[WebAssembly] Use section index in relocation section header

[sbc100]

Summary

Rather than referring to sections my their code, use the
absolute index of the target section within the module.

This is non-backwards compatible change to the linking format.

See https://github.com/WebAssembly/tool-conventions/issues/52

[sbc100]

• cleanup
• regenerate test inputs

[sbc100]

lld-side change is one-liner: https://reviews.llvm.org/D45795

[yurydelendik]

Looks good to me. Does it make sense to check/assert in parseRelocSection what type of section is referred to -- just to ensure we are dealing with CODE and DATA sections (while we have multiple reloc section formats in the wild)?

[ncw]

Looks good :)

include/llvm/Object/Wasm.h

private:
   bool isValidGlobalSymbol(uint32_t Index) const;
   bool isValidDataSymbol(uint32_t Index) const;
   wasm::WasmFunction &getDefinedFunction(uint32_t Index);
   wasm::WasmGlobal &getDefinedGlobal(uint32_t Index);
 
   const WasmSection &getWasmSection(DataRefImpl Ref) const;
   const wasm::WasmRelocation &getWasmRelocation(DataRefImpl Ref) const;
 
-  WasmSection* findCustomSectionByName(StringRef Name);
-  WasmSection* findSectionByType(uint32_t Type);
-
   const uint8_t *getPtr(size_t Offset) const;
   Error parseSection(WasmSection &Sec);
   Error parseCustomSection(WasmSection &Sec, const uint8_t *Ptr,
                            const uint8_t *End);
 
   // Standard section types
   Error parseTypeSection(const uint8_t *Ptr, const uint8_t *End);
   Error parseImportSection(const uint8_t *Ptr, const uint8_t *End);

lib/MC/WasmObjectWriter.cpp

 // For patching purposes, we need to remember where each section starts, both
 // for patching up the section size field, and for patching up references to
 // locations within the section.
 struct SectionBookkeeping {
   // Where the size of the section is written.
   uint64_t SizeOffset;
   // Where the contents of the section starts (after the header).
   uint64_t ContentsOffset;
+  uint32_t Index;
 };
 
 // The signature of a wasm function, in a struct capable of being used as a
 // DenseMap key.
 struct WasmFunctionType {
   // Support empty and tombstone instances, needed by DenseMap.
   enum { Plain, Empty, Tombstone } State;
 
Out << "Off=" << Offset << ", Sym=" << *Symbol << ", Addend=" << Addend
         << ", FixupSection=" << FixupSection->getSectionName();
   }
 
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
   LLVM_DUMP_METHOD void dump() const { print(dbgs()); }
 #endif
 };
 
+#if !defined(NDEBUG)
+raw_ostream &operator<<(raw_ostream &OS, const WasmRelocationEntry &Rel) {
+  Rel.print(OS);
+  return OS;
+}
+#endif
+
+struct RelocationSection {
+  uint32_t TargetSection;
+  std::vector<WasmRelocationEntry> Relocations;
+};
+
 struct WasmCustomSection {
   StringRef Name;
   const SmallVectorImpl<char> &Contents;
 
   WasmCustomSection(StringRef Name, const SmallVectorImpl<char> &Contents)
       : Name(Name), Contents(Contents) {}
 };
 
-#if !defined(NDEBUG)
-raw_ostream &operator<<(raw_ostream &OS, const WasmRelocationEntry &Rel) {
-  Rel.print(OS);
-  return OS;
-}
-#endif
-
 class WasmObjectWriter : public MCObjectWriter {
   /// The target specific Wasm writer instance.
   std::unique_ptr<MCWasmObjectTargetWriter> TargetObjectWriter;
 
   // Relocations for fixing up references in the code section.
   std::vector<WasmRelocationEntry> CodeRelocations;
+  uint32_t CodeSectionIndex;
 
   // Relocations for fixing up references in the data section.
   std::vector<WasmRelocationEntry> DataRelocations;
+  uint32_t DataSectionIndex;
 
   // Index values to use for fixing up call_indirect type indices.
   // Maps function symbols to the index of the type of the function
   DenseMap<const MCSymbolWasm *, uint32_t> TypeIndices;
   // Maps function symbols to the table element index space. Used
   // for TABLE_INDEX relocation types (i.e. address taken functions).
   DenseMap<const MCSymbolWasm *, uint32_t> TableIndices;
   // Maps function/global symbols to the (shared) Symbol index space.
   DenseMap<const MCSymbolWasm *, uint32_t> SymbolIndices;
   // Maps function/global symbols to the function/global Wasm index space.
   DenseMap<const MCSymbolWasm *, uint32_t> WasmIndices;
   // Maps data symbols to the Wasm segment and offset/size with the segment.
   DenseMap<const MCSymbolWasm *, wasm::WasmDataReference> DataLocations;
 
   DenseMap<WasmFunctionType, int32_t, WasmFunctionTypeDenseMapInfo>
       FunctionTypeIndices;
   SmallVector<WasmFunctionType, 4> FunctionTypes;
   SmallVector<WasmGlobal, 4> Globals;
   SmallVector<WasmDataSegment, 4> DataSegments;
   std::vector<WasmCustomSection> CustomSections;
   unsigned NumFunctionImports = 0;
   unsigned NumGlobalImports = 0;
+  uint32_t SectionCount;
 
   // TargetObjectWriter wrappers.
   bool is64Bit() const { return TargetObjectWriter->is64Bit(); }
   unsigned getRelocType(const MCValue &Target, const MCFixup &Fixup) const {
     return TargetObjectWriter->getRelocType(Target, Fixup);
   }
 
   void startSection(SectionBookkeeping &Section, unsigned SectionId);
void writeImportSection(ArrayRef<wasm::WasmImport> Imports, uint32_t DataSize,
                           uint32_t NumElements);
   void writeFunctionSection(ArrayRef<WasmFunction> Functions);
   void writeGlobalSection();
   void writeExportSection(ArrayRef<wasm::WasmExport> Exports);
   void writeElemSection(ArrayRef<uint32_t> TableElems);
   void writeCodeSection(const MCAssembler &Asm, const MCAsmLayout &Layout,
                         ArrayRef<WasmFunction> Functions);
   void writeDataSection();
-  void writeCodeRelocSection();
-  void writeDataRelocSection();
+  void writeRelocSection(uint32_t SectionIndex, StringRef Name,
+                         ArrayRef<WasmRelocationEntry> Relocations);
   void writeLinkingMetaDataSection(
       ArrayRef<wasm::WasmSymbolInfo> SymbolInfos,
       ArrayRef<std::pair<uint16_t, uint32_t>> InitFuncs,
       const std::map<StringRef, std::vector<WasmComdatEntry>> &Comdats);
   void writeUserCustomSections(ArrayRef<WasmCustomSection> CustomSections);
 
   uint32_t getProvisionalValue(const WasmRelocationEntry &RelEntry);
   void applyRelocations(ArrayRef<WasmRelocationEntry> Relocations,
                         uint64_t ContentsOffset);
 
-  void writeRelocations(ArrayRef<WasmRelocationEntry> Relocations);
   uint32_t getRelocationIndexValue(const WasmRelocationEntry &RelEntry);
-  uint32_t getFunctionType(const MCSymbolWasm& Symbol);
-  uint32_t registerFunctionType(const MCSymbolWasm& Symbol);
+  uint32_t getFunctionType(const MCSymbolWasm &Symbol);
+  uint32_t registerFunctionType(const MCSymbolWasm &Symbol);
 };
 
 } // end anonymous namespace
 
 WasmObjectWriter::~WasmObjectWriter() {}
 
 // Write out a section header and a patchable section size field.
 void WasmObjectWriter::startSection(SectionBookkeeping &Section,
                                     unsigned SectionId) {
   DEBUG(dbgs() << "startSection " << SectionId << "\n");
   write8(SectionId);
 
   Section.SizeOffset = getStream().tell();
 
   // The section size. We don't know the size yet, so reserve enough space
   // for any 32-bit value; we'll patch it later.
   encodeULEB128(UINT32_MAX, getStream());
 
   // The position where the section starts, for measuring its size.
   Section.ContentsOffset = getStream().tell();
+  Section.Index = SectionCount++;
 }
 
 void WasmObjectWriter::startCustomSection(SectionBookkeeping &Section,
                                           StringRef Name) {
   DEBUG(dbgs() << "startCustomSection " << Name << "\n");
   startSection(Section, wasm::WASM_SEC_CUSTOM);
   // Custom sections in wasm also have a string identifier.
   writeString(Name);
case wasm::R_WEBASSEMBLY_MEMORY_ADDR_SLEB:
       WritePatchableSLEB(Stream, Value, Offset);
       break;
     default:
       llvm_unreachable("invalid relocation type");
     }
   }
 }
 
-// Write out the portions of the relocation records that the linker will
-// need to handle.
-void WasmObjectWriter::writeRelocations(
-    ArrayRef<WasmRelocationEntry> Relocations) {
-  raw_pwrite_stream &Stream = getStream();
-  for (const WasmRelocationEntry& RelEntry : Relocations) {
-
-    uint64_t Offset = RelEntry.Offset +
-                      RelEntry.FixupSection->getSectionOffset();
-    uint32_t Index = getRelocationIndexValue(RelEntry);
-
-    write8(RelEntry.Type);
-    encodeULEB128(Offset, Stream);
-    encodeULEB128(Index, Stream);
-    if (RelEntry.hasAddend())
-      encodeSLEB128(RelEntry.Addend, Stream);
-  }
-}
-
 void WasmObjectWriter::writeTypeSection(
     ArrayRef<WasmFunctionType> FunctionTypes) {
   if (FunctionTypes.empty())
     return;
 
   SectionBookkeeping Section;
   startSection(Section, wasm::WASM_SEC_TYPE);
 

 void WasmObjectWriter::writeCodeSection(const MCAssembler &Asm,
                                         const MCAsmLayout &Layout,
                                         ArrayRef<WasmFunction> Functions) {
   if (Functions.empty())
     return;
 
   SectionBookkeeping Section;
   startSection(Section, wasm::WASM_SEC_CODE);
+  CodeSectionIndex = Section.Index;
 
   encodeULEB128(Functions.size(), getStream());
 
   for (const WasmFunction &Func : Functions) {
     auto &FuncSection = static_cast<MCSectionWasm &>(Func.Sym->getSection());
 
     int64_t Size = 0;
     if (!Func.Sym->getSize()->evaluateAsAbsolute(Size, Layout))

 }
 
 void WasmObjectWriter::writeDataSection() {
   if (DataSegments.empty())
     return;
 
   SectionBookkeeping Section;
   startSection(Section, wasm::WASM_SEC_DATA);
+  DataSectionIndex = Section.Index;
 
   encodeULEB128(DataSegments.size(), getStream()); // count
 
   for (const WasmDataSegment &Segment : DataSegments) {
     encodeULEB128(0, getStream()); // memory index
     write8(wasm::WASM_OPCODE_I32_CONST);
     encodeSLEB128(Segment.Offset, getStream()); // offset
     write8(wasm::WASM_OPCODE_END);
     encodeULEB128(Segment.Data.size(), getStream()); // size
     Segment.Section->setSectionOffset(getStream().tell() - Section.ContentsOffset);
     writeBytes(Segment.Data); // data
   }
 
   // Apply fixups.
   applyRelocations(DataRelocations, Section.ContentsOffset);
 
   endSection(Section);
 }
 
-void WasmObjectWriter::writeCodeRelocSection() {
+void WasmObjectWriter::writeRelocSection(
+    uint32_t SectionIndex, StringRef Name,
+    ArrayRef<WasmRelocationEntry> Relocations) {
   // See: https://github.com/WebAssembly/tool-conventions/blob/master/Linking.md
   // for descriptions of the reloc sections.
 
-  if (CodeRelocations.empty())
+  if (Relocations.empty())
     return;
 
   SectionBookkeeping Section;
-  startCustomSection(Section, "reloc.CODE");
+  startCustomSection(Section, std::string("reloc.") + Name.str());
 
-  encodeULEB128(wasm::WASM_SEC_CODE, getStream());
-  encodeULEB128(CodeRelocations.size(), getStream());
+  raw_pwrite_stream &Stream = getStream();
 
-  writeRelocations(CodeRelocations);
+  encodeULEB128(SectionIndex, Stream);
+  encodeULEB128(Relocations.size(), Stream);
+  for (const WasmRelocationEntry& RelEntry : Relocations) {
+    uint64_t Offset = RelEntry.Offset +
+                      RelEntry.FixupSection->getSectionOffset();
+    uint32_t Index = getRelocationIndexValue(RelEntry);
 
-  endSection(Section);
+    write8(RelEntry.Type);
+    encodeULEB128(Offset, Stream);
+    encodeULEB128(Index, Stream);
+    if (RelEntry.hasAddend())
+      encodeSLEB128(RelEntry.Addend, Stream);
   }
 
-void WasmObjectWriter::writeDataRelocSection() {
-  // See: https://github.com/WebAssembly/tool-conventions/blob/master/Linking.md
-  // for descriptions of the reloc sections.
-
-  if (DataRelocations.empty())
-    return;
-
-  SectionBookkeeping Section;
-  startCustomSection(Section, "reloc.DATA");
-
-  encodeULEB128(wasm::WASM_SEC_DATA, getStream());
-  encodeULEB128(DataRelocations.size(), getStream());
-
-  writeRelocations(DataRelocations);
-
   endSection(Section);
 }
 
 void WasmObjectWriter::writeLinkingMetaDataSection(
     ArrayRef<wasm::WasmSymbolInfo> SymbolInfos,
     ArrayRef<std::pair<uint16_t, uint32_t>> InitFuncs,
     const std::map<StringRef, std::vector<WasmComdatEntry>> &Comdats) {
   SectionBookkeeping Section;
void WasmObjectWriter::writeObject(MCAssembler &Asm,
   // Skip the "memory" section; we import the memory instead.
   writeGlobalSection();
   writeExportSection(Exports);
   writeElemSection(TableElems);
   writeCodeSection(Asm, Layout, Functions);
   writeDataSection();
   writeUserCustomSections(CustomSections);
   writeLinkingMetaDataSection(SymbolInfos, InitFuncs, Comdats);
-  writeCodeRelocSection();
-  writeDataRelocSection();
+  writeRelocSection(CodeSectionIndex, "CODE", CodeRelocations);
+  writeRelocSection(DataSectionIndex, "DATA", DataRelocations);
 
   // TODO: Translate the .comment section to the output.
   // TODO: Translate debug sections to the output.
 }
 
 std::unique_ptr<MCObjectWriter>
 llvm::createWasmObjectWriter(std::unique_ptr<MCWasmObjectTargetWriter> MOTW,
                              raw_pwrite_stream &OS) {
   return llvm::make_unique<WasmObjectWriter>(std::move(MOTW), OS);
 }

lib/Object/WasmObjectFile.cpp

while (EntryCount--) {
         getDefinedFunction(Index).Comdat = ComdatIndex;
         break;
       }
     }
   }
   return Error::success();
 }
 
-WasmSection* WasmObjectFile::findCustomSectionByName(StringRef Name) {
-  for (WasmSection& Section : Sections) {
-    if (Section.Type == wasm::WASM_SEC_CUSTOM && Section.Name == Name)
-      return &Section;
-  }
-  return nullptr;
-}
-
-WasmSection* WasmObjectFile::findSectionByType(uint32_t Type) {
-  assert(Type != wasm::WASM_SEC_CUSTOM);
-  for (WasmSection& Section : Sections) {
-    if (Section.Type == Type)
-      return &Section;
-  }
-  return nullptr;
-}
-
 Error WasmObjectFile::parseRelocSection(StringRef Name, const uint8_t *Ptr,
                                         const uint8_t *End) {
-  uint8_t SectionCode = readUint8(Ptr);
-  WasmSection* Section = nullptr;
-  if (SectionCode == wasm::WASM_SEC_CUSTOM) {
-    StringRef Name = readString(Ptr);
-    Section = findCustomSectionByName(Name);
-  } else {
-    Section = findSectionByType(SectionCode);
-  }
-  if (!Section)
-    return make_error<GenericBinaryError>("Invalid section code",
+  uint8_t SectionIndex = readUint8(Ptr);
+  if (SectionIndex >= Sections.size())
+    return make_error<GenericBinaryError>("Invalid section index",
                                           object_error::parse_failed);
+  WasmSection& Section = Sections[SectionIndex];
   uint32_t RelocCount = readVaruint32(Ptr);
-  uint32_t EndOffset = Section->Content.size();
+  uint32_t EndOffset = Section.Content.size();
   while (RelocCount--) {
     wasm::WasmRelocation Reloc = {};
     Reloc.Type = readVaruint32(Ptr);
     Reloc.Offset = readVaruint32(Ptr);
     Reloc.Index = readVaruint32(Ptr);
     switch (Reloc.Type) {
     case wasm::R_WEBASSEMBLY_FUNCTION_INDEX_LEB:
     case wasm::R_WEBASSEMBLY_TABLE_INDEX_SLEB:
while (RelocCount--) {
     uint64_t Size = 5;
     if (Reloc.Type == wasm::R_WEBASSEMBLY_TABLE_INDEX_I32 ||
         Reloc.Type == wasm::R_WEBASSEMBLY_MEMORY_ADDR_I32)
       Size = 4;
     if (Reloc.Offset + Size > EndOffset)
       return make_error<GenericBinaryError>("Bad relocation offset",
                                             object_error::parse_failed);
 
-    Section->Relocations.push_back(Reloc);
+    Section.Relocations.push_back(Reloc);
   }
   if (Ptr != End)
     return make_error<GenericBinaryError>("Reloc section ended prematurely",
                                           object_error::parse_failed);
   return Error::success();
 }
 
 Error WasmObjectFile::parseCustomSection(WasmSection &Sec,