // Note that, we may notify the main thread even though the monitor is not // blocked. This is okay, since we always check for pending events before // blocking again. #ifdef MOZ_TASK_TRACER aMsg.TaskTracerDispatch(); #endif RefPtr task = new MessageTask(this, std::move(aMsg)); mPending.insertBack(task); if (!alwaysDeferred) { mMaybeDeferredPendingCount++; } if (shouldWakeUp) { NotifyWorkerThread(); } if (shouldPostTask) { task->Post(); } } void MessageChannel::PeekMessages( const std::function& aInvoke) { // FIXME: We shouldn't be holding the lock for aInvoke! MonitorAutoLock lock(*mMonitor); for (MessageTask* it : mPending) { const Message& msg = it->Msg(); if (!aInvoke(msg)) { break; } } } void MessageChannel::ProcessPendingRequests( AutoEnterTransaction& aTransaction) { mMonitor->AssertCurrentThreadOwns(); AssertMaybeDeferredCountCorrect(); if (mMaybeDeferredPendingCount == 0) { return; } IPC_LOG("ProcessPendingRequests for seqno=%d, xid=%d", aTransaction.SequenceNumber(), aTransaction.TransactionID()); // Loop until there aren't any more nested messages to process. for (;;) { // If we canceled during ProcessPendingRequest, then we need to leave // immediately because the results of ShouldDeferMessage will be // operating with weird state (as if no Send is in progress). That could // cause even NOT_NESTED sync messages to be processed (but not // NOT_NESTED async messages), which would break message ordering. if (aTransaction.IsCanceled()) { return; } mozilla::Vector toProcess; for (MessageTask* p = mPending.getFirst(); p;) { Message& msg = p->Msg(); MOZ_RELEASE_ASSERT(!aTransaction.IsCanceled(), "Calling ShouldDeferMessage when cancelled"); bool defer = ShouldDeferMessage(msg); // Only log the interesting messages. if (msg.is_sync() || msg.nested_level() == IPC::Message::NESTED_INSIDE_CPOW) { IPC_LOG("ShouldDeferMessage(seqno=%d) = %d", msg.seqno(), defer); } if (!defer) { MOZ_ASSERT(!IsAlwaysDeferred(msg)); if (!toProcess.append(std::move(msg))) MOZ_CRASH(); mMaybeDeferredPendingCount--; p = p->removeAndGetNext(); continue; } p = p->getNext(); } if (toProcess.empty()) { break; } // Processing these messages could result in more messages, so we // loop around to check for more afterwards. for (auto it = toProcess.begin(); it != toProcess.end(); it++) { ProcessPendingRequest(std::move(*it)); } } AssertMaybeDeferredCountCorrect(); } bool MessageChannel::Send(UniquePtr aMsg, Message* aReply) { mozilla::TimeStamp start = TimeStamp::Now(); if (aMsg->size() >= kMinTelemetryMessageSize) { Telemetry::Accumulate(Telemetry::IPC_MESSAGE_SIZE2, aMsg->size()); } // Sanity checks. AssertWorkerThread(); mMonitor->AssertNotCurrentThreadOwns(); MOZ_RELEASE_ASSERT(!mIsSameThreadChannel, "sync send over same-thread channel will deadlock!"); #ifdef OS_WIN SyncStackFrame frame(this, false); NeuteredWindowRegion neuteredRgn(mFlags & REQUIRE_DEFERRED_MESSAGE_PROTECTION); #endif #ifdef MOZ_TASK_TRACER AutoScopedLabel autolabel("sync message %s", aMsg->name()); #endif CxxStackFrame f(*this, OUT_MESSAGE, aMsg.get()); MonitorAutoLock lock(*mMonitor); if (mTimedOutMessageSeqno) { // Don't bother sending another sync message if a previous one timed out // and we haven't received a reply for it. Once the original timed-out // message receives a reply, we'll be able to send more sync messages // again. IPC_LOG("Send() failed due to previous timeout"); mLastSendError = SyncSendError::PreviousTimeout; return false; } if (DispatchingSyncMessageNestedLevel() == IPC::Message::NOT_NESTED && aMsg->nested_level() > IPC::Message::NOT_NESTED) { // Don't allow sending CPOWs while we're dispatching a sync message. IPC_LOG("Nested level forbids send"); mLastSendError = SyncSendError::SendingCPOWWhileDispatchingSync; return false; } if (DispatchingSyncMessageNestedLevel() == IPC::Message::NESTED_INSIDE_CPOW || DispatchingAsyncMessageNestedLevel() == IPC::Message::NESTED_INSIDE_CPOW) { // Generally only the parent dispatches urgent messages. And the only // sync messages it can send are NESTED_INSIDE_SYNC. Mainly we want to // ensure here that we don't return false for non-CPOW messages. MOZ_RELEASE_ASSERT(aMsg->nested_level() == IPC::Message::NESTED_INSIDE_SYNC); IPC_LOG("Sending while dispatching urgent message"); mLastSendError = SyncSendError::SendingCPOWWhileDispatchingUrgent; return false; } if (aMsg->nested_level() < DispatchingSyncMessageNestedLevel() || aMsg->nested_level() < AwaitingSyncReplyNestedLevel()) { MOZ_RELEASE_ASSERT(DispatchingSyncMessage() || DispatchingAsyncMessage()); MOZ_RELEASE_ASSERT(!mIsPostponingSends); IPC_LOG("Cancel from Send"); auto cancel = MakeUnique(CurrentNestedInsideSyncTransaction()); CancelTransaction(CurrentNestedInsideSyncTransaction()); mLink->SendMessage(std::move(cancel)); } IPC_ASSERT(aMsg->is_sync(), "can only Send() sync messages here"); IPC_ASSERT(aMsg->nested_level() >= DispatchingSyncMessageNestedLevel(), "can't send sync message of a lesser nested level than what's " "being dispatched"); IPC_ASSERT(AwaitingSyncReplyNestedLevel() <= aMsg->nested_level(), "nested sync message sends must be of increasing nested level"); IPC_ASSERT( DispatchingSyncMessageNestedLevel() != IPC::Message::NESTED_INSIDE_CPOW, "not allowed to send messages while dispatching urgent messages"); IPC_ASSERT( DispatchingAsyncMessageNestedLevel() != IPC::Message::NESTED_INSIDE_CPOW, "not allowed to send messages while dispatching urgent messages"); if (!Connected()) { ReportConnectionError("MessageChannel::SendAndWait", aMsg.get()); mLastSendError = SyncSendError::NotConnectedBeforeSend; return false; } aMsg->set_seqno(NextSeqno()); int32_t seqno = aMsg->seqno(); int nestedLevel = aMsg->nested_level(); msgid_t replyType = aMsg->type() + 1; AutoEnterTransaction* stackTop = mTransactionStack; // If the most recent message on the stack is NESTED_INSIDE_SYNC, then our // message should nest inside that and we use the same transaction // ID. Otherwise we need a new transaction ID (so we use the seqno of the // message we're sending). bool nest = stackTop && stackTop->NestedLevel() == IPC::Message::NESTED_INSIDE_SYNC; int32_t transaction = nest ? stackTop->TransactionID() : seqno; aMsg->set_transaction_id(transaction); bool handleWindowsMessages = mListener->HandleWindowsMessages(*aMsg.get()); AutoEnterTransaction transact(this, seqno, transaction, nestedLevel); IPC_LOG("Send seqno=%d, xid=%d", seqno, transaction); // aMsg will be destroyed soon, but name() is not owned by aMsg. const char* msgName = aMsg->name(); AddProfilerMarker(*aMsg, MessageDirection::eSending); SendMessageToLink(std::move(aMsg)); while (true) { MOZ_RELEASE_ASSERT(!transact.IsCanceled()); ProcessPendingRequests(transact); if (transact.IsComplete()) { break; } if (!Connected()) { ReportConnectionError("MessageChannel::Send"); mLastSendError = SyncSendError::DisconnectedDuringSend; return false; } MOZ_RELEASE_ASSERT(!mTimedOutMessageSeqno); MOZ_RELEASE_ASSERT(!transact.IsComplete()); MOZ_RELEASE_ASSERT(mTransactionStack == &transact); bool maybeTimedOut = !WaitForSyncNotify(handleWindowsMessages); if (mListener->NeedArtificialSleep()) { MonitorAutoUnlock unlock(*mMonitor); mListener->ArtificialSleep(); } if (!Connected()) { ReportConnectionError("MessageChannel::SendAndWait"); mLastSendError = SyncSendError::DisconnectedDuringSend; return false; } if (transact.IsCanceled()) { break; } MOZ_RELEASE_ASSERT(mTransactionStack == &transact); // We only time out a message if it initiated a new transaction (i.e., // if neither side has any other message Sends on the stack). bool canTimeOut = transact.IsBottom(); if (maybeTimedOut && canTimeOut && !ShouldContinueFromTimeout()) { // Since ShouldContinueFromTimeout drops the lock, we need to // re-check all our conditions here. We shouldn't time out if any of // these things happen because there won't be a reply to the timed // out message in these cases. if (transact.IsComplete()) { break; } IPC_LOG("Timing out Send: xid=%d", transaction); mTimedOutMessageSeqno = seqno; mTimedOutMessageNestedLevel = nestedLevel; mLastSendError = SyncSendError::TimedOut; return false; } if (transact.IsCanceled()) { break; } } if (transact.IsCanceled()) { IPC_LOG("Other side canceled seqno=%d, xid=%d", seqno, transaction); mLastSendError = SyncSendError::CancelledAfterSend; return false; } if (transact.IsError()) { IPC_LOG("Error: seqno=%d, xid=%d", seqno, transaction); mLastSendError = SyncSendError::ReplyError; return false; } uint32_t latencyMs = round((TimeStamp::Now() - start).ToMilliseconds()); IPC_LOG("Got reply: seqno=%d, xid=%d, msgName=%s, latency=%ums", seqno, transaction, msgName, latencyMs); UniquePtr reply = transact.GetReply(); MOZ_RELEASE_ASSERT(reply); MOZ_RELEASE_ASSERT(reply->is_reply(), "expected reply"); MOZ_RELEASE_ASSERT(!reply->is_reply_error()); MOZ_RELEASE_ASSERT(reply->seqno() == seqno); MOZ_RELEASE_ASSERT(reply->type() == replyType, "wrong reply type"); MOZ_RELEASE_ASSERT(reply->is_sync()); AddProfilerMarker(*reply, MessageDirection::eReceiving); *aReply = std::move(*reply); if (aReply->size() >= kMinTelemetryMessageSize) { Telemetry::Accumulate(Telemetry::IPC_REPLY_SIZE, nsDependentCString(msgName), aReply->size()); } // NOTE: Only collect IPC_SYNC_MAIN_LATENCY_MS on the main thread (bug // 1343729) if (NS_IsMainThread() && latencyMs >= kMinTelemetrySyncIPCLatencyMs) { Telemetry::Accumulate(Telemetry::IPC_SYNC_MAIN_LATENCY_MS, nsDependentCString(msgName), latencyMs); } return true; } bool MessageChannel::Call(UniquePtr aMsg, Message* aReply) { AssertWorkerThread(); mMonitor->AssertNotCurrentThreadOwns(); MOZ_RELEASE_ASSERT(!mIsSameThreadChannel, "intr call send over same-thread channel will deadlock!"); #ifdef OS_WIN SyncStackFrame frame(this, true); #endif #ifdef MOZ_TASK_TRACER AutoScopedLabel autolabel("sync message %s", aMsg->name()); #endif // This must come before MonitorAutoLock, as its destructor acquires the // monitor lock. CxxStackFrame cxxframe(*this, OUT_MESSAGE, aMsg.get()); MonitorAutoLock lock(*mMonitor); if (!Connected()) { ReportConnectionError("MessageChannel::Call", aMsg.get()); return false; } // Sanity checks. IPC_ASSERT(!AwaitingSyncReply(), "cannot issue Interrupt call while blocked on sync request"); IPC_ASSERT(!DispatchingSyncMessage(), "violation of sync handler invariant"); IPC_ASSERT(aMsg->is_interrupt(), "can only Call() Interrupt messages here"); IPC_ASSERT(!mIsPostponingSends, "not postponing sends"); aMsg->set_seqno(NextSeqno()); aMsg->set_interrupt_remote_stack_depth_guess(mRemoteStackDepthGuess); aMsg->set_interrupt_local_stack_depth(1 + InterruptStackDepth()); mInterruptStack.push(MessageInfo(*aMsg)); AddProfilerMarker(*aMsg, MessageDirection::eSending); mLink->SendMessage(std::move(aMsg)); while (true) { // if a handler invoked by *Dispatch*() spun a nested event // loop, and the connection was broken during that loop, we // might have already processed the OnError event. if so, // trying another loop iteration will be futile because // channel state will have been cleared if (!Connected()) { ReportConnectionError("MessageChannel::Call"); return false; } #ifdef OS_WIN // We need to limit the scoped of neuteredRgn to this spot in the code. // Window neutering can't be enabled during some plugin calls because // we then risk the neutered window procedure being subclassed by a // plugin. { NeuteredWindowRegion neuteredRgn(mFlags & REQUIRE_DEFERRED_MESSAGE_PROTECTION); /* We should pump messages at this point to ensure that the IPC peer does not become deadlocked on a pending inter-thread SendMessage() */ neuteredRgn.PumpOnce(); } #endif // Now might be the time to process a message deferred because of race // resolution. MaybeUndeferIncall(); // Wait for an event to occur. while (!InterruptEventOccurred()) { bool maybeTimedOut = !WaitForInterruptNotify(); // We might have received a "subtly deferred" message in a nested // loop that it's now time to process. if (InterruptEventOccurred() || (!maybeTimedOut && (!mDeferred.empty() || !mOutOfTurnReplies.empty()))) { break; } if (maybeTimedOut && !ShouldContinueFromTimeout()) return false; } Message recvd; MessageMap::iterator it; if ((it = mOutOfTurnReplies.find(mInterruptStack.top().seqno())) != mOutOfTurnReplies.end()) { recvd = std::move(it->second); mOutOfTurnReplies.erase(it); } else if (!mPending.isEmpty()) { RefPtr task = mPending.popFirst(); recvd = std::move(task->Msg()); if (!IsAlwaysDeferred(recvd)) { mMaybeDeferredPendingCount--; } } else { // because of subtleties with nested event loops, it's possible // that we got here and nothing happened. or, we might have a // deferred in-call that needs to be processed. either way, we // won't break the inner while loop again until something new // happens. continue; } // If the message is not Interrupt, we can dispatch it as normal. if (!recvd.is_interrupt()) { DispatchMessage(std::move(recvd)); if (!Connected()) { ReportConnectionError("MessageChannel::DispatchMessage"); return false; } continue; } // If the message is an Interrupt reply, either process it as a reply to our // call, or add it to the list of out-of-turn replies we've received. if (recvd.is_reply()) { IPC_ASSERT(!mInterruptStack.empty(), "invalid Interrupt stack"); // If this is not a reply the call we've initiated, add it to our // out-of-turn replies and keep polling for events. { const MessageInfo& outcall = mInterruptStack.top(); // Note, In the parent, sequence numbers increase from 0, and // in the child, they decrease from 0. if ((mSide == ChildSide && recvd.seqno() > outcall.seqno()) || (mSide != ChildSide && recvd.seqno() < outcall.seqno())) { mOutOfTurnReplies[recvd.seqno()] = std::move(recvd); continue; } IPC_ASSERT( recvd.is_reply_error() || (recvd.type() == (outcall.type() + 1) && recvd.seqno() == outcall.seqno()), "somebody's misbehavin'", true); } // We received a reply to our most recent outstanding call. Pop // this frame and return the reply. mInterruptStack.pop(); AddProfilerMarker(recvd, MessageDirection::eReceiving); bool is_reply_error = recvd.is_reply_error(); if (!is_reply_error) { *aReply = std::move(recvd); } // If we have no more pending out calls waiting on replies, then // the reply queue should be empty. IPC_ASSERT(!mInterruptStack.empty() || mOutOfTurnReplies.empty(), "still have pending replies with no pending out-calls", true); return !is_reply_error; } // Dispatch an Interrupt in-call. Snapshot the current stack depth while we // own the monitor. size_t stackDepth = InterruptStackDepth(); { #ifdef MOZ_TASK_TRACER Message::AutoTaskTracerRun tasktracerRun(recvd); #endif MonitorAutoUnlock unlock(*mMonitor); CxxStackFrame frame(*this, IN_MESSAGE, &recvd); RefPtr listenerProxy = mListener->GetLifecycleProxy(); DispatchInterruptMessage(listenerProxy, std::move(recvd), stackDepth); } if (!Connected()) { ReportConnectionError("MessageChannel::DispatchInterruptMessage"); return false; } } return true; } bool MessageChannel::WaitForIncomingMessage() { #ifdef OS_WIN SyncStackFrame frame(this, true); NeuteredWindowRegion neuteredRgn(mFlags & REQUIRE_DEFERRED_MESSAGE_PROTECTION); #endif MonitorAutoLock lock(*mMonitor); AutoEnterWaitForIncoming waitingForIncoming(*this); if (mChannelState != ChannelConnected) { return false; } if (!HasPendingEvents()) { return WaitForInterruptNotify(); } MOZ_RELEASE_ASSERT(!mPending.isEmpty()); RefPtr task = mPending.getFirst(); RunMessage(*task); return true; } bool MessageChannel::HasPendingEvents() { AssertWorkerThread(); mMonitor->AssertCurrentThreadOwns(); return Connected() && !mPending.isEmpty(); } bool MessageChannel::InterruptEventOccurred() { AssertWorkerThread(); mMonitor->AssertCurrentThreadOwns(); IPC_ASSERT(InterruptStackDepth() > 0, "not in wait loop"); return (!Connected() || !mPending.isEmpty() || (!mOutOfTurnReplies.empty() && mOutOfTurnReplies.find(mInterruptStack.top().seqno()) != mOutOfTurnReplies.end())); } bool MessageChannel::ProcessPendingRequest(Message&& aUrgent) { AssertWorkerThread(); mMonitor->AssertCurrentThreadOwns(); IPC_LOG("Process pending: seqno=%d, xid=%d", aUrgent.seqno(), aUrgent.transaction_id()); DispatchMessage(std::move(aUrgent)); if (!Connected()) { ReportConnectionError("MessageChannel::ProcessPendingRequest"); return false; } return true; } bool MessageChannel::ShouldRunMessage(const Message& aMsg) { if (!mTimedOutMessageSeqno) { return true; } // If we've timed out a message and we're awaiting the reply to the timed // out message, we have to be careful what messages we process. Here's what // can go wrong: // 1. child sends a NOT_NESTED sync message S // 2. parent sends a NESTED_INSIDE_SYNC sync message H at the same time // 3. parent times out H // 4. child starts processing H and sends a NESTED_INSIDE_SYNC message H' // nested within the same transaction // 5. parent dispatches S and sends reply // 6. child asserts because it instead expected a reply to H'. // // To solve this, we refuse to process S in the parent until we get a reply // to H. More generally, let the timed out message be M. We don't process a // message unless the child would need the response to that message in order // to process M. Those messages are the ones that have a higher nested level // than M or that are part of the same transaction as M. if (aMsg.nested_level() < mTimedOutMessageNestedLevel || (aMsg.nested_level() == mTimedOutMessageNestedLevel && aMsg.transaction_id() != mTimedOutMessageSeqno)) { return false; } return true; } void MessageChannel::RunMessage(MessageTask& aTask) { AssertWorkerThread(); mMonitor->AssertCurrentThreadOwns(); Message& msg = aTask.Msg(); if (!Connected()) { ReportConnectionError("RunMessage"); return; } // Check that we're going to run the first message that's valid to run. #if 0 # ifdef DEBUG nsCOMPtr messageTarget = mListener->GetMessageEventTarget(msg); for (MessageTask* task : mPending) { if (task == &aTask) { break; } nsCOMPtr taskTarget = mListener->GetMessageEventTarget(task->Msg()); MOZ_ASSERT(!ShouldRunMessage(task->Msg()) || taskTarget != messageTarget || aTask.Msg().priority() != task->Msg().priority()); } # endif #endif if (!mDeferred.empty()) { MaybeUndeferIncall(); } if (!ShouldRunMessage(msg)) { return; } MOZ_RELEASE_ASSERT(aTask.isInList()); aTask.remove(); if (!IsAlwaysDeferred(msg)) { mMaybeDeferredPendingCount--; } if (IsOnCxxStack() && msg.is_interrupt() && msg.is_reply()) { // We probably just received a reply in a nested loop for an // Interrupt call sent before entering that loop. mOutOfTurnReplies[msg.seqno()] = std::move(msg); return; } DispatchMessage(std::move(msg)); } NS_IMPL_ISUPPORTS_INHERITED(MessageChannel::MessageTask, CancelableRunnable, nsIRunnablePriority, nsIRunnableIPCMessageType) MessageChannel::MessageTask::MessageTask(MessageChannel* aChannel, Message&& aMessage) : CancelableRunnable(aMessage.name()), mChannel(aChannel), mMessage(std::move(aMessage)), mScheduled(false) {} nsresult MessageChannel::MessageTask::Run() { if (!mChannel) { return NS_OK; } mChannel->AssertWorkerThread(); mChannel->mMonitor->AssertNotCurrentThreadOwns(); MonitorAutoLock lock(*mChannel->mMonitor); // In case we choose not to run this message, we may need to be able to Post // it again. mScheduled = false; if (!isInList()) { return NS_OK; } mChannel->RunMessage(*this); return NS_OK; } // Warning: This method removes the receiver from whatever list it might be in. nsresult MessageChannel::MessageTask::Cancel() { if (!mChannel) { return NS_OK; } mChannel->AssertWorkerThread(); mChannel->mMonitor->AssertNotCurrentThreadOwns(); MonitorAutoLock lock(*mChannel->mMonitor); if (!isInList()) { return NS_OK; } remove(); if (!IsAlwaysDeferred(Msg())) { mChannel->mMaybeDeferredPendingCount--; } return NS_OK; } void MessageChannel::MessageTask::Post() { MOZ_RELEASE_ASSERT(!mScheduled); MOZ_RELEASE_ASSERT(isInList()); mScheduled = true; RefPtr self = this; nsCOMPtr eventTarget = mChannel->mListener->GetMessageEventTarget(mMessage); if (eventTarget) { eventTarget->Dispatch(self.forget(), NS_DISPATCH_NORMAL); } else { mChannel->mWorkerThread->Dispatch(self.forget()); } } void MessageChannel::MessageTask::Clear() { mChannel->AssertWorkerThread(); mChannel = nullptr; } NS_IMETHODIMP MessageChannel::MessageTask::GetPriority(uint32_t* aPriority) { switch (mMessage.priority()) { case Message::NORMAL_PRIORITY: *aPriority = PRIORITY_NORMAL; break; case Message::INPUT_PRIORITY: *aPriority = PRIORITY_INPUT_HIGH; break; case Message::HIGH_PRIORITY: *aPriority = PRIORITY_HIGH; break; case Message::MEDIUMHIGH_PRIORITY: *aPriority = PRIORITY_MEDIUMHIGH; break; default: MOZ_ASSERT(false); break; } return NS_OK; } NS_IMETHODIMP MessageChannel::MessageTask::GetType(uint32_t* aType) { if (!Msg().is_valid()) { // If mMessage has been moved already elsewhere, we can't know what the type // has been. return NS_ERROR_FAILURE; } *aType = Msg().type(); return NS_OK; } void MessageChannel::DispatchMessage(Message&& aMsg) { AssertWorkerThread(); mMonitor->AssertCurrentThreadOwns(); RefPtr listenerProxy = mListener->GetLifecycleProxy(); Maybe nojsapi; if (NS_IsMainThread() && CycleCollectedJSContext::Get()) { nojsapi.emplace(); } UniquePtr reply; IPC_LOG("DispatchMessage: seqno=%d, xid=%d", aMsg.seqno(), aMsg.transaction_id()); AddProfilerMarker(aMsg, MessageDirection::eReceiving); { AutoEnterTransaction transaction(this, aMsg); int id = aMsg.transaction_id(); MOZ_RELEASE_ASSERT(!aMsg.is_sync() || id == transaction.TransactionID()); { #ifdef MOZ_TASK_TRACER Message::AutoTaskTracerRun tasktracerRun(aMsg); #endif MonitorAutoUnlock unlock(*mMonitor); CxxStackFrame frame(*this, IN_MESSAGE, &aMsg); mListener->ArtificialSleep(); if (aMsg.is_sync()) { DispatchSyncMessage(listenerProxy, aMsg, *getter_Transfers(reply)); } else if (aMsg.is_interrupt()) { DispatchInterruptMessage(listenerProxy, std::move(aMsg), 0); } else { DispatchAsyncMessage(listenerProxy, aMsg); } mListener->ArtificialSleep(); } if (reply && transaction.IsCanceled()) { // The transaction has been canceled. Don't send a reply. IPC_LOG("Nulling out reply due to cancellation, seqno=%d, xid=%d", aMsg.seqno(), id); reply = nullptr; } } if (reply && ChannelConnected == mChannelState) { IPC_LOG("Sending reply seqno=%d, xid=%d", aMsg.seqno(), aMsg.transaction_id()); AddProfilerMarker(*reply, MessageDirection::eSending); mLink->SendMessage(std::move(reply)); } } void MessageChannel::DispatchSyncMessage(ActorLifecycleProxy* aProxy, const Message& aMsg, Message*& aReply) { AssertWorkerThread(); mozilla::TimeStamp start = TimeStamp::Now(); int nestedLevel = aMsg.nested_level(); MOZ_RELEASE_ASSERT(nestedLevel == IPC::Message::NOT_NESTED || NS_IsMainThread()); #ifdef MOZ_TASK_TRACER AutoScopedLabel autolabel("sync message %s", aMsg.name()); #endif MessageChannel* dummy; MessageChannel*& blockingVar = mSide == ChildSide && NS_IsMainThread() ? gParentProcessBlocker : dummy; Result rv; { AutoSetValue blocked(blockingVar, this); rv = aProxy->Get()->OnMessageReceived(aMsg, aReply); } uint32_t latencyMs = round((TimeStamp::Now() - start).ToMilliseconds()); if (latencyMs >= kMinTelemetrySyncIPCLatencyMs) { Telemetry::Accumulate(Telemetry::IPC_SYNC_RECEIVE_MS, nsDependentCString(aMsg.name()), latencyMs); } if (!MaybeHandleError(rv, aMsg, "DispatchSyncMessage")) { aReply = Message::ForSyncDispatchError(aMsg.nested_level()); } aReply->set_seqno(aMsg.seqno()); aReply->set_transaction_id(aMsg.transaction_id()); } void MessageChannel::DispatchAsyncMessage(ActorLifecycleProxy* aProxy, const Message& aMsg) { AssertWorkerThread(); MOZ_RELEASE_ASSERT(!aMsg.is_interrupt() && !aMsg.is_sync()); if (aMsg.routing_id() == MSG_ROUTING_NONE) { MOZ_CRASH("unhandled special message!"); } Result rv; { int nestedLevel = aMsg.nested_level(); AutoSetValue async(mDispatchingAsyncMessage, true); AutoSetValue nestedLevelSet(mDispatchingAsyncMessageNestedLevel, nestedLevel); rv = aProxy->Get()->OnMessageReceived(aMsg); } MaybeHandleError(rv, aMsg, "DispatchAsyncMessage"); } void MessageChannel::DispatchInterruptMessage(ActorLifecycleProxy* aProxy, Message&& aMsg, size_t stackDepth) { AssertWorkerThread(); mMonitor->AssertNotCurrentThreadOwns(); IPC_ASSERT(aMsg.is_interrupt() && !aMsg.is_reply(), "wrong message type"); if (ShouldDeferInterruptMessage(aMsg, stackDepth)) { // We now know the other side's stack has one more frame // than we thought. ++mRemoteStackDepthGuess; // decremented in MaybeProcessDeferred() mDeferred.push(std::move(aMsg)); return; } // If we "lost" a race and need to process the other side's in-call, we // don't need to fix up the mRemoteStackDepthGuess here, because we're just // about to increment it, which will make it correct again. #ifdef OS_WIN SyncStackFrame frame(this, true); #endif UniquePtr reply; ++mRemoteStackDepthGuess; Result rv = aProxy->Get()->OnCallReceived(aMsg, *getter_Transfers(reply)); --mRemoteStackDepthGuess; if (!MaybeHandleError(rv, aMsg, "DispatchInterruptMessage")) { reply = WrapUnique(Message::ForInterruptDispatchError()); } reply->set_seqno(aMsg.seqno()); MonitorAutoLock lock(*mMonitor); if (ChannelConnected == mChannelState) { AddProfilerMarker(*reply, MessageDirection::eSending); mLink->SendMessage(std::move(reply)); } } bool MessageChannel::ShouldDeferInterruptMessage(const Message& aMsg, size_t aStackDepth) { AssertWorkerThread(); // We may or may not own the lock in this function, so don't access any // channel state. IPC_ASSERT(aMsg.is_interrupt() && !aMsg.is_reply(), "wrong message type"); // Race detection: see the long comment near mRemoteStackDepthGuess in // MessageChannel.h. "Remote" stack depth means our side, and "local" means // the other side. if (aMsg.interrupt_remote_stack_depth_guess() == RemoteViewOfStackDepth(aStackDepth)) { return false; } // Interrupt in-calls have raced. The winner, if there is one, gets to defer // processing of the other side's in-call. bool defer; const char* winner; const MessageInfo parentMsgInfo = (mSide == ChildSide) ? MessageInfo(aMsg) : mInterruptStack.top(); const MessageInfo childMsgInfo = (mSide == ChildSide) ? mInterruptStack.top() : MessageInfo(aMsg); switch (mListener->MediateInterruptRace(parentMsgInfo, childMsgInfo)) { case RIPChildWins: winner = "child"; defer = (mSide == ChildSide); break; case RIPParentWins: winner = "parent"; defer = (mSide != ChildSide); break; case RIPError: MOZ_CRASH("NYI: 'Error' Interrupt race policy"); default: MOZ_CRASH("not reached"); } IPC_LOG("race in %s: %s won", (mSide == ChildSide) ? "child" : "parent", winner); return defer; } void MessageChannel::MaybeUndeferIncall() { AssertWorkerThread(); mMonitor->AssertCurrentThreadOwns(); if (mDeferred.empty()) return; size_t stackDepth = InterruptStackDepth(); Message& deferred = mDeferred.top(); // the other side can only *under*-estimate our actual stack depth IPC_ASSERT(deferred.interrupt_remote_stack_depth_guess() <= stackDepth, "fatal logic error"); if (ShouldDeferInterruptMessage(deferred, stackDepth)) { return; } // maybe time to process this message Message call(std::move(deferred)); mDeferred.pop(); // fix up fudge factor we added to account for race IPC_ASSERT(0 < mRemoteStackDepthGuess, "fatal logic error"); --mRemoteStackDepthGuess; MOZ_RELEASE_ASSERT(call.nested_level() == IPC::Message::NOT_NESTED); RefPtr task = new MessageTask(this, std::move(call)); mPending.insertBack(task); MOZ_ASSERT(IsAlwaysDeferred(task->Msg())); task->Post(); } void MessageChannel::EnteredCxxStack() { mListener->EnteredCxxStack(); } void MessageChannel::ExitedCxxStack() { mListener->ExitedCxxStack(); if (mSawInterruptOutMsg) { MonitorAutoLock lock(*mMonitor); // see long comment in OnMaybeDequeueOne() EnqueuePendingMessages(); mSawInterruptOutMsg = false; } } void MessageChannel::EnteredCall() { mListener->EnteredCall(); } void MessageChannel::ExitedCall() { mListener->ExitedCall(); } void MessageChannel::EnteredSyncSend() { mListener->OnEnteredSyncSend(); } void MessageChannel::ExitedSyncSend() { mListener->OnExitedSyncSend(); } void MessageChannel::EnqueuePendingMessages() { AssertWorkerThread(); mMonitor->AssertCurrentThreadOwns(); MaybeUndeferIncall(); // XXX performance tuning knob: could process all or k pending // messages here, rather than enqueuing for later processing RepostAllMessages(); } bool MessageChannel::WaitResponse(bool aWaitTimedOut) { if (aWaitTimedOut) { if (mInTimeoutSecondHalf) { // We've really timed out this time. return false; } // Try a second time. mInTimeoutSecondHalf = true; } else { mInTimeoutSecondHalf = false; } return true; } #ifndef OS_WIN bool MessageChannel::WaitForSyncNotify(bool /* aHandleWindowsMessages */) { # ifdef DEBUG // WARNING: We don't release the lock here. We can't because the link thread // could signal at this time and we would miss it. Instead we require // ArtificialTimeout() to be extremely simple. if (mListener->ArtificialTimeout()) { return false; } # endif MOZ_RELEASE_ASSERT(!mIsSameThreadChannel, "Wait on same-thread channel will deadlock!"); TimeDuration timeout = (kNoTimeout == mTimeoutMs) ? TimeDuration::Forever() : TimeDuration::FromMilliseconds(mTimeoutMs); CVStatus status = mMonitor->Wait(timeout); // If the timeout didn't expire, we know we received an event. The // converse is not true. return WaitResponse(status == CVStatus::Timeout); } bool MessageChannel::WaitForInterruptNotify() { return WaitForSyncNotify(true); } void MessageChannel::NotifyWorkerThread() { mMonitor->Notify(); } #endif bool MessageChannel::ShouldContinueFromTimeout() { AssertWorkerThread(); mMonitor->AssertCurrentThreadOwns(); bool cont; { MonitorAutoUnlock unlock(*mMonitor); cont = mListener->ShouldContinueFromReplyTimeout(); mListener->ArtificialSleep(); } static enum { UNKNOWN, NOT_DEBUGGING, DEBUGGING } sDebuggingChildren = UNKNOWN; if (sDebuggingChildren == UNKNOWN) { sDebuggingChildren = getenv("MOZ_DEBUG_CHILD_PROCESS") || getenv("MOZ_DEBUG_CHILD_PAUSE") ? DEBUGGING : NOT_DEBUGGING; } if (sDebuggingChildren == DEBUGGING) { return true; } return cont; } void MessageChannel::SetReplyTimeoutMs(int32_t aTimeoutMs) { // Set channel timeout value. Since this is broken up into // two period, the minimum timeout value is 2ms. AssertWorkerThread(); mTimeoutMs = (aTimeoutMs <= 0) ? kNoTimeout : (int32_t)ceil((double)aTimeoutMs / 2.0); } void MessageChannel::OnChannelConnected(int32_t peer_id) { MOZ_RELEASE_ASSERT(!mPeerPidSet); mPeerPidSet = true; mPeerPid = peer_id; RefPtr task = mOnChannelConnectedTask; mWorkerThread->Dispatch(task.forget()); } void MessageChannel::DispatchOnChannelConnected() { AssertWorkerThread(); MOZ_RELEASE_ASSERT(mPeerPidSet); mListener->OnChannelConnected(mPeerPid); } void MessageChannel::ReportMessageRouteError(const char* channelName) const { PrintErrorMessage(mSide, channelName, "Need a route"); mListener->ProcessingError(MsgRouteError, "MsgRouteError"); } void MessageChannel::ReportConnectionError(const char* aChannelName, Message* aMsg) const { AssertWorkerThread(); mMonitor->AssertCurrentThreadOwns(); const char* errorMsg = nullptr; switch (mChannelState) { case ChannelClosed: errorMsg = "Closed channel: cannot send/recv"; break; case ChannelOpening: errorMsg = "Opening channel: not yet ready for send/recv"; break; case ChannelTimeout: errorMsg = "Channel timeout: cannot send/recv"; break; case ChannelClosing: errorMsg = "Channel closing: too late to send/recv, messages will be lost"; break; case ChannelError: errorMsg = "Channel error: cannot send/recv"; break; default: MOZ_CRASH("unreached"); } if (aMsg) { char reason[512]; SprintfLiteral(reason, "(msgtype=0x%X,name=%s) %s", aMsg->type(), aMsg->name(), errorMsg); PrintErrorMessage(mSide, aChannelName, reason); } else { PrintErrorMessage(mSide, aChannelName, errorMsg); } MonitorAutoUnlock unlock(*mMonitor); mListener->ProcessingError(MsgDropped, errorMsg); } bool MessageChannel::MaybeHandleError(Result code, const Message& aMsg, const char* channelName) { if (MsgProcessed == code) return true; const char* errorMsg = nullptr; switch (code) { case MsgNotKnown: errorMsg = "Unknown message: not processed"; break; case MsgNotAllowed: errorMsg = "Message not allowed: cannot be sent/recvd in this state"; break; case MsgPayloadError: errorMsg = "Payload error: message could not be deserialized"; break; case MsgProcessingError: errorMsg = "Processing error: message was deserialized, but the handler " "returned false (indicating failure)"; break; case MsgRouteError: errorMsg = "Route error: message sent to unknown actor ID"; break; case MsgValueError: errorMsg = "Value error: message was deserialized, but contained an illegal " "value"; break; default: MOZ_CRASH("unknown Result code"); return false; } char reason[512]; const char* msgname = aMsg.name(); if (msgname[0] == '?') { SprintfLiteral(reason, "(msgtype=0x%X) %s", aMsg.type(), errorMsg); } else { SprintfLiteral(reason, "%s %s", msgname, errorMsg); } PrintErrorMessage(mSide, channelName, reason); // Error handled in mozilla::ipc::IPCResult. if (code == MsgProcessingError) { return false; } mListener->ProcessingError(code, reason); return false; } void MessageChannel::OnChannelErrorFromLink() { AssertLinkThread(); mMonitor->AssertCurrentThreadOwns(); IPC_LOG("OnChannelErrorFromLink"); if (InterruptStackDepth() > 0) NotifyWorkerThread(); if (AwaitingSyncReply() || AwaitingIncomingMessage()) NotifyWorkerThread(); if (ChannelClosing != mChannelState) { if (mAbortOnError) { // mAbortOnError is set by main actors (e.g., ContentChild) to ensure // that the process terminates even if normal shutdown is prevented. // A MOZ_CRASH() here is not helpful because crash reporting relies // on the parent process which we know is dead or otherwise unusable. // // Additionally, the parent process can (and often is) killed on Android // when apps are backgrounded. We don't need to report a crash for // normal behavior in that case. printf_stderr("Exiting due to channel error.\n"); ProcessChild::QuickExit(); } mChannelState = ChannelError; mMonitor->Notify(); } PostErrorNotifyTask(); } void MessageChannel::NotifyMaybeChannelError() { mMonitor->AssertNotCurrentThreadOwns(); // TODO sort out Close() on this side racing with Close() on the other side if (ChannelClosing == mChannelState) { // the channel closed, but we received a "Goodbye" message warning us // about it. no worries mChannelState = ChannelClosed; NotifyChannelClosed(); return; } Clear(); // Oops, error! Let the listener know about it. mChannelState = ChannelError; // IPDL assumes these notifications do not fire twice, so we do not let // that happen. if (mNotifiedChannelDone) { return; } mNotifiedChannelDone = true; // After this, the channel may be deleted. Based on the premise that // mListener owns this channel, any calls back to this class that may // work with mListener should still work on living objects. mListener->OnChannelError(); } void MessageChannel::OnNotifyMaybeChannelError() { AssertWorkerThread(); mMonitor->AssertNotCurrentThreadOwns(); mChannelErrorTask = nullptr; // OnChannelError holds mMonitor when it posts this task and this // task cannot be allowed to run until OnChannelError has // exited. We enforce that order by grabbing the mutex here which // should only continue once OnChannelError has completed. { MonitorAutoLock lock(*mMonitor); // nothing to do here } if (IsOnCxxStack()) { mChannelErrorTask = NewNonOwningCancelableRunnableMethod( "ipc::MessageChannel::OnNotifyMaybeChannelError", this, &MessageChannel::OnNotifyMaybeChannelError); RefPtr task = mChannelErrorTask; // This used to post a 10ms delayed patch; however not all // nsISerialEventTarget implementations support delayed dispatch. // The delay being completely arbitrary, we may not as well have any. mWorkerThread->Dispatch(task.forget()); return; } NotifyMaybeChannelError(); } void MessageChannel::PostErrorNotifyTask() { mMonitor->AssertCurrentThreadOwns(); if (mChannelErrorTask) return; // This must be the last code that runs on this thread! mChannelErrorTask = NewNonOwningCancelableRunnableMethod( "ipc::MessageChannel::OnNotifyMaybeChannelError", this, &MessageChannel::OnNotifyMaybeChannelError); RefPtr task = mChannelErrorTask; mWorkerThread->Dispatch(task.forget()); } // Special async message. class GoodbyeMessage : public IPC::Message { public: GoodbyeMessage() : IPC::Message(MSG_ROUTING_NONE, GOODBYE_MESSAGE_TYPE) {} static bool Read(const Message* msg) { return true; } void Log(const std::string& aPrefix, FILE* aOutf) const { fputs("(special `Goodbye' message)", aOutf); } }; void MessageChannel::SynchronouslyClose() { AssertWorkerThread(); mMonitor->AssertCurrentThreadOwns(); mLink->SendClose(); MOZ_RELEASE_ASSERT(!mIsSameThreadChannel || ChannelClosed == mChannelState, "same-thread channel failed to synchronously close?"); while (ChannelClosed != mChannelState) mMonitor->Wait(); } void MessageChannel::CloseWithError() { AssertWorkerThread(); MonitorAutoLock lock(*mMonitor); if (ChannelConnected != mChannelState) { return; } SynchronouslyClose(); mChannelState = ChannelError; PostErrorNotifyTask(); } void MessageChannel::CloseWithTimeout() { AssertWorkerThread(); MonitorAutoLock lock(*mMonitor); if (ChannelConnected != mChannelState) { return; } SynchronouslyClose(); mChannelState = ChannelTimeout; } void MessageChannel::NotifyImpendingShutdown() { UniquePtr msg = MakeUnique(MSG_ROUTING_NONE, IMPENDING_SHUTDOWN_MESSAGE_TYPE); MonitorAutoLock lock(*mMonitor); if (Connected()) { MOZ_DIAGNOSTIC_ASSERT(mIsCrossProcess); mLink->SendMessage(std::move(msg)); } } void MessageChannel::Close() { AssertWorkerThread(); { // We don't use MonitorAutoLock here as that causes some sort of // deadlock in the error/timeout-with-a-listener state below when // compiling an optimized msvc build. mMonitor->Lock(); // Instead just use a ScopeExit to manage the unlock. RefPtr monitor(mMonitor); auto exit = MakeScopeExit([m = std::move(monitor)]() { m->Unlock(); }); if (ChannelError == mChannelState || ChannelTimeout == mChannelState) { // See bug 538586: if the listener gets deleted while the // IO thread's NotifyChannelError event is still enqueued // and subsequently deletes us, then the error event will // also be deleted and the listener will never be notified // of the channel error. if (mListener) { exit.release(); // Explicitly unlocking, clear scope exit. mMonitor->Unlock(); NotifyMaybeChannelError(); } return; } if (ChannelOpening == mChannelState) { // SynchronouslyClose() waits for an ack from the other side, so // the opening sequence should complete before this returns. SynchronouslyClose(); mChannelState = ChannelError; NotifyMaybeChannelError(); return; } if (ChannelClosed == mChannelState) { // Slightly unexpected but harmless; ignore. See bug 1554244. return; } // Notify the other side that we're about to close our socket. If we've // already received a Goodbye from the other side (and our state is // ChannelClosing), there's no reason to send one. if (ChannelConnected == mChannelState) { mLink->SendMessage(MakeUnique()); } SynchronouslyClose(); } NotifyChannelClosed(); } void MessageChannel::NotifyChannelClosed() { mMonitor->AssertNotCurrentThreadOwns(); if (ChannelClosed != mChannelState) MOZ_CRASH("channel should have been closed!"); Clear(); // IPDL assumes these notifications do not fire twice, so we do not let // that happen. if (mNotifiedChannelDone) { return; } mNotifiedChannelDone = true; // OK, the IO thread just closed the channel normally. Let the // listener know about it. After this point the channel may be // deleted. mListener->OnChannelClose(); } void MessageChannel::DebugAbort(const char* file, int line, const char* cond, const char* why, bool reply) { printf_stderr( "###!!! [MessageChannel][%s][%s:%d] " "Assertion (%s) failed. %s %s\n", mSide == ChildSide ? "Child" : "Parent", file, line, cond, why, reply ? "(reply)" : ""); // technically we need the mutex for this, but we're dying anyway DumpInterruptStack(" "); printf_stderr(" remote Interrupt stack guess: %zu\n", mRemoteStackDepthGuess); printf_stderr(" deferred stack size: %zu\n", mDeferred.size()); printf_stderr(" out-of-turn Interrupt replies stack size: %zu\n", mOutOfTurnReplies.size()); MessageQueue pending = std::move(mPending); while (!pending.isEmpty()) { printf_stderr( " [ %s%s ]\n", pending.getFirst()->Msg().is_interrupt() ? "intr" : (pending.getFirst()->Msg().is_sync() ? "sync" : "async"), pending.getFirst()->Msg().is_reply() ? "reply" : ""); pending.popFirst(); } MOZ_CRASH_UNSAFE(why); } void MessageChannel::DumpInterruptStack(const char* const pfx) const { NS_WARNING_ASSERTION(!mWorkerThread->IsOnCurrentThread(), "The worker thread had better be paused in a debugger!"); printf_stderr("%sMessageChannel 'backtrace':\n", pfx); // print a python-style backtrace, first frame to last for (uint32_t i = 0; i < mCxxStackFrames.length(); ++i) { int32_t id; const char* dir; const char* sems; const char* name; mCxxStackFrames[i].Describe(&id, &dir, &sems, &name); printf_stderr("%s[(%u) %s %s %s(actor=%d) ]\n", pfx, i, dir, sems, name, id); } } void MessageChannel::AddProfilerMarker(const IPC::Message& aMessage, MessageDirection aDirection) { mMonitor->AssertCurrentThreadOwns(); #ifdef MOZ_GECKO_PROFILER if (profiler_feature_active(ProfilerFeature::IPCMessages)) { int32_t pid = mListener->OtherPidMaybeInvalid(); // Only record markers for IPCs with a valid pid. // And if one of the profiler mutexes is locked on this thread, don't record // markers, because we don't want to expose profiler IPCs due to the // profiler itself, and also to avoid possible re-entrancy issues. if (pid != kInvalidProcessId && !profiler_is_locked_on_current_thread()) { // The current timestamp must be given to the `IPCMarker` payload. const TimeStamp now = TimeStamp::NowUnfuzzed(); PROFILER_MARKER("IPC", IPC, MarkerTiming::InstantAt(now), IPCMarker, now, now, pid, aMessage.seqno(), aMessage.type(), mSide, aDirection, MessagePhase::Endpoint, aMessage.is_sync()); } } #endif } int32_t MessageChannel::GetTopmostMessageRoutingId() const { AssertWorkerThread(); if (mCxxStackFrames.empty()) { return MSG_ROUTING_NONE; } const InterruptFrame& frame = mCxxStackFrames.back(); return frame.GetRoutingId(); } void MessageChannel::EndTimeout() { mMonitor->AssertCurrentThreadOwns(); IPC_LOG("Ending timeout of seqno=%d", mTimedOutMessageSeqno); mTimedOutMessageSeqno = 0; mTimedOutMessageNestedLevel = 0; RepostAllMessages(); } void MessageChannel::RepostAllMessages() { bool needRepost = false; for (MessageTask* task : mPending) { if (!task->IsScheduled()) { needRepost = true; break; } } if (!needRepost) { // If everything is already scheduled to run, do nothing. return; } // In some cases we may have deferred dispatch of some messages in the // queue. Now we want to run them again. However, we can't just re-post // those messages since the messages after them in mPending would then be // before them in the event queue. So instead we cancel everything and // re-post all messages in the correct order. MessageQueue queue = std::move(mPending); while (RefPtr task = queue.popFirst()) { RefPtr newTask = new MessageTask(this, std::move(task->Msg())); mPending.insertBack(newTask); newTask->Post(); } AssertMaybeDeferredCountCorrect(); } void MessageChannel::CancelTransaction(int transaction) { mMonitor->AssertCurrentThreadOwns(); // When we cancel a transaction, we need to behave as if there's no longer // any IPC on the stack. Anything we were dispatching or sending will get // canceled. Consequently, we have to update the state variables below. // // We also need to ensure that when any IPC functions on the stack return, // they don't reset these values using an RAII class like AutoSetValue. To // avoid that, these RAII classes check if the variable they set has been // tampered with (by us). If so, they don't reset the variable to the old // value. IPC_LOG("CancelTransaction: xid=%d", transaction); // An unusual case: We timed out a transaction which the other side then // cancelled. In this case we just leave the timedout state and try to // forget this ever happened. if (transaction == mTimedOutMessageSeqno) { IPC_LOG("Cancelled timed out message %d", mTimedOutMessageSeqno); EndTimeout(); // Normally mCurrentTransaction == 0 here. But it can be non-zero if: // 1. Parent sends NESTED_INSIDE_SYNC message H. // 2. Parent times out H. // 3. Child dispatches H and sends nested message H' (same transaction). // 4. Parent dispatches H' and cancels. MOZ_RELEASE_ASSERT(!mTransactionStack || mTransactionStack->TransactionID() == transaction); if (mTransactionStack) { mTransactionStack->Cancel(); } } else { MOZ_RELEASE_ASSERT(mTransactionStack->TransactionID() == transaction); mTransactionStack->Cancel(); } bool foundSync = false; for (MessageTask* p = mPending.getFirst(); p;) { Message& msg = p->Msg(); // If there was a race between the parent and the child, then we may // have a queued sync message. We want to drop this message from the // queue since if will get cancelled along with the transaction being // cancelled. This happens if the message in the queue is // NESTED_INSIDE_SYNC. if (msg.is_sync() && msg.nested_level() != IPC::Message::NOT_NESTED) { MOZ_RELEASE_ASSERT(!foundSync); MOZ_RELEASE_ASSERT(msg.transaction_id() != transaction); IPC_LOG("Removing msg from queue seqno=%d xid=%d", msg.seqno(), msg.transaction_id()); foundSync = true; if (!IsAlwaysDeferred(msg)) { mMaybeDeferredPendingCount--; } p = p->removeAndGetNext(); continue; } p = p->getNext(); } AssertMaybeDeferredCountCorrect(); } void MessageChannel::CancelCurrentTransaction() { MonitorAutoLock lock(*mMonitor); if (DispatchingSyncMessageNestedLevel() >= IPC::Message::NESTED_INSIDE_SYNC) { if (DispatchingSyncMessageNestedLevel() == IPC::Message::NESTED_INSIDE_CPOW || DispatchingAsyncMessageNestedLevel() == IPC::Message::NESTED_INSIDE_CPOW) { mListener->IntentionalCrash(); } IPC_LOG("Cancel requested: current xid=%d", CurrentNestedInsideSyncTransaction()); MOZ_RELEASE_ASSERT(DispatchingSyncMessage()); auto cancel = MakeUnique(CurrentNestedInsideSyncTransaction()); CancelTransaction(CurrentNestedInsideSyncTransaction()); mLink->SendMessage(std::move(cancel)); } } void CancelCPOWs() { if (gParentProcessBlocker) { mozilla::Telemetry::Accumulate(mozilla::Telemetry::IPC_TRANSACTION_CANCEL, true); gParentProcessBlocker->CancelCurrentTransaction(); } } } // namespace ipc } // namespace mozilla