diff options
Diffstat (limited to 'lib/Transforms/Instrumentation')
-rw-r--r-- | lib/Transforms/Instrumentation/AddressSanitizer.cpp | 361 | ||||
-rw-r--r-- | lib/Transforms/Instrumentation/Android.mk | 1 | ||||
-rw-r--r-- | lib/Transforms/Instrumentation/CMakeLists.txt | 1 | ||||
-rw-r--r-- | lib/Transforms/Instrumentation/DataFlowSanitizer.cpp | 282 | ||||
-rw-r--r-- | lib/Transforms/Instrumentation/DebugIR.cpp | 8 | ||||
-rw-r--r-- | lib/Transforms/Instrumentation/DebugIR.h | 6 | ||||
-rw-r--r-- | lib/Transforms/Instrumentation/GCOVProfiling.cpp | 14 | ||||
-rw-r--r-- | lib/Transforms/Instrumentation/Instrumentation.cpp | 1 | ||||
-rw-r--r-- | lib/Transforms/Instrumentation/MemorySanitizer.cpp | 140 | ||||
-rw-r--r-- | lib/Transforms/Instrumentation/SanitizerCoverage.cpp | 294 | ||||
-rw-r--r-- | lib/Transforms/Instrumentation/ThreadSanitizer.cpp | 42 |
11 files changed, 791 insertions, 359 deletions
diff --git a/lib/Transforms/Instrumentation/AddressSanitizer.cpp b/lib/Transforms/Instrumentation/AddressSanitizer.cpp index 5e5ddc1..38f587f 100644 --- a/lib/Transforms/Instrumentation/AddressSanitizer.cpp +++ b/lib/Transforms/Instrumentation/AddressSanitizer.cpp @@ -40,6 +40,7 @@ #include "llvm/Support/DataTypes.h" #include "llvm/Support/Debug.h" #include "llvm/Support/Endian.h" +#include "llvm/Transforms/Scalar.h" #include "llvm/Transforms/Utils/ASanStackFrameLayout.h" #include "llvm/Transforms/Utils/BasicBlockUtils.h" #include "llvm/Transforms/Utils/Cloning.h" @@ -59,7 +60,8 @@ static const uint64_t kIOSShadowOffset32 = 1ULL << 30; static const uint64_t kDefaultShadowOffset64 = 1ULL << 44; 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 kMIPS32_ShadowOffset32 = 0x0aaa0000; +static const uint64_t kMIPS64_ShadowOffset64 = 1ULL << 36; static const uint64_t kFreeBSD_ShadowOffset32 = 1ULL << 30; static const uint64_t kFreeBSD_ShadowOffset64 = 1ULL << 46; @@ -70,7 +72,7 @@ static const uintptr_t kRetiredStackFrameMagic = 0x45E0360E; static const char *const kAsanModuleCtorName = "asan.module_ctor"; static const char *const kAsanModuleDtorName = "asan.module_dtor"; -static const int kAsanCtorAndDtorPriority = 1; +static const uint64_t 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"; @@ -80,8 +82,6 @@ static const char *const kAsanUnregisterGlobalsName = 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"; @@ -89,6 +89,7 @@ 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 kSanCovGenPrefix = "__sancov_gen_"; static const char *const kAsanPoisonStackMemoryName = "__asan_poison_stack_memory"; static const char *const kAsanUnpoisonStackMemoryName = @@ -133,13 +134,6 @@ static cl::opt<bool> ClUseAfterReturn("asan-use-after-return", // This flag may need to be replaced with -f[no]asan-globals. static cl::opt<bool> ClGlobals("asan-globals", cl::desc("Handle global objects"), cl::Hidden, cl::init(true)); -static cl::opt<int> ClCoverage("asan-coverage", - cl::desc("ASan coverage. 0: none, 1: entry block, 2: all blocks"), - cl::Hidden, cl::init(false)); -static cl::opt<int> 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<bool> ClInitializers("asan-initialization-order", cl::desc("Handle C++ initializer order"), cl::Hidden, cl::init(true)); static cl::opt<bool> ClInvalidPointerPairs("asan-detect-invalid-pointer-pair", @@ -212,10 +206,40 @@ STATISTIC(NumOptimizedAccessesToGlobalVar, "Number of optimized accesses to global vars"); namespace { +/// 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<MDString>(MDN->getOperand(0)); + Filename = MDFilename->getString(); + LineNo = cast<ConstantInt>(MDN->getOperand(1))->getLimitedValue(); + ColumnNo = cast<ConstantInt>(MDN->getOperand(2))->getLimitedValue(); + } +}; + /// Frontend-provided metadata for global variables. class GlobalsMetadata { public: + 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; @@ -223,76 +247,38 @@ class GlobalsMetadata { if (!Globals) return; for (auto MDN : Globals->operands()) { - // Format of the metadata node for the global: - // { - // global, - // source_location, - // i1 is_dynamically_initialized, - // i1 is_blacklisted - // } - assert(MDN->getNumOperands() == 4); + // 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; GlobalVariable *GV = cast<GlobalVariable>(V); - if (Value *Loc = MDN->getOperand(1)) { - GlobalVariable *GVLoc = cast<GlobalVariable>(Loc); - // We may already know the source location for GV, if it was merged - // with another global. - if (SourceLocation.insert(std::make_pair(GV, GVLoc)).second) - addSourceLocationGlobal(GVLoc); + // 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<MDNode>(Loc)); + if (Value *Name = MDN->getOperand(2)) { + MDString *MDName = cast<MDString>(Name); + E.Name = MDName->getString(); } - ConstantInt *IsDynInit = cast<ConstantInt>(MDN->getOperand(2)); - if (IsDynInit->isOne()) - DynInitGlobals.insert(GV); - ConstantInt *IsBlacklisted = cast<ConstantInt>(MDN->getOperand(3)); - if (IsBlacklisted->isOne()) - BlacklistedGlobals.insert(GV); + ConstantInt *IsDynInit = cast<ConstantInt>(MDN->getOperand(3)); + E.IsDynInit |= IsDynInit->isOne(); + ConstantInt *IsBlacklisted = cast<ConstantInt>(MDN->getOperand(4)); + E.IsBlacklisted |= IsBlacklisted->isOne(); } } - GlobalVariable *getSourceLocation(GlobalVariable *G) const { - auto Pos = SourceLocation.find(G); - return (Pos != SourceLocation.end()) ? Pos->second : nullptr; - } - - /// Check if the global is dynamically initialized. - bool isDynInit(GlobalVariable *G) const { - return DynInitGlobals.count(G); - } - - /// Check if the global was blacklisted. - bool isBlacklisted(GlobalVariable *G) const { - return BlacklistedGlobals.count(G); - } - - /// Check if the global was generated to describe source location of another - /// global (we don't want to instrument them). - bool isSourceLocationGlobal(GlobalVariable *G) const { - return LocationGlobals.count(G); + /// 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: bool inited_; - DenseMap<GlobalVariable*, GlobalVariable*> SourceLocation; - DenseSet<GlobalVariable*> DynInitGlobals; - DenseSet<GlobalVariable*> BlacklistedGlobals; - DenseSet<GlobalVariable*> LocationGlobals; - - void addSourceLocationGlobal(GlobalVariable *SourceLocGV) { - // Source location global is a struct with layout: - // { - // filename, - // i32 line_number, - // i32 column_number, - // } - LocationGlobals.insert(SourceLocGV); - ConstantStruct *Contents = - cast<ConstantStruct>(SourceLocGV->getInitializer()); - GlobalVariable *FilenameGV = cast<GlobalVariable>(Contents->getOperand(0)); - LocationGlobals.insert(FilenameGV); - } + DenseMap<GlobalVariable*, Entry> Entries; }; /// This struct defines the shadow mapping using the rule: @@ -306,7 +292,7 @@ struct ShadowMapping { static ShadowMapping getShadowMapping(const Module &M, int LongSize) { llvm::Triple TargetTriple(M.getTargetTriple()); bool IsAndroid = TargetTriple.getEnvironment() == llvm::Triple::Android; - bool IsIOS = TargetTriple.getOS() == llvm::Triple::IOS; + bool IsIOS = TargetTriple.isiOS(); bool IsFreeBSD = TargetTriple.getOS() == llvm::Triple::FreeBSD; bool IsLinux = TargetTriple.getOS() == llvm::Triple::Linux; bool IsPPC64 = TargetTriple.getArch() == llvm::Triple::ppc64 || @@ -314,6 +300,8 @@ static ShadowMapping getShadowMapping(const Module &M, int LongSize) { bool IsX86_64 = TargetTriple.getArch() == llvm::Triple::x86_64; bool IsMIPS32 = TargetTriple.getArch() == llvm::Triple::mips || TargetTriple.getArch() == llvm::Triple::mipsel; + bool IsMIPS64 = TargetTriple.getArch() == llvm::Triple::mips64 || + TargetTriple.getArch() == llvm::Triple::mips64el; ShadowMapping Mapping; @@ -335,6 +323,8 @@ static ShadowMapping getShadowMapping(const Module &M, int LongSize) { Mapping.Offset = kFreeBSD_ShadowOffset64; else if (IsLinux && IsX86_64) Mapping.Offset = kSmallX86_64ShadowOffset; + else if (IsMIPS64) + Mapping.Offset = kMIPS64_ShadowOffset64; else Mapping.Offset = kDefaultShadowOffset64; } @@ -386,8 +376,6 @@ struct AddressSanitizer : public FunctionPass { bool LooksLikeCodeInBug11395(Instruction *I); bool GlobalIsLinkerInitialized(GlobalVariable *G); - bool InjectCoverage(Function &F, const ArrayRef<BasicBlock*> AllBlocks); - void InjectCoverageAtBlock(Function &F, BasicBlock &BB); LLVMContext *C; const DataLayout *DL; @@ -397,7 +385,6 @@ struct AddressSanitizer : public FunctionPass { Function *AsanCtorFunction; Function *AsanInitFunction; Function *AsanHandleNoReturnFunc; - Function *AsanCovFunction; Function *AsanPtrCmpFunction, *AsanPtrSubFunction; // This array is indexed by AccessIsWrite and log2(AccessSize). Function *AsanErrorCallback[2][kNumberOfAccessSizes]; @@ -441,7 +428,6 @@ class AddressSanitizerModule : public ModulePass { Function *AsanUnpoisonGlobals; Function *AsanRegisterGlobals; Function *AsanUnregisterGlobals; - Function *AsanCovModuleInit; }; // Stack poisoning does not play well with exception handling. @@ -570,7 +556,7 @@ struct FunctionStackPoisoner : public InstVisitor<FunctionStackPoisoner> { } /// Finds alloca where the value comes from. AllocaInst *findAllocaForValue(Value *V); - void poisonRedZones(const ArrayRef<uint8_t> ShadowBytes, IRBuilder<> &IRB, + void poisonRedZones(ArrayRef<uint8_t> ShadowBytes, IRBuilder<> &IRB, Value *ShadowBase, bool DoPoison); void poisonAlloca(Value *V, uint64_t Size, IRBuilder<> &IRB, bool DoPoison); @@ -617,8 +603,25 @@ static GlobalVariable *createPrivateGlobalForString( 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; + return G->getName().find(kAsanGenPrefix) == 0 || + G->getName().find(kSanCovGenPrefix) == 0; } Value *AddressSanitizer::memToShadow(Value *Shadow, IRBuilder<> &IRB) { @@ -653,9 +656,12 @@ void AddressSanitizer::instrumentMemIntrinsic(MemIntrinsic *MI) { } // If I is an interesting memory access, return the PointerOperand -// and set IsWrite/Alignment. Otherwise return NULL. +// and set IsWrite/Alignment. Otherwise return nullptr. 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<LoadInst>(I)) { if (!ClInstrumentReads) return nullptr; *IsWrite = false; @@ -710,7 +716,7 @@ 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.isDynInit(G); + return G->hasInitializer() && !GlobalsMD.get(G).IsDynInit; } void @@ -859,8 +865,11 @@ void AddressSanitizer::instrumentAddress(Instruction *OrigIns, 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(Cmp, InsertBefore, false); + SplitBlockAndInsertIfThen(Cmp, InsertBefore, false, + MDBuilder(*C).createBranchWeights(1, 100000)); assert(dyn_cast<BranchInst>(CheckTerm)->isUnconditional()); BasicBlock *NextBB = CheckTerm->getSuccessor(0); IRB.SetInsertPoint(CheckTerm); @@ -905,10 +914,12 @@ void AddressSanitizerModule::createInitializerPoisonCalls( ConstantStruct *CS = cast<ConstantStruct>(OP); // Must have a function or null ptr. - // (CS->getOperand(0) is the init priority.) if (Function* F = dyn_cast<Function>(CS->getOperand(1))) { - if (F->getName() != kAsanModuleCtorName) - poisonOneInitializer(*F, ModuleName); + if (F->getName() == kAsanModuleCtorName) continue; + ConstantInt *Priority = dyn_cast<ConstantInt>(CS->getOperand(0)); + // Don't instrument CTORs that will run before asan.module_ctor. + if (Priority->getLimitedValue() <= kAsanCtorAndDtorPriority) continue; + poisonOneInitializer(*F, ModuleName); } } } @@ -917,8 +928,7 @@ bool AddressSanitizerModule::ShouldInstrumentGlobal(GlobalVariable *G) { Type *Ty = cast<PointerType>(G->getType())->getElementType(); DEBUG(dbgs() << "GLOBAL: " << *G << "\n"); - if (GlobalsMD.isBlacklisted(G)) return false; - if (GlobalsMD.isSourceLocationGlobal(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. @@ -939,16 +949,6 @@ bool AddressSanitizerModule::ShouldInstrumentGlobal(GlobalVariable *G) { // 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 - // that section by removing the spare \0s after the string terminator, so - // our redzones get broken. - if ((G->getName().find("\01L_OBJC_") == 0) || - (G->getName().find("\01l_OBJC_") == 0)) { - DEBUG(dbgs() << "Ignoring \\01L_OBJC_* global: " << *G << "\n"); - return false; - } - if (G->hasSection()) { StringRef Section(G->getSection()); // Ignore the globals from the __OBJC section. The ObjC runtime assumes @@ -977,6 +977,11 @@ bool AddressSanitizerModule::ShouldInstrumentGlobal(GlobalVariable *G) { DEBUG(dbgs() << "Ignoring a cstring literal: " << *G << "\n"); return false; } + if (Section.startswith("__TEXT,__objc_methname,cstring_literals")) { + DEBUG(dbgs() << "Ignoring objc_methname cstring global: " << *G << "\n"); + return false; + } + // Callbacks put into the CRT initializer/terminator sections // should not be instrumented. @@ -998,24 +1003,20 @@ void AddressSanitizerModule::initializeCallbacks(Module &M) { IRBuilder<> IRB(*C); // Declare our poisoning and unpoisoning functions. AsanPoisonGlobals = checkInterfaceFunction(M.getOrInsertFunction( - kAsanPoisonGlobalsName, IRB.getVoidTy(), IntptrTy, NULL)); + kAsanPoisonGlobalsName, IRB.getVoidTy(), IntptrTy, nullptr)); AsanPoisonGlobals->setLinkage(Function::ExternalLinkage); AsanUnpoisonGlobals = checkInterfaceFunction(M.getOrInsertFunction( - kAsanUnpoisonGlobalsName, IRB.getVoidTy(), NULL)); + kAsanUnpoisonGlobalsName, IRB.getVoidTy(), nullptr)); AsanUnpoisonGlobals->setLinkage(Function::ExternalLinkage); // Declare functions that register/unregister globals. AsanRegisterGlobals = checkInterfaceFunction(M.getOrInsertFunction( kAsanRegisterGlobalsName, IRB.getVoidTy(), - IntptrTy, IntptrTy, NULL)); + IntptrTy, IntptrTy, nullptr)); AsanRegisterGlobals->setLinkage(Function::ExternalLinkage); AsanUnregisterGlobals = checkInterfaceFunction(M.getOrInsertFunction( kAsanUnregisterGlobalsName, - IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + IRB.getVoidTy(), IntptrTy, IntptrTy, nullptr)); 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 @@ -1045,7 +1046,7 @@ bool AddressSanitizerModule::InstrumentGlobals(IRBuilder<> &IRB, Module &M) { // 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, IntptrTy, NULL); + IntptrTy, IntptrTy, nullptr); SmallVector<Constant *, 16> Initializers(n); bool HasDynamicallyInitializedGlobals = false; @@ -1058,6 +1059,14 @@ bool AddressSanitizerModule::InstrumentGlobals(IRBuilder<> &IRB, Module &M) { 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<PointerType>(G->getType()); Type *Ty = PtrTy->getElementType(); uint64_t SizeInBytes = DL->getTypeAllocSize(Ty); @@ -1074,13 +1083,10 @@ bool AddressSanitizerModule::InstrumentGlobals(IRBuilder<> &IRB, Module &M) { assert(((RightRedzoneSize + SizeInBytes) % MinRZ) == 0); Type *RightRedZoneTy = ArrayType::get(IRB.getInt8Ty(), RightRedzoneSize); - StructType *NewTy = StructType::get(Ty, RightRedZoneTy, NULL); + StructType *NewTy = StructType::get(Ty, RightRedZoneTy, nullptr); Constant *NewInitializer = ConstantStruct::get( NewTy, G->getInitializer(), - Constant::getNullValue(RightRedZoneTy), NULL); - - GlobalVariable *Name = - createPrivateGlobalForString(M, G->getName(), /*AllowMerging*/true); + Constant::getNullValue(RightRedZoneTy), nullptr); // Create a new global variable with enough space for a redzone. GlobalValue::LinkageTypes Linkage = G->getLinkage(); @@ -1101,8 +1107,13 @@ bool AddressSanitizerModule::InstrumentGlobals(IRBuilder<> &IRB, Module &M) { NewGlobal->takeName(G); G->eraseFromParent(); - bool GlobalHasDynamicInitializer = GlobalsMD.isDynInit(G); - GlobalVariable *SourceLoc = GlobalsMD.getSourceLocation(G); + 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), @@ -1110,12 +1121,9 @@ bool AddressSanitizerModule::InstrumentGlobals(IRBuilder<> &IRB, Module &M) { ConstantInt::get(IntptrTy, SizeInBytes + RightRedzoneSize), ConstantExpr::getPointerCast(Name, IntptrTy), ConstantExpr::getPointerCast(ModuleName, IntptrTy), - ConstantInt::get(IntptrTy, GlobalHasDynamicInitializer), - SourceLoc ? ConstantExpr::getPointerCast(SourceLoc, IntptrTy) - : ConstantInt::get(IntptrTy, 0), - NULL); + ConstantInt::get(IntptrTy, MD.IsDynInit), SourceLoc, nullptr); - if (ClInitializers && GlobalHasDynamicInitializer) + if (ClInitializers && MD.IsDynInit) HasDynamicallyInitializedGlobals = true; DEBUG(dbgs() << "NEW GLOBAL: " << *NewGlobal << "\n"); @@ -1166,13 +1174,6 @@ bool AddressSanitizerModule::runOnModule(Module &M) { 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); @@ -1191,43 +1192,42 @@ void AddressSanitizer::initializeCallbacks(Module &M) { AsanErrorCallback[AccessIsWrite][AccessSizeIndex] = checkInterfaceFunction( M.getOrInsertFunction(kAsanReportErrorTemplate + Suffix, - IRB.getVoidTy(), IntptrTy, NULL)); + IRB.getVoidTy(), IntptrTy, nullptr)); AsanMemoryAccessCallback[AccessIsWrite][AccessSizeIndex] = checkInterfaceFunction( M.getOrInsertFunction(ClMemoryAccessCallbackPrefix + Suffix, - IRB.getVoidTy(), IntptrTy, NULL)); + IRB.getVoidTy(), IntptrTy, nullptr)); } } AsanErrorCallbackSized[0] = checkInterfaceFunction(M.getOrInsertFunction( - kAsanReportLoadN, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + kAsanReportLoadN, IRB.getVoidTy(), IntptrTy, IntptrTy, nullptr)); AsanErrorCallbackSized[1] = checkInterfaceFunction(M.getOrInsertFunction( - kAsanReportStoreN, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + kAsanReportStoreN, IRB.getVoidTy(), IntptrTy, IntptrTy, nullptr)); AsanMemoryAccessCallbackSized[0] = checkInterfaceFunction( M.getOrInsertFunction(ClMemoryAccessCallbackPrefix + "loadN", - IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + IRB.getVoidTy(), IntptrTy, IntptrTy, nullptr)); AsanMemoryAccessCallbackSized[1] = checkInterfaceFunction( M.getOrInsertFunction(ClMemoryAccessCallbackPrefix + "storeN", - IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + IRB.getVoidTy(), IntptrTy, IntptrTy, nullptr)); AsanMemmove = checkInterfaceFunction(M.getOrInsertFunction( ClMemoryAccessCallbackPrefix + "memmove", IRB.getInt8PtrTy(), - IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IntptrTy, NULL)); + IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IntptrTy, nullptr)); AsanMemcpy = checkInterfaceFunction(M.getOrInsertFunction( ClMemoryAccessCallbackPrefix + "memcpy", IRB.getInt8PtrTy(), - IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IntptrTy, NULL)); + IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IntptrTy, nullptr)); AsanMemset = checkInterfaceFunction(M.getOrInsertFunction( ClMemoryAccessCallbackPrefix + "memset", IRB.getInt8PtrTy(), - IRB.getInt8PtrTy(), IRB.getInt32Ty(), IntptrTy, NULL)); + IRB.getInt8PtrTy(), IRB.getInt32Ty(), IntptrTy, nullptr)); AsanHandleNoReturnFunc = checkInterfaceFunction( - M.getOrInsertFunction(kAsanHandleNoReturnName, IRB.getVoidTy(), NULL)); - AsanCovFunction = checkInterfaceFunction(M.getOrInsertFunction( - kAsanCovName, IRB.getVoidTy(), NULL)); + M.getOrInsertFunction(kAsanHandleNoReturnName, IRB.getVoidTy(), nullptr)); + AsanPtrCmpFunction = checkInterfaceFunction(M.getOrInsertFunction( - kAsanPtrCmp, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + kAsanPtrCmp, IRB.getVoidTy(), IntptrTy, IntptrTy, nullptr)); AsanPtrSubFunction = checkInterfaceFunction(M.getOrInsertFunction( - kAsanPtrSub, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + kAsanPtrSub, IRB.getVoidTy(), IntptrTy, IntptrTy, nullptr)); // We insert an empty inline asm after __asan_report* to avoid callback merge. EmptyAsm = InlineAsm::get(FunctionType::get(IRB.getVoidTy(), false), StringRef(""), StringRef(""), @@ -1255,7 +1255,7 @@ bool AddressSanitizer::doInitialization(Module &M) { // call __asan_init in the module ctor. IRBuilder<> IRB(ReturnInst::Create(*C, AsanCtorBB)); AsanInitFunction = checkInterfaceFunction( - M.getOrInsertFunction(kAsanInitName, IRB.getVoidTy(), NULL)); + M.getOrInsertFunction(kAsanInitName, IRB.getVoidTy(), nullptr)); AsanInitFunction->setLinkage(Function::ExternalLinkage); IRB.CreateCall(AsanInitFunction); @@ -1281,74 +1281,6 @@ 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<AllocaInst>(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); - // 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. - 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(&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<BasicBlock *> 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 (&F == AsanCtorFunction) return false; if (F.getLinkage() == GlobalValue::AvailableExternallyLinkage) return false; @@ -1385,7 +1317,7 @@ bool AddressSanitizer::runOnFunction(Function &F) { if (Value *Addr = isInterestingMemoryAccess(&Inst, &IsWrite, &Alignment)) { if (ClOpt && ClOptSameTemp) { - if (!TempsToInstrument.insert(Addr)) + if (!TempsToInstrument.insert(Addr).second) continue; // We've seen this temp in the current BB. } } else if (ClInvalidPointerPairs && @@ -1459,9 +1391,6 @@ bool AddressSanitizer::runOnFunction(Function &F) { bool res = NumInstrumented > 0 || ChangedStack || !NoReturnCalls.empty(); - if (InjectCoverage(F, AllBlocks)) - res = true; - DEBUG(dbgs() << "ASAN done instrumenting: " << res << " " << F << "\n"); if (ClKeepUninstrumented) { @@ -1499,19 +1428,21 @@ void FunctionStackPoisoner::initializeCallbacks(Module &M) { std::string Suffix = itostr(i); AsanStackMallocFunc[i] = checkInterfaceFunction( M.getOrInsertFunction(kAsanStackMallocNameTemplate + Suffix, IntptrTy, - IntptrTy, IntptrTy, NULL)); + IntptrTy, IntptrTy, nullptr)); AsanStackFreeFunc[i] = checkInterfaceFunction(M.getOrInsertFunction( kAsanStackFreeNameTemplate + Suffix, IRB.getVoidTy(), IntptrTy, - IntptrTy, IntptrTy, NULL)); + IntptrTy, IntptrTy, nullptr)); } - AsanPoisonStackMemoryFunc = checkInterfaceFunction(M.getOrInsertFunction( - kAsanPoisonStackMemoryName, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); - AsanUnpoisonStackMemoryFunc = checkInterfaceFunction(M.getOrInsertFunction( - kAsanUnpoisonStackMemoryName, IRB.getVoidTy(), IntptrTy, IntptrTy, NULL)); + AsanPoisonStackMemoryFunc = checkInterfaceFunction( + M.getOrInsertFunction(kAsanPoisonStackMemoryName, IRB.getVoidTy(), + IntptrTy, IntptrTy, nullptr)); + AsanUnpoisonStackMemoryFunc = checkInterfaceFunction( + M.getOrInsertFunction(kAsanUnpoisonStackMemoryName, IRB.getVoidTy(), + IntptrTy, IntptrTy, nullptr)); } void -FunctionStackPoisoner::poisonRedZones(const ArrayRef<uint8_t> ShadowBytes, +FunctionStackPoisoner::poisonRedZones(ArrayRef<uint8_t> ShadowBytes, IRBuilder<> &IRB, Value *ShadowBase, bool DoPoison) { size_t n = ShadowBytes.size(); diff --git a/lib/Transforms/Instrumentation/Android.mk b/lib/Transforms/Instrumentation/Android.mk index f9a55c7..1f21028 100644 --- a/lib/Transforms/Instrumentation/Android.mk +++ b/lib/Transforms/Instrumentation/Android.mk @@ -8,6 +8,7 @@ instrumentation_SRC_FILES := \ GCOVProfiling.cpp \ Instrumentation.cpp \ MemorySanitizer.cpp \ + SanitizerCoverage.cpp \ ThreadSanitizer.cpp # For the host diff --git a/lib/Transforms/Instrumentation/CMakeLists.txt b/lib/Transforms/Instrumentation/CMakeLists.txt index 3563593..139e514 100644 --- a/lib/Transforms/Instrumentation/CMakeLists.txt +++ b/lib/Transforms/Instrumentation/CMakeLists.txt @@ -6,6 +6,7 @@ add_llvm_library(LLVMInstrumentation GCOVProfiling.cpp MemorySanitizer.cpp Instrumentation.cpp + SanitizerCoverage.cpp ThreadSanitizer.cpp ) diff --git a/lib/Transforms/Instrumentation/DataFlowSanitizer.cpp b/lib/Transforms/Instrumentation/DataFlowSanitizer.cpp index 799e14b..c5a4860 100644 --- a/lib/Transforms/Instrumentation/DataFlowSanitizer.cpp +++ b/lib/Transforms/Instrumentation/DataFlowSanitizer.cpp @@ -50,6 +50,8 @@ #include "llvm/ADT/DepthFirstIterator.h" #include "llvm/ADT/StringExtras.h" #include "llvm/Analysis/ValueTracking.h" +#include "llvm/IR/Dominators.h" +#include "llvm/IR/DebugInfo.h" #include "llvm/IR/IRBuilder.h" #include "llvm/IR/InlineAsm.h" #include "llvm/IR/InstVisitor.h" @@ -62,7 +64,10 @@ #include "llvm/Support/SpecialCaseList.h" #include "llvm/Transforms/Utils/BasicBlockUtils.h" #include "llvm/Transforms/Utils/Local.h" +#include <algorithm> #include <iterator> +#include <set> +#include <utility> using namespace llvm; @@ -135,11 +140,11 @@ class DFSanABIList { std::unique_ptr<SpecialCaseList> SCL; public: - DFSanABIList(SpecialCaseList *SCL) : SCL(SCL) {} + DFSanABIList(std::unique_ptr<SpecialCaseList> SCL) : SCL(std::move(SCL)) {} /// Returns whether either this function or its source file are listed in the /// given category. - bool isIn(const Function &F, const StringRef Category) const { + bool isIn(const Function &F, StringRef Category) const { return isIn(*F.getParent(), Category) || SCL->inSection("fun", F.getName(), Category); } @@ -148,7 +153,7 @@ class DFSanABIList { /// /// If GA aliases a function, the alias's name is matched as a function name /// would be. Similarly, aliases of globals are matched like globals. - bool isIn(const GlobalAlias &GA, const StringRef Category) const { + bool isIn(const GlobalAlias &GA, StringRef Category) const { if (isIn(*GA.getParent(), Category)) return true; @@ -160,7 +165,7 @@ class DFSanABIList { } /// Returns whether this module is listed in the given category. - bool isIn(const Module &M, const StringRef Category) const { + bool isIn(const Module &M, StringRef Category) const { return SCL->inSection("src", M.getModuleIdentifier(), Category); } }; @@ -229,18 +234,21 @@ class DataFlowSanitizer : public ModulePass { FunctionType *DFSanUnimplementedFnTy; FunctionType *DFSanSetLabelFnTy; FunctionType *DFSanNonzeroLabelFnTy; + FunctionType *DFSanVarargWrapperFnTy; Constant *DFSanUnionFn; + Constant *DFSanCheckedUnionFn; Constant *DFSanUnionLoadFn; Constant *DFSanUnimplementedFn; Constant *DFSanSetLabelFn; Constant *DFSanNonzeroLabelFn; + Constant *DFSanVarargWrapperFn; MDNode *ColdCallWeights; DFSanABIList ABIList; DenseMap<Value *, Function *> UnwrappedFnMap; AttributeSet ReadOnlyNoneAttrs; + DenseMap<const Function *, DISubprogram> FunctionDIs; Value *getShadowAddress(Value *Addr, Instruction *Pos); - Value *combineShadows(Value *V1, Value *V2, Instruction *Pos); bool isInstrumented(const Function *F); bool isInstrumented(const GlobalAlias *GA); FunctionType *getArgsFunctionType(FunctionType *T); @@ -266,6 +274,7 @@ class DataFlowSanitizer : public ModulePass { struct DFSanFunction { DataFlowSanitizer &DFS; Function *F; + DominatorTree DT; DataFlowSanitizer::InstrumentedABI IA; bool IsNativeABI; Value *ArgTLSPtr; @@ -275,17 +284,32 @@ struct DFSanFunction { DenseMap<AllocaInst *, AllocaInst *> AllocaShadowMap; std::vector<std::pair<PHINode *, PHINode *> > PHIFixups; DenseSet<Instruction *> SkipInsts; - DenseSet<Value *> NonZeroChecks; + std::vector<Value *> NonZeroChecks; + bool AvoidNewBlocks; + + struct CachedCombinedShadow { + BasicBlock *Block; + Value *Shadow; + }; + DenseMap<std::pair<Value *, Value *>, CachedCombinedShadow> + CachedCombinedShadows; + DenseMap<Value *, std::set<Value *>> ShadowElements; DFSanFunction(DataFlowSanitizer &DFS, Function *F, bool IsNativeABI) : DFS(DFS), F(F), IA(DFS.getInstrumentedABI()), IsNativeABI(IsNativeABI), ArgTLSPtr(nullptr), RetvalTLSPtr(nullptr), - LabelReturnAlloca(nullptr) {} + LabelReturnAlloca(nullptr) { + DT.recalculate(*F); + // FIXME: Need to track down the register allocator issue which causes poor + // performance in pathological cases with large numbers of basic blocks. + AvoidNewBlocks = F->size() > 1000; + } Value *getArgTLSPtr(); Value *getArgTLS(unsigned Index, Instruction *Pos); Value *getRetvalTLS(); Value *getShadow(Value *V); void setShadow(Instruction *I, Value *Shadow); + Value *combineShadows(Value *V1, Value *V2, Instruction *Pos); Value *combineOperandShadows(Instruction *Inst); Value *loadShadow(Value *ShadowAddr, uint64_t Size, uint64_t Align, Instruction *Pos); @@ -367,7 +391,6 @@ FunctionType *DataFlowSanitizer::getTrampolineFunctionType(FunctionType *T) { } FunctionType *DataFlowSanitizer::getCustomFunctionType(FunctionType *T) { - assert(!T->isVarArg()); llvm::SmallVector<Type *, 4> ArgTypes; for (FunctionType::param_iterator i = T->param_begin(), e = T->param_end(); i != e; ++i) { @@ -382,10 +405,12 @@ FunctionType *DataFlowSanitizer::getCustomFunctionType(FunctionType *T) { } for (unsigned i = 0, e = T->getNumParams(); i != e; ++i) ArgTypes.push_back(ShadowTy); + if (T->isVarArg()) + ArgTypes.push_back(ShadowPtrTy); Type *RetType = T->getReturnType(); if (!RetType->isVoidTy()) ArgTypes.push_back(ShadowPtrTy); - return FunctionType::get(T->getReturnType(), ArgTypes, false); + return FunctionType::get(T->getReturnType(), ArgTypes, T->isVarArg()); } bool DataFlowSanitizer::doInitialization(Module &M) { @@ -415,7 +440,9 @@ bool DataFlowSanitizer::doInitialization(Module &M) { DFSanSetLabelFnTy = FunctionType::get(Type::getVoidTy(*Ctx), DFSanSetLabelArgs, /*isVarArg=*/false); DFSanNonzeroLabelFnTy = FunctionType::get( - Type::getVoidTy(*Ctx), ArrayRef<Type *>(), /*isVarArg=*/false); + Type::getVoidTy(*Ctx), None, /*isVarArg=*/false); + DFSanVarargWrapperFnTy = FunctionType::get( + Type::getVoidTy(*Ctx), Type::getInt8PtrTy(*Ctx), /*isVarArg=*/false); if (GetArgTLSPtr) { Type *ArgTLSTy = ArrayType::get(ShadowTy, 64); @@ -495,15 +522,26 @@ DataFlowSanitizer::buildWrapperFunction(Function *F, StringRef NewFName, AttributeSet::ReturnIndex)); BasicBlock *BB = BasicBlock::Create(*Ctx, "entry", NewF); - std::vector<Value *> Args; - unsigned n = FT->getNumParams(); - for (Function::arg_iterator ai = NewF->arg_begin(); n != 0; ++ai, --n) - Args.push_back(&*ai); - CallInst *CI = CallInst::Create(F, Args, "", BB); - if (FT->getReturnType()->isVoidTy()) - ReturnInst::Create(*Ctx, BB); - else - ReturnInst::Create(*Ctx, CI, BB); + if (F->isVarArg()) { + NewF->removeAttributes( + AttributeSet::FunctionIndex, + AttributeSet().addAttribute(*Ctx, AttributeSet::FunctionIndex, + "split-stack")); + CallInst::Create(DFSanVarargWrapperFn, + IRBuilder<>(BB).CreateGlobalStringPtr(F->getName()), "", + BB); + new UnreachableInst(*Ctx, BB); + } else { + std::vector<Value *> Args; + unsigned n = FT->getNumParams(); + for (Function::arg_iterator ai = NewF->arg_begin(); n != 0; ++ai, --n) + Args.push_back(&*ai); + CallInst *CI = CallInst::Create(F, Args, "", BB); + if (FT->getReturnType()->isVoidTy()) + ReturnInst::Create(*Ctx, BB); + else + ReturnInst::Create(*Ctx, CI, BB); + } return NewF; } @@ -548,6 +586,8 @@ bool DataFlowSanitizer::runOnModule(Module &M) { if (ABIList.isIn(M, "skip")) return false; + FunctionDIs = makeSubprogramMap(M); + if (!GetArgTLSPtr) { Type *ArgTLSTy = ArrayType::get(ShadowTy, 64); ArgTLS = Mod->getOrInsertGlobal("__dfsan_arg_tls", ArgTLSTy); @@ -562,6 +602,15 @@ bool DataFlowSanitizer::runOnModule(Module &M) { DFSanUnionFn = Mod->getOrInsertFunction("__dfsan_union", DFSanUnionFnTy); if (Function *F = dyn_cast<Function>(DFSanUnionFn)) { + F->addAttribute(AttributeSet::FunctionIndex, Attribute::NoUnwind); + F->addAttribute(AttributeSet::FunctionIndex, Attribute::ReadNone); + F->addAttribute(AttributeSet::ReturnIndex, Attribute::ZExt); + F->addAttribute(1, Attribute::ZExt); + F->addAttribute(2, Attribute::ZExt); + } + DFSanCheckedUnionFn = Mod->getOrInsertFunction("dfsan_union", DFSanUnionFnTy); + if (Function *F = dyn_cast<Function>(DFSanCheckedUnionFn)) { + F->addAttribute(AttributeSet::FunctionIndex, Attribute::NoUnwind); F->addAttribute(AttributeSet::FunctionIndex, Attribute::ReadNone); F->addAttribute(AttributeSet::ReturnIndex, Attribute::ZExt); F->addAttribute(1, Attribute::ZExt); @@ -570,6 +619,7 @@ bool DataFlowSanitizer::runOnModule(Module &M) { DFSanUnionLoadFn = Mod->getOrInsertFunction("__dfsan_union_load", DFSanUnionLoadFnTy); if (Function *F = dyn_cast<Function>(DFSanUnionLoadFn)) { + F->addAttribute(AttributeSet::FunctionIndex, Attribute::NoUnwind); F->addAttribute(AttributeSet::FunctionIndex, Attribute::ReadOnly); F->addAttribute(AttributeSet::ReturnIndex, Attribute::ZExt); } @@ -582,16 +632,20 @@ bool DataFlowSanitizer::runOnModule(Module &M) { } DFSanNonzeroLabelFn = Mod->getOrInsertFunction("__dfsan_nonzero_label", DFSanNonzeroLabelFnTy); + DFSanVarargWrapperFn = Mod->getOrInsertFunction("__dfsan_vararg_wrapper", + DFSanVarargWrapperFnTy); std::vector<Function *> FnsToInstrument; llvm::SmallPtrSet<Function *, 2> FnsWithNativeABI; for (Module::iterator i = M.begin(), e = M.end(); i != e; ++i) { if (!i->isIntrinsic() && i != DFSanUnionFn && + i != DFSanCheckedUnionFn && i != DFSanUnionLoadFn && i != DFSanUnimplementedFn && i != DFSanSetLabelFn && - i != DFSanNonzeroLabelFn) + i != DFSanNonzeroLabelFn && + i != DFSanVarargWrapperFn) FnsToInstrument.push_back(&*i); } @@ -673,11 +727,6 @@ bool DataFlowSanitizer::runOnModule(Module &M) { } else { addGlobalNamePrefix(&F); } - // Hopefully, nobody will try to indirectly call a vararg - // function... yet. - } else if (FT->isVarArg()) { - UnwrappedFnMap[&F] = &F; - *i = nullptr; } else if (!IsZeroArgsVoidRet || getWrapperKind(&F) == WK_Custom) { // Build a wrapper function for F. The wrapper simply calls F, and is // added to FnsToInstrument so that any instrumentation according to its @@ -694,6 +743,12 @@ bool DataFlowSanitizer::runOnModule(Module &M) { Value *WrappedFnCst = ConstantExpr::getBitCast(NewF, PointerType::getUnqual(FT)); F.replaceAllUsesWith(WrappedFnCst); + + // Patch the pointer to LLVM function in debug info descriptor. + auto DI = FunctionDIs.find(&F); + if (DI != FunctionDIs.end()) + DI->second.replaceFunction(&F); + UnwrappedFnMap[WrappedFnCst] = &F; *i = NewF; @@ -713,6 +768,11 @@ bool DataFlowSanitizer::runOnModule(Module &M) { i = FnsToInstrument.begin() + N; e = FnsToInstrument.begin() + Count; } + // Hopefully, nobody will try to indirectly call a vararg + // function... yet. + } else if (FT->isVarArg()) { + UnwrappedFnMap[&F] = &F; + *i = nullptr; } } @@ -771,18 +831,16 @@ bool DataFlowSanitizer::runOnModule(Module &M) { // yet). To make our life easier, do this work in a pass after the main // instrumentation. if (ClDebugNonzeroLabels) { - for (DenseSet<Value *>::iterator i = DFSF.NonZeroChecks.begin(), - e = DFSF.NonZeroChecks.end(); - i != e; ++i) { + for (Value *V : DFSF.NonZeroChecks) { Instruction *Pos; - if (Instruction *I = dyn_cast<Instruction>(*i)) + if (Instruction *I = dyn_cast<Instruction>(V)) Pos = I->getNextNode(); else Pos = DFSF.F->getEntryBlock().begin(); while (isa<PHINode>(Pos) || isa<AllocaInst>(Pos)) Pos = Pos->getNextNode(); IRBuilder<> IRB(Pos); - Value *Ne = IRB.CreateICmpNE(*i, DFSF.DFS.ZeroShadow); + Value *Ne = IRB.CreateICmpNE(V, DFSF.DFS.ZeroShadow); BranchInst *BI = cast<BranchInst>(SplitBlockAndInsertIfThen( Ne, Pos, /*Unreachable=*/false, ColdCallWeights)); IRBuilder<> ThenIRB(BI); @@ -847,7 +905,7 @@ Value *DFSanFunction::getShadow(Value *V) { break; } } - NonZeroChecks.insert(Shadow); + NonZeroChecks.push_back(Shadow); } else { Shadow = DFS.ZeroShadow; } @@ -873,30 +931,82 @@ Value *DataFlowSanitizer::getShadowAddress(Value *Addr, Instruction *Pos) { // Generates IR to compute the union of the two given shadows, inserting it // before Pos. Returns the computed union Value. -Value *DataFlowSanitizer::combineShadows(Value *V1, Value *V2, - Instruction *Pos) { - if (V1 == ZeroShadow) +Value *DFSanFunction::combineShadows(Value *V1, Value *V2, Instruction *Pos) { + if (V1 == DFS.ZeroShadow) return V2; - if (V2 == ZeroShadow) + if (V2 == DFS.ZeroShadow) return V1; if (V1 == V2) return V1; + + auto V1Elems = ShadowElements.find(V1); + auto V2Elems = ShadowElements.find(V2); + if (V1Elems != ShadowElements.end() && V2Elems != ShadowElements.end()) { + if (std::includes(V1Elems->second.begin(), V1Elems->second.end(), + V2Elems->second.begin(), V2Elems->second.end())) { + return V1; + } else if (std::includes(V2Elems->second.begin(), V2Elems->second.end(), + V1Elems->second.begin(), V1Elems->second.end())) { + return V2; + } + } else if (V1Elems != ShadowElements.end()) { + if (V1Elems->second.count(V2)) + return V1; + } else if (V2Elems != ShadowElements.end()) { + if (V2Elems->second.count(V1)) + return V2; + } + + auto Key = std::make_pair(V1, V2); + if (V1 > V2) + std::swap(Key.first, Key.second); + CachedCombinedShadow &CCS = CachedCombinedShadows[Key]; + if (CCS.Block && DT.dominates(CCS.Block, Pos->getParent())) + return CCS.Shadow; + IRBuilder<> IRB(Pos); - BasicBlock *Head = Pos->getParent(); - Value *Ne = IRB.CreateICmpNE(V1, V2); - BranchInst *BI = cast<BranchInst>(SplitBlockAndInsertIfThen( - Ne, Pos, /*Unreachable=*/false, ColdCallWeights)); - IRBuilder<> ThenIRB(BI); - CallInst *Call = ThenIRB.CreateCall2(DFSanUnionFn, V1, V2); - Call->addAttribute(AttributeSet::ReturnIndex, Attribute::ZExt); - Call->addAttribute(1, Attribute::ZExt); - Call->addAttribute(2, Attribute::ZExt); - - BasicBlock *Tail = BI->getSuccessor(0); - PHINode *Phi = PHINode::Create(ShadowTy, 2, "", Tail->begin()); - Phi->addIncoming(Call, Call->getParent()); - Phi->addIncoming(V1, Head); - return Phi; + if (AvoidNewBlocks) { + CallInst *Call = IRB.CreateCall2(DFS.DFSanCheckedUnionFn, V1, V2); + Call->addAttribute(AttributeSet::ReturnIndex, Attribute::ZExt); + Call->addAttribute(1, Attribute::ZExt); + Call->addAttribute(2, Attribute::ZExt); + + CCS.Block = Pos->getParent(); + CCS.Shadow = Call; + } else { + BasicBlock *Head = Pos->getParent(); + Value *Ne = IRB.CreateICmpNE(V1, V2); + BranchInst *BI = cast<BranchInst>(SplitBlockAndInsertIfThen( + Ne, Pos, /*Unreachable=*/false, DFS.ColdCallWeights, &DT)); + IRBuilder<> ThenIRB(BI); + CallInst *Call = ThenIRB.CreateCall2(DFS.DFSanUnionFn, V1, V2); + Call->addAttribute(AttributeSet::ReturnIndex, Attribute::ZExt); + Call->addAttribute(1, Attribute::ZExt); + Call->addAttribute(2, Attribute::ZExt); + + BasicBlock *Tail = BI->getSuccessor(0); + PHINode *Phi = PHINode::Create(DFS.ShadowTy, 2, "", Tail->begin()); + Phi->addIncoming(Call, Call->getParent()); + Phi->addIncoming(V1, Head); + + CCS.Block = Tail; + CCS.Shadow = Phi; + } + + std::set<Value *> UnionElems; + if (V1Elems != ShadowElements.end()) { + UnionElems = V1Elems->second; + } else { + UnionElems.insert(V1); + } + if (V2Elems != ShadowElements.end()) { + UnionElems.insert(V2Elems->second.begin(), V2Elems->second.end()); + } else { + UnionElems.insert(V2); + } + ShadowElements[CCS.Shadow] = std::move(UnionElems); + + return CCS.Shadow; } // A convenience function which folds the shadows of each of the operands @@ -908,7 +1018,7 @@ Value *DFSanFunction::combineOperandShadows(Instruction *Inst) { Value *Shadow = getShadow(Inst->getOperand(0)); for (unsigned i = 1, n = Inst->getNumOperands(); i != n; ++i) { - Shadow = DFS.combineShadows(Shadow, getShadow(Inst->getOperand(i)), Inst); + Shadow = combineShadows(Shadow, getShadow(Inst->getOperand(i)), Inst); } return Shadow; } @@ -961,12 +1071,11 @@ Value *DFSanFunction::loadShadow(Value *Addr, uint64_t Size, uint64_t Align, IRBuilder<> IRB(Pos); Value *ShadowAddr1 = IRB.CreateGEP(ShadowAddr, ConstantInt::get(DFS.IntptrTy, 1)); - return DFS.combineShadows(IRB.CreateAlignedLoad(ShadowAddr, ShadowAlign), - IRB.CreateAlignedLoad(ShadowAddr1, ShadowAlign), - Pos); + return combineShadows(IRB.CreateAlignedLoad(ShadowAddr, ShadowAlign), + IRB.CreateAlignedLoad(ShadowAddr1, ShadowAlign), Pos); } } - if (Size % (64 / DFS.ShadowWidth) == 0) { + if (!AvoidNewBlocks && Size % (64 / DFS.ShadowWidth) == 0) { // Fast path for the common case where each byte has identical shadow: load // shadow 64 bits at a time, fall out to a __dfsan_union_load call if any // shadow is non-equal. @@ -990,16 +1099,27 @@ Value *DFSanFunction::loadShadow(Value *Addr, uint64_t Size, uint64_t Align, BasicBlock *Head = Pos->getParent(); BasicBlock *Tail = Head->splitBasicBlock(Pos); + + if (DomTreeNode *OldNode = DT.getNode(Head)) { + std::vector<DomTreeNode *> Children(OldNode->begin(), OldNode->end()); + + DomTreeNode *NewNode = DT.addNewBlock(Tail, Head); + for (auto Child : Children) + DT.changeImmediateDominator(Child, NewNode); + } + // In the following code LastBr will refer to the previous basic block's // conditional branch instruction, whose true successor is fixed up to point // to the next block during the loop below or to the tail after the final // iteration. BranchInst *LastBr = BranchInst::Create(FallbackBB, FallbackBB, ShadowsEq); ReplaceInstWithInst(Head->getTerminator(), LastBr); + DT.addNewBlock(FallbackBB, Head); for (uint64_t Ofs = 64 / DFS.ShadowWidth; Ofs != Size; Ofs += 64 / DFS.ShadowWidth) { BasicBlock *NextBB = BasicBlock::Create(*DFS.Ctx, "", F); + DT.addNewBlock(NextBB, LastBr->getParent()); IRBuilder<> NextIRB(NextBB); WideAddr = NextIRB.CreateGEP(WideAddr, ConstantInt::get(DFS.IntptrTy, 1)); Value *NextWideShadow = NextIRB.CreateAlignedLoad(WideAddr, ShadowAlign); @@ -1025,6 +1145,11 @@ Value *DFSanFunction::loadShadow(Value *Addr, uint64_t Size, uint64_t Align, void DFSanVisitor::visitLoadInst(LoadInst &LI) { uint64_t Size = DFSF.DFS.DL->getTypeStoreSize(LI.getType()); + if (Size == 0) { + DFSF.setShadow(&LI, DFSF.DFS.ZeroShadow); + return; + } + uint64_t Align; if (ClPreserveAlignment) { Align = LI.getAlignment(); @@ -1037,10 +1162,10 @@ void DFSanVisitor::visitLoadInst(LoadInst &LI) { Value *Shadow = DFSF.loadShadow(LI.getPointerOperand(), Size, Align, &LI); if (ClCombinePointerLabelsOnLoad) { Value *PtrShadow = DFSF.getShadow(LI.getPointerOperand()); - Shadow = DFSF.DFS.combineShadows(Shadow, PtrShadow, &LI); + Shadow = DFSF.combineShadows(Shadow, PtrShadow, &LI); } if (Shadow != DFSF.DFS.ZeroShadow) - DFSF.NonZeroChecks.insert(Shadow); + DFSF.NonZeroChecks.push_back(Shadow); DFSF.setShadow(&LI, Shadow); } @@ -1099,6 +1224,9 @@ void DFSanFunction::storeShadow(Value *Addr, uint64_t Size, uint64_t Align, void DFSanVisitor::visitStoreInst(StoreInst &SI) { uint64_t Size = DFSF.DFS.DL->getTypeStoreSize(SI.getValueOperand()->getType()); + if (Size == 0) + return; + uint64_t Align; if (ClPreserveAlignment) { Align = SI.getAlignment(); @@ -1111,7 +1239,7 @@ void DFSanVisitor::visitStoreInst(StoreInst &SI) { Value* Shadow = DFSF.getShadow(SI.getValueOperand()); if (ClCombinePointerLabelsOnStore) { Value *PtrShadow = DFSF.getShadow(SI.getPointerOperand()); - Shadow = DFSF.DFS.combineShadows(Shadow, PtrShadow, &SI); + Shadow = DFSF.combineShadows(Shadow, PtrShadow, &SI); } DFSF.storeShadow(SI.getPointerOperand(), Size, Align, Shadow, &SI); } @@ -1176,9 +1304,9 @@ void DFSanVisitor::visitSelectInst(SelectInst &I) { if (isa<VectorType>(I.getCondition()->getType())) { DFSF.setShadow( - &I, DFSF.DFS.combineShadows( - CondShadow, - DFSF.DFS.combineShadows(TrueShadow, FalseShadow, &I), &I)); + &I, + DFSF.combineShadows( + CondShadow, DFSF.combineShadows(TrueShadow, FalseShadow, &I), &I)); } else { Value *ShadowSel; if (TrueShadow == FalseShadow) { @@ -1187,7 +1315,7 @@ void DFSanVisitor::visitSelectInst(SelectInst &I) { ShadowSel = SelectInst::Create(I.getCondition(), TrueShadow, FalseShadow, "", &I); } - DFSF.setShadow(&I, DFSF.DFS.combineShadows(CondShadow, ShadowSel, &I)); + DFSF.setShadow(&I, DFSF.combineShadows(CondShadow, ShadowSel, &I)); } } @@ -1253,6 +1381,15 @@ void DFSanVisitor::visitCallSite(CallSite CS) { return; } + // Calls to this function are synthesized in wrappers, and we shouldn't + // instrument them. + if (F == DFSF.DFS.DFSanVarargWrapperFn) + return; + + assert(!(cast<FunctionType>( + CS.getCalledValue()->getType()->getPointerElementType())->isVarArg() && + dyn_cast<InvokeInst>(CS.getInstruction()))); + IRBuilder<> IRB(CS.getInstruction()); DenseMap<Value *, Function *>::iterator i = @@ -1324,6 +1461,20 @@ void DFSanVisitor::visitCallSite(CallSite CS) { for (unsigned n = FT->getNumParams(); n != 0; ++i, --n) Args.push_back(DFSF.getShadow(*i)); + if (FT->isVarArg()) { + auto LabelVAAlloca = + new AllocaInst(ArrayType::get(DFSF.DFS.ShadowTy, + CS.arg_size() - FT->getNumParams()), + "labelva", DFSF.F->getEntryBlock().begin()); + + for (unsigned n = 0; i != CS.arg_end(); ++i, ++n) { + auto LabelVAPtr = IRB.CreateStructGEP(LabelVAAlloca, n); + IRB.CreateStore(DFSF.getShadow(*i), LabelVAPtr); + } + + Args.push_back(IRB.CreateStructGEP(LabelVAAlloca, 0)); + } + if (!FT->getReturnType()->isVoidTy()) { if (!DFSF.LabelReturnAlloca) { DFSF.LabelReturnAlloca = @@ -1333,6 +1484,9 @@ void DFSanVisitor::visitCallSite(CallSite CS) { Args.push_back(DFSF.LabelReturnAlloca); } + for (i = CS.arg_begin() + FT->getNumParams(); i != CS.arg_end(); ++i) + Args.push_back(*i); + CallInst *CustomCI = IRB.CreateCall(CustomF, Args); CustomCI->setCallingConv(CI->getCallingConv()); CustomCI->setAttributes(CI->getAttributes()); @@ -1379,7 +1533,7 @@ void DFSanVisitor::visitCallSite(CallSite CS) { LoadInst *LI = NextIRB.CreateLoad(DFSF.getRetvalTLS()); DFSF.SkipInsts.insert(LI); DFSF.setShadow(CS.getInstruction(), LI); - DFSF.NonZeroChecks.insert(LI); + DFSF.NonZeroChecks.push_back(LI); } } @@ -1433,7 +1587,7 @@ void DFSanVisitor::visitCallSite(CallSite CS) { ExtractValueInst::Create(NewCS.getInstruction(), 1, "", Next); DFSF.SkipInsts.insert(ExShadow); DFSF.setShadow(ExVal, ExShadow); - DFSF.NonZeroChecks.insert(ExShadow); + DFSF.NonZeroChecks.push_back(ExShadow); CS.getInstruction()->replaceAllUsesWith(ExVal); } diff --git a/lib/Transforms/Instrumentation/DebugIR.cpp b/lib/Transforms/Instrumentation/DebugIR.cpp index f2f1738..5234341 100644 --- a/lib/Transforms/Instrumentation/DebugIR.cpp +++ b/lib/Transforms/Instrumentation/DebugIR.cpp @@ -396,7 +396,7 @@ private: Elements.push_back(getOrCreateType(T->getStructElementType(i))); // set struct elements - StructDescriptor.setTypeArray(Builder.getOrCreateArray(Elements)); + StructDescriptor.setArrays(Builder.getOrCreateArray(Elements)); } else if (T->isPointerTy()) { Type *PointeeTy = T->getPointerElementType(); if (!(N = getType(PointeeTy))) @@ -440,7 +440,7 @@ private: Params.push_back(getOrCreateType(T)); } - DIArray ParamArray = Builder.getOrCreateArray(Params); + DITypeArray ParamArray = Builder.getOrCreateTypeArray(Params); return Builder.createSubroutineType(DIFile(FileNode), ParamArray); } @@ -525,11 +525,11 @@ std::string DebugIR::getPath() { void DebugIR::writeDebugBitcode(const Module *M, int *fd) { std::unique_ptr<raw_fd_ostream> Out; - std::string error; + std::error_code EC; if (!fd) { std::string Path = getPath(); - Out.reset(new raw_fd_ostream(Path.c_str(), error, sys::fs::F_Text)); + Out.reset(new raw_fd_ostream(Path, EC, sys::fs::F_Text)); DEBUG(dbgs() << "WRITING debug bitcode from Module " << M << " to file " << Path << "\n"); } else { diff --git a/lib/Transforms/Instrumentation/DebugIR.h b/lib/Transforms/Instrumentation/DebugIR.h index 02831ed..8d74a4d 100644 --- a/lib/Transforms/Instrumentation/DebugIR.h +++ b/lib/Transforms/Instrumentation/DebugIR.h @@ -13,8 +13,8 @@ // //===----------------------------------------------------------------------===// -#ifndef LLVM_TRANSFORMS_INSTRUMENTATION_DEBUGIR_H -#define LLVM_TRANSFORMS_INSTRUMENTATION_DEBUGIR_H +#ifndef LLVM_LIB_TRANSFORMS_INSTRUMENTATION_DEBUGIR_H +#define LLVM_LIB_TRANSFORMS_INSTRUMENTATION_DEBUGIR_H #include "llvm/Pass.h" @@ -95,4 +95,4 @@ private: } // llvm namespace -#endif // LLVM_TRANSFORMS_INSTRUMENTATION_DEBUGIR_H +#endif diff --git a/lib/Transforms/Instrumentation/GCOVProfiling.cpp b/lib/Transforms/Instrumentation/GCOVProfiling.cpp index cfeb62e..220d7f8 100644 --- a/lib/Transforms/Instrumentation/GCOVProfiling.cpp +++ b/lib/Transforms/Instrumentation/GCOVProfiling.cpp @@ -480,12 +480,12 @@ void GCOVProfiler::emitProfileNotes() { // LTO, we'll generate the same .gcno files. DICompileUnit CU(CU_Nodes->getOperand(i)); - std::string ErrorInfo; - raw_fd_ostream out(mangleName(CU, "gcno").c_str(), ErrorInfo, - sys::fs::F_None); + std::error_code EC; + raw_fd_ostream out(mangleName(CU, "gcno"), EC, sys::fs::F_None); std::string EdgeDestinations; DIArray SPs = CU.getSubprograms(); + unsigned FunctionIdent = 0; for (unsigned i = 0, e = SPs.getNumElements(); i != e; ++i) { DISubprogram SP(SPs.getElement(i)); assert((!SP || SP.isSubprogram()) && @@ -505,8 +505,8 @@ void GCOVProfiler::emitProfileNotes() { ++It; EntryBlock.splitBasicBlock(It); - Funcs.push_back( - make_unique<GCOVFunction>(SP, &out, i, Options.UseCfgChecksum)); + Funcs.push_back(make_unique<GCOVFunction>(SP, &out, FunctionIdent++, + Options.UseCfgChecksum)); GCOVFunction &Func = *Funcs.back(); for (Function::iterator BB = F->begin(), E = F->end(); BB != E; ++BB) { @@ -738,11 +738,11 @@ GlobalVariable *GCOVProfiler::buildEdgeLookupTable( Edge += Successors; } - ArrayRef<Constant*> V(&EdgeTable[0], TableSize); GlobalVariable *EdgeTableGV = new GlobalVariable( *M, EdgeTableTy, true, GlobalValue::InternalLinkage, - ConstantArray::get(EdgeTableTy, V), + ConstantArray::get(EdgeTableTy, + makeArrayRef(&EdgeTable[0],TableSize)), "__llvm_gcda_edge_table"); EdgeTableGV->setUnnamedAddr(true); return EdgeTableGV; diff --git a/lib/Transforms/Instrumentation/Instrumentation.cpp b/lib/Transforms/Instrumentation/Instrumentation.cpp index ac1dd43..8e95367 100644 --- a/lib/Transforms/Instrumentation/Instrumentation.cpp +++ b/lib/Transforms/Instrumentation/Instrumentation.cpp @@ -27,6 +27,7 @@ void llvm::initializeInstrumentation(PassRegistry &Registry) { initializeGCOVProfilerPass(Registry); initializeMemorySanitizerPass(Registry); initializeThreadSanitizerPass(Registry); + initializeSanitizerCoverageModulePass(Registry); initializeDataFlowSanitizerPass(Registry); } diff --git a/lib/Transforms/Instrumentation/MemorySanitizer.cpp b/lib/Transforms/Instrumentation/MemorySanitizer.cpp index 496ab48..1261259 100644 --- a/lib/Transforms/Instrumentation/MemorySanitizer.cpp +++ b/lib/Transforms/Instrumentation/MemorySanitizer.cpp @@ -127,6 +127,10 @@ static const uint64_t kOriginOffset64 = 1ULL << 45; static const unsigned kMinOriginAlignment = 4; static const unsigned kShadowTLSAlignment = 8; +// These constants must be kept in sync with the ones in msan.h. +static const unsigned kParamTLSSize = 800; +static const unsigned kRetvalTLSSize = 800; + // Accesses sizes are powers of two: 1, 2, 4, 8. static const size_t kNumberOfAccessSizes = 4; @@ -195,6 +199,13 @@ static cl::opt<bool> ClWrapIndirectCallsFast("msan-wrap-indirect-calls-fast", cl::desc("Do not wrap indirect calls with target in the same module"), cl::Hidden, cl::init(true)); +// This is an experiment to enable handling of cases where shadow is a non-zero +// compile-time constant. For some unexplainable reason they were silently +// ignored in the instrumentation. +static cl::opt<bool> ClCheckConstantShadow("msan-check-constant-shadow", + cl::desc("Insert checks for constant shadow values"), + cl::Hidden, cl::init(false)); + namespace { /// \brief An instrumentation pass implementing detection of uninitialized @@ -321,7 +332,7 @@ void MemorySanitizer::initializeCallbacks(Module &M) { // which is not yet implemented. StringRef WarningFnName = ClKeepGoing ? "__msan_warning" : "__msan_warning_noreturn"; - WarningFn = M.getOrInsertFunction(WarningFnName, IRB.getVoidTy(), NULL); + WarningFn = M.getOrInsertFunction(WarningFnName, IRB.getVoidTy(), nullptr); for (size_t AccessSizeIndex = 0; AccessSizeIndex < kNumberOfAccessSizes; AccessSizeIndex++) { @@ -329,34 +340,35 @@ void MemorySanitizer::initializeCallbacks(Module &M) { std::string FunctionName = "__msan_maybe_warning_" + itostr(AccessSize); MaybeWarningFn[AccessSizeIndex] = M.getOrInsertFunction( FunctionName, IRB.getVoidTy(), IRB.getIntNTy(AccessSize * 8), - IRB.getInt32Ty(), NULL); + IRB.getInt32Ty(), nullptr); FunctionName = "__msan_maybe_store_origin_" + itostr(AccessSize); MaybeStoreOriginFn[AccessSizeIndex] = M.getOrInsertFunction( FunctionName, IRB.getVoidTy(), IRB.getIntNTy(AccessSize * 8), - IRB.getInt8PtrTy(), IRB.getInt32Ty(), NULL); + IRB.getInt8PtrTy(), IRB.getInt32Ty(), nullptr); } MsanSetAllocaOrigin4Fn = M.getOrInsertFunction( "__msan_set_alloca_origin4", IRB.getVoidTy(), IRB.getInt8PtrTy(), IntptrTy, - IRB.getInt8PtrTy(), IntptrTy, NULL); - MsanPoisonStackFn = M.getOrInsertFunction( - "__msan_poison_stack", IRB.getVoidTy(), IRB.getInt8PtrTy(), IntptrTy, NULL); + IRB.getInt8PtrTy(), IntptrTy, nullptr); + MsanPoisonStackFn = + M.getOrInsertFunction("__msan_poison_stack", IRB.getVoidTy(), + IRB.getInt8PtrTy(), IntptrTy, nullptr); MsanChainOriginFn = M.getOrInsertFunction( - "__msan_chain_origin", IRB.getInt32Ty(), IRB.getInt32Ty(), NULL); + "__msan_chain_origin", IRB.getInt32Ty(), IRB.getInt32Ty(), nullptr); MemmoveFn = M.getOrInsertFunction( "__msan_memmove", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), - IRB.getInt8PtrTy(), IntptrTy, NULL); + IRB.getInt8PtrTy(), IntptrTy, nullptr); MemcpyFn = M.getOrInsertFunction( "__msan_memcpy", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), - IntptrTy, NULL); + IntptrTy, nullptr); MemsetFn = M.getOrInsertFunction( "__msan_memset", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IRB.getInt32Ty(), - IntptrTy, NULL); + IntptrTy, nullptr); // Create globals. RetvalTLS = new GlobalVariable( - M, ArrayType::get(IRB.getInt64Ty(), 8), false, + M, ArrayType::get(IRB.getInt64Ty(), kRetvalTLSSize / 8), false, GlobalVariable::ExternalLinkage, nullptr, "__msan_retval_tls", nullptr, GlobalVariable::InitialExecTLSModel); RetvalOriginTLS = new GlobalVariable( @@ -364,16 +376,16 @@ void MemorySanitizer::initializeCallbacks(Module &M) { "__msan_retval_origin_tls", nullptr, GlobalVariable::InitialExecTLSModel); ParamTLS = new GlobalVariable( - M, ArrayType::get(IRB.getInt64Ty(), 1000), false, + M, ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8), false, GlobalVariable::ExternalLinkage, nullptr, "__msan_param_tls", nullptr, GlobalVariable::InitialExecTLSModel); ParamOriginTLS = new GlobalVariable( - M, ArrayType::get(OriginTy, 1000), false, GlobalVariable::ExternalLinkage, - nullptr, "__msan_param_origin_tls", nullptr, - GlobalVariable::InitialExecTLSModel); + M, ArrayType::get(OriginTy, kParamTLSSize / 4), false, + GlobalVariable::ExternalLinkage, nullptr, "__msan_param_origin_tls", + nullptr, GlobalVariable::InitialExecTLSModel); VAArgTLS = new GlobalVariable( - M, ArrayType::get(IRB.getInt64Ty(), 1000), false, + M, ArrayType::get(IRB.getInt64Ty(), kParamTLSSize / 8), false, GlobalVariable::ExternalLinkage, nullptr, "__msan_va_arg_tls", nullptr, GlobalVariable::InitialExecTLSModel); VAArgOverflowSizeTLS = new GlobalVariable( @@ -393,7 +405,7 @@ void MemorySanitizer::initializeCallbacks(Module &M) { AnyFunctionPtrTy = PointerType::getUnqual(FunctionType::get(IRB.getVoidTy(), false)); IndirectCallWrapperFn = M.getOrInsertFunction( - ClWrapIndirectCalls, AnyFunctionPtrTy, AnyFunctionPtrTy, NULL); + ClWrapIndirectCalls, AnyFunctionPtrTy, AnyFunctionPtrTy, nullptr); } if (WrapIndirectCalls && ClWrapIndirectCallsFast) { @@ -442,7 +454,7 @@ bool MemorySanitizer::doInitialization(Module &M) { // Insert a call to __msan_init/__msan_track_origins into the module's CTORs. appendToGlobalCtors(M, cast<Function>(M.getOrInsertFunction( - "__msan_init", IRB.getVoidTy(), NULL)), 0); + "__msan_init", IRB.getVoidTy(), nullptr)), 0); if (TrackOrigins) new GlobalVariable(M, IRB.getInt32Ty(), true, GlobalValue::WeakODRLinkage, @@ -559,7 +571,8 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> { // TODO(eugenis): handle non-zero constant shadow by inserting an // unconditional check (can not simply fail compilation as this could // be in the dead code). - if (isa<Constant>(ConvertedShadow)) return; + if (!ClCheckConstantShadow) + if (isa<Constant>(ConvertedShadow)) return; unsigned TypeSizeInBits = MS.DL->getTypeSizeInBits(ConvertedShadow->getType()); unsigned SizeIndex = TypeSizeToSizeIndex(TypeSizeInBits); @@ -615,8 +628,9 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> { DEBUG(dbgs() << " SHAD0 : " << *Shadow << "\n"); Value *ConvertedShadow = convertToShadowTyNoVec(Shadow, IRB); DEBUG(dbgs() << " SHAD1 : " << *ConvertedShadow << "\n"); - // See the comment in materializeStores(). - if (isa<Constant>(ConvertedShadow)) return; + // See the comment in storeOrigin(). + if (!ClCheckConstantShadow) + if (isa<Constant>(ConvertedShadow)) return; unsigned TypeSizeInBits = MS.DL->getTypeSizeInBits(ConvertedShadow->getType()); unsigned SizeIndex = TypeSizeToSizeIndex(TypeSizeInBits); @@ -763,6 +777,10 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> { return VectorType::get(IntegerType::get(*MS.C, EltSize), VT->getNumElements()); } + if (ArrayType *AT = dyn_cast<ArrayType>(OrigTy)) { + return ArrayType::get(getShadowTy(AT->getElementType()), + AT->getNumElements()); + } if (StructType *ST = dyn_cast<StructType>(OrigTy)) { SmallVector<Type*, 4> Elements; for (unsigned i = 0, n = ST->getNumElements(); i < n; i++) @@ -882,11 +900,18 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> { assert(ShadowTy); if (isa<IntegerType>(ShadowTy) || isa<VectorType>(ShadowTy)) return Constant::getAllOnesValue(ShadowTy); - StructType *ST = cast<StructType>(ShadowTy); - SmallVector<Constant *, 4> Vals; - for (unsigned i = 0, n = ST->getNumElements(); i < n; i++) - Vals.push_back(getPoisonedShadow(ST->getElementType(i))); - return ConstantStruct::get(ST, Vals); + if (ArrayType *AT = dyn_cast<ArrayType>(ShadowTy)) { + SmallVector<Constant *, 4> Vals(AT->getNumElements(), + getPoisonedShadow(AT->getElementType())); + return ConstantArray::get(AT, Vals); + } + if (StructType *ST = dyn_cast<StructType>(ShadowTy)) { + SmallVector<Constant *, 4> Vals; + for (unsigned i = 0, n = ST->getNumElements(); i < n; i++) + Vals.push_back(getPoisonedShadow(ST->getElementType(i))); + return ConstantStruct::get(ST, Vals); + } + llvm_unreachable("Unexpected shadow type"); } /// \brief Create a dirty shadow for a given value. @@ -941,6 +966,7 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> { ? MS.DL->getTypeAllocSize(FArg.getType()->getPointerElementType()) : MS.DL->getTypeAllocSize(FArg.getType()); if (A == &FArg) { + bool Overflow = ArgOffset + Size > kParamTLSSize; Value *Base = getShadowPtrForArgument(&FArg, EntryIRB, ArgOffset); if (FArg.hasByValAttr()) { // ByVal pointer itself has clean shadow. We copy the actual @@ -951,25 +977,38 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> { Type *EltType = A->getType()->getPointerElementType(); ArgAlign = MS.DL->getABITypeAlignment(EltType); } - unsigned CopyAlign = std::min(ArgAlign, kShadowTLSAlignment); - Value *Cpy = EntryIRB.CreateMemCpy( - getShadowPtr(V, EntryIRB.getInt8Ty(), EntryIRB), Base, Size, - CopyAlign); - DEBUG(dbgs() << " ByValCpy: " << *Cpy << "\n"); - (void)Cpy; + if (Overflow) { + // ParamTLS overflow. + EntryIRB.CreateMemSet( + getShadowPtr(V, EntryIRB.getInt8Ty(), EntryIRB), + Constant::getNullValue(EntryIRB.getInt8Ty()), Size, ArgAlign); + } else { + unsigned CopyAlign = std::min(ArgAlign, kShadowTLSAlignment); + Value *Cpy = EntryIRB.CreateMemCpy( + getShadowPtr(V, EntryIRB.getInt8Ty(), EntryIRB), Base, Size, + CopyAlign); + DEBUG(dbgs() << " ByValCpy: " << *Cpy << "\n"); + (void)Cpy; + } *ShadowPtr = getCleanShadow(V); } else { - *ShadowPtr = EntryIRB.CreateAlignedLoad(Base, kShadowTLSAlignment); + if (Overflow) { + // ParamTLS overflow. + *ShadowPtr = getCleanShadow(V); + } else { + *ShadowPtr = + EntryIRB.CreateAlignedLoad(Base, kShadowTLSAlignment); + } } DEBUG(dbgs() << " ARG: " << FArg << " ==> " << **ShadowPtr << "\n"); - if (MS.TrackOrigins) { + if (MS.TrackOrigins && !Overflow) { Value *OriginPtr = getOriginPtrForArgument(&FArg, EntryIRB, ArgOffset); setOrigin(A, EntryIRB.CreateLoad(OriginPtr)); } } - ArgOffset += DataLayout::RoundUpAlignment(Size, kShadowTLSAlignment); + ArgOffset += RoundUpToAlignment(Size, kShadowTLSAlignment); } assert(*ShadowPtr && "Could not find shadow for an argument"); return *ShadowPtr; @@ -1024,9 +1063,16 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> { /// UMR warning in runtime if the value is not fully defined. void insertShadowCheck(Value *Val, Instruction *OrigIns) { assert(Val); - Instruction *Shadow = dyn_cast_or_null<Instruction>(getShadow(Val)); - if (!Shadow) return; - Instruction *Origin = dyn_cast_or_null<Instruction>(getOrigin(Val)); + Value *Shadow, *Origin; + if (ClCheckConstantShadow) { + Shadow = getShadow(Val); + if (!Shadow) return; + Origin = getOrigin(Val); + } else { + Shadow = dyn_cast_or_null<Instruction>(getShadow(Val)); + if (!Shadow) return; + Origin = dyn_cast_or_null<Instruction>(getOrigin(Val)); + } insertShadowCheck(Shadow, Origin, OrigIns); } @@ -1859,7 +1905,7 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> { Value *Op = I.getArgOperand(0); Type *OpType = Op->getType(); Function *BswapFunc = Intrinsic::getDeclaration( - F.getParent(), Intrinsic::bswap, ArrayRef<Type*>(&OpType, 1)); + F.getParent(), Intrinsic::bswap, makeArrayRef(&OpType, 1)); setShadow(&I, IRB.CreateCall(BswapFunc, getShadow(Op))); setOrigin(&I, getOrigin(Op)); } @@ -2313,26 +2359,32 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> { Value *ArgShadowBase = getShadowPtrForArgument(A, IRB, ArgOffset); DEBUG(dbgs() << " Arg#" << i << ": " << *A << " Shadow: " << *ArgShadow << "\n"); + bool ArgIsInitialized = false; if (CS.paramHasAttr(i + 1, Attribute::ByVal)) { assert(A->getType()->isPointerTy() && "ByVal argument is not a pointer!"); Size = MS.DL->getTypeAllocSize(A->getType()->getPointerElementType()); - unsigned Alignment = CS.getParamAlignment(i + 1); + if (ArgOffset + Size > kParamTLSSize) break; + unsigned ParamAlignment = CS.getParamAlignment(i + 1); + unsigned Alignment = std::min(ParamAlignment, kShadowTLSAlignment); Store = IRB.CreateMemCpy(ArgShadowBase, getShadowPtr(A, Type::getInt8Ty(*MS.C), IRB), Size, Alignment); } else { Size = MS.DL->getTypeAllocSize(A->getType()); + if (ArgOffset + Size > kParamTLSSize) break; Store = IRB.CreateAlignedStore(ArgShadow, ArgShadowBase, kShadowTLSAlignment); + Constant *Cst = dyn_cast<Constant>(ArgShadow); + if (Cst && Cst->isNullValue()) ArgIsInitialized = true; } - if (MS.TrackOrigins) + if (MS.TrackOrigins && !ArgIsInitialized) IRB.CreateStore(getOrigin(A), getOriginPtrForArgument(A, IRB, ArgOffset)); (void)Store; assert(Size != 0 && Store != nullptr); DEBUG(dbgs() << " Param:" << *Store << "\n"); - ArgOffset += DataLayout::RoundUpAlignment(Size, 8); + ArgOffset += RoundUpToAlignment(Size, 8); } DEBUG(dbgs() << " done with call args\n"); @@ -2613,7 +2665,7 @@ struct VarArgAMD64Helper : public VarArgHelper { Type *RealTy = A->getType()->getPointerElementType(); uint64_t ArgSize = MS.DL->getTypeAllocSize(RealTy); Value *Base = getShadowPtrForVAArgument(RealTy, IRB, OverflowOffset); - OverflowOffset += DataLayout::RoundUpAlignment(ArgSize, 8); + OverflowOffset += RoundUpToAlignment(ArgSize, 8); IRB.CreateMemCpy(Base, MSV.getShadowPtr(A, IRB.getInt8Ty(), IRB), ArgSize, kShadowTLSAlignment); } else { @@ -2635,7 +2687,7 @@ struct VarArgAMD64Helper : public VarArgHelper { case AK_Memory: uint64_t ArgSize = MS.DL->getTypeAllocSize(A->getType()); Base = getShadowPtrForVAArgument(A->getType(), IRB, OverflowOffset); - OverflowOffset += DataLayout::RoundUpAlignment(ArgSize, 8); + OverflowOffset += RoundUpToAlignment(ArgSize, 8); } IRB.CreateAlignedStore(MSV.getShadow(A), Base, kShadowTLSAlignment); } diff --git a/lib/Transforms/Instrumentation/SanitizerCoverage.cpp b/lib/Transforms/Instrumentation/SanitizerCoverage.cpp new file mode 100644 index 0000000..f882072 --- /dev/null +++ b/lib/Transforms/Instrumentation/SanitizerCoverage.cpp @@ -0,0 +1,294 @@ +//===-- SanitizerCoverage.cpp - coverage instrumentation for sanitizers ---===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// Coverage instrumentation that works with AddressSanitizer +// and potentially with other Sanitizers. +// +// We create a Guard boolean variable with the same linkage +// as the function and inject this code into the entry block (CoverageLevel=1) +// or all blocks (CoverageLevel>=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). +// +// With CoverageLevel>=3 we also split critical edges this effectively +// instrumenting all edges. +// +// CoverageLevel>=4 add indirect call profiling implented as a function call. +// +// This coverage implementation provides very limited data: +// it only tells if a given function (block) was ever executed. No counters. +// But for many use cases this is what we need and the added slowdown small. +// +//===----------------------------------------------------------------------===// + +#include "llvm/Transforms/Instrumentation.h" +#include "llvm/ADT/ArrayRef.h" +#include "llvm/ADT/SmallVector.h" +#include "llvm/IR/CallSite.h" +#include "llvm/IR/DataLayout.h" +#include "llvm/IR/Function.h" +#include "llvm/IR/IRBuilder.h" +#include "llvm/IR/LLVMContext.h" +#include "llvm/IR/MDBuilder.h" +#include "llvm/IR/Module.h" +#include "llvm/IR/Type.h" +#include "llvm/Support/CommandLine.h" +#include "llvm/Support/Debug.h" +#include "llvm/Support/raw_ostream.h" +#include "llvm/Transforms/Scalar.h" +#include "llvm/Transforms/Utils/BasicBlockUtils.h" +#include "llvm/Transforms/Utils/ModuleUtils.h" + +using namespace llvm; + +#define DEBUG_TYPE "sancov" + +static const char *const kSanCovModuleInitName = "__sanitizer_cov_module_init"; +static const char *const kSanCovName = "__sanitizer_cov"; +static const char *const kSanCovIndirCallName = "__sanitizer_cov_indir_call16"; +static const char *const kSanCovTraceEnter = "__sanitizer_cov_trace_func_enter"; +static const char *const kSanCovTraceBB = "__sanitizer_cov_trace_basic_block"; +static const char *const kSanCovModuleCtorName = "sancov.module_ctor"; +static const uint64_t kSanCtorAndDtorPriority = 1; + +static cl::opt<int> ClCoverageLevel("sanitizer-coverage-level", + cl::desc("Sanitizer Coverage. 0: none, 1: entry block, 2: all blocks, " + "3: all blocks and critical edges, " + "4: above plus indirect calls"), + cl::Hidden, cl::init(0)); + +static cl::opt<int> ClCoverageBlockThreshold( + "sanitizer-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<bool> + ClExperimentalTracing("sanitizer-coverage-experimental-tracing", + cl::desc("Experimental basic-block tracing: insert " + "callbacks at every basic block"), + cl::Hidden, cl::init(false)); + +namespace { + +class SanitizerCoverageModule : public ModulePass { + public: + SanitizerCoverageModule(int CoverageLevel = 0) + : ModulePass(ID), + CoverageLevel(std::max(CoverageLevel, (int)ClCoverageLevel)) {} + bool runOnModule(Module &M) override; + bool runOnFunction(Function &F); + static char ID; // Pass identification, replacement for typeid + const char *getPassName() const override { + return "SanitizerCoverageModule"; + } + + void getAnalysisUsage(AnalysisUsage &AU) const override { + AU.addRequired<DataLayoutPass>(); + } + + private: + void InjectCoverageForIndirectCalls(Function &F, + ArrayRef<Instruction *> IndirCalls); + bool InjectCoverage(Function &F, ArrayRef<BasicBlock *> AllBlocks, + ArrayRef<Instruction *> IndirCalls); + bool InjectTracing(Function &F, ArrayRef<BasicBlock *> AllBlocks); + void InjectCoverageAtBlock(Function &F, BasicBlock &BB); + Function *SanCovFunction; + Function *SanCovIndirCallFunction; + Function *SanCovModuleInit; + Function *SanCovTraceEnter, *SanCovTraceBB; + Type *IntptrTy; + LLVMContext *C; + + int CoverageLevel; +}; + +} // namespace + +static Function *checkInterfaceFunction(Constant *FuncOrBitcast) { + if (Function *F = dyn_cast<Function>(FuncOrBitcast)) + return F; + std::string Err; + raw_string_ostream Stream(Err); + Stream << "SanitizerCoverage interface function redefined: " + << *FuncOrBitcast; + report_fatal_error(Err); +} + +bool SanitizerCoverageModule::runOnModule(Module &M) { + if (!CoverageLevel) return false; + C = &(M.getContext()); + DataLayoutPass *DLP = &getAnalysis<DataLayoutPass>(); + IntptrTy = Type::getIntNTy(*C, DLP->getDataLayout().getPointerSizeInBits()); + Type *VoidTy = Type::getVoidTy(*C); + + Function *CtorFunc = + Function::Create(FunctionType::get(VoidTy, false), + GlobalValue::InternalLinkage, kSanCovModuleCtorName, &M); + ReturnInst::Create(*C, BasicBlock::Create(*C, "", CtorFunc)); + appendToGlobalCtors(M, CtorFunc, kSanCtorAndDtorPriority); + + SanCovFunction = + checkInterfaceFunction(M.getOrInsertFunction(kSanCovName, VoidTy, nullptr)); + SanCovIndirCallFunction = checkInterfaceFunction(M.getOrInsertFunction( + kSanCovIndirCallName, VoidTy, IntptrTy, IntptrTy, nullptr)); + SanCovModuleInit = checkInterfaceFunction(M.getOrInsertFunction( + kSanCovModuleInitName, Type::getVoidTy(*C), IntptrTy, nullptr)); + SanCovModuleInit->setLinkage(Function::ExternalLinkage); + + if (ClExperimentalTracing) { + SanCovTraceEnter = checkInterfaceFunction( + M.getOrInsertFunction(kSanCovTraceEnter, VoidTy, IntptrTy, nullptr)); + SanCovTraceBB = checkInterfaceFunction( + M.getOrInsertFunction(kSanCovTraceBB, VoidTy, IntptrTy, nullptr)); + } + + for (auto &F : M) + runOnFunction(F); + + IRBuilder<> IRB(CtorFunc->getEntryBlock().getTerminator()); + IRB.CreateCall(SanCovModuleInit, + ConstantInt::get(IntptrTy, SanCovFunction->getNumUses())); + return true; +} + +bool SanitizerCoverageModule::runOnFunction(Function &F) { + if (F.empty()) return false; + // For now instrument only functions that will also be asan-instrumented. + if (!F.hasFnAttribute(Attribute::SanitizeAddress)) + return false; + if (CoverageLevel >= 3) + SplitAllCriticalEdges(F, this); + SmallVector<Instruction*, 8> IndirCalls; + SmallVector<BasicBlock*, 16> AllBlocks; + for (auto &BB : F) { + AllBlocks.push_back(&BB); + if (CoverageLevel >= 4) + for (auto &Inst : BB) { + CallSite CS(&Inst); + if (CS && !CS.getCalledFunction()) + IndirCalls.push_back(&Inst); + } + } + InjectCoverage(F, AllBlocks, IndirCalls); + InjectTracing(F, AllBlocks); + return true; +} + +// Experimental support for tracing. +// Basicaly, insert a callback at the beginning of every basic block. +// Every callback gets a pointer to a uniqie global for internal storage. +bool SanitizerCoverageModule::InjectTracing(Function &F, + ArrayRef<BasicBlock *> AllBlocks) { + if (!ClExperimentalTracing) return false; + Type *Ty = ArrayType::get(IntptrTy, 1); // May need to use more words later. + for (auto BB : AllBlocks) { + IRBuilder<> IRB(BB->getFirstInsertionPt()); + GlobalVariable *TraceCache = new GlobalVariable( + *F.getParent(), Ty, false, GlobalValue::PrivateLinkage, + Constant::getNullValue(Ty), "__sancov_gen_trace_cache"); + IRB.CreateCall(&F.getEntryBlock() == BB ? SanCovTraceEnter : SanCovTraceBB, + IRB.CreatePointerCast(TraceCache, IntptrTy)); + } + return true; +} + +bool +SanitizerCoverageModule::InjectCoverage(Function &F, + ArrayRef<BasicBlock *> AllBlocks, + ArrayRef<Instruction *> IndirCalls) { + if (!CoverageLevel) return false; + + if (CoverageLevel == 1 || + (unsigned)ClCoverageBlockThreshold < AllBlocks.size()) { + InjectCoverageAtBlock(F, F.getEntryBlock()); + } else { + for (auto BB : AllBlocks) + InjectCoverageAtBlock(F, *BB); + } + InjectCoverageForIndirectCalls(F, IndirCalls); + return true; +} + +// On every indirect call we call a run-time function +// __sanitizer_cov_indir_call* with two parameters: +// - callee address, +// - global cache array that contains kCacheSize pointers (zero-initialized). +// The cache is used to speed up recording the caller-callee pairs. +// The address of the caller is passed implicitly via caller PC. +// kCacheSize is encoded in the name of the run-time function. +void SanitizerCoverageModule::InjectCoverageForIndirectCalls( + Function &F, ArrayRef<Instruction *> IndirCalls) { + if (IndirCalls.empty()) return; + const int kCacheSize = 16; + const int kCacheAlignment = 64; // Align for better performance. + Type *Ty = ArrayType::get(IntptrTy, kCacheSize); + for (auto I : IndirCalls) { + IRBuilder<> IRB(I); + CallSite CS(I); + Value *Callee = CS.getCalledValue(); + if (dyn_cast<InlineAsm>(Callee)) continue; + GlobalVariable *CalleeCache = new GlobalVariable( + *F.getParent(), Ty, false, GlobalValue::PrivateLinkage, + Constant::getNullValue(Ty), "__sancov_gen_callee_cache"); + CalleeCache->setAlignment(kCacheAlignment); + IRB.CreateCall2(SanCovIndirCallFunction, + IRB.CreatePointerCast(Callee, IntptrTy), + IRB.CreatePointerCast(CalleeCache, IntptrTy)); + } +} + +void SanitizerCoverageModule::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<AllocaInst>(IP); + if (!AI || !AI->isStaticAlloca()) + break; + } + + DebugLoc EntryLoc = &BB == &F.getEntryBlock() + ? IP->getDebugLoc().getFnDebugLoc(*C) + : IP->getDebugLoc(); + IRBuilder<> IRB(IP); + IRB.SetCurrentDebugLocation(EntryLoc); + Type *Int8Ty = IRB.getInt8Ty(); + GlobalVariable *Guard = new GlobalVariable( + *F.getParent(), Int8Ty, false, GlobalValue::PrivateLinkage, + Constant::getNullValue(Int8Ty), "__sancov_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(SanCovFunction); + StoreInst *Store = IRB.CreateStore(ConstantInt::get(Int8Ty, 1), Guard); + Store->setAtomic(Monotonic); + Store->setAlignment(1); +} + +char SanitizerCoverageModule::ID = 0; +INITIALIZE_PASS(SanitizerCoverageModule, "sancov", + "SanitizerCoverage: TODO." + "ModulePass", false, false) +ModulePass *llvm::createSanitizerCoverageModulePass(int CoverageLevel) { + return new SanitizerCoverageModule(CoverageLevel); +} diff --git a/lib/Transforms/Instrumentation/ThreadSanitizer.cpp b/lib/Transforms/Instrumentation/ThreadSanitizer.cpp index 89386a6..8a56a1f 100644 --- a/lib/Transforms/Instrumentation/ThreadSanitizer.cpp +++ b/lib/Transforms/Instrumentation/ThreadSanitizer.cpp @@ -135,33 +135,33 @@ void ThreadSanitizer::initializeCallbacks(Module &M) { IRBuilder<> IRB(M.getContext()); // Initialize the callbacks. TsanFuncEntry = checkInterfaceFunction(M.getOrInsertFunction( - "__tsan_func_entry", IRB.getVoidTy(), IRB.getInt8PtrTy(), NULL)); + "__tsan_func_entry", IRB.getVoidTy(), IRB.getInt8PtrTy(), nullptr)); TsanFuncExit = checkInterfaceFunction(M.getOrInsertFunction( - "__tsan_func_exit", IRB.getVoidTy(), NULL)); + "__tsan_func_exit", IRB.getVoidTy(), nullptr)); OrdTy = IRB.getInt32Ty(); for (size_t i = 0; i < kNumberOfAccessSizes; ++i) { const size_t ByteSize = 1 << i; const size_t BitSize = ByteSize * 8; SmallString<32> ReadName("__tsan_read" + itostr(ByteSize)); TsanRead[i] = checkInterfaceFunction(M.getOrInsertFunction( - ReadName, IRB.getVoidTy(), IRB.getInt8PtrTy(), NULL)); + ReadName, IRB.getVoidTy(), IRB.getInt8PtrTy(), nullptr)); SmallString<32> WriteName("__tsan_write" + itostr(ByteSize)); TsanWrite[i] = checkInterfaceFunction(M.getOrInsertFunction( - WriteName, IRB.getVoidTy(), IRB.getInt8PtrTy(), NULL)); + WriteName, IRB.getVoidTy(), IRB.getInt8PtrTy(), nullptr)); Type *Ty = Type::getIntNTy(M.getContext(), BitSize); Type *PtrTy = Ty->getPointerTo(); SmallString<32> AtomicLoadName("__tsan_atomic" + itostr(BitSize) + "_load"); TsanAtomicLoad[i] = checkInterfaceFunction(M.getOrInsertFunction( - AtomicLoadName, Ty, PtrTy, OrdTy, NULL)); + AtomicLoadName, Ty, PtrTy, OrdTy, nullptr)); SmallString<32> AtomicStoreName("__tsan_atomic" + itostr(BitSize) + "_store"); TsanAtomicStore[i] = checkInterfaceFunction(M.getOrInsertFunction( AtomicStoreName, IRB.getVoidTy(), PtrTy, Ty, OrdTy, - NULL)); + nullptr)); for (int op = AtomicRMWInst::FIRST_BINOP; op <= AtomicRMWInst::LAST_BINOP; ++op) { @@ -185,33 +185,33 @@ void ThreadSanitizer::initializeCallbacks(Module &M) { continue; SmallString<32> RMWName("__tsan_atomic" + itostr(BitSize) + NamePart); TsanAtomicRMW[op][i] = checkInterfaceFunction(M.getOrInsertFunction( - RMWName, Ty, PtrTy, Ty, OrdTy, NULL)); + RMWName, Ty, PtrTy, Ty, OrdTy, nullptr)); } SmallString<32> AtomicCASName("__tsan_atomic" + itostr(BitSize) + "_compare_exchange_val"); TsanAtomicCAS[i] = checkInterfaceFunction(M.getOrInsertFunction( - AtomicCASName, Ty, PtrTy, Ty, Ty, OrdTy, OrdTy, NULL)); + AtomicCASName, Ty, PtrTy, Ty, Ty, OrdTy, OrdTy, nullptr)); } TsanVptrUpdate = checkInterfaceFunction(M.getOrInsertFunction( "__tsan_vptr_update", IRB.getVoidTy(), IRB.getInt8PtrTy(), - IRB.getInt8PtrTy(), NULL)); + IRB.getInt8PtrTy(), nullptr)); TsanVptrLoad = checkInterfaceFunction(M.getOrInsertFunction( - "__tsan_vptr_read", IRB.getVoidTy(), IRB.getInt8PtrTy(), NULL)); + "__tsan_vptr_read", IRB.getVoidTy(), IRB.getInt8PtrTy(), nullptr)); TsanAtomicThreadFence = checkInterfaceFunction(M.getOrInsertFunction( - "__tsan_atomic_thread_fence", IRB.getVoidTy(), OrdTy, NULL)); + "__tsan_atomic_thread_fence", IRB.getVoidTy(), OrdTy, nullptr)); TsanAtomicSignalFence = checkInterfaceFunction(M.getOrInsertFunction( - "__tsan_atomic_signal_fence", IRB.getVoidTy(), OrdTy, NULL)); + "__tsan_atomic_signal_fence", IRB.getVoidTy(), OrdTy, nullptr)); MemmoveFn = checkInterfaceFunction(M.getOrInsertFunction( "memmove", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), - IRB.getInt8PtrTy(), IntptrTy, NULL)); + IRB.getInt8PtrTy(), IntptrTy, nullptr)); MemcpyFn = checkInterfaceFunction(M.getOrInsertFunction( "memcpy", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), - IntptrTy, NULL)); + IntptrTy, nullptr)); MemsetFn = checkInterfaceFunction(M.getOrInsertFunction( "memset", IRB.getInt8PtrTy(), IRB.getInt8PtrTy(), IRB.getInt32Ty(), - IntptrTy, NULL)); + IntptrTy, nullptr)); } bool ThreadSanitizer::doInitialization(Module &M) { @@ -224,7 +224,7 @@ bool ThreadSanitizer::doInitialization(Module &M) { IRBuilder<> IRB(M.getContext()); IntptrTy = IRB.getIntPtrTy(DL); Value *TsanInit = M.getOrInsertFunction("__tsan_init", - IRB.getVoidTy(), NULL); + IRB.getVoidTy(), nullptr); appendToGlobalCtors(M, cast<Function>(TsanInit), 0); return true; @@ -481,8 +481,7 @@ bool ThreadSanitizer::instrumentAtomic(Instruction *I) { Type *PtrTy = Ty->getPointerTo(); Value *Args[] = {IRB.CreatePointerCast(Addr, PtrTy), createOrdering(&IRB, LI->getOrdering())}; - CallInst *C = CallInst::Create(TsanAtomicLoad[Idx], - ArrayRef<Value*>(Args)); + CallInst *C = CallInst::Create(TsanAtomicLoad[Idx], Args); ReplaceInstWithInst(I, C); } else if (StoreInst *SI = dyn_cast<StoreInst>(I)) { @@ -497,8 +496,7 @@ bool ThreadSanitizer::instrumentAtomic(Instruction *I) { Value *Args[] = {IRB.CreatePointerCast(Addr, PtrTy), IRB.CreateIntCast(SI->getValueOperand(), Ty, false), createOrdering(&IRB, SI->getOrdering())}; - CallInst *C = CallInst::Create(TsanAtomicStore[Idx], - ArrayRef<Value*>(Args)); + CallInst *C = CallInst::Create(TsanAtomicStore[Idx], Args); ReplaceInstWithInst(I, C); } else if (AtomicRMWInst *RMWI = dyn_cast<AtomicRMWInst>(I)) { Value *Addr = RMWI->getPointerOperand(); @@ -515,7 +513,7 @@ bool ThreadSanitizer::instrumentAtomic(Instruction *I) { Value *Args[] = {IRB.CreatePointerCast(Addr, PtrTy), IRB.CreateIntCast(RMWI->getValOperand(), Ty, false), createOrdering(&IRB, RMWI->getOrdering())}; - CallInst *C = CallInst::Create(F, ArrayRef<Value*>(Args)); + CallInst *C = CallInst::Create(F, Args); ReplaceInstWithInst(I, C); } else if (AtomicCmpXchgInst *CASI = dyn_cast<AtomicCmpXchgInst>(I)) { Value *Addr = CASI->getPointerOperand(); @@ -543,7 +541,7 @@ bool ThreadSanitizer::instrumentAtomic(Instruction *I) { Value *Args[] = {createOrdering(&IRB, FI->getOrdering())}; Function *F = FI->getSynchScope() == SingleThread ? TsanAtomicSignalFence : TsanAtomicThreadFence; - CallInst *C = CallInst::Create(F, ArrayRef<Value*>(Args)); + CallInst *C = CallInst::Create(F, Args); ReplaceInstWithInst(I, C); } return true; |