Move UMEntryThunk to StubPrecode infrastructure (#112460)

* Move UMEntryThunk to StubPrecode infrastructure
- Use the StubPrecode logic to create UMEntryThunk instances
- Allocate them all on the Global LoaderAllocator
- Unify the creation of UMEntryThunks between the C++/CLI fixup generation path and the delegate creation path
- Remove ThunkHeapStubManager as it is now subsumed by RangeSectionStubManager
- Tweak code paths around all of this so that we don't have unnecessary indirections (This change will add 2 memory loads to the UMEntryThunk path on x86/x64, but that should be fairly immaterial compared to the cost of an actual reverse p/invoke)

Author: davidwrighton

src/coreclr/debug/daccess/daccess.cpp

@@ -5825,7 +5825,7 @@ ClrDataAccess::RawGetMethodName(
                 {
                     // Try to find matching precode entrypoint
                     Precode* pPrecode = Precode::GetPrecodeFromEntryPoint(alignedAddress, TRUE);
-                    if (pPrecode != NULL)
+                    if (pPrecode != NULL && pPrecode->GetType() != PRECODE_UMENTRY_THUNK)
                     {
                         methodDesc = pPrecode->GetMethodDesc();
                         if (methodDesc != NULL)

src/coreclr/debug/daccess/request.cpp

@@ -1783,7 +1783,6 @@ ClrDataAccess::GetModuleData(CLRDATA_ADDRESS addr, struct DacpModuleData *Module
     ModuleData->dwModuleIndex = 0; // CoreCLR no longer has this concept
     ModuleData->dwTransientFlags = pModule->m_dwTransientFlags;
     ModuleData->LoaderAllocator = HOST_CDADDR(pModule->m_loaderAllocator);
-    ModuleData->ThunkHeap = HOST_CDADDR(pModule->m_pThunkHeap);
 
     EX_TRY
     {

src/coreclr/debug/runtimeinfo/datadescriptor.h

@@ -223,7 +223,6 @@ CDAC_TYPE_FIELD(Module, /*pointer*/, Assembly, cdac_data<Module>::Assembly)
 CDAC_TYPE_FIELD(Module, /*pointer*/, Base, cdac_data<Module>::Base)
 CDAC_TYPE_FIELD(Module, /*uint32*/, Flags, cdac_data<Module>::Flags)
 CDAC_TYPE_FIELD(Module, /*pointer*/, LoaderAllocator, cdac_data<Module>::LoaderAllocator)
-CDAC_TYPE_FIELD(Module, /*pointer*/, ThunkHeap, cdac_data<Module>::ThunkHeap)
 CDAC_TYPE_FIELD(Module, /*pointer*/, DynamicMetadata, cdac_data<Module>::DynamicMetadata)
 CDAC_TYPE_FIELD(Module, /*pointer*/, Path, cdac_data<Module>::Path)
 CDAC_TYPE_FIELD(Module, /*pointer*/, FileName, cdac_data<Module>::FileName)
@@ -493,7 +492,7 @@ CDAC_TYPE_END(PlatformMetadata)
 
 CDAC_TYPE_BEGIN(StubPrecodeData)
 CDAC_TYPE_INDETERMINATE(StubPrecodeData)
-CDAC_TYPE_FIELD(StubPrecodeData, /*pointer*/, MethodDesc, offsetof(StubPrecodeData, MethodDesc))
+CDAC_TYPE_FIELD(StubPrecodeData, /*pointer*/, MethodDesc, offsetof(StubPrecodeData, SecretParam))
 CDAC_TYPE_FIELD(StubPrecodeData, /*uint8*/, Type, offsetof(StubPrecodeData, Type))
 CDAC_TYPE_END(StubPrecodeData)
 

src/coreclr/inc/CrstTypes.def

@@ -175,9 +175,6 @@ End
 Crst DbgTransport
 End
 
-Crst DelegateToFPtrHash
-End
-
 Crst GenericDictionaryExpansion
     AcquiredBefore PinnedHeapHandleTable IbcProfile LoaderHeap SystemDomainDelayedUnloadList UniqueStack
 End

src/coreclr/inc/crsttypes_generated.h

@@ -38,88 +38,87 @@ enum CrstType
     CrstDebuggerHeapLock = 20,
     CrstDebuggerJitInfo = 21,
     CrstDebuggerMutex = 22,
-    CrstDelegateToFPtrHash = 23,
-    CrstDynamicIL = 24,
-    CrstDynamicMT = 25,
-    CrstEtwTypeLogHash = 26,
-    CrstEventPipe = 27,
-    CrstEventStore = 28,
-    CrstException = 29,
-    CrstExecutableAllocatorLock = 30,
-    CrstFCall = 31,
-    CrstFrozenObjectHeap = 32,
-    CrstFuncPtrStubs = 33,
-    CrstFusionAppCtx = 34,
-    CrstGCCover = 35,
-    CrstGenericDictionaryExpansion = 36,
-    CrstGlobalStrLiteralMap = 37,
-    CrstHandleTable = 38,
-    CrstIbcProfile = 39,
-    CrstIJWFixupData = 40,
-    CrstIJWHash = 41,
-    CrstILStubGen = 42,
-    CrstInlineTrackingMap = 43,
-    CrstInstMethodHashTable = 44,
-    CrstInterop = 45,
-    CrstInteropData = 46,
-    CrstIsJMCMethod = 47,
-    CrstISymUnmanagedReader = 48,
-    CrstJit = 49,
-    CrstJitInlineTrackingMap = 50,
-    CrstJitPatchpoint = 51,
-    CrstJumpStubCache = 52,
-    CrstLeafLock = 53,
-    CrstListLock = 54,
-    CrstLoaderAllocator = 55,
-    CrstLoaderAllocatorReferences = 56,
-    CrstLoaderHeap = 57,
-    CrstManagedObjectWrapperMap = 58,
-    CrstMethodDescBackpatchInfoTracker = 59,
-    CrstMethodTableExposedObject = 60,
-    CrstModule = 61,
-    CrstModuleLookupTable = 62,
-    CrstMulticoreJitHash = 63,
-    CrstMulticoreJitManager = 64,
-    CrstNativeImageEagerFixups = 65,
-    CrstNativeImageLoad = 66,
-    CrstNotifyGdb = 67,
-    CrstPEImage = 68,
-    CrstPendingTypeLoadEntry = 69,
-    CrstPerfMap = 70,
-    CrstPgoData = 71,
-    CrstPinnedByrefValidation = 72,
-    CrstPinnedHeapHandleTable = 73,
-    CrstProfilerGCRefDataFreeList = 74,
-    CrstProfilingAPIStatus = 75,
-    CrstRCWCache = 76,
-    CrstRCWCleanupList = 77,
-    CrstReadyToRunEntryPointToMethodDescMap = 78,
-    CrstReflection = 79,
-    CrstReJITGlobalRequest = 80,
-    CrstRetThunkCache = 81,
-    CrstSigConvert = 82,
-    CrstSingleUseLock = 83,
-    CrstStressLog = 84,
-    CrstStubCache = 85,
-    CrstStubDispatchCache = 86,
-    CrstSyncBlockCache = 87,
-    CrstSyncHashLock = 88,
-    CrstSystemDomain = 89,
-    CrstSystemDomainDelayedUnloadList = 90,
-    CrstThreadIdDispenser = 91,
-    CrstThreadLocalStorageLock = 92,
-    CrstThreadStore = 93,
-    CrstTieredCompilation = 94,
-    CrstTypeEquivalenceMap = 95,
-    CrstTypeIDMap = 96,
-    CrstUMEntryThunkCache = 97,
-    CrstUMEntryThunkFreeListLock = 98,
-    CrstUniqueStack = 99,
-    CrstUnresolvedClassLock = 100,
-    CrstUnwindInfoTableLock = 101,
-    CrstVSDIndirectionCellLock = 102,
-    CrstWrapperTemplate = 103,
-    kNumberOfCrstTypes = 104
+    CrstDynamicIL = 23,
+    CrstDynamicMT = 24,
+    CrstEtwTypeLogHash = 25,
+    CrstEventPipe = 26,
+    CrstEventStore = 27,
+    CrstException = 28,
+    CrstExecutableAllocatorLock = 29,
+    CrstFCall = 30,
+    CrstFrozenObjectHeap = 31,
+    CrstFuncPtrStubs = 32,
+    CrstFusionAppCtx = 33,
+    CrstGCCover = 34,
+    CrstGenericDictionaryExpansion = 35,
+    CrstGlobalStrLiteralMap = 36,
+    CrstHandleTable = 37,
+    CrstIbcProfile = 38,
+    CrstIJWFixupData = 39,
+    CrstIJWHash = 40,
+    CrstILStubGen = 41,
+    CrstInlineTrackingMap = 42,
+    CrstInstMethodHashTable = 43,
+    CrstInterop = 44,
+    CrstInteropData = 45,
+    CrstIsJMCMethod = 46,
+    CrstISymUnmanagedReader = 47,
+    CrstJit = 48,
+    CrstJitInlineTrackingMap = 49,
+    CrstJitPatchpoint = 50,
+    CrstJumpStubCache = 51,
+    CrstLeafLock = 52,
+    CrstListLock = 53,
+    CrstLoaderAllocator = 54,
+    CrstLoaderAllocatorReferences = 55,
+    CrstLoaderHeap = 56,
+    CrstManagedObjectWrapperMap = 57,
+    CrstMethodDescBackpatchInfoTracker = 58,
+    CrstMethodTableExposedObject = 59,
+    CrstModule = 60,
+    CrstModuleLookupTable = 61,
+    CrstMulticoreJitHash = 62,
+    CrstMulticoreJitManager = 63,
+    CrstNativeImageEagerFixups = 64,
+    CrstNativeImageLoad = 65,
+    CrstNotifyGdb = 66,
+    CrstPEImage = 67,
+    CrstPendingTypeLoadEntry = 68,
+    CrstPerfMap = 69,
+    CrstPgoData = 70,
+    CrstPinnedByrefValidation = 71,
+    CrstPinnedHeapHandleTable = 72,
+    CrstProfilerGCRefDataFreeList = 73,
+    CrstProfilingAPIStatus = 74,
+    CrstRCWCache = 75,
+    CrstRCWCleanupList = 76,
+    CrstReadyToRunEntryPointToMethodDescMap = 77,
+    CrstReflection = 78,
+    CrstReJITGlobalRequest = 79,
+    CrstRetThunkCache = 80,
+    CrstSigConvert = 81,
+    CrstSingleUseLock = 82,
+    CrstStressLog = 83,
+    CrstStubCache = 84,
+    CrstStubDispatchCache = 85,
+    CrstSyncBlockCache = 86,
+    CrstSyncHashLock = 87,
+    CrstSystemDomain = 88,
+    CrstSystemDomainDelayedUnloadList = 89,
+    CrstThreadIdDispenser = 90,
+    CrstThreadLocalStorageLock = 91,
+    CrstThreadStore = 92,
+    CrstTieredCompilation = 93,
+    CrstTypeEquivalenceMap = 94,
+    CrstTypeIDMap = 95,
+    CrstUMEntryThunkCache = 96,
+    CrstUMEntryThunkFreeListLock = 97,
+    CrstUniqueStack = 98,
+    CrstUnresolvedClassLock = 99,
+    CrstUnwindInfoTableLock = 100,
+    CrstVSDIndirectionCellLock = 101,
+    CrstWrapperTemplate = 102,
+    kNumberOfCrstTypes = 103
 };
 
 #endif // __CRST_TYPES_INCLUDED
@@ -153,7 +152,6 @@ int g_rgCrstLevelMap[] =
     0,          // CrstDebuggerHeapLock
     3,          // CrstDebuggerJitInfo
     12,         // CrstDebuggerMutex
-    0,          // CrstDelegateToFPtrHash
     0,          // CrstDynamicIL
     9,          // CrstDynamicMT
     0,          // CrstEtwTypeLogHash
@@ -262,7 +260,6 @@ LPCSTR g_rgCrstNameMap[] =
     "CrstDebuggerHeapLock",
     "CrstDebuggerJitInfo",
     "CrstDebuggerMutex",
-    "CrstDelegateToFPtrHash",
     "CrstDynamicIL",
     "CrstDynamicMT",
     "CrstEtwTypeLogHash",

src/coreclr/inc/dacvars.h

@@ -91,7 +91,6 @@ DEFINE_DACVAR(VMHELPDEF *, dac__hlpDynamicFuncTable, ::hlpDynamicFuncTable)
 DEFINE_DACVAR(PTR_StubManager, StubManager__g_pFirstManager, StubManager::g_pFirstManager)
 DEFINE_DACVAR(PTR_PrecodeStubManager, PrecodeStubManager__g_pManager, PrecodeStubManager::g_pManager)
 DEFINE_DACVAR(PTR_StubLinkStubManager, StubLinkStubManager__g_pManager, StubLinkStubManager::g_pManager)
-DEFINE_DACVAR(PTR_ThunkHeapStubManager, ThunkHeapStubManager__g_pManager, ThunkHeapStubManager::g_pManager)
 DEFINE_DACVAR(PTR_JumpStubStubManager, JumpStubStubManager__g_pManager, JumpStubStubManager::g_pManager)
 DEFINE_DACVAR(PTR_RangeSectionStubManager, RangeSectionStubManager__g_pManager, RangeSectionStubManager::g_pManager)
 DEFINE_DACVAR(PTR_VirtualCallStubManagerManager, VirtualCallStubManagerManager__g_pManager, VirtualCallStubManagerManager::g_pManager)

src/coreclr/inc/vptr_list.h

@@ -24,7 +24,6 @@ VPTR_CLASS(ReflectionModule)
 VPTR_CLASS(PrecodeStubManager)
 VPTR_CLASS(StubLinkStubManager)
 VPTR_CLASS(ThePreStubManager)
-VPTR_CLASS(ThunkHeapStubManager)
 VPTR_CLASS(VirtualCallStubManager)
 VPTR_CLASS(VirtualCallStubManagerManager)
 VPTR_CLASS(JumpStubStubManager)

src/coreclr/vm/amd64/asmconstants.h

@@ -526,8 +526,8 @@ ASMCONSTANTS_C_ASSERT(FixupPrecodeData__PrecodeFixupThunk == offsetof(FixupPreco
 #define StubPrecodeData__Target 0x08
 ASMCONSTANTS_C_ASSERT(StubPrecodeData__Target            == offsetof(StubPrecodeData, Target))
 
-#define StubPrecodeData__MethodDesc 0x00
-ASMCONSTANTS_C_ASSERT(StubPrecodeData__MethodDesc        == offsetof(StubPrecodeData, MethodDesc))
+#define StubPrecodeData__SecretParam 0x00
+ASMCONSTANTS_C_ASSERT(StubPrecodeData__SecretParam        == offsetof(StubPrecodeData, SecretParam))
 
 #define CallCountingStubData__RemainingCallCountCell 0x00
 ASMCONSTANTS_C_ASSERT(CallCountingStubData__RemainingCallCountCell == offsetof(CallCountingStubData, RemainingCallCountCell))

src/coreclr/vm/amd64/cgenamd64.cpp

@@ -553,77 +553,6 @@ void emitJump(LPBYTE pBufferRX, LPBYTE pBufferRW, LPVOID target)
     _ASSERTE(DbgIsExecutable(pBufferRX, 12));
 }
 
-void UMEntryThunkCode::Encode(UMEntryThunkCode *pEntryThunkCodeRX, BYTE* pTargetCode, void* pvSecretParam)
-{
-    CONTRACTL
-    {
-        NOTHROW;
-        GC_NOTRIGGER;
-        MODE_ANY;
-    }
-    CONTRACTL_END;
-
-    // padding                  // CC CC CC CC
-    // mov r10, pUMEntryThunk   // 49 ba xx xx xx xx xx xx xx xx    // METHODDESC_REGISTER
-    // mov rax, pJmpDest        // 48 b8 xx xx xx xx xx xx xx xx    // need to ensure this imm64 is qword aligned
-    // TAILJMP_RAX              // 48 FF E0
-
-#ifdef _DEBUG
-    m_padding[0] = X86_INSTR_INT3;
-    m_padding[1] = X86_INSTR_INT3;
-    m_padding[2] = X86_INSTR_INT3;
-    m_padding[3] = X86_INSTR_INT3;
-#endif // _DEBUG
-    m_movR10[0]  = REX_PREFIX_BASE | REX_OPERAND_SIZE_64BIT | REX_OPCODE_REG_EXT;
-    m_movR10[1]  = 0xBA;
-    m_uet        = pvSecretParam;
-    m_movRAX[0]  = REX_PREFIX_BASE | REX_OPERAND_SIZE_64BIT;
-    m_movRAX[1]  = 0xB8;
-    m_execstub   = pTargetCode;
-    m_jmpRAX[0]  = REX_PREFIX_BASE | REX_OPERAND_SIZE_64BIT;
-    m_jmpRAX[1]  = 0xFF;
-    m_jmpRAX[2]  = 0xE0;
-
-    _ASSERTE(DbgIsExecutable(&pEntryThunkCodeRX->m_movR10[0], &pEntryThunkCodeRX->m_jmpRAX[3]-&pEntryThunkCodeRX->m_movR10[0]));
-    FlushInstructionCache(GetCurrentProcess(),pEntryThunkCodeRX,sizeof(UMEntryThunkCode));
-}
-
-void UMEntryThunkCode::Poison()
-{
-    CONTRACTL
-    {
-        NOTHROW;
-        GC_NOTRIGGER;
-        MODE_ANY;
-    }
-    CONTRACTL_END;
-
-    ExecutableWriterHolder<UMEntryThunkCode> thunkWriterHolder(this, sizeof(UMEntryThunkCode));
-    UMEntryThunkCode *pThisRW = thunkWriterHolder.GetRW();
-
-    pThisRW->m_execstub    = (BYTE *)UMEntryThunk::ReportViolation;
-
-    pThisRW->m_movR10[0]  = REX_PREFIX_BASE | REX_OPERAND_SIZE_64BIT;
-#ifdef _WIN32
-    // mov rcx, pUMEntryThunk // 48 b9 xx xx xx xx xx xx xx xx
-    pThisRW->m_movR10[1]  = 0xB9;
-#else
-    // mov rdi, pUMEntryThunk // 48 bf xx xx xx xx xx xx xx xx
-    pThisRW->m_movR10[1]  = 0xBF;
-#endif
-
-    ClrFlushInstructionCache(&m_movR10[0], &m_jmpRAX[3]-&m_movR10[0], /* hasCodeExecutedBefore */ true);
-}
-
-UMEntryThunk* UMEntryThunk::Decode(LPVOID pCallback)
-{
-    LIMITED_METHOD_CONTRACT;
-
-    UMEntryThunkCode *pThunkCode = (UMEntryThunkCode*)((BYTE*)pCallback - UMEntryThunkCode::GetEntryPointOffset());
-
-    return (UMEntryThunk*)pThunkCode->m_uet;
-}
-
 INT32 rel32UsingJumpStub(INT32 UNALIGNED * pRel32, PCODE target, MethodDesc *pMethod,
     LoaderAllocator *pLoaderAllocator /* = NULL */, bool throwOnOutOfMemoryWithinRange /*= true*/)
 {

src/coreclr/vm/amd64/cgencpu.h

@@ -505,43 +505,6 @@ inline PCODE decodeBackToBackJump(PCODE pCode)
 extern "C" void setFPReturn(int fpSize, INT64 retVal);
 extern "C" void getFPReturn(int fpSize, INT64 *retval);
 
-
-#include <pshpack1.h>
-struct DECLSPEC_ALIGN(8) UMEntryThunkCode
-{
-    // padding                  // CC CC CC CC
-    // mov r10, pUMEntryThunk   // 49 ba xx xx xx xx xx xx xx xx    // METHODDESC_REGISTER
-    // mov rax, pJmpDest        // 48 b8 xx xx xx xx xx xx xx xx    // need to ensure this imm64 is qword aligned
-    // TAILJMP_RAX              // 48 FF E0
-
-    BYTE            m_padding[4];
-    BYTE            m_movR10[2];    // MOV R10,
-    LPVOID          m_uet;          //          pointer to start of this structure
-    BYTE            m_movRAX[2];    // MOV RAX,
-    DECLSPEC_ALIGN(8)
-    const BYTE*     m_execstub;     //          pointer to destination code // ensure this is qword aligned
-    BYTE            m_jmpRAX[3];    // JMP RAX
-    BYTE            m_padding2[5];
-
-    void Encode(UMEntryThunkCode *pEntryThunkCodeRX, BYTE* pTargetCode, void* pvSecretParam);
-    void Poison();
-
-    LPCBYTE GetEntryPoint() const
-    {
-        LIMITED_METHOD_CONTRACT;
-
-        return (LPCBYTE)&m_movR10;
-    }
-
-    static int GetEntryPointOffset()
-    {
-        LIMITED_METHOD_CONTRACT;
-
-        return offsetof(UMEntryThunkCode, m_movR10);
-    }
-};
-#include <poppack.h>
-
 struct HijackArgs
 {
 #ifndef FEATURE_MULTIREG_RETURN

src/coreclr/vm/amd64/thunktemplates.S

@@ -11,7 +11,7 @@ STUB_PAGE_SIZE = 16384
 #define DATA_SLOT(stub, field) C_FUNC(stub##Code) + STUB_PAGE_SIZE + stub##Data__##field
 
 LEAF_ENTRY StubPrecodeCode, _TEXT
-        mov    r10, [rip + DATA_SLOT(StubPrecode, MethodDesc)]
+        mov    r10, [rip + DATA_SLOT(StubPrecode, SecretParam)]
         jmp    [rip + DATA_SLOT(StubPrecode, Target)]
 LEAF_END_MARKED StubPrecodeCode, _TEXT
 

src/coreclr/vm/amd64/thunktemplates.asm

@@ -12,7 +12,7 @@ DATA_SLOT macro stub, field
 endm
 
 LEAF_ENTRY StubPrecodeCode, _TEXT
-        mov    r10, QWORD PTR [DATA_SLOT(StubPrecode, MethodDesc)]
+        mov    r10, QWORD PTR [DATA_SLOT(StubPrecode, SecretParam)]
         jmp    QWORD PTR [DATA_SLOT(StubPrecode, Target)]
 LEAF_END_MARKED StubPrecodeCode, _TEXT