X-Git-Url: http://plrg.eecs.uci.edu/git/?p=oota-llvm.git;a=blobdiff_plain;f=lib%2FTransforms%2FInstrumentation%2FAddressSanitizer.cpp;h=3ff0a616f4a688ccce6adfba162ec14eb1d4b44e;hp=28c5622863bd9c2e4256cc8f6894ba417ad1c889;hb=891198b358c79d9464bd40d94f5ba6249080741a;hpb=405515d55f470d04ef75f653b7f1994329c9066b diff --git a/lib/Transforms/Instrumentation/AddressSanitizer.cpp b/lib/Transforms/Instrumentation/AddressSanitizer.cpp index 28c5622863b..3ff0a616f4a 100644 --- a/lib/Transforms/Instrumentation/AddressSanitizer.cpp +++ b/lib/Transforms/Instrumentation/AddressSanitizer.cpp @@ -13,81 +13,93 @@ // //===----------------------------------------------------------------------===// -#define DEBUG_TYPE "asan" - #include "llvm/Transforms/Instrumentation.h" #include "llvm/ADT/ArrayRef.h" #include "llvm/ADT/DenseMap.h" +#include "llvm/ADT/DenseSet.h" #include "llvm/ADT/DepthFirstIterator.h" -#include "llvm/ADT/OwningPtr.h" #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" +#include "llvm/IR/CallSite.h" +#include "llvm/IR/DIBuilder.h" #include "llvm/IR/DataLayout.h" #include "llvm/IR/Function.h" #include "llvm/IR/IRBuilder.h" #include "llvm/IR/InlineAsm.h" +#include "llvm/IR/InstVisitor.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/Endian.h" -#include "llvm/Support/raw_ostream.h" -#include "llvm/Support/system_error.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 +#include using namespace llvm; +#define DEBUG_TYPE "asan" + static const uint64_t kDefaultShadowScale = 3; static const uint64_t kDefaultShadowOffset32 = 1ULL << 29; +static const uint64_t kIOSShadowOffset32 = 1ULL << 30; static const uint64_t kDefaultShadowOffset64 = 1ULL << 44; -static const uint64_t kDefaultShort64bitShadowOffset = 0x7FFF8000; // < 2G. +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 *kAsanReportLoadN = "__asan_report_load_n"; -static const char *kAsanReportStoreN = "__asan_report_store_n"; -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_v3"; -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 kAsanCtorAndDtorPriority = 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_v4"; +static const char *const kAsanCovModuleInitName = "__sanitizer_cov_module_init"; +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; @@ -116,21 +128,36 @@ static cl::opt ClMaxInsnsToInstrumentPerBB("asan-max-ins-per-bb", // This flag may need to be replaced with -f[no]asan-stack. static cl::opt ClStack("asan-stack", cl::desc("Handle stack memory"), cl::Hidden, cl::init(true)); -// This flag may need to be replaced with -f[no]asan-use-after-return. static cl::opt ClUseAfterReturn("asan-use-after-return", - cl::desc("Check return-after-free"), cl::Hidden, cl::init(false)); + cl::desc("Check return-after-free"), cl::Hidden, cl::init(true)); // 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 ClCoverageBlockThreshold("asan-coverage-block-threshold", + cl::desc("Add coverage instrumentation only to the entry block if there " + "are more than this number of blocks."), + cl::Hidden, cl::init(1500)); 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 ClBlacklistFile("asan-blacklist", - cl::desc("File containing the list of objects to ignore " - "during instrumentation"), cl::Hidden); + cl::desc("Handle C++ initializer order"), 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 ClInstrumentationWithCallsThreshold( + "asan-instrumentation-with-call-threshold", + cl::desc("If the function being instrumented contains more than " + "this number of memory accesses, use callbacks instead of " + "inline checks (-1 means never use callbacks)."), + cl::Hidden, cl::init(7000)); +static cl::opt ClMemoryAccessCallbackPrefix( + "asan-memory-access-callback-prefix", + cl::desc("Prefix for memory access callbacks"), cl::Hidden, + cl::init("__asan_")); // This is an experimental feature that will allow to choose between // instrumented and non-instrumented code at link-time. @@ -150,11 +177,6 @@ static cl::opt ClKeepUninstrumented("asan-keep-uninstrumented-functions", // 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)); -static cl::opt ClShort64BitOffset("asan-short-64bit-mapping-offset", - cl::desc("Use short immediate constant as the mapping offset for 64bit"), - cl::Hidden, cl::init(true)); // Optimization flags. Not user visible, used mostly for testing // and benchmarking the tool. @@ -182,30 +204,87 @@ 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 { +/// Frontend-provided metadata for source location. +struct LocationMetadata { + StringRef Filename; + int LineNo; + int ColumnNo; + + LocationMetadata() : Filename(), LineNo(0), ColumnNo(0) {} + + bool empty() const { return Filename.empty(); } + + void parse(MDNode *MDN) { + assert(MDN->getNumOperands() == 3); + MDString *MDFilename = cast(MDN->getOperand(0)); + Filename = MDFilename->getString(); + LineNo = cast(MDN->getOperand(1))->getLimitedValue(); + ColumnNo = cast(MDN->getOperand(2))->getLimitedValue(); + } +}; + +/// Frontend-provided metadata for global variables. +class GlobalsMetadata { public: - void Init(Module& M) { - // Clang generates metadata identifying all dynamically initialized globals. - NamedMDNode *DynamicGlobals = - M.getNamedMetadata("llvm.asan.dynamically_initialized_globals"); - if (!DynamicGlobals) + struct Entry { + Entry() + : SourceLoc(), Name(), IsDynInit(false), + IsBlacklisted(false) {} + LocationMetadata SourceLoc; + StringRef Name; + bool IsDynInit; + bool IsBlacklisted; + }; + + GlobalsMetadata() : inited_(false) {} + + void init(Module& M) { + assert(!inited_); + inited_ = true; + NamedMDNode *Globals = M.getNamedMetadata("llvm.asan.globals"); + if (!Globals) return; - for (int i = 0, n = DynamicGlobals->getNumOperands(); i < n; ++i) { - MDNode *MDN = DynamicGlobals->getOperand(i); - assert(MDN->getNumOperands() == 1); - Value *VG = MDN->getOperand(0); - // The optimizer may optimize away a global entirely, in which case we - // cannot instrument access to it. - if (!VG) + for (auto MDN : Globals->operands()) { + // Metadata node contains the global and the fields of "Entry". + assert(MDN->getNumOperands() == 5); + Value *V = MDN->getOperand(0); + // The optimizer may optimize away a global entirely. + if (!V) continue; - DynInitGlobals.insert(cast(VG)); + GlobalVariable *GV = cast(V); + // We can already have an entry for GV if it was merged with another + // global. + Entry &E = Entries[GV]; + if (Value *Loc = MDN->getOperand(1)) + E.SourceLoc.parse(cast(Loc)); + if (Value *Name = MDN->getOperand(2)) { + MDString *MDName = cast(Name); + E.Name = MDName->getString(); + } + ConstantInt *IsDynInit = cast(MDN->getOperand(3)); + E.IsDynInit |= IsDynInit->isOne(); + ConstantInt *IsBlacklisted = cast(MDN->getOperand(4)); + E.IsBlacklisted |= IsBlacklisted->isOne(); } } - bool Contains(GlobalVariable *G) { return DynInitGlobals.count(G) != 0; } + + /// Returns metadata entry for a given global. + Entry get(GlobalVariable *G) const { + auto Pos = Entries.find(G); + return (Pos != Entries.end()) ? Pos->second : Entry(); + } + private: - SmallSet DynInitGlobals; + bool inited_; + DenseMap Entries; }; /// This struct defines the shadow mapping using the rule: @@ -216,34 +295,40 @@ struct ShadowMapping { bool OrShadowOffset; }; -static ShadowMapping getShadowMapping(const Module &M, int LongSize, - bool ZeroBaseShadow) { +static ShadowMapping getShadowMapping(const Module &M, int LongSize) { llvm::Triple TargetTriple(M.getTargetTriple()); bool IsAndroid = TargetTriple.getEnvironment() == llvm::Triple::Android; - bool IsMacOSX = TargetTriple.getOS() == llvm::Triple::MacOSX; - bool IsPPC64 = TargetTriple.getArch() == llvm::Triple::ppc64; + bool IsIOS = TargetTriple.getOS() == llvm::Triple::IOS; + 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; - // OR-ing shadow offset if more efficient (at least on x86), - // but on ppc64 we have to use add since the shadow offset is not neccesary - // 1/8-th of the address space. - Mapping.OrShadowOffset = !IsPPC64 && !ClShort64BitOffset; - - Mapping.Offset = (IsAndroid || ZeroBaseShadow) ? 0 : - (LongSize == 32 ? - (IsMIPS32 ? kMIPS32_ShadowOffset32 : kDefaultShadowOffset32) : - IsPPC64 ? kPPC64_ShadowOffset64 : kDefaultShadowOffset64); - if (!ZeroBaseShadow && ClShort64BitOffset && IsX86_64 && !IsMacOSX) { - assert(LongSize == 64); - Mapping.Offset = kDefaultShort64bitShadowOffset; - } - if (!ZeroBaseShadow && 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 if (IsIOS) + Mapping.Offset = kIOSShadowOffset32; + 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; @@ -251,6 +336,11 @@ 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; } @@ -262,112 +352,88 @@ static size_t RedzoneSizeForScale(int MappingScale) { /// AddressSanitizer: instrument the code in module to find memory bugs. struct AddressSanitizer : public FunctionPass { - AddressSanitizer(bool CheckInitOrder = true, - bool CheckUseAfterReturn = false, - bool CheckLifetime = false, - StringRef BlacklistFile = StringRef(), - bool ZeroBaseShadow = false) - : FunctionPass(ID), - CheckInitOrder(CheckInitOrder || ClInitializers), - CheckUseAfterReturn(CheckUseAfterReturn || ClUseAfterReturn), - CheckLifetime(CheckLifetime || ClCheckLifetime), - BlacklistFile(BlacklistFile.empty() ? ClBlacklistFile - : BlacklistFile), - ZeroBaseShadow(ZeroBaseShadow) {} - virtual const char *getPassName() const { + AddressSanitizer() : FunctionPass(ID) {} + const char *getPassName() const override { return "AddressSanitizerFunctionPass"; } - void instrumentMop(Instruction *I); + void instrumentMop(Instruction *I, bool UseCalls); + void instrumentPointerComparisonOrSubtraction(Instruction *I); void instrumentAddress(Instruction *OrigIns, Instruction *InsertBefore, Value *Addr, uint32_t TypeSize, bool IsWrite, - Value *SizeArgument); + Value *SizeArgument, bool UseCalls); Value *createSlowPathCmp(IRBuilder<> &IRB, Value *AddrLong, Value *ShadowValue, uint32_t TypeSize); Instruction *generateCrashCode(Instruction *InsertBefore, Value *Addr, bool IsWrite, size_t AccessSizeIndex, Value *SizeArgument); - bool instrumentMemIntrinsic(MemIntrinsic *MI); - void instrumentMemIntrinsicParam(Instruction *OrigIns, Value *Addr, - Value *Size, - Instruction *InsertBefore, bool IsWrite); + void instrumentMemIntrinsic(MemIntrinsic *MI); Value *memToShadow(Value *Shadow, IRBuilder<> &IRB); - bool runOnFunction(Function &F); + bool runOnFunction(Function &F) override; bool maybeInsertAsanInitAtFunctionEntry(Function &F); - void emitShadowMapping(Module &M, IRBuilder<> &IRB) const; - virtual bool doInitialization(Module &M); + bool doInitialization(Module &M) override; static char ID; // Pass identification, replacement for typeid private: void initializeCallbacks(Module &M); - bool ShouldInstrumentGlobal(GlobalVariable *G); bool LooksLikeCodeInBug11395(Instruction *I); - void FindDynamicInitializers(Module &M); - - bool CheckInitOrder; - bool CheckUseAfterReturn; - bool CheckLifetime; - SmallString<64> BlacklistFile; - bool ZeroBaseShadow; + bool GlobalIsLinkerInitialized(GlobalVariable *G); + bool InjectCoverage(Function &F, ArrayRef AllBlocks); + void InjectCoverageAtBlock(Function &F, BasicBlock &BB); LLVMContext *C; - DataLayout *TD; + const DataLayout *DL; int LongSize; Type *IntptrTy; ShadowMapping Mapping; Function *AsanCtorFunction; Function *AsanInitFunction; Function *AsanHandleNoReturnFunc; - OwningPtr BL; + Function *AsanCovFunction; + Function *AsanPtrCmpFunction, *AsanPtrSubFunction; // This array is indexed by AccessIsWrite and log2(AccessSize). Function *AsanErrorCallback[2][kNumberOfAccessSizes]; + Function *AsanMemoryAccessCallback[2][kNumberOfAccessSizes]; // This array is indexed by AccessIsWrite. - Function *AsanErrorCallbackSized[2]; + Function *AsanErrorCallbackSized[2], + *AsanMemoryAccessCallbackSized[2]; + Function *AsanMemmove, *AsanMemcpy, *AsanMemset; InlineAsm *EmptyAsm; - SetOfDynamicallyInitializedGlobals DynamicallyInitializedGlobals; + GlobalsMetadata GlobalsMD; friend struct FunctionStackPoisoner; }; class AddressSanitizerModule : public ModulePass { public: - AddressSanitizerModule(bool CheckInitOrder = true, - StringRef BlacklistFile = StringRef(), - bool ZeroBaseShadow = false) - : ModulePass(ID), - CheckInitOrder(CheckInitOrder || ClInitializers), - BlacklistFile(BlacklistFile.empty() ? ClBlacklistFile - : BlacklistFile), - ZeroBaseShadow(ZeroBaseShadow) {} - bool runOnModule(Module &M); + AddressSanitizerModule() : ModulePass(ID) {} + bool runOnModule(Module &M) override; static char ID; // Pass identification, replacement for typeid - virtual const char *getPassName() const { + const char *getPassName() const override { return "AddressSanitizerModule"; } private: void initializeCallbacks(Module &M); + bool InstrumentGlobals(IRBuilder<> &IRB, Module &M); bool ShouldInstrumentGlobal(GlobalVariable *G); + void poisonOneInitializer(Function &GlobalInit, GlobalValue *ModuleName); void createInitializerPoisonCalls(Module &M, GlobalValue *ModuleName); - size_t RedzoneSize() const { + size_t MinRedzoneSizeForGlobal() const { return RedzoneSizeForScale(Mapping.Scale); } - bool CheckInitOrder; - SmallString<64> BlacklistFile; - bool ZeroBaseShadow; - - OwningPtr BL; - SetOfDynamicallyInitializedGlobals DynamicallyInitializedGlobals; + GlobalsMetadata GlobalsMD; Type *IntptrTy; LLVMContext *C; - DataLayout *TD; + const DataLayout *DL; ShadowMapping Mapping; Function *AsanPoisonGlobals; Function *AsanUnpoisonGlobals; Function *AsanRegisterGlobals; Function *AsanUnregisterGlobals; + Function *AsanCovModuleInit; }; // Stack poisoning does not play well with exception handling. @@ -390,15 +456,16 @@ 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; }; @@ -412,16 +479,14 @@ 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; // Collect alloca, ret, lifetime instructions etc. - for (df_iterator DI = df_begin(&F.getEntryBlock()), - DE = df_end(&F.getEntryBlock()); DI != DE; ++DI) { - BasicBlock *BB = *DI; + for (BasicBlock *BB : depth_first(&F.getEntryBlock())) visit(*BB); - } + if (AllocaVec.empty()) return false; initializeCallbacks(*F.getParent()); @@ -451,14 +516,12 @@ 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 /// errors. void visitIntrinsicInst(IntrinsicInst &II) { - if (!ASan.CheckLifetime) return; + if (!ClCheckLifetime) return; Intrinsic::ID ID = II.getIntrinsicID(); if (ID != Intrinsic::lifetime_start && ID != Intrinsic::lifetime_end) @@ -477,7 +540,7 @@ 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); } @@ -485,34 +548,26 @@ 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.getAlignment() <= RedzoneSize() && - 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(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 @@ -521,21 +576,16 @@ char AddressSanitizer::ID = 0; INITIALIZE_PASS(AddressSanitizer, "asan", "AddressSanitizer: detects use-after-free and out-of-bounds bugs.", false, false) -FunctionPass *llvm::createAddressSanitizerFunctionPass( - bool CheckInitOrder, bool CheckUseAfterReturn, bool CheckLifetime, - StringRef BlacklistFile, bool ZeroBaseShadow) { - return new AddressSanitizer(CheckInitOrder, CheckUseAfterReturn, - CheckLifetime, BlacklistFile, ZeroBaseShadow); +FunctionPass *llvm::createAddressSanitizerFunctionPass() { + return new AddressSanitizer(); } char AddressSanitizerModule::ID = 0; INITIALIZE_PASS(AddressSanitizerModule, "asan-module", "AddressSanitizer: detects use-after-free and out-of-bounds bugs." "ModulePass", false, false) -ModulePass *llvm::createAddressSanitizerModulePass( - bool CheckInitOrder, StringRef BlacklistFile, bool ZeroBaseShadow) { - return new AddressSanitizerModule(CheckInitOrder, BlacklistFile, - ZeroBaseShadow); +ModulePass *llvm::createAddressSanitizerModulePass() { + return new AddressSanitizerModule(); } static size_t TypeSizeToSizeIndex(uint32_t TypeSize) { @@ -544,17 +594,37 @@ static size_t TypeSizeToSizeIndex(uint32_t TypeSize) { 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); - GlobalVariable *GV = new GlobalVariable(M, StrConst->getType(), true, - GlobalValue::PrivateLinkage, StrConst, - kAsanGenPrefix); - GV->setUnnamedAddr(true); // Ok to merge these. + // 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; } +/// \brief Create a global describing a source location. +static GlobalVariable *createPrivateGlobalForSourceLoc(Module &M, + LocationMetadata MD) { + Constant *LocData[] = { + createPrivateGlobalForString(M, MD.Filename, true), + ConstantInt::get(Type::getInt32Ty(M.getContext()), MD.LineNo), + ConstantInt::get(Type::getInt32Ty(M.getContext()), MD.ColumnNo), + }; + auto LocStruct = ConstantStruct::getAnon(LocData); + auto GV = new GlobalVariable(M, LocStruct->getType(), true, + GlobalValue::PrivateLinkage, LocStruct, + kAsanGenPrefix); + GV->setUnnamedAddr(true); + return GV; +} + static bool GlobalWasGeneratedByAsan(GlobalVariable *G) { return G->getName().find(kAsanGenPrefix) == 0; } @@ -571,89 +641,123 @@ Value *AddressSanitizer::memToShadow(Value *Shadow, IRBuilder<> &IRB) { 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. - instrumentAddress(OrigIns, InsertBefore, Addr, 8, IsWrite, Size); - // Check the last byte. - 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 -bool AddressSanitizer::instrumentMemIntrinsic(MemIntrinsic *MI) { - Value *Dst = MI->getDest(); - MemTransferInst *MemTran = dyn_cast(MI); - Value *Src = MemTran ? MemTran->getSource() : 0; - Value *Length = MI->getLength(); - - Constant *ConstLength = dyn_cast(Length); - Instruction *InsertBefore = MI; - if (ConstLength) { - if (ConstLength->isNullValue()) return false; - } else { - // The size is not a constant so it could be zero -- check at run-time. - IRBuilder<> IRB(InsertBefore); - - Value *Cmp = IRB.CreateICmpNE(Length, - Constant::getNullValue(Length->getType())); - InsertBefore = SplitBlockAndInsertIfThen(cast(Cmp), false); +void AddressSanitizer::instrumentMemIntrinsic(MemIntrinsic *MI) { + IRBuilder<> IRB(MI); + if (isa(MI)) { + IRB.CreateCall3( + isa(MI) ? AsanMemmove : AsanMemcpy, + IRB.CreatePointerCast(MI->getOperand(0), IRB.getInt8PtrTy()), + IRB.CreatePointerCast(MI->getOperand(1), IRB.getInt8PtrTy()), + IRB.CreateIntCast(MI->getOperand(2), IntptrTy, false)); + } else if (isa(MI)) { + IRB.CreateCall3( + AsanMemset, + IRB.CreatePointerCast(MI->getOperand(0), IRB.getInt8PtrTy()), + IRB.CreateIntCast(MI->getOperand(1), IRB.getInt32Ty(), false), + IRB.CreateIntCast(MI->getOperand(2), IntptrTy, false)); } - - instrumentMemIntrinsicParam(MI, Dst, Length, InsertBefore, true); - if (Src) - instrumentMemIntrinsicParam(MI, Src, Length, InsertBefore, false); - return true; + MI->eraseFromParent(); } // If I is an interesting memory access, return the PointerOperand -// and set IsWrite. Otherwise return NULL. -static Value *isInterestingMemoryAccess(Instruction *I, bool *IsWrite) { +// and set IsWrite/Alignment. Otherwise return NULL. +static Value *isInterestingMemoryAccess(Instruction *I, bool *IsWrite, + unsigned *Alignment) { + // Skip memory accesses inserted by another instrumentation. + if (I->getMetadata("nosanitize")) + return nullptr; if (LoadInst *LI = dyn_cast(I)) { - if (!ClInstrumentReads) return NULL; + if (!ClInstrumentReads) return nullptr; *IsWrite = false; + *Alignment = LI->getAlignment(); return LI->getPointerOperand(); } if (StoreInst *SI = dyn_cast(I)) { - if (!ClInstrumentWrites) return NULL; + if (!ClInstrumentWrites) return nullptr; *IsWrite = true; + *Alignment = SI->getAlignment(); return SI->getPointerOperand(); } if (AtomicRMWInst *RMW = dyn_cast(I)) { - if (!ClInstrumentAtomics) return NULL; + if (!ClInstrumentAtomics) return nullptr; *IsWrite = true; + *Alignment = 0; return RMW->getPointerOperand(); } if (AtomicCmpXchgInst *XCHG = dyn_cast(I)) { - if (!ClInstrumentAtomics) return NULL; + if (!ClInstrumentAtomics) return nullptr; *IsWrite = true; + *Alignment = 0; return XCHG->getPointerOperand(); } - return NULL; + return nullptr; +} + +static bool isPointerOperand(Value *V) { + return V->getType()->isPointerTy() || isa(V); } -void AddressSanitizer::instrumentMop(Instruction *I) { +// 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() && !GlobalsMD.get(G).IsDynInit; +} + +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 UseCalls) { bool IsWrite = false; - Value *Addr = isInterestingMemoryAccess(I, &IsWrite); + unsigned Alignment = 0; + Value *Addr = isInterestingMemoryAccess(I, &IsWrite, &Alignment); assert(Addr); if (ClOpt && ClOptGlobals) { 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 (!ClInitializers || 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; + } + } } } @@ -661,26 +765,38 @@ void AddressSanitizer::instrumentMop(Instruction *I) { Type *OrigTy = cast(OrigPtrTy)->getElementType(); assert(OrigTy->isSized()); - uint32_t TypeSize = TD->getTypeStoreSizeInBits(OrigTy); + uint32_t TypeSize = DL->getTypeStoreSizeInBits(OrigTy); assert((TypeSize % 8) == 0); - // 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). + if (IsWrite) + NumInstrumentedWrites++; + else + NumInstrumentedReads++; + + unsigned Granularity = 1 << Mapping.Scale; + // Instrument a 1-, 2-, 4-, 8-, or 16- byte access with one check + // if the data is properly aligned. + if ((TypeSize == 8 || TypeSize == 16 || TypeSize == 32 || TypeSize == 64 || + TypeSize == 128) && + (Alignment >= Granularity || Alignment == 0 || Alignment >= TypeSize / 8)) + return instrumentAddress(I, I, Addr, TypeSize, IsWrite, nullptr, UseCalls); + // Instrument unusual size or unusual alignment. // 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); - 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); + Value *AddrLong = IRB.CreatePointerCast(Addr, IntptrTy); + if (UseCalls) { + IRB.CreateCall2(AsanMemoryAccessCallbackSized[IsWrite], AddrLong, Size); + } else { + Value *LastByte = IRB.CreateIntToPtr( + IRB.CreateAdd(AddrLong, ConstantInt::get(IntptrTy, TypeSize / 8 - 1)), + OrigPtrTy); + instrumentAddress(I, I, Addr, 8, IsWrite, Size, false); + instrumentAddress(I, I, LastByte, 8, IsWrite, Size, false); + } } // Validate the result of Module::getOrInsertFunction called for an interface @@ -728,11 +844,18 @@ Value *AddressSanitizer::createSlowPathCmp(IRBuilder<> &IRB, Value *AddrLong, } void AddressSanitizer::instrumentAddress(Instruction *OrigIns, - Instruction *InsertBefore, - Value *Addr, uint32_t TypeSize, - bool IsWrite, Value *SizeArgument) { + Instruction *InsertBefore, Value *Addr, + uint32_t TypeSize, bool IsWrite, + Value *SizeArgument, bool UseCalls) { IRBuilder<> IRB(InsertBefore); Value *AddrLong = IRB.CreatePointerCast(Addr, IntptrTy); + size_t AccessSizeIndex = TypeSizeToSizeIndex(TypeSize); + + if (UseCalls) { + IRB.CreateCall(AsanMemoryAccessCallback[IsWrite][AccessSizeIndex], + AddrLong); + return; + } Type *ShadowTy = IntegerType::get( *C, std::max(8U, TypeSize >> Mapping.Scale)); @@ -743,13 +866,15 @@ void AddressSanitizer::instrumentAddress(Instruction *OrigIns, IRB.CreateIntToPtr(ShadowPtr, ShadowPtrTy)); Value *Cmp = IRB.CreateICmpNE(ShadowValue, CmpVal); - size_t AccessSizeIndex = TypeSizeToSizeIndex(TypeSize); size_t Granularity = 1 << Mapping.Scale; - TerminatorInst *CrashTerm = 0; + TerminatorInst *CrashTerm = nullptr; if (ClAlwaysSlowPath || (TypeSize < 8 * Granularity)) { + // We use branch weights for the slow path check, to indicate that the slow + // path is rarely taken. This seems to be the case for SPEC benchmarks. TerminatorInst *CheckTerm = - SplitBlockAndInsertIfThen(cast(Cmp), false); + SplitBlockAndInsertIfThen(Cmp, InsertBefore, false, + MDBuilder(*C).createBranchWeights(1, 100000)); assert(dyn_cast(CheckTerm)->isUnconditional()); BasicBlock *NextBB = CheckTerm->getSuccessor(0); IRB.SetInsertPoint(CheckTerm); @@ -760,7 +885,7 @@ 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( @@ -768,27 +893,36 @@ void AddressSanitizer::instrumentAddress(Instruction *OrigIns, Crash->setDebugLoc(OrigIns->getDebugLoc()); } -void AddressSanitizerModule::createInitializerPoisonCalls( - 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 - // all been optimized away - if (!GlobalInit) - return; - +void AddressSanitizerModule::poisonOneInitializer(Function &GlobalInit, + GlobalValue *ModuleName) { // Set up the arguments to our poison/unpoison functions. - IRBuilder<> IRB(GlobalInit->begin()->getFirstInsertionPt()); + IRBuilder<> IRB(GlobalInit.begin()->getFirstInsertionPt()); // Add a call to poison all external globals before the given function starts. 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(); - I != E; ++I) { - if (ReturnInst *RI = dyn_cast(I->getTerminator())) { + for (auto &BB : GlobalInit.getBasicBlockList()) + if (ReturnInst *RI = dyn_cast(BB.getTerminator())) CallInst::Create(AsanUnpoisonGlobals, "", RI); +} + +void AddressSanitizerModule::createInitializerPoisonCalls( + Module &M, GlobalValue *ModuleName) { + GlobalVariable *GV = M.getGlobalVariable("llvm.global_ctors"); + + ConstantArray *CA = cast(GV->getInitializer()); + for (Use &OP : CA->operands()) { + if (isa(OP)) + continue; + ConstantStruct *CS = cast(OP); + + // Must have a function or null ptr. + // (CS->getOperand(0) is the init priority.) + if (Function* F = dyn_cast(CS->getOperand(1))) { + if (F->getName() != kAsanModuleCtorName) + poisonOneInitializer(*F, ModuleName); } } } @@ -797,23 +931,26 @@ bool AddressSanitizerModule::ShouldInstrumentGlobal(GlobalVariable *G) { Type *Ty = cast(G->getType())->getElementType(); DEBUG(dbgs() << "GLOBAL: " << *G << "\n"); - if (BL->isIn(*G)) return false; + if (GlobalsMD.get(G).IsBlacklisted) return false; if (!Ty->isSized()) return false; if (!G->hasInitializer()) return false; if (GlobalWasGeneratedByAsan(G)) return false; // Our own global. // Touch only those globals that will not be defined in other modules. - // Don't handle ODR type linkages since other modules may be built w/o asan. + // Don't handle ODR linkage types and COMDATs since other modules may be built + // without ASan. if (G->getLinkage() != GlobalVariable::ExternalLinkage && G->getLinkage() != GlobalVariable::PrivateLinkage && G->getLinkage() != GlobalVariable::InternalLinkage) return false; + if (G->hasComdat()) + return false; // Two problems with thread-locals: // - The address of the main thread's copy can't be computed at link-time. // - 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 @@ -821,7 +958,7 @@ bool AddressSanitizerModule::ShouldInstrumentGlobal(GlobalVariable *G) { // our redzones get broken. if ((G->getName().find("\01L_OBJC_") == 0) || (G->getName().find("\01l_OBJC_") == 0)) { - DEBUG(dbgs() << "Ignoring \\01L_OBJC_* global: " << *G); + DEBUG(dbgs() << "Ignoring \\01L_OBJC_* global: " << *G << "\n"); return false; } @@ -830,9 +967,9 @@ bool AddressSanitizerModule::ShouldInstrumentGlobal(GlobalVariable *G) { // Ignore the globals from the __OBJC section. The ObjC runtime assumes // those conform to /usr/lib/objc/runtime.h, so we can't add redzones to // them. - if ((Section.find("__OBJC,") == 0) || - (Section.find("__DATA, __objc_") == 0)) { - DEBUG(dbgs() << "Ignoring ObjC runtime global: " << *G); + if (Section.startswith("__OBJC,") || + Section.startswith("__DATA, __objc_")) { + DEBUG(dbgs() << "Ignoring ObjC runtime global: " << *G << "\n"); return false; } // See http://code.google.com/p/address-sanitizer/issues/detail?id=32 @@ -843,10 +980,28 @@ bool AddressSanitizerModule::ShouldInstrumentGlobal(GlobalVariable *G) { // is placed into __DATA,__cfstring // Therefore there's no point in placing redzones into __DATA,__cfstring. // Moreover, it causes the linker to crash on OS X 10.7 - if (Section.find("__DATA,__cfstring") == 0) { - DEBUG(dbgs() << "Ignoring CFString: " << *G); + if (Section.startswith("__DATA,__cfstring")) { + DEBUG(dbgs() << "Ignoring CFString: " << *G << "\n"); + return false; + } + // The linker merges the contents of cstring_literals and removes the + // trailing zeroes. + if (Section.startswith("__TEXT,__cstring,cstring_literals")) { + DEBUG(dbgs() << "Ignoring a cstring literal: " << *G << "\n"); + return false; + } + + // Callbacks put into the CRT initializer/terminator sections + // should not be instrumented. + // See https://code.google.com/p/address-sanitizer/issues/detail?id=305 + // and http://msdn.microsoft.com/en-US/en-en/library/bb918180(v=vs.120).aspx + if (Section.startswith(".CRT")) { + DEBUG(dbgs() << "Ignoring a global initializer callback: " << *G << "\n"); return false; } + + // Globals from llvm.metadata aren't emitted, do not instrument them. + if (Section == "llvm.metadata") return false; } return true; @@ -870,31 +1025,23 @@ void AddressSanitizerModule::initializeCallbacks(Module &M) { kAsanUnregisterGlobalsName, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); AsanUnregisterGlobals->setLinkage(Function::ExternalLinkage); + AsanCovModuleInit = checkInterfaceFunction(M.getOrInsertFunction( + kAsanCovModuleInitName, + IRB.getVoidTy(), IntptrTy, NULL)); + AsanCovModuleInit->setLinkage(Function::ExternalLinkage); } // This function replaces all global variables with new variables that have // trailing redzones. It also creates a function that poisons // redzones and inserts this function into llvm.global_ctors. -bool AddressSanitizerModule::runOnModule(Module &M) { - if (!ClGlobals) return false; - TD = getAnalysisIfAvailable(); - if (!TD) - return false; - BL.reset(new SpecialCaseList(BlacklistFile)); - if (BL->isIn(M)) return false; - C = &(M.getContext()); - int LongSize = TD->getPointerSizeInBits(); - IntptrTy = Type::getIntNTy(*C, LongSize); - Mapping = getShadowMapping(M, LongSize, ZeroBaseShadow); - initializeCallbacks(M); - DynamicallyInitializedGlobals.Init(M); +bool AddressSanitizerModule::InstrumentGlobals(IRBuilder<> &IRB, Module &M) { + GlobalsMD.init(M); SmallVector GlobalsToChange; - for (Module::GlobalListType::iterator G = M.global_begin(), - E = M.global_end(); G != E; ++G) { - if (ShouldInstrumentGlobal(G)) - GlobalsToChange.push_back(G); + for (auto &G : M.globals()) { + if (ShouldInstrumentGlobal(&G)) + GlobalsToChange.push_back(&G); } size_t n = GlobalsToChange.size(); @@ -907,32 +1054,35 @@ bool AddressSanitizerModule::runOnModule(Module &M) { // const char *name; // const char *module_name; // size_t has_dynamic_init; + // void *source_location; // We initialize an array of such structures and pass it to a run-time call. - StructType *GlobalStructTy = StructType::get(IntptrTy, IntptrTy, - IntptrTy, IntptrTy, - IntptrTy, IntptrTy, NULL); - SmallVector Initializers(n), DynamicInit; - - - Function *CtorFunc = M.getFunction(kAsanModuleCtorName); - assert(CtorFunc); - IRBuilder<> IRB(CtorFunc->getEntryBlock().getTerminator()); + StructType *GlobalStructTy = + StructType::get(IntptrTy, IntptrTy, IntptrTy, IntptrTy, IntptrTy, + IntptrTy, IntptrTy, NULL); + SmallVector Initializers(n); bool HasDynamicallyInitializedGlobals = false; - GlobalVariable *ModuleName = createPrivateGlobalForString( - M, M.getModuleIdentifier()); // We shouldn't merge same module names, as this string serves as unique // module ID in runtime. - ModuleName->setUnnamedAddr(false); + 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]; + + auto MD = GlobalsMD.get(G); + // Create string holding the global name (use global name from metadata + // if it's available, otherwise just write the name of global variable). + GlobalVariable *Name = createPrivateGlobalForString( + M, MD.Name.empty() ? G->getName() : MD.Name, + /*AllowMerging*/ true); + PointerType *PtrTy = cast(G->getType()); Type *Ty = PtrTy->getElementType(); - uint64_t SizeInBytes = TD->getTypeAllocSize(Ty); - uint64_t MinRZ = RedzoneSize(); + 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, @@ -944,22 +1094,18 @@ bool AddressSanitizerModule::runOnModule(Module &M) { 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); StructType *NewTy = StructType::get(Ty, RightRedZoneTy, NULL); Constant *NewInitializer = ConstantStruct::get( NewTy, G->getInitializer(), Constant::getNullValue(RightRedZoneTy), NULL); - GlobalVariable *Name = createPrivateGlobalForString(M, G->getName()); - // 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(MinRZ); @@ -973,18 +1119,23 @@ bool AddressSanitizerModule::runOnModule(Module &M) { NewGlobal->takeName(G); G->eraseFromParent(); + Constant *SourceLoc; + if (!MD.SourceLoc.empty()) { + auto SourceLocGlobal = createPrivateGlobalForSourceLoc(M, MD.SourceLoc); + SourceLoc = ConstantExpr::getPointerCast(SourceLocGlobal, IntptrTy); + } else { + SourceLoc = ConstantInt::get(IntptrTy, 0); + } + Initializers[i] = ConstantStruct::get( - GlobalStructTy, - ConstantExpr::getPointerCast(NewGlobal, IntptrTy), + GlobalStructTy, ConstantExpr::getPointerCast(NewGlobal, IntptrTy), ConstantInt::get(IntptrTy, SizeInBytes), ConstantInt::get(IntptrTy, SizeInBytes + RightRedzoneSize), ConstantExpr::getPointerCast(Name, IntptrTy), ConstantExpr::getPointerCast(ModuleName, IntptrTy), - ConstantInt::get(IntptrTy, GlobalHasDynamicInitializer), - NULL); + ConstantInt::get(IntptrTy, MD.IsDynInit), SourceLoc, NULL); - // Populate the first and last globals declared in this TU. - if (CheckInitOrder && GlobalHasDynamicInitializer) + if (ClInitializers && MD.IsDynInit) HasDynamicallyInitializedGlobals = true; DEBUG(dbgs() << "NEW GLOBAL: " << *NewGlobal << "\n"); @@ -992,11 +1143,11 @@ bool AddressSanitizerModule::runOnModule(Module &M) { 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 && HasDynamicallyInitializedGlobals) + if (HasDynamicallyInitializedGlobals) createInitializerPoisonCalls(M, ModuleName); IRB.CreateCall2(AsanRegisterGlobals, IRB.CreatePointerCast(AllGlobals, IntptrTy), @@ -1012,12 +1163,42 @@ bool AddressSanitizerModule::runOnModule(Module &M) { IRB_Dtor.CreateCall2(AsanUnregisterGlobals, IRB.CreatePointerCast(AllGlobals, IntptrTy), ConstantInt::get(IntptrTy, n)); - appendToGlobalDtors(M, AsanDtorFunction, kAsanCtorAndCtorPriority); + appendToGlobalDtors(M, AsanDtorFunction, kAsanCtorAndDtorPriority); DEBUG(dbgs() << M); return true; } +bool AddressSanitizerModule::runOnModule(Module &M) { + DataLayoutPass *DLP = getAnalysisIfAvailable(); + if (!DLP) + return false; + DL = &DLP->getDataLayout(); + C = &(M.getContext()); + int LongSize = DL->getPointerSizeInBits(); + IntptrTy = Type::getIntNTy(*C, LongSize); + Mapping = getShadowMapping(M, LongSize); + initializeCallbacks(M); + + bool Changed = false; + + Function *CtorFunc = M.getFunction(kAsanModuleCtorName); + assert(CtorFunc); + IRBuilder<> IRB(CtorFunc->getEntryBlock().getTerminator()); + + if (ClCoverage > 0) { + Function *CovFunc = M.getFunction(kAsanCovName); + int nCov = CovFunc ? CovFunc->getNumUses() : 0; + IRB.CreateCall(AsanCovModuleInit, ConstantInt::get(IntptrTy, nCov)); + Changed = true; + } + + if (ClGlobals) + Changed |= InstrumentGlobals(IRB, M); + + return Changed; +} + void AddressSanitizer::initializeCallbacks(Module &M) { IRBuilder<> IRB(*C); // Create __asan_report* callbacks. @@ -1025,12 +1206,16 @@ void AddressSanitizer::initializeCallbacks(Module &M) { for (size_t AccessSizeIndex = 0; AccessSizeIndex < kNumberOfAccessSizes; AccessSizeIndex++) { // IsWrite and TypeSize are encoded in the function name. - std::string FunctionName = std::string(kAsanReportErrorTemplate) + + std::string Suffix = (AccessIsWrite ? "store" : "load") + itostr(1 << AccessSizeIndex); - // If we are merging crash callbacks, they have two parameters. AsanErrorCallback[AccessIsWrite][AccessSizeIndex] = - checkInterfaceFunction(M.getOrInsertFunction( - FunctionName, IRB.getVoidTy(), IntptrTy, NULL)); + checkInterfaceFunction( + M.getOrInsertFunction(kAsanReportErrorTemplate + Suffix, + IRB.getVoidTy(), IntptrTy, NULL)); + AsanMemoryAccessCallback[AccessIsWrite][AccessSizeIndex] = + checkInterfaceFunction( + M.getOrInsertFunction(ClMemoryAccessCallbackPrefix + Suffix, + IRB.getVoidTy(), IntptrTy, NULL)); } } AsanErrorCallbackSized[0] = checkInterfaceFunction(M.getOrInsertFunction( @@ -1038,43 +1223,49 @@ void AddressSanitizer::initializeCallbacks(Module &M) { AsanErrorCallbackSized[1] = checkInterfaceFunction(M.getOrInsertFunction( kAsanReportStoreN, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); - AsanHandleNoReturnFunc = checkInterfaceFunction(M.getOrInsertFunction( - kAsanHandleNoReturnName, IRB.getVoidTy(), NULL)); + AsanMemoryAccessCallbackSized[0] = checkInterfaceFunction( + M.getOrInsertFunction(ClMemoryAccessCallbackPrefix + "loadN", + IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + AsanMemoryAccessCallbackSized[1] = checkInterfaceFunction( + M.getOrInsertFunction(ClMemoryAccessCallbackPrefix + "storeN", + IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + + AsanMemmove = checkInterfaceFunction(M.getOrInsertFunction( + ClMemoryAccessCallbackPrefix + "memmove", IRB.getInt8PtrTy(), + IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IntptrTy, NULL)); + AsanMemcpy = checkInterfaceFunction(M.getOrInsertFunction( + ClMemoryAccessCallbackPrefix + "memcpy", IRB.getInt8PtrTy(), + IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IntptrTy, NULL)); + AsanMemset = checkInterfaceFunction(M.getOrInsertFunction( + ClMemoryAccessCallbackPrefix + "memset", IRB.getInt8PtrTy(), + IRB.getInt8PtrTy(), IRB.getInt32Ty(), 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. - 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); - - 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(); + DataLayoutPass *DLP = getAnalysisIfAvailable(); + if (!DLP) + report_fatal_error("data layout missing"); + DL = &DLP->getDataLayout(); - if (!TD) - return false; - BL.reset(new SpecialCaseList(BlacklistFile)); - DynamicallyInitializedGlobals.Init(M); + GlobalsMD.init(M); C = &(M.getContext()); - LongSize = TD->getPointerSizeInBits(); + LongSize = DL->getPointerSizeInBits(); IntptrTy = Type::getIntNTy(*C, LongSize); AsanCtorFunction = Function::Create( @@ -1088,10 +1279,9 @@ bool AddressSanitizer::doInitialization(Module &M) { AsanInitFunction->setLinkage(Function::ExternalLinkage); IRB.CreateCall(AsanInitFunction); - Mapping = getShadowMapping(M, LongSize, ZeroBaseShadow); - emitShadowMapping(M, IRB); + Mapping = getShadowMapping(M, LongSize); - appendToGlobalCtors(M, AsanCtorFunction, kAsanCtorAndCtorPriority); + appendToGlobalCtors(M, AsanCtorFunction, kAsanCtorAndDtorPriority); return true; } @@ -1111,8 +1301,74 @@ 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; + } + + DebugLoc EntryLoc = IP->getDebugLoc().getFnDebugLoc(*C); + IRBuilder<> IRB(IP); + IRB.SetCurrentDebugLocation(EntryLoc); + 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); + IRB.SetCurrentDebugLocation(EntryLoc); + // __sanitizer_cov gets the PC of the instruction using GET_CALLER_PC. + IRB.CreateCall(AsanCovFunction); + 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(); +// *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, + ArrayRef AllBlocks) { + if (!ClCoverage) return false; + + if (ClCoverage == 1 || + (unsigned)ClCoverageBlockThreshold < AllBlocks.size()) { + InjectCoverageAtBlock(F, F.getEntryBlock()); + } else { + for (auto BB : AllBlocks) + InjectCoverageAtBlock(F, *BB); + } + 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"); @@ -1132,28 +1388,35 @@ bool AddressSanitizer::runOnFunction(Function &F) { SmallSet TempsToInstrument; SmallVector ToInstrument; SmallVector NoReturnCalls; + SmallVector AllBlocks; + SmallVector PointerComparisonsOrSubtracts; int NumAllocas = 0; bool IsWrite; + unsigned Alignment; // Fill the set of memory operations to instrument. - for (Function::iterator FI = F.begin(), FE = F.end(); - FI != FE; ++FI) { + for (auto &BB : F) { + AllBlocks.push_back(&BB); TempsToInstrument.clear(); int NumInsnsPerBB = 0; - for (BasicBlock::iterator BI = FI->begin(), BE = FI->end(); - BI != BE; ++BI) { - if (LooksLikeCodeInBug11395(BI)) return false; - if (Value *Addr = isInterestingMemoryAccess(BI, &IsWrite)) { + for (auto &Inst : BB) { + if (LooksLikeCodeInBug11395(&Inst)) return false; + if (Value *Addr = + isInterestingMemoryAccess(&Inst, &IsWrite, &Alignment)) { if (ClOpt && ClOptSameTemp) { if (!TempsToInstrument.insert(Addr)) continue; // We've seen this temp in the current BB. } - } else if (isa(BI) && ClMemIntrin) { + } else if (ClInvalidPointerPairs && + isInterestingPointerComparisonOrSubtraction(&Inst)) { + PointerComparisonsOrSubtracts.push_back(&Inst); + continue; + } else if (isa(Inst)) { // ok, take it. } else { - if (isa(BI)) + if (isa(Inst)) NumAllocas++; - CallSite CS(BI); + CallSite CS(&Inst); if (CS) { // A call inside BB. TempsToInstrument.clear(); @@ -1162,14 +1425,14 @@ bool AddressSanitizer::runOnFunction(Function &F) { } continue; } - ToInstrument.push_back(BI); + ToInstrument.push_back(&Inst); NumInsnsPerBB++; if (NumInsnsPerBB >= ClMaxInsnsToInstrumentPerBB) break; } } - Function *UninstrumentedDuplicate = 0; + Function *UninstrumentedDuplicate = nullptr; bool LikelyToInstrument = !NoReturnCalls.empty() || !ToInstrument.empty() || (NumAllocas > 0); if (ClKeepUninstrumented && LikelyToInstrument) { @@ -1180,14 +1443,18 @@ bool AddressSanitizer::runOnFunction(Function &F) { F.getParent()->getFunctionList().push_back(UninstrumentedDuplicate); } + bool UseCalls = false; + if (ClInstrumentationWithCallsThreshold >= 0 && + ToInstrument.size() > (unsigned)ClInstrumentationWithCallsThreshold) + UseCalls = true; + // Instrument. int NumInstrumented = 0; - for (size_t i = 0, n = ToInstrument.size(); i != n; i++) { - Instruction *Inst = ToInstrument[i]; + for (auto Inst : ToInstrument) { if (ClDebugMin < 0 || ClDebugMax < 0 || (NumInstrumented >= ClDebugMin && NumInstrumented <= ClDebugMax)) { - if (isInterestingMemoryAccess(Inst, &IsWrite)) - instrumentMop(Inst); + if (isInterestingMemoryAccess(Inst, &IsWrite, &Alignment)) + instrumentMop(Inst, UseCalls); else instrumentMemIntrinsic(cast(Inst)); } @@ -1199,13 +1466,21 @@ bool AddressSanitizer::runOnFunction(Function &F) { // We must unpoison the stack before every NoReturn call (throw, _exit, etc). // See e.g. http://code.google.com/p/address-sanitizer/issues/detail?id=37 - for (size_t i = 0, n = NoReturnCalls.size(); i != n; i++) { - Instruction *CI = NoReturnCalls[i]; + for (auto CI : NoReturnCalls) { IRBuilder<> IRB(CI); IRB.CreateCall(AsanHandleNoReturnFunc); } + for (auto Inst : PointerComparisonsOrSubtracts) { + instrumentPointerComparisonOrSubtraction(Inst); + NumInstrumented++; + } + bool res = NumInstrumented > 0 || ChangedStack || !NoReturnCalls.empty(); + + if (InjectCoverage(F, AllBlocks)) + res = true; + DEBUG(dbgs() << "ASAN done instrumenting: " << res << " " << F << "\n"); if (ClKeepUninstrumented) { @@ -1225,32 +1500,6 @@ bool AddressSanitizer::runOnFunction(Function &F) { return res; } -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"); -} - -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 - } else { - *Shadow = Size - i; // first Size-i bytes are addressable - } - } -} - // Workaround for bug 11395: we don't want to instrument stack in functions // with large assembly blobs (32-bit only), otherwise reg alloc may crash. // FIXME: remove once the bug 11395 is fixed. @@ -1265,141 +1514,166 @@ 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); - Poison = - ASan.TD->isLittleEndian() - ? support::endian::byte_swap(Poison) - : support::endian::byte_swap(Poison); +void +FunctionStackPoisoner::poisonRedZones(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(); +static DebugLoc getFunctionEntryDebugLocation(Function &F) { + for (const auto &Inst : F.getEntryBlock()) + if (!isa(Inst)) + return Inst.getDebugLoc(); + return DebugLoc(); +} - bool DoStackMalloc = ASan.CheckUseAfterReturn - && LocalStackSize <= kMaxStackMallocSize; +void FunctionStackPoisoner::poisonStack() { + int StackMallocIdx = -1; + DebugLoc EntryDebugLocation = getFunctionEntryDebugLocation(F); assert(AllocaVec.size() > 0); Instruction *InsBefore = AllocaVec[0]; IRBuilder<> IRB(InsBefore); - + IRB.SetCurrentDebugLocation(EntryDebugLocation); + + SmallVector SVD; + SVD.reserve(AllocaVec.size()); + for (AllocaInst *AI : AllocaVec) { + 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 = + ClUseAfterReturn && 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); + MyAlloca->setDebugLoc(EntryDebugLocation); + 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); + IRBIf.SetCurrentDebugLocation(EntryDebugLocation); + LocalStackBase = IRBIf.CreateCall2( + AsanStackMallocFunc[StackMallocIdx], ConstantInt::get(IntptrTy, LocalStackSize), OrigStackBase); + BasicBlock *SetBlock = cast(LocalStackBase)->getParent(); + IRB.SetInsertPoint(InsBefore); + IRB.SetCurrentDebugLocation(EntryDebugLocation); + 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 (DescribeAddressIfStack). - SmallString<2048> StackDescriptionStorage; - raw_svector_ostream StackDescription(StackDescriptionStorage); - StackDescription << 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); + for (const auto &APC : AllocaPoisonCallVec) { + 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 (const auto &Desc : SVD) { + AllocaInst *AI = Desc.AI; Value *NewAllocaPtr = IRB.CreateIntToPtr( - IRB.CreateAdd(LocalStackBase, ConstantInt::get(IntptrTy, Pos)), - AI->getType()); + IRB.CreateAdd(LocalStackBase, ConstantInt::get(IntptrTy, Desc.Offset)), + AI->getType()); replaceDbgDeclareForAlloca(AI, NewAllocaPtr, DIB); AI->replaceAllUsesWith(NewAllocaPtr); - Pos += AlignedSize + RedzoneSize(); } - assert(Pos == LocalStackSize); // The left-most redzone has enough space for at least 4 pointers. // Write the Magic value to redzone[0]. @@ -1411,7 +1685,8 @@ void FunctionStackPoisoner::poisonStack() { 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); @@ -1424,37 +1699,69 @@ void FunctionStackPoisoner::poisonStack() { // 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. - for (size_t i = 0, n = RetVec.size(); i < n; i++) { - Instruction *Ret = RetVec[i]; + // (Un)poison the stack before all ret instructions. + for (auto Ret : RetVec) { 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); } } // We are done. Remove the old unused alloca instructions. - for (size_t i = 0, n = AllocaVec.size(); i < n; i++) - AllocaVec[i]->eraseFromParent(); + for (auto AI : AllocaVec) + AI->eraseFromParent(); } 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); @@ -1475,7 +1782,7 @@ void FunctionStackPoisoner::poisonAlloca(Value *V, uint64_t Size, AllocaInst *FunctionStackPoisoner::findAllocaForValue(Value *V) { if (AllocaInst *AI = dyn_cast(V)) // We're intested only in allocas we can handle. - return isInterestingAlloca(*AI) ? AI : 0; + return isInterestingAlloca(*AI) ? AI : nullptr; // See if we've already calculated (or started to calculate) alloca for a // given value. AllocaForValueMapTy::iterator I = AllocaForValue.find(V); @@ -1483,8 +1790,8 @@ AllocaInst *FunctionStackPoisoner::findAllocaForValue(Value *V) { return I->second; // Store 0 while we're calculating alloca for value V to avoid // infinite recursion if the value references itself. - AllocaForValue[V] = 0; - AllocaInst *Res = 0; + AllocaForValue[V] = nullptr; + AllocaInst *Res = nullptr; if (CastInst *CI = dyn_cast(V)) Res = findAllocaForValue(CI->getOperand(0)); else if (PHINode *PN = dyn_cast(V)) { @@ -1494,12 +1801,12 @@ AllocaInst *FunctionStackPoisoner::findAllocaForValue(Value *V) { if (IncValue == PN) continue; AllocaInst *IncValueAI = findAllocaForValue(IncValue); // AI for incoming values should exist and should all be equal. - if (IncValueAI == 0 || (Res != 0 && IncValueAI != Res)) - return 0; + if (IncValueAI == nullptr || (Res != nullptr && IncValueAI != Res)) + return nullptr; Res = IncValueAI; } } - if (Res != 0) + if (Res) AllocaForValue[V] = Res; return Res; }