#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/DenseSet.h"
#include "llvm/ADT/DepthFirstIterator.h"
+#include "llvm/ADT/StringExtras.h"
+#include "llvm/ADT/Triple.h"
#include "llvm/Analysis/ValueTracking.h"
-#include "llvm/IR/InlineAsm.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"
#include "llvm/IR/LLVMContext.h"
#include "llvm/IR/MDBuilder.h"
#include "llvm/IR/Type.h"
#include "llvm/IR/Value.h"
-#include "llvm/InstVisitor.h"
#include "llvm/Pass.h"
#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/SpecialCaseList.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/Local.h"
-#include "llvm/Transforms/Utils/SpecialCaseList.h"
+#include <algorithm>
#include <iterator>
+#include <set>
+#include <utility>
using namespace llvm;
cl::desc("respect alignment requirements provided by input IR"), cl::Hidden,
cl::init(false));
-// The ABI list file controls how shadow parameters are passed. The pass treats
+// The ABI list files control how shadow parameters are passed. The pass treats
// every function labelled "uninstrumented" in the ABI list file as conforming
// to the "native" (i.e. unsanitized) ABI. Unless the ABI list contains
// additional annotations for those functions, a call to one of those functions
// will produce a warning message, as the labelling behaviour of the function is
// unknown. The other supported annotations are "functional" and "discard",
// which are described below under DataFlowSanitizer::WrapperKind.
-static cl::opt<std::string> ClABIListFile(
+static cl::list<std::string> ClABIListFiles(
"dfsan-abilist",
cl::desc("File listing native ABI functions and how the pass treats them"),
cl::Hidden);
cl::desc("Use the argument ABI rather than the TLS ABI"),
cl::Hidden);
+// Controls whether the pass includes or ignores the labels of pointers in load
+// instructions.
+static cl::opt<bool> ClCombinePointerLabelsOnLoad(
+ "dfsan-combine-pointer-labels-on-load",
+ cl::desc("Combine the label of the pointer with the label of the data when "
+ "loading from memory."),
+ cl::Hidden, cl::init(true));
+
+// Controls whether the pass includes or ignores the labels of pointers in
+// stores instructions.
+static cl::opt<bool> ClCombinePointerLabelsOnStore(
+ "dfsan-combine-pointer-labels-on-store",
+ cl::desc("Combine the label of the pointer with the label of the data when "
+ "storing in memory."),
+ cl::Hidden, cl::init(false));
+
+static cl::opt<bool> ClDebugNonzeroLabels(
+ "dfsan-debug-nonzero-labels",
+ cl::desc("Insert calls to __dfsan_nonzero_label on observing a parameter, "
+ "load or return with a nonzero label"),
+ cl::Hidden);
+
namespace {
+StringRef GetGlobalTypeString(const GlobalValue &G) {
+ // Types of GlobalVariables are always pointer types.
+ Type *GType = G.getType()->getElementType();
+ // For now we support blacklisting struct types only.
+ if (StructType *SGType = dyn_cast<StructType>(GType)) {
+ if (!SGType->isLiteral())
+ return SGType->getName();
+ }
+ return "<unknown type>";
+}
+
+class DFSanABIList {
+ std::unique_ptr<SpecialCaseList> SCL;
+
+ public:
+ DFSanABIList() {}
+
+ void set(std::unique_ptr<SpecialCaseList> List) { SCL = std::move(List); }
+
+ /// Returns whether either this function or its source file are listed in the
+ /// given category.
+ bool isIn(const Function &F, StringRef Category) const {
+ return isIn(*F.getParent(), Category) ||
+ SCL->inSection("fun", F.getName(), Category);
+ }
+
+ /// Returns whether this global alias is listed in the given category.
+ ///
+ /// 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, StringRef Category) const {
+ if (isIn(*GA.getParent(), Category))
+ return true;
+
+ if (isa<FunctionType>(GA.getType()->getElementType()))
+ return SCL->inSection("fun", GA.getName(), Category);
+
+ return SCL->inSection("global", GA.getName(), Category) ||
+ SCL->inSection("type", GetGlobalTypeString(GA), Category);
+ }
+
+ /// Returns whether this module is listed in the given category.
+ bool isIn(const Module &M, StringRef Category) const {
+ return SCL->inSection("src", M.getModuleIdentifier(), Category);
+ }
+};
+
class DataFlowSanitizer : public ModulePass {
friend struct DFSanFunction;
friend class DFSanVisitor;
WK_Custom
};
- DataLayout *DL;
+ const DataLayout *DL;
Module *Mod;
LLVMContext *Ctx;
IntegerType *ShadowTy;
FunctionType *DFSanUnionFnTy;
FunctionType *DFSanUnionLoadFnTy;
FunctionType *DFSanUnimplementedFnTy;
+ FunctionType *DFSanSetLabelFnTy;
+ FunctionType *DFSanNonzeroLabelFnTy;
+ FunctionType *DFSanVarargWrapperFnTy;
Constant *DFSanUnionFn;
+ Constant *DFSanCheckedUnionFn;
Constant *DFSanUnionLoadFn;
Constant *DFSanUnimplementedFn;
+ Constant *DFSanSetLabelFn;
+ Constant *DFSanNonzeroLabelFn;
+ Constant *DFSanVarargWrapperFn;
MDNode *ColdCallWeights;
- OwningPtr<SpecialCaseList> ABIList;
+ 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(Function *F);
+ bool isInstrumented(const Function *F);
+ bool isInstrumented(const GlobalAlias *GA);
FunctionType *getArgsFunctionType(FunctionType *T);
+ FunctionType *getTrampolineFunctionType(FunctionType *T);
FunctionType *getCustomFunctionType(FunctionType *T);
InstrumentedABI getInstrumentedABI();
WrapperKind getWrapperKind(Function *F);
+ void addGlobalNamePrefix(GlobalValue *GV);
+ Function *buildWrapperFunction(Function *F, StringRef NewFName,
+ GlobalValue::LinkageTypes NewFLink,
+ FunctionType *NewFT);
+ Constant *getOrBuildTrampolineFunction(FunctionType *FT, StringRef FName);
public:
- DataFlowSanitizer(StringRef ABIListFile = StringRef(),
- void *(*getArgTLS)() = 0, void *(*getRetValTLS)() = 0);
+ DataFlowSanitizer(
+ const std::vector<std::string> &ABIListFiles = std::vector<std::string>(),
+ void *(*getArgTLS)() = nullptr, void *(*getRetValTLS)() = nullptr);
static char ID;
- bool doInitialization(Module &M);
- bool runOnModule(Module &M);
+ bool doInitialization(Module &M) override;
+ bool runOnModule(Module &M) override;
};
struct DFSanFunction {
DataFlowSanitizer &DFS;
Function *F;
+ DominatorTree DT;
DataFlowSanitizer::InstrumentedABI IA;
bool IsNativeABI;
Value *ArgTLSPtr;
DenseMap<AllocaInst *, AllocaInst *> AllocaShadowMap;
std::vector<std::pair<PHINode *, PHINode *> > PHIFixups;
DenseSet<Instruction *> SkipInsts;
+ 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(0), RetvalTLSPtr(0),
- LabelReturnAlloca(0) {}
+ IsNativeABI(IsNativeABI), ArgTLSPtr(nullptr), RetvalTLSPtr(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);
void visitInsertValueInst(InsertValueInst &I);
void visitAllocaInst(AllocaInst &I);
void visitSelectInst(SelectInst &I);
+ void visitMemSetInst(MemSetInst &I);
void visitMemTransferInst(MemTransferInst &I);
};
INITIALIZE_PASS(DataFlowSanitizer, "dfsan",
"DataFlowSanitizer: dynamic data flow analysis.", false, false)
-ModulePass *llvm::createDataFlowSanitizerPass(StringRef ABIListFile,
- void *(*getArgTLS)(),
- void *(*getRetValTLS)()) {
- return new DataFlowSanitizer(ABIListFile, getArgTLS, getRetValTLS);
+ModulePass *
+llvm::createDataFlowSanitizerPass(const std::vector<std::string> &ABIListFiles,
+ void *(*getArgTLS)(),
+ void *(*getRetValTLS)()) {
+ return new DataFlowSanitizer(ABIListFiles, getArgTLS, getRetValTLS);
}
-DataFlowSanitizer::DataFlowSanitizer(StringRef ABIListFile,
- void *(*getArgTLS)(),
- void *(*getRetValTLS)())
- : ModulePass(ID), GetArgTLSPtr(getArgTLS), GetRetvalTLSPtr(getRetValTLS),
- ABIList(SpecialCaseList::createOrDie(ABIListFile.empty() ? ClABIListFile
- : ABIListFile)) {
+DataFlowSanitizer::DataFlowSanitizer(
+ const std::vector<std::string> &ABIListFiles, void *(*getArgTLS)(),
+ void *(*getRetValTLS)())
+ : ModulePass(ID), GetArgTLSPtr(getArgTLS), GetRetvalTLSPtr(getRetValTLS) {
+ std::vector<std::string> AllABIListFiles(std::move(ABIListFiles));
+ AllABIListFiles.insert(AllABIListFiles.end(), ClABIListFiles.begin(),
+ ClABIListFiles.end());
+ ABIList.set(SpecialCaseList::createOrDie(AllABIListFiles));
}
FunctionType *DataFlowSanitizer::getArgsFunctionType(FunctionType *T) {
ArgTypes.push_back(ShadowPtrTy);
Type *RetType = T->getReturnType();
if (!RetType->isVoidTy())
- RetType = StructType::get(RetType, ShadowTy, (Type *)0);
+ RetType = StructType::get(RetType, ShadowTy, (Type *)nullptr);
return FunctionType::get(RetType, ArgTypes, T->isVarArg());
}
-FunctionType *DataFlowSanitizer::getCustomFunctionType(FunctionType *T) {
+FunctionType *DataFlowSanitizer::getTrampolineFunctionType(FunctionType *T) {
assert(!T->isVarArg());
llvm::SmallVector<Type *, 4> ArgTypes;
+ ArgTypes.push_back(T->getPointerTo());
std::copy(T->param_begin(), T->param_end(), std::back_inserter(ArgTypes));
for (unsigned i = 0, e = T->getNumParams(); i != e; ++i)
ArgTypes.push_back(ShadowTy);
return FunctionType::get(T->getReturnType(), ArgTypes, false);
}
+FunctionType *DataFlowSanitizer::getCustomFunctionType(FunctionType *T) {
+ llvm::SmallVector<Type *, 4> ArgTypes;
+ for (FunctionType::param_iterator i = T->param_begin(), e = T->param_end();
+ i != e; ++i) {
+ FunctionType *FT;
+ if (isa<PointerType>(*i) && (FT = dyn_cast<FunctionType>(cast<PointerType>(
+ *i)->getElementType()))) {
+ ArgTypes.push_back(getTrampolineFunctionType(FT)->getPointerTo());
+ ArgTypes.push_back(Type::getInt8PtrTy(*Ctx));
+ } else {
+ ArgTypes.push_back(*i);
+ }
+ }
+ 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, T->isVarArg());
+}
+
bool DataFlowSanitizer::doInitialization(Module &M) {
- DL = getAnalysisIfAvailable<DataLayout>();
- if (!DL)
- return false;
+ llvm::Triple TargetTriple(M.getTargetTriple());
+ bool IsX86_64 = TargetTriple.getArch() == llvm::Triple::x86_64;
+ bool IsMIPS64 = TargetTriple.getArch() == llvm::Triple::mips64 ||
+ TargetTriple.getArch() == llvm::Triple::mips64el;
+
+ DataLayoutPass *DLP = getAnalysisIfAvailable<DataLayoutPass>();
+ if (!DLP)
+ report_fatal_error("data layout missing");
+ DL = &DLP->getDataLayout();
Mod = &M;
Ctx = &M.getContext();
ShadowPtrTy = PointerType::getUnqual(ShadowTy);
IntptrTy = DL->getIntPtrType(*Ctx);
ZeroShadow = ConstantInt::getSigned(ShadowTy, 0);
- ShadowPtrMask = ConstantInt::getSigned(IntptrTy, ~0x700000000000LL);
ShadowPtrMul = ConstantInt::getSigned(IntptrTy, ShadowWidth / 8);
+ if (IsX86_64)
+ ShadowPtrMask = ConstantInt::getSigned(IntptrTy, ~0x700000000000LL);
+ else if (IsMIPS64)
+ ShadowPtrMask = ConstantInt::getSigned(IntptrTy, ~0xF000000000LL);
+ else
+ report_fatal_error("unsupported triple");
Type *DFSanUnionArgs[2] = { ShadowTy, ShadowTy };
DFSanUnionFnTy =
FunctionType::get(ShadowTy, DFSanUnionLoadArgs, /*isVarArg=*/ false);
DFSanUnimplementedFnTy = FunctionType::get(
Type::getVoidTy(*Ctx), Type::getInt8PtrTy(*Ctx), /*isVarArg=*/false);
+ Type *DFSanSetLabelArgs[3] = { ShadowTy, Type::getInt8PtrTy(*Ctx), IntptrTy };
+ DFSanSetLabelFnTy = FunctionType::get(Type::getVoidTy(*Ctx),
+ DFSanSetLabelArgs, /*isVarArg=*/false);
+ DFSanNonzeroLabelFnTy = FunctionType::get(
+ 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);
- ArgTLS = 0;
+ ArgTLS = nullptr;
GetArgTLS = ConstantExpr::getIntToPtr(
ConstantInt::get(IntptrTy, uintptr_t(GetArgTLSPtr)),
PointerType::getUnqual(
- FunctionType::get(PointerType::getUnqual(ArgTLSTy), (Type *)0)));
+ FunctionType::get(PointerType::getUnqual(ArgTLSTy),
+ (Type *)nullptr)));
}
if (GetRetvalTLSPtr) {
- RetvalTLS = 0;
+ RetvalTLS = nullptr;
GetRetvalTLS = ConstantExpr::getIntToPtr(
ConstantInt::get(IntptrTy, uintptr_t(GetRetvalTLSPtr)),
PointerType::getUnqual(
- FunctionType::get(PointerType::getUnqual(ShadowTy), (Type *)0)));
+ FunctionType::get(PointerType::getUnqual(ShadowTy),
+ (Type *)nullptr)));
}
ColdCallWeights = MDBuilder(*Ctx).createBranchWeights(1, 1000);
return true;
}
-bool DataFlowSanitizer::isInstrumented(Function *F) {
- return !ABIList->isIn(*F, "uninstrumented");
+bool DataFlowSanitizer::isInstrumented(const Function *F) {
+ return !ABIList.isIn(*F, "uninstrumented");
+}
+
+bool DataFlowSanitizer::isInstrumented(const GlobalAlias *GA) {
+ return !ABIList.isIn(*GA, "uninstrumented");
}
DataFlowSanitizer::InstrumentedABI DataFlowSanitizer::getInstrumentedABI() {
}
DataFlowSanitizer::WrapperKind DataFlowSanitizer::getWrapperKind(Function *F) {
- if (ABIList->isIn(*F, "functional"))
+ if (ABIList.isIn(*F, "functional"))
return WK_Functional;
- if (ABIList->isIn(*F, "discard"))
+ if (ABIList.isIn(*F, "discard"))
return WK_Discard;
- if (ABIList->isIn(*F, "custom"))
+ if (ABIList.isIn(*F, "custom"))
return WK_Custom;
return WK_Warning;
}
+void DataFlowSanitizer::addGlobalNamePrefix(GlobalValue *GV) {
+ std::string GVName = GV->getName(), Prefix = "dfs$";
+ GV->setName(Prefix + GVName);
+
+ // Try to change the name of the function in module inline asm. We only do
+ // this for specific asm directives, currently only ".symver", to try to avoid
+ // corrupting asm which happens to contain the symbol name as a substring.
+ // Note that the substitution for .symver assumes that the versioned symbol
+ // also has an instrumented name.
+ std::string Asm = GV->getParent()->getModuleInlineAsm();
+ std::string SearchStr = ".symver " + GVName + ",";
+ size_t Pos = Asm.find(SearchStr);
+ if (Pos != std::string::npos) {
+ Asm.replace(Pos, SearchStr.size(),
+ ".symver " + Prefix + GVName + "," + Prefix);
+ GV->getParent()->setModuleInlineAsm(Asm);
+ }
+}
+
+Function *
+DataFlowSanitizer::buildWrapperFunction(Function *F, StringRef NewFName,
+ GlobalValue::LinkageTypes NewFLink,
+ FunctionType *NewFT) {
+ FunctionType *FT = F->getFunctionType();
+ Function *NewF = Function::Create(NewFT, NewFLink, NewFName,
+ F->getParent());
+ NewF->copyAttributesFrom(F);
+ NewF->removeAttributes(
+ AttributeSet::ReturnIndex,
+ AttributeFuncs::typeIncompatible(NewFT->getReturnType(),
+ AttributeSet::ReturnIndex));
+
+ BasicBlock *BB = BasicBlock::Create(*Ctx, "entry", NewF);
+ 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;
+}
+
+Constant *DataFlowSanitizer::getOrBuildTrampolineFunction(FunctionType *FT,
+ StringRef FName) {
+ FunctionType *FTT = getTrampolineFunctionType(FT);
+ Constant *C = Mod->getOrInsertFunction(FName, FTT);
+ Function *F = dyn_cast<Function>(C);
+ if (F && F->isDeclaration()) {
+ F->setLinkage(GlobalValue::LinkOnceODRLinkage);
+ BasicBlock *BB = BasicBlock::Create(*Ctx, "entry", F);
+ std::vector<Value *> Args;
+ Function::arg_iterator AI = F->arg_begin(); ++AI;
+ for (unsigned N = FT->getNumParams(); N != 0; ++AI, --N)
+ Args.push_back(&*AI);
+ CallInst *CI =
+ CallInst::Create(&F->getArgumentList().front(), Args, "", BB);
+ ReturnInst *RI;
+ if (FT->getReturnType()->isVoidTy())
+ RI = ReturnInst::Create(*Ctx, BB);
+ else
+ RI = ReturnInst::Create(*Ctx, CI, BB);
+
+ DFSanFunction DFSF(*this, F, /*IsNativeABI=*/true);
+ Function::arg_iterator ValAI = F->arg_begin(), ShadowAI = AI; ++ValAI;
+ for (unsigned N = FT->getNumParams(); N != 0; ++ValAI, ++ShadowAI, --N)
+ DFSF.ValShadowMap[ValAI] = ShadowAI;
+ DFSanVisitor(DFSF).visitCallInst(*CI);
+ if (!FT->getReturnType()->isVoidTy())
+ new StoreInst(DFSF.getShadow(RI->getReturnValue()),
+ &F->getArgumentList().back(), RI);
+ }
+
+ return C;
+}
+
bool DataFlowSanitizer::runOnModule(Module &M) {
if (!DL)
return false;
- if (ABIList->isIn(M, "skip"))
+ 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);
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);
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);
}
DFSanUnimplementedFn =
Mod->getOrInsertFunction("__dfsan_unimplemented", DFSanUnimplementedFnTy);
+ DFSanSetLabelFn =
+ Mod->getOrInsertFunction("__dfsan_set_label", DFSanSetLabelFnTy);
+ if (Function *F = dyn_cast<Function>(DFSanSetLabelFn)) {
+ F->addAttribute(1, Attribute::ZExt);
+ }
+ 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 != DFSanUnimplementedFn &&
+ i != DFSanSetLabelFn &&
+ i != DFSanNonzeroLabelFn &&
+ i != DFSanVarargWrapperFn)
FnsToInstrument.push_back(&*i);
}
+ // Give function aliases prefixes when necessary, and build wrappers where the
+ // instrumentedness is inconsistent.
+ for (Module::alias_iterator i = M.alias_begin(), e = M.alias_end(); i != e;) {
+ GlobalAlias *GA = &*i;
+ ++i;
+ // Don't stop on weak. We assume people aren't playing games with the
+ // instrumentedness of overridden weak aliases.
+ if (auto F = dyn_cast<Function>(GA->getBaseObject())) {
+ bool GAInst = isInstrumented(GA), FInst = isInstrumented(F);
+ if (GAInst && FInst) {
+ addGlobalNamePrefix(GA);
+ } else if (GAInst != FInst) {
+ // Non-instrumented alias of an instrumented function, or vice versa.
+ // Replace the alias with a native-ABI wrapper of the aliasee. The pass
+ // below will take care of instrumenting it.
+ Function *NewF =
+ buildWrapperFunction(F, "", GA->getLinkage(), F->getFunctionType());
+ GA->replaceAllUsesWith(ConstantExpr::getBitCast(NewF, GA->getType()));
+ NewF->takeName(GA);
+ GA->eraseFromParent();
+ FnsToInstrument.push_back(NewF);
+ }
+ }
+ }
+
AttrBuilder B;
B.addAttribute(Attribute::ReadOnly).addAttribute(Attribute::ReadNone);
ReadOnlyNoneAttrs = AttributeSet::get(*Ctx, AttributeSet::FunctionIndex, B);
Function &F = **i;
FunctionType *FT = F.getFunctionType();
- if (FT->getNumParams() == 0 && !FT->isVarArg() &&
- FT->getReturnType()->isVoidTy())
- continue;
+ bool IsZeroArgsVoidRet = (FT->getNumParams() == 0 && !FT->isVarArg() &&
+ FT->getReturnType()->isVoidTy());
if (isInstrumented(&F)) {
- if (getInstrumentedABI() == IA_Args) {
+ // Instrumented functions get a 'dfs$' prefix. This allows us to more
+ // easily identify cases of mismatching ABIs.
+ if (getInstrumentedABI() == IA_Args && !IsZeroArgsVoidRet) {
FunctionType *NewFT = getArgsFunctionType(FT);
Function *NewF = Function::Create(NewFT, F.getLinkage(), "", &M);
NewF->copyAttributesFrom(&F);
}
NewF->getBasicBlockList().splice(NewF->begin(), F.getBasicBlockList());
- for (Function::use_iterator ui = F.use_begin(), ue = F.use_end();
- ui != ue;) {
- BlockAddress *BA = dyn_cast<BlockAddress>(ui.getUse().getUser());
- ++ui;
+ for (Function::user_iterator UI = F.user_begin(), UE = F.user_end();
+ UI != UE;) {
+ BlockAddress *BA = dyn_cast<BlockAddress>(*UI);
+ ++UI;
if (BA) {
BA->replaceAllUsesWith(
BlockAddress::get(NewF, BA->getBasicBlock()));
NewF->takeName(&F);
F.eraseFromParent();
*i = NewF;
+ addGlobalNamePrefix(NewF);
+ } else {
+ addGlobalNamePrefix(&F);
}
- // Hopefully, nobody will try to indirectly call a vararg
- // function... yet.
- } else if (FT->isVarArg()) {
- UnwrappedFnMap[&F] = &F;
- *i = 0;
- } else {
+ } 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
// WrapperKind is done in the second pass below.
FunctionType *NewFT = getInstrumentedABI() == IA_Args
? getArgsFunctionType(FT)
: FT;
- Function *NewF =
- Function::Create(NewFT, GlobalValue::LinkOnceODRLinkage,
- std::string("dfsw$") + F.getName(), &M);
- NewF->copyAttributesFrom(&F);
- NewF->removeAttributes(
- AttributeSet::ReturnIndex,
- AttributeFuncs::typeIncompatible(NewFT->getReturnType(),
- AttributeSet::ReturnIndex));
+ Function *NewF = buildWrapperFunction(
+ &F, std::string("dfsw$") + std::string(F.getName()),
+ GlobalValue::LinkOnceODRLinkage, NewFT);
if (getInstrumentedABI() == IA_TLS)
- NewF->removeAttributes(AttributeSet::FunctionIndex,
- ReadOnlyNoneAttrs);
-
- 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);
+ NewF->removeAttributes(AttributeSet::FunctionIndex, ReadOnlyNoneAttrs);
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;
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;
}
}
// DFSanVisitor may create new basic blocks, which confuses df_iterator.
// Build a copy of the list before iterating over it.
- llvm::SmallVector<BasicBlock *, 4> BBList;
- std::copy(df_begin(&(*i)->getEntryBlock()), df_end(&(*i)->getEntryBlock()),
- std::back_inserter(BBList));
+ llvm::SmallVector<BasicBlock *, 4> BBList(
+ depth_first(&(*i)->getEntryBlock()));
for (llvm::SmallVector<BasicBlock *, 4>::iterator i = BBList.begin(),
e = BBList.end();
val, DFSF.getShadow(i->first->getIncomingValue(val)));
}
}
+
+ // -dfsan-debug-nonzero-labels will split the CFG in all kinds of crazy
+ // places (i.e. instructions in basic blocks we haven't even begun visiting
+ // yet). To make our life easier, do this work in a pass after the main
+ // instrumentation.
+ if (ClDebugNonzeroLabels) {
+ for (Value *V : DFSF.NonZeroChecks) {
+ Instruction *Pos;
+ 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(V, DFSF.DFS.ZeroShadow);
+ BranchInst *BI = cast<BranchInst>(SplitBlockAndInsertIfThen(
+ Ne, Pos, /*Unreachable=*/false, ColdCallWeights));
+ IRBuilder<> ThenIRB(BI);
+ ThenIRB.CreateCall(DFSF.DFS.DFSanNonzeroLabelFn);
+ }
+ }
}
return false;
break;
}
}
+ NonZeroChecks.push_back(Shadow);
} else {
Shadow = DFS.ZeroShadow;
}
// 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);
- Instruction *NeInst = dyn_cast<Instruction>(Ne);
- if (NeInst) {
+ 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(
- NeInst, /*Unreachable=*/ false, ColdCallWeights));
+ Ne, Pos, /*Unreachable=*/false, DFS.ColdCallWeights, &DT));
IRBuilder<> ThenIRB(BI);
- CallInst *Call = ThenIRB.CreateCall2(DFSanUnionFn, V1, V2);
+ 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(ShadowTy, 2, "", Tail->begin());
+ PHINode *Phi = PHINode::Create(DFS.ShadowTy, 2, "", Tail->begin());
Phi->addIncoming(Call, Call->getParent());
- Phi->addIncoming(ZeroShadow, Head);
- Pos = Phi;
- return Phi;
+ Phi->addIncoming(V1, Head);
+
+ CCS.Block = Tail;
+ CCS.Shadow = Phi;
+ }
+
+ std::set<Value *> UnionElems;
+ if (V1Elems != ShadowElements.end()) {
+ UnionElems = V1Elems->second;
} else {
- assert(0 && "todo");
- return 0;
+ 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
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;
}
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.
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);
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();
Align = 1;
}
IRBuilder<> IRB(&LI);
- Value *LoadedShadow =
- DFSF.loadShadow(LI.getPointerOperand(), Size, Align, &LI);
- Value *PtrShadow = DFSF.getShadow(LI.getPointerOperand());
- DFSF.setShadow(&LI, DFSF.DFS.combineShadows(LoadedShadow, PtrShadow, &LI));
+ Value *Shadow = DFSF.loadShadow(LI.getPointerOperand(), Size, Align, &LI);
+ if (ClCombinePointerLabelsOnLoad) {
+ Value *PtrShadow = DFSF.getShadow(LI.getPointerOperand());
+ Shadow = DFSF.combineShadows(Shadow, PtrShadow, &LI);
+ }
+ if (Shadow != DFSF.DFS.ZeroShadow)
+ DFSF.NonZeroChecks.push_back(Shadow);
+
+ DFSF.setShadow(&LI, Shadow);
}
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();
} else {
Align = 1;
}
- DFSF.storeShadow(SI.getPointerOperand(), Size, Align,
- DFSF.getShadow(SI.getValueOperand()), &SI);
+
+ Value* Shadow = DFSF.getShadow(SI.getValueOperand());
+ if (ClCombinePointerLabelsOnStore) {
+ Value *PtrShadow = DFSF.getShadow(SI.getPointerOperand());
+ Shadow = DFSF.combineShadows(Shadow, PtrShadow, &SI);
+ }
+ DFSF.storeShadow(SI.getPointerOperand(), Size, Align, Shadow, &SI);
}
void DFSanVisitor::visitBinaryOperator(BinaryOperator &BO) {
void DFSanVisitor::visitAllocaInst(AllocaInst &I) {
bool AllLoadsStores = true;
- for (Instruction::use_iterator i = I.use_begin(), e = I.use_end(); i != e;
- ++i) {
- if (isa<LoadInst>(*i))
+ for (User *U : I.users()) {
+ if (isa<LoadInst>(U))
continue;
- if (StoreInst *SI = dyn_cast<StoreInst>(*i)) {
+ if (StoreInst *SI = dyn_cast<StoreInst>(U)) {
if (SI->getPointerOperand() == &I)
continue;
}
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) {
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));
}
}
+void DFSanVisitor::visitMemSetInst(MemSetInst &I) {
+ IRBuilder<> IRB(&I);
+ Value *ValShadow = DFSF.getShadow(I.getValue());
+ IRB.CreateCall3(
+ DFSF.DFS.DFSanSetLabelFn, ValShadow,
+ IRB.CreateBitCast(I.getDest(), Type::getInt8PtrTy(*DFSF.DFS.Ctx)),
+ IRB.CreateZExtOrTrunc(I.getLength(), DFSF.DFS.IntptrTy));
+}
+
void DFSanVisitor::visitMemTransferInst(MemTransferInst &I) {
IRBuilder<> IRB(&I);
Value *DestShadow = DFSF.DFS.getShadowAddress(I.getDest(), &I);
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 =
std::vector<Value *> Args;
CallSite::arg_iterator i = CS.arg_begin();
- for (unsigned n = FT->getNumParams(); n != 0; ++i, --n)
- Args.push_back(*i);
+ for (unsigned n = FT->getNumParams(); n != 0; ++i, --n) {
+ Type *T = (*i)->getType();
+ FunctionType *ParamFT;
+ if (isa<PointerType>(T) &&
+ (ParamFT = dyn_cast<FunctionType>(
+ cast<PointerType>(T)->getElementType()))) {
+ std::string TName = "dfst";
+ TName += utostr(FT->getNumParams() - n);
+ TName += "$";
+ TName += F->getName();
+ Constant *T = DFSF.DFS.getOrBuildTrampolineFunction(ParamFT, TName);
+ Args.push_back(T);
+ Args.push_back(
+ IRB.CreateBitCast(*i, Type::getInt8PtrTy(*DFSF.DFS.Ctx)));
+ } else {
+ Args.push_back(*i);
+ }
+ }
i = CS.arg_begin();
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 =
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());
}
}
- Instruction *Next = 0;
+ Instruction *Next = nullptr;
if (!CS.getType()->isVoidTy()) {
if (InvokeInst *II = dyn_cast<InvokeInst>(CS.getInstruction())) {
if (II->getNormalDest()->getSinglePredecessor()) {
Next = II->getNormalDest()->begin();
} else {
BasicBlock *NewBB =
- SplitEdge(II->getParent(), II->getNormalDest(), &DFSF.DFS);
+ SplitEdge(II->getParent(), II->getNormalDest(), &DFSF.DT);
Next = NewBB->begin();
}
} else {
LoadInst *LI = NextIRB.CreateLoad(DFSF.getRetvalTLS());
DFSF.SkipInsts.insert(LI);
DFSF.setShadow(CS.getInstruction(), LI);
+ DFSF.NonZeroChecks.push_back(LI);
}
}
ExtractValueInst::Create(NewCS.getInstruction(), 1, "", Next);
DFSF.SkipInsts.insert(ExShadow);
DFSF.setShadow(ExVal, ExShadow);
+ DFSF.NonZeroChecks.push_back(ExShadow);
CS.getInstruction()->replaceAllUsesWith(ExVal);
}