X-Git-Url: http://plrg.eecs.uci.edu/git/?p=oota-llvm.git;a=blobdiff_plain;f=lib%2FIR%2FVerifier.cpp;h=1557ca0dbb78821bfecf64e093cf521ce7da1e35;hp=c47a0ff1d80dd93fb4e248971c455ad2d78ae63e;hb=fad8ff3134d32996a6ce6beba7ee70d5e1c51d8b;hpb=f84a6504a617cf3d632a2bee7d7d93d9c6d78ccf diff --git a/lib/IR/Verifier.cpp b/lib/IR/Verifier.cpp index c47a0ff1d80..1557ca0dbb7 100644 --- a/lib/IR/Verifier.cpp +++ b/lib/IR/Verifier.cpp @@ -39,13 +39,13 @@ // only by the unwind edge of an invoke instruction. // * A landingpad instruction must be the first non-PHI instruction in the // block. -// * All landingpad instructions must use the same personality function with -// the same function. +// * Landingpad instructions must be in a function with a personality function. // * All other things that are tested by asserts spread about the code... // //===----------------------------------------------------------------------===// #include "llvm/IR/Verifier.h" +#include "llvm/ADT/MapVector.h" #include "llvm/ADT/STLExtras.h" #include "llvm/ADT/SetVector.h" #include "llvm/ADT/SmallPtrSet.h" @@ -92,6 +92,16 @@ struct VerifierSupport { : OS(OS), M(nullptr), Broken(false) {} private: + template void Write(const ilist_iterator &I) { + Write(&*I); + } + + void Write(const Module *M) { + if (!M) + return; + OS << "; ModuleID = '" << M->getModuleIdentifier() << "'\n"; + } + void Write(const Value *V) { if (!V) return; @@ -102,10 +112,8 @@ private: OS << '\n'; } } - void Write(const CallSite *CS) { - if (!CS) - return; - Write(CS->getInstruction()); + void Write(ImmutableCallSite CS) { + Write(CS.getInstruction()); } void Write(const Metadata *MD) { @@ -138,6 +146,11 @@ private: OS << *C; } + template void Write(ArrayRef Vs) { + for (const T &V : Vs) + Write(V); + } + template void WriteTs(const T1 &V1, const Ts &... Vs) { Write(V1); @@ -186,17 +199,30 @@ class Verifier : public InstVisitor, VerifierSupport { /// \brief Track unresolved string-based type references. SmallDenseMap UnresolvedTypeRefs; - /// \brief Whether we've seen a call to @llvm.frameescape in this function + /// \brief The result type for a landingpad. + Type *LandingPadResultTy; + + /// \brief Whether we've seen a call to @llvm.localescape in this function /// already. bool SawFrameEscape; - /// Stores the count of how many objects were passed to llvm.frameescape for a - /// given function and the largest index passed to llvm.framerecover. + /// Stores the count of how many objects were passed to llvm.localescape for a + /// given function and the largest index passed to llvm.localrecover. DenseMap> FrameEscapeInfo; + // Maps catchswitches and cleanuppads that unwind to siblings to the + // terminators that indicate the unwind, used to detect cycles therein. + MapVector SiblingFuncletInfo; + + /// Cache of constants visited in search of ConstantExprs. + SmallPtrSet ConstantExprVisited; + + void checkAtomicMemAccessSize(const Module *M, Type *Ty, + const Instruction *I); public: explicit Verifier(raw_ostream &OS) - : VerifierSupport(OS), Context(nullptr), SawFrameEscape(false) {} + : VerifierSupport(OS), Context(nullptr), LandingPadResultTy(nullptr), + SawFrameEscape(false) {} bool verify(const Function &F) { M = F.getParent(); @@ -229,8 +255,11 @@ public: Broken = false; // FIXME: We strip const here because the inst visitor strips const. visit(const_cast(F)); + verifySiblingFuncletUnwinds(); InstsInThisBlock.clear(); + LandingPadResultTy = nullptr; SawFrameEscape = false; + SiblingFuncletInfo.clear(); return !Broken; } @@ -299,12 +328,12 @@ private: void visitFunction(const Function &F); void visitBasicBlock(BasicBlock &BB); void visitRangeMetadata(Instruction& I, MDNode* Range, Type* Ty); + void visitDereferenceableMetadata(Instruction& I, MDNode* MD); template bool isValidMetadataArray(const MDTuple &N); #define HANDLE_SPECIALIZED_MDNODE_LEAF(CLASS) void visit##CLASS(const CLASS &N); #include "llvm/IR/Metadata.def" void visitDIScope(const DIScope &N); - void visitDIDerivedTypeBase(const DIDerivedTypeBase &N); void visitDIVariable(const DIVariable &N); void visitDILexicalBlockBase(const DILexicalBlockBase &N); void visitDITemplateParameter(const DITemplateParameter &N); @@ -372,7 +401,7 @@ private: void visitSelectInst(SelectInst &SI); void visitUserOp1(Instruction &I); void visitUserOp2(Instruction &I) { visitUserOp1(I); } - void visitIntrinsicFunctionCall(Intrinsic::ID ID, CallSite CS); + void visitIntrinsicCallSite(Intrinsic::ID ID, CallSite CS); template void visitDbgIntrinsic(StringRef Kind, DbgIntrinsicTy &DII); void visitAtomicCmpXchgInst(AtomicCmpXchgInst &CXI); @@ -381,7 +410,14 @@ private: void visitAllocaInst(AllocaInst &AI); void visitExtractValueInst(ExtractValueInst &EVI); void visitInsertValueInst(InsertValueInst &IVI); + void visitEHPadPredecessors(Instruction &I); void visitLandingPadInst(LandingPadInst &LPI); + void visitCatchPadInst(CatchPadInst &CPI); + void visitCatchReturnInst(CatchReturnInst &CatchReturn); + void visitCleanupPadInst(CleanupPadInst &CPI); + void visitFuncletPadInst(FuncletPadInst &FPI); + void visitCatchSwitchInst(CatchSwitchInst &CatchSwitch); + void visitCleanupReturnInst(CleanupReturnInst &CRI); void VerifyCallSite(CallSite CS); void verifyMustTailCall(CallInst &CI); @@ -401,15 +437,16 @@ private: void VerifyFunctionMetadata( const SmallVector, 4> MDs); - void VerifyConstantExprBitcastType(const ConstantExpr *CE); + void visitConstantExprsRecursively(const Constant *EntryC); + void visitConstantExpr(const ConstantExpr *CE); void VerifyStatepoint(ImmutableCallSite CS); void verifyFrameRecoverIndices(); + void verifySiblingFuncletUnwinds(); // Module-level debug info verification... void verifyTypeRefs(); template - void verifyBitPieceExpression(const DbgInfoIntrinsic &I, - const MapTy &TypeRefs); + void verifyDIExpression(const DbgInfoIntrinsic &I, const MapTy &TypeRefs); void visitUnresolvedTypeRef(const MDString *S, const MDNode *N); }; } // End anonymous namespace @@ -440,6 +477,9 @@ void Verifier::visitGlobalValue(const GlobalValue &GV) { Assert(GVar && GVar->getValueType()->isArrayTy(), "Only global arrays can have appending linkage!", GVar); } + + if (GV.isDeclarationForLinker()) + Assert(!GV.hasComdat(), "Declaration may not be in a Comdat!", &GV); } void Verifier::visitGlobalVariable(const GlobalVariable &GV) { @@ -523,25 +563,7 @@ void Verifier::visitGlobalVariable(const GlobalVariable &GV) { } // Walk any aggregate initializers looking for bitcasts between address spaces - SmallPtrSet Visited; - SmallVector WorkStack; - WorkStack.push_back(cast(GV.getInitializer())); - - while (!WorkStack.empty()) { - const Value *V = WorkStack.pop_back_val(); - if (!Visited.insert(V).second) - continue; - - if (const User *U = dyn_cast(V)) { - WorkStack.append(U->op_begin(), U->op_end()); - } - - if (const ConstantExpr *CE = dyn_cast(V)) { - VerifyConstantExprBitcastType(CE); - if (Broken) - return; - } - } + visitConstantExprsRecursively(GV.getInitializer()); visitGlobalValue(GV); } @@ -555,7 +577,8 @@ void Verifier::visitAliaseeSubExpr(const GlobalAlias &GA, const Constant &C) { void Verifier::visitAliaseeSubExpr(SmallPtrSetImpl &Visited, const GlobalAlias &GA, const Constant &C) { if (const auto *GV = dyn_cast(&C)) { - Assert(!GV->isDeclaration(), "Alias must point to a definition", &GA); + Assert(!GV->isDeclarationForLinker(), "Alias must point to a definition", + &GA); if (const auto *GA2 = dyn_cast(GV)) { Assert(Visited.insert(GA2).second, "Aliases cannot form a cycle", &GA); @@ -570,7 +593,7 @@ void Verifier::visitAliaseeSubExpr(SmallPtrSetImpl &Visited, } if (const auto *CE = dyn_cast(&C)) - VerifyConstantExprBitcastType(CE); + visitConstantExprsRecursively(CE); for (const Use &U : C.operands()) { Value *V = &*U; @@ -778,39 +801,10 @@ void Verifier::visitDIBasicType(const DIBasicType &N) { "invalid tag", &N); } -void Verifier::visitDIDerivedTypeBase(const DIDerivedTypeBase &N) { +void Verifier::visitDIDerivedType(const DIDerivedType &N) { // Common scope checks. visitDIScope(N); - Assert(isScopeRef(N, N.getScope()), "invalid scope", &N, N.getScope()); - Assert(isTypeRef(N, N.getBaseType()), "invalid base type", &N, - N.getBaseType()); - - // FIXME: Sink this into the subclass verifies. - if (!N.getFile() || N.getFile()->getFilename().empty()) { - // Check whether the filename is allowed to be empty. - uint16_t Tag = N.getTag(); - Assert( - Tag == dwarf::DW_TAG_const_type || Tag == dwarf::DW_TAG_volatile_type || - Tag == dwarf::DW_TAG_pointer_type || - Tag == dwarf::DW_TAG_ptr_to_member_type || - Tag == dwarf::DW_TAG_reference_type || - Tag == dwarf::DW_TAG_rvalue_reference_type || - Tag == dwarf::DW_TAG_restrict_type || - Tag == dwarf::DW_TAG_array_type || - Tag == dwarf::DW_TAG_enumeration_type || - Tag == dwarf::DW_TAG_subroutine_type || - Tag == dwarf::DW_TAG_inheritance || Tag == dwarf::DW_TAG_friend || - Tag == dwarf::DW_TAG_structure_type || - Tag == dwarf::DW_TAG_member || Tag == dwarf::DW_TAG_typedef, - "derived/composite type requires a filename", &N, N.getFile()); - } -} - -void Verifier::visitDIDerivedType(const DIDerivedType &N) { - // Common derived type checks. - visitDIDerivedTypeBase(N); - Assert(N.getTag() == dwarf::DW_TAG_typedef || N.getTag() == dwarf::DW_TAG_pointer_type || N.getTag() == dwarf::DW_TAG_ptr_to_member_type || @@ -827,6 +821,10 @@ void Verifier::visitDIDerivedType(const DIDerivedType &N) { Assert(isTypeRef(N, N.getExtraData()), "invalid pointer to member type", &N, N.getExtraData()); } + + Assert(isScopeRef(N, N.getScope()), "invalid scope", &N, N.getScope()); + Assert(isTypeRef(N, N.getBaseType()), "invalid base type", &N, + N.getBaseType()); } static bool hasConflictingReferenceFlags(unsigned Flags) { @@ -844,27 +842,34 @@ void Verifier::visitTemplateParams(const MDNode &N, const Metadata &RawParams) { } void Verifier::visitDICompositeType(const DICompositeType &N) { - // Common derived type checks. - visitDIDerivedTypeBase(N); + // Common scope checks. + visitDIScope(N); Assert(N.getTag() == dwarf::DW_TAG_array_type || N.getTag() == dwarf::DW_TAG_structure_type || N.getTag() == dwarf::DW_TAG_union_type || N.getTag() == dwarf::DW_TAG_enumeration_type || - N.getTag() == dwarf::DW_TAG_subroutine_type || N.getTag() == dwarf::DW_TAG_class_type, "invalid tag", &N); + Assert(isScopeRef(N, N.getScope()), "invalid scope", &N, N.getScope()); + Assert(isTypeRef(N, N.getBaseType()), "invalid base type", &N, + N.getBaseType()); + Assert(!N.getRawElements() || isa(N.getRawElements()), "invalid composite elements", &N, N.getRawElements()); Assert(isTypeRef(N, N.getRawVTableHolder()), "invalid vtable holder", &N, N.getRawVTableHolder()); - Assert(!N.getRawElements() || isa(N.getRawElements()), - "invalid composite elements", &N, N.getRawElements()); Assert(!hasConflictingReferenceFlags(N.getFlags()), "invalid reference flags", &N); if (auto *Params = N.getRawTemplateParams()) visitTemplateParams(N, *Params); + + if (N.getTag() == dwarf::DW_TAG_class_type || + N.getTag() == dwarf::DW_TAG_union_type) { + Assert(N.getFile() && !N.getFile()->getFilename().empty(), + "class/union requires a filename", &N, N.getFile()); + } } void Verifier::visitDISubroutineType(const DISubroutineType &N) { @@ -884,6 +889,7 @@ void Verifier::visitDIFile(const DIFile &N) { } void Verifier::visitDICompileUnit(const DICompileUnit &N) { + Assert(N.isDistinct(), "compile units must be distinct", &N); Assert(N.getTag() == dwarf::DW_TAG_compile_unit, "invalid tag", &N); // Don't bother verifying the compilation directory or producer string @@ -927,6 +933,12 @@ void Verifier::visitDICompileUnit(const DICompileUnit &N) { Op); } } + if (auto *Array = N.getRawMacros()) { + Assert(isa(Array), "invalid macro list", &N, Array); + for (Metadata *Op : N.getMacros()->operands()) { + Assert(Op && isa(Op), "invalid macro ref", &N, Op); + } + } } void Verifier::visitDISubprogram(const DISubprogram &N) { @@ -936,13 +948,6 @@ void Verifier::visitDISubprogram(const DISubprogram &N) { Assert(isa(T), "invalid subroutine type", &N, T); Assert(isTypeRef(N, N.getRawContainingType()), "invalid containing type", &N, N.getRawContainingType()); - if (auto *RawF = N.getRawFunction()) { - auto *FMD = dyn_cast(RawF); - auto *F = FMD ? FMD->getValue() : nullptr; - auto *FT = F ? dyn_cast(F->getType()) : nullptr; - Assert(F && FT && isa(FT->getElementType()), - "invalid function", &N, F, FT); - } if (auto *Params = N.getRawTemplateParams()) visitTemplateParams(N, *Params); if (auto *S = N.getRawDeclaration()) { @@ -960,40 +965,8 @@ void Verifier::visitDISubprogram(const DISubprogram &N) { Assert(!hasConflictingReferenceFlags(N.getFlags()), "invalid reference flags", &N); - auto *F = N.getFunction(); - if (!F) - return; - - // Check that all !dbg attachments lead to back to N (or, at least, another - // subprogram that describes the same function). - // - // FIXME: Check this incrementally while visiting !dbg attachments. - // FIXME: Only check when N is the canonical subprogram for F. - SmallPtrSet Seen; - for (auto &BB : *F) - for (auto &I : BB) { - // Be careful about using DILocation here since we might be dealing with - // broken code (this is the Verifier after all). - DILocation *DL = - dyn_cast_or_null(I.getDebugLoc().getAsMDNode()); - if (!DL) - continue; - if (!Seen.insert(DL).second) - continue; - - DILocalScope *Scope = DL->getInlinedAtScope(); - if (Scope && !Seen.insert(Scope).second) - continue; - - DISubprogram *SP = Scope ? Scope->getSubprogram() : nullptr; - if (SP && !Seen.insert(SP).second) - continue; - - // FIXME: Once N is canonical, check "SP == &N". - Assert(SP->describes(F), - "!dbg attachment points at wrong subprogram for function", &N, F, - &I, DL, Scope, SP); - } + if (N.isDefinition()) + Assert(N.isDistinct(), "subprogram definitions must be distinct", &N); } void Verifier::visitDILexicalBlockBase(const DILexicalBlockBase &N) { @@ -1019,6 +992,35 @@ void Verifier::visitDINamespace(const DINamespace &N) { Assert(isa(S), "invalid scope ref", &N, S); } +void Verifier::visitDIMacro(const DIMacro &N) { + Assert(N.getMacinfoType() == dwarf::DW_MACINFO_define || + N.getMacinfoType() == dwarf::DW_MACINFO_undef, + "invalid macinfo type", &N); + Assert(!N.getName().empty(), "anonymous macro", &N); + if (!N.getValue().empty()) { + assert(N.getValue().data()[0] != ' ' && "Macro value has a space prefix"); + } +} + +void Verifier::visitDIMacroFile(const DIMacroFile &N) { + Assert(N.getMacinfoType() == dwarf::DW_MACINFO_start_file, + "invalid macinfo type", &N); + if (auto *F = N.getRawFile()) + Assert(isa(F), "invalid file", &N, F); + + if (auto *Array = N.getRawElements()) { + Assert(isa(Array), "invalid macro list", &N, Array); + for (Metadata *Op : N.getElements()->operands()) { + Assert(Op && isa(Op), "invalid macro ref", &N, Op); + } + } +} + +void Verifier::visitDIModule(const DIModule &N) { + Assert(N.getTag() == dwarf::DW_TAG_module, "invalid tag", &N); + Assert(!N.getName().empty(), "anonymous module", &N); +} + void Verifier::visitDITemplateParameter(const DITemplateParameter &N) { Assert(isTypeRef(N, N.getType()), "invalid type ref", &N, N.getType()); } @@ -1069,9 +1071,7 @@ void Verifier::visitDILocalVariable(const DILocalVariable &N) { // Checks common to all variables. visitDIVariable(N); - Assert(N.getTag() == dwarf::DW_TAG_auto_variable || - N.getTag() == dwarf::DW_TAG_arg_variable, - "invalid tag", &N); + Assert(N.getTag() == dwarf::DW_TAG_variable, "invalid tag", &N); Assert(N.getRawScope() && isa(N.getRawScope()), "local variable requires a valid scope", &N, N.getRawScope()); } @@ -1267,7 +1267,11 @@ void Verifier::VerifyAttributeTypes(AttributeSet Attrs, unsigned Idx, I->getKindAsEnum() == Attribute::Cold || I->getKindAsEnum() == Attribute::OptimizeNone || I->getKindAsEnum() == Attribute::JumpTable || - I->getKindAsEnum() == Attribute::Convergent) { + I->getKindAsEnum() == Attribute::Convergent || + I->getKindAsEnum() == Attribute::ArgMemOnly || + I->getKindAsEnum() == Attribute::NoRecurse || + I->getKindAsEnum() == Attribute::InaccessibleMemOnly || + I->getKindAsEnum() == Attribute::InaccessibleMemOrArgMemOnly) { if (!isFunction) { CheckFailed("Attribute '" + I->getAsString() + "' only applies to functions!", V); @@ -1358,7 +1362,7 @@ void Verifier::VerifyParameterAttrs(AttributeSet Attrs, unsigned Idx, Type *Ty, V); if (PointerType *PTy = dyn_cast(Ty)) { - SmallPtrSet Visited; + SmallPtrSet Visited; if (!PTy->getElementType()->isSized(&Visited)) { Assert(!Attrs.hasAttribute(Idx, Attribute::ByVal) && !Attrs.hasAttribute(Idx, Attribute::InAlloca), @@ -1437,6 +1441,18 @@ void Verifier::VerifyFunctionAttrs(FunctionType *FT, AttributeSet Attrs, Attrs.hasAttribute(AttributeSet::FunctionIndex, Attribute::ReadOnly)), "Attributes 'readnone and readonly' are incompatible!", V); + Assert( + !(Attrs.hasAttribute(AttributeSet::FunctionIndex, Attribute::ReadNone) && + Attrs.hasAttribute(AttributeSet::FunctionIndex, + Attribute::InaccessibleMemOrArgMemOnly)), + "Attributes 'readnone and inaccessiblemem_or_argmemonly' are incompatible!", V); + + Assert( + !(Attrs.hasAttribute(AttributeSet::FunctionIndex, Attribute::ReadNone) && + Attrs.hasAttribute(AttributeSet::FunctionIndex, + Attribute::InaccessibleMemOnly)), + "Attributes 'readnone and inaccessiblememonly' are incompatible!", V); + Assert( !(Attrs.hasAttribute(AttributeSet::FunctionIndex, Attribute::NoInline) && Attrs.hasAttribute(AttributeSet::FunctionIndex, @@ -1494,7 +1510,35 @@ void Verifier::VerifyFunctionMetadata( } } -void Verifier::VerifyConstantExprBitcastType(const ConstantExpr *CE) { +void Verifier::visitConstantExprsRecursively(const Constant *EntryC) { + if (!ConstantExprVisited.insert(EntryC).second) + return; + + SmallVector Stack; + Stack.push_back(EntryC); + + while (!Stack.empty()) { + const Constant *C = Stack.pop_back_val(); + + // Check this constant expression. + if (const auto *CE = dyn_cast(C)) + visitConstantExpr(CE); + + // Visit all sub-expressions. + for (const Use &U : C->operands()) { + const auto *OpC = dyn_cast(U); + if (!OpC) + continue; + if (isa(OpC)) + continue; // Global values get visited separately. + if (!ConstantExprVisited.insert(OpC).second) + continue; + Stack.push_back(OpC); + } + } +} + +void Verifier::visitConstantExpr(const ConstantExpr *CE) { if (CE->getOpcode() != Instruction::BitCast) return; @@ -1525,8 +1569,9 @@ void Verifier::VerifyStatepoint(ImmutableCallSite CS) { const Instruction &CI = *CS.getInstruction(); - Assert(!CS.doesNotAccessMemory() && !CS.onlyReadsMemory(), - "gc.statepoint must read and write memory to preserve " + Assert(!CS.doesNotAccessMemory() && !CS.onlyReadsMemory() && + !CS.onlyAccessesArgMemory(), + "gc.statepoint must read and write all memory to preserve " "reordering restrictions required by safepoint semantics", &CI); @@ -1546,17 +1591,11 @@ void Verifier::VerifyStatepoint(ImmutableCallSite CS) { &CI); const Value *Target = CS.getArgument(2); - const PointerType *PT = dyn_cast(Target->getType()); + auto *PT = dyn_cast(Target->getType()); Assert(PT && PT->getElementType()->isFunctionTy(), "gc.statepoint callee must be of function pointer type", &CI, Target); FunctionType *TargetFuncType = cast(PT->getElementType()); - if (NumPatchBytes) - Assert(isa(Target->stripPointerCasts()), - "gc.statepoint must have null as call target if number of patchable " - "bytes is non zero", - &CI); - const Value *NumCallArgsV = CS.getArgument(3); Assert(isa(NumCallArgsV), "gc.statepoint number of arguments to underlying call " @@ -1634,14 +1673,14 @@ void Verifier::VerifyStatepoint(ImmutableCallSite CS) { const CallInst *Call = dyn_cast(U); Assert(Call, "illegal use of statepoint token", &CI, U); if (!Call) continue; - Assert(isGCRelocate(Call) || isGCResult(Call), + Assert(isa(Call) || isGCResult(Call), "gc.result or gc.relocate are the only value uses" "of a gc.statepoint", &CI, U); if (isGCResult(Call)) { Assert(Call->getArgOperand(0) == &CI, "gc.result connected to wrong gc.statepoint", &CI, Call); - } else if (isGCRelocate(Call)) { + } else if (isa(Call)) { Assert(Call->getArgOperand(0) == &CI, "gc.relocate connected to wrong gc.statepoint", &CI, Call); } @@ -1663,13 +1702,66 @@ void Verifier::verifyFrameRecoverIndices() { unsigned EscapedObjectCount = Counts.second.first; unsigned MaxRecoveredIndex = Counts.second.second; Assert(MaxRecoveredIndex <= EscapedObjectCount, - "all indices passed to llvm.framerecover must be less than the " - "number of arguments passed ot llvm.frameescape in the parent " + "all indices passed to llvm.localrecover must be less than the " + "number of arguments passed ot llvm.localescape in the parent " "function", F); } } +static Instruction *getSuccPad(TerminatorInst *Terminator) { + BasicBlock *UnwindDest; + if (auto *II = dyn_cast(Terminator)) + UnwindDest = II->getUnwindDest(); + else if (auto *CSI = dyn_cast(Terminator)) + UnwindDest = CSI->getUnwindDest(); + else + UnwindDest = cast(Terminator)->getUnwindDest(); + return UnwindDest->getFirstNonPHI(); +} + +void Verifier::verifySiblingFuncletUnwinds() { + SmallPtrSet Visited; + SmallPtrSet Active; + for (const auto &Pair : SiblingFuncletInfo) { + Instruction *PredPad = Pair.first; + if (Visited.count(PredPad)) + continue; + Active.insert(PredPad); + TerminatorInst *Terminator = Pair.second; + do { + Instruction *SuccPad = getSuccPad(Terminator); + if (Active.count(SuccPad)) { + // Found a cycle; report error + Instruction *CyclePad = SuccPad; + SmallVector CycleNodes; + do { + CycleNodes.push_back(CyclePad); + TerminatorInst *CycleTerminator = SiblingFuncletInfo[CyclePad]; + if (CycleTerminator != CyclePad) + CycleNodes.push_back(CycleTerminator); + CyclePad = getSuccPad(CycleTerminator); + } while (CyclePad != SuccPad); + Assert(false, "EH pads can't handle each other's exceptions", + ArrayRef(CycleNodes)); + } + // Don't re-walk a node we've already checked + if (!Visited.insert(SuccPad).second) + break; + // Walk to this successor if it has a map entry. + PredPad = SuccPad; + auto TermI = SiblingFuncletInfo.find(PredPad); + if (TermI == SiblingFuncletInfo.end()) + break; + Terminator = TermI->second; + Active.insert(PredPad); + } while (true); + // Each node only has one successor, so we've walked all the active + // nodes' successors. + Active.clear(); + } +} + // visitFunction - Verify that a function is ok. // void Verifier::visitFunction(const Function &F) { @@ -1735,17 +1827,33 @@ void Verifier::visitFunction(const Function &F) { FT->getParamType(i)); Assert(I->getType()->isFirstClassType(), "Function arguments must have first-class types!", I); - if (!isLLVMdotName) + if (!isLLVMdotName) { Assert(!I->getType()->isMetadataTy(), "Function takes metadata but isn't an intrinsic", I, &F); + Assert(!I->getType()->isTokenTy(), + "Function takes token but isn't an intrinsic", I, &F); + } } + if (!isLLVMdotName) + Assert(!F.getReturnType()->isTokenTy(), + "Functions returns a token but isn't an intrinsic", &F); + // Get the function metadata attachments. SmallVector, 4> MDs; F.getAllMetadata(MDs); assert(F.hasMetadata() != MDs.empty() && "Bit out-of-sync"); VerifyFunctionMetadata(MDs); + // Check validity of the personality function + if (F.hasPersonalityFn()) { + auto *Per = dyn_cast(F.getPersonalityFn()->stripPointerCasts()); + if (Per) + Assert(Per->getParent() == F.getParent(), + "Referencing personality function in another module!", + &F, F.getParent(), Per, Per->getParent()); + } + if (F.isMaterializable()) { // Function has a body somewhere we can't see. Assert(MDs.empty(), "unmaterialized function cannot have metadata", &F, @@ -1774,13 +1882,27 @@ void Verifier::visitFunction(const Function &F) { } // Visit metadata attachments. - for (const auto &I : MDs) + for (const auto &I : MDs) { + // Verify that the attachment is legal. + switch (I.first) { + default: + break; + case LLVMContext::MD_dbg: + Assert(isa(I.second), + "function !dbg attachment must be a subprogram", &F, I.second); + break; + } + + // Verify the metadata itself. visitMDNode(*I.second); + } } // If this function is actually an intrinsic, verify that it is only used in // direct call/invokes, never having its "address taken". - if (F.getIntrinsicID()) { + // Only do this if the module is materialized, otherwise we don't have all the + // uses. + if (F.getIntrinsicID() && F.getParent()->isMaterialized()) { const User *U; if (F.hasAddressTaken(&U)) Assert(0, "Invalid user of intrinsic instruction!", U); @@ -1790,6 +1912,44 @@ void Verifier::visitFunction(const Function &F) { (F.isDeclaration() && F.hasExternalLinkage()) || F.hasAvailableExternallyLinkage(), "Function is marked as dllimport, but not external.", &F); + + auto *N = F.getSubprogram(); + if (!N) + return; + + // Check that all !dbg attachments lead to back to N (or, at least, another + // subprogram that describes the same function). + // + // FIXME: Check this incrementally while visiting !dbg attachments. + // FIXME: Only check when N is the canonical subprogram for F. + SmallPtrSet Seen; + for (auto &BB : F) + for (auto &I : BB) { + // Be careful about using DILocation here since we might be dealing with + // broken code (this is the Verifier after all). + DILocation *DL = + dyn_cast_or_null(I.getDebugLoc().getAsMDNode()); + if (!DL) + continue; + if (!Seen.insert(DL).second) + continue; + + DILocalScope *Scope = DL->getInlinedAtScope(); + if (Scope && !Seen.insert(Scope).second) + continue; + + DISubprogram *SP = Scope ? Scope->getSubprogram() : nullptr; + + // Scope and SP could be the same MDNode and we don't want to skip + // validation in that case + if (SP && ((Scope != SP) && !Seen.insert(SP).second)) + continue; + + // FIXME: Once N is canonical, check "SP == &N". + Assert(SP->describes(&F), + "!dbg attachment points at wrong subprogram for function", N, &F, + &I, DL, Scope, SP); + } } // verifyBasicBlock - Verify that a basic block is well formed... @@ -2186,6 +2346,9 @@ void Verifier::visitPHINode(PHINode &PN) { isa(--BasicBlock::iterator(&PN)), "PHI nodes not grouped at top of basic block!", &PN, PN.getParent()); + // Check that a PHI doesn't yield a Token. + Assert(!PN.getType()->isTokenTy(), "PHI nodes cannot have token type!"); + // Check that all of the values of the PHI node have the same type as the // result, and that the incoming blocks are really basic blocks. for (Value *IncValue : PN.incoming_values()) { @@ -2288,15 +2451,43 @@ void Verifier::VerifyCallSite(CallSite CS) { // Verify that there's no metadata unless it's a direct call to an intrinsic. if (CS.getCalledFunction() == nullptr || !CS.getCalledFunction()->getName().startswith("llvm.")) { - for (FunctionType::param_iterator PI = FTy->param_begin(), - PE = FTy->param_end(); PI != PE; ++PI) - Assert(!(*PI)->isMetadataTy(), + for (Type *ParamTy : FTy->params()) { + Assert(!ParamTy->isMetadataTy(), "Function has metadata parameter but isn't an intrinsic", I); + Assert(!ParamTy->isTokenTy(), + "Function has token parameter but isn't an intrinsic", I); + } } + // Verify that indirect calls don't return tokens. + if (CS.getCalledFunction() == nullptr) + Assert(!FTy->getReturnType()->isTokenTy(), + "Return type cannot be token for indirect call!"); + if (Function *F = CS.getCalledFunction()) if (Intrinsic::ID ID = (Intrinsic::ID)F->getIntrinsicID()) - visitIntrinsicFunctionCall(ID, CS); + visitIntrinsicCallSite(ID, CS); + + // Verify that a callsite has at most one "deopt" and one "funclet" operand + // bundle. + bool FoundDeoptBundle = false, FoundFuncletBundle = false; + for (unsigned i = 0, e = CS.getNumOperandBundles(); i < e; ++i) { + OperandBundleUse BU = CS.getOperandBundleAt(i); + uint32_t Tag = BU.getTagID(); + if (Tag == LLVMContext::OB_deopt) { + Assert(!FoundDeoptBundle, "Multiple deopt operand bundles", I); + FoundDeoptBundle = true; + } + if (Tag == LLVMContext::OB_funclet) { + Assert(!FoundFuncletBundle, "Multiple funclet operand bundles", I); + FoundFuncletBundle = true; + Assert(BU.Inputs.size() == 1, + "Expected exactly one funclet bundle operand", I); + Assert(isa(BU.Inputs.front()), + "Funclet bundle operands should correspond to a FuncletPadInst", + I); + } + } visitInstruction(*I); } @@ -2398,10 +2589,12 @@ void Verifier::visitCallInst(CallInst &CI) { void Verifier::visitInvokeInst(InvokeInst &II) { VerifyCallSite(&II); - // Verify that there is a landingpad instruction as the first non-PHI - // instruction of the 'unwind' destination. - Assert(II.getUnwindDest()->isLandingPad(), - "The unwind destination does not have a landingpad instruction!", &II); + // Verify that the first non-PHI instruction of the unwind destination is an + // exception handling instruction. + Assert( + II.getUnwindDest()->isEHPad(), + "The unwind destination does not have an exception handling instruction!", + &II); visitTerminatorInst(II); } @@ -2532,10 +2725,6 @@ void Verifier::visitGetElementPtrInst(GetElementPtrInst &GEP) { Assert(isa(TargetTy), "GEP base pointer is not a vector or a vector of pointers", &GEP); Assert(GEP.getSourceElementType()->isSized(), "GEP into unsized type!", &GEP); - Assert(GEP.getPointerOperandType()->isVectorTy() == - GEP.getType()->isVectorTy(), - "Vector GEP must return a vector value", &GEP); - SmallVector Idxs(GEP.idx_begin(), GEP.idx_end()); Type *ElTy = GetElementPtrInst::getIndexedType(GEP.getSourceElementType(), Idxs); @@ -2545,17 +2734,20 @@ void Verifier::visitGetElementPtrInst(GetElementPtrInst &GEP) { GEP.getResultElementType() == ElTy, "GEP is not of right type for indices!", &GEP, ElTy); - if (GEP.getPointerOperandType()->isVectorTy()) { + if (GEP.getType()->isVectorTy()) { // Additional checks for vector GEPs. - unsigned GepWidth = GEP.getPointerOperandType()->getVectorNumElements(); - Assert(GepWidth == GEP.getType()->getVectorNumElements(), - "Vector GEP result width doesn't match operand's", &GEP); + unsigned GEPWidth = GEP.getType()->getVectorNumElements(); + if (GEP.getPointerOperandType()->isVectorTy()) + Assert(GEPWidth == GEP.getPointerOperandType()->getVectorNumElements(), + "Vector GEP result width doesn't match operand's", &GEP); for (unsigned i = 0, e = Idxs.size(); i != e; ++i) { Type *IndexTy = Idxs[i]->getType(); - Assert(IndexTy->isVectorTy(), "Vector GEP must have vector indices!", - &GEP); - unsigned IndexWidth = IndexTy->getVectorNumElements(); - Assert(IndexWidth == GepWidth, "Invalid GEP index vector width", &GEP); + if (IndexTy->isVectorTy()) { + unsigned IndexWidth = IndexTy->getVectorNumElements(); + Assert(IndexWidth == GEPWidth, "Invalid GEP index vector width", &GEP); + } + Assert(IndexTy->getScalarType()->isIntegerTy(), + "All GEP indices should be of integer type"); } } visitInstruction(GEP); @@ -2615,6 +2807,14 @@ void Verifier::visitRangeMetadata(Instruction& I, } } +void Verifier::checkAtomicMemAccessSize(const Module *M, Type *Ty, + const Instruction *I) { + unsigned Size = M->getDataLayout().getTypeSizeInBits(Ty); + Assert(Size >= 8, "atomic memory access' size must be byte-sized", Ty, I); + Assert(!(Size & (Size - 1)), + "atomic memory access' operand must have a power-of-two size", Ty, I); +} + void Verifier::visitLoadInst(LoadInst &LI) { PointerType *PTy = dyn_cast(LI.getOperand(0)->getType()); Assert(PTy, "Load operand must be a pointer.", &LI); @@ -2626,14 +2826,12 @@ void Verifier::visitLoadInst(LoadInst &LI) { "Load cannot have Release ordering", &LI); Assert(LI.getAlignment() != 0, "Atomic load must specify explicit alignment", &LI); - if (!ElTy->isPointerTy()) { - Assert(ElTy->isIntegerTy(), "atomic load operand must have integer type!", - &LI, ElTy); - unsigned Size = ElTy->getPrimitiveSizeInBits(); - Assert(Size >= 8 && !(Size & (Size - 1)), - "atomic load operand must be power-of-two byte-sized integer", &LI, - ElTy); - } + Assert(ElTy->isIntegerTy() || ElTy->isPointerTy() || + ElTy->isFloatingPointTy(), + "atomic load operand must have integer, pointer, or floating point " + "type!", + ElTy, &LI); + checkAtomicMemAccessSize(M, ElTy, &LI); } else { Assert(LI.getSynchScope() == CrossThread, "Non-atomic load cannot have SynchronizationScope specified", &LI); @@ -2655,14 +2853,12 @@ void Verifier::visitStoreInst(StoreInst &SI) { "Store cannot have Acquire ordering", &SI); Assert(SI.getAlignment() != 0, "Atomic store must specify explicit alignment", &SI); - if (!ElTy->isPointerTy()) { - Assert(ElTy->isIntegerTy(), - "atomic store operand must have integer type!", &SI, ElTy); - unsigned Size = ElTy->getPrimitiveSizeInBits(); - Assert(Size >= 8 && !(Size & (Size - 1)), - "atomic store operand must be power-of-two byte-sized integer", - &SI, ElTy); - } + Assert(ElTy->isIntegerTy() || ElTy->isPointerTy() || + ElTy->isFloatingPointTy(), + "atomic store operand must have integer, pointer, or floating point " + "type!", + ElTy, &SI); + checkAtomicMemAccessSize(M, ElTy, &SI); } else { Assert(SI.getSynchScope() == CrossThread, "Non-atomic store cannot have SynchronizationScope specified", &SI); @@ -2671,7 +2867,7 @@ void Verifier::visitStoreInst(StoreInst &SI) { } void Verifier::visitAllocaInst(AllocaInst &AI) { - SmallPtrSet Visited; + SmallPtrSet Visited; PointerType *PTy = AI.getType(); Assert(PTy->getAddressSpace() == 0, "Allocation instruction pointer not in the generic address space!", @@ -2709,9 +2905,7 @@ void Verifier::visitAtomicCmpXchgInst(AtomicCmpXchgInst &CXI) { Type *ElTy = PTy->getElementType(); Assert(ElTy->isIntegerTy(), "cmpxchg operand must have integer type!", &CXI, ElTy); - unsigned Size = ElTy->getPrimitiveSizeInBits(); - Assert(Size >= 8 && !(Size & (Size - 1)), - "cmpxchg operand must be power-of-two byte-sized integer", &CXI, ElTy); + checkAtomicMemAccessSize(M, ElTy, &CXI); Assert(ElTy == CXI.getOperand(1)->getType(), "Expected value type does not match pointer operand type!", &CXI, ElTy); @@ -2730,10 +2924,7 @@ void Verifier::visitAtomicRMWInst(AtomicRMWInst &RMWI) { Type *ElTy = PTy->getElementType(); Assert(ElTy->isIntegerTy(), "atomicrmw operand must have integer type!", &RMWI, ElTy); - unsigned Size = ElTy->getPrimitiveSizeInBits(); - Assert(Size >= 8 && !(Size & (Size - 1)), - "atomicrmw operand must be power-of-two byte-sized integer", &RMWI, - ElTy); + checkAtomicMemAccessSize(M, ElTy, &RMWI); Assert(ElTy == RMWI.getOperand(1)->getType(), "Argument value type does not match pointer operand type!", &RMWI, ElTy); @@ -2770,23 +2961,98 @@ void Verifier::visitInsertValueInst(InsertValueInst &IVI) { visitInstruction(IVI); } -void Verifier::visitLandingPadInst(LandingPadInst &LPI) { - BasicBlock *BB = LPI.getParent(); +static Value *getParentPad(Value *EHPad) { + if (auto *FPI = dyn_cast(EHPad)) + return FPI->getParentPad(); + + return cast(EHPad)->getParentPad(); +} + +void Verifier::visitEHPadPredecessors(Instruction &I) { + assert(I.isEHPad()); + + BasicBlock *BB = I.getParent(); + Function *F = BB->getParent(); + + Assert(BB != &F->getEntryBlock(), "EH pad cannot be in entry block.", &I); + + if (auto *LPI = dyn_cast(&I)) { + // The landingpad instruction defines its parent as a landing pad block. The + // landing pad block may be branched to only by the unwind edge of an + // invoke. + for (BasicBlock *PredBB : predecessors(BB)) { + const auto *II = dyn_cast(PredBB->getTerminator()); + Assert(II && II->getUnwindDest() == BB && II->getNormalDest() != BB, + "Block containing LandingPadInst must be jumped to " + "only by the unwind edge of an invoke.", + LPI); + } + return; + } + if (auto *CPI = dyn_cast(&I)) { + if (!pred_empty(BB)) + Assert(BB->getUniquePredecessor() == CPI->getCatchSwitch()->getParent(), + "Block containg CatchPadInst must be jumped to " + "only by its catchswitch.", + CPI); + Assert(BB != CPI->getCatchSwitch()->getUnwindDest(), + "Catchswitch cannot unwind to one of its catchpads", + CPI->getCatchSwitch(), CPI); + return; + } + // Verify that each pred has a legal terminator with a legal to/from EH + // pad relationship. + Instruction *ToPad = &I; + Value *ToPadParent = getParentPad(ToPad); + for (BasicBlock *PredBB : predecessors(BB)) { + TerminatorInst *TI = PredBB->getTerminator(); + Value *FromPad; + if (auto *II = dyn_cast(TI)) { + Assert(II->getUnwindDest() == BB && II->getNormalDest() != BB, + "EH pad must be jumped to via an unwind edge", ToPad, II); + if (auto Bundle = II->getOperandBundle(LLVMContext::OB_funclet)) + FromPad = Bundle->Inputs[0]; + else + FromPad = ConstantTokenNone::get(II->getContext()); + } else if (auto *CRI = dyn_cast(TI)) { + FromPad = CRI->getCleanupPad(); + Assert(FromPad != ToPadParent, "A cleanupret must exit its cleanup", CRI); + } else if (auto *CSI = dyn_cast(TI)) { + FromPad = CSI; + } else { + Assert(false, "EH pad must be jumped to via an unwind edge", ToPad, TI); + } + + // The edge may exit from zero or more nested pads. + for (;; FromPad = getParentPad(FromPad)) { + Assert(FromPad != ToPad, + "EH pad cannot handle exceptions raised within it", FromPad, TI); + if (FromPad == ToPadParent) { + // This is a legal unwind edge. + break; + } + Assert(!isa(FromPad), + "A single unwind edge may only enter one EH pad", TI); + } + } +} + +void Verifier::visitLandingPadInst(LandingPadInst &LPI) { // The landingpad instruction is ill-formed if it doesn't have any clauses and // isn't a cleanup. Assert(LPI.getNumClauses() > 0 || LPI.isCleanup(), "LandingPadInst needs at least one clause or to be a cleanup.", &LPI); - // The landingpad instruction defines its parent as a landing pad block. The - // landing pad block may be branched to only by the unwind edge of an invoke. - for (pred_iterator I = pred_begin(BB), E = pred_end(BB); I != E; ++I) { - const InvokeInst *II = dyn_cast((*I)->getTerminator()); - Assert(II && II->getUnwindDest() == BB && II->getNormalDest() != BB, - "Block containing LandingPadInst must be jumped to " - "only by the unwind edge of an invoke.", + visitEHPadPredecessors(LPI); + + if (!LandingPadResultTy) + LandingPadResultTy = LPI.getType(); + else + Assert(LandingPadResultTy == LPI.getType(), + "The landingpad instruction should have a consistent result type " + "inside a function.", &LPI); - } Function *F = LPI.getParent()->getParent(); Assert(F->hasPersonalityFn(), @@ -2813,6 +3079,269 @@ void Verifier::visitLandingPadInst(LandingPadInst &LPI) { visitInstruction(LPI); } +void Verifier::visitCatchPadInst(CatchPadInst &CPI) { + visitEHPadPredecessors(CPI); + + BasicBlock *BB = CPI.getParent(); + + Function *F = BB->getParent(); + Assert(F->hasPersonalityFn(), + "CatchPadInst needs to be in a function with a personality.", &CPI); + + Assert(isa(CPI.getParentPad()), + "CatchPadInst needs to be directly nested in a CatchSwitchInst.", + CPI.getParentPad()); + + // The catchpad instruction must be the first non-PHI instruction in the + // block. + Assert(BB->getFirstNonPHI() == &CPI, + "CatchPadInst not the first non-PHI instruction in the block.", &CPI); + + visitFuncletPadInst(CPI); +} + +void Verifier::visitCatchReturnInst(CatchReturnInst &CatchReturn) { + Assert(isa(CatchReturn.getOperand(0)), + "CatchReturnInst needs to be provided a CatchPad", &CatchReturn, + CatchReturn.getOperand(0)); + + visitTerminatorInst(CatchReturn); +} + +void Verifier::visitCleanupPadInst(CleanupPadInst &CPI) { + visitEHPadPredecessors(CPI); + + BasicBlock *BB = CPI.getParent(); + + Function *F = BB->getParent(); + Assert(F->hasPersonalityFn(), + "CleanupPadInst needs to be in a function with a personality.", &CPI); + + // The cleanuppad instruction must be the first non-PHI instruction in the + // block. + Assert(BB->getFirstNonPHI() == &CPI, + "CleanupPadInst not the first non-PHI instruction in the block.", + &CPI); + + auto *ParentPad = CPI.getParentPad(); + Assert(isa(ParentPad) || isa(ParentPad), + "CleanupPadInst has an invalid parent.", &CPI); + + visitFuncletPadInst(CPI); +} + +void Verifier::visitFuncletPadInst(FuncletPadInst &FPI) { + User *FirstUser = nullptr; + Value *FirstUnwindPad = nullptr; + SmallVector Worklist({&FPI}); + while (!Worklist.empty()) { + FuncletPadInst *CurrentPad = Worklist.pop_back_val(); + Value *UnresolvedAncestorPad = nullptr; + for (User *U : CurrentPad->users()) { + BasicBlock *UnwindDest; + if (auto *CRI = dyn_cast(U)) { + UnwindDest = CRI->getUnwindDest(); + } else if (auto *CSI = dyn_cast(U)) { + // We allow catchswitch unwind to caller to nest + // within an outer pad that unwinds somewhere else, + // because catchswitch doesn't have a nounwind variant. + // See e.g. SimplifyCFGOpt::SimplifyUnreachable. + if (CSI->unwindsToCaller()) + continue; + UnwindDest = CSI->getUnwindDest(); + } else if (auto *II = dyn_cast(U)) { + UnwindDest = II->getUnwindDest(); + } else if (isa(U)) { + // Calls which don't unwind may be found inside funclet + // pads that unwind somewhere else. We don't *require* + // such calls to be annotated nounwind. + continue; + } else if (auto *CPI = dyn_cast(U)) { + // The unwind dest for a cleanup can only be found by + // recursive search. Add it to the worklist, and we'll + // search for its first use that determines where it unwinds. + Worklist.push_back(CPI); + continue; + } else { + Assert(isa(U), "Bogus funclet pad use", U); + continue; + } + + Value *UnwindPad; + bool ExitsFPI; + if (UnwindDest) { + UnwindPad = UnwindDest->getFirstNonPHI(); + Value *UnwindParent = getParentPad(UnwindPad); + // Ignore unwind edges that don't exit CurrentPad. + if (UnwindParent == CurrentPad) + continue; + // Determine whether the original funclet pad is exited, + // and if we are scanning nested pads determine how many + // of them are exited so we can stop searching their + // children. + Value *ExitedPad = CurrentPad; + ExitsFPI = false; + do { + if (ExitedPad == &FPI) { + ExitsFPI = true; + // Now we can resolve any ancestors of CurrentPad up to + // FPI, but not including FPI since we need to make sure + // to check all direct users of FPI for consistency. + UnresolvedAncestorPad = &FPI; + break; + } + Value *ExitedParent = getParentPad(ExitedPad); + if (ExitedParent == UnwindParent) { + // ExitedPad is the ancestor-most pad which this unwind + // edge exits, so we can resolve up to it, meaning that + // ExitedParent is the first ancestor still unresolved. + UnresolvedAncestorPad = ExitedParent; + break; + } + ExitedPad = ExitedParent; + } while (!isa(ExitedPad)); + } else { + // Unwinding to caller exits all pads. + UnwindPad = ConstantTokenNone::get(FPI.getContext()); + ExitsFPI = true; + UnresolvedAncestorPad = &FPI; + } + + if (ExitsFPI) { + // This unwind edge exits FPI. Make sure it agrees with other + // such edges. + if (FirstUser) { + Assert(UnwindPad == FirstUnwindPad, "Unwind edges out of a funclet " + "pad must have the same unwind " + "dest", + &FPI, U, FirstUser); + } else { + FirstUser = U; + FirstUnwindPad = UnwindPad; + // Record cleanup sibling unwinds for verifySiblingFuncletUnwinds + if (isa(&FPI) && !isa(UnwindPad) && + getParentPad(UnwindPad) == getParentPad(&FPI)) + SiblingFuncletInfo[&FPI] = cast(U); + } + } + // Make sure we visit all uses of FPI, but for nested pads stop as + // soon as we know where they unwind to. + if (CurrentPad != &FPI) + break; + } + if (UnresolvedAncestorPad) { + if (CurrentPad == UnresolvedAncestorPad) { + // When CurrentPad is FPI itself, we don't mark it as resolved even if + // we've found an unwind edge that exits it, because we need to verify + // all direct uses of FPI. + assert(CurrentPad == &FPI); + continue; + } + // Pop off the worklist any nested pads that we've found an unwind + // destination for. The pads on the worklist are the uncles, + // great-uncles, etc. of CurrentPad. We've found an unwind destination + // for all ancestors of CurrentPad up to but not including + // UnresolvedAncestorPad. + Value *ResolvedPad = CurrentPad; + while (!Worklist.empty()) { + Value *UnclePad = Worklist.back(); + Value *AncestorPad = getParentPad(UnclePad); + // Walk ResolvedPad up the ancestor list until we either find the + // uncle's parent or the last resolved ancestor. + while (ResolvedPad != AncestorPad) { + Value *ResolvedParent = getParentPad(ResolvedPad); + if (ResolvedParent == UnresolvedAncestorPad) { + break; + } + ResolvedPad = ResolvedParent; + } + // If the resolved ancestor search didn't find the uncle's parent, + // then the uncle is not yet resolved. + if (ResolvedPad != AncestorPad) + break; + // This uncle is resolved, so pop it from the worklist. + Worklist.pop_back(); + } + } + } + + if (FirstUnwindPad) { + if (auto *CatchSwitch = dyn_cast(FPI.getParentPad())) { + BasicBlock *SwitchUnwindDest = CatchSwitch->getUnwindDest(); + Value *SwitchUnwindPad; + if (SwitchUnwindDest) + SwitchUnwindPad = SwitchUnwindDest->getFirstNonPHI(); + else + SwitchUnwindPad = ConstantTokenNone::get(FPI.getContext()); + Assert(SwitchUnwindPad == FirstUnwindPad, + "Unwind edges out of a catch must have the same unwind dest as " + "the parent catchswitch", + &FPI, FirstUser, CatchSwitch); + } + } + + visitInstruction(FPI); +} + +void Verifier::visitCatchSwitchInst(CatchSwitchInst &CatchSwitch) { + visitEHPadPredecessors(CatchSwitch); + + BasicBlock *BB = CatchSwitch.getParent(); + + Function *F = BB->getParent(); + Assert(F->hasPersonalityFn(), + "CatchSwitchInst needs to be in a function with a personality.", + &CatchSwitch); + + // The catchswitch instruction must be the first non-PHI instruction in the + // block. + Assert(BB->getFirstNonPHI() == &CatchSwitch, + "CatchSwitchInst not the first non-PHI instruction in the block.", + &CatchSwitch); + + auto *ParentPad = CatchSwitch.getParentPad(); + Assert(isa(ParentPad) || isa(ParentPad), + "CatchSwitchInst has an invalid parent.", ParentPad); + + if (BasicBlock *UnwindDest = CatchSwitch.getUnwindDest()) { + Instruction *I = UnwindDest->getFirstNonPHI(); + Assert(I->isEHPad() && !isa(I), + "CatchSwitchInst must unwind to an EH block which is not a " + "landingpad.", + &CatchSwitch); + + // Record catchswitch sibling unwinds for verifySiblingFuncletUnwinds + if (getParentPad(I) == ParentPad) + SiblingFuncletInfo[&CatchSwitch] = &CatchSwitch; + } + + Assert(CatchSwitch.getNumHandlers() != 0, + "CatchSwitchInst cannot have empty handler list", &CatchSwitch); + + for (BasicBlock *Handler : CatchSwitch.handlers()) { + Assert(isa(Handler->getFirstNonPHI()), + "CatchSwitchInst handlers must be catchpads", &CatchSwitch, Handler); + } + + visitTerminatorInst(CatchSwitch); +} + +void Verifier::visitCleanupReturnInst(CleanupReturnInst &CRI) { + Assert(isa(CRI.getOperand(0)), + "CleanupReturnInst needs to be provided a CleanupPad", &CRI, + CRI.getOperand(0)); + + if (BasicBlock *UnwindDest = CRI.getUnwindDest()) { + Instruction *I = UnwindDest->getFirstNonPHI(); + Assert(I->isEHPad() && !isa(I), + "CleanupReturnInst must unwind to an EH block which is not a " + "landingpad.", + &CRI); + } + + visitTerminatorInst(CRI); +} + void Verifier::verifyDominatesUse(Instruction &I, unsigned i) { Instruction *Op = cast(I.getOperand(i)); // If the we have an invalid invoke, don't try to compute the dominance. @@ -2828,6 +3357,19 @@ void Verifier::verifyDominatesUse(Instruction &I, unsigned i) { "Instruction does not dominate all uses!", Op, &I); } +void Verifier::visitDereferenceableMetadata(Instruction& I, MDNode* MD) { + Assert(I.getType()->isPointerTy(), "dereferenceable, dereferenceable_or_null " + "apply only to pointer types", &I); + Assert(isa(I), + "dereferenceable, dereferenceable_or_null apply only to load" + " instructions, use attributes for calls or invokes", &I); + Assert(MD->getNumOperands() == 1, "dereferenceable, dereferenceable_or_null " + "take one operand!", &I); + ConstantInt *CI = mdconst::dyn_extract(MD->getOperand(0)); + Assert(CI && CI->getType()->isIntegerTy(64), "dereferenceable, " + "dereferenceable_or_null metadata value must be an i64!", &I); +} + /// verifyInstruction - Verify that an instruction is well formed. /// void Verifier::visitInstruction(Instruction &I) { @@ -2896,7 +3438,7 @@ void Verifier::visitInstruction(Instruction &I) { " donothing or patchpoint", &I); Assert(F->getParent() == M, "Referencing function in another module!", - &I); + &I, M, F, F->getParent()); } else if (BasicBlock *OpBB = dyn_cast(I.getOperand(i))) { Assert(OpBB->getParent() == BB->getParent(), "Referring to a basic block in another function!", &I); @@ -2904,7 +3446,7 @@ void Verifier::visitInstruction(Instruction &I) { Assert(OpArg->getParent() == BB->getParent(), "Referring to an argument in another function!", &I); } else if (GlobalValue *GV = dyn_cast(I.getOperand(i))) { - Assert(GV->getParent() == M, "Referencing global in another module!", &I); + Assert(GV->getParent() == M, "Referencing global in another module!", &I, M, GV, GV->getParent()); } else if (isa(I.getOperand(i))) { verifyDominatesUse(I, i); } else if (isa(I.getOperand(i))) { @@ -2915,22 +3457,7 @@ void Verifier::visitInstruction(Instruction &I) { if (CE->getType()->isPtrOrPtrVectorTy()) { // If we have a ConstantExpr pointer, we need to see if it came from an // illegal bitcast (inttoptr ) - SmallVector Stack; - SmallPtrSet Visited; - Stack.push_back(CE); - - while (!Stack.empty()) { - const ConstantExpr *V = Stack.pop_back_val(); - if (!Visited.insert(V).second) - continue; - - VerifyConstantExprBitcastType(V); - - for (unsigned I = 0, N = V->getNumOperands(); I != N; ++I) { - if (ConstantExpr *Op = dyn_cast(V->getOperand(I))) - Stack.push_back(Op); - } - } + visitConstantExprsRecursively(CE); } } } @@ -2964,6 +3491,28 @@ void Verifier::visitInstruction(Instruction &I) { &I); } + if (MDNode *MD = I.getMetadata(LLVMContext::MD_dereferenceable)) + visitDereferenceableMetadata(I, MD); + + if (MDNode *MD = I.getMetadata(LLVMContext::MD_dereferenceable_or_null)) + visitDereferenceableMetadata(I, MD); + + if (MDNode *AlignMD = I.getMetadata(LLVMContext::MD_align)) { + Assert(I.getType()->isPointerTy(), "align applies only to pointer types", + &I); + Assert(isa(I), "align applies only to load instructions, " + "use attributes for calls or invokes", &I); + Assert(AlignMD->getNumOperands() == 1, "align takes one operand!", &I); + ConstantInt *CI = mdconst::dyn_extract(AlignMD->getOperand(0)); + Assert(CI && CI->getType()->isIntegerTy(64), + "align metadata value must be an i64!", &I); + uint64_t Align = CI->getZExtValue(); + Assert(isPowerOf2_64(Align), + "align metadata value must be a power of 2!", &I); + Assert(Align <= Value::MaximumAlignment, + "alignment is larger that implementation defined limit", &I); + } + if (MDNode *N = I.getDebugLoc().getAsMDNode()) { Assert(isa(N), "invalid !dbg metadata attachment", &I, N); visitMDNode(*N); @@ -2991,6 +3540,7 @@ bool Verifier::VerifyIntrinsicType(Type *Ty, case IITDescriptor::Void: return !Ty->isVoidTy(); case IITDescriptor::VarArg: return true; case IITDescriptor::MMX: return !Ty->isX86_MMXTy(); + case IITDescriptor::Token: return !Ty->isTokenTy(); case IITDescriptor::Metadata: return !Ty->isMetadataTy(); case IITDescriptor::Half: return !Ty->isHalfTy(); case IITDescriptor::Float: return !Ty->isFloatTy(); @@ -3142,10 +3692,9 @@ Verifier::VerifyIntrinsicIsVarArg(bool isVarArg, return true; } -/// visitIntrinsicFunction - Allow intrinsics to be verified in different ways. -/// -void Verifier::visitIntrinsicFunctionCall(Intrinsic::ID ID, CallSite CI) { - Function *IF = CI.getCalledFunction(); +/// Allow intrinsics to be verified in different ways. +void Verifier::visitIntrinsicCallSite(Intrinsic::ID ID, CallSite CS) { + Function *IF = CS.getCalledFunction(); Assert(IF->isDeclaration(), "Intrinsic functions should never be defined!", IF); @@ -3189,41 +3738,41 @@ void Verifier::visitIntrinsicFunctionCall(Intrinsic::ID ID, CallSite CI) { // If the intrinsic takes MDNode arguments, verify that they are either global // or are local to *this* function. - for (unsigned i = 0, e = CI.getNumArgOperands(); i != e; ++i) - if (auto *MD = dyn_cast(CI.getArgOperand(i))) - visitMetadataAsValue(*MD, CI.getParent()->getParent()); + for (Value *V : CS.args()) + if (auto *MD = dyn_cast(V)) + visitMetadataAsValue(*MD, CS.getCaller()); switch (ID) { default: break; case Intrinsic::ctlz: // llvm.ctlz case Intrinsic::cttz: // llvm.cttz - Assert(isa(CI.getArgOperand(1)), + Assert(isa(CS.getArgOperand(1)), "is_zero_undef argument of bit counting intrinsics must be a " "constant int", - &CI); + CS); break; case Intrinsic::dbg_declare: // llvm.dbg.declare - Assert(isa(CI.getArgOperand(0)), - "invalid llvm.dbg.declare intrinsic call 1", &CI); - visitDbgIntrinsic("declare", cast(*CI.getInstruction())); + Assert(isa(CS.getArgOperand(0)), + "invalid llvm.dbg.declare intrinsic call 1", CS); + visitDbgIntrinsic("declare", cast(*CS.getInstruction())); break; case Intrinsic::dbg_value: // llvm.dbg.value - visitDbgIntrinsic("value", cast(*CI.getInstruction())); + visitDbgIntrinsic("value", cast(*CS.getInstruction())); break; case Intrinsic::memcpy: case Intrinsic::memmove: case Intrinsic::memset: { - ConstantInt *AlignCI = dyn_cast(CI.getArgOperand(3)); + ConstantInt *AlignCI = dyn_cast(CS.getArgOperand(3)); Assert(AlignCI, "alignment argument of memory intrinsics must be a constant int", - &CI); + CS); const APInt &AlignVal = AlignCI->getValue(); Assert(AlignCI->isZero() || AlignVal.isPowerOf2(), - "alignment argument of memory intrinsics must be a power of 2", &CI); - Assert(isa(CI.getArgOperand(4)), + "alignment argument of memory intrinsics must be a power of 2", CS); + Assert(isa(CS.getArgOperand(4)), "isvolatile argument of memory intrinsics must be a constant int", - &CI); + CS); break; } case Intrinsic::gcroot: @@ -3231,76 +3780,76 @@ void Verifier::visitIntrinsicFunctionCall(Intrinsic::ID ID, CallSite CI) { case Intrinsic::gcread: if (ID == Intrinsic::gcroot) { AllocaInst *AI = - dyn_cast(CI.getArgOperand(0)->stripPointerCasts()); - Assert(AI, "llvm.gcroot parameter #1 must be an alloca.", &CI); - Assert(isa(CI.getArgOperand(1)), - "llvm.gcroot parameter #2 must be a constant.", &CI); + dyn_cast(CS.getArgOperand(0)->stripPointerCasts()); + Assert(AI, "llvm.gcroot parameter #1 must be an alloca.", CS); + Assert(isa(CS.getArgOperand(1)), + "llvm.gcroot parameter #2 must be a constant.", CS); if (!AI->getAllocatedType()->isPointerTy()) { - Assert(!isa(CI.getArgOperand(1)), + Assert(!isa(CS.getArgOperand(1)), "llvm.gcroot parameter #1 must either be a pointer alloca, " "or argument #2 must be a non-null constant.", - &CI); + CS); } } - Assert(CI.getParent()->getParent()->hasGC(), - "Enclosing function does not use GC.", &CI); + Assert(CS.getParent()->getParent()->hasGC(), + "Enclosing function does not use GC.", CS); break; case Intrinsic::init_trampoline: - Assert(isa(CI.getArgOperand(1)->stripPointerCasts()), + Assert(isa(CS.getArgOperand(1)->stripPointerCasts()), "llvm.init_trampoline parameter #2 must resolve to a function.", - &CI); + CS); break; case Intrinsic::prefetch: - Assert(isa(CI.getArgOperand(1)) && - isa(CI.getArgOperand(2)) && - cast(CI.getArgOperand(1))->getZExtValue() < 2 && - cast(CI.getArgOperand(2))->getZExtValue() < 4, - "invalid arguments to llvm.prefetch", &CI); + Assert(isa(CS.getArgOperand(1)) && + isa(CS.getArgOperand(2)) && + cast(CS.getArgOperand(1))->getZExtValue() < 2 && + cast(CS.getArgOperand(2))->getZExtValue() < 4, + "invalid arguments to llvm.prefetch", CS); break; case Intrinsic::stackprotector: - Assert(isa(CI.getArgOperand(1)->stripPointerCasts()), - "llvm.stackprotector parameter #2 must resolve to an alloca.", &CI); + Assert(isa(CS.getArgOperand(1)->stripPointerCasts()), + "llvm.stackprotector parameter #2 must resolve to an alloca.", CS); break; case Intrinsic::lifetime_start: case Intrinsic::lifetime_end: case Intrinsic::invariant_start: - Assert(isa(CI.getArgOperand(0)), + Assert(isa(CS.getArgOperand(0)), "size argument of memory use markers must be a constant integer", - &CI); + CS); break; case Intrinsic::invariant_end: - Assert(isa(CI.getArgOperand(1)), - "llvm.invariant.end parameter #2 must be a constant integer", &CI); + Assert(isa(CS.getArgOperand(1)), + "llvm.invariant.end parameter #2 must be a constant integer", CS); break; - case Intrinsic::frameescape: { - BasicBlock *BB = CI.getParent(); + case Intrinsic::localescape: { + BasicBlock *BB = CS.getParent(); Assert(BB == &BB->getParent()->front(), - "llvm.frameescape used outside of entry block", &CI); + "llvm.localescape used outside of entry block", CS); Assert(!SawFrameEscape, - "multiple calls to llvm.frameescape in one function", &CI); - for (Value *Arg : CI.args()) { + "multiple calls to llvm.localescape in one function", CS); + for (Value *Arg : CS.args()) { if (isa(Arg)) continue; // Null values are allowed as placeholders. auto *AI = dyn_cast(Arg->stripPointerCasts()); Assert(AI && AI->isStaticAlloca(), - "llvm.frameescape only accepts static allocas", &CI); + "llvm.localescape only accepts static allocas", CS); } - FrameEscapeInfo[BB->getParent()].first = CI.getNumArgOperands(); + FrameEscapeInfo[BB->getParent()].first = CS.getNumArgOperands(); SawFrameEscape = true; break; } - case Intrinsic::framerecover: { - Value *FnArg = CI.getArgOperand(0)->stripPointerCasts(); + case Intrinsic::localrecover: { + Value *FnArg = CS.getArgOperand(0)->stripPointerCasts(); Function *Fn = dyn_cast(FnArg); Assert(Fn && !Fn->isDeclaration(), - "llvm.framerecover first " + "llvm.localrecover first " "argument must be function defined in this module", - &CI); - auto *IdxArg = dyn_cast(CI.getArgOperand(2)); - Assert(IdxArg, "idx argument of llvm.framerecover must be a constant int", - &CI); + CS); + auto *IdxArg = dyn_cast(CS.getArgOperand(2)); + Assert(IdxArg, "idx argument of llvm.localrecover must be a constant int", + CS); auto &Entry = FrameEscapeInfo[Fn]; Entry.second = unsigned( std::max(uint64_t(Entry.second), IdxArg->getLimitedValue(~0U) + 1)); @@ -3308,57 +3857,53 @@ void Verifier::visitIntrinsicFunctionCall(Intrinsic::ID ID, CallSite CI) { } case Intrinsic::experimental_gc_statepoint: - Assert(!CI.isInlineAsm(), - "gc.statepoint support for inline assembly unimplemented", &CI); - Assert(CI.getParent()->getParent()->hasGC(), - "Enclosing function does not use GC.", &CI); + Assert(!CS.isInlineAsm(), + "gc.statepoint support for inline assembly unimplemented", CS); + Assert(CS.getParent()->getParent()->hasGC(), + "Enclosing function does not use GC.", CS); - VerifyStatepoint(ImmutableCallSite(CI)); + VerifyStatepoint(CS); break; - case Intrinsic::experimental_gc_result_int: - case Intrinsic::experimental_gc_result_float: - case Intrinsic::experimental_gc_result_ptr: case Intrinsic::experimental_gc_result: { - Assert(CI.getParent()->getParent()->hasGC(), - "Enclosing function does not use GC.", &CI); + Assert(CS.getParent()->getParent()->hasGC(), + "Enclosing function does not use GC.", CS); // Are we tied to a statepoint properly? - CallSite StatepointCS(CI.getArgOperand(0)); + CallSite StatepointCS(CS.getArgOperand(0)); const Function *StatepointFn = StatepointCS.getInstruction() ? StatepointCS.getCalledFunction() : nullptr; Assert(StatepointFn && StatepointFn->isDeclaration() && StatepointFn->getIntrinsicID() == Intrinsic::experimental_gc_statepoint, - "gc.result operand #1 must be from a statepoint", &CI, - CI.getArgOperand(0)); + "gc.result operand #1 must be from a statepoint", CS, + CS.getArgOperand(0)); // Assert that result type matches wrapped callee. const Value *Target = StatepointCS.getArgument(2); - const PointerType *PT = cast(Target->getType()); - const FunctionType *TargetFuncType = - cast(PT->getElementType()); - Assert(CI.getType() == TargetFuncType->getReturnType(), - "gc.result result type does not match wrapped callee", &CI); + auto *PT = cast(Target->getType()); + auto *TargetFuncType = cast(PT->getElementType()); + Assert(CS.getType() == TargetFuncType->getReturnType(), + "gc.result result type does not match wrapped callee", CS); break; } case Intrinsic::experimental_gc_relocate: { - Assert(CI.getNumArgOperands() == 3, "wrong number of arguments", &CI); + Assert(CS.getNumArgOperands() == 3, "wrong number of arguments", CS); + + Assert(isa(CS.getType()->getScalarType()), + "gc.relocate must return a pointer or a vector of pointers", CS); // Check that this relocate is correctly tied to the statepoint // This is case for relocate on the unwinding path of an invoke statepoint - if (ExtractValueInst *ExtractValue = - dyn_cast(CI.getArgOperand(0))) { - Assert(isa(ExtractValue->getAggregateOperand()), - "gc relocate on unwind path incorrectly linked to the statepoint", - &CI); + if (LandingPadInst *LandingPad = + dyn_cast(CS.getArgOperand(0))) { const BasicBlock *InvokeBB = - ExtractValue->getParent()->getUniquePredecessor(); + LandingPad->getParent()->getUniquePredecessor(); // Landingpad relocates should have only one predecessor with invoke // statepoint terminator Assert(InvokeBB, "safepoints should have unique landingpads", - ExtractValue->getParent()); + LandingPad->getParent()); Assert(InvokeBB->getTerminator(), "safepoint block should be well formed", InvokeBB); Assert(isStatepoint(InvokeBB->getTerminator()), @@ -3368,32 +3913,32 @@ void Verifier::visitIntrinsicFunctionCall(Intrinsic::ID ID, CallSite CI) { // In all other cases relocate should be tied to the statepoint directly. // This covers relocates on a normal return path of invoke statepoint and // relocates of a call statepoint - auto Token = CI.getArgOperand(0); + auto Token = CS.getArgOperand(0); Assert(isa(Token) && isStatepoint(cast(Token)), - "gc relocate is incorrectly tied to the statepoint", &CI, Token); + "gc relocate is incorrectly tied to the statepoint", CS, Token); } // Verify rest of the relocate arguments - GCRelocateOperands Ops(CI); - ImmutableCallSite StatepointCS(Ops.getStatepoint()); + ImmutableCallSite StatepointCS( + cast(*CS.getInstruction()).getStatepoint()); // Both the base and derived must be piped through the safepoint - Value* Base = CI.getArgOperand(1); + Value* Base = CS.getArgOperand(1); Assert(isa(Base), - "gc.relocate operand #2 must be integer offset", &CI); + "gc.relocate operand #2 must be integer offset", CS); - Value* Derived = CI.getArgOperand(2); + Value* Derived = CS.getArgOperand(2); Assert(isa(Derived), - "gc.relocate operand #3 must be integer offset", &CI); + "gc.relocate operand #3 must be integer offset", CS); const int BaseIndex = cast(Base)->getZExtValue(); const int DerivedIndex = cast(Derived)->getZExtValue(); // Check the bounds Assert(0 <= BaseIndex && BaseIndex < (int)StatepointCS.arg_size(), - "gc.relocate: statepoint base index out of bounds", &CI); + "gc.relocate: statepoint base index out of bounds", CS); Assert(0 <= DerivedIndex && DerivedIndex < (int)StatepointCS.arg_size(), - "gc.relocate: statepoint derived index out of bounds", &CI); + "gc.relocate: statepoint derived index out of bounds", CS); // Check that BaseIndex and DerivedIndex fall within the 'gc parameters' // section of the statepoint's argument @@ -3422,24 +3967,33 @@ void Verifier::visitIntrinsicFunctionCall(Intrinsic::ID ID, CallSite CI) { Assert(GCParamArgsStart <= BaseIndex && BaseIndex < GCParamArgsEnd, "gc.relocate: statepoint base index doesn't fall within the " "'gc parameters' section of the statepoint call", - &CI); + CS); Assert(GCParamArgsStart <= DerivedIndex && DerivedIndex < GCParamArgsEnd, "gc.relocate: statepoint derived index doesn't fall within the " "'gc parameters' section of the statepoint call", - &CI); - - // Relocated value must be a pointer type, but gc_relocate does not need to return the - // same pointer type as the relocated pointer. It can be casted to the correct type later - // if it's desired. However, they must have the same address space. - GCRelocateOperands Operands(CI); - Assert(Operands.getDerivedPtr()->getType()->isPointerTy(), - "gc.relocate: relocated value must be a gc pointer", &CI); - - // gc_relocate return type must be a pointer type, and is verified earlier in - // VerifyIntrinsicType(). - Assert(cast(CI.getType())->getAddressSpace() == - cast(Operands.getDerivedPtr()->getType())->getAddressSpace(), - "gc.relocate: relocating a pointer shouldn't change its address space", &CI); + CS); + + // Relocated value must be either a pointer type or vector-of-pointer type, + // but gc_relocate does not need to return the same pointer type as the + // relocated pointer. It can be casted to the correct type later if it's + // desired. However, they must have the same address space and 'vectorness' + GCRelocateInst &Relocate = cast(*CS.getInstruction()); + Assert(Relocate.getDerivedPtr()->getType()->getScalarType()->isPointerTy(), + "gc.relocate: relocated value must be a gc pointer", CS); + + auto ResultType = CS.getType(); + auto DerivedType = Relocate.getDerivedPtr()->getType(); + Assert(ResultType->isVectorTy() == DerivedType->isVectorTy(), + "gc.relocate: vector relocates to vector and pointer to pointer", CS); + Assert(ResultType->getPointerAddressSpace() == + DerivedType->getPointerAddressSpace(), + "gc.relocate: relocating a pointer shouldn't change its address space", CS); + break; + } + case Intrinsic::eh_exceptioncode: + case Intrinsic::eh_exceptionpointer: { + Assert(isa(CS.getArgOperand(0)), + "eh.exceptionpointer argument must be a catchpad", CS); break; } }; @@ -3533,15 +4087,34 @@ static uint64_t getVariableSize(const DILocalVariable &V, const MapTy &Map) { } template -void Verifier::verifyBitPieceExpression(const DbgInfoIntrinsic &I, - const MapTy &TypeRefs) { +void Verifier::verifyDIExpression(const DbgInfoIntrinsic &I, + const MapTy &TypeRefs) { DILocalVariable *V; DIExpression *E; + const Value *Arg; + uint64_t ArgumentTypeSizeInBits = 0; if (auto *DVI = dyn_cast(&I)) { + Arg = DVI->getValue(); + if (Arg) + ArgumentTypeSizeInBits = + M->getDataLayout().getTypeAllocSizeInBits(Arg->getType()); V = dyn_cast_or_null(DVI->getRawVariable()); E = dyn_cast_or_null(DVI->getRawExpression()); } else { auto *DDI = cast(&I); + // For declare intrinsics, get the total size of the alloca, to allow + // case where the variable may span more than one element. + Arg = DDI->getAddress(); + if (Arg) + Arg = Arg->stripPointerCasts(); + const AllocaInst *AI = dyn_cast_or_null(Arg); + if (AI) { + // We can only say something about constant size allocations + if (const ConstantInt *CI = dyn_cast(AI->getArraySize())) + ArgumentTypeSizeInBits = + CI->getLimitedValue() * + M->getDataLayout().getTypeAllocSizeInBits(AI->getAllocatedType()); + } V = dyn_cast_or_null(DDI->getRawVariable()); E = dyn_cast_or_null(DDI->getRawExpression()); } @@ -3550,10 +4123,6 @@ void Verifier::verifyBitPieceExpression(const DbgInfoIntrinsic &I, if (!V || !E || !E->isValid()) return; - // Nothing to do if this isn't a bit piece expression. - if (!E->isBitPiece()) - return; - // The frontend helps out GDB by emitting the members of local anonymous // unions as artificial local variables with shared storage. When SROA splits // the storage for artificial local variables that are smaller than the entire @@ -3569,11 +4138,33 @@ void Verifier::verifyBitPieceExpression(const DbgInfoIntrinsic &I, if (!VarSize) return; - unsigned PieceSize = E->getBitPieceSize(); - unsigned PieceOffset = E->getBitPieceOffset(); - Assert(PieceSize + PieceOffset <= VarSize, - "piece is larger than or outside of variable", &I, V, E); - Assert(PieceSize != VarSize, "piece covers entire variable", &I, V, E); + if (E->isBitPiece()) { + unsigned PieceSize = E->getBitPieceSize(); + unsigned PieceOffset = E->getBitPieceOffset(); + Assert(PieceSize + PieceOffset <= VarSize, + "piece is larger than or outside of variable", &I, V, E); + Assert(PieceSize != VarSize, "piece covers entire variable", &I, V, E); + return; + } + + if (!ArgumentTypeSizeInBits) + return; // We were unable to determine the size of the argument + + if (E->getNumElements() == 0) { + // In the case where the expression is empty, verify the size of the + // argument. Doing this in the general case would require looking through + // any dereferences that may be in the expression. + Assert(ArgumentTypeSizeInBits == VarSize, + "size of passed value (" + utostr(ArgumentTypeSizeInBits) + + ") does not match size of declared variable (" + + utostr(VarSize) + ")", + &I, Arg, V, V->getType(), E); + } else if (E->getElement(0) == dwarf::DW_OP_deref) { + Assert(ArgumentTypeSizeInBits == M->getDataLayout().getPointerSizeInBits(), + "the operation of the expression is a deref, but the passed value " + "is not pointer sized", + &I, Arg, V, V->getType(), E); + } } void Verifier::visitUnresolvedTypeRef(const MDString *S, const MDNode *N) { @@ -3592,7 +4183,7 @@ void Verifier::verifyTypeRefs() { for (auto *CU : CUs->operands()) if (auto Ts = cast(CU)->getRetainedTypes()) for (DIType *Op : Ts) - if (auto *T = dyn_cast(Op)) + if (auto *T = dyn_cast_or_null(Op)) if (auto *S = T->getRawIdentifier()) { UnresolvedTypeRefs.erase(S); TypeRefs.insert(std::make_pair(S, T)); @@ -3606,7 +4197,7 @@ void Verifier::verifyTypeRefs() { for (const BasicBlock &BB : F) for (const Instruction &I : BB) if (auto *DII = dyn_cast(&I)) - verifyBitPieceExpression(*DII, TypeRefs); + verifyDIExpression(*DII, TypeRefs); // Return early if all typerefs were resolved. if (UnresolvedTypeRefs.empty())