diff options
author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-07 16:49:04 +0000 |
---|---|---|
committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-07 16:49:04 +0000 |
commit | 16f504a9dca3fe3b70568f67b7d41241ae485288 (patch) | |
tree | c60f36ada0496ba928b7161059ba5ab1ab224f9d /src/VBox/Devices/Network/DrvIntNet.cpp | |
parent | Initial commit. (diff) | |
download | virtualbox-upstream.tar.xz virtualbox-upstream.zip |
Adding upstream version 7.0.6-dfsg.upstream/7.0.6-dfsgupstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'src/VBox/Devices/Network/DrvIntNet.cpp')
-rw-r--r-- | src/VBox/Devices/Network/DrvIntNet.cpp | 2136 |
1 files changed, 2136 insertions, 0 deletions
diff --git a/src/VBox/Devices/Network/DrvIntNet.cpp b/src/VBox/Devices/Network/DrvIntNet.cpp new file mode 100644 index 00000000..76d429ec --- /dev/null +++ b/src/VBox/Devices/Network/DrvIntNet.cpp @@ -0,0 +1,2136 @@ +/* $Id: DrvIntNet.cpp $ */ +/** @file + * DrvIntNet - Internal network transport driver. + */ + +/* + * Copyright (C) 2006-2022 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * SPDX-License-Identifier: GPL-3.0-only + */ + + +/********************************************************************************************************************************* +* Header Files * +*********************************************************************************************************************************/ +#define LOG_GROUP LOG_GROUP_DRV_INTNET +#if defined(RT_OS_DARWIN) && defined(VBOX_WITH_INTNET_SERVICE_IN_R3) +# include <xpc/xpc.h> /* This needs to be here because it drags PVM in and cdefs.h needs to undefine it... */ +#endif +#include <iprt/cdefs.h> + +#include <VBox/vmm/pdmdrv.h> +#include <VBox/vmm/pdmnetinline.h> +#include <VBox/vmm/pdmnetifs.h> +#include <VBox/vmm/cfgm.h> +#include <VBox/intnet.h> +#include <VBox/intnetinline.h> +#include <VBox/vmm/vmm.h> +#include <VBox/sup.h> +#include <VBox/err.h> + +#include <VBox/param.h> +#include <VBox/log.h> +#include <iprt/asm.h> +#include <iprt/assert.h> +#include <iprt/ctype.h> +#include <iprt/memcache.h> +#include <iprt/net.h> +#include <iprt/semaphore.h> +#include <iprt/string.h> +#include <iprt/time.h> +#include <iprt/thread.h> +#include <iprt/uuid.h> +#if defined(RT_OS_DARWIN) && defined(IN_RING3) +# include <iprt/system.h> +#endif + +#include "VBoxDD.h" + + +/********************************************************************************************************************************* +* Defined Constants And Macros * +*********************************************************************************************************************************/ +#if 0 +/** Enables the ring-0 part. */ +#define VBOX_WITH_DRVINTNET_IN_R0 +#endif + + +/********************************************************************************************************************************* +* Structures and Typedefs * +*********************************************************************************************************************************/ +/** + * The state of the asynchronous thread. + */ +typedef enum RECVSTATE +{ + /** The thread is suspended. */ + RECVSTATE_SUSPENDED = 1, + /** The thread is running. */ + RECVSTATE_RUNNING, + /** The thread must (/has) terminate. */ + RECVSTATE_TERMINATE, + /** The usual 32-bit type blowup. */ + RECVSTATE_32BIT_HACK = 0x7fffffff +} RECVSTATE; + +/** + * Internal networking driver instance data. + * + * @implements PDMINETWORKUP + */ +typedef struct DRVINTNET +{ + /** The network interface. */ + PDMINETWORKUP INetworkUpR3; + /** The network interface. */ + R3PTRTYPE(PPDMINETWORKDOWN) pIAboveNet; + /** The network config interface. + * Can (in theory at least) be NULL. */ + R3PTRTYPE(PPDMINETWORKCONFIG) pIAboveConfigR3; + /** Pointer to the driver instance (ring-3). */ + PPDMDRVINSR3 pDrvInsR3; + /** Pointer to the communication buffer (ring-3). */ + R3PTRTYPE(PINTNETBUF) pBufR3; +#ifdef VBOX_WITH_DRVINTNET_IN_R0 + /** Ring-3 base interface for the ring-0 context. */ + PDMIBASER0 IBaseR0; + /** Ring-3 base interface for the raw-mode context. */ + PDMIBASERC IBaseRC; + RTR3PTR R3PtrAlignment; + + /** The network interface for the ring-0 context. */ + PDMINETWORKUPR0 INetworkUpR0; + /** Pointer to the driver instance (ring-0). */ + PPDMDRVINSR0 pDrvInsR0; + /** Pointer to the communication buffer (ring-0). */ + R0PTRTYPE(PINTNETBUF) pBufR0; + + /** The network interface for the raw-mode context. */ + PDMINETWORKUPRC INetworkUpRC; + /** Pointer to the driver instance. */ + PPDMDRVINSRC pDrvInsRC; + RTRCPTR RCPtrAlignment; +#endif + + /** The transmit lock. */ + PDMCRITSECT XmitLock; + /** Interface handle. */ + INTNETIFHANDLE hIf; + /** The receive thread state. */ + RECVSTATE volatile enmRecvState; + /** The receive thread. */ + RTTHREAD hRecvThread; + /** The event semaphore that the receive thread waits on. */ + RTSEMEVENT hRecvEvt; + /** The transmit thread. */ + PPDMTHREAD pXmitThread; + /** The event semaphore that the transmit thread waits on. */ + SUPSEMEVENT hXmitEvt; + /** The support driver session handle. */ + PSUPDRVSESSION pSupDrvSession; + /** Scatter/gather descriptor cache. */ + RTMEMCACHE hSgCache; + /** Set if the link is down. + * When the link is down all incoming packets will be dropped. */ + bool volatile fLinkDown; + /** Set when the xmit thread has been signalled. (atomic) */ + bool volatile fXmitSignalled; + /** Set if the transmit thread the one busy transmitting. */ + bool volatile fXmitOnXmitThread; + /** The xmit thread should process the ring ASAP. */ + bool fXmitProcessRing; + /** Set if data transmission should start immediately and deactivate + * as late as possible. */ + bool fActivateEarlyDeactivateLate; + /** Padding. */ + bool afReserved[HC_ARCH_BITS == 64 ? 3 : 3]; + /** Scratch space for holding the ring-0 scatter / gather descriptor. + * The PDMSCATTERGATHER::fFlags member is used to indicate whether it is in + * use or not. Always accessed while owning the XmitLock. */ + union + { + PDMSCATTERGATHER Sg; + uint8_t padding[8 * sizeof(RTUINTPTR)]; + } u; + /** The network name. */ + char szNetwork[INTNET_MAX_NETWORK_NAME]; + + /** Number of GSO packets sent. */ + STAMCOUNTER StatSentGso; + /** Number of GSO packets received. */ + STAMCOUNTER StatReceivedGso; + /** Number of packets send from ring-0. */ + STAMCOUNTER StatSentR0; + /** The number of times we've had to wake up the xmit thread to continue the + * ring-0 job. */ + STAMCOUNTER StatXmitWakeupR0; + /** The number of times we've had to wake up the xmit thread to continue the + * ring-3 job. */ + STAMCOUNTER StatXmitWakeupR3; + /** The times the xmit thread has been told to process the ring. */ + STAMCOUNTER StatXmitProcessRing; +#ifdef VBOX_WITH_STATISTICS + /** Profiling packet transmit runs. */ + STAMPROFILE StatTransmit; + /** Profiling packet receive runs. */ + STAMPROFILEADV StatReceive; +#endif /* VBOX_WITH_STATISTICS */ +#ifdef LOG_ENABLED + /** The nano ts of the last transfer. */ + uint64_t u64LastTransferTS; + /** The nano ts of the last receive. */ + uint64_t u64LastReceiveTS; +#endif +#if defined(RT_OS_DARWIN) && defined(VBOX_WITH_INTNET_SERVICE_IN_R3) + /** XPC connection handle to the R3 internal network switch service. */ + xpc_connection_t hXpcCon; + /** Flag whether the R3 internal network service is being used. */ + bool fIntNetR3Svc; + /** Size of the communication buffer in bytes. */ + size_t cbBuf; +#endif +} DRVINTNET; +AssertCompileMemberAlignment(DRVINTNET, XmitLock, 8); +AssertCompileMemberAlignment(DRVINTNET, StatSentGso, 8); +/** Pointer to instance data of the internal networking driver. */ +typedef DRVINTNET *PDRVINTNET; + +/** + * Config value to flag translation structure. + */ +typedef struct DRVINTNETFLAG +{ + /** The value. */ + const char *pszChoice; + /** The corresponding flag. */ + uint32_t fFlag; +} DRVINTNETFLAG; +/** Pointer to a const flag value translation. */ +typedef DRVINTNETFLAG const *PCDRVINTNETFLAG; + + +#ifdef IN_RING3 + + +/** + * Calls the internal networking switch service living in either R0 or in another R3 process. + * + * @returns VBox status code. + * @param pThis The internal network driver instance data. + * @param uOperation The operation to execute. + * @param pvArg Pointer to the argument data. + * @param cbArg Size of the argument data in bytes. + */ +static int drvR3IntNetCallSvc(PDRVINTNET pThis, uint32_t uOperation, void *pvArg, unsigned cbArg) +{ +#if defined(RT_OS_DARWIN) && defined(VBOX_WITH_INTNET_SERVICE_IN_R3) + if (pThis->fIntNetR3Svc) + { + xpc_object_t hObj = xpc_dictionary_create(NULL, NULL, 0); + xpc_dictionary_set_uint64(hObj, "req-id", uOperation); + xpc_dictionary_set_data(hObj, "req", pvArg, cbArg); + xpc_object_t hObjReply = xpc_connection_send_message_with_reply_sync(pThis->hXpcCon, hObj); + uint64_t u64Rc = xpc_dictionary_get_uint64(hObjReply, "rc"); + if (INTNET_R3_SVC_IS_VALID_RC(u64Rc)) + { + size_t cbReply = 0; + const void *pvData = xpc_dictionary_get_data(hObjReply, "reply", &cbReply); + AssertRelease(cbReply == cbArg); + memcpy(pvArg, pvData, cbArg); + xpc_release(hObjReply); + + return INTNET_R3_SVC_GET_RC(u64Rc); + } + + xpc_release(hObjReply); + return VERR_INVALID_STATE; + } + else +#endif + return PDMDrvHlpSUPCallVMMR0Ex(pThis->pDrvInsR3, uOperation, pvArg, cbArg); +} + + +#if defined(RT_OS_DARWIN) && defined(VBOX_WITH_INTNET_SERVICE_IN_R3) +/** + * Calls the internal networking switch service living in either R0 or in another R3 process. + * + * @returns VBox status code. + * @param pThis The internal network driver instance data. + * @param uOperation The operation to execute. + * @param pvArg Pointer to the argument data. + * @param cbArg Size of the argument data in bytes. + */ +static int drvR3IntNetCallSvcAsync(PDRVINTNET pThis, uint32_t uOperation, void *pvArg, unsigned cbArg) +{ + if (pThis->fIntNetR3Svc) + { + xpc_object_t hObj = xpc_dictionary_create(NULL, NULL, 0); + xpc_dictionary_set_uint64(hObj, "req-id", uOperation); + xpc_dictionary_set_data(hObj, "req", pvArg, cbArg); + xpc_connection_send_message(pThis->hXpcCon, hObj); + return VINF_SUCCESS; + } + else + return PDMDrvHlpSUPCallVMMR0Ex(pThis->pDrvInsR3, uOperation, pvArg, cbArg); +} +#endif + + +/** + * Map the ring buffer pointer into this process R3 address space. + * + * @returns VBox status code. + * @param pThis The internal network driver instance data. + */ +static int drvR3IntNetMapBufferPointers(PDRVINTNET pThis) +{ + int rc = VINF_SUCCESS; + + INTNETIFGETBUFFERPTRSREQ GetBufferPtrsReq; + GetBufferPtrsReq.Hdr.u32Magic = SUPVMMR0REQHDR_MAGIC; + GetBufferPtrsReq.Hdr.cbReq = sizeof(GetBufferPtrsReq); + GetBufferPtrsReq.pSession = NIL_RTR0PTR; + GetBufferPtrsReq.hIf = pThis->hIf; + GetBufferPtrsReq.pRing3Buf = NULL; + GetBufferPtrsReq.pRing0Buf = NIL_RTR0PTR; + +#if defined(RT_OS_DARWIN) && defined(VBOX_WITH_INTNET_SERVICE_IN_R3) + if (pThis->fIntNetR3Svc) + { + xpc_object_t hObj = xpc_dictionary_create(NULL, NULL, 0); + xpc_dictionary_set_uint64(hObj, "req-id", VMMR0_DO_INTNET_IF_GET_BUFFER_PTRS); + xpc_dictionary_set_data(hObj, "req", &GetBufferPtrsReq, sizeof(GetBufferPtrsReq)); + xpc_object_t hObjReply = xpc_connection_send_message_with_reply_sync(pThis->hXpcCon, hObj); + uint64_t u64Rc = xpc_dictionary_get_uint64(hObjReply, "rc"); + if (INTNET_R3_SVC_IS_VALID_RC(u64Rc)) + rc = INTNET_R3_SVC_GET_RC(u64Rc); + else + rc = VERR_INVALID_STATE; + + if (RT_SUCCESS(rc)) + { + /* Get the shared memory object. */ + xpc_object_t hObjShMem = xpc_dictionary_get_value(hObjReply, "buf-ptr"); + size_t cbMem = xpc_shmem_map(hObjShMem, (void **)&pThis->pBufR3); + if (!cbMem) + rc = VERR_NO_MEMORY; + else + pThis->cbBuf = cbMem; + } + + xpc_release(hObjReply); + } + else +#endif + { + rc = PDMDrvHlpSUPCallVMMR0Ex(pThis->pDrvInsR3, VMMR0_DO_INTNET_IF_GET_BUFFER_PTRS, &GetBufferPtrsReq, sizeof(GetBufferPtrsReq)); + if (RT_SUCCESS(rc)) + { + AssertRelease(RT_VALID_PTR(GetBufferPtrsReq.pRing3Buf)); + pThis->pBufR3 = GetBufferPtrsReq.pRing3Buf; +#ifdef VBOX_WITH_DRVINTNET_IN_R0 + pThis->pBufR0 = GetBufferPtrsReq.pRing0Buf; +#endif + } + } + + return rc; +} + + +/** + * Updates the MAC address on the kernel side. + * + * @returns VBox status code. + * @param pThis The driver instance. + */ +static int drvR3IntNetUpdateMacAddress(PDRVINTNET pThis) +{ + if (!pThis->pIAboveConfigR3) + return VINF_SUCCESS; + + INTNETIFSETMACADDRESSREQ SetMacAddressReq; + SetMacAddressReq.Hdr.u32Magic = SUPVMMR0REQHDR_MAGIC; + SetMacAddressReq.Hdr.cbReq = sizeof(SetMacAddressReq); + SetMacAddressReq.pSession = NIL_RTR0PTR; + SetMacAddressReq.hIf = pThis->hIf; + int rc = pThis->pIAboveConfigR3->pfnGetMac(pThis->pIAboveConfigR3, &SetMacAddressReq.Mac); + if (RT_SUCCESS(rc)) + rc = drvR3IntNetCallSvc(pThis, VMMR0_DO_INTNET_IF_SET_MAC_ADDRESS, + &SetMacAddressReq, sizeof(SetMacAddressReq)); + + Log(("drvR3IntNetUpdateMacAddress: %.*Rhxs rc=%Rrc\n", sizeof(SetMacAddressReq.Mac), &SetMacAddressReq.Mac, rc)); + return rc; +} + + +/** + * Sets the kernel interface active or inactive. + * + * Worker for poweron, poweroff, suspend and resume. + * + * @returns VBox status code. + * @param pThis The driver instance. + * @param fActive The new state. + */ +static int drvR3IntNetSetActive(PDRVINTNET pThis, bool fActive) +{ + if (!pThis->pIAboveConfigR3) + return VINF_SUCCESS; + + INTNETIFSETACTIVEREQ SetActiveReq; + SetActiveReq.Hdr.u32Magic = SUPVMMR0REQHDR_MAGIC; + SetActiveReq.Hdr.cbReq = sizeof(SetActiveReq); + SetActiveReq.pSession = NIL_RTR0PTR; + SetActiveReq.hIf = pThis->hIf; + SetActiveReq.fActive = fActive; + int rc = drvR3IntNetCallSvc(pThis, VMMR0_DO_INTNET_IF_SET_ACTIVE, + &SetActiveReq, sizeof(SetActiveReq)); + + Log(("drvR3IntNetSetActive: fActive=%d rc=%Rrc\n", fActive, rc)); + AssertRC(rc); + return rc; +} + +#endif /* IN_RING3 */ + +/* -=-=-=-=- PDMINETWORKUP -=-=-=-=- */ + +#ifndef IN_RING3 +/** + * Helper for signalling the xmit thread. + * + * @returns VERR_TRY_AGAIN (convenience). + * @param pThis The instance data.. + */ +DECLINLINE(int) drvR0IntNetSignalXmit(PDRVINTNET pThis) +{ + /// @todo if (!ASMAtomicXchgBool(&pThis->fXmitSignalled, true)) - needs careful optimizing. + { + int rc = SUPSemEventSignal(pThis->pSupDrvSession, pThis->hXmitEvt); + AssertRC(rc); + STAM_REL_COUNTER_INC(&pThis->CTX_SUFF(StatXmitWakeup)); + } + return VERR_TRY_AGAIN; +} +#endif /* !IN_RING3 */ + + +/** + * Helper for processing the ring-0 consumer side of the xmit ring. + * + * The caller MUST own the xmit lock. + * + * @returns Status code from IntNetR0IfSend, except for VERR_TRY_AGAIN. + * @param pThis The instance data.. + */ +DECLINLINE(int) drvIntNetProcessXmit(PDRVINTNET pThis) +{ + Assert(PDMDrvHlpCritSectIsOwner(pThis->CTX_SUFF(pDrvIns), &pThis->XmitLock)); + +#ifdef IN_RING3 + INTNETIFSENDREQ SendReq; + SendReq.Hdr.u32Magic = SUPVMMR0REQHDR_MAGIC; + SendReq.Hdr.cbReq = sizeof(SendReq); + SendReq.pSession = NIL_RTR0PTR; + SendReq.hIf = pThis->hIf; + int rc = drvR3IntNetCallSvc(pThis, VMMR0_DO_INTNET_IF_SEND, &SendReq, sizeof(SendReq)); +#else + int rc = IntNetR0IfSend(pThis->hIf, pThis->pSupDrvSession); + if (rc == VERR_TRY_AGAIN) + { + ASMAtomicUoWriteBool(&pThis->fXmitProcessRing, true); + drvR0IntNetSignalXmit(pThis); + rc = VINF_SUCCESS; + } +#endif + return rc; +} + + + +/** + * @interface_method_impl{PDMINETWORKUP,pfnBeginXmit} + */ +PDMBOTHCBDECL(int) drvIntNetUp_BeginXmit(PPDMINETWORKUP pInterface, bool fOnWorkerThread) +{ + PDRVINTNET pThis = RT_FROM_MEMBER(pInterface, DRVINTNET, CTX_SUFF(INetworkUp)); +#ifndef IN_RING3 + Assert(!fOnWorkerThread); +#endif + + int rc = PDMDrvHlpCritSectTryEnter(pThis->CTX_SUFF(pDrvIns), &pThis->XmitLock); + if (RT_SUCCESS(rc)) + { + if (fOnWorkerThread) + { + ASMAtomicUoWriteBool(&pThis->fXmitOnXmitThread, true); + ASMAtomicWriteBool(&pThis->fXmitSignalled, false); + } + } + else if (rc == VERR_SEM_BUSY) + { + /** @todo Does this actually make sense if the other dude is an EMT and so + * forth? I seriously think this is ring-0 only... + * We might end up waking up the xmit thread unnecessarily here, even when in + * ring-0... This needs some more thought and optimizations when the ring-0 bits + * are working. */ +#ifdef IN_RING3 + if ( !fOnWorkerThread + /*&& !ASMAtomicUoReadBool(&pThis->fXmitOnXmitThread) + && ASMAtomicCmpXchgBool(&pThis->fXmitSignalled, true, false)*/) + { + rc = SUPSemEventSignal(pThis->pSupDrvSession, pThis->hXmitEvt); + AssertRC(rc); + } + rc = VERR_TRY_AGAIN; +#else /* IN_RING0 */ + rc = drvR0IntNetSignalXmit(pThis); +#endif /* IN_RING0 */ + } + return rc; +} + + +/** + * @interface_method_impl{PDMINETWORKUP,pfnAllocBuf} + */ +PDMBOTHCBDECL(int) drvIntNetUp_AllocBuf(PPDMINETWORKUP pInterface, size_t cbMin, + PCPDMNETWORKGSO pGso, PPPDMSCATTERGATHER ppSgBuf) +{ + PDRVINTNET pThis = RT_FROM_MEMBER(pInterface, DRVINTNET, CTX_SUFF(INetworkUp)); + int rc = VINF_SUCCESS; + Assert(cbMin < UINT32_MAX / 2); + Assert(PDMDrvHlpCritSectIsOwner(pThis->CTX_SUFF(pDrvIns), &pThis->XmitLock)); + + /* + * Allocate a S/G descriptor. + * This shouldn't normally fail as the NICs usually won't allocate more + * than one buffer at a time and the SG gets freed on sending. + */ +#ifdef IN_RING3 + PPDMSCATTERGATHER pSgBuf = (PPDMSCATTERGATHER)RTMemCacheAlloc(pThis->hSgCache); + if (!pSgBuf) + return VERR_NO_MEMORY; +#else + PPDMSCATTERGATHER pSgBuf = &pThis->u.Sg; + if (RT_UNLIKELY(pSgBuf->fFlags != 0)) + return drvR0IntNetSignalXmit(pThis); +#endif + + /* + * Allocate room in the ring buffer. + * + * In ring-3 we may have to process the xmit ring before there is + * sufficient buffer space since we might have stacked up a few frames to the + * trunk while in ring-0. (There is not point of doing this in ring-0.) + */ + PINTNETHDR pHdr = NULL; /* gcc silliness */ + if (pGso) + rc = IntNetRingAllocateGsoFrame(&pThis->CTX_SUFF(pBuf)->Send, (uint32_t)cbMin, pGso, + &pHdr, &pSgBuf->aSegs[0].pvSeg); + else + rc = IntNetRingAllocateFrame(&pThis->CTX_SUFF(pBuf)->Send, (uint32_t)cbMin, + &pHdr, &pSgBuf->aSegs[0].pvSeg); +#ifdef IN_RING3 + if ( RT_FAILURE(rc) + && pThis->CTX_SUFF(pBuf)->cbSend >= cbMin * 2 + sizeof(INTNETHDR)) + { + drvIntNetProcessXmit(pThis); + if (pGso) + rc = IntNetRingAllocateGsoFrame(&pThis->CTX_SUFF(pBuf)->Send, (uint32_t)cbMin, pGso, + &pHdr, &pSgBuf->aSegs[0].pvSeg); + else + rc = IntNetRingAllocateFrame(&pThis->CTX_SUFF(pBuf)->Send, (uint32_t)cbMin, + &pHdr, &pSgBuf->aSegs[0].pvSeg); + } +#endif + if (RT_SUCCESS(rc)) + { + /* + * Set up the S/G descriptor and return successfully. + */ + pSgBuf->fFlags = PDMSCATTERGATHER_FLAGS_MAGIC | PDMSCATTERGATHER_FLAGS_OWNER_1; + pSgBuf->cbUsed = 0; + pSgBuf->cbAvailable = cbMin; + pSgBuf->pvAllocator = pHdr; + pSgBuf->pvUser = pGso ? (PPDMNETWORKGSO)pSgBuf->aSegs[0].pvSeg - 1 : NULL; + pSgBuf->cSegs = 1; + pSgBuf->aSegs[0].cbSeg = cbMin; + + *ppSgBuf = pSgBuf; + return VINF_SUCCESS; + } + +#ifdef IN_RING3 + /* + * If the above fails, then we're really out of space. There are nobody + * competing with us here because of the xmit lock. + */ + rc = VERR_NO_MEMORY; + RTMemCacheFree(pThis->hSgCache, pSgBuf); + +#else /* IN_RING0 */ + /* + * If the request is reasonable, kick the xmit thread and tell it to + * process the xmit ring ASAP. + */ + if (pThis->CTX_SUFF(pBuf)->cbSend >= cbMin * 2 + sizeof(INTNETHDR)) + { + pThis->fXmitProcessRing = true; + rc = drvR0IntNetSignalXmit(pThis); + } + else + rc = VERR_NO_MEMORY; + pSgBuf->fFlags = 0; +#endif /* IN_RING0 */ + return rc; +} + + +/** + * @interface_method_impl{PDMINETWORKUP,pfnFreeBuf} + */ +PDMBOTHCBDECL(int) drvIntNetUp_FreeBuf(PPDMINETWORKUP pInterface, PPDMSCATTERGATHER pSgBuf) +{ + PDRVINTNET pThis = RT_FROM_MEMBER(pInterface, DRVINTNET, CTX_SUFF(INetworkUp)); + PINTNETHDR pHdr = (PINTNETHDR)pSgBuf->pvAllocator; +#ifdef IN_RING0 + Assert(pSgBuf == &pThis->u.Sg); +#endif + Assert(pSgBuf->fFlags == (PDMSCATTERGATHER_FLAGS_MAGIC | PDMSCATTERGATHER_FLAGS_OWNER_1)); + Assert(pSgBuf->cbUsed <= pSgBuf->cbAvailable); + Assert( pHdr->u8Type == INTNETHDR_TYPE_FRAME + || pHdr->u8Type == INTNETHDR_TYPE_GSO); + Assert(PDMDrvHlpCritSectIsOwner(pThis->CTX_SUFF(pDrvIns), &pThis->XmitLock)); + + /** @todo LATER: try unalloc the frame. */ + pHdr->u8Type = INTNETHDR_TYPE_PADDING; + IntNetRingCommitFrame(&pThis->CTX_SUFF(pBuf)->Send, pHdr); + +#ifdef IN_RING3 + RTMemCacheFree(pThis->hSgCache, pSgBuf); +#else + pSgBuf->fFlags = 0; +#endif + return VINF_SUCCESS; +} + + +/** + * @interface_method_impl{PDMINETWORKUP,pfnSendBuf} + */ +PDMBOTHCBDECL(int) drvIntNetUp_SendBuf(PPDMINETWORKUP pInterface, PPDMSCATTERGATHER pSgBuf, bool fOnWorkerThread) +{ + PDRVINTNET pThis = RT_FROM_MEMBER(pInterface, DRVINTNET, CTX_SUFF(INetworkUp)); + STAM_PROFILE_START(&pThis->StatTransmit, a); + RT_NOREF_PV(fOnWorkerThread); + + AssertPtr(pSgBuf); + Assert(pSgBuf->fFlags == (PDMSCATTERGATHER_FLAGS_MAGIC | PDMSCATTERGATHER_FLAGS_OWNER_1)); + Assert(pSgBuf->cbUsed <= pSgBuf->cbAvailable); + Assert(PDMDrvHlpCritSectIsOwner(pThis->CTX_SUFF(pDrvIns), &pThis->XmitLock)); + + if (pSgBuf->pvUser) + STAM_COUNTER_INC(&pThis->StatSentGso); + + /* + * Commit the frame and push it thru the switch. + */ + PINTNETHDR pHdr = (PINTNETHDR)pSgBuf->pvAllocator; + IntNetRingCommitFrameEx(&pThis->CTX_SUFF(pBuf)->Send, pHdr, pSgBuf->cbUsed); + int rc = drvIntNetProcessXmit(pThis); + STAM_PROFILE_STOP(&pThis->StatTransmit, a); + + /* + * Free the descriptor and return. + */ +#ifdef IN_RING3 + RTMemCacheFree(pThis->hSgCache, pSgBuf); +#else + STAM_REL_COUNTER_INC(&pThis->StatSentR0); + pSgBuf->fFlags = 0; +#endif + return rc; +} + + +/** + * @interface_method_impl{PDMINETWORKUP,pfnEndXmit} + */ +PDMBOTHCBDECL(void) drvIntNetUp_EndXmit(PPDMINETWORKUP pInterface) +{ + PDRVINTNET pThis = RT_FROM_MEMBER(pInterface, DRVINTNET, CTX_SUFF(INetworkUp)); + ASMAtomicUoWriteBool(&pThis->fXmitOnXmitThread, false); + PDMDrvHlpCritSectLeave(pThis->CTX_SUFF(pDrvIns), &pThis->XmitLock); +} + + +/** + * @interface_method_impl{PDMINETWORKUP,pfnSetPromiscuousMode} + */ +PDMBOTHCBDECL(void) drvIntNetUp_SetPromiscuousMode(PPDMINETWORKUP pInterface, bool fPromiscuous) +{ + PDRVINTNET pThis = RT_FROM_MEMBER(pInterface, DRVINTNET, CTX_SUFF(INetworkUp)); + +#ifdef IN_RING3 + INTNETIFSETPROMISCUOUSMODEREQ Req; + Req.Hdr.u32Magic = SUPVMMR0REQHDR_MAGIC; + Req.Hdr.cbReq = sizeof(Req); + Req.pSession = NIL_RTR0PTR; + Req.hIf = pThis->hIf; + Req.fPromiscuous = fPromiscuous; + int rc = drvR3IntNetCallSvc(pThis, VMMR0_DO_INTNET_IF_SET_PROMISCUOUS_MODE, &Req, sizeof(Req)); +#else /* IN_RING0 */ + int rc = IntNetR0IfSetPromiscuousMode(pThis->hIf, pThis->pSupDrvSession, fPromiscuous); +#endif /* IN_RING0 */ + + LogFlow(("drvIntNetUp_SetPromiscuousMode: fPromiscuous=%RTbool\n", fPromiscuous)); + AssertRC(rc); +} + +#ifdef IN_RING3 + +/** + * @interface_method_impl{PDMINETWORKUP,pfnNotifyLinkChanged} + */ +static DECLCALLBACK(void) drvR3IntNetUp_NotifyLinkChanged(PPDMINETWORKUP pInterface, PDMNETWORKLINKSTATE enmLinkState) +{ + PDRVINTNET pThis = RT_FROM_MEMBER(pInterface, DRVINTNET, CTX_SUFF(INetworkUp)); + bool fLinkDown; + switch (enmLinkState) + { + case PDMNETWORKLINKSTATE_DOWN: + case PDMNETWORKLINKSTATE_DOWN_RESUME: + fLinkDown = true; + break; + default: + AssertMsgFailed(("enmLinkState=%d\n", enmLinkState)); + RT_FALL_THRU(); + case PDMNETWORKLINKSTATE_UP: + fLinkDown = false; + break; + } + LogFlow(("drvR3IntNetUp_NotifyLinkChanged: enmLinkState=%d %d->%d\n", enmLinkState, pThis->fLinkDown, fLinkDown)); + ASMAtomicXchgSize(&pThis->fLinkDown, fLinkDown); +} + + +/* -=-=-=-=- Transmit Thread -=-=-=-=- */ + +/** + * Async I/O thread for deferred packet transmission. + * + * @returns VBox status code. Returning failure will naturally terminate the thread. + * @param pDrvIns The internal networking driver instance. + * @param pThread The thread. + */ +static DECLCALLBACK(int) drvR3IntNetXmitThread(PPDMDRVINS pDrvIns, PPDMTHREAD pThread) +{ + PDRVINTNET pThis = PDMINS_2_DATA(pDrvIns, PDRVINTNET); + + while (pThread->enmState == PDMTHREADSTATE_RUNNING) + { + /* + * Transmit any pending packets. + */ + /** @todo Optimize this. We shouldn't call pfnXmitPending unless asked for. + * Also there is no need to call drvIntNetProcessXmit if we also + * called pfnXmitPending and send one or more frames. */ + if (ASMAtomicXchgBool(&pThis->fXmitProcessRing, false)) + { + STAM_REL_COUNTER_INC(&pThis->StatXmitProcessRing); + PDMDrvHlpCritSectEnter(pDrvIns, &pThis->XmitLock, VERR_IGNORED); + drvIntNetProcessXmit(pThis); + PDMDrvHlpCritSectLeave(pDrvIns, &pThis->XmitLock); + } + + pThis->pIAboveNet->pfnXmitPending(pThis->pIAboveNet); + + if (ASMAtomicXchgBool(&pThis->fXmitProcessRing, false)) + { + STAM_REL_COUNTER_INC(&pThis->StatXmitProcessRing); + PDMDrvHlpCritSectEnter(pDrvIns, &pThis->XmitLock, VERR_IGNORED); + drvIntNetProcessXmit(pThis); + PDMDrvHlpCritSectLeave(pDrvIns, &pThis->XmitLock); + } + + /* + * Block until we've got something to send or is supposed + * to leave the running state. + */ + int rc = SUPSemEventWaitNoResume(pThis->pSupDrvSession, pThis->hXmitEvt, RT_INDEFINITE_WAIT); + AssertLogRelMsgReturn(RT_SUCCESS(rc) || rc == VERR_INTERRUPTED, ("%Rrc\n", rc), rc); + if (RT_UNLIKELY(pThread->enmState != PDMTHREADSTATE_RUNNING)) + break; + + } + + /* The thread is being initialized, suspended or terminated. */ + return VINF_SUCCESS; +} + + +/** + * @copydoc FNPDMTHREADWAKEUPDRV + */ +static DECLCALLBACK(int) drvR3IntNetXmitWakeUp(PPDMDRVINS pDrvIns, PPDMTHREAD pThread) +{ + RT_NOREF(pThread); + PDRVINTNET pThis = PDMINS_2_DATA(pDrvIns, PDRVINTNET); + return SUPSemEventSignal(pThis->pSupDrvSession, pThis->hXmitEvt); +} + + +/* -=-=-=-=- Receive Thread -=-=-=-=- */ + +/** + * Wait for space to become available up the driver/device chain. + * + * @returns VINF_SUCCESS if space is available. + * @returns VERR_STATE_CHANGED if the state changed. + * @returns VBox status code on other errors. + * @param pThis Pointer to the instance data. + */ +static int drvR3IntNetRecvWaitForSpace(PDRVINTNET pThis) +{ + LogFlow(("drvR3IntNetRecvWaitForSpace:\n")); + STAM_PROFILE_ADV_STOP(&pThis->StatReceive, a); + int rc = pThis->pIAboveNet->pfnWaitReceiveAvail(pThis->pIAboveNet, RT_INDEFINITE_WAIT); + STAM_PROFILE_ADV_START(&pThis->StatReceive, a); + LogFlow(("drvR3IntNetRecvWaitForSpace: returns %Rrc\n", rc)); + return rc; +} + + +/** + * Executes async I/O (RUNNING mode). + * + * @returns VERR_STATE_CHANGED if the state changed. + * @returns Appropriate VBox status code (error) on fatal error. + * @param pThis The driver instance data. + */ +static int drvR3IntNetRecvRun(PDRVINTNET pThis) +{ + LogFlow(("drvR3IntNetRecvRun: pThis=%p\n", pThis)); + + /* + * The running loop - processing received data and waiting for more to arrive. + */ + STAM_PROFILE_ADV_START(&pThis->StatReceive, a); + PINTNETBUF pBuf = pThis->CTX_SUFF(pBuf); + PINTNETRINGBUF pRingBuf = &pBuf->Recv; + for (;;) + { + /* + * Process the receive buffer. + */ + PINTNETHDR pHdr; + while ((pHdr = IntNetRingGetNextFrameToRead(pRingBuf)) != NULL) + { + /* + * Check the state and then inspect the packet. + */ + if (pThis->enmRecvState != RECVSTATE_RUNNING) + { + STAM_PROFILE_ADV_STOP(&pThis->StatReceive, a); + LogFlow(("drvR3IntNetRecvRun: returns VERR_STATE_CHANGED (state changed - #0)\n")); + return VERR_STATE_CHANGED; + } + + Log2(("pHdr=%p offRead=%#x: %.8Rhxs\n", pHdr, pRingBuf->offReadX, pHdr)); + uint8_t u8Type = pHdr->u8Type; + if ( ( u8Type == INTNETHDR_TYPE_FRAME + || u8Type == INTNETHDR_TYPE_GSO) + && !pThis->fLinkDown) + { + /* + * Check if there is room for the frame and pass it up. + */ + size_t cbFrame = pHdr->cbFrame; + int rc = pThis->pIAboveNet->pfnWaitReceiveAvail(pThis->pIAboveNet, 0); + if (rc == VINF_SUCCESS) + { + if (u8Type == INTNETHDR_TYPE_FRAME) + { + /* + * Normal frame. + */ +#ifdef LOG_ENABLED + if (LogIsEnabled()) + { + uint64_t u64Now = RTTimeProgramNanoTS(); + LogFlow(("drvR3IntNetRecvRun: %-4d bytes at %llu ns deltas: r=%llu t=%llu\n", + cbFrame, u64Now, u64Now - pThis->u64LastReceiveTS, u64Now - pThis->u64LastTransferTS)); + pThis->u64LastReceiveTS = u64Now; + Log2(("drvR3IntNetRecvRun: cbFrame=%#x\n" + "%.*Rhxd\n", + cbFrame, cbFrame, IntNetHdrGetFramePtr(pHdr, pBuf))); + } +#endif + rc = pThis->pIAboveNet->pfnReceive(pThis->pIAboveNet, IntNetHdrGetFramePtr(pHdr, pBuf), cbFrame); + AssertRC(rc); + + /* skip to the next frame. */ + IntNetRingSkipFrame(pRingBuf); + } + else + { + /* + * Generic segment offload frame (INTNETHDR_TYPE_GSO). + */ + STAM_COUNTER_INC(&pThis->StatReceivedGso); + PCPDMNETWORKGSO pGso = IntNetHdrGetGsoContext(pHdr, pBuf); + if (PDMNetGsoIsValid(pGso, cbFrame, cbFrame - sizeof(PDMNETWORKGSO))) + { + if ( !pThis->pIAboveNet->pfnReceiveGso + || RT_FAILURE(pThis->pIAboveNet->pfnReceiveGso(pThis->pIAboveNet, + (uint8_t *)(pGso + 1), + pHdr->cbFrame - sizeof(PDMNETWORKGSO), + pGso))) + { + /* + * This is where we do the offloading since this NIC + * does not support large receive offload (LRO). + */ + cbFrame -= sizeof(PDMNETWORKGSO); + + uint8_t abHdrScratch[256]; + uint32_t const cSegs = PDMNetGsoCalcSegmentCount(pGso, cbFrame); +#ifdef LOG_ENABLED + if (LogIsEnabled()) + { + uint64_t u64Now = RTTimeProgramNanoTS(); + LogFlow(("drvR3IntNetRecvRun: %-4d bytes at %llu ns deltas: r=%llu t=%llu; GSO - %u segs\n", + cbFrame, u64Now, u64Now - pThis->u64LastReceiveTS, u64Now - pThis->u64LastTransferTS, cSegs)); + pThis->u64LastReceiveTS = u64Now; + Log2(("drvR3IntNetRecvRun: cbFrame=%#x type=%d cbHdrsTotal=%#x cbHdrsSeg=%#x Hdr1=%#x Hdr2=%#x MMS=%#x\n" + "%.*Rhxd\n", + cbFrame, pGso->u8Type, pGso->cbHdrsTotal, pGso->cbHdrsSeg, pGso->offHdr1, pGso->offHdr2, pGso->cbMaxSeg, + cbFrame - sizeof(*pGso), pGso + 1)); + } +#endif + for (uint32_t iSeg = 0; iSeg < cSegs; iSeg++) + { + uint32_t cbSegFrame; + void *pvSegFrame = PDMNetGsoCarveSegmentQD(pGso, (uint8_t *)(pGso + 1), cbFrame, + abHdrScratch, iSeg, cSegs, &cbSegFrame); + rc = drvR3IntNetRecvWaitForSpace(pThis); + if (RT_FAILURE(rc)) + { + Log(("drvR3IntNetRecvRun: drvR3IntNetRecvWaitForSpace -> %Rrc; iSeg=%u cSegs=%u\n", rc, iSeg, cSegs)); + break; /* we drop the rest. */ + } + rc = pThis->pIAboveNet->pfnReceive(pThis->pIAboveNet, pvSegFrame, cbSegFrame); + AssertRC(rc); + } + } + } + else + { + AssertMsgFailed(("cbFrame=%#x type=%d cbHdrsTotal=%#x cbHdrsSeg=%#x Hdr1=%#x Hdr2=%#x MMS=%#x\n", + cbFrame, pGso->u8Type, pGso->cbHdrsTotal, pGso->cbHdrsSeg, pGso->offHdr1, pGso->offHdr2, pGso->cbMaxSeg)); + STAM_REL_COUNTER_INC(&pBuf->cStatBadFrames); + } + + IntNetRingSkipFrame(pRingBuf); + } + } + else + { + /* + * Wait for sufficient space to become available and then retry. + */ + rc = drvR3IntNetRecvWaitForSpace(pThis); + if (RT_FAILURE(rc)) + { + if (rc == VERR_INTERRUPTED) + { + /* + * NIC is going down, likely because the VM is being reset. Skip the frame. + */ + AssertMsg(IntNetIsValidFrameType(pHdr->u8Type), ("Unknown frame type %RX16! offRead=%#x\n", pHdr->u8Type, pRingBuf->offReadX)); + IntNetRingSkipFrame(pRingBuf); + } + else + { + STAM_PROFILE_ADV_STOP(&pThis->StatReceive, a); + LogFlow(("drvR3IntNetRecvRun: returns %Rrc (wait-for-space)\n", rc)); + return rc; + } + } + } + } + else + { + /* + * Link down or unknown frame - skip to the next frame. + */ + AssertMsg(IntNetIsValidFrameType(pHdr->u8Type), ("Unknown frame type %RX16! offRead=%#x\n", pHdr->u8Type, pRingBuf->offReadX)); + IntNetRingSkipFrame(pRingBuf); + STAM_REL_COUNTER_INC(&pBuf->cStatBadFrames); + } + } /* while more received data */ + + /* + * Wait for data, checking the state before we block. + */ + if (pThis->enmRecvState != RECVSTATE_RUNNING) + { + STAM_PROFILE_ADV_STOP(&pThis->StatReceive, a); + LogFlow(("drvR3IntNetRecvRun: returns VINF_SUCCESS (state changed - #1)\n")); + return VERR_STATE_CHANGED; + } + INTNETIFWAITREQ WaitReq; + WaitReq.Hdr.u32Magic = SUPVMMR0REQHDR_MAGIC; + WaitReq.Hdr.cbReq = sizeof(WaitReq); + WaitReq.pSession = NIL_RTR0PTR; + WaitReq.hIf = pThis->hIf; + WaitReq.cMillies = 30000; /* 30s - don't wait forever, timeout now and then. */ + STAM_PROFILE_ADV_STOP(&pThis->StatReceive, a); + +#if defined(RT_OS_DARWIN) && defined(VBOX_WITH_INTNET_SERVICE_IN_R3) + if (pThis->fIntNetR3Svc) + { + /* Send an asynchronous message. */ + int rc = drvR3IntNetCallSvcAsync(pThis, VMMR0_DO_INTNET_IF_WAIT, &WaitReq, sizeof(WaitReq)); + if (RT_SUCCESS(rc)) + { + /* Wait on the receive semaphore. */ + rc = RTSemEventWait(pThis->hRecvEvt, 30 * RT_MS_1SEC); + if ( RT_FAILURE(rc) + && rc != VERR_TIMEOUT + && rc != VERR_INTERRUPTED) + { + LogFlow(("drvR3IntNetRecvRun: returns %Rrc\n", rc)); + return rc; + } + } + } + else +#endif + { + int rc = PDMDrvHlpSUPCallVMMR0Ex(pThis->pDrvInsR3, VMMR0_DO_INTNET_IF_WAIT, &WaitReq, sizeof(WaitReq)); + if ( RT_FAILURE(rc) + && rc != VERR_TIMEOUT + && rc != VERR_INTERRUPTED) + { + LogFlow(("drvR3IntNetRecvRun: returns %Rrc\n", rc)); + return rc; + } + } + STAM_PROFILE_ADV_START(&pThis->StatReceive, a); + } +} + + +/** + * Asynchronous I/O thread for handling receive. + * + * @returns VINF_SUCCESS (ignored). + * @param hThreadSelf Thread handle. + * @param pvUser Pointer to a DRVINTNET structure. + */ +static DECLCALLBACK(int) drvR3IntNetRecvThread(RTTHREAD hThreadSelf, void *pvUser) +{ + RT_NOREF(hThreadSelf); + PDRVINTNET pThis = (PDRVINTNET)pvUser; + LogFlow(("drvR3IntNetRecvThread: pThis=%p\n", pThis)); + STAM_PROFILE_ADV_START(&pThis->StatReceive, a); + + /* + * The main loop - acting on state. + */ + for (;;) + { + RECVSTATE enmRecvState = pThis->enmRecvState; + switch (enmRecvState) + { + case RECVSTATE_SUSPENDED: + { + int rc = RTSemEventWait(pThis->hRecvEvt, 30000); + if ( RT_FAILURE(rc) + && rc != VERR_TIMEOUT) + { + LogFlow(("drvR3IntNetRecvThread: returns %Rrc\n", rc)); + return rc; + } + break; + } + + case RECVSTATE_RUNNING: + { + int rc = drvR3IntNetRecvRun(pThis); + if ( rc != VERR_STATE_CHANGED + && RT_FAILURE(rc)) + { + LogFlow(("drvR3IntNetRecvThread: returns %Rrc\n", rc)); + return rc; + } + break; + } + + default: + AssertMsgFailed(("Invalid state %d\n", enmRecvState)); + RT_FALL_THRU(); + case RECVSTATE_TERMINATE: + LogFlow(("drvR3IntNetRecvThread: returns VINF_SUCCESS\n")); + return VINF_SUCCESS; + } + } +} + + +#ifdef VBOX_WITH_DRVINTNET_IN_R0 + +/* -=-=-=-=- PDMIBASERC -=-=-=-=- */ + +/** + * @interface_method_impl{PDMIBASERC,pfnQueryInterface} + */ +static DECLCALLBACK(RTRCPTR) drvR3IntNetIBaseRC_QueryInterface(PPDMIBASERC pInterface, const char *pszIID) +{ + PDRVINTNET pThis = RT_FROM_MEMBER(pInterface, DRVINTNET, IBaseRC); +#if 0 + PDMIBASERC_RETURN_INTERFACE(pThis->pDrvInsR3, pszIID, PDMINETWORKUP, &pThis->INetworkUpRC); +#else + RT_NOREF(pThis, pszIID); +#endif + return NIL_RTRCPTR; +} + + +/* -=-=-=-=- PDMIBASER0 -=-=-=-=- */ + +/** + * @interface_method_impl{PDMIBASER0,pfnQueryInterface} + */ +static DECLCALLBACK(RTR0PTR) drvR3IntNetIBaseR0_QueryInterface(PPDMIBASER0 pInterface, const char *pszIID) +{ + PDRVINTNET pThis = RT_FROM_MEMBER(pInterface, DRVINTNET, IBaseR0); + PDMIBASER0_RETURN_INTERFACE(pThis->pDrvInsR3, pszIID, PDMINETWORKUP, &pThis->INetworkUpR0); + return NIL_RTR0PTR; +} + +#endif /* VBOX_WITH_DRVINTNET_IN_R0 */ + +/* -=-=-=-=- PDMIBASE -=-=-=-=- */ + + +/** + * @interface_method_impl{PDMIBASE,pfnQueryInterface} + */ +static DECLCALLBACK(void *) drvR3IntNetIBase_QueryInterface(PPDMIBASE pInterface, const char *pszIID) +{ + PPDMDRVINS pDrvIns = PDMIBASE_2_PDMDRV(pInterface); + PDRVINTNET pThis = PDMINS_2_DATA(pDrvIns, PDRVINTNET); + + PDMIBASE_RETURN_INTERFACE(pszIID, PDMIBASE, &pDrvIns->IBase); +#ifdef VBOX_WITH_DRVINTNET_IN_R0 + PDMIBASE_RETURN_INTERFACE(pszIID, PDMIBASER0, &pThis->IBaseR0); + PDMIBASE_RETURN_INTERFACE(pszIID, PDMIBASERC, &pThis->IBaseRC); +#endif + PDMIBASE_RETURN_INTERFACE(pszIID, PDMINETWORKUP, &pThis->INetworkUpR3); + return NULL; +} + + +/* -=-=-=-=- PDMDRVREG -=-=-=-=- */ + +/** + * Power Off notification. + * + * @param pDrvIns The driver instance. + */ +static DECLCALLBACK(void) drvR3IntNetPowerOff(PPDMDRVINS pDrvIns) +{ + LogFlow(("drvR3IntNetPowerOff\n")); + PDRVINTNET pThis = PDMINS_2_DATA(pDrvIns, PDRVINTNET); + if (!pThis->fActivateEarlyDeactivateLate) + { + ASMAtomicXchgSize(&pThis->enmRecvState, RECVSTATE_SUSPENDED); + drvR3IntNetSetActive(pThis, false /* fActive */); + } +} + + +/** + * drvR3IntNetResume helper. + */ +static int drvR3IntNetResumeSend(PDRVINTNET pThis, const void *pvBuf, size_t cb) +{ + /* + * Add the frame to the send buffer and push it onto the network. + */ + int rc = IntNetRingWriteFrame(&pThis->pBufR3->Send, pvBuf, (uint32_t)cb); + if ( rc == VERR_BUFFER_OVERFLOW + && pThis->pBufR3->cbSend < cb) + { + INTNETIFSENDREQ SendReq; + SendReq.Hdr.u32Magic = SUPVMMR0REQHDR_MAGIC; + SendReq.Hdr.cbReq = sizeof(SendReq); + SendReq.pSession = NIL_RTR0PTR; + SendReq.hIf = pThis->hIf; + drvR3IntNetCallSvc(pThis, VMMR0_DO_INTNET_IF_SEND, &SendReq, sizeof(SendReq)); + + rc = IntNetRingWriteFrame(&pThis->pBufR3->Send, pvBuf, (uint32_t)cb); + } + + if (RT_SUCCESS(rc)) + { + INTNETIFSENDREQ SendReq; + SendReq.Hdr.u32Magic = SUPVMMR0REQHDR_MAGIC; + SendReq.Hdr.cbReq = sizeof(SendReq); + SendReq.pSession = NIL_RTR0PTR; + SendReq.hIf = pThis->hIf; + rc = drvR3IntNetCallSvc(pThis, VMMR0_DO_INTNET_IF_SEND, &SendReq, sizeof(SendReq)); + } + + AssertRC(rc); + return rc; +} + + +/** + * Resume notification. + * + * @param pDrvIns The driver instance. + */ +static DECLCALLBACK(void) drvR3IntNetResume(PPDMDRVINS pDrvIns) +{ + LogFlow(("drvR3IntNetPowerResume\n")); + PDRVINTNET pThis = PDMINS_2_DATA(pDrvIns, PDRVINTNET); + VMRESUMEREASON enmReason = PDMDrvHlpVMGetResumeReason(pDrvIns); + + if (!pThis->fActivateEarlyDeactivateLate) + { + ASMAtomicXchgSize(&pThis->enmRecvState, RECVSTATE_RUNNING); + RTSemEventSignal(pThis->hRecvEvt); + drvR3IntNetUpdateMacAddress(pThis); /* (could be a state restore) */ + drvR3IntNetSetActive(pThis, true /* fActive */); + } + + switch (enmReason) + { + case VMRESUMEREASON_HOST_RESUME: + { + uint32_t u32TrunkType; + int rc = pDrvIns->pHlpR3->pfnCFGMQueryU32(pDrvIns->pCfg, "TrunkType", &u32TrunkType); + AssertRC(rc); + + /* + * Only do the disconnect for bridged networking. Host-only and + * internal networks are not affected by a host resume. + */ + if ( RT_SUCCESS(rc) + && u32TrunkType == kIntNetTrunkType_NetFlt) + { + rc = pThis->pIAboveConfigR3->pfnSetLinkState(pThis->pIAboveConfigR3, + PDMNETWORKLINKSTATE_DOWN_RESUME); + AssertRC(rc); + } + break; + } + case VMRESUMEREASON_TELEPORTED: + case VMRESUMEREASON_TELEPORT_FAILED: + { + if ( PDMDrvHlpVMTeleportedAndNotFullyResumedYet(pDrvIns) + && pThis->pIAboveConfigR3) + { + /* + * We've just been teleported and need to drop a hint to the switch + * since we're likely to have changed to a different port. We just + * push out some ethernet frame that doesn't mean anything to anyone. + * For this purpose ethertype 0x801e was chosen since it was registered + * to Sun (dunno what it is/was used for though). + */ + union + { + RTNETETHERHDR Hdr; + uint8_t ab[128]; + } Frame; + RT_ZERO(Frame); + Frame.Hdr.DstMac.au16[0] = 0xffff; + Frame.Hdr.DstMac.au16[1] = 0xffff; + Frame.Hdr.DstMac.au16[2] = 0xffff; + Frame.Hdr.EtherType = RT_H2BE_U16_C(0x801e); + int rc = pThis->pIAboveConfigR3->pfnGetMac(pThis->pIAboveConfigR3, + &Frame.Hdr.SrcMac); + if (RT_SUCCESS(rc)) + rc = drvR3IntNetResumeSend(pThis, &Frame, sizeof(Frame)); + if (RT_FAILURE(rc)) + LogRel(("IntNet#%u: Sending dummy frame failed: %Rrc\n", + pDrvIns->iInstance, rc)); + } + break; + } + default: /* ignore every other resume reason else */ + break; + } /* end of switch(enmReason) */ +} + + +/** + * Suspend notification. + * + * @param pDrvIns The driver instance. + */ +static DECLCALLBACK(void) drvR3IntNetSuspend(PPDMDRVINS pDrvIns) +{ + LogFlow(("drvR3IntNetPowerSuspend\n")); + PDRVINTNET pThis = PDMINS_2_DATA(pDrvIns, PDRVINTNET); + if (!pThis->fActivateEarlyDeactivateLate) + { + ASMAtomicXchgSize(&pThis->enmRecvState, RECVSTATE_SUSPENDED); + drvR3IntNetSetActive(pThis, false /* fActive */); + } +} + + +/** + * Power On notification. + * + * @param pDrvIns The driver instance. + */ +static DECLCALLBACK(void) drvR3IntNetPowerOn(PPDMDRVINS pDrvIns) +{ + LogFlow(("drvR3IntNetPowerOn\n")); + PDRVINTNET pThis = PDMINS_2_DATA(pDrvIns, PDRVINTNET); + if (!pThis->fActivateEarlyDeactivateLate) + { + ASMAtomicXchgSize(&pThis->enmRecvState, RECVSTATE_RUNNING); + RTSemEventSignal(pThis->hRecvEvt); + drvR3IntNetUpdateMacAddress(pThis); + drvR3IntNetSetActive(pThis, true /* fActive */); + } +} + + +/** + * @interface_method_impl{PDMDRVREG,pfnRelocate} + */ +static DECLCALLBACK(void) drvR3IntNetRelocate(PPDMDRVINS pDrvIns, RTGCINTPTR offDelta) +{ + /* nothing to do here yet */ + RT_NOREF(pDrvIns, offDelta); +} + + +/** + * Destruct a driver instance. + * + * Most VM resources are freed by the VM. This callback is provided so that any non-VM + * resources can be freed correctly. + * + * @param pDrvIns The driver instance data. + */ +static DECLCALLBACK(void) drvR3IntNetDestruct(PPDMDRVINS pDrvIns) +{ + LogFlow(("drvR3IntNetDestruct\n")); + PDRVINTNET pThis = PDMINS_2_DATA(pDrvIns, PDRVINTNET); + PDMDRV_CHECK_VERSIONS_RETURN_VOID(pDrvIns); + + /* + * Indicate to the receive thread that it's time to quit. + */ + ASMAtomicXchgSize(&pThis->enmRecvState, RECVSTATE_TERMINATE); + ASMAtomicXchgSize(&pThis->fLinkDown, true); + RTSEMEVENT hRecvEvt = pThis->hRecvEvt; + pThis->hRecvEvt = NIL_RTSEMEVENT; + + if (hRecvEvt != NIL_RTSEMEVENT) + RTSemEventSignal(hRecvEvt); + + if (pThis->hIf != INTNET_HANDLE_INVALID) + { +#if defined(RT_OS_DARWIN) && defined(VBOX_WITH_INTNET_SERVICE_IN_R3) + if (!pThis->fIntNetR3Svc) /* The R3 service case is handled b the hRecEvt event semaphore. */ +#endif + { + INTNETIFABORTWAITREQ AbortWaitReq; + AbortWaitReq.Hdr.u32Magic = SUPVMMR0REQHDR_MAGIC; + AbortWaitReq.Hdr.cbReq = sizeof(AbortWaitReq); + AbortWaitReq.pSession = NIL_RTR0PTR; + AbortWaitReq.hIf = pThis->hIf; + AbortWaitReq.fNoMoreWaits = true; + int rc = drvR3IntNetCallSvc(pThis, VMMR0_DO_INTNET_IF_ABORT_WAIT, &AbortWaitReq, sizeof(AbortWaitReq)); + AssertMsg(RT_SUCCESS(rc) || rc == VERR_SEM_DESTROYED, ("%Rrc\n", rc)); RT_NOREF_PV(rc); + } + } + + /* + * Wait for the threads to terminate. + */ + if (pThis->pXmitThread) + { + int rc = PDMDrvHlpThreadDestroy(pDrvIns, pThis->pXmitThread, NULL); + AssertRC(rc); + pThis->pXmitThread = NULL; + } + + if (pThis->hRecvThread != NIL_RTTHREAD) + { + int rc = RTThreadWait(pThis->hRecvThread, 5000, NULL); + AssertRC(rc); + pThis->hRecvThread = NIL_RTTHREAD; + } + + /* + * Deregister statistics in case we're being detached. + */ + if (pThis->pBufR3) + { + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->Recv.cStatFrames); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->Recv.cbStatWritten); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->Recv.cOverflows); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->Send.cStatFrames); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->Send.cbStatWritten); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->Send.cOverflows); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->cStatYieldsOk); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->cStatYieldsNok); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->cStatLost); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->cStatBadFrames); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->StatSend1); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->StatSend2); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->StatRecv1); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->StatRecv2); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->pBufR3->StatReserved); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->StatReceivedGso); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->StatSentGso); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->StatSentR0); +#ifdef VBOX_WITH_STATISTICS + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->StatReceive); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->StatTransmit); +#endif + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->StatXmitWakeupR0); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->StatXmitWakeupR3); + PDMDrvHlpSTAMDeregister(pDrvIns, &pThis->StatXmitProcessRing); + } + + /* + * Close the interface + */ + if (pThis->hIf != INTNET_HANDLE_INVALID) + { + INTNETIFCLOSEREQ CloseReq; + CloseReq.Hdr.u32Magic = SUPVMMR0REQHDR_MAGIC; + CloseReq.Hdr.cbReq = sizeof(CloseReq); + CloseReq.pSession = NIL_RTR0PTR; + CloseReq.hIf = pThis->hIf; + pThis->hIf = INTNET_HANDLE_INVALID; + int rc = drvR3IntNetCallSvc(pThis, VMMR0_DO_INTNET_IF_CLOSE, &CloseReq, sizeof(CloseReq)); + AssertRC(rc); + } + +#if defined(RT_OS_DARWIN) && defined(VBOX_WITH_INTNET_SERVICE_IN_R3) + if (pThis->fIntNetR3Svc) + { + /* Unmap the shared buffer. */ + munmap(pThis->pBufR3, pThis->cbBuf); + xpc_connection_cancel(pThis->hXpcCon); + pThis->fIntNetR3Svc = false; + pThis->hXpcCon = NULL; + } +#endif + + /* + * Destroy the semaphores, S/G cache and xmit lock. + */ + if (hRecvEvt != NIL_RTSEMEVENT) + RTSemEventDestroy(hRecvEvt); + + if (pThis->hXmitEvt != NIL_SUPSEMEVENT) + { + SUPSemEventClose(pThis->pSupDrvSession, pThis->hXmitEvt); + pThis->hXmitEvt = NIL_SUPSEMEVENT; + } + + RTMemCacheDestroy(pThis->hSgCache); + pThis->hSgCache = NIL_RTMEMCACHE; + + if (PDMDrvHlpCritSectIsInitialized(pDrvIns, &pThis->XmitLock)) + PDMDrvHlpCritSectDelete(pDrvIns, &pThis->XmitLock); +} + + +/** + * Queries a policy config value and translates it into open network flag. + * + * @returns VBox status code (error set on failure). + * @param pDrvIns The driver instance. + * @param pszName The value name. + * @param paFlags The open network flag descriptors. + * @param cFlags The number of descriptors. + * @param fFlags The fixed flag. + * @param pfFlags The flags variable to update. + */ +static int drvIntNetR3CfgGetPolicy(PPDMDRVINS pDrvIns, const char *pszName, PCDRVINTNETFLAG paFlags, size_t cFlags, + uint32_t fFixedFlag, uint32_t *pfFlags) +{ + PCPDMDRVHLPR3 pHlp = pDrvIns->pHlpR3; + + char szValue[64]; + int rc = pHlp->pfnCFGMQueryString(pDrvIns->pCfg, pszName, szValue, sizeof(szValue)); + if (RT_FAILURE(rc)) + { + if (rc == VERR_CFGM_VALUE_NOT_FOUND) + return VINF_SUCCESS; + return PDMDrvHlpVMSetError(pDrvIns, rc, RT_SRC_POS, + N_("Configuration error: Failed to query value of \"%s\""), pszName); + } + + /* + * Check for +fixed first, so it can be stripped off. + */ + char *pszSep = strpbrk(szValue, "+,;"); + if (pszSep) + { + *pszSep++ = '\0'; + const char *pszFixed = RTStrStripL(pszSep); + if (strcmp(pszFixed, "fixed")) + { + *pszSep = '+'; + return PDMDrvHlpVMSetError(pDrvIns, VERR_INVALID_PARAMETER, RT_SRC_POS, + N_("Configuration error: The value of \"%s\" is unknown: \"%s\""), pszName, szValue); + } + *pfFlags |= fFixedFlag; + RTStrStripR(szValue); + } + + /* + * Match against the flag values. + */ + size_t i = cFlags; + while (i-- > 0) + if (!strcmp(paFlags[i].pszChoice, szValue)) + { + *pfFlags |= paFlags[i].fFlag; + return VINF_SUCCESS; + } + + if (!strcmp(szValue, "none")) + return VINF_SUCCESS; + + if (!strcmp(szValue, "fixed")) + { + *pfFlags |= fFixedFlag; + return VINF_SUCCESS; + } + + return PDMDrvHlpVMSetError(pDrvIns, VERR_INVALID_PARAMETER, RT_SRC_POS, + N_("Configuration error: The value of \"%s\" is unknown: \"%s\""), pszName, szValue); +} + + +/** + * Construct a TAP network transport driver instance. + * + * @copydoc FNPDMDRVCONSTRUCT + */ +static DECLCALLBACK(int) drvR3IntNetConstruct(PPDMDRVINS pDrvIns, PCFGMNODE pCfg, uint32_t fFlags) +{ + RT_NOREF(fFlags); + PDMDRV_CHECK_VERSIONS_RETURN(pDrvIns); + PDRVINTNET pThis = PDMINS_2_DATA(pDrvIns, PDRVINTNET); + PCPDMDRVHLPR3 pHlp = pDrvIns->pHlpR3; + bool f; + + /* + * Init the static parts. + */ + pThis->pDrvInsR3 = pDrvIns; +#ifdef VBOX_WITH_DRVINTNET_IN_R0 + pThis->pDrvInsR0 = PDMDRVINS_2_R0PTR(pDrvIns); +#endif + pThis->hIf = INTNET_HANDLE_INVALID; + pThis->hRecvThread = NIL_RTTHREAD; + pThis->hRecvEvt = NIL_RTSEMEVENT; + pThis->pXmitThread = NULL; + pThis->hXmitEvt = NIL_SUPSEMEVENT; + pThis->pSupDrvSession = PDMDrvHlpGetSupDrvSession(pDrvIns); + pThis->hSgCache = NIL_RTMEMCACHE; + pThis->enmRecvState = RECVSTATE_SUSPENDED; + pThis->fActivateEarlyDeactivateLate = false; + /* IBase* */ + pDrvIns->IBase.pfnQueryInterface = drvR3IntNetIBase_QueryInterface; +#ifdef VBOX_WITH_DRVINTNET_IN_R0 + pThis->IBaseR0.pfnQueryInterface = drvR3IntNetIBaseR0_QueryInterface; + pThis->IBaseRC.pfnQueryInterface = drvR3IntNetIBaseRC_QueryInterface; +#endif + /* INetworkUp */ + pThis->INetworkUpR3.pfnBeginXmit = drvIntNetUp_BeginXmit; + pThis->INetworkUpR3.pfnAllocBuf = drvIntNetUp_AllocBuf; + pThis->INetworkUpR3.pfnFreeBuf = drvIntNetUp_FreeBuf; + pThis->INetworkUpR3.pfnSendBuf = drvIntNetUp_SendBuf; + pThis->INetworkUpR3.pfnEndXmit = drvIntNetUp_EndXmit; + pThis->INetworkUpR3.pfnSetPromiscuousMode = drvIntNetUp_SetPromiscuousMode; + pThis->INetworkUpR3.pfnNotifyLinkChanged = drvR3IntNetUp_NotifyLinkChanged; + + /* + * Validate the config. + */ + PDMDRV_VALIDATE_CONFIG_RETURN(pDrvIns, + "Network" + "|Trunk" + "|TrunkType" + "|ReceiveBufferSize" + "|SendBufferSize" + "|SharedMacOnWire" + "|RestrictAccess" + "|RequireExactPolicyMatch" + "|RequireAsRestrictivePolicy" + "|AccessPolicy" + "|PromiscPolicyClients" + "|PromiscPolicyHost" + "|PromiscPolicyWire" + "|IfPolicyPromisc" + "|TrunkPolicyHost" + "|TrunkPolicyWire" + "|IsService" + "|IgnoreConnectFailure" + "|Workaround1", + ""); + + /* + * Check that no-one is attached to us. + */ + AssertMsgReturn(PDMDrvHlpNoAttach(pDrvIns) == VERR_PDM_NO_ATTACHED_DRIVER, + ("Configuration error: Not possible to attach anything to this driver!\n"), + VERR_PDM_DRVINS_NO_ATTACH); + + /* + * Query the network port interface. + */ + pThis->pIAboveNet = PDMIBASE_QUERY_INTERFACE(pDrvIns->pUpBase, PDMINETWORKDOWN); + if (!pThis->pIAboveNet) + { + AssertMsgFailed(("Configuration error: the above device/driver didn't export the network port interface!\n")); + return VERR_PDM_MISSING_INTERFACE_ABOVE; + } + pThis->pIAboveConfigR3 = PDMIBASE_QUERY_INTERFACE(pDrvIns->pUpBase, PDMINETWORKCONFIG); + + /* + * Read the configuration. + */ + INTNETOPENREQ OpenReq; + RT_ZERO(OpenReq); + OpenReq.Hdr.cbReq = sizeof(OpenReq); + OpenReq.Hdr.u32Magic = SUPVMMR0REQHDR_MAGIC; + OpenReq.pSession = NIL_RTR0PTR; + + /** @cfgm{Network, string} + * The name of the internal network to connect to. + */ + int rc = pHlp->pfnCFGMQueryString(pCfg, "Network", OpenReq.szNetwork, sizeof(OpenReq.szNetwork)); + if (RT_FAILURE(rc)) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: Failed to get the \"Network\" value")); + strcpy(pThis->szNetwork, OpenReq.szNetwork); + + /** @cfgm{TrunkType, uint32_t, kIntNetTrunkType_None} + * The trunk connection type see INTNETTRUNKTYPE. + */ + uint32_t u32TrunkType; + rc = pHlp->pfnCFGMQueryU32(pCfg, "TrunkType", &u32TrunkType); + if (rc == VERR_CFGM_VALUE_NOT_FOUND) + u32TrunkType = kIntNetTrunkType_None; + else if (RT_FAILURE(rc)) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: Failed to get the \"TrunkType\" value")); + OpenReq.enmTrunkType = (INTNETTRUNKTYPE)u32TrunkType; + + /** @cfgm{Trunk, string, ""} + * The name of the trunk connection. + */ + rc = pHlp->pfnCFGMQueryString(pCfg, "Trunk", OpenReq.szTrunk, sizeof(OpenReq.szTrunk)); + if (rc == VERR_CFGM_VALUE_NOT_FOUND) + OpenReq.szTrunk[0] = '\0'; + else if (RT_FAILURE(rc)) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: Failed to get the \"Trunk\" value")); + + OpenReq.fFlags = 0; + + /** @cfgm{SharedMacOnWire, boolean, false} + * Whether to shared the MAC address of the host interface when using the wire. When + * attaching to a wireless NIC this option is usually a requirement. + */ + bool fSharedMacOnWire; + rc = pHlp->pfnCFGMQueryBoolDef(pCfg, "SharedMacOnWire", &fSharedMacOnWire, false); + if (RT_FAILURE(rc)) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: Failed to get the \"SharedMacOnWire\" value")); + if (fSharedMacOnWire) + OpenReq.fFlags |= INTNET_OPEN_FLAGS_SHARED_MAC_ON_WIRE; + + /** @cfgm{RestrictAccess, boolean, true} + * Whether to restrict the access to the network or if it should be public. + * Everyone on the computer can connect to a public network. + * @deprecated Use AccessPolicy instead. + */ + rc = pHlp->pfnCFGMQueryBool(pCfg, "RestrictAccess", &f); + if (RT_SUCCESS(rc)) + { + if (f) + OpenReq.fFlags |= INTNET_OPEN_FLAGS_ACCESS_RESTRICTED; + else + OpenReq.fFlags |= INTNET_OPEN_FLAGS_ACCESS_PUBLIC; + OpenReq.fFlags |= INTNET_OPEN_FLAGS_ACCESS_FIXED; + } + else if (rc != VERR_CFGM_VALUE_NOT_FOUND) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: Failed to get the \"RestrictAccess\" value")); + + /** @cfgm{RequireExactPolicyMatch, boolean, false} + * Whether to require that the current security and promiscuous policies of + * the network is exactly as the ones specified in this open network + * request. Use this with RequireAsRestrictivePolicy to prevent + * restrictions from being lifted. If no further policy changes are + * desired, apply the relevant fixed flags. */ + rc = pHlp->pfnCFGMQueryBoolDef(pCfg, "RequireExactPolicyMatch", &f, false); + if (RT_FAILURE(rc)) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: Failed to get the \"RequireExactPolicyMatch\" value")); + if (f) + OpenReq.fFlags |= INTNET_OPEN_FLAGS_REQUIRE_EXACT; + + /** @cfgm{RequireAsRestrictivePolicy, boolean, false} + * Whether to require that the security and promiscuous policies of the + * network is at least as restrictive as specified this request specifies + * and prevent them being lifted later on. + */ + rc = pHlp->pfnCFGMQueryBoolDef(pCfg, "RequireAsRestrictivePolicy", &f, false); + if (RT_FAILURE(rc)) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: Failed to get the \"RequireAsRestrictivePolicy\" value")); + if (f) + OpenReq.fFlags |= INTNET_OPEN_FLAGS_REQUIRE_AS_RESTRICTIVE_POLICIES; + + /** @cfgm{AccessPolicy, string, "none"} + * The access policy of the network: + * public, public+fixed, restricted, restricted+fixed, none or fixed. + * + * A "public" network is accessible to everyone on the same host, while a + * "restricted" one is only accessible to VMs & services started by the + * same user. The "none" policy, which is the default, means no policy + * change or choice is made and that the current (existing network) or + * default (new) policy should be used. */ + static const DRVINTNETFLAG s_aAccessPolicyFlags[] = + { + { "public", INTNET_OPEN_FLAGS_ACCESS_PUBLIC }, + { "restricted", INTNET_OPEN_FLAGS_ACCESS_RESTRICTED } + }; + rc = drvIntNetR3CfgGetPolicy(pDrvIns, "AccessPolicy", &s_aAccessPolicyFlags[0], RT_ELEMENTS(s_aAccessPolicyFlags), + INTNET_OPEN_FLAGS_ACCESS_FIXED, &OpenReq.fFlags); + AssertRCReturn(rc, rc); + + /** @cfgm{PromiscPolicyClients, string, "none"} + * The network wide promiscuous mode policy for client (non-trunk) + * interfaces: allow, allow+fixed, deny, deny+fixed, none or fixed. */ + static const DRVINTNETFLAG s_aPromiscPolicyClient[] = + { + { "allow", INTNET_OPEN_FLAGS_PROMISC_ALLOW_CLIENTS }, + { "deny", INTNET_OPEN_FLAGS_PROMISC_DENY_CLIENTS } + }; + rc = drvIntNetR3CfgGetPolicy(pDrvIns, "PromiscPolicyClients", &s_aPromiscPolicyClient[0], RT_ELEMENTS(s_aPromiscPolicyClient), + INTNET_OPEN_FLAGS_PROMISC_FIXED, &OpenReq.fFlags); + AssertRCReturn(rc, rc); + /** @cfgm{PromiscPolicyHost, string, "none"} + * The promiscuous mode policy for the trunk-host + * connection: allow, allow+fixed, deny, deny+fixed, none or fixed. */ + static const DRVINTNETFLAG s_aPromiscPolicyHost[] = + { + { "allow", INTNET_OPEN_FLAGS_PROMISC_ALLOW_TRUNK_HOST }, + { "deny", INTNET_OPEN_FLAGS_PROMISC_DENY_TRUNK_HOST } + }; + rc = drvIntNetR3CfgGetPolicy(pDrvIns, "PromiscPolicyHost", &s_aPromiscPolicyHost[0], RT_ELEMENTS(s_aPromiscPolicyHost), + INTNET_OPEN_FLAGS_PROMISC_FIXED, &OpenReq.fFlags); + AssertRCReturn(rc, rc); + /** @cfgm{PromiscPolicyWire, string, "none"} + * The promiscuous mode policy for the trunk-host + * connection: allow, allow+fixed, deny, deny+fixed, none or fixed. */ + static const DRVINTNETFLAG s_aPromiscPolicyWire[] = + { + { "allow", INTNET_OPEN_FLAGS_PROMISC_ALLOW_TRUNK_WIRE }, + { "deny", INTNET_OPEN_FLAGS_PROMISC_DENY_TRUNK_WIRE } + }; + rc = drvIntNetR3CfgGetPolicy(pDrvIns, "PromiscPolicyWire", &s_aPromiscPolicyWire[0], RT_ELEMENTS(s_aPromiscPolicyWire), + INTNET_OPEN_FLAGS_PROMISC_FIXED, &OpenReq.fFlags); + AssertRCReturn(rc, rc); + + + /** @cfgm{IfPolicyPromisc, string, "none"} + * The promiscuous mode policy for this + * interface: deny, deny+fixed, allow-all, allow-all+fixed, allow-network, + * allow-network+fixed, none or fixed. */ + static const DRVINTNETFLAG s_aIfPolicyPromisc[] = + { + { "allow-all", INTNET_OPEN_FLAGS_IF_PROMISC_ALLOW | INTNET_OPEN_FLAGS_IF_PROMISC_SEE_TRUNK }, + { "allow-network", INTNET_OPEN_FLAGS_IF_PROMISC_ALLOW | INTNET_OPEN_FLAGS_IF_PROMISC_NO_TRUNK }, + { "deny", INTNET_OPEN_FLAGS_IF_PROMISC_DENY } + }; + rc = drvIntNetR3CfgGetPolicy(pDrvIns, "IfPolicyPromisc", &s_aIfPolicyPromisc[0], RT_ELEMENTS(s_aIfPolicyPromisc), + INTNET_OPEN_FLAGS_IF_FIXED, &OpenReq.fFlags); + AssertRCReturn(rc, rc); + + + /** @cfgm{TrunkPolicyHost, string, "none"} + * The trunk-host policy: promisc, promisc+fixed, enabled, enabled+fixed, + * disabled, disabled+fixed, none or fixed + * + * This can be used to prevent packages to be routed to the host. */ + static const DRVINTNETFLAG s_aTrunkPolicyHost[] = + { + { "promisc", INTNET_OPEN_FLAGS_TRUNK_HOST_ENABLED | INTNET_OPEN_FLAGS_TRUNK_HOST_PROMISC_MODE }, + { "enabled", INTNET_OPEN_FLAGS_TRUNK_HOST_ENABLED }, + { "disabled", INTNET_OPEN_FLAGS_TRUNK_HOST_DISABLED } + }; + rc = drvIntNetR3CfgGetPolicy(pDrvIns, "TrunkPolicyHost", &s_aTrunkPolicyHost[0], RT_ELEMENTS(s_aTrunkPolicyHost), + INTNET_OPEN_FLAGS_TRUNK_FIXED, &OpenReq.fFlags); + AssertRCReturn(rc, rc); + /** @cfgm{TrunkPolicyWire, string, "none"} + * The trunk-host policy: promisc, promisc+fixed, enabled, enabled+fixed, + * disabled, disabled+fixed, none or fixed. + * + * This can be used to prevent packages to be routed to the wire. */ + static const DRVINTNETFLAG s_aTrunkPolicyWire[] = + { + { "promisc", INTNET_OPEN_FLAGS_TRUNK_WIRE_ENABLED | INTNET_OPEN_FLAGS_TRUNK_WIRE_PROMISC_MODE }, + { "enabled", INTNET_OPEN_FLAGS_TRUNK_WIRE_ENABLED }, + { "disabled", INTNET_OPEN_FLAGS_TRUNK_WIRE_DISABLED } + }; + rc = drvIntNetR3CfgGetPolicy(pDrvIns, "TrunkPolicyWire", &s_aTrunkPolicyWire[0], RT_ELEMENTS(s_aTrunkPolicyWire), + INTNET_OPEN_FLAGS_TRUNK_FIXED, &OpenReq.fFlags); + AssertRCReturn(rc, rc); + + + /** @cfgm{ReceiveBufferSize, uint32_t, 318 KB} + * The size of the receive buffer. + */ + rc = pHlp->pfnCFGMQueryU32(pCfg, "ReceiveBufferSize", &OpenReq.cbRecv); + if (rc == VERR_CFGM_VALUE_NOT_FOUND) + OpenReq.cbRecv = 318 * _1K ; + else if (RT_FAILURE(rc)) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: Failed to get the \"ReceiveBufferSize\" value")); + + /** @cfgm{SendBufferSize, uint32_t, 196 KB} + * The size of the send (transmit) buffer. + * This should be more than twice the size of the larges frame size because + * the ring buffer is very simple and doesn't support splitting up frames + * nor inserting padding. So, if this is too close to the frame size the + * header will fragment the buffer such that the frame won't fit on either + * side of it and the code will get very upset about it all. + */ + rc = pHlp->pfnCFGMQueryU32(pCfg, "SendBufferSize", &OpenReq.cbSend); + if (rc == VERR_CFGM_VALUE_NOT_FOUND) + OpenReq.cbSend = RT_ALIGN_Z(VBOX_MAX_GSO_SIZE * 3, _1K); + else if (RT_FAILURE(rc)) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: Failed to get the \"SendBufferSize\" value")); + if (OpenReq.cbSend < 128) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: The \"SendBufferSize\" value is too small")); + if (OpenReq.cbSend < VBOX_MAX_GSO_SIZE * 3) + LogRel(("DrvIntNet: Warning! SendBufferSize=%u, Recommended minimum size %u butes.\n", OpenReq.cbSend, VBOX_MAX_GSO_SIZE * 4)); + + /** @cfgm{IsService, boolean, true} + * This alterns the way the thread is suspended and resumed. When it's being used by + * a service such as LWIP/iSCSI it shouldn't suspend immediately like for a NIC. + */ + rc = pHlp->pfnCFGMQueryBool(pCfg, "IsService", &pThis->fActivateEarlyDeactivateLate); + if (rc == VERR_CFGM_VALUE_NOT_FOUND) + pThis->fActivateEarlyDeactivateLate = false; + else if (RT_FAILURE(rc)) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: Failed to get the \"IsService\" value")); + + + /** @cfgm{IgnoreConnectFailure, boolean, false} + * When set only raise a runtime error if we cannot connect to the internal + * network. */ + bool fIgnoreConnectFailure; + rc = pHlp->pfnCFGMQueryBoolDef(pCfg, "IgnoreConnectFailure", &fIgnoreConnectFailure, false); + if (RT_FAILURE(rc)) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: Failed to get the \"IgnoreConnectFailure\" value")); + + /** @cfgm{Workaround1, boolean, depends} + * Enables host specific workarounds, the default is depends on the whether + * we think the host requires it or not. + */ + bool fWorkaround1 = false; +#ifdef RT_OS_DARWIN + if (OpenReq.fFlags & INTNET_OPEN_FLAGS_SHARED_MAC_ON_WIRE) + { + char szKrnlVer[256]; + RTSystemQueryOSInfo(RTSYSOSINFO_RELEASE, szKrnlVer, sizeof(szKrnlVer)); + if (strcmp(szKrnlVer, "10.7.0") >= 0) + { + LogRel(("IntNet#%u: Enables the workaround (ip_tos=0) for the little endian ip header checksum problem\n")); + fWorkaround1 = true; + } + } +#endif + rc = pHlp->pfnCFGMQueryBoolDef(pCfg, "Workaround1", &fWorkaround1, fWorkaround1); + if (RT_FAILURE(rc)) + return PDMDRV_SET_ERROR(pDrvIns, rc, + N_("Configuration error: Failed to get the \"Workaround1\" value")); + if (fWorkaround1) + OpenReq.fFlags |= INTNET_OPEN_FLAGS_WORKAROUND_1; + + LogRel(("IntNet#%u: szNetwork={%s} enmTrunkType=%d szTrunk={%s} fFlags=%#x cbRecv=%u cbSend=%u fIgnoreConnectFailure=%RTbool\n", + pDrvIns->iInstance, OpenReq.szNetwork, OpenReq.enmTrunkType, OpenReq.szTrunk, OpenReq.fFlags, + OpenReq.cbRecv, OpenReq.cbSend, fIgnoreConnectFailure)); + +#ifdef RT_OS_DARWIN + /* Temporary hack: attach to a network with the name 'if=en0' and you're hitting the wire. */ + if ( !OpenReq.szTrunk[0] + && OpenReq.enmTrunkType == kIntNetTrunkType_None + && !strncmp(pThis->szNetwork, RT_STR_TUPLE("if=en")) + && RT_C_IS_DIGIT(pThis->szNetwork[sizeof("if=en") - 1]) + && !pThis->szNetwork[sizeof("if=en")]) + { + OpenReq.enmTrunkType = kIntNetTrunkType_NetFlt; + strcpy(OpenReq.szTrunk, &pThis->szNetwork[sizeof("if=") - 1]); + } + /* Temporary hack: attach to a network with the name 'wif=en0' and you're on the air. */ + if ( !OpenReq.szTrunk[0] + && OpenReq.enmTrunkType == kIntNetTrunkType_None + && !strncmp(pThis->szNetwork, RT_STR_TUPLE("wif=en")) + && RT_C_IS_DIGIT(pThis->szNetwork[sizeof("wif=en") - 1]) + && !pThis->szNetwork[sizeof("wif=en")]) + { + OpenReq.enmTrunkType = kIntNetTrunkType_NetFlt; + OpenReq.fFlags |= INTNET_OPEN_FLAGS_SHARED_MAC_ON_WIRE; + strcpy(OpenReq.szTrunk, &pThis->szNetwork[sizeof("wif=") - 1]); + } +#endif /* DARWIN */ + + /* + * Create the event semaphore, S/G cache and xmit critsect. + */ + rc = RTSemEventCreate(&pThis->hRecvEvt); + if (RT_FAILURE(rc)) + return rc; + rc = RTMemCacheCreate(&pThis->hSgCache, sizeof(PDMSCATTERGATHER), 0, UINT32_MAX, NULL, NULL, pThis, 0); + if (RT_FAILURE(rc)) + return rc; + rc = PDMDrvHlpCritSectInit(pDrvIns, &pThis->XmitLock, RT_SRC_POS, "IntNetXmit"); + if (RT_FAILURE(rc)) + return rc; + + /* + * Create the interface. + */ + if (SUPR3IsDriverless()) + { +#if defined(RT_OS_DARWIN) && defined(VBOX_WITH_INTNET_SERVICE_IN_R3) + xpc_connection_t hXpcCon = xpc_connection_create(INTNET_R3_SVC_NAME, NULL); + xpc_connection_set_event_handler(hXpcCon, ^(xpc_object_t hObj) { + if (xpc_get_type(hObj) == XPC_TYPE_ERROR) + { + /** @todo Error handling - reconnecting. */ + } + else + { + /* Out of band messages should only come when there is something to receive. */ + RTSemEventSignal(pThis->hRecvEvt); + } + }); + + xpc_connection_resume(hXpcCon); + pThis->hXpcCon = hXpcCon; + pThis->fIntNetR3Svc = true; +#else + /** @todo This is probably not good enough for doing fuzz testing, but later... */ + return PDMDrvHlpVMSetError(pDrvIns, VERR_SUP_DRIVERLESS, RT_SRC_POS, + N_("Cannot attach to '%s' in driverless mode"), pThis->szNetwork); +#endif + } + OpenReq.hIf = INTNET_HANDLE_INVALID; + rc = drvR3IntNetCallSvc(pThis, VMMR0_DO_INTNET_OPEN, &OpenReq, sizeof(OpenReq)); + if (RT_FAILURE(rc)) + { + if (fIgnoreConnectFailure) + { + /* + * During VM restore it is fatal if the network is not available because the + * VM settings are locked and the user has no chance to fix network settings. + * Therefore don't abort but just raise a runtime warning. + */ + PDMDrvHlpVMSetRuntimeError(pDrvIns, 0 /*fFlags*/, "HostIfNotConnecting", + N_ ("Cannot connect to the network interface '%s'. The virtual " + "network card will appear to work but the guest will not " + "be able to connect. Please choose a different network in the " + "network settings"), OpenReq.szTrunk); + + return VERR_PDM_NO_ATTACHED_DRIVER; + } + return PDMDrvHlpVMSetError(pDrvIns, rc, RT_SRC_POS, + N_("Failed to open/create the internal network '%s'"), pThis->szNetwork); + } + + AssertRelease(OpenReq.hIf != INTNET_HANDLE_INVALID); + pThis->hIf = OpenReq.hIf; + Log(("IntNet%d: hIf=%RX32 '%s'\n", pDrvIns->iInstance, pThis->hIf, pThis->szNetwork)); + + /* + * Get default buffer. + */ + rc = drvR3IntNetMapBufferPointers(pThis); + if (RT_FAILURE(rc)) + return PDMDrvHlpVMSetError(pDrvIns, rc, RT_SRC_POS, + N_("Failed to get ring-3 buffer for the newly created interface to '%s'"), pThis->szNetwork); + + /* + * Register statistics. + */ + PDMDrvHlpSTAMRegCounterEx(pDrvIns, &pThis->pBufR3->Recv.cbStatWritten, "Bytes/Received", STAMUNIT_BYTES, "Number of received bytes."); + PDMDrvHlpSTAMRegCounterEx(pDrvIns, &pThis->pBufR3->Send.cbStatWritten, "Bytes/Sent", STAMUNIT_BYTES, "Number of sent bytes."); + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->pBufR3->Recv.cOverflows, "Overflows/Recv", "Number overflows."); + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->pBufR3->Send.cOverflows, "Overflows/Sent", "Number overflows."); + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->pBufR3->Recv.cStatFrames, "Packets/Received", "Number of received packets."); + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->pBufR3->Send.cStatFrames, "Packets/Sent", "Number of sent packets."); + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->StatReceivedGso, "Packets/Received-Gso", "The GSO portion of the received packets."); + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->StatSentGso, "Packets/Sent-Gso", "The GSO portion of the sent packets."); + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->StatSentR0, "Packets/Sent-R0", "The ring-0 portion of the sent packets."); + + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->pBufR3->cStatLost, "Packets/Lost", "Number of lost packets."); + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->pBufR3->cStatYieldsNok, "YieldOk", "Number of times yielding helped fix an overflow."); + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->pBufR3->cStatYieldsOk, "YieldNok", "Number of times yielding didn't help fix an overflow."); + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->pBufR3->cStatBadFrames, "BadFrames", "Number of bad frames seed by the consumers."); + PDMDrvHlpSTAMRegProfile(pDrvIns, &pThis->pBufR3->StatSend1, "Send1", "Profiling IntNetR0IfSend."); + PDMDrvHlpSTAMRegProfile(pDrvIns, &pThis->pBufR3->StatSend2, "Send2", "Profiling sending to the trunk."); + PDMDrvHlpSTAMRegProfile(pDrvIns, &pThis->pBufR3->StatRecv1, "Recv1", "Reserved for future receive profiling."); + PDMDrvHlpSTAMRegProfile(pDrvIns, &pThis->pBufR3->StatRecv2, "Recv2", "Reserved for future receive profiling."); + PDMDrvHlpSTAMRegProfile(pDrvIns, &pThis->pBufR3->StatReserved, "Reserved", "Reserved for future use."); +#ifdef VBOX_WITH_STATISTICS + PDMDrvHlpSTAMRegProfileAdv(pDrvIns, &pThis->StatReceive, "Receive", "Profiling packet receive runs."); + PDMDrvHlpSTAMRegProfile(pDrvIns, &pThis->StatTransmit, "Transmit", "Profiling packet transmit runs."); +#endif + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->StatXmitWakeupR0, "XmitWakeup-R0", "Xmit thread wakeups from ring-0."); + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->StatXmitWakeupR3, "XmitWakeup-R3", "Xmit thread wakeups from ring-3."); + PDMDrvHlpSTAMRegCounter(pDrvIns, &pThis->StatXmitProcessRing, "XmitProcessRing", "Time xmit thread was told to process the ring."); + + /* + * Create the async I/O threads. + * Note! Using a PDM thread here doesn't fit with the IsService=true operation. + */ + rc = RTThreadCreate(&pThis->hRecvThread, drvR3IntNetRecvThread, pThis, 0, + RTTHREADTYPE_IO, RTTHREADFLAGS_WAITABLE, "INTNET-RECV"); + if (RT_FAILURE(rc)) + { + AssertRC(rc); + return rc; + } + + rc = SUPSemEventCreate(pThis->pSupDrvSession, &pThis->hXmitEvt); + AssertRCReturn(rc, rc); + + rc = PDMDrvHlpThreadCreate(pDrvIns, &pThis->pXmitThread, pThis, + drvR3IntNetXmitThread, drvR3IntNetXmitWakeUp, 0, RTTHREADTYPE_IO, "INTNET-XMIT"); + AssertRCReturn(rc, rc); + +#ifdef VBOX_WITH_DRVINTNET_IN_R0 + /* + * Resolve the ring-0 context interface addresses. + */ + rc = pDrvIns->pHlpR3->pfnLdrGetR0InterfaceSymbols(pDrvIns, &pThis->INetworkUpR0, sizeof(pThis->INetworkUpR0), + "drvIntNetUp_", PDMINETWORKUP_SYM_LIST); + AssertLogRelRCReturn(rc, rc); +#endif + + /* + * Activate data transmission as early as possible + */ + if (pThis->fActivateEarlyDeactivateLate) + { + ASMAtomicXchgSize(&pThis->enmRecvState, RECVSTATE_RUNNING); + RTSemEventSignal(pThis->hRecvEvt); + + drvR3IntNetUpdateMacAddress(pThis); + drvR3IntNetSetActive(pThis, true /* fActive */); + } + + return rc; +} + + + +/** + * Internal networking transport driver registration record. + */ +const PDMDRVREG g_DrvIntNet = +{ + /* u32Version */ + PDM_DRVREG_VERSION, + /* szName */ + "IntNet", + /* szRCMod */ + "VBoxDDRC.rc", + /* szR0Mod */ + "VBoxDDR0.r0", + /* pszDescription */ + "Internal Networking Transport Driver", + /* fFlags */ +#ifdef VBOX_WITH_DRVINTNET_IN_R0 + PDM_DRVREG_FLAGS_HOST_BITS_DEFAULT | PDM_DRVREG_FLAGS_R0, +#else + PDM_DRVREG_FLAGS_HOST_BITS_DEFAULT, +#endif + /* fClass. */ + PDM_DRVREG_CLASS_NETWORK, + /* cMaxInstances */ + ~0U, + /* cbInstance */ + sizeof(DRVINTNET), + /* pfnConstruct */ + drvR3IntNetConstruct, + /* pfnDestruct */ + drvR3IntNetDestruct, + /* pfnRelocate */ + drvR3IntNetRelocate, + /* pfnIOCtl */ + NULL, + /* pfnPowerOn */ + drvR3IntNetPowerOn, + /* pfnReset */ + NULL, + /* pfnSuspend */ + drvR3IntNetSuspend, + /* pfnResume */ + drvR3IntNetResume, + /* pfnAttach */ + NULL, + /* pfnDetach */ + NULL, + /* pfnPowerOff */ + drvR3IntNetPowerOff, + /* pfnSoftReset */ + NULL, + /* u32EndVersion */ + PDM_DRVREG_VERSION +}; + +#endif /* IN_RING3 */ + |