[CodeGen] Collect information about sizes of accesses and access types for TBAA

[kosarev]

Summary

The information about access and type sizes is necessary for producing TBAA metadata in the new size-aware format. With this patch, D39955 and D39956 in place we should be able to change CodeGenTBAA::createScalarTypeNode() and CodeGenTBAA::getBaseTypeInfo() to generate metadata in the new format under the -new-struct-path-tbaa command-line option. For now, this new information remains unused.

[kosarev]

lib/CodeGen/CGClass.cpp

2426–2427

Now that type and access descriptors include information about sizes, the function needs to know the type of the virtual table pointer to access.

lib/CodeGen/CodeGenModule.cpp

139

To clarify why we pass the module instead of LLVM context: we need it to compute the size of virtual table pointers as their types are llvm::Type types and not clang::Type or QualType ones.

585

getAccessTagInfo() will raise an assertion failure if such an access descriptor is passed to it. This way we explicitly claim that generating access tags for incomplete objects is not allowed.

[kosarev]

Rebased.

[rjmccall]

Alright, I guess this all makes sense. We do have some extensions that are playing around with v-table pointers, so it's probably fair to require the v-table pointer type to be passed down instead of assuming it's just the generic pointer size.

lib/CodeGen/CGClass.cpp

void CodeGenFunction::InitializeVTablePointer(const VPtr &Vptr) {
   llvm::Type *VTablePtrTy =
       llvm::FunctionType::get(CGM.Int32Ty, /*isVarArg=*/true)
           ->getPointerTo()
           ->getPointerTo();
   VTableField = Builder.CreateBitCast(VTableField, VTablePtrTy->getPointerTo());
   VTableAddressPoint = Builder.CreateBitCast(VTableAddressPoint, VTablePtrTy);
 
   llvm::StoreInst *Store = Builder.CreateStore(VTableAddressPoint, VTableField);
-  CGM.DecorateInstructionWithTBAA(Store, CGM.getTBAAVTablePtrAccessInfo());
+  TBAAAccessInfo TBAAInfo = CGM.getTBAAVTablePtrAccessInfo(VTablePtrTy);
+  CGM.DecorateInstructionWithTBAA(Store, TBAAInfo);
   if (CGM.getCodeGenOpts().OptimizationLevel > 0 &&
       CGM.getCodeGenOpts().StrictVTablePointers)
     CGM.DecorateInstructionWithInvariantGroup(Store, Vptr.VTableClass);
 }
 
 CodeGenFunction::VPtrsVector
 CodeGenFunction::getVTablePointers(const CXXRecordDecl *VTableClass) {
   CodeGenFunction::VPtrsVector VPtrsResult;
if (RD->getNumVBases())
     CGM.getCXXABI().initializeHiddenVirtualInheritanceMembers(*this, RD);
 }
 
 llvm::Value *CodeGenFunction::GetVTablePtr(Address This,
                                            llvm::Type *VTableTy,
                                            const CXXRecordDecl *RD) {
   Address VTablePtrSrc = Builder.CreateElementBitCast(This, VTableTy);
   llvm::Instruction *VTable = Builder.CreateLoad(VTablePtrSrc, "vtable");
-  CGM.DecorateInstructionWithTBAA(VTable, CGM.getTBAAVTablePtrAccessInfo());
+  TBAAAccessInfo TBAAInfo = CGM.getTBAAVTablePtrAccessInfo(VTableTy);
+  CGM.DecorateInstructionWithTBAA(VTable, TBAAInfo);
 
   if (CGM.getCodeGenOpts().OptimizationLevel > 0 &&
       CGM.getCodeGenOpts().StrictVTablePointers)
     CGM.DecorateInstructionWithInvariantGroup(VTable, RD);
 
   return VTable;
 }
 

lib/CodeGen/CodeGenModule.h

public:
   llvm::MDNode *getTBAATypeInfo(QualType QTy);
 
   /// getTBAAAccessInfo - Get TBAA information that describes an access to
   /// an object of the given type.
   TBAAAccessInfo getTBAAAccessInfo(QualType AccessType);
 
   /// getTBAAVTablePtrAccessInfo - Get the TBAA information that describes an
   /// access to a virtual table pointer.
-  TBAAAccessInfo getTBAAVTablePtrAccessInfo();
+  TBAAAccessInfo getTBAAVTablePtrAccessInfo(llvm::Type *VTablePtrType);
 
   llvm::MDNode *getTBAAStructInfo(QualType QTy);
 
   /// getTBAABaseTypeInfo - Get metadata that describes the given base access
   /// type. Return null if the type is not suitable for use in TBAA access tags.
   llvm::MDNode *getTBAABaseTypeInfo(QualType QTy);
 
   /// getTBAAAccessTagInfo - Get TBAA tag for a given memory access.

lib/CodeGen/CodeGenModule.cpp

CodeGenModule::CodeGenModule(ASTContext &C, const HeaderSearchOptions &HSO,
   if (LangOpts.OpenMP)
     createOpenMPRuntime();
   if (LangOpts.CUDA)
     createCUDARuntime();
 
   // Enable TBAA unless it's suppressed. ThreadSanitizer needs TBAA even at O0.
   if (LangOpts.Sanitize.has(SanitizerKind::Thread) ||
       (!CodeGenOpts.RelaxedAliasing && CodeGenOpts.OptimizationLevel > 0))
-    TBAA.reset(new CodeGenTBAA(Context, VMContext, CodeGenOpts, getLangOpts(),
+    TBAA.reset(new CodeGenTBAA(Context, TheModule, CodeGenOpts, getLangOpts(),
                                getCXXABI().getMangleContext()));
 
   // If debug info or coverage generation is enabled, create the CGDebugInfo
   // object.
   if (CodeGenOpts.getDebugInfo() != codegenoptions::NoDebugInfo ||
       CodeGenOpts.EmitGcovArcs || CodeGenOpts.EmitGcovNotes)
     DebugInfo.reset(new CGDebugInfo(*this));
 

 
 llvm::MDNode *CodeGenModule::getTBAATypeInfo(QualType QTy) {
   if (!TBAA)
     return nullptr;
   return TBAA->getTypeInfo(QTy);
 }
 
 TBAAAccessInfo CodeGenModule::getTBAAAccessInfo(QualType AccessType) {
-  return TBAAAccessInfo(getTBAATypeInfo(AccessType));
+  // Pointee values may have incomplete types, but they shall never be
+  // dereferenced.
+  if (AccessType->isIncompleteType())
+    return TBAAAccessInfo::getIncompleteInfo();
+
+  uint64_t Size = Context.getTypeSizeInChars(AccessType).getQuantity();
+  return TBAAAccessInfo(getTBAATypeInfo(AccessType), Size);
 }
 
-TBAAAccessInfo CodeGenModule::getTBAAVTablePtrAccessInfo() {
+TBAAAccessInfo
+CodeGenModule::getTBAAVTablePtrAccessInfo(llvm::Type *VTablePtrType) {
   if (!TBAA)
     return TBAAAccessInfo();
-  return TBAA->getVTablePtrAccessInfo();
+  return TBAA->getVTablePtrAccessInfo(VTablePtrType);
 }
 
 llvm::MDNode *CodeGenModule::getTBAAStructInfo(QualType QTy) {
   if (!TBAA)
     return nullptr;
   return TBAA->getTBAAStructInfo(QTy);
 }
 

lib/CodeGen/CodeGenTBAA.h

 
 namespace CodeGen {
 class CGRecordLayout;
 
 // TBAAAccessKind - A kind of TBAA memory access descriptor.
 enum class TBAAAccessKind : unsigned {
   Ordinary,
   MayAlias,
+  Incomplete,
 };
 
 // TBAAAccessInfo - Describes a memory access in terms of TBAA.
 struct TBAAAccessInfo {
   TBAAAccessInfo(TBAAAccessKind Kind, llvm::MDNode *BaseType,
-                 llvm::MDNode *AccessType, uint64_t Offset)
-    : Kind(Kind), BaseType(BaseType), AccessType(AccessType), Offset(Offset)
+                 llvm::MDNode *AccessType, uint64_t Offset, uint64_t Size)
+    : Kind(Kind), BaseType(BaseType), AccessType(AccessType),
+      Offset(Offset), Size(Size)
   {}
 
   TBAAAccessInfo(llvm::MDNode *BaseType, llvm::MDNode *AccessType,
-                 uint64_t Offset)
-    : TBAAAccessInfo(TBAAAccessKind::Ordinary, BaseType, AccessType, Offset)
+                 uint64_t Offset, uint64_t Size)
+    : TBAAAccessInfo(TBAAAccessKind::Ordinary, BaseType, AccessType,
+                     Offset, Size)
   {}
 
-  explicit TBAAAccessInfo(llvm::MDNode *AccessType)
-    : TBAAAccessInfo(/* BaseType= */ nullptr, AccessType, /* Offset= */ 0)
+  explicit TBAAAccessInfo(llvm::MDNode *AccessType, uint64_t Size)
+    : TBAAAccessInfo(/* BaseType= */ nullptr, AccessType, /* Offset= */ 0, Size)
   {}
 
   TBAAAccessInfo()
-    : TBAAAccessInfo(/* AccessType= */ nullptr)
+    : TBAAAccessInfo(/* AccessType= */ nullptr, /* Size= */ 0)
   {}
 
   static TBAAAccessInfo getMayAliasInfo() {
-    return TBAAAccessInfo(TBAAAccessKind::MayAlias, /* BaseType= */ nullptr,
-                          /* AccessType= */ nullptr, /* Offset= */ 0);
+    return TBAAAccessInfo(TBAAAccessKind::MayAlias,
+                          /* BaseType= */ nullptr, /* AccessType= */ nullptr,
+                          /* Offset= */ 0, /* Size= */ 0);
   }
 
   bool isMayAlias() const { return Kind == TBAAAccessKind::MayAlias; }
 
+  static TBAAAccessInfo getIncompleteInfo() {
+    return TBAAAccessInfo(TBAAAccessKind::Incomplete,
+                          /* BaseType= */ nullptr, /* AccessType= */ nullptr,
+                          /* Offset= */ 0, /* Size= */ 0);
+  }
+
+  bool isIncomplete() const { return Kind == TBAAAccessKind::Incomplete; }
+
   bool operator==(const TBAAAccessInfo &Other) const {
     return Kind == Other.Kind &&
            BaseType == Other.BaseType &&
            AccessType == Other.AccessType &&
-           Offset == Other.Offset;
+           Offset == Other.Offset &&
+           Size == Other.Size;
   }
 
   bool operator!=(const TBAAAccessInfo &Other) const {
     return !(*this == Other);
   }
 
   explicit operator bool() const {
     return *this != TBAAAccessInfo();
struct TBAAAccessInfo {
 
   /// AccessType - The final access type. May be null if there is no TBAA
   /// information available about this access.
   llvm::MDNode *AccessType;
 
   /// Offset - The byte offset of the final access within the base one. Must be
   /// zero if the base access type is not specified.
   uint64_t Offset;
+
+  /// Size - The size of access, in bytes.
+  uint64_t Size;
 };
 
 /// CodeGenTBAA - This class organizes the cross-module state that is used
 /// while lowering AST types to LLVM types.
 class CodeGenTBAA {
   ASTContext &Context;
+  llvm::Module &Module;
   const CodeGenOptions &CodeGenOpts;
   const LangOptions &Features;
   MangleContext &MContext;
 
   // MDHelper - Helper for creating metadata.
   llvm::MDBuilder MDHelper;
 
   /// MetadataCache - This maps clang::Types to scalar llvm::MDNodes describing
class CodeGenTBAA {
 
   /// CollectFields - Collect information about the fields of a type for
   /// !tbaa.struct metadata formation. Return false for an unsupported type.
   bool CollectFields(uint64_t BaseOffset,
                      QualType Ty,
                      SmallVectorImpl<llvm::MDBuilder::TBAAStructField> &Fields,
                      bool MayAlias);
 
-  /// A wrapper function to create a scalar type. For struct-path aware TBAA,
-  /// the scalar type has the same format as the struct type: name, offset,
-  /// pointer to another node in the type DAG.
-  llvm::MDNode *createTBAAScalarType(StringRef Name, llvm::MDNode *Parent);
+  /// createScalarTypeNode - A wrapper function to create a metadata node
+  /// describing a scalar type.
+  llvm::MDNode *createScalarTypeNode(StringRef Name, llvm::MDNode *Parent,
+                                     uint64_t Size);
 
   /// getTypeInfoHelper - An internal helper function to generate metadata used
   /// to describe accesses to objects of the given type.
   llvm::MDNode *getTypeInfoHelper(const Type *Ty);
 
   /// getBaseTypeInfoHelper - An internal helper function to generate metadata
   /// used to describe accesses to objects of the given base type.
   llvm::MDNode *getBaseTypeInfoHelper(const Type *Ty);
 
 public:
-  CodeGenTBAA(ASTContext &Ctx, llvm::LLVMContext &VMContext,
-              const CodeGenOptions &CGO,
-              const LangOptions &Features,
-              MangleContext &MContext);
+  CodeGenTBAA(ASTContext &Ctx, llvm::Module &M, const CodeGenOptions &CGO,
+              const LangOptions &Features, MangleContext &MContext);
   ~CodeGenTBAA();
 
   /// getTypeInfo - Get metadata used to describe accesses to objects of the
   /// given type.
   llvm::MDNode *getTypeInfo(QualType QTy);
 
   /// getVTablePtrAccessInfo - Get the TBAA information that describes an
   /// access to a virtual table pointer.
-  TBAAAccessInfo getVTablePtrAccessInfo();
+  TBAAAccessInfo getVTablePtrAccessInfo(llvm::Type *VTablePtrType);
 
   /// getTBAAStructInfo - Get the TBAAStruct MDNode to be used for a memcpy of
   /// the given type.
   llvm::MDNode *getTBAAStructInfo(QualType QTy);
 
   /// getBaseTypeInfo - Get metadata that describes the given base access type.
   /// Return null if the type is not suitable for use in TBAA access tags.
   llvm::MDNode *getBaseTypeInfo(QualType QTy);

 
 template<> struct DenseMapInfo<clang::CodeGen::TBAAAccessInfo> {
   static clang::CodeGen::TBAAAccessInfo getEmptyKey() {
     unsigned UnsignedKey = DenseMapInfo<unsigned>::getEmptyKey();
     return clang::CodeGen::TBAAAccessInfo(
       static_cast<clang::CodeGen::TBAAAccessKind>(UnsignedKey),
       DenseMapInfo<MDNode *>::getEmptyKey(),
       DenseMapInfo<MDNode *>::getEmptyKey(),
+      DenseMapInfo<uint64_t>::getEmptyKey(),
       DenseMapInfo<uint64_t>::getEmptyKey());
   }
 
   static clang::CodeGen::TBAAAccessInfo getTombstoneKey() {
     unsigned UnsignedKey = DenseMapInfo<unsigned>::getTombstoneKey();
     return clang::CodeGen::TBAAAccessInfo(
       static_cast<clang::CodeGen::TBAAAccessKind>(UnsignedKey),
       DenseMapInfo<MDNode *>::getTombstoneKey(),
       DenseMapInfo<MDNode *>::getTombstoneKey(),
+      DenseMapInfo<uint64_t>::getTombstoneKey(),
       DenseMapInfo<uint64_t>::getTombstoneKey());
   }
 
   static unsigned getHashValue(const clang::CodeGen::TBAAAccessInfo &Val) {
     auto KindValue = static_cast<unsigned>(Val.Kind);
     return DenseMapInfo<unsigned>::getHashValue(KindValue) ^
            DenseMapInfo<MDNode *>::getHashValue(Val.BaseType) ^
            DenseMapInfo<MDNode *>::getHashValue(Val.AccessType) ^
-           DenseMapInfo<uint64_t>::getHashValue(Val.Offset);
+           DenseMapInfo<uint64_t>::getHashValue(Val.Offset) ^
+           DenseMapInfo<uint64_t>::getHashValue(Val.Size);
   }
 
   static bool isEqual(const clang::CodeGen::TBAAAccessInfo &LHS,
                       const clang::CodeGen::TBAAAccessInfo &RHS) {
     return LHS == RHS;
   }
 };
 
 }  // end namespace llvm
 
 #endif

lib/CodeGen/CodeGenTBAA.cpp

 #include "clang/AST/Attr.h"
 #include "clang/AST/Mangle.h"
 #include "clang/AST/RecordLayout.h"
 #include "clang/Frontend/CodeGenOptions.h"
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/Metadata.h"
+#include "llvm/IR/Module.h"
 #include "llvm/IR/Type.h"
 using namespace clang;
 using namespace CodeGen;
 
-CodeGenTBAA::CodeGenTBAA(ASTContext &Ctx, llvm::LLVMContext& VMContext,
+CodeGenTBAA::CodeGenTBAA(ASTContext &Ctx, llvm::Module &M,
                          const CodeGenOptions &CGO,
                          const LangOptions &Features, MangleContext &MContext)
-  : Context(Ctx), CodeGenOpts(CGO), Features(Features), MContext(MContext),
-    MDHelper(VMContext), Root(nullptr), Char(nullptr) {
-}
+  : Context(Ctx), Module(M), CodeGenOpts(CGO),
+    Features(Features), MContext(MContext), MDHelper(M.getContext()),
+    Root(nullptr), Char(nullptr)
+{}
 
 CodeGenTBAA::~CodeGenTBAA() {
 }
 
 llvm::MDNode *CodeGenTBAA::getRoot() {
   // Define the root of the tree. This identifies the tree, so that
   // if our LLVM IR is linked with LLVM IR from a different front-end
   // (or a different version of this front-end), their TBAA trees will
   // remain distinct, and the optimizer will treat them conservatively.
   if (!Root) {
     if (Features.CPlusPlus)
       Root = MDHelper.createTBAARoot("Simple C++ TBAA");
     else
       Root = MDHelper.createTBAARoot("Simple C/C++ TBAA");
   }
 
   return Root;
 }
 
-// For both scalar TBAA and struct-path aware TBAA, the scalar type has the
-// same format: name, parent node, and offset.
-llvm::MDNode *CodeGenTBAA::createTBAAScalarType(StringRef Name,
-                                                llvm::MDNode *Parent) {
+llvm::MDNode *CodeGenTBAA::createScalarTypeNode(StringRef Name,
+                                                llvm::MDNode *Parent,
+                                                uint64_t Size) {
+  (void)Size; // TODO: Support generation of size-aware type nodes.
   return MDHelper.createTBAAScalarTypeNode(Name, Parent);
 }
 
 llvm::MDNode *CodeGenTBAA::getChar() {
   // Define the root of the tree for user-accessible memory. C and C++
   // give special powers to char and certain similar types. However,
   // these special powers only cover user-accessible memory, and doesn't
   // include things like vtables.
   if (!Char)
-    Char = createTBAAScalarType("omnipotent char", getRoot());
+    Char = createScalarTypeNode("omnipotent char", getRoot(), /* Size= */ 1);
 
   return Char;
 }
 
 static bool TypeHasMayAlias(QualType QTy) {
   // Tagged types have declarations, and therefore may have attributes.
   if (const TagType *TTy = dyn_cast<TagType>(QTy))
     return TTy->getDecl()->hasAttr<MayAliasAttr>();
if (const RecordType *TTy = QTy->getAs<RecordType>()) {
     // For now, we only handle struct and class.
     if (RD->isStruct() || RD->isClass())
       return true;
   }
   return false;
 }
 
 llvm::MDNode *CodeGenTBAA::getTypeInfoHelper(const Type *Ty) {
+  uint64_t Size = Context.getTypeSizeInChars(Ty).getQuantity();
+
   // Handle builtin types.
   if (const BuiltinType *BTy = dyn_cast<BuiltinType>(Ty)) {
     switch (BTy->getKind()) {
     // Character types are special and can alias anything.
     // In C++, this technically only includes "char" and "unsigned char",
     // and not "signed char". In C, it includes all three. For now,
     // the risk of exploiting this detail in C++ seems likely to outweigh
     // the benefit.
case BuiltinType::ULongLong:
       return getTypeInfo(Context.LongLongTy);
     case BuiltinType::UInt128:
       return getTypeInfo(Context.Int128Ty);
 
     // Treat all other builtin types as distinct types. This includes
     // treating wchar_t, char16_t, and char32_t as distinct from their
     // "underlying types".
     default:
-      return createTBAAScalarType(BTy->getName(Features), getChar());
+      return createScalarTypeNode(BTy->getName(Features), getChar(), Size);
     }
   }
 
   // C++1z [basic.lval]p10: "If a program attempts to access the stored value of
   // an object through a glvalue of other than one of the following types the
   // behavior is undefined: [...] a char, unsigned char, or std::byte type."
   if (Ty->isStdByteType())
     return getChar();
 
   // Handle pointers and references.
   // TODO: Implement C++'s type "similarity" and consider dis-"similar"
   // pointers distinct.
   if (Ty->isPointerType() || Ty->isReferenceType())
-    return createTBAAScalarType("any pointer", getChar());
+    return createScalarTypeNode("any pointer", getChar(), Size);
 
   // Enum types are distinct types. In C++ they have "underlying types",
   // however they aren't related for TBAA.
   if (const EnumType *ETy = dyn_cast<EnumType>(Ty)) {
     // In C++ mode, types have linkage, so we can rely on the ODR and
     // on their mangled names, if they're external.
     // TODO: Is there a way to get a program-wide unique name for a
     // decl with local linkage or no linkage?
     if (!Features.CPlusPlus || !ETy->getDecl()->isExternallyVisible())
       return getChar();
 
     SmallString<256> OutName;
     llvm::raw_svector_ostream Out(OutName);
     MContext.mangleTypeName(QualType(ETy, 0), Out);
-    return createTBAAScalarType(OutName, getChar());
+    return createScalarTypeNode(OutName, getChar(), Size);
   }
 
   // For now, handle any other kind of type conservatively.
   return getChar();
 }
 
 llvm::MDNode *CodeGenTBAA::getTypeInfo(QualType QTy) {
   // At -O0 or relaxed aliasing, TBAA is not emitted for regular types.
llvm::MDNode *CodeGenTBAA::getTypeInfo(QualType QTy) {
 
   // Note that the following helper call is allowed to add new nodes to the
   // cache, which invalidates all its previously obtained iterators. So we
   // first generate the node for the type and then add that node to the cache.
   llvm::MDNode *TypeNode = getTypeInfoHelper(Ty);
   return MetadataCache[Ty] = TypeNode;
 }
 
-TBAAAccessInfo CodeGenTBAA::getVTablePtrAccessInfo() {
-  return TBAAAccessInfo(createTBAAScalarType("vtable pointer", getRoot()));
+TBAAAccessInfo CodeGenTBAA::getVTablePtrAccessInfo(llvm::Type *VTablePtrType) {
+  llvm::DataLayout DL(&Module);
+  unsigned Size = DL.getPointerTypeSize(VTablePtrType);
+  return TBAAAccessInfo(createScalarTypeNode("vtable pointer", getRoot(), Size),
+                        Size);
 }
 
 bool
 CodeGenTBAA::CollectFields(uint64_t BaseOffset,
                            QualType QTy,
                            SmallVectorImpl<llvm::MDBuilder::TBAAStructField> &
                              Fields,
                            bool MayAlias) {
if (const RecordType *TTy = QTy->getAs<RecordType>()) {
     }
     return true;
   }
 
   /* Otherwise, treat whatever it is as a field. */
   uint64_t Offset = BaseOffset;
   uint64_t Size = Context.getTypeSizeInChars(QTy).getQuantity();
   llvm::MDNode *TBAAType = MayAlias ? getChar() : getTypeInfo(QTy);
-  llvm::MDNode *TBAATag = getAccessTagInfo(TBAAAccessInfo(TBAAType));
+  llvm::MDNode *TBAATag = getAccessTagInfo(TBAAAccessInfo(TBAAType, Size));
   Fields.push_back(llvm::MDBuilder::TBAAStructField(Offset, Size, TBAATag));
   return true;
 }
 
 llvm::MDNode *
 CodeGenTBAA::getTBAAStructInfo(QualType QTy) {
   const Type *Ty = Context.getCanonicalType(QTy).getTypePtr();
 
   if (llvm::MDNode *N = StructMetadataCache[Ty])
     return N;
 
   SmallVector<llvm::MDBuilder::TBAAStructField, 4> Fields;
   if (CollectFields(0, QTy, Fields, TypeHasMayAlias(QTy)))
     return MDHelper.createTBAAStructNode(Fields);
 
   // For now, handle any other kind of type conservatively.
   return StructMetadataCache[Ty] = nullptr;
 }
 
 llvm::MDNode *CodeGenTBAA::getBaseTypeInfoHelper(const Type *Ty) {
   if (auto *TTy = dyn_cast<RecordType>(Ty)) {
     const RecordDecl *RD = TTy->getDecl()->getDefinition();
-
     const ASTRecordLayout &Layout = Context.getASTRecordLayout(RD);
-    SmallVector <std::pair<llvm::MDNode*, uint64_t>, 4> Fields;
-    unsigned idx = 0;
-    for (RecordDecl::field_iterator i = RD->field_begin(),
-         e = RD->field_end(); i != e; ++i, ++idx) {
-      QualType FieldQTy = i->getType();
-      llvm::MDNode *FieldNode = isValidBaseType(FieldQTy) ?
+    SmallVector<llvm::MDBuilder::TBAAStructField, 4> Fields;
+    for (FieldDecl *Field : RD->fields()) {
+      QualType FieldQTy = Field->getType();
+      llvm::MDNode *TypeNode = isValidBaseType(FieldQTy) ?
           getBaseTypeInfo(FieldQTy) : getTypeInfo(FieldQTy);
-      if (!FieldNode)
+      if (!TypeNode)
         return BaseTypeMetadataCache[Ty] = nullptr;
-      Fields.push_back(std::make_pair(
-          FieldNode, Layout.getFieldOffset(idx) / Context.getCharWidth()));
+
+      uint64_t BitOffset = Layout.getFieldOffset(Field->getFieldIndex());
+      uint64_t Offset = Context.toCharUnitsFromBits(BitOffset).getQuantity();
+      uint64_t Size = Context.getTypeSizeInChars(FieldQTy).getQuantity();
+      Fields.push_back(llvm::MDBuilder::TBAAStructField(Offset, Size,
+                                                        TypeNode));
     }
 
     SmallString<256> OutName;
     if (Features.CPlusPlus) {
       // Don't use the mangler for C code.
       llvm::raw_svector_ostream Out(OutName);
       MContext.mangleTypeName(QualType(Ty, 0), Out);
     } else {
       OutName = RD->getName();
     }
+
+    // TODO: Support size-aware type nodes and create one here for the
+    // given aggregate type.
+
     // Create the struct type node with a vector of pairs (offset, type).
-    return MDHelper.createTBAAStructTypeNode(OutName, Fields);
+    SmallVector<std::pair<llvm::MDNode*, uint64_t>, 4> OffsetsAndTypes;
+    for (const auto &Field : Fields)
+        OffsetsAndTypes.push_back(std::make_pair(Field.TBAA, Field.Offset));
+    return MDHelper.createTBAAStructTypeNode(OutName, OffsetsAndTypes);
   }
 
   return nullptr;
 }
 
 llvm::MDNode *CodeGenTBAA::getBaseTypeInfo(QualType QTy) {
   if (!isValidBaseType(QTy))
     return nullptr;
 
   const Type *Ty = Context.getCanonicalType(QTy).getTypePtr();
   if (llvm::MDNode *N = BaseTypeMetadataCache[Ty])
     return N;
 
   // Note that the following helper call is allowed to add new nodes to the
   // cache, which invalidates all its previously obtained iterators. So we
   // first generate the node for the type and then add that node to the cache.
   llvm::MDNode *TypeNode = getBaseTypeInfoHelper(Ty);
   return BaseTypeMetadataCache[Ty] = TypeNode;
 }
 
 llvm::MDNode *CodeGenTBAA::getAccessTagInfo(TBAAAccessInfo Info) {
+  assert(!Info.isIncomplete() && "Access to an object of an incomplete type!");
+
   if (Info.isMayAlias())
-    Info = TBAAAccessInfo(getChar());
+    Info = TBAAAccessInfo(getChar(), Info.Size);
 
   if (!Info.AccessType)
     return nullptr;
 
   if (!CodeGenOpts.StructPathTBAA)
-    Info = TBAAAccessInfo(Info.AccessType);
+    Info = TBAAAccessInfo(Info.AccessType, Info.Size);
 
   llvm::MDNode *&N = AccessTagMetadataCache[Info];
   if (N)
     return N;
 
   if (!Info.BaseType) {
     Info.BaseType = Info.AccessType;
     assert(!Info.Offset && "Nonzero offset for an access with no base type!");