X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTransforms%2FInstrumentation%2FAddressSanitizer.cpp;h=27c97f25044a8dac5ebad520665a01ab5c689dd8;hb=9aef4445a34dc182556ee84b4d446a5afe9daac0;hp=25ca1659c6b73dd37eb6bdcea545e91eee52f74e;hpb=19cd7e9ce28ed7f3326ebcd386eec215ab3763f9;p=oota-llvm.git diff --git a/lib/Transforms/Instrumentation/AddressSanitizer.cpp b/lib/Transforms/Instrumentation/AddressSanitizer.cpp index 25ca1659c6b..27c97f25044 100644 --- a/lib/Transforms/Instrumentation/AddressSanitizer.cpp +++ b/lib/Transforms/Instrumentation/AddressSanitizer.cpp @@ -16,7 +16,6 @@ #define DEBUG_TYPE "asan" #include "llvm/Transforms/Instrumentation.h" -#include "BlackList.h" #include "llvm/ADT/ArrayRef.h" #include "llvm/ADT/DenseMap.h" #include "llvm/ADT/DepthFirstIterator.h" @@ -24,6 +23,7 @@ #include "llvm/ADT/SmallSet.h" #include "llvm/ADT/SmallString.h" #include "llvm/ADT/SmallVector.h" +#include "llvm/ADT/Statistic.h" #include "llvm/ADT/StringExtras.h" #include "llvm/ADT/Triple.h" #include "llvm/DIBuilder.h" @@ -33,18 +33,22 @@ #include "llvm/IR/InlineAsm.h" #include "llvm/IR/IntrinsicInst.h" #include "llvm/IR/LLVMContext.h" +#include "llvm/IR/MDBuilder.h" #include "llvm/IR/Module.h" #include "llvm/IR/Type.h" #include "llvm/InstVisitor.h" +#include "llvm/Support/CallSite.h" #include "llvm/Support/CommandLine.h" #include "llvm/Support/DataTypes.h" #include "llvm/Support/Debug.h" -#include "llvm/Support/raw_ostream.h" +#include "llvm/Support/Endian.h" #include "llvm/Support/system_error.h" -#include "llvm/Target/TargetMachine.h" +#include "llvm/Transforms/Utils/ASanStackFrameLayout.h" #include "llvm/Transforms/Utils/BasicBlockUtils.h" +#include "llvm/Transforms/Utils/Cloning.h" #include "llvm/Transforms/Utils/Local.h" #include "llvm/Transforms/Utils/ModuleUtils.h" +#include "llvm/Transforms/Utils/SpecialCaseList.h" #include #include @@ -53,35 +57,48 @@ using namespace llvm; static const uint64_t kDefaultShadowScale = 3; static const uint64_t kDefaultShadowOffset32 = 1ULL << 29; static const uint64_t kDefaultShadowOffset64 = 1ULL << 44; -static const uint64_t kDefaultShadowOffsetPie = 0; +static const uint64_t kSmallX86_64ShadowOffset = 0x7FFF8000; // < 2G. +static const uint64_t kPPC64_ShadowOffset64 = 1ULL << 41; +static const uint64_t kMIPS32_ShadowOffset32 = 0x0aaa8000; +static const uint64_t kFreeBSD_ShadowOffset32 = 1ULL << 30; +static const uint64_t kFreeBSD_ShadowOffset64 = 1ULL << 46; +static const size_t kMinStackMallocSize = 1 << 6; // 64B static const size_t kMaxStackMallocSize = 1 << 16; // 64K static const uintptr_t kCurrentStackFrameMagic = 0x41B58AB3; static const uintptr_t kRetiredStackFrameMagic = 0x45E0360E; -static const char *kAsanModuleCtorName = "asan.module_ctor"; -static const char *kAsanModuleDtorName = "asan.module_dtor"; -static const int kAsanCtorAndCtorPriority = 1; -static const char *kAsanReportErrorTemplate = "__asan_report_"; -static const char *kAsanRegisterGlobalsName = "__asan_register_globals"; -static const char *kAsanUnregisterGlobalsName = "__asan_unregister_globals"; -static const char *kAsanPoisonGlobalsName = "__asan_before_dynamic_init"; -static const char *kAsanUnpoisonGlobalsName = "__asan_after_dynamic_init"; -static const char *kAsanInitName = "__asan_init"; -static const char *kAsanHandleNoReturnName = "__asan_handle_no_return"; -static const char *kAsanMappingOffsetName = "__asan_mapping_offset"; -static const char *kAsanMappingScaleName = "__asan_mapping_scale"; -static const char *kAsanStackMallocName = "__asan_stack_malloc"; -static const char *kAsanStackFreeName = "__asan_stack_free"; -static const char *kAsanGenPrefix = "__asan_gen_"; -static const char *kAsanPoisonStackMemoryName = "__asan_poison_stack_memory"; -static const char *kAsanUnpoisonStackMemoryName = +static const char *const kAsanModuleCtorName = "asan.module_ctor"; +static const char *const kAsanModuleDtorName = "asan.module_dtor"; +static const int kAsanCtorAndCtorPriority = 1; +static const char *const kAsanReportErrorTemplate = "__asan_report_"; +static const char *const kAsanReportLoadN = "__asan_report_load_n"; +static const char *const kAsanReportStoreN = "__asan_report_store_n"; +static const char *const kAsanRegisterGlobalsName = "__asan_register_globals"; +static const char *const kAsanUnregisterGlobalsName = + "__asan_unregister_globals"; +static const char *const kAsanPoisonGlobalsName = "__asan_before_dynamic_init"; +static const char *const kAsanUnpoisonGlobalsName = "__asan_after_dynamic_init"; +static const char *const kAsanInitName = "__asan_init_v3"; +static const char *const kAsanCovName = "__sanitizer_cov"; +static const char *const kAsanPtrCmp = "__sanitizer_ptr_cmp"; +static const char *const kAsanPtrSub = "__sanitizer_ptr_sub"; +static const char *const kAsanHandleNoReturnName = "__asan_handle_no_return"; +static const int kMaxAsanStackMallocSizeClass = 10; +static const char *const kAsanStackMallocNameTemplate = "__asan_stack_malloc_"; +static const char *const kAsanStackFreeNameTemplate = "__asan_stack_free_"; +static const char *const kAsanGenPrefix = "__asan_gen_"; +static const char *const kAsanPoisonStackMemoryName = + "__asan_poison_stack_memory"; +static const char *const kAsanUnpoisonStackMemoryName = "__asan_unpoison_stack_memory"; -static const int kAsanStackLeftRedzoneMagic = 0xf1; -static const int kAsanStackMidRedzoneMagic = 0xf2; -static const int kAsanStackRightRedzoneMagic = 0xf3; -static const int kAsanStackPartialRedzoneMagic = 0xf4; +static const char *const kAsanOptionDetectUAR = + "__asan_option_detect_stack_use_after_return"; + +#ifndef NDEBUG +static const int kAsanStackAfterReturnMagic = 0xf5; +#endif // Accesses sizes are powers of two: 1, 2, 4, 8, 16. static const size_t kNumberOfAccessSizes = 5; @@ -116,23 +133,41 @@ static cl::opt ClUseAfterReturn("asan-use-after-return", // This flag may need to be replaced with -f[no]asan-globals. static cl::opt ClGlobals("asan-globals", cl::desc("Handle global objects"), cl::Hidden, cl::init(true)); +static cl::opt ClCoverage("asan-coverage", + cl::desc("ASan coverage. 0: none, 1: entry block, 2: all blocks"), + cl::Hidden, cl::init(false)); static cl::opt ClInitializers("asan-initialization-order", cl::desc("Handle C++ initializer order"), cl::Hidden, cl::init(false)); static cl::opt ClMemIntrin("asan-memintrin", cl::desc("Handle memset/memcpy/memmove"), cl::Hidden, cl::init(true)); -static cl::opt ClRealignStack("asan-realign-stack", - cl::desc("Realign stack to 32"), cl::Hidden, cl::init(true)); +static cl::opt ClInvalidPointerPairs("asan-detect-invalid-pointer-pair", + cl::desc("Instrument <, <=, >, >=, - with pointer operands"), + cl::Hidden, cl::init(false)); +static cl::opt ClRealignStack("asan-realign-stack", + cl::desc("Realign stack to the value of this flag (power of two)"), + cl::Hidden, cl::init(32)); static cl::opt ClBlacklistFile("asan-blacklist", cl::desc("File containing the list of objects to ignore " "during instrumentation"), cl::Hidden); +// This is an experimental feature that will allow to choose between +// instrumented and non-instrumented code at link-time. +// If this option is on, just before instrumenting a function we create its +// clone; if the function is not changed by asan the clone is deleted. +// If we end up with a clone, we put the instrumented function into a section +// called "ASAN" and the uninstrumented function into a section called "NOASAN". +// +// This is still a prototype, we need to figure out a way to keep two copies of +// a function so that the linker can easily choose one of them. +static cl::opt ClKeepUninstrumented("asan-keep-uninstrumented-functions", + cl::desc("Keep uninstrumented copies of functions"), + cl::Hidden, cl::init(false)); + // These flags allow to change the shadow mapping. // The shadow mapping looks like // Shadow = (Mem >> scale) + (1 << offset_log) static cl::opt ClMappingScale("asan-mapping-scale", cl::desc("scale of asan shadow mapping"), cl::Hidden, cl::init(0)); -static cl::opt ClMappingOffsetLog("asan-mapping-offset-log", - cl::desc("offset of asan shadow mapping"), cl::Hidden, cl::init(-1)); // Optimization flags. Not user visible, used mostly for testing // and benchmarking the tool. @@ -160,6 +195,13 @@ static cl::opt ClDebugMin("asan-debug-min", cl::desc("Debug min inst"), static cl::opt ClDebugMax("asan-debug-max", cl::desc("Debug man inst"), cl::Hidden, cl::init(-1)); +STATISTIC(NumInstrumentedReads, "Number of instrumented reads"); +STATISTIC(NumInstrumentedWrites, "Number of instrumented writes"); +STATISTIC(NumOptimizedAccessesToGlobalArray, + "Number of optimized accesses to global arrays"); +STATISTIC(NumOptimizedAccessesToGlobalVar, + "Number of optimized accesses to global vars"); + namespace { /// A set of dynamically initialized globals extracted from metadata. class SetOfDynamicallyInitializedGlobals { @@ -187,23 +229,45 @@ class SetOfDynamicallyInitializedGlobals { }; /// This struct defines the shadow mapping using the rule: -/// shadow = (mem >> Scale) + Offset. +/// shadow = (mem >> Scale) ADD-or-OR Offset. struct ShadowMapping { int Scale; uint64_t Offset; + bool OrShadowOffset; }; static ShadowMapping getShadowMapping(const Module &M, int LongSize) { - llvm::Triple targetTriple(M.getTargetTriple()); - bool isAndroid = targetTriple.getEnvironment() == llvm::Triple::Android; + llvm::Triple TargetTriple(M.getTargetTriple()); + bool IsAndroid = TargetTriple.getEnvironment() == llvm::Triple::Android; + // bool IsMacOSX = TargetTriple.getOS() == llvm::Triple::MacOSX; + bool IsFreeBSD = TargetTriple.getOS() == llvm::Triple::FreeBSD; + bool IsLinux = TargetTriple.getOS() == llvm::Triple::Linux; + bool IsPPC64 = TargetTriple.getArch() == llvm::Triple::ppc64 || + TargetTriple.getArch() == llvm::Triple::ppc64le; + bool IsX86_64 = TargetTriple.getArch() == llvm::Triple::x86_64; + bool IsMIPS32 = TargetTriple.getArch() == llvm::Triple::mips || + TargetTriple.getArch() == llvm::Triple::mipsel; ShadowMapping Mapping; - Mapping.Offset = isAndroid ? kDefaultShadowOffsetPie : - (LongSize == 32 ? kDefaultShadowOffset32 : kDefaultShadowOffset64); - if (ClMappingOffsetLog >= 0) { - // Zero offset log is the special case. - Mapping.Offset = (ClMappingOffsetLog == 0) ? 0 : 1ULL << ClMappingOffsetLog; + if (LongSize == 32) { + if (IsAndroid) + Mapping.Offset = 0; + else if (IsMIPS32) + Mapping.Offset = kMIPS32_ShadowOffset32; + else if (IsFreeBSD) + Mapping.Offset = kFreeBSD_ShadowOffset32; + else + Mapping.Offset = kDefaultShadowOffset32; + } else { // LongSize == 64 + if (IsPPC64) + Mapping.Offset = kPPC64_ShadowOffset64; + else if (IsFreeBSD) + Mapping.Offset = kFreeBSD_ShadowOffset64; + else if (IsLinux && IsX86_64) + Mapping.Offset = kSmallX86_64ShadowOffset; + else + Mapping.Offset = kDefaultShadowOffset64; } Mapping.Scale = kDefaultShadowScale; @@ -211,6 +275,11 @@ static ShadowMapping getShadowMapping(const Module &M, int LongSize) { Mapping.Scale = ClMappingScale; } + // OR-ing shadow offset if more efficient (at least on x86) if the offset + // is a power of two, but on ppc64 we have to use add since the shadow + // offset is not necessary 1/8-th of the address space. + Mapping.OrShadowOffset = !IsPPC64 && !(Mapping.Offset & (Mapping.Offset - 1)); + return Mapping; } @@ -222,7 +291,7 @@ static size_t RedzoneSizeForScale(int MappingScale) { /// AddressSanitizer: instrument the code in module to find memory bugs. struct AddressSanitizer : public FunctionPass { - AddressSanitizer(bool CheckInitOrder = false, + AddressSanitizer(bool CheckInitOrder = true, bool CheckUseAfterReturn = false, bool CheckLifetime = false, StringRef BlacklistFile = StringRef()) @@ -236,22 +305,22 @@ struct AddressSanitizer : public FunctionPass { return "AddressSanitizerFunctionPass"; } void instrumentMop(Instruction *I); - void instrumentAddress(Instruction *OrigIns, IRBuilder<> &IRB, - Value *Addr, uint32_t TypeSize, bool IsWrite); + void instrumentPointerComparisonOrSubtraction(Instruction *I); + void instrumentAddress(Instruction *OrigIns, Instruction *InsertBefore, + Value *Addr, uint32_t TypeSize, bool IsWrite, + Value *SizeArgument); Value *createSlowPathCmp(IRBuilder<> &IRB, Value *AddrLong, Value *ShadowValue, uint32_t TypeSize); Instruction *generateCrashCode(Instruction *InsertBefore, Value *Addr, - bool IsWrite, size_t AccessSizeIndex); + bool IsWrite, size_t AccessSizeIndex, + Value *SizeArgument); bool instrumentMemIntrinsic(MemIntrinsic *MI); void instrumentMemIntrinsicParam(Instruction *OrigIns, Value *Addr, Value *Size, Instruction *InsertBefore, bool IsWrite); Value *memToShadow(Value *Shadow, IRBuilder<> &IRB); bool runOnFunction(Function &F); - void createInitializerPoisonCalls(Module &M, - Value *FirstAddr, Value *LastAddr); bool maybeInsertAsanInitAtFunctionEntry(Function &F); - void emitShadowMapping(Module &M, IRBuilder<> &IRB) const; virtual bool doInitialization(Module &M); static char ID; // Pass identification, replacement for typeid @@ -261,22 +330,30 @@ struct AddressSanitizer : public FunctionPass { bool ShouldInstrumentGlobal(GlobalVariable *G); bool LooksLikeCodeInBug11395(Instruction *I); void FindDynamicInitializers(Module &M); + bool GlobalIsLinkerInitialized(GlobalVariable *G); + bool InjectCoverage(Function &F, const ArrayRef AllBlocks); + void InjectCoverageAtBlock(Function &F, BasicBlock &BB); bool CheckInitOrder; bool CheckUseAfterReturn; bool CheckLifetime; + SmallString<64> BlacklistFile; + LLVMContext *C; - DataLayout *TD; + const DataLayout *DL; int LongSize; Type *IntptrTy; ShadowMapping Mapping; Function *AsanCtorFunction; Function *AsanInitFunction; Function *AsanHandleNoReturnFunc; - SmallString<64> BlacklistFile; - OwningPtr BL; + Function *AsanCovFunction; + Function *AsanPtrCmpFunction, *AsanPtrSubFunction; + OwningPtr BL; // This array is indexed by AccessIsWrite and log2(AccessSize). Function *AsanErrorCallback[2][kNumberOfAccessSizes]; + // This array is indexed by AccessIsWrite. + Function *AsanErrorCallbackSized[2]; InlineAsm *EmptyAsm; SetOfDynamicallyInitializedGlobals DynamicallyInitializedGlobals; @@ -285,7 +362,7 @@ struct AddressSanitizer : public FunctionPass { class AddressSanitizerModule : public ModulePass { public: - AddressSanitizerModule(bool CheckInitOrder = false, + AddressSanitizerModule(bool CheckInitOrder = true, StringRef BlacklistFile = StringRef()) : ModulePass(ID), CheckInitOrder(CheckInitOrder || ClInitializers), @@ -301,19 +378,19 @@ class AddressSanitizerModule : public ModulePass { void initializeCallbacks(Module &M); bool ShouldInstrumentGlobal(GlobalVariable *G); - void createInitializerPoisonCalls(Module &M, Value *FirstAddr, - Value *LastAddr); - size_t RedzoneSize() const { + void createInitializerPoisonCalls(Module &M, GlobalValue *ModuleName); + size_t MinRedzoneSizeForGlobal() const { return RedzoneSizeForScale(Mapping.Scale); } bool CheckInitOrder; SmallString<64> BlacklistFile; - OwningPtr BL; + + OwningPtr BL; SetOfDynamicallyInitializedGlobals DynamicallyInitializedGlobals; Type *IntptrTy; LLVMContext *C; - DataLayout *TD; + const DataLayout *DL; ShadowMapping Mapping; Function *AsanPoisonGlobals; Function *AsanUnpoisonGlobals; @@ -341,15 +418,16 @@ struct FunctionStackPoisoner : public InstVisitor { SmallVector AllocaVec; SmallVector RetVec; - uint64_t TotalStackSize; unsigned StackAlignment; - Function *AsanStackMallocFunc, *AsanStackFreeFunc; + Function *AsanStackMallocFunc[kMaxAsanStackMallocSizeClass + 1], + *AsanStackFreeFunc[kMaxAsanStackMallocSizeClass + 1]; Function *AsanPoisonStackMemoryFunc, *AsanUnpoisonStackMemoryFunc; // Stores a place and arguments of poisoning/unpoisoning call for alloca. struct AllocaPoisonCall { IntrinsicInst *InsBefore; + AllocaInst *AI; uint64_t Size; bool DoPoison; }; @@ -363,7 +441,7 @@ struct FunctionStackPoisoner : public InstVisitor { : F(F), ASan(ASan), DIB(*F.getParent()), C(ASan.C), IntptrTy(ASan.IntptrTy), IntptrPtrTy(PointerType::get(IntptrTy, 0)), Mapping(ASan.Mapping), - TotalStackSize(0), StackAlignment(1 << Mapping.Scale) {} + StackAlignment(1 << Mapping.Scale) {} bool runOnFunction() { if (!ClStack) return false; @@ -402,8 +480,6 @@ struct FunctionStackPoisoner : public InstVisitor { StackAlignment = std::max(StackAlignment, AI.getAlignment()); AllocaVec.push_back(&AI); - uint64_t AlignedSize = getAlignedAllocaSize(&AI); - TotalStackSize += AlignedSize; } /// \brief Collect lifetime intrinsic calls to check for use-after-scope @@ -428,7 +504,7 @@ struct FunctionStackPoisoner : public InstVisitor { AllocaInst *AI = findAllocaForValue(II.getArgOperand(1)); if (!AI) return; bool DoPoison = (ID == Intrinsic::lifetime_end); - AllocaPoisonCall APC = {&II, SizeValue, DoPoison}; + AllocaPoisonCall APC = {&II, AI, SizeValue, DoPoison}; AllocaPoisonCallVec.push_back(APC); } @@ -436,33 +512,26 @@ struct FunctionStackPoisoner : public InstVisitor { void initializeCallbacks(Module &M); // Check if we want (and can) handle this alloca. - bool isInterestingAlloca(AllocaInst &AI) { - return (!AI.isArrayAllocation() && - AI.isStaticAlloca() && - AI.getAllocatedType()->isSized()); + bool isInterestingAlloca(AllocaInst &AI) const { + return (!AI.isArrayAllocation() && AI.isStaticAlloca() && + AI.getAllocatedType()->isSized() && + // alloca() may be called with 0 size, ignore it. + getAllocaSizeInBytes(&AI) > 0); } - size_t RedzoneSize() const { - return RedzoneSizeForScale(Mapping.Scale); - } - uint64_t getAllocaSizeInBytes(AllocaInst *AI) { + uint64_t getAllocaSizeInBytes(AllocaInst *AI) const { Type *Ty = AI->getAllocatedType(); - uint64_t SizeInBytes = ASan.TD->getTypeAllocSize(Ty); + uint64_t SizeInBytes = ASan.DL->getTypeAllocSize(Ty); return SizeInBytes; } - uint64_t getAlignedSize(uint64_t SizeInBytes) { - size_t RZ = RedzoneSize(); - return ((SizeInBytes + RZ - 1) / RZ) * RZ; - } - uint64_t getAlignedAllocaSize(AllocaInst *AI) { - uint64_t SizeInBytes = getAllocaSizeInBytes(AI); - return getAlignedSize(SizeInBytes); - } /// Finds alloca where the value comes from. AllocaInst *findAllocaForValue(Value *V); - void poisonRedZones(const ArrayRef &AllocaVec, IRBuilder<> IRB, + void poisonRedZones(const ArrayRef ShadowBytes, IRBuilder<> &IRB, Value *ShadowBase, bool DoPoison); - void poisonAlloca(Value *V, uint64_t Size, IRBuilder<> IRB, bool DoPoison); + void poisonAlloca(Value *V, uint64_t Size, IRBuilder<> &IRB, bool DoPoison); + + void SetShadowToStackAfterReturnInlined(IRBuilder<> &IRB, Value *ShadowBase, + int Size); }; } // namespace @@ -488,17 +557,24 @@ ModulePass *llvm::createAddressSanitizerModulePass( } static size_t TypeSizeToSizeIndex(uint32_t TypeSize) { - size_t Res = CountTrailingZeros_32(TypeSize / 8); + size_t Res = countTrailingZeros(TypeSize / 8); assert(Res < kNumberOfAccessSizes); return Res; } -// Create a constant for Str so that we can pass it to the run-time lib. -static GlobalVariable *createPrivateGlobalForString(Module &M, StringRef Str) { +// \brief Create a constant for Str so that we can pass it to the run-time lib. +static GlobalVariable *createPrivateGlobalForString( + Module &M, StringRef Str, bool AllowMerging) { Constant *StrConst = ConstantDataArray::getString(M.getContext(), Str); - return new GlobalVariable(M, StrConst->getType(), true, - GlobalValue::PrivateLinkage, StrConst, - kAsanGenPrefix); + // We use private linkage for module-local strings. If they can be merged + // with another one, we set the unnamed_addr attribute. + GlobalVariable *GV = + new GlobalVariable(M, StrConst->getType(), true, + GlobalValue::PrivateLinkage, StrConst, kAsanGenPrefix); + if (AllowMerging) + GV->setUnnamedAddr(true); + GV->setAlignment(1); // Strings may not be merged w/o setting align 1. + return GV; } static bool GlobalWasGeneratedByAsan(GlobalVariable *G) { @@ -511,28 +587,26 @@ Value *AddressSanitizer::memToShadow(Value *Shadow, IRBuilder<> &IRB) { if (Mapping.Offset == 0) return Shadow; // (Shadow >> scale) | offset - return IRB.CreateOr(Shadow, ConstantInt::get(IntptrTy, - Mapping.Offset)); + if (Mapping.OrShadowOffset) + return IRB.CreateOr(Shadow, ConstantInt::get(IntptrTy, Mapping.Offset)); + else + return IRB.CreateAdd(Shadow, ConstantInt::get(IntptrTy, Mapping.Offset)); } void AddressSanitizer::instrumentMemIntrinsicParam( Instruction *OrigIns, Value *Addr, Value *Size, Instruction *InsertBefore, bool IsWrite) { + IRBuilder<> IRB(InsertBefore); + if (Size->getType() != IntptrTy) + Size = IRB.CreateIntCast(Size, IntptrTy, false); // Check the first byte. - { - IRBuilder<> IRB(InsertBefore); - instrumentAddress(OrigIns, IRB, Addr, 8, IsWrite); - } + instrumentAddress(OrigIns, InsertBefore, Addr, 8, IsWrite, Size); // Check the last byte. - { - IRBuilder<> IRB(InsertBefore); - Value *SizeMinusOne = IRB.CreateSub( - Size, ConstantInt::get(Size->getType(), 1)); - SizeMinusOne = IRB.CreateIntCast(SizeMinusOne, IntptrTy, false); - Value *AddrLong = IRB.CreatePointerCast(Addr, IntptrTy); - Value *AddrPlusSizeMinisOne = IRB.CreateAdd(AddrLong, SizeMinusOne); - instrumentAddress(OrigIns, IRB, AddrPlusSizeMinisOne, 8, IsWrite); - } + IRB.SetInsertPoint(InsertBefore); + Value *SizeMinusOne = IRB.CreateSub(Size, ConstantInt::get(IntptrTy, 1)); + Value *AddrLong = IRB.CreatePointerCast(Addr, IntptrTy); + Value *AddrLast = IRB.CreateAdd(AddrLong, SizeMinusOne); + instrumentAddress(OrigIns, InsertBefore, AddrLast, 8, IsWrite, Size); } // Instrument memset/memmove/memcpy @@ -552,7 +626,7 @@ bool AddressSanitizer::instrumentMemIntrinsic(MemIntrinsic *MI) { Value *Cmp = IRB.CreateICmpNE(Length, Constant::getNullValue(Length->getType())); - InsertBefore = SplitBlockAndInsertIfThen(cast(Cmp), false); + InsertBefore = SplitBlockAndInsertIfThen(Cmp, InsertBefore, false); } instrumentMemIntrinsicParam(MI, Dst, Length, InsertBefore, true); @@ -587,6 +661,48 @@ static Value *isInterestingMemoryAccess(Instruction *I, bool *IsWrite) { return NULL; } +static bool isPointerOperand(Value *V) { + return V->getType()->isPointerTy() || isa(V); +} + +// This is a rough heuristic; it may cause both false positives and +// false negatives. The proper implementation requires cooperation with +// the frontend. +static bool isInterestingPointerComparisonOrSubtraction(Instruction *I) { + if (ICmpInst *Cmp = dyn_cast(I)) { + if (!Cmp->isRelational()) + return false; + } else if (BinaryOperator *BO = dyn_cast(I)) { + if (BO->getOpcode() != Instruction::Sub) + return false; + } else { + return false; + } + if (!isPointerOperand(I->getOperand(0)) || + !isPointerOperand(I->getOperand(1))) + return false; + return true; +} + +bool AddressSanitizer::GlobalIsLinkerInitialized(GlobalVariable *G) { + // If a global variable does not have dynamic initialization we don't + // have to instrument it. However, if a global does not have initializer + // at all, we assume it has dynamic initializer (in other TU). + return G->hasInitializer() && !DynamicallyInitializedGlobals.Contains(G); +} + +void +AddressSanitizer::instrumentPointerComparisonOrSubtraction(Instruction *I) { + IRBuilder<> IRB(I); + Function *F = isa(I) ? AsanPtrCmpFunction : AsanPtrSubFunction; + Value *Param[2] = {I->getOperand(0), I->getOperand(1)}; + for (int i = 0; i < 2; i++) { + if (Param[i]->getType()->isPointerTy()) + Param[i] = IRB.CreatePointerCast(Param[i], IntptrTy); + } + IRB.CreateCall2(F, Param[0], Param[1]); +} + void AddressSanitizer::instrumentMop(Instruction *I) { bool IsWrite = false; Value *Addr = isInterestingMemoryAccess(I, &IsWrite); @@ -595,13 +711,19 @@ void AddressSanitizer::instrumentMop(Instruction *I) { if (GlobalVariable *G = dyn_cast(Addr)) { // If initialization order checking is disabled, a simple access to a // dynamically initialized global is always valid. - if (!CheckInitOrder) - return; - // If a global variable does not have dynamic initialization we don't - // have to instrument it. However, if a global does not have initailizer - // at all, we assume it has dynamic initializer (in other TU). - if (G->hasInitializer() && !DynamicallyInitializedGlobals.Contains(G)) + if (!CheckInitOrder || GlobalIsLinkerInitialized(G)) { + NumOptimizedAccessesToGlobalVar++; return; + } + } + ConstantExpr *CE = dyn_cast(Addr); + if (CE && CE->isGEPWithNoNotionalOverIndexing()) { + if (GlobalVariable *G = dyn_cast(CE->getOperand(0))) { + if (CE->getOperand(1)->isNullValue() && GlobalIsLinkerInitialized(G)) { + NumOptimizedAccessesToGlobalArray++; + return; + } + } } } @@ -609,16 +731,31 @@ void AddressSanitizer::instrumentMop(Instruction *I) { Type *OrigTy = cast(OrigPtrTy)->getElementType(); assert(OrigTy->isSized()); - uint32_t TypeSize = TD->getTypeStoreSizeInBits(OrigTy); - - if (TypeSize != 8 && TypeSize != 16 && - TypeSize != 32 && TypeSize != 64 && TypeSize != 128) { - // Ignore all unusual sizes. - return; - } - + uint32_t TypeSize = DL->getTypeStoreSizeInBits(OrigTy); + + assert((TypeSize % 8) == 0); + + if (IsWrite) + NumInstrumentedWrites++; + else + NumInstrumentedReads++; + + // Instrument a 1-, 2-, 4-, 8-, or 16- byte access with one check. + if (TypeSize == 8 || TypeSize == 16 || + TypeSize == 32 || TypeSize == 64 || TypeSize == 128) + return instrumentAddress(I, I, Addr, TypeSize, IsWrite, 0); + // Instrument unusual size (but still multiple of 8). + // We can not do it with a single check, so we do 1-byte check for the first + // and the last bytes. We call __asan_report_*_n(addr, real_size) to be able + // to report the actual access size. IRBuilder<> IRB(I); - instrumentAddress(I, IRB, Addr, TypeSize, IsWrite); + Value *LastByte = IRB.CreateIntToPtr( + IRB.CreateAdd(IRB.CreatePointerCast(Addr, IntptrTy), + ConstantInt::get(IntptrTy, TypeSize / 8 - 1)), + OrigPtrTy); + Value *Size = ConstantInt::get(IntptrTy, TypeSize / 8); + instrumentAddress(I, I, Addr, 8, IsWrite, Size); + instrumentAddress(I, I, LastByte, 8, IsWrite, Size); } // Validate the result of Module::getOrInsertFunction called for an interface @@ -634,10 +771,12 @@ static Function *checkInterfaceFunction(Constant *FuncOrBitcast) { Instruction *AddressSanitizer::generateCrashCode( Instruction *InsertBefore, Value *Addr, - bool IsWrite, size_t AccessSizeIndex) { + bool IsWrite, size_t AccessSizeIndex, Value *SizeArgument) { IRBuilder<> IRB(InsertBefore); - CallInst *Call = IRB.CreateCall(AsanErrorCallback[IsWrite][AccessSizeIndex], - Addr); + CallInst *Call = SizeArgument + ? IRB.CreateCall2(AsanErrorCallbackSized[IsWrite], Addr, SizeArgument) + : IRB.CreateCall(AsanErrorCallback[IsWrite][AccessSizeIndex], Addr); + // We don't do Call->setDoesNotReturn() because the BB already has // UnreachableInst at the end. // This EmptyAsm is required to avoid callback merge. @@ -664,8 +803,10 @@ Value *AddressSanitizer::createSlowPathCmp(IRBuilder<> &IRB, Value *AddrLong, } void AddressSanitizer::instrumentAddress(Instruction *OrigIns, - IRBuilder<> &IRB, Value *Addr, - uint32_t TypeSize, bool IsWrite) { + Instruction *InsertBefore, + Value *Addr, uint32_t TypeSize, + bool IsWrite, Value *SizeArgument) { + IRBuilder<> IRB(InsertBefore); Value *AddrLong = IRB.CreatePointerCast(Addr, IntptrTy); Type *ShadowTy = IntegerType::get( @@ -683,7 +824,7 @@ void AddressSanitizer::instrumentAddress(Instruction *OrigIns, if (ClAlwaysSlowPath || (TypeSize < 8 * Granularity)) { TerminatorInst *CheckTerm = - SplitBlockAndInsertIfThen(cast(Cmp), false); + SplitBlockAndInsertIfThen(Cmp, InsertBefore, false); assert(dyn_cast(CheckTerm)->isUnconditional()); BasicBlock *NextBB = CheckTerm->getSuccessor(0); IRB.SetInsertPoint(CheckTerm); @@ -694,16 +835,16 @@ void AddressSanitizer::instrumentAddress(Instruction *OrigIns, BranchInst *NewTerm = BranchInst::Create(CrashBlock, NextBB, Cmp2); ReplaceInstWithInst(CheckTerm, NewTerm); } else { - CrashTerm = SplitBlockAndInsertIfThen(cast(Cmp), true); + CrashTerm = SplitBlockAndInsertIfThen(Cmp, InsertBefore, true); } - Instruction *Crash = - generateCrashCode(CrashTerm, AddrLong, IsWrite, AccessSizeIndex); + Instruction *Crash = generateCrashCode( + CrashTerm, AddrLong, IsWrite, AccessSizeIndex, SizeArgument); Crash->setDebugLoc(OrigIns->getDebugLoc()); } void AddressSanitizerModule::createInitializerPoisonCalls( - Module &M, Value *FirstAddr, Value *LastAddr) { + Module &M, GlobalValue *ModuleName) { // We do all of our poisoning and unpoisoning within _GLOBAL__I_a. Function *GlobalInit = M.getFunction("_GLOBAL__I_a"); // If that function is not present, this TU contains no globals, or they have @@ -715,7 +856,8 @@ void AddressSanitizerModule::createInitializerPoisonCalls( IRBuilder<> IRB(GlobalInit->begin()->getFirstInsertionPt()); // Add a call to poison all external globals before the given function starts. - IRB.CreateCall2(AsanPoisonGlobals, FirstAddr, LastAddr); + Value *ModuleNameAddr = ConstantExpr::getPointerCast(ModuleName, IntptrTy); + IRB.CreateCall(AsanPoisonGlobals, ModuleNameAddr); // Add calls to unpoison all globals before each return instruction. for (Function::iterator I = GlobalInit->begin(), E = GlobalInit->end(); @@ -745,8 +887,8 @@ bool AddressSanitizerModule::ShouldInstrumentGlobal(GlobalVariable *G) { // - Need to poison all copies, not just the main thread's one. if (G->isThreadLocal()) return false; - // For now, just ignore this Alloca if the alignment is large. - if (G->getAlignment() > RedzoneSize()) return false; + // For now, just ignore this Global if the alignment is large. + if (G->getAlignment() > MinRedzoneSizeForGlobal()) return false; // Ignore all the globals with the names starting with "\01L_OBJC_". // Many of those are put into the .cstring section. The linker compresses @@ -789,7 +931,7 @@ void AddressSanitizerModule::initializeCallbacks(Module &M) { IRBuilder<> IRB(*C); // Declare our poisoning and unpoisoning functions. AsanPoisonGlobals = checkInterfaceFunction(M.getOrInsertFunction( - kAsanPoisonGlobalsName, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + kAsanPoisonGlobalsName, IRB.getVoidTy(), IntptrTy, NULL)); AsanPoisonGlobals->setLinkage(Function::ExternalLinkage); AsanUnpoisonGlobals = checkInterfaceFunction(M.getOrInsertFunction( kAsanUnpoisonGlobalsName, IRB.getVoidTy(), NULL)); @@ -810,13 +952,16 @@ void AddressSanitizerModule::initializeCallbacks(Module &M) { // redzones and inserts this function into llvm.global_ctors. bool AddressSanitizerModule::runOnModule(Module &M) { if (!ClGlobals) return false; - TD = getAnalysisIfAvailable(); - if (!TD) + + DataLayoutPass *DLP = getAnalysisIfAvailable(); + if (!DLP) return false; - BL.reset(new BlackList(BlacklistFile)); + DL = &DLP->getDataLayout(); + + BL.reset(SpecialCaseList::createOrDie(BlacklistFile)); if (BL->isIn(M)) return false; C = &(M.getContext()); - int LongSize = TD->getPointerSizeInBits(); + int LongSize = DL->getPointerSizeInBits(); IntptrTy = Type::getIntNTy(*C, LongSize); Mapping = getShadowMapping(M, LongSize); initializeCallbacks(M); @@ -838,53 +983,66 @@ bool AddressSanitizerModule::runOnModule(Module &M) { // size_t size; // size_t size_with_redzone; // const char *name; + // const char *module_name; // size_t has_dynamic_init; // We initialize an array of such structures and pass it to a run-time call. StructType *GlobalStructTy = StructType::get(IntptrTy, IntptrTy, IntptrTy, IntptrTy, - IntptrTy, NULL); - SmallVector Initializers(n), DynamicInit; - + IntptrTy, IntptrTy, NULL); + SmallVector Initializers(n); Function *CtorFunc = M.getFunction(kAsanModuleCtorName); assert(CtorFunc); IRBuilder<> IRB(CtorFunc->getEntryBlock().getTerminator()); - // The addresses of the first and last dynamically initialized globals in - // this TU. Used in initialization order checking. - Value *FirstDynamic = 0, *LastDynamic = 0; + bool HasDynamicallyInitializedGlobals = false; + + // We shouldn't merge same module names, as this string serves as unique + // module ID in runtime. + GlobalVariable *ModuleName = createPrivateGlobalForString( + M, M.getModuleIdentifier(), /*AllowMerging*/false); for (size_t i = 0; i < n; i++) { + static const uint64_t kMaxGlobalRedzone = 1 << 18; GlobalVariable *G = GlobalsToChange[i]; PointerType *PtrTy = cast(G->getType()); Type *Ty = PtrTy->getElementType(); - uint64_t SizeInBytes = TD->getTypeAllocSize(Ty); - size_t RZ = RedzoneSize(); - uint64_t RightRedzoneSize = RZ + (RZ - (SizeInBytes % RZ)); + uint64_t SizeInBytes = DL->getTypeAllocSize(Ty); + uint64_t MinRZ = MinRedzoneSizeForGlobal(); + // MinRZ <= RZ <= kMaxGlobalRedzone + // and trying to make RZ to be ~ 1/4 of SizeInBytes. + uint64_t RZ = std::max(MinRZ, + std::min(kMaxGlobalRedzone, + (SizeInBytes / MinRZ / 4) * MinRZ)); + uint64_t RightRedzoneSize = RZ; + // Round up to MinRZ + if (SizeInBytes % MinRZ) + RightRedzoneSize += MinRZ - (SizeInBytes % MinRZ); + assert(((RightRedzoneSize + SizeInBytes) % MinRZ) == 0); Type *RightRedZoneTy = ArrayType::get(IRB.getInt8Ty(), RightRedzoneSize); // Determine whether this global should be poisoned in initialization. bool GlobalHasDynamicInitializer = DynamicallyInitializedGlobals.Contains(G); // Don't check initialization order if this global is blacklisted. - GlobalHasDynamicInitializer &= !BL->isInInit(*G); + GlobalHasDynamicInitializer &= !BL->isIn(*G, "init"); StructType *NewTy = StructType::get(Ty, RightRedZoneTy, NULL); Constant *NewInitializer = ConstantStruct::get( NewTy, G->getInitializer(), Constant::getNullValue(RightRedZoneTy), NULL); - SmallString<2048> DescriptionOfGlobal = G->getName(); - DescriptionOfGlobal += " ("; - DescriptionOfGlobal += M.getModuleIdentifier(); - DescriptionOfGlobal += ")"; - GlobalVariable *Name = createPrivateGlobalForString(M, DescriptionOfGlobal); + GlobalVariable *Name = + createPrivateGlobalForString(M, G->getName(), /*AllowMerging*/true); // Create a new global variable with enough space for a redzone. + GlobalValue::LinkageTypes Linkage = G->getLinkage(); + if (G->isConstant() && Linkage == GlobalValue::PrivateLinkage) + Linkage = GlobalValue::InternalLinkage; GlobalVariable *NewGlobal = new GlobalVariable( - M, NewTy, G->isConstant(), G->getLinkage(), + M, NewTy, G->isConstant(), Linkage, NewInitializer, "", G, G->getThreadLocalMode()); NewGlobal->copyAttributesFrom(G); - NewGlobal->setAlignment(RZ); + NewGlobal->setAlignment(MinRZ); Value *Indices2[2]; Indices2[0] = IRB.getInt32(0); @@ -901,27 +1059,25 @@ bool AddressSanitizerModule::runOnModule(Module &M) { ConstantInt::get(IntptrTy, SizeInBytes), ConstantInt::get(IntptrTy, SizeInBytes + RightRedzoneSize), ConstantExpr::getPointerCast(Name, IntptrTy), + ConstantExpr::getPointerCast(ModuleName, IntptrTy), ConstantInt::get(IntptrTy, GlobalHasDynamicInitializer), NULL); // Populate the first and last globals declared in this TU. - if (CheckInitOrder && GlobalHasDynamicInitializer) { - LastDynamic = ConstantExpr::getPointerCast(NewGlobal, IntptrTy); - if (FirstDynamic == 0) - FirstDynamic = LastDynamic; - } + if (CheckInitOrder && GlobalHasDynamicInitializer) + HasDynamicallyInitializedGlobals = true; DEBUG(dbgs() << "NEW GLOBAL: " << *NewGlobal << "\n"); } ArrayType *ArrayOfGlobalStructTy = ArrayType::get(GlobalStructTy, n); GlobalVariable *AllGlobals = new GlobalVariable( - M, ArrayOfGlobalStructTy, false, GlobalVariable::PrivateLinkage, + M, ArrayOfGlobalStructTy, false, GlobalVariable::InternalLinkage, ConstantArray::get(ArrayOfGlobalStructTy, Initializers), ""); // Create calls for poisoning before initializers run and unpoisoning after. - if (CheckInitOrder && FirstDynamic && LastDynamic) - createInitializerPoisonCalls(M, FirstDynamic, LastDynamic); + if (CheckInitOrder && HasDynamicallyInitializedGlobals) + createInitializerPoisonCalls(M, ModuleName); IRB.CreateCall2(AsanRegisterGlobals, IRB.CreatePointerCast(AllGlobals, IntptrTy), ConstantInt::get(IntptrTy, n)); @@ -957,49 +1113,38 @@ void AddressSanitizer::initializeCallbacks(Module &M) { FunctionName, IRB.getVoidTy(), IntptrTy, NULL)); } } + AsanErrorCallbackSized[0] = checkInterfaceFunction(M.getOrInsertFunction( + kAsanReportLoadN, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + AsanErrorCallbackSized[1] = checkInterfaceFunction(M.getOrInsertFunction( + kAsanReportStoreN, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); AsanHandleNoReturnFunc = checkInterfaceFunction(M.getOrInsertFunction( kAsanHandleNoReturnName, IRB.getVoidTy(), NULL)); + AsanCovFunction = checkInterfaceFunction(M.getOrInsertFunction( + kAsanCovName, IRB.getVoidTy(), NULL)); + AsanPtrCmpFunction = checkInterfaceFunction(M.getOrInsertFunction( + kAsanPtrCmp, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + AsanPtrSubFunction = checkInterfaceFunction(M.getOrInsertFunction( + kAsanPtrSub, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); // We insert an empty inline asm after __asan_report* to avoid callback merge. EmptyAsm = InlineAsm::get(FunctionType::get(IRB.getVoidTy(), false), StringRef(""), StringRef(""), /*hasSideEffects=*/true); } -void AddressSanitizer::emitShadowMapping(Module &M, IRBuilder<> &IRB) const { - // Tell the values of mapping offset and scale to the run-time if they are - // specified by command-line flags. - if (ClMappingOffsetLog >= 0) { - GlobalValue *asan_mapping_offset = - new GlobalVariable(M, IntptrTy, true, GlobalValue::LinkOnceODRLinkage, - ConstantInt::get(IntptrTy, Mapping.Offset), - kAsanMappingOffsetName); - // Read the global, otherwise it may be optimized away. - IRB.CreateLoad(asan_mapping_offset, true); - } - - if (ClMappingScale) { - GlobalValue *asan_mapping_scale = - new GlobalVariable(M, IntptrTy, true, GlobalValue::LinkOnceODRLinkage, - ConstantInt::get(IntptrTy, Mapping.Scale), - kAsanMappingScaleName); - // Read the global, otherwise it may be optimized away. - IRB.CreateLoad(asan_mapping_scale, true); - } -} - // virtual bool AddressSanitizer::doInitialization(Module &M) { // Initialize the private fields. No one has accessed them before. - TD = getAnalysisIfAvailable(); - - if (!TD) + DataLayoutPass *DLP = getAnalysisIfAvailable(); + if (!DLP) return false; - BL.reset(new BlackList(BlacklistFile)); + DL = &DLP->getDataLayout(); + + BL.reset(SpecialCaseList::createOrDie(BlacklistFile)); DynamicallyInitializedGlobals.Init(M); C = &(M.getContext()); - LongSize = TD->getPointerSizeInBits(); + LongSize = DL->getPointerSizeInBits(); IntptrTy = Type::getIntNTy(*C, LongSize); AsanCtorFunction = Function::Create( @@ -1014,7 +1159,6 @@ bool AddressSanitizer::doInitialization(Module &M) { IRB.CreateCall(AsanInitFunction); Mapping = getShadowMapping(M, LongSize); - emitShadowMapping(M, IRB); appendToGlobalCtors(M, AsanCtorFunction, kAsanCtorAndCtorPriority); return true; @@ -1036,17 +1180,82 @@ bool AddressSanitizer::maybeInsertAsanInitAtFunctionEntry(Function &F) { return false; } +void AddressSanitizer::InjectCoverageAtBlock(Function &F, BasicBlock &BB) { + BasicBlock::iterator IP = BB.getFirstInsertionPt(), BE = BB.end(); + // Skip static allocas at the top of the entry block so they don't become + // dynamic when we split the block. If we used our optimized stack layout, + // then there will only be one alloca and it will come first. + for (; IP != BE; ++IP) { + AllocaInst *AI = dyn_cast(IP); + if (!AI || !AI->isStaticAlloca()) + break; + } + + IRBuilder<> IRB(IP); + Type *Int8Ty = IRB.getInt8Ty(); + GlobalVariable *Guard = new GlobalVariable( + *F.getParent(), Int8Ty, false, GlobalValue::PrivateLinkage, + Constant::getNullValue(Int8Ty), "__asan_gen_cov_" + F.getName()); + LoadInst *Load = IRB.CreateLoad(Guard); + Load->setAtomic(Monotonic); + Load->setAlignment(1); + Value *Cmp = IRB.CreateICmpEQ(Constant::getNullValue(Int8Ty), Load); + Instruction *Ins = SplitBlockAndInsertIfThen( + Cmp, IP, false, MDBuilder(*C).createBranchWeights(1, 100000)); + IRB.SetInsertPoint(Ins); + // We pass &F to __sanitizer_cov. We could avoid this and rely on + // GET_CALLER_PC, but having the PC of the first instruction is just nice. + Instruction *Call = IRB.CreateCall(AsanCovFunction); + Call->setDebugLoc(IP->getDebugLoc()); + StoreInst *Store = IRB.CreateStore(ConstantInt::get(Int8Ty, 1), Guard); + Store->setAtomic(Monotonic); + Store->setAlignment(1); +} + +// Poor man's coverage that works with ASan. +// We create a Guard boolean variable with the same linkage +// as the function and inject this code into the entry block (-asan-coverage=1) +// or all blocks (-asan-coverage=2): +// if (*Guard) { +// __sanitizer_cov(&F); +// *Guard = 1; +// } +// The accesses to Guard are atomic. The rest of the logic is +// in __sanitizer_cov (it's fine to call it more than once). +// +// This coverage implementation provides very limited data: +// it only tells if a given function (block) was ever executed. +// No counters, no per-edge data. +// But for many use cases this is what we need and the added slowdown +// is negligible. This simple implementation will probably be obsoleted +// by the upcoming Clang-based coverage implementation. +// By having it here and now we hope to +// a) get the functionality to users earlier and +// b) collect usage statistics to help improve Clang coverage design. +bool AddressSanitizer::InjectCoverage(Function &F, + const ArrayRef AllBlocks) { + if (!ClCoverage) return false; + + if (ClCoverage == 1) { + InjectCoverageAtBlock(F, F.getEntryBlock()); + } else { + for (size_t i = 0, n = AllBlocks.size(); i < n; i++) + InjectCoverageAtBlock(F, *AllBlocks[i]); + } + return true; +} + bool AddressSanitizer::runOnFunction(Function &F) { if (BL->isIn(F)) return false; if (&F == AsanCtorFunction) return false; + if (F.getLinkage() == GlobalValue::AvailableExternallyLinkage) return false; DEBUG(dbgs() << "ASAN instrumenting:\n" << F << "\n"); initializeCallbacks(*F.getParent()); - // If needed, insert __asan_init before checking for AddressSafety attr. + // If needed, insert __asan_init before checking for SanitizeAddress attr. maybeInsertAsanInitAtFunctionEntry(F); - if (!F.getAttributes().hasAttribute(AttributeSet::FunctionIndex, - Attribute::AddressSafety)) + if (!F.hasFnAttribute(Attribute::SanitizeAddress)) return false; if (!ClDebugFunc.empty() && ClDebugFunc != F.getName()) @@ -1057,11 +1266,15 @@ bool AddressSanitizer::runOnFunction(Function &F) { SmallSet TempsToInstrument; SmallVector ToInstrument; SmallVector NoReturnCalls; + SmallVector AllBlocks; + SmallVector PointerComparisonsOrSubtracts; + int NumAllocas = 0; bool IsWrite; // Fill the set of memory operations to instrument. for (Function::iterator FI = F.begin(), FE = F.end(); FI != FE; ++FI) { + AllBlocks.push_back(FI); TempsToInstrument.clear(); int NumInsnsPerBB = 0; for (BasicBlock::iterator BI = FI->begin(), BE = FI->end(); @@ -1072,15 +1285,21 @@ bool AddressSanitizer::runOnFunction(Function &F) { if (!TempsToInstrument.insert(Addr)) continue; // We've seen this temp in the current BB. } + } else if (ClInvalidPointerPairs && + isInterestingPointerComparisonOrSubtraction(BI)) { + PointerComparisonsOrSubtracts.push_back(BI); + continue; } else if (isa(BI) && ClMemIntrin) { // ok, take it. } else { - if (CallInst *CI = dyn_cast(BI)) { + if (isa(BI)) + NumAllocas++; + CallSite CS(BI); + if (CS) { // A call inside BB. TempsToInstrument.clear(); - if (CI->doesNotReturn()) { - NoReturnCalls.push_back(CI); - } + if (CS.doesNotReturn()) + NoReturnCalls.push_back(CS.getInstruction()); } continue; } @@ -1091,6 +1310,17 @@ bool AddressSanitizer::runOnFunction(Function &F) { } } + Function *UninstrumentedDuplicate = 0; + bool LikelyToInstrument = + !NoReturnCalls.empty() || !ToInstrument.empty() || (NumAllocas > 0); + if (ClKeepUninstrumented && LikelyToInstrument) { + ValueToValueMapTy VMap; + UninstrumentedDuplicate = CloneFunction(&F, VMap, false); + UninstrumentedDuplicate->removeFnAttr(Attribute::SanitizeAddress); + UninstrumentedDuplicate->setName("NOASAN_" + F.getName()); + F.getParent()->getFunctionList().push_back(UninstrumentedDuplicate); + } + // Instrument. int NumInstrumented = 0; for (size_t i = 0, n = ToInstrument.size(); i != n; i++) { @@ -1115,35 +1345,34 @@ bool AddressSanitizer::runOnFunction(Function &F) { IRBuilder<> IRB(CI); IRB.CreateCall(AsanHandleNoReturnFunc); } - DEBUG(dbgs() << "ASAN done instrumenting:\n" << F << "\n"); - return NumInstrumented > 0 || ChangedStack || !NoReturnCalls.empty(); -} + for (size_t i = 0, n = PointerComparisonsOrSubtracts.size(); i != n; i++) { + instrumentPointerComparisonOrSubtraction(PointerComparisonsOrSubtracts[i]); + NumInstrumented++; + } -static uint64_t ValueForPoison(uint64_t PoisonByte, size_t ShadowRedzoneSize) { - if (ShadowRedzoneSize == 1) return PoisonByte; - if (ShadowRedzoneSize == 2) return (PoisonByte << 8) + PoisonByte; - if (ShadowRedzoneSize == 4) - return (PoisonByte << 24) + (PoisonByte << 16) + - (PoisonByte << 8) + (PoisonByte); - llvm_unreachable("ShadowRedzoneSize is either 1, 2 or 4"); -} + bool res = NumInstrumented > 0 || ChangedStack || !NoReturnCalls.empty(); + + if (InjectCoverage(F, AllBlocks)) + res = true; + + DEBUG(dbgs() << "ASAN done instrumenting: " << res << " " << F << "\n"); -static void PoisonShadowPartialRightRedzone(uint8_t *Shadow, - size_t Size, - size_t RZSize, - size_t ShadowGranularity, - uint8_t Magic) { - for (size_t i = 0; i < RZSize; - i+= ShadowGranularity, Shadow++) { - if (i + ShadowGranularity <= Size) { - *Shadow = 0; // fully addressable - } else if (i >= Size) { - *Shadow = Magic; // unaddressable + if (ClKeepUninstrumented) { + if (!res) { + // No instrumentation is done, no need for the duplicate. + if (UninstrumentedDuplicate) + UninstrumentedDuplicate->eraseFromParent(); } else { - *Shadow = Size - i; // first Size-i bytes are addressable + // The function was instrumented. We must have the duplicate. + assert(UninstrumentedDuplicate); + UninstrumentedDuplicate->setSection("NOASAN"); + assert(!F.hasSection()); + F.setSection("ASAN"); } } + + return res; } // Workaround for bug 11395: we don't want to instrument stack in functions @@ -1160,175 +1389,236 @@ bool AddressSanitizer::LooksLikeCodeInBug11395(Instruction *I) { void FunctionStackPoisoner::initializeCallbacks(Module &M) { IRBuilder<> IRB(*C); - AsanStackMallocFunc = checkInterfaceFunction(M.getOrInsertFunction( - kAsanStackMallocName, IntptrTy, IntptrTy, IntptrTy, NULL)); - AsanStackFreeFunc = checkInterfaceFunction(M.getOrInsertFunction( - kAsanStackFreeName, IRB.getVoidTy(), - IntptrTy, IntptrTy, IntptrTy, NULL)); + for (int i = 0; i <= kMaxAsanStackMallocSizeClass; i++) { + std::string Suffix = itostr(i); + AsanStackMallocFunc[i] = checkInterfaceFunction( + M.getOrInsertFunction(kAsanStackMallocNameTemplate + Suffix, IntptrTy, + IntptrTy, IntptrTy, NULL)); + AsanStackFreeFunc[i] = checkInterfaceFunction(M.getOrInsertFunction( + kAsanStackFreeNameTemplate + Suffix, IRB.getVoidTy(), IntptrTy, + IntptrTy, IntptrTy, NULL)); + } AsanPoisonStackMemoryFunc = checkInterfaceFunction(M.getOrInsertFunction( kAsanPoisonStackMemoryName, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); AsanUnpoisonStackMemoryFunc = checkInterfaceFunction(M.getOrInsertFunction( kAsanUnpoisonStackMemoryName, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); } -void FunctionStackPoisoner::poisonRedZones( - const ArrayRef &AllocaVec, IRBuilder<> IRB, Value *ShadowBase, - bool DoPoison) { - size_t ShadowRZSize = RedzoneSize() >> Mapping.Scale; - assert(ShadowRZSize >= 1 && ShadowRZSize <= 4); - Type *RZTy = Type::getIntNTy(*C, ShadowRZSize * 8); - Type *RZPtrTy = PointerType::get(RZTy, 0); - - Value *PoisonLeft = ConstantInt::get(RZTy, - ValueForPoison(DoPoison ? kAsanStackLeftRedzoneMagic : 0LL, ShadowRZSize)); - Value *PoisonMid = ConstantInt::get(RZTy, - ValueForPoison(DoPoison ? kAsanStackMidRedzoneMagic : 0LL, ShadowRZSize)); - Value *PoisonRight = ConstantInt::get(RZTy, - ValueForPoison(DoPoison ? kAsanStackRightRedzoneMagic : 0LL, ShadowRZSize)); - - // poison the first red zone. - IRB.CreateStore(PoisonLeft, IRB.CreateIntToPtr(ShadowBase, RZPtrTy)); - - // poison all other red zones. - uint64_t Pos = RedzoneSize(); - for (size_t i = 0, n = AllocaVec.size(); i < n; i++) { - AllocaInst *AI = AllocaVec[i]; - uint64_t SizeInBytes = getAllocaSizeInBytes(AI); - uint64_t AlignedSize = getAlignedAllocaSize(AI); - assert(AlignedSize - SizeInBytes < RedzoneSize()); - Value *Ptr = NULL; - - Pos += AlignedSize; - - assert(ShadowBase->getType() == IntptrTy); - if (SizeInBytes < AlignedSize) { - // Poison the partial redzone at right - Ptr = IRB.CreateAdd( - ShadowBase, ConstantInt::get(IntptrTy, - (Pos >> Mapping.Scale) - ShadowRZSize)); - size_t AddressableBytes = RedzoneSize() - (AlignedSize - SizeInBytes); - uint32_t Poison = 0; - if (DoPoison) { - PoisonShadowPartialRightRedzone((uint8_t*)&Poison, AddressableBytes, - RedzoneSize(), - 1ULL << Mapping.Scale, - kAsanStackPartialRedzoneMagic); +void +FunctionStackPoisoner::poisonRedZones(const ArrayRef ShadowBytes, + IRBuilder<> &IRB, Value *ShadowBase, + bool DoPoison) { + size_t n = ShadowBytes.size(); + size_t i = 0; + // We need to (un)poison n bytes of stack shadow. Poison as many as we can + // using 64-bit stores (if we are on 64-bit arch), then poison the rest + // with 32-bit stores, then with 16-byte stores, then with 8-byte stores. + for (size_t LargeStoreSizeInBytes = ASan.LongSize / 8; + LargeStoreSizeInBytes != 0; LargeStoreSizeInBytes /= 2) { + for (; i + LargeStoreSizeInBytes - 1 < n; i += LargeStoreSizeInBytes) { + uint64_t Val = 0; + for (size_t j = 0; j < LargeStoreSizeInBytes; j++) { + if (ASan.DL->isLittleEndian()) + Val |= (uint64_t)ShadowBytes[i + j] << (8 * j); + else + Val = (Val << 8) | ShadowBytes[i + j]; } - Value *PartialPoison = ConstantInt::get(RZTy, Poison); - IRB.CreateStore(PartialPoison, IRB.CreateIntToPtr(Ptr, RZPtrTy)); + if (!Val) continue; + Value *Ptr = IRB.CreateAdd(ShadowBase, ConstantInt::get(IntptrTy, i)); + Type *StoreTy = Type::getIntNTy(*C, LargeStoreSizeInBytes * 8); + Value *Poison = ConstantInt::get(StoreTy, DoPoison ? Val : 0); + IRB.CreateStore(Poison, IRB.CreateIntToPtr(Ptr, StoreTy->getPointerTo())); } + } +} - // Poison the full redzone at right. - Ptr = IRB.CreateAdd(ShadowBase, - ConstantInt::get(IntptrTy, Pos >> Mapping.Scale)); - bool LastAlloca = (i == AllocaVec.size() - 1); - Value *Poison = LastAlloca ? PoisonRight : PoisonMid; - IRB.CreateStore(Poison, IRB.CreateIntToPtr(Ptr, RZPtrTy)); +// Fake stack allocator (asan_fake_stack.h) has 11 size classes +// for every power of 2 from kMinStackMallocSize to kMaxAsanStackMallocSizeClass +static int StackMallocSizeClass(uint64_t LocalStackSize) { + assert(LocalStackSize <= kMaxStackMallocSize); + uint64_t MaxSize = kMinStackMallocSize; + for (int i = 0; ; i++, MaxSize *= 2) + if (LocalStackSize <= MaxSize) + return i; + llvm_unreachable("impossible LocalStackSize"); +} - Pos += RedzoneSize(); +// Set Size bytes starting from ShadowBase to kAsanStackAfterReturnMagic. +// We can not use MemSet intrinsic because it may end up calling the actual +// memset. Size is a multiple of 8. +// Currently this generates 8-byte stores on x86_64; it may be better to +// generate wider stores. +void FunctionStackPoisoner::SetShadowToStackAfterReturnInlined( + IRBuilder<> &IRB, Value *ShadowBase, int Size) { + assert(!(Size % 8)); + assert(kAsanStackAfterReturnMagic == 0xf5); + for (int i = 0; i < Size; i += 8) { + Value *p = IRB.CreateAdd(ShadowBase, ConstantInt::get(IntptrTy, i)); + IRB.CreateStore(ConstantInt::get(IRB.getInt64Ty(), 0xf5f5f5f5f5f5f5f5ULL), + IRB.CreateIntToPtr(p, IRB.getInt64Ty()->getPointerTo())); } } void FunctionStackPoisoner::poisonStack() { - uint64_t LocalStackSize = TotalStackSize + - (AllocaVec.size() + 1) * RedzoneSize(); - - bool DoStackMalloc = ASan.CheckUseAfterReturn - && LocalStackSize <= kMaxStackMallocSize; + int StackMallocIdx = -1; assert(AllocaVec.size() > 0); Instruction *InsBefore = AllocaVec[0]; IRBuilder<> IRB(InsBefore); + SmallVector SVD; + SVD.reserve(AllocaVec.size()); + for (size_t i = 0, n = AllocaVec.size(); i < n; i++) { + AllocaInst *AI = AllocaVec[i]; + ASanStackVariableDescription D = { AI->getName().data(), + getAllocaSizeInBytes(AI), + AI->getAlignment(), AI, 0}; + SVD.push_back(D); + } + // Minimal header size (left redzone) is 4 pointers, + // i.e. 32 bytes on 64-bit platforms and 16 bytes in 32-bit platforms. + size_t MinHeaderSize = ASan.LongSize / 2; + ASanStackFrameLayout L; + ComputeASanStackFrameLayout(SVD, 1UL << Mapping.Scale, MinHeaderSize, &L); + DEBUG(dbgs() << L.DescriptionString << " --- " << L.FrameSize << "\n"); + uint64_t LocalStackSize = L.FrameSize; + bool DoStackMalloc = + ASan.CheckUseAfterReturn && LocalStackSize <= kMaxStackMallocSize; Type *ByteArrayTy = ArrayType::get(IRB.getInt8Ty(), LocalStackSize); AllocaInst *MyAlloca = new AllocaInst(ByteArrayTy, "MyAlloca", InsBefore); - if (ClRealignStack && StackAlignment < RedzoneSize()) - StackAlignment = RedzoneSize(); - MyAlloca->setAlignment(StackAlignment); + assert((ClRealignStack & (ClRealignStack - 1)) == 0); + size_t FrameAlignment = std::max(L.FrameAlignment, (size_t)ClRealignStack); + MyAlloca->setAlignment(FrameAlignment); assert(MyAlloca->isStaticAlloca()); Value *OrigStackBase = IRB.CreatePointerCast(MyAlloca, IntptrTy); Value *LocalStackBase = OrigStackBase; if (DoStackMalloc) { - LocalStackBase = IRB.CreateCall2(AsanStackMallocFunc, + // LocalStackBase = OrigStackBase + // if (__asan_option_detect_stack_use_after_return) + // LocalStackBase = __asan_stack_malloc_N(LocalStackBase, OrigStackBase); + StackMallocIdx = StackMallocSizeClass(LocalStackSize); + assert(StackMallocIdx <= kMaxAsanStackMallocSizeClass); + Constant *OptionDetectUAR = F.getParent()->getOrInsertGlobal( + kAsanOptionDetectUAR, IRB.getInt32Ty()); + Value *Cmp = IRB.CreateICmpNE(IRB.CreateLoad(OptionDetectUAR), + Constant::getNullValue(IRB.getInt32Ty())); + Instruction *Term = SplitBlockAndInsertIfThen(Cmp, InsBefore, false); + BasicBlock *CmpBlock = cast(Cmp)->getParent(); + IRBuilder<> IRBIf(Term); + LocalStackBase = IRBIf.CreateCall2( + AsanStackMallocFunc[StackMallocIdx], ConstantInt::get(IntptrTy, LocalStackSize), OrigStackBase); + BasicBlock *SetBlock = cast(LocalStackBase)->getParent(); + IRB.SetInsertPoint(InsBefore); + PHINode *Phi = IRB.CreatePHI(IntptrTy, 2); + Phi->addIncoming(OrigStackBase, CmpBlock); + Phi->addIncoming(LocalStackBase, SetBlock); + LocalStackBase = Phi; } - // This string will be parsed by the run-time (DescribeStackAddress). - SmallString<2048> StackDescriptionStorage; - raw_svector_ostream StackDescription(StackDescriptionStorage); - StackDescription << F.getName() << " " << AllocaVec.size() << " "; - // Insert poison calls for lifetime intrinsics for alloca. bool HavePoisonedAllocas = false; for (size_t i = 0, n = AllocaPoisonCallVec.size(); i < n; i++) { const AllocaPoisonCall &APC = AllocaPoisonCallVec[i]; - IntrinsicInst *II = APC.InsBefore; - AllocaInst *AI = findAllocaForValue(II->getArgOperand(1)); - assert(AI); - IRBuilder<> IRB(II); - poisonAlloca(AI, APC.Size, IRB, APC.DoPoison); + assert(APC.InsBefore); + assert(APC.AI); + IRBuilder<> IRB(APC.InsBefore); + poisonAlloca(APC.AI, APC.Size, IRB, APC.DoPoison); HavePoisonedAllocas |= APC.DoPoison; } - uint64_t Pos = RedzoneSize(); // Replace Alloca instructions with base+offset. - for (size_t i = 0, n = AllocaVec.size(); i < n; i++) { - AllocaInst *AI = AllocaVec[i]; - uint64_t SizeInBytes = getAllocaSizeInBytes(AI); - StringRef Name = AI->getName(); - StackDescription << Pos << " " << SizeInBytes << " " - << Name.size() << " " << Name << " "; - uint64_t AlignedSize = getAlignedAllocaSize(AI); - assert((AlignedSize % RedzoneSize()) == 0); + for (size_t i = 0, n = SVD.size(); i < n; i++) { + AllocaInst *AI = SVD[i].AI; Value *NewAllocaPtr = IRB.CreateIntToPtr( - IRB.CreateAdd(LocalStackBase, ConstantInt::get(IntptrTy, Pos)), - AI->getType()); + IRB.CreateAdd(LocalStackBase, + ConstantInt::get(IntptrTy, SVD[i].Offset)), + AI->getType()); replaceDbgDeclareForAlloca(AI, NewAllocaPtr, DIB); AI->replaceAllUsesWith(NewAllocaPtr); - Pos += AlignedSize + RedzoneSize(); } - assert(Pos == LocalStackSize); - // Write the Magic value and the frame description constant to the redzone. + // The left-most redzone has enough space for at least 4 pointers. + // Write the Magic value to redzone[0]. Value *BasePlus0 = IRB.CreateIntToPtr(LocalStackBase, IntptrPtrTy); IRB.CreateStore(ConstantInt::get(IntptrTy, kCurrentStackFrameMagic), BasePlus0); - Value *BasePlus1 = IRB.CreateAdd(LocalStackBase, - ConstantInt::get(IntptrTy, - ASan.LongSize/8)); - BasePlus1 = IRB.CreateIntToPtr(BasePlus1, IntptrPtrTy); + // Write the frame description constant to redzone[1]. + Value *BasePlus1 = IRB.CreateIntToPtr( + IRB.CreateAdd(LocalStackBase, ConstantInt::get(IntptrTy, ASan.LongSize/8)), + IntptrPtrTy); GlobalVariable *StackDescriptionGlobal = - createPrivateGlobalForString(*F.getParent(), StackDescription.str()); + createPrivateGlobalForString(*F.getParent(), L.DescriptionString, + /*AllowMerging*/true); Value *Description = IRB.CreatePointerCast(StackDescriptionGlobal, IntptrTy); IRB.CreateStore(Description, BasePlus1); + // Write the PC to redzone[2]. + Value *BasePlus2 = IRB.CreateIntToPtr( + IRB.CreateAdd(LocalStackBase, ConstantInt::get(IntptrTy, + 2 * ASan.LongSize/8)), + IntptrPtrTy); + IRB.CreateStore(IRB.CreatePointerCast(&F, IntptrTy), BasePlus2); // Poison the stack redzones at the entry. Value *ShadowBase = ASan.memToShadow(LocalStackBase, IRB); - poisonRedZones(AllocaVec, IRB, ShadowBase, true); + poisonRedZones(L.ShadowBytes, IRB, ShadowBase, true); - // Unpoison the stack before all ret instructions. + // (Un)poison the stack before all ret instructions. for (size_t i = 0, n = RetVec.size(); i < n; i++) { Instruction *Ret = RetVec[i]; IRBuilder<> IRBRet(Ret); // Mark the current frame as retired. IRBRet.CreateStore(ConstantInt::get(IntptrTy, kRetiredStackFrameMagic), BasePlus0); - // Unpoison the stack. - poisonRedZones(AllocaVec, IRBRet, ShadowBase, false); if (DoStackMalloc) { - // In use-after-return mode, mark the whole stack frame unaddressable. - IRBRet.CreateCall3(AsanStackFreeFunc, LocalStackBase, - ConstantInt::get(IntptrTy, LocalStackSize), - OrigStackBase); + assert(StackMallocIdx >= 0); + // if LocalStackBase != OrigStackBase: + // // In use-after-return mode, poison the whole stack frame. + // if StackMallocIdx <= 4 + // // For small sizes inline the whole thing: + // memset(ShadowBase, kAsanStackAfterReturnMagic, ShadowSize); + // **SavedFlagPtr(LocalStackBase) = 0 + // else + // __asan_stack_free_N(LocalStackBase, OrigStackBase) + // else + // + Value *Cmp = IRBRet.CreateICmpNE(LocalStackBase, OrigStackBase); + TerminatorInst *ThenTerm, *ElseTerm; + SplitBlockAndInsertIfThenElse(Cmp, Ret, &ThenTerm, &ElseTerm); + + IRBuilder<> IRBPoison(ThenTerm); + if (StackMallocIdx <= 4) { + int ClassSize = kMinStackMallocSize << StackMallocIdx; + SetShadowToStackAfterReturnInlined(IRBPoison, ShadowBase, + ClassSize >> Mapping.Scale); + Value *SavedFlagPtrPtr = IRBPoison.CreateAdd( + LocalStackBase, + ConstantInt::get(IntptrTy, ClassSize - ASan.LongSize / 8)); + Value *SavedFlagPtr = IRBPoison.CreateLoad( + IRBPoison.CreateIntToPtr(SavedFlagPtrPtr, IntptrPtrTy)); + IRBPoison.CreateStore( + Constant::getNullValue(IRBPoison.getInt8Ty()), + IRBPoison.CreateIntToPtr(SavedFlagPtr, IRBPoison.getInt8PtrTy())); + } else { + // For larger frames call __asan_stack_free_*. + IRBPoison.CreateCall3(AsanStackFreeFunc[StackMallocIdx], LocalStackBase, + ConstantInt::get(IntptrTy, LocalStackSize), + OrigStackBase); + } + + IRBuilder<> IRBElse(ElseTerm); + poisonRedZones(L.ShadowBytes, IRBElse, ShadowBase, false); } else if (HavePoisonedAllocas) { // If we poisoned some allocas in llvm.lifetime analysis, // unpoison whole stack frame now. assert(LocalStackBase == OrigStackBase); poisonAlloca(LocalStackBase, LocalStackSize, IRBRet, false); + } else { + poisonRedZones(L.ShadowBytes, IRBRet, ShadowBase, false); } } @@ -1338,7 +1628,7 @@ void FunctionStackPoisoner::poisonStack() { } void FunctionStackPoisoner::poisonAlloca(Value *V, uint64_t Size, - IRBuilder<> IRB, bool DoPoison) { + IRBuilder<> &IRB, bool DoPoison) { // For now just insert the call to ASan runtime. Value *AddrArg = IRB.CreatePointerCast(V, IntptrTy); Value *SizeArg = ConstantInt::get(IntptrTy, Size);