/* $Id: VMInternal.h 62478 2016-07-22 18:29:06Z vboxsync $ */ /** @file * VM - Internal header file. */ /* * Copyright (C) 2006-2016 Oracle Corporation * * This file is part of VirtualBox Open Source Edition (OSE), as * available from http://www.virtualbox.org. This file is free software; * you can redistribute it and/or modify it under the terms of the GNU * General Public License (GPL) as published by the Free Software * Foundation, in version 2 as it comes in the "COPYING" file of the * VirtualBox OSE distribution. VirtualBox OSE is distributed in the * hope that it will be useful, but WITHOUT ANY WARRANTY of any kind. */ #ifndef ___VMInternal_h #define ___VMInternal_h #include #include #include #include #include /** @defgroup grp_vm_int Internals * @ingroup grp_vm * @internal * @{ */ /** * VM state change callback. */ typedef struct VMATSTATE { /** Pointer to the next one. */ struct VMATSTATE *pNext; /** Pointer to the callback. */ PFNVMATSTATE pfnAtState; /** The user argument. */ void *pvUser; } VMATSTATE; /** Pointer to a VM state change callback. */ typedef VMATSTATE *PVMATSTATE; /** * VM error callback. */ typedef struct VMATERROR { /** Pointer to the next one. */ struct VMATERROR *pNext; /** Pointer to the callback. */ PFNVMATERROR pfnAtError; /** The user argument. */ void *pvUser; } VMATERROR; /** Pointer to a VM error callback. */ typedef VMATERROR *PVMATERROR; /** * Chunk of memory allocated off the hypervisor heap in which * we copy the error details. */ typedef struct VMERROR { /** The size of the chunk. */ uint32_t cbAllocated; /** The current offset into the chunk. * We start by putting the filename and function immediately * after the end of the buffer. */ uint32_t off; /** Offset from the start of this structure to the file name. */ uint32_t offFile; /** The line number. */ uint32_t iLine; /** Offset from the start of this structure to the function name. */ uint32_t offFunction; /** Offset from the start of this structure to the formatted message text. */ uint32_t offMessage; /** The VBox status code. */ int32_t rc; } VMERROR, *PVMERROR; /** * VM runtime error callback. */ typedef struct VMATRUNTIMEERROR { /** Pointer to the next one. */ struct VMATRUNTIMEERROR *pNext; /** Pointer to the callback. */ PFNVMATRUNTIMEERROR pfnAtRuntimeError; /** The user argument. */ void *pvUser; } VMATRUNTIMEERROR; /** Pointer to a VM error callback. */ typedef VMATRUNTIMEERROR *PVMATRUNTIMEERROR; /** * Chunk of memory allocated off the hypervisor heap in which * we copy the runtime error details. */ typedef struct VMRUNTIMEERROR { /** The size of the chunk. */ uint32_t cbAllocated; /** The current offset into the chunk. * We start by putting the error ID immediately * after the end of the buffer. */ uint32_t off; /** Offset from the start of this structure to the error ID. */ uint32_t offErrorId; /** Offset from the start of this structure to the formatted message text. */ uint32_t offMessage; /** Error flags. */ uint32_t fFlags; } VMRUNTIMEERROR, *PVMRUNTIMEERROR; /** The halt method. */ typedef enum { /** The usual invalid value. */ VMHALTMETHOD_INVALID = 0, /** Use the method used during bootstrapping. */ VMHALTMETHOD_BOOTSTRAP, /** Use the default method. */ VMHALTMETHOD_DEFAULT, /** The old spin/yield/block method. */ VMHALTMETHOD_OLD, /** The first go at a block/spin method. */ VMHALTMETHOD_1, /** The first go at a more global approach. */ VMHALTMETHOD_GLOBAL_1, /** The end of valid methods. (not inclusive of course) */ VMHALTMETHOD_END, /** The usual 32-bit max value. */ VMHALTMETHOD_32BIT_HACK = 0x7fffffff } VMHALTMETHOD; /** * VM Internal Data (part of the VM structure). * * @todo Move this and all related things to VMM. The VM component was, to some * extent at least, a bad ad hoc design which should all have been put in * VMM. @see pg_vm. */ typedef struct VMINT { /** VM Error Message. */ R3PTRTYPE(PVMERROR) pErrorR3; /** VM Runtime Error Message. */ R3PTRTYPE(PVMRUNTIMEERROR) pRuntimeErrorR3; /** The VM was/is-being teleported and has not yet been fully resumed. */ bool fTeleportedAndNotFullyResumedYet; /** The VM should power off instead of reset. */ bool fPowerOffInsteadOfReset; } VMINT; /** Pointer to the VM Internal Data (part of the VM structure). */ typedef VMINT *PVMINT; #ifdef IN_RING3 /** * VM internal data kept in the UVM. */ typedef struct VMINTUSERPERVM { /** Head of the standard request queue. Atomic. */ volatile PVMREQ pNormalReqs; /** Head of the priority request queue. Atomic. */ volatile PVMREQ pPriorityReqs; /** The last index used during alloc/free. */ volatile uint32_t iReqFree; /** Number of free request packets. */ volatile uint32_t cReqFree; /** Array of pointers to lists of free request packets. Atomic. */ volatile PVMREQ apReqFree[16 - (HC_ARCH_BITS == 32 ? 5 : 4)]; /** The reference count of the UVM handle. */ volatile uint32_t cUvmRefs; # ifdef VBOX_WITH_STATISTICS /** Number of VMR3ReqAlloc returning a new packet. */ STAMCOUNTER StatReqAllocNew; /** Number of VMR3ReqAlloc causing races. */ STAMCOUNTER StatReqAllocRaces; /** Number of VMR3ReqAlloc returning a recycled packet. */ STAMCOUNTER StatReqAllocRecycled; /** Number of VMR3ReqFree calls. */ STAMCOUNTER StatReqFree; /** Number of times the request was actually freed. */ STAMCOUNTER StatReqFreeOverflow; /** Number of requests served. */ STAMCOUNTER StatReqProcessed; /** Number of times there are more than one request and the others needed to be * pushed back onto the list. */ STAMCOUNTER StatReqMoreThan1; /** Number of times we've raced someone when pushing the other requests back * onto the list. */ STAMCOUNTER StatReqPushBackRaces; # endif /** Pointer to the support library session. * Mainly for creation and destruction. */ PSUPDRVSESSION pSession; /** Force EMT to terminate. */ bool volatile fTerminateEMT; /** Critical section for pAtState and enmPrevVMState. */ RTCRITSECT AtStateCritSect; /** List of registered state change callbacks. */ PVMATSTATE pAtState; /** List of registered state change callbacks. */ PVMATSTATE *ppAtStateNext; /** The previous VM state. * This is mainly used for the 'Resetting' state, but may come in handy later * and when debugging. */ VMSTATE enmPrevVMState; /** Reason for the most recent suspend operation. */ VMSUSPENDREASON enmSuspendReason; /** Reason for the most recent operation. */ VMRESUMEREASON enmResumeReason; /** Critical section for pAtError and pAtRuntimeError. */ RTCRITSECT AtErrorCritSect; /** List of registered error callbacks. */ PVMATERROR pAtError; /** List of registered error callbacks. */ PVMATERROR *ppAtErrorNext; /** The error message count. * This is incremented every time an error is raised. */ uint32_t volatile cErrors; /** The runtime error message count. * This is incremented every time a runtime error is raised. */ uint32_t volatile cRuntimeErrors; /** List of registered error callbacks. */ PVMATRUNTIMEERROR pAtRuntimeError; /** List of registered error callbacks. */ PVMATRUNTIMEERROR *ppAtRuntimeErrorNext; /** @name Generic Halt data * @{ */ /** The current halt method. * Can be selected by CFGM option 'VM/HaltMethod'. */ VMHALTMETHOD enmHaltMethod; /** The index into g_aHaltMethods of the current halt method. */ uint32_t volatile iHaltMethod; /** @} */ /** @todo Do NOT add new members here or reuse the current, we need to store the config for * each halt method separately because we're racing on SMP guest rigs. */ union { /** * Method 1 & 2 - Block whenever possible, and when lagging behind * switch to spinning with regular blocking every 5-200ms (defaults) * depending on the accumulated lag. The blocking interval is adjusted * with the average oversleeping of the last 64 times. * * The difference between 1 and 2 is that we use native absolute * time APIs for the blocking instead of the millisecond based IPRT * interface. */ struct { /** The max interval without blocking (when spinning). */ uint32_t u32MinBlockIntervalCfg; /** The minimum interval between blocking (when spinning). */ uint32_t u32MaxBlockIntervalCfg; /** The value to divide the current lag by to get the raw blocking interval (when spinning). */ uint32_t u32LagBlockIntervalDivisorCfg; /** When to start spinning (lag / nano secs). */ uint32_t u32StartSpinningCfg; /** When to stop spinning (lag / nano secs). */ uint32_t u32StopSpinningCfg; } Method12; /** * The GVMM manages halted and waiting EMTs. */ struct { /** The threshold between spinning and blocking. */ uint32_t cNsSpinBlockThresholdCfg; } Global1; } Halt; /** Pointer to the DBGC instance data. */ void *pvDBGC; /** TLS index for the VMINTUSERPERVMCPU pointer. */ RTTLS idxTLS; /** The VM name. (Set after the config constructure has been called.) */ char *pszName; /** The VM UUID. (Set after the config constructure has been called.) */ RTUUID Uuid; } VMINTUSERPERVM; # ifdef VBOX_WITH_STATISTICS AssertCompileMemberAlignment(VMINTUSERPERVM, StatReqAllocNew, 8); # endif /** Pointer to the VM internal data kept in the UVM. */ typedef VMINTUSERPERVM *PVMINTUSERPERVM; /** * VMCPU internal data kept in the UVM. * * Almost a copy of VMINTUSERPERVM. Separate data properly later on. */ typedef struct VMINTUSERPERVMCPU { /** Head of the normal request queue. Atomic. */ volatile PVMREQ pNormalReqs; /** Head of the priority request queue. Atomic. */ volatile PVMREQ pPriorityReqs; /** The handle to the EMT thread. */ RTTHREAD ThreadEMT; /** The native of the EMT thread. */ RTNATIVETHREAD NativeThreadEMT; /** Wait event semaphore. */ RTSEMEVENT EventSemWait; /** Wait/Idle indicator. */ bool volatile fWait; /** Align the next bit. */ bool afAlignment[HC_ARCH_BITS == 32 ? 3 : 7]; /** @name Generic Halt data * @{ */ /** The average time (ns) between two halts in the last second. (updated once per second) */ uint32_t HaltInterval; /** The average halt frequency for the last second. (updated once per second) */ uint32_t HaltFrequency; /** The number of halts in the current period. */ uint32_t cHalts; uint32_t padding; /**< alignment padding. */ /** When we started counting halts in cHalts (RTTimeNanoTS). */ uint64_t u64HaltsStartTS; /** @} */ /** Union containing data and config for the different halt algorithms. */ union { /** * Method 1 & 2 - Block whenever possible, and when lagging behind * switch to spinning with regular blocking every 5-200ms (defaults) * depending on the accumulated lag. The blocking interval is adjusted * with the average oversleeping of the last 64 times. * * The difference between 1 and 2 is that we use native absolute * time APIs for the blocking instead of the millisecond based IPRT * interface. */ struct { /** How many times we've blocked while cBlockedNS and cBlockedTooLongNS has been accumulating. */ uint32_t cBlocks; /** Align the next member. */ uint32_t u32Alignment; /** Avg. time spend oversleeping when blocking. (Re-calculated every so often.) */ uint64_t cNSBlockedTooLongAvg; /** Total time spend oversleeping when blocking. */ uint64_t cNSBlockedTooLong; /** Total time spent blocking. */ uint64_t cNSBlocked; /** The timestamp (RTTimeNanoTS) of the last block. */ uint64_t u64LastBlockTS; /** When we started spinning relentlessly in order to catch up some of the oversleeping. * This is 0 when we're not spinning. */ uint64_t u64StartSpinTS; } Method12; # if 0 /** * Method 3 & 4 - Same as method 1 & 2 respectivly, except that we * sprinkle it with yields. */ struct { /** How many times we've blocked while cBlockedNS and cBlockedTooLongNS has been accumulating. */ uint32_t cBlocks; /** Avg. time spend oversleeping when blocking. (Re-calculated every so often.) */ uint64_t cBlockedTooLongNSAvg; /** Total time spend oversleeping when blocking. */ uint64_t cBlockedTooLongNS; /** Total time spent blocking. */ uint64_t cBlockedNS; /** The timestamp (RTTimeNanoTS) of the last block. */ uint64_t u64LastBlockTS; /** How many times we've yielded while cBlockedNS and cBlockedTooLongNS has been accumulating. */ uint32_t cYields; /** Avg. time spend oversleeping when yielding. */ uint32_t cYieldTooLongNSAvg; /** Total time spend oversleeping when yielding. */ uint64_t cYieldTooLongNS; /** Total time spent yielding. */ uint64_t cYieldedNS; /** The timestamp (RTTimeNanoTS) of the last block. */ uint64_t u64LastYieldTS; /** When we started spinning relentlessly in order to catch up some of the oversleeping. */ uint64_t u64StartSpinTS; } Method34; # endif } Halt; /** Profiling the halted state; yielding vs blocking. * @{ */ STAMPROFILE StatHaltYield; STAMPROFILE StatHaltBlock; STAMPROFILE StatHaltBlockOverslept; STAMPROFILE StatHaltBlockInsomnia; STAMPROFILE StatHaltBlockOnTime; STAMPROFILE StatHaltTimers; STAMPROFILE StatHaltPoll; /** @} */ } VMINTUSERPERVMCPU; AssertCompileMemberAlignment(VMINTUSERPERVMCPU, u64HaltsStartTS, 8); AssertCompileMemberAlignment(VMINTUSERPERVMCPU, Halt.Method12.cNSBlockedTooLongAvg, 8); AssertCompileMemberAlignment(VMINTUSERPERVMCPU, StatHaltYield, 8); /** Pointer to the VM internal data kept in the UVM. */ typedef VMINTUSERPERVMCPU *PVMINTUSERPERVMCPU; #endif /* IN_RING3 */ RT_C_DECLS_BEGIN DECLCALLBACK(int) vmR3EmulationThread(RTTHREAD ThreadSelf, void *pvArg); int vmR3SetHaltMethodU(PUVM pUVM, VMHALTMETHOD enmHaltMethod); DECLCALLBACK(int) vmR3Destroy(PVM pVM); DECLCALLBACK(void) vmR3SetErrorUV(PUVM pUVM, int rc, RT_SRC_POS_DECL, const char *pszFormat, va_list *args); void vmSetErrorCopy(PVM pVM, int rc, RT_SRC_POS_DECL, const char *pszFormat, va_list args); DECLCALLBACK(int) vmR3SetRuntimeError(PVM pVM, uint32_t fFlags, const char *pszErrorId, char *pszMessage); DECLCALLBACK(int) vmR3SetRuntimeErrorV(PVM pVM, uint32_t fFlags, const char *pszErrorId, const char *pszFormat, va_list *pVa); void vmSetRuntimeErrorCopy(PVM pVM, uint32_t fFlags, const char *pszErrorId, const char *pszFormat, va_list va); void vmR3SetTerminated(PVM pVM); RT_C_DECLS_END /** @} */ #endif