[flang] Support OPEN(..., FORM="BINARY") (#124657)

[llvm-project.git] / lld / ELF / InputFiles.cpp

//===- InputFiles.cpp -----------------------------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//

#include "InputFiles.h"
#include "Config.h"
#include "DWARF.h"
#include "Driver.h"
#include "InputSection.h"
#include "LinkerScript.h"
#include "SymbolTable.h"
#include "Symbols.h"
#include "SyntheticSections.h"
#include "Target.h"
#include "lld/Common/CommonLinkerContext.h"
#include "lld/Common/DWARF.h"
#include "llvm/ADT/CachedHashString.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/LTO/LTO.h"
#include "llvm/Object/IRObjectFile.h"
#include "llvm/Support/ARMAttributeParser.h"
#include "llvm/Support/ARMBuildAttributes.h"
#include "llvm/Support/Endian.h"
#include "llvm/Support/FileSystem.h"
#include "llvm/Support/Path.h"
#include "llvm/Support/RISCVAttributeParser.h"
#include "llvm/Support/TimeProfiler.h"
#include "llvm/Support/raw_ostream.h"
#include <optional>

using namespace llvm;
using namespace llvm::ELF;
using namespace llvm::object;
using namespace llvm::sys;
using namespace llvm::sys::fs;
using namespace llvm::support::endian;
using namespace lld;
using namespace lld::elf;

// This function is explicitly instantiated in ARM.cpp, don't do it here to
// avoid warnings with MSVC.
extern template void ObjFile<ELF32LE>::importCmseSymbols();
extern template void ObjFile<ELF32BE>::importCmseSymbols();
extern template void ObjFile<ELF64LE>::importCmseSymbols();
extern template void ObjFile<ELF64BE>::importCmseSymbols();

// Returns "<internal>", "foo.a(bar.o)" or "baz.o".
std::string elf::toStr(Ctx &ctx, const InputFile *f) {
  static std::mutex mu;
  if (!f)
    return "<internal>";

  {
    std::lock_guard<std::mutex> lock(mu);
    if (f->toStringCache.empty()) {
      if (f->archiveName.empty())
        f->toStringCache = f->getName();
      else
        (f->archiveName + "(" + f->getName() + ")").toVector(f->toStringCache);
    }
  }
  return std::string(f->toStringCache);
}

const ELFSyncStream &elf::operator<<(const ELFSyncStream &s,
                                     const InputFile *f) {
  return s << toStr(s.ctx, f);
}

static ELFKind getELFKind(Ctx &ctx, MemoryBufferRef mb, StringRef archiveName) {
  unsigned char size;
  unsigned char endian;
  std::tie(size, endian) = getElfArchType(mb.getBuffer());

  auto report = [&](StringRef msg) {
    StringRef filename = mb.getBufferIdentifier();
    if (archiveName.empty())
      Fatal(ctx) << filename << ": " << msg;
    else
      Fatal(ctx) << archiveName << "(" << filename << "): " << msg;
  };

  if (!mb.getBuffer().starts_with(ElfMagic))
    report("not an ELF file");
  if (endian != ELFDATA2LSB && endian != ELFDATA2MSB)
    report("corrupted ELF file: invalid data encoding");
  if (size != ELFCLASS32 && size != ELFCLASS64)
    report("corrupted ELF file: invalid file class");

  size_t bufSize = mb.getBuffer().size();
  if ((size == ELFCLASS32 && bufSize < sizeof(Elf32_Ehdr)) ||
      (size == ELFCLASS64 && bufSize < sizeof(Elf64_Ehdr)))
    report("corrupted ELF file: file is too short");

  if (size == ELFCLASS32)
    return (endian == ELFDATA2LSB) ? ELF32LEKind : ELF32BEKind;
  return (endian == ELFDATA2LSB) ? ELF64LEKind : ELF64BEKind;
}

// For ARM only, to set the EF_ARM_ABI_FLOAT_SOFT or EF_ARM_ABI_FLOAT_HARD
// flag in the ELF Header we need to look at Tag_ABI_VFP_args to find out how
// the input objects have been compiled.
static void updateARMVFPArgs(Ctx &ctx, const ARMAttributeParser &attributes,
                             const InputFile *f) {
  std::optional<unsigned> attr =
      attributes.getAttributeValue(ARMBuildAttrs::ABI_VFP_args);
  if (!attr)
    // If an ABI tag isn't present then it is implicitly given the value of 0
    // which maps to ARMBuildAttrs::BaseAAPCS. However many assembler files,
    // including some in glibc that don't use FP args (and should have value 3)
    // don't have the attribute so we do not consider an implicit value of 0
    // as a clash.
    return;

  unsigned vfpArgs = *attr;
  ARMVFPArgKind arg;
  switch (vfpArgs) {
  case ARMBuildAttrs::BaseAAPCS:
    arg = ARMVFPArgKind::Base;
    break;
  case ARMBuildAttrs::HardFPAAPCS:
    arg = ARMVFPArgKind::VFP;
    break;
  case ARMBuildAttrs::ToolChainFPPCS:
    // Tool chain specific convention that conforms to neither AAPCS variant.
    arg = ARMVFPArgKind::ToolChain;
    break;
  case ARMBuildAttrs::CompatibleFPAAPCS:
    // Object compatible with all conventions.
    return;
  default:
    ErrAlways(ctx) << f << ": unknown Tag_ABI_VFP_args value: " << vfpArgs;
    return;
  }
  // Follow ld.bfd and error if there is a mix of calling conventions.
  if (ctx.arg.armVFPArgs != arg && ctx.arg.armVFPArgs != ARMVFPArgKind::Default)
    ErrAlways(ctx) << f << ": incompatible Tag_ABI_VFP_args";
  else
    ctx.arg.armVFPArgs = arg;
}

// The ARM support in lld makes some use of instructions that are not available
// on all ARM architectures. Namely:
// - Use of BLX instruction for interworking between ARM and Thumb state.
// - Use of the extended Thumb branch encoding in relocation.
// - Use of the MOVT/MOVW instructions in Thumb Thunks.
// The ARM Attributes section contains information about the architecture chosen
// at compile time. We follow the convention that if at least one input object
// is compiled with an architecture that supports these features then lld is
// permitted to use them.
static void updateSupportedARMFeatures(Ctx &ctx,
                                       const ARMAttributeParser &attributes) {
  std::optional<unsigned> attr =
      attributes.getAttributeValue(ARMBuildAttrs::CPU_arch);
  if (!attr)
    return;
  auto arch = *attr;
  switch (arch) {
  case ARMBuildAttrs::Pre_v4:
  case ARMBuildAttrs::v4:
  case ARMBuildAttrs::v4T:
    // Architectures prior to v5 do not support BLX instruction
    break;
  case ARMBuildAttrs::v5T:
  case ARMBuildAttrs::v5TE:
  case ARMBuildAttrs::v5TEJ:
  case ARMBuildAttrs::v6:
  case ARMBuildAttrs::v6KZ:
  case ARMBuildAttrs::v6K:
    ctx.arg.armHasBlx = true;
    // Architectures used in pre-Cortex processors do not support
    // The J1 = 1 J2 = 1 Thumb branch range extension, with the exception
    // of Architecture v6T2 (arm1156t2-s and arm1156t2f-s) that do.
    break;
  default:
    // All other Architectures have BLX and extended branch encoding
    ctx.arg.armHasBlx = true;
    ctx.arg.armJ1J2BranchEncoding = true;
    if (arch != ARMBuildAttrs::v6_M && arch != ARMBuildAttrs::v6S_M)
      // All Architectures used in Cortex processors with the exception
      // of v6-M and v6S-M have the MOVT and MOVW instructions.
      ctx.arg.armHasMovtMovw = true;
    break;
  }

  // Only ARMv8-M or later architectures have CMSE support.
  std::optional<unsigned> profile =
      attributes.getAttributeValue(ARMBuildAttrs::CPU_arch_profile);
  if (!profile)
    return;
  if (arch >= ARMBuildAttrs::CPUArch::v8_M_Base &&
      profile == ARMBuildAttrs::MicroControllerProfile)
    ctx.arg.armCMSESupport = true;

  // The thumb PLT entries require Thumb2 which can be used on multiple archs.
  // For now, let's limit it to ones where ARM isn't available and we know have
  // Thumb2.
  std::optional<unsigned> armISA =
      attributes.getAttributeValue(ARMBuildAttrs::ARM_ISA_use);
  std::optional<unsigned> thumb =
      attributes.getAttributeValue(ARMBuildAttrs::THUMB_ISA_use);
  ctx.arg.armHasArmISA |= armISA && *armISA >= ARMBuildAttrs::Allowed;
  ctx.arg.armHasThumb2ISA |= thumb && *thumb >= ARMBuildAttrs::AllowThumb32;
}

InputFile::InputFile(Ctx &ctx, Kind k, MemoryBufferRef m)
    : ctx(ctx), mb(m), groupId(ctx.driver.nextGroupId), fileKind(k) {
  // All files within the same --{start,end}-group get the same group ID.
  // Otherwise, a new file will get a new group ID.
  if (!ctx.driver.isInGroup)
    ++ctx.driver.nextGroupId;
}

InputFile::~InputFile() {}

std::optional<MemoryBufferRef> elf::readFile(Ctx &ctx, StringRef path) {
  llvm::TimeTraceScope timeScope("Load input files", path);

  // The --chroot option changes our virtual root directory.
  // This is useful when you are dealing with files created by --reproduce.
  if (!ctx.arg.chroot.empty() && path.starts_with("/"))
    path = ctx.saver.save(ctx.arg.chroot + path);

  bool remapped = false;
  auto it = ctx.arg.remapInputs.find(path);
  if (it != ctx.arg.remapInputs.end()) {
    path = it->second;
    remapped = true;
  } else {
    for (const auto &[pat, toFile] : ctx.arg.remapInputsWildcards) {
      if (pat.match(path)) {
        path = toFile;
        remapped = true;
        break;
      }
    }
  }
  if (remapped) {
    // Use /dev/null to indicate an input file that should be ignored. Change
    // the path to NUL on Windows.
#ifdef _WIN32
    if (path == "/dev/null")
      path = "NUL";
#endif
  }

  Log(ctx) << path;
  ctx.arg.dependencyFiles.insert(llvm::CachedHashString(path));

  auto mbOrErr = MemoryBuffer::getFile(path, /*IsText=*/false,
                                       /*RequiresNullTerminator=*/false);
  if (auto ec = mbOrErr.getError()) {
    ErrAlways(ctx) << "cannot open " << path << ": " << ec.message();
    return std::nullopt;
  }

  MemoryBufferRef mbref = (*mbOrErr)->getMemBufferRef();
  ctx.memoryBuffers.push_back(std::move(*mbOrErr)); // take MB ownership

  if (ctx.tar)
    ctx.tar->append(relativeToRoot(path), mbref.getBuffer());
  return mbref;
}

// All input object files must be for the same architecture
// (e.g. it does not make sense to link x86 object files with
// MIPS object files.) This function checks for that error.
static bool isCompatible(Ctx &ctx, InputFile *file) {
  if (!file->isElf() && !isa<BitcodeFile>(file))
    return true;

  if (file->ekind == ctx.arg.ekind && file->emachine == ctx.arg.emachine) {
    if (ctx.arg.emachine != EM_MIPS)
      return true;
    if (isMipsN32Abi(ctx, *file) == ctx.arg.mipsN32Abi)
      return true;
  }

  StringRef target =
      !ctx.arg.bfdname.empty() ? ctx.arg.bfdname : ctx.arg.emulation;
  if (!target.empty()) {
    Err(ctx) << file << " is incompatible with " << target;
    return false;
  }

  InputFile *existing = nullptr;
  if (!ctx.objectFiles.empty())
    existing = ctx.objectFiles[0];
  else if (!ctx.sharedFiles.empty())
    existing = ctx.sharedFiles[0];
  else if (!ctx.bitcodeFiles.empty())
    existing = ctx.bitcodeFiles[0];
  auto diag = Err(ctx);
  diag << file << " is incompatible";
  if (existing)
    diag << " with " << existing;
  return false;
}

template <class ELFT> static void doParseFile(Ctx &ctx, InputFile *file) {
  if (!isCompatible(ctx, file))
    return;

  // Lazy object file
  if (file->lazy) {
    if (auto *f = dyn_cast<BitcodeFile>(file)) {
      ctx.lazyBitcodeFiles.push_back(f);
      f->parseLazy();
    } else {
      cast<ObjFile<ELFT>>(file)->parseLazy();
    }
    return;
  }

  if (ctx.arg.trace)
    Msg(ctx) << file;

  if (file->kind() == InputFile::ObjKind) {
    ctx.objectFiles.push_back(cast<ELFFileBase>(file));
    cast<ObjFile<ELFT>>(file)->parse();
  } else if (auto *f = dyn_cast<SharedFile>(file)) {
    f->parse<ELFT>();
  } else if (auto *f = dyn_cast<BitcodeFile>(file)) {
    ctx.bitcodeFiles.push_back(f);
    f->parse();
  } else {
    ctx.binaryFiles.push_back(cast<BinaryFile>(file));
    cast<BinaryFile>(file)->parse();
  }
}

// Add symbols in File to the symbol table.
void elf::parseFile(Ctx &ctx, InputFile *file) {
  invokeELFT(doParseFile, ctx, file);
}

// This function is explicitly instantiated in ARM.cpp. Mark it extern here,
// to avoid warnings when building with MSVC.
extern template void ObjFile<ELF32LE>::importCmseSymbols();
extern template void ObjFile<ELF32BE>::importCmseSymbols();
extern template void ObjFile<ELF64LE>::importCmseSymbols();
extern template void ObjFile<ELF64BE>::importCmseSymbols();

template <class ELFT>
static void
doParseFiles(Ctx &ctx,
             const SmallVector<std::unique_ptr<InputFile>, 0> &files) {
  // Add all files to the symbol table. This will add almost all symbols that we
  // need to the symbol table. This process might add files to the link due to
  // addDependentLibrary.
  for (size_t i = 0; i < files.size(); ++i) {
    llvm::TimeTraceScope timeScope("Parse input files", files[i]->getName());
    doParseFile<ELFT>(ctx, files[i].get());
  }
  if (ctx.driver.armCmseImpLib)
    cast<ObjFile<ELFT>>(*ctx.driver.armCmseImpLib).importCmseSymbols();
}

void elf::parseFiles(Ctx &ctx,
                     const SmallVector<std::unique_ptr<InputFile>, 0> &files) {
  llvm::TimeTraceScope timeScope("Parse input files");
  invokeELFT(doParseFiles, ctx, files);
}

// Concatenates arguments to construct a string representing an error location.
StringRef InputFile::getNameForScript() const {
  if (archiveName.empty())
    return getName();

  if (nameForScriptCache.empty())
    nameForScriptCache = (archiveName + Twine(':') + getName()).str();

  return nameForScriptCache;
}

// An ELF object file may contain a `.deplibs` section. If it exists, the
// section contains a list of library specifiers such as `m` for libm. This
// function resolves a given name by finding the first matching library checking
// the various ways that a library can be specified to LLD. This ELF extension
// is a form of autolinking and is called `dependent libraries`. It is currently
// unique to LLVM and lld.
static void addDependentLibrary(Ctx &ctx, StringRef specifier,
                                const InputFile *f) {
  if (!ctx.arg.dependentLibraries)
    return;
  if (std::optional<std::string> s = searchLibraryBaseName(ctx, specifier))
    ctx.driver.addFile(ctx.saver.save(*s), /*withLOption=*/true);
  else if (std::optional<std::string> s = findFromSearchPaths(ctx, specifier))
    ctx.driver.addFile(ctx.saver.save(*s), /*withLOption=*/true);
  else if (fs::exists(specifier))
    ctx.driver.addFile(specifier, /*withLOption=*/false);
  else
    ErrAlways(ctx)
        << f << ": unable to find library from dependent library specifier: "
        << specifier;
}

// Record the membership of a section group so that in the garbage collection
// pass, section group members are kept or discarded as a unit.
template <class ELFT>
static void handleSectionGroup(ArrayRef<InputSectionBase *> sections,
                               ArrayRef<typename ELFT::Word> entries) {
  bool hasAlloc = false;
  for (uint32_t index : entries.slice(1)) {
    if (index >= sections.size())
      return;
    if (InputSectionBase *s = sections[index])
      if (s != &InputSection::discarded && s->flags & SHF_ALLOC)
        hasAlloc = true;
  }

  // If any member has the SHF_ALLOC flag, the whole group is subject to garbage
  // collection. See the comment in markLive(). This rule retains .debug_types
  // and .rela.debug_types.
  if (!hasAlloc)
    return;

  // Connect the members in a circular doubly-linked list via
  // nextInSectionGroup.
  InputSectionBase *head;
  InputSectionBase *prev = nullptr;
  for (uint32_t index : entries.slice(1)) {
    InputSectionBase *s = sections[index];
    if (!s || s == &InputSection::discarded)
      continue;
    if (prev)
      prev->nextInSectionGroup = s;
    else
      head = s;
    prev = s;
  }
  if (prev)
    prev->nextInSectionGroup = head;
}

template <class ELFT> void ObjFile<ELFT>::initDwarf() {
  dwarf = std::make_unique<DWARFCache>(std::make_unique<DWARFContext>(
      std::make_unique<LLDDwarfObj<ELFT>>(this), "",
      [&](Error err) { Warn(ctx) << getName() + ": " << std::move(err); },
      [&](Error warning) {
        Warn(ctx) << getName() << ": " << std::move(warning);
      }));
}

DWARFCache *ELFFileBase::getDwarf() {
  assert(fileKind == ObjKind);
  llvm::call_once(initDwarf, [this]() {
    switch (ekind) {
    default:
      llvm_unreachable("");
    case ELF32LEKind:
      return cast<ObjFile<ELF32LE>>(this)->initDwarf();
    case ELF32BEKind:
      return cast<ObjFile<ELF32BE>>(this)->initDwarf();
    case ELF64LEKind:
      return cast<ObjFile<ELF64LE>>(this)->initDwarf();
    case ELF64BEKind:
      return cast<ObjFile<ELF64BE>>(this)->initDwarf();
    }
  });
  return dwarf.get();
}

ELFFileBase::ELFFileBase(Ctx &ctx, Kind k, ELFKind ekind, MemoryBufferRef mb)
    : InputFile(ctx, k, mb) {
  this->ekind = ekind;
}

ELFFileBase::~ELFFileBase() {}

template <typename Elf_Shdr>
static const Elf_Shdr *findSection(ArrayRef<Elf_Shdr> sections, uint32_t type) {
  for (const Elf_Shdr &sec : sections)
    if (sec.sh_type == type)
      return &sec;
  return nullptr;
}

void ELFFileBase::init() {
  switch (ekind) {
  case ELF32LEKind:
    init<ELF32LE>(fileKind);
    break;
  case ELF32BEKind:
    init<ELF32BE>(fileKind);
    break;
  case ELF64LEKind:
    init<ELF64LE>(fileKind);
    break;
  case ELF64BEKind:
    init<ELF64BE>(fileKind);
    break;
  default:
    llvm_unreachable("getELFKind");
  }
}

template <class ELFT> void ELFFileBase::init(InputFile::Kind k) {
  using Elf_Shdr = typename ELFT::Shdr;
  using Elf_Sym = typename ELFT::Sym;

  // Initialize trivial attributes.
  const ELFFile<ELFT> &obj = getObj<ELFT>();
  emachine = obj.getHeader().e_machine;
  osabi = obj.getHeader().e_ident[llvm::ELF::EI_OSABI];
  abiVersion = obj.getHeader().e_ident[llvm::ELF::EI_ABIVERSION];

  ArrayRef<Elf_Shdr> sections = CHECK2(obj.sections(), this);
  elfShdrs = sections.data();
  numELFShdrs = sections.size();

  // Find a symbol table.
  const Elf_Shdr *symtabSec =
      findSection(sections, k == SharedKind ? SHT_DYNSYM : SHT_SYMTAB);

  if (!symtabSec)
    return;

  // Initialize members corresponding to a symbol table.
  firstGlobal = symtabSec->sh_info;

  ArrayRef<Elf_Sym> eSyms = CHECK2(obj.symbols(symtabSec), this);
  if (firstGlobal == 0 || firstGlobal > eSyms.size())
    Fatal(ctx) << this << ": invalid sh_info in symbol table";

  elfSyms = reinterpret_cast<const void *>(eSyms.data());
  numSymbols = eSyms.size();
  stringTable = CHECK2(obj.getStringTableForSymtab(*symtabSec, sections), this);
}

template <class ELFT>
uint32_t ObjFile<ELFT>::getSectionIndex(const Elf_Sym &sym) const {
  return CHECK2(
      this->getObj().getSectionIndex(sym, getELFSyms<ELFT>(), shndxTable),
      this);
}

template <class ELFT> void ObjFile<ELFT>::parse(bool ignoreComdats) {
  object::ELFFile<ELFT> obj = this->getObj();
  // Read a section table. justSymbols is usually false.
  if (this->justSymbols) {
    initializeJustSymbols();
    initializeSymbols(obj);
    return;
  }

  // Handle dependent libraries and selection of section groups as these are not
  // done in parallel.
  ArrayRef<Elf_Shdr> objSections = getELFShdrs<ELFT>();
  StringRef shstrtab = CHECK2(obj.getSectionStringTable(objSections), this);
  uint64_t size = objSections.size();
  sections.resize(size);
  for (size_t i = 0; i != size; ++i) {
    const Elf_Shdr &sec = objSections[i];
    if (LLVM_LIKELY(sec.sh_type == SHT_PROGBITS))
      continue;
    if (LLVM_LIKELY(sec.sh_type == SHT_GROUP)) {
      StringRef signature = getShtGroupSignature(objSections, sec);
      ArrayRef<Elf_Word> entries =
          CHECK2(obj.template getSectionContentsAsArray<Elf_Word>(sec), this);
      if (entries.empty())
        Fatal(ctx) << this << ": empty SHT_GROUP";

      Elf_Word flag = entries[0];
      if (flag && flag != GRP_COMDAT)
        Fatal(ctx) << this << ": unsupported SHT_GROUP format";

      bool keepGroup = !flag || ignoreComdats ||
                       ctx.symtab->comdatGroups
                           .try_emplace(CachedHashStringRef(signature), this)
                           .second;
      if (keepGroup) {
        if (!ctx.arg.resolveGroups)
          sections[i] = createInputSection(
              i, sec, check(obj.getSectionName(sec, shstrtab)));
      } else {
        // Otherwise, discard group members.
        for (uint32_t secIndex : entries.slice(1)) {
          if (secIndex >= size)
            Fatal(ctx) << this
                       << ": invalid section index in group: " << secIndex;
          sections[secIndex] = &InputSection::discarded;
        }
      }
      continue;
    }

    if (sec.sh_type == SHT_LLVM_DEPENDENT_LIBRARIES && !ctx.arg.relocatable) {
      StringRef name = check(obj.getSectionName(sec, shstrtab));
      ArrayRef<char> data = CHECK2(
          this->getObj().template getSectionContentsAsArray<char>(sec), this);
      if (!data.empty() && data.back() != '\0') {
        Err(ctx)
            << this
            << ": corrupted dependent libraries section (unterminated string): "
            << name;
      } else {
        for (const char *d = data.begin(), *e = data.end(); d < e;) {
          StringRef s(d);
          addDependentLibrary(ctx, s, this);
          d += s.size() + 1;
        }
      }
      sections[i] = &InputSection::discarded;
      continue;
    }

    switch (ctx.arg.emachine) {
    case EM_ARM:
      if (sec.sh_type == SHT_ARM_ATTRIBUTES) {
        ARMAttributeParser attributes;
        ArrayRef<uint8_t> contents =
            check(this->getObj().getSectionContents(sec));
        StringRef name = check(obj.getSectionName(sec, shstrtab));
        sections[i] = &InputSection::discarded;
        if (Error e = attributes.parse(contents, ekind == ELF32LEKind
                                                     ? llvm::endianness::little
                                                     : llvm::endianness::big)) {
          InputSection isec(*this, sec, name);
          Warn(ctx) << &isec << ": " << std::move(e);
        } else {
          updateSupportedARMFeatures(ctx, attributes);
          updateARMVFPArgs(ctx, attributes, this);

          // FIXME: Retain the first attribute section we see. The eglibc ARM
          // dynamic loaders require the presence of an attribute section for
          // dlopen to work. In a full implementation we would merge all
          // attribute sections.
          if (ctx.in.attributes == nullptr) {
            ctx.in.attributes =
                std::make_unique<InputSection>(*this, sec, name);
            sections[i] = ctx.in.attributes.get();
          }
        }
      }
      break;
    case EM_AARCH64:
      // Producing a static binary with MTE globals is not currently supported,
      // remove all SHT_AARCH64_MEMTAG_GLOBALS_STATIC sections as they're unused
      // medatada, and we don't want them to end up in the output file for
      // static executables.
      if (sec.sh_type == SHT_AARCH64_MEMTAG_GLOBALS_STATIC &&
          !canHaveMemtagGlobals(ctx))
        sections[i] = &InputSection::discarded;
      break;
    }
  }

  // Read a symbol table.
  initializeSymbols(obj);
}

// Sections with SHT_GROUP and comdat bits define comdat section groups.
// They are identified and deduplicated by group name. This function
// returns a group name.
template <class ELFT>
StringRef ObjFile<ELFT>::getShtGroupSignature(ArrayRef<Elf_Shdr> sections,
                                              const Elf_Shdr &sec) {
  typename ELFT::SymRange symbols = this->getELFSyms<ELFT>();
  if (sec.sh_info >= symbols.size())
    Fatal(ctx) << this << ": invalid symbol index";
  const typename ELFT::Sym &sym = symbols[sec.sh_info];
  return CHECK2(sym.getName(this->stringTable), this);
}

template <class ELFT>
bool ObjFile<ELFT>::shouldMerge(const Elf_Shdr &sec, StringRef name) {
  // On a regular link we don't merge sections if -O0 (default is -O1). This
  // sometimes makes the linker significantly faster, although the output will
  // be bigger.
  //
  // Doing the same for -r would create a problem as it would combine sections
  // with different sh_entsize. One option would be to just copy every SHF_MERGE
  // section as is to the output. While this would produce a valid ELF file with
  // usable SHF_MERGE sections, tools like (llvm-)?dwarfdump get confused when
  // they see two .debug_str. We could have separate logic for combining
  // SHF_MERGE sections based both on their name and sh_entsize, but that seems
  // to be more trouble than it is worth. Instead, we just use the regular (-O1)
  // logic for -r.
  if (ctx.arg.optimize == 0 && !ctx.arg.relocatable)
    return false;

  // A mergeable section with size 0 is useless because they don't have
  // any data to merge. A mergeable string section with size 0 can be
  // argued as invalid because it doesn't end with a null character.
  // We'll avoid a mess by handling them as if they were non-mergeable.
  if (sec.sh_size == 0)
    return false;

  // Check for sh_entsize. The ELF spec is not clear about the zero
  // sh_entsize. It says that "the member [sh_entsize] contains 0 if
  // the section does not hold a table of fixed-size entries". We know
  // that Rust 1.13 produces a string mergeable section with a zero
  // sh_entsize. Here we just accept it rather than being picky about it.
  uint64_t entSize = sec.sh_entsize;
  if (entSize == 0)
    return false;
  if (sec.sh_size % entSize)
    ErrAlways(ctx) << this << ":(" << name << "): SHF_MERGE section size ("
                   << uint64_t(sec.sh_size)
                   << ") must be a multiple of sh_entsize (" << entSize << ")";
  if (sec.sh_flags & SHF_WRITE)
    Err(ctx) << this << ":(" << name
             << "): writable SHF_MERGE section is not supported";

  return true;
}

// This is for --just-symbols.
//
// --just-symbols is a very minor feature that allows you to link your
// output against other existing program, so that if you load both your
// program and the other program into memory, your output can refer the
// other program's symbols.
//
// When the option is given, we link "just symbols". The section table is
// initialized with null pointers.
template <class ELFT> void ObjFile<ELFT>::initializeJustSymbols() {
  sections.resize(numELFShdrs);
}

static bool isKnownSpecificSectionType(uint32_t t, uint32_t flags) {
  if (SHT_LOUSER <= t && t <= SHT_HIUSER && !(flags & SHF_ALLOC))
    return true;
  if (SHT_LOOS <= t && t <= SHT_HIOS && !(flags & SHF_OS_NONCONFORMING))
    return true;
  // Allow all processor-specific types. This is different from GNU ld.
  return SHT_LOPROC <= t && t <= SHT_HIPROC;
}

template <class ELFT>
void ObjFile<ELFT>::initializeSections(bool ignoreComdats,
                                       const llvm::object::ELFFile<ELFT> &obj) {
  ArrayRef<Elf_Shdr> objSections = getELFShdrs<ELFT>();
  StringRef shstrtab = CHECK2(obj.getSectionStringTable(objSections), this);
  uint64_t size = objSections.size();
  SmallVector<ArrayRef<Elf_Word>, 0> selectedGroups;
  for (size_t i = 0; i != size; ++i) {
    if (this->sections[i] == &InputSection::discarded)
      continue;
    const Elf_Shdr &sec = objSections[i];
    const uint32_t type = sec.sh_type;

    // SHF_EXCLUDE'ed sections are discarded by the linker. However,
    // if -r is given, we'll let the final link discard such sections.
    // This is compatible with GNU.
    if ((sec.sh_flags & SHF_EXCLUDE) && !ctx.arg.relocatable) {
      if (type == SHT_LLVM_CALL_GRAPH_PROFILE)
        cgProfileSectionIndex = i;
      if (type == SHT_LLVM_ADDRSIG) {
        // We ignore the address-significance table if we know that the object
        // file was created by objcopy or ld -r. This is because these tools
        // will reorder the symbols in the symbol table, invalidating the data
        // in the address-significance table, which refers to symbols by index.
        if (sec.sh_link != 0)
          this->addrsigSec = &sec;
        else if (ctx.arg.icf == ICFLevel::Safe)
          Warn(ctx) << this
                    << ": --icf=safe conservatively ignores "
                       "SHT_LLVM_ADDRSIG [index "
                    << i
                    << "] with sh_link=0 "
                       "(likely created using objcopy or ld -r)";
      }
      this->sections[i] = &InputSection::discarded;
      continue;
    }

    switch (type) {
    case SHT_GROUP: {
      if (!ctx.arg.relocatable)
        sections[i] = &InputSection::discarded;
      StringRef signature =
          cantFail(this->getELFSyms<ELFT>()[sec.sh_info].getName(stringTable));
      ArrayRef<Elf_Word> entries =
          cantFail(obj.template getSectionContentsAsArray<Elf_Word>(sec));
      if ((entries[0] & GRP_COMDAT) == 0 || ignoreComdats ||
          ctx.symtab->comdatGroups.find(CachedHashStringRef(signature))
                  ->second == this)
        selectedGroups.push_back(entries);
      break;
    }
    case SHT_SYMTAB_SHNDX:
      shndxTable = CHECK2(obj.getSHNDXTable(sec, objSections), this);
      break;
    case SHT_SYMTAB:
    case SHT_STRTAB:
    case SHT_REL:
    case SHT_RELA:
    case SHT_CREL:
    case SHT_NULL:
      break;
    case SHT_PROGBITS:
    case SHT_NOTE:
    case SHT_NOBITS:
    case SHT_INIT_ARRAY:
    case SHT_FINI_ARRAY:
    case SHT_PREINIT_ARRAY:
      this->sections[i] =
          createInputSection(i, sec, check(obj.getSectionName(sec, shstrtab)));
      break;
    case SHT_LLVM_LTO:
      // Discard .llvm.lto in a relocatable link that does not use the bitcode.
      // The concatenated output does not properly reflect the linking
      // semantics. In addition, since we do not use the bitcode wrapper format,
      // the concatenated raw bitcode would be invalid.
      if (ctx.arg.relocatable && !ctx.arg.fatLTOObjects) {
        sections[i] = &InputSection::discarded;
        break;
      }
      [[fallthrough]];
    default:
      this->sections[i] =
          createInputSection(i, sec, check(obj.getSectionName(sec, shstrtab)));
      if (type == SHT_LLVM_SYMPART)
        ctx.hasSympart.store(true, std::memory_order_relaxed);
      else if (ctx.arg.rejectMismatch &&
               !isKnownSpecificSectionType(type, sec.sh_flags))
        Err(ctx) << this->sections[i] << ": unknown section type 0x"
                 << Twine::utohexstr(type);
      break;
    }
  }

  // We have a second loop. It is used to:
  // 1) handle SHF_LINK_ORDER sections.
  // 2) create relocation sections. In some cases the section header index of a
  //    relocation section may be smaller than that of the relocated section. In
  //    such cases, the relocation section would attempt to reference a target
  //    section that has not yet been created. For simplicity, delay creation of
  //    relocation sections until now.
  for (size_t i = 0; i != size; ++i) {
    if (this->sections[i] == &InputSection::discarded)
      continue;
    const Elf_Shdr &sec = objSections[i];

    if (isStaticRelSecType(sec.sh_type)) {
      // Find a relocation target section and associate this section with that.
      // Target may have been discarded if it is in a different section group
      // and the group is discarded, even though it's a violation of the spec.
      // We handle that situation gracefully by discarding dangling relocation
      // sections.
      const uint32_t info = sec.sh_info;
      InputSectionBase *s = getRelocTarget(i, info);
      if (!s)
        continue;

      // ELF spec allows mergeable sections with relocations, but they are rare,
      // and it is in practice hard to merge such sections by contents, because
      // applying relocations at end of linking changes section contents. So, we
      // simply handle such sections as non-mergeable ones. Degrading like this
      // is acceptable because section merging is optional.
      if (auto *ms = dyn_cast<MergeInputSection>(s)) {
        s = makeThreadLocal<InputSection>(ms->file, ms->name, ms->type,
                                          ms->flags, ms->addralign, ms->entsize,
                                          ms->contentMaybeDecompress());
        sections[info] = s;
      }

      if (s->relSecIdx != 0)
        ErrAlways(ctx) << s
                       << ": multiple relocation sections to one section are "
                          "not supported";
      s->relSecIdx = i;

      // Relocation sections are usually removed from the output, so return
      // `nullptr` for the normal case. However, if -r or --emit-relocs is
      // specified, we need to copy them to the output. (Some post link analysis
      // tools specify --emit-relocs to obtain the information.)
      if (ctx.arg.copyRelocs) {
        auto *isec = makeThreadLocal<InputSection>(
            *this, sec, check(obj.getSectionName(sec, shstrtab)));
        // If the relocated section is discarded (due to /DISCARD/ or
        // --gc-sections), the relocation section should be discarded as well.
        s->dependentSections.push_back(isec);
        sections[i] = isec;
      }
      continue;
    }

    // A SHF_LINK_ORDER section with sh_link=0 is handled as if it did not have
    // the flag.
    if (!sec.sh_link || !(sec.sh_flags & SHF_LINK_ORDER))
      continue;

    InputSectionBase *linkSec = nullptr;
    if (sec.sh_link < size)
      linkSec = this->sections[sec.sh_link];
    if (!linkSec) {
      ErrAlways(ctx) << this
                     << ": invalid sh_link index: " << uint32_t(sec.sh_link);
      continue;
    }

    // A SHF_LINK_ORDER section is discarded if its linked-to section is
    // discarded.
    InputSection *isec = cast<InputSection>(this->sections[i]);
    linkSec->dependentSections.push_back(isec);
    if (!isa<InputSection>(linkSec))
      ErrAlways(ctx)
          << "a section " << isec->name
          << " with SHF_LINK_ORDER should not refer a non-regular section: "
          << linkSec;
  }

  for (ArrayRef<Elf_Word> entries : selectedGroups)
    handleSectionGroup<ELFT>(this->sections, entries);
}

// Read the following info from the .note.gnu.property section and write it to
// the corresponding fields in `ObjFile`:
// - Feature flags (32 bits) representing x86 or AArch64 features for
//   hardware-assisted call flow control;
// - AArch64 PAuth ABI core info (16 bytes).
template <class ELFT>
static void readGnuProperty(Ctx &ctx, const InputSection &sec,
                            ObjFile<ELFT> &f) {
  using Elf_Nhdr = typename ELFT::Nhdr;
  using Elf_Note = typename ELFT::Note;

  ArrayRef<uint8_t> data = sec.content();
  auto err = [&](const uint8_t *place) -> ELFSyncStream {
    auto diag = Err(ctx);
    diag << sec.file << ":(" << sec.name << "+0x"
         << Twine::utohexstr(place - sec.content().data()) << "): ";
    return diag;
  };
  while (!data.empty()) {
    // Read one NOTE record.
    auto *nhdr = reinterpret_cast<const Elf_Nhdr *>(data.data());
    if (data.size() < sizeof(Elf_Nhdr) ||
        data.size() < nhdr->getSize(sec.addralign))
      return void(err(data.data()) << "data is too short");

    Elf_Note note(*nhdr);
    if (nhdr->n_type != NT_GNU_PROPERTY_TYPE_0 || note.getName() != "GNU") {
      data = data.slice(nhdr->getSize(sec.addralign));
      continue;
    }

    uint32_t featureAndType = ctx.arg.emachine == EM_AARCH64
                                  ? GNU_PROPERTY_AARCH64_FEATURE_1_AND
                                  : GNU_PROPERTY_X86_FEATURE_1_AND;

    // Read a body of a NOTE record, which consists of type-length-value fields.
    ArrayRef<uint8_t> desc = note.getDesc(sec.addralign);
    while (!desc.empty()) {
      const uint8_t *place = desc.data();
      if (desc.size() < 8)
        return void(err(place) << "program property is too short");
      uint32_t type = read32<ELFT::Endianness>(desc.data());
      uint32_t size = read32<ELFT::Endianness>(desc.data() + 4);
      desc = desc.slice(8);
      if (desc.size() < size)
        return void(err(place) << "program property is too short");

      if (type == featureAndType) {
        // We found a FEATURE_1_AND field. There may be more than one of these
        // in a .note.gnu.property section, for a relocatable object we
        // accumulate the bits set.
        if (size < 4)
          return void(err(place) << "FEATURE_1_AND entry is too short");
        f.andFeatures |= read32<ELFT::Endianness>(desc.data());
      } else if (ctx.arg.emachine == EM_AARCH64 &&
                 type == GNU_PROPERTY_AARCH64_FEATURE_PAUTH) {
        if (!f.aarch64PauthAbiCoreInfo.empty()) {
          return void(
              err(data.data())
              << "multiple GNU_PROPERTY_AARCH64_FEATURE_PAUTH entries are "
                 "not supported");
        } else if (size != 16) {
          return void(err(data.data())
                      << "GNU_PROPERTY_AARCH64_FEATURE_PAUTH entry "
                         "is invalid: expected 16 bytes, but got "
                      << size);
        }
        f.aarch64PauthAbiCoreInfo = desc;
      }

      // Padding is present in the note descriptor, if necessary.
      desc = desc.slice(alignTo<(ELFT::Is64Bits ? 8 : 4)>(size));
    }

    // Go to next NOTE record to look for more FEATURE_1_AND descriptions.
    data = data.slice(nhdr->getSize(sec.addralign));
  }
}

template <class ELFT>
InputSectionBase *ObjFile<ELFT>::getRelocTarget(uint32_t idx, uint32_t info) {
  if (info < this->sections.size()) {
    InputSectionBase *target = this->sections[info];

    // Strictly speaking, a relocation section must be included in the
    // group of the section it relocates. However, LLVM 3.3 and earlier
    // would fail to do so, so we gracefully handle that case.
    if (target == &InputSection::discarded)
      return nullptr;

    if (target != nullptr)
      return target;
  }

  Err(ctx) << this << ": relocation section (index " << idx
           << ") has invalid sh_info (" << info << ')';
  return nullptr;
}

// The function may be called concurrently for different input files. For
// allocation, prefer makeThreadLocal which does not require holding a lock.
template <class ELFT>
InputSectionBase *ObjFile<ELFT>::createInputSection(uint32_t idx,
                                                    const Elf_Shdr &sec,
                                                    StringRef name) {
  if (name.starts_with(".n")) {
    // The GNU linker uses .note.GNU-stack section as a marker indicating
    // that the code in the object file does not expect that the stack is
    // executable (in terms of NX bit). If all input files have the marker,
    // the GNU linker adds a PT_GNU_STACK segment to tells the loader to
    // make the stack non-executable. Most object files have this section as
    // of 2017.
    //
    // But making the stack non-executable is a norm today for security
    // reasons. Failure to do so may result in a serious security issue.
    // Therefore, we make LLD always add PT_GNU_STACK unless it is
    // explicitly told to do otherwise (by -z execstack). Because the stack
    // executable-ness is controlled solely by command line options,
    // .note.GNU-stack sections are, with one exception, ignored. Report
    // an error if we encounter an executable .note.GNU-stack to force the
    // user to explicitly request an executable stack.
    if (name == ".note.GNU-stack") {
      if ((sec.sh_flags & SHF_EXECINSTR) && !ctx.arg.relocatable &&
          ctx.arg.zGnustack != GnuStackKind::Exec) {
        Err(ctx) << this
                 << ": requires an executable stack, but -z execstack is not "
                    "specified";
      }
      return &InputSection::discarded;
    }

    // Object files that use processor features such as Intel Control-Flow
    // Enforcement (CET) or AArch64 Branch Target Identification BTI, use a
    // .note.gnu.property section containing a bitfield of feature bits like the
    // GNU_PROPERTY_X86_FEATURE_1_IBT flag. Read a bitmap containing the flag.
    //
    // Since we merge bitmaps from multiple object files to create a new
    // .note.gnu.property containing a single AND'ed bitmap, we discard an input
    // file's .note.gnu.property section.
    if (name == ".note.gnu.property") {
      readGnuProperty<ELFT>(ctx, InputSection(*this, sec, name), *this);
      return &InputSection::discarded;
    }

    // Split stacks is a feature to support a discontiguous stack,
    // commonly used in the programming language Go. For the details,
    // see https://gcc.gnu.org/wiki/SplitStacks. An object file compiled
    // for split stack will include a .note.GNU-split-stack section.
    if (name == ".note.GNU-split-stack") {
      if (ctx.arg.relocatable) {
        ErrAlways(ctx) << "cannot mix split-stack and non-split-stack in a "
                          "relocatable link";
        return &InputSection::discarded;
      }
      this->splitStack = true;
      return &InputSection::discarded;
    }

    // An object file compiled for split stack, but where some of the
    // functions were compiled with the no_split_stack_attribute will
    // include a .note.GNU-no-split-stack section.
    if (name == ".note.GNU-no-split-stack") {
      this->someNoSplitStack = true;
      return &InputSection::discarded;
    }

    // Strip existing .note.gnu.build-id sections so that the output won't have
    // more than one build-id. This is not usually a problem because input
    // object files normally don't have .build-id sections, but you can create
    // such files by "ld.{bfd,gold,lld} -r --build-id", and we want to guard
    // against it.
    if (name == ".note.gnu.build-id")
      return &InputSection::discarded;
  }

  // The linker merges EH (exception handling) frames and creates a
  // .eh_frame_hdr section for runtime. So we handle them with a special
  // class. For relocatable outputs, they are just passed through.
  if (name == ".eh_frame" && !ctx.arg.relocatable)
    return makeThreadLocal<EhInputSection>(*this, sec, name);

  if ((sec.sh_flags & SHF_MERGE) && shouldMerge(sec, name))
    return makeThreadLocal<MergeInputSection>(*this, sec, name);
  return makeThreadLocal<InputSection>(*this, sec, name);
}

// Initialize symbols. symbols is a parallel array to the corresponding ELF
// symbol table.
template <class ELFT>
void ObjFile<ELFT>::initializeSymbols(const object::ELFFile<ELFT> &obj) {
  ArrayRef<Elf_Sym> eSyms = this->getELFSyms<ELFT>();
  if (!symbols)
    symbols = std::make_unique<Symbol *[]>(numSymbols);

  // Some entries have been filled by LazyObjFile.
  auto *symtab = ctx.symtab.get();
  for (size_t i = firstGlobal, end = eSyms.size(); i != end; ++i)
    if (!symbols[i])
      symbols[i] = symtab->insert(CHECK2(eSyms[i].getName(stringTable), this));

  // Perform symbol resolution on non-local symbols.
  SmallVector<unsigned, 32> undefineds;
  for (size_t i = firstGlobal, end = eSyms.size(); i != end; ++i) {
    const Elf_Sym &eSym = eSyms[i];
    uint32_t secIdx = eSym.st_shndx;
    if (secIdx == SHN_UNDEF) {
      undefineds.push_back(i);
      continue;
    }

    uint8_t binding = eSym.getBinding();
    uint8_t stOther = eSym.st_other;
    uint8_t type = eSym.getType();
    uint64_t value = eSym.st_value;
    uint64_t size = eSym.st_size;

    Symbol *sym = symbols[i];
    sym->isUsedInRegularObj = true;
    if (LLVM_UNLIKELY(eSym.st_shndx == SHN_COMMON)) {
      if (value == 0 || value >= UINT32_MAX)
        Err(ctx) << this << ": common symbol '" << sym->getName()
                 << "' has invalid alignment: " << value;
      hasCommonSyms = true;
      sym->resolve(ctx, CommonSymbol{ctx, this, StringRef(), binding, stOther,
                                     type, value, size});
      continue;
    }

    // Handle global defined symbols. Defined::section will be set in postParse.
    sym->resolve(ctx, Defined{ctx, this, StringRef(), binding, stOther, type,
                              value, size, nullptr});
  }

  // Undefined symbols (excluding those defined relative to non-prevailing
  // sections) can trigger recursive extract. Process defined symbols first so
  // that the relative order between a defined symbol and an undefined symbol
  // does not change the symbol resolution behavior. In addition, a set of
  // interconnected symbols will all be resolved to the same file, instead of
  // being resolved to different files.
  for (unsigned i : undefineds) {
    const Elf_Sym &eSym = eSyms[i];
    Symbol *sym = symbols[i];
    sym->resolve(ctx, Undefined{this, StringRef(), eSym.getBinding(),
                                eSym.st_other, eSym.getType()});
    sym->isUsedInRegularObj = true;
    sym->referenced = true;
  }
}

template <class ELFT>
void ObjFile<ELFT>::initSectionsAndLocalSyms(bool ignoreComdats) {
  if (!justSymbols)
    initializeSections(ignoreComdats, getObj());

  if (!firstGlobal)
    return;
  SymbolUnion *locals = makeThreadLocalN<SymbolUnion>(firstGlobal);
  memset(locals, 0, sizeof(SymbolUnion) * firstGlobal);

  ArrayRef<Elf_Sym> eSyms = this->getELFSyms<ELFT>();
  for (size_t i = 0, end = firstGlobal; i != end; ++i) {
    const Elf_Sym &eSym = eSyms[i];
    uint32_t secIdx = eSym.st_shndx;
    if (LLVM_UNLIKELY(secIdx == SHN_XINDEX))
      secIdx = check(getExtendedSymbolTableIndex<ELFT>(eSym, i, shndxTable));
    else if (secIdx >= SHN_LORESERVE)
      secIdx = 0;
    if (LLVM_UNLIKELY(secIdx >= sections.size())) {
      Err(ctx) << this << ": invalid section index: " << secIdx;
      secIdx = 0;
    }
    if (LLVM_UNLIKELY(eSym.getBinding() != STB_LOCAL))
      ErrAlways(ctx) << this << ": non-local symbol (" << i
                     << ") found at index < .symtab's sh_info (" << end << ")";

    InputSectionBase *sec = sections[secIdx];
    uint8_t type = eSym.getType();
    if (type == STT_FILE)
      sourceFile = CHECK2(eSym.getName(stringTable), this);
    unsigned stName = eSym.st_name;
    if (LLVM_UNLIKELY(stringTable.size() <= stName)) {
      Err(ctx) << this << ": invalid symbol name offset";
      stName = 0;
    }
    StringRef name(stringTable.data() + stName);

    symbols[i] = reinterpret_cast<Symbol *>(locals + i);
    if (eSym.st_shndx == SHN_UNDEF || sec == &InputSection::discarded)
      new (symbols[i]) Undefined(this, name, STB_LOCAL, eSym.st_other, type,
                                 /*discardedSecIdx=*/secIdx);
    else
      new (symbols[i]) Defined(ctx, this, name, STB_LOCAL, eSym.st_other, type,
                               eSym.st_value, eSym.st_size, sec);
    symbols[i]->partition = 1;
    symbols[i]->isUsedInRegularObj = true;
  }
}

// Called after all ObjFile::parse is called for all ObjFiles. This checks
// duplicate symbols and may do symbol property merge in the future.
template <class ELFT> void ObjFile<ELFT>::postParse() {
  static std::mutex mu;
  ArrayRef<Elf_Sym> eSyms = this->getELFSyms<ELFT>();
  for (size_t i = firstGlobal, end = eSyms.size(); i != end; ++i) {
    const Elf_Sym &eSym = eSyms[i];
    Symbol &sym = *symbols[i];
    uint32_t secIdx = eSym.st_shndx;
    uint8_t binding = eSym.getBinding();
    if (LLVM_UNLIKELY(binding != STB_GLOBAL && binding != STB_WEAK &&
                      binding != STB_GNU_UNIQUE))
      Err(ctx) << this << ": symbol (" << i
               << ") has invalid binding: " << (int)binding;

    // st_value of STT_TLS represents the assigned offset, not the actual
    // address which is used by STT_FUNC and STT_OBJECT. STT_TLS symbols can
    // only be referenced by special TLS relocations. It is usually an error if
    // a STT_TLS symbol is replaced by a non-STT_TLS symbol, vice versa.
    if (LLVM_UNLIKELY(sym.isTls()) && eSym.getType() != STT_TLS &&
        eSym.getType() != STT_NOTYPE)
      Err(ctx) << "TLS attribute mismatch: " << &sym << "\n>>> in " << sym.file
               << "\n>>> in " << this;

    // Handle non-COMMON defined symbol below. !sym.file allows a symbol
    // assignment to redefine a symbol without an error.
    if (!sym.isDefined() || secIdx == SHN_UNDEF)
      continue;
    if (LLVM_UNLIKELY(secIdx >= SHN_LORESERVE)) {
      if (secIdx == SHN_COMMON)
        continue;
      if (secIdx == SHN_XINDEX)
        secIdx = check(getExtendedSymbolTableIndex<ELFT>(eSym, i, shndxTable));
      else
        secIdx = 0;
    }

    if (LLVM_UNLIKELY(secIdx >= sections.size())) {
      Err(ctx) << this << ": invalid section index: " << secIdx;
      continue;
    }
    InputSectionBase *sec = sections[secIdx];
    if (sec == &InputSection::discarded) {
      if (sym.traced) {
        printTraceSymbol(Undefined{this, sym.getName(), sym.binding,
                                   sym.stOther, sym.type, secIdx},
                         sym.getName());
      }
      if (sym.file == this) {
        std::lock_guard<std::mutex> lock(mu);
        ctx.nonPrevailingSyms.emplace_back(&sym, secIdx);
      }
      continue;
    }

    if (sym.file == this) {
      cast<Defined>(sym).section = sec;
      continue;
    }

    if (sym.binding == STB_WEAK || binding == STB_WEAK)
      continue;
    std::lock_guard<std::mutex> lock(mu);
    ctx.duplicates.push_back({&sym, this, sec, eSym.st_value});
  }
}

// The handling of tentative definitions (COMMON symbols) in archives is murky.
// A tentative definition will be promoted to a global definition if there are
// no non-tentative definitions to dominate it. When we hold a tentative
// definition to a symbol and are inspecting archive members for inclusion
// there are 2 ways we can proceed:
//
// 1) Consider the tentative definition a 'real' definition (ie promotion from
//    tentative to real definition has already happened) and not inspect
//    archive members for Global/Weak definitions to replace the tentative
//    definition. An archive member would only be included if it satisfies some
//    other undefined symbol. This is the behavior Gold uses.
//
// 2) Consider the tentative definition as still undefined (ie the promotion to
//    a real definition happens only after all symbol resolution is done).
//    The linker searches archive members for STB_GLOBAL definitions to
//    replace the tentative definition with. This is the behavior used by
//    GNU ld.
//
//  The second behavior is inherited from SysVR4, which based it on the FORTRAN
//  COMMON BLOCK model. This behavior is needed for proper initialization in old
//  (pre F90) FORTRAN code that is packaged into an archive.
//
//  The following functions search archive members for definitions to replace
//  tentative definitions (implementing behavior 2).
static bool isBitcodeNonCommonDef(MemoryBufferRef mb, StringRef symName,
                                  StringRef archiveName) {
  IRSymtabFile symtabFile = check(readIRSymtab(mb));
  for (const irsymtab::Reader::SymbolRef &sym :
       symtabFile.TheReader.symbols()) {
    if (sym.isGlobal() && sym.getName() == symName)
      return !sym.isUndefined() && !sym.isWeak() && !sym.isCommon();
  }
  return false;
}

template <class ELFT>
static bool isNonCommonDef(Ctx &ctx, ELFKind ekind, MemoryBufferRef mb,
                           StringRef symName, StringRef archiveName) {
  ObjFile<ELFT> *obj = make<ObjFile<ELFT>>(ctx, ekind, mb, archiveName);
  obj->init();
  StringRef stringtable = obj->getStringTable();

  for (auto sym : obj->template getGlobalELFSyms<ELFT>()) {
    Expected<StringRef> name = sym.getName(stringtable);
    if (name && name.get() == symName)
      return sym.isDefined() && sym.getBinding() == STB_GLOBAL &&
             !sym.isCommon();
  }
  return false;
}

static bool isNonCommonDef(Ctx &ctx, MemoryBufferRef mb, StringRef symName,
                           StringRef archiveName) {
  switch (getELFKind(ctx, mb, archiveName)) {
  case ELF32LEKind:
    return isNonCommonDef<ELF32LE>(ctx, ELF32LEKind, mb, symName, archiveName);
  case ELF32BEKind:
    return isNonCommonDef<ELF32BE>(ctx, ELF32BEKind, mb, symName, archiveName);
  case ELF64LEKind:
    return isNonCommonDef<ELF64LE>(ctx, ELF64LEKind, mb, symName, archiveName);
  case ELF64BEKind:
    return isNonCommonDef<ELF64BE>(ctx, ELF64BEKind, mb, symName, archiveName);
  default:
    llvm_unreachable("getELFKind");
  }
}

SharedFile::SharedFile(Ctx &ctx, MemoryBufferRef m, StringRef defaultSoName)
    : ELFFileBase(ctx, SharedKind, getELFKind(ctx, m, ""), m),
      soName(defaultSoName), isNeeded(!ctx.arg.asNeeded) {}

// Parse the version definitions in the object file if present, and return a
// vector whose nth element contains a pointer to the Elf_Verdef for version
// identifier n. Version identifiers that are not definitions map to nullptr.
template <typename ELFT>
static SmallVector<const void *, 0>
parseVerdefs(const uint8_t *base, const typename ELFT::Shdr *sec) {
  if (!sec)
    return {};

  // Build the Verdefs array by following the chain of Elf_Verdef objects
  // from the start of the .gnu.version_d section.
  SmallVector<const void *, 0> verdefs;
  const uint8_t *verdef = base + sec->sh_offset;
  for (unsigned i = 0, e = sec->sh_info; i != e; ++i) {
    auto *curVerdef = reinterpret_cast<const typename ELFT::Verdef *>(verdef);
    verdef += curVerdef->vd_next;
    unsigned verdefIndex = curVerdef->vd_ndx;
    if (verdefIndex >= verdefs.size())
      verdefs.resize(verdefIndex + 1);
    verdefs[verdefIndex] = curVerdef;
  }
  return verdefs;
}

// Parse SHT_GNU_verneed to properly set the name of a versioned undefined
// symbol. We detect fatal issues which would cause vulnerabilities, but do not
// implement sophisticated error checking like in llvm-readobj because the value
// of such diagnostics is low.
template <typename ELFT>
std::vector<uint32_t> SharedFile::parseVerneed(const ELFFile<ELFT> &obj,
                                               const typename ELFT::Shdr *sec) {
  if (!sec)
    return {};
  std::vector<uint32_t> verneeds;
  ArrayRef<uint8_t> data = CHECK2(obj.getSectionContents(*sec), this);
  const uint8_t *verneedBuf = data.begin();
  for (unsigned i = 0; i != sec->sh_info; ++i) {
    if (verneedBuf + sizeof(typename ELFT::Verneed) > data.end()) {
      Err(ctx) << this << " has an invalid Verneed";
      break;
    }
    auto *vn = reinterpret_cast<const typename ELFT::Verneed *>(verneedBuf);
    const uint8_t *vernauxBuf = verneedBuf + vn->vn_aux;
    for (unsigned j = 0; j != vn->vn_cnt; ++j) {
      if (vernauxBuf + sizeof(typename ELFT::Vernaux) > data.end()) {
        Err(ctx) << this << " has an invalid Vernaux";
        break;
      }
      auto *aux = reinterpret_cast<const typename ELFT::Vernaux *>(vernauxBuf);
      if (aux->vna_name >= this->stringTable.size()) {
        Err(ctx) << this << " has a Vernaux with an invalid vna_name";
        break;
      }
      uint16_t version = aux->vna_other & VERSYM_VERSION;
      if (version >= verneeds.size())
        verneeds.resize(version + 1);
      verneeds[version] = aux->vna_name;
      vernauxBuf += aux->vna_next;
    }
    verneedBuf += vn->vn_next;
  }
  return verneeds;
}

// We do not usually care about alignments of data in shared object
// files because the loader takes care of it. However, if we promote a
// DSO symbol to point to .bss due to copy relocation, we need to keep
// the original alignment requirements. We infer it in this function.
template <typename ELFT>
static uint64_t getAlignment(ArrayRef<typename ELFT::Shdr> sections,
                             const typename ELFT::Sym &sym) {
  uint64_t ret = UINT64_MAX;
  if (sym.st_value)
    ret = 1ULL << llvm::countr_zero((uint64_t)sym.st_value);
  if (0 < sym.st_shndx && sym.st_shndx < sections.size())
    ret = std::min<uint64_t>(ret, sections[sym.st_shndx].sh_addralign);
  return (ret > UINT32_MAX) ? 0 : ret;
}

// Fully parse the shared object file.
//
// This function parses symbol versions. If a DSO has version information,
// the file has a ".gnu.version_d" section which contains symbol version
// definitions. Each symbol is associated to one version through a table in
// ".gnu.version" section. That table is a parallel array for the symbol
// table, and each table entry contains an index in ".gnu.version_d".
//
// The special index 0 is reserved for VERF_NDX_LOCAL and 1 is for
// VER_NDX_GLOBAL. There's no table entry for these special versions in
// ".gnu.version_d".
//
// The file format for symbol versioning is perhaps a bit more complicated
// than necessary, but you can easily understand the code if you wrap your
// head around the data structure described above.
template <class ELFT> void SharedFile::parse() {
  using Elf_Dyn = typename ELFT::Dyn;
  using Elf_Shdr = typename ELFT::Shdr;
  using Elf_Sym = typename ELFT::Sym;
  using Elf_Verdef = typename ELFT::Verdef;
  using Elf_Versym = typename ELFT::Versym;

  ArrayRef<Elf_Dyn> dynamicTags;
  const ELFFile<ELFT> obj = this->getObj<ELFT>();
  ArrayRef<Elf_Shdr> sections = getELFShdrs<ELFT>();

  const Elf_Shdr *versymSec = nullptr;
  const Elf_Shdr *verdefSec = nullptr;
  const Elf_Shdr *verneedSec = nullptr;
  symbols = std::make_unique<Symbol *[]>(numSymbols);

  // Search for .dynsym, .dynamic, .symtab, .gnu.version and .gnu.version_d.
  for (const Elf_Shdr &sec : sections) {
    switch (sec.sh_type) {
    default:
      continue;
    case SHT_DYNAMIC:
      dynamicTags =
          CHECK2(obj.template getSectionContentsAsArray<Elf_Dyn>(sec), this);
      break;
    case SHT_GNU_versym:
      versymSec = &sec;
      break;
    case SHT_GNU_verdef:
      verdefSec = &sec;
      break;
    case SHT_GNU_verneed:
      verneedSec = &sec;
      break;
    }
  }

  if (versymSec && numSymbols == 0) {
    ErrAlways(ctx) << "SHT_GNU_versym should be associated with symbol table";
    return;
  }

  // Search for a DT_SONAME tag to initialize this->soName.
  for (const Elf_Dyn &dyn : dynamicTags) {
    if (dyn.d_tag == DT_NEEDED) {
      uint64_t val = dyn.getVal();
      if (val >= this->stringTable.size()) {
        Err(ctx) << this << ": invalid DT_NEEDED entry";
        return;
      }
      dtNeeded.push_back(this->stringTable.data() + val);
    } else if (dyn.d_tag == DT_SONAME) {
      uint64_t val = dyn.getVal();
      if (val >= this->stringTable.size()) {
        Err(ctx) << this << ": invalid DT_SONAME entry";
        return;
      }
      soName = this->stringTable.data() + val;
    }
  }

  // DSOs are uniquified not by filename but by soname.
  StringSaver &ss = ctx.saver;
  DenseMap<CachedHashStringRef, SharedFile *>::iterator it;
  bool wasInserted;
  std::tie(it, wasInserted) =
      ctx.symtab->soNames.try_emplace(CachedHashStringRef(soName), this);

  // If a DSO appears more than once on the command line with and without
  // --as-needed, --no-as-needed takes precedence over --as-needed because a
  // user can add an extra DSO with --no-as-needed to force it to be added to
  // the dependency list.
  it->second->isNeeded |= isNeeded;
  if (!wasInserted)
    return;

  ctx.sharedFiles.push_back(this);

  verdefs = parseVerdefs<ELFT>(obj.base(), verdefSec);
  std::vector<uint32_t> verneeds = parseVerneed<ELFT>(obj, verneedSec);

  // Parse ".gnu.version" section which is a parallel array for the symbol
  // table. If a given file doesn't have a ".gnu.version" section, we use
  // VER_NDX_GLOBAL.
  size_t size = numSymbols - firstGlobal;
  std::vector<uint16_t> versyms(size, VER_NDX_GLOBAL);
  if (versymSec) {
    ArrayRef<Elf_Versym> versym =
        CHECK2(obj.template getSectionContentsAsArray<Elf_Versym>(*versymSec),
               this)
            .slice(firstGlobal);
    for (size_t i = 0; i < size; ++i)
      versyms[i] = versym[i].vs_index;
  }

  // System libraries can have a lot of symbols with versions. Using a
  // fixed buffer for computing the versions name (foo@ver) can save a
  // lot of allocations.
  SmallString<0> versionedNameBuffer;

  // Add symbols to the symbol table.
  ArrayRef<Elf_Sym> syms = this->getGlobalELFSyms<ELFT>();
  for (size_t i = 0, e = syms.size(); i != e; ++i) {
    const Elf_Sym &sym = syms[i];

    // ELF spec requires that all local symbols precede weak or global
    // symbols in each symbol table, and the index of first non-local symbol
    // is stored to sh_info. If a local symbol appears after some non-local
    // symbol, that's a violation of the spec.
    StringRef name = CHECK2(sym.getName(stringTable), this);
    if (sym.getBinding() == STB_LOCAL) {
      Err(ctx) << this << ": invalid local symbol '" << name
               << "' in global part of symbol table";
      continue;
    }

    const uint16_t ver = versyms[i], idx = ver & ~VERSYM_HIDDEN;
    if (sym.isUndefined()) {
      // For unversioned undefined symbols, VER_NDX_GLOBAL makes more sense but
      // as of binutils 2.34, GNU ld produces VER_NDX_LOCAL.
      if (ver != VER_NDX_LOCAL && ver != VER_NDX_GLOBAL) {
        if (idx >= verneeds.size()) {
          ErrAlways(ctx) << "corrupt input file: version need index " << idx
                         << " for symbol " << name
                         << " is out of bounds\n>>> defined in " << this;
          continue;
        }
        StringRef verName = stringTable.data() + verneeds[idx];
        versionedNameBuffer.clear();
        name = ss.save((name + "@" + verName).toStringRef(versionedNameBuffer));
      }
      Symbol *s = ctx.symtab->addSymbol(
          Undefined{this, name, sym.getBinding(), sym.st_other, sym.getType()});
      s->isExported = true;
      if (sym.getBinding() != STB_WEAK &&
          ctx.arg.unresolvedSymbolsInShlib != UnresolvedPolicy::Ignore)
        requiredSymbols.push_back(s);
      continue;
    }

    if (ver == VER_NDX_LOCAL ||
        (ver != VER_NDX_GLOBAL && idx >= verdefs.size())) {
      // In GNU ld < 2.31 (before 3be08ea4728b56d35e136af4e6fd3086ade17764), the
      // MIPS port puts _gp_disp symbol into DSO files and incorrectly assigns
      // VER_NDX_LOCAL. Workaround this bug.
      if (ctx.arg.emachine == EM_MIPS && name == "_gp_disp")
        continue;
      ErrAlways(ctx) << "corrupt input file: version definition index " << idx
                     << " for symbol " << name
                     << " is out of bounds\n>>> defined in " << this;
      continue;
    }

    uint32_t alignment = getAlignment<ELFT>(sections, sym);
    if (ver == idx) {
      auto *s = ctx.symtab->addSymbol(
          SharedSymbol{*this, name, sym.getBinding(), sym.st_other,
                       sym.getType(), sym.st_value, sym.st_size, alignment});
      s->dsoDefined = true;
      if (s->file == this)
        s->versionId = ver;
    }

    // Also add the symbol with the versioned name to handle undefined symbols
    // with explicit versions.
    if (ver == VER_NDX_GLOBAL)
      continue;

    StringRef verName =
        stringTable.data() +
        reinterpret_cast<const Elf_Verdef *>(verdefs[idx])->getAux()->vda_name;
    versionedNameBuffer.clear();
    name = (name + "@" + verName).toStringRef(versionedNameBuffer);
    auto *s = ctx.symtab->addSymbol(
        SharedSymbol{*this, ss.save(name), sym.getBinding(), sym.st_other,
                     sym.getType(), sym.st_value, sym.st_size, alignment});
    s->dsoDefined = true;
    if (s->file == this)
      s->versionId = idx;
  }
}

static ELFKind getBitcodeELFKind(const Triple &t) {
  if (t.isLittleEndian())
    return t.isArch64Bit() ? ELF64LEKind : ELF32LEKind;
  return t.isArch64Bit() ? ELF64BEKind : ELF32BEKind;
}

static uint16_t getBitcodeMachineKind(Ctx &ctx, StringRef path,
                                      const Triple &t) {
  switch (t.getArch()) {
  case Triple::aarch64:
  case Triple::aarch64_be:
    return EM_AARCH64;
  case Triple::amdgcn:
  case Triple::r600:
    return EM_AMDGPU;
  case Triple::arm:
  case Triple::armeb:
  case Triple::thumb:
  case Triple::thumbeb:
    return EM_ARM;
  case Triple::avr:
    return EM_AVR;
  case Triple::hexagon:
    return EM_HEXAGON;
  case Triple::loongarch32:
  case Triple::loongarch64:
    return EM_LOONGARCH;
  case Triple::mips:
  case Triple::mipsel:
  case Triple::mips64:
  case Triple::mips64el:
    return EM_MIPS;
  case Triple::msp430:
    return EM_MSP430;
  case Triple::ppc:
  case Triple::ppcle:
    return EM_PPC;
  case Triple::ppc64:
  case Triple::ppc64le:
    return EM_PPC64;
  case Triple::riscv32:
  case Triple::riscv64:
    return EM_RISCV;
  case Triple::sparcv9:
    return EM_SPARCV9;
  case Triple::systemz:
    return EM_S390;
  case Triple::x86:
    return t.isOSIAMCU() ? EM_IAMCU : EM_386;
  case Triple::x86_64:
    return EM_X86_64;
  default:
    ErrAlways(ctx) << path
                   << ": could not infer e_machine from bitcode target triple "
                   << t.str();
    return EM_NONE;
  }
}

static uint8_t getOsAbi(const Triple &t) {
  switch (t.getOS()) {
  case Triple::AMDHSA:
    return ELF::ELFOSABI_AMDGPU_HSA;
  case Triple::AMDPAL:
    return ELF::ELFOSABI_AMDGPU_PAL;
  case Triple::Mesa3D:
    return ELF::ELFOSABI_AMDGPU_MESA3D;
  default:
    return ELF::ELFOSABI_NONE;
  }
}

BitcodeFile::BitcodeFile(Ctx &ctx, MemoryBufferRef mb, StringRef archiveName,
                         uint64_t offsetInArchive, bool lazy)
    : InputFile(ctx, BitcodeKind, mb) {
  this->archiveName = archiveName;
  this->lazy = lazy;

  std::string path = mb.getBufferIdentifier().str();
  if (ctx.arg.thinLTOIndexOnly)
    path = replaceThinLTOSuffix(ctx, mb.getBufferIdentifier());

  // ThinLTO assumes that all MemoryBufferRefs given to it have a unique
  // name. If two archives define two members with the same name, this
  // causes a collision which result in only one of the objects being taken
  // into consideration at LTO time (which very likely causes undefined
  // symbols later in the link stage). So we append file offset to make
  // filename unique.
  StringSaver &ss = ctx.saver;
  StringRef name = archiveName.empty()
                       ? ss.save(path)
                       : ss.save(archiveName + "(" + path::filename(path) +
                                 " at " + utostr(offsetInArchive) + ")");
  MemoryBufferRef mbref(mb.getBuffer(), name);

  obj = CHECK2(lto::InputFile::create(mbref), this);

  Triple t(obj->getTargetTriple());
  ekind = getBitcodeELFKind(t);
  emachine = getBitcodeMachineKind(ctx, mb.getBufferIdentifier(), t);
  osabi = getOsAbi(t);
}

static uint8_t mapVisibility(GlobalValue::VisibilityTypes gvVisibility) {
  switch (gvVisibility) {
  case GlobalValue::DefaultVisibility:
    return STV_DEFAULT;
  case GlobalValue::HiddenVisibility:
    return STV_HIDDEN;
  case GlobalValue::ProtectedVisibility:
    return STV_PROTECTED;
  }
  llvm_unreachable("unknown visibility");
}

static void createBitcodeSymbol(Ctx &ctx, Symbol *&sym,
                                const lto::InputFile::Symbol &objSym,
                                BitcodeFile &f) {
  uint8_t binding = objSym.isWeak() ? STB_WEAK : STB_GLOBAL;
  uint8_t type = objSym.isTLS() ? STT_TLS : STT_NOTYPE;
  uint8_t visibility = mapVisibility(objSym.getVisibility());

  if (!sym) {
    // Symbols can be duplicated in bitcode files because of '#include' and
    // linkonce_odr. Use uniqueSaver to save symbol names for de-duplication.
    // Update objSym.Name to reference (via StringRef) the string saver's copy;
    // this way LTO can reference the same string saver's copy rather than
    // keeping copies of its own.
    objSym.Name = ctx.uniqueSaver.save(objSym.getName());
    sym = ctx.symtab->insert(objSym.getName());
  }

  if (objSym.isUndefined()) {
    Undefined newSym(&f, StringRef(), binding, visibility, type);
    sym->resolve(ctx, newSym);
    sym->referenced = true;
    return;
  }

  if (objSym.isCommon()) {
    sym->resolve(ctx, CommonSymbol{ctx, &f, StringRef(), binding, visibility,
                                   STT_OBJECT, objSym.getCommonAlignment(),
                                   objSym.getCommonSize()});
  } else {
    Defined newSym(ctx, &f, StringRef(), binding, visibility, type, 0, 0,
                   nullptr);
    // The definition can be omitted if all bitcode definitions satisfy
    // `canBeOmittedFromSymbolTable()` and isUsedInRegularObj is false.
    // The latter condition is tested in parseVersionAndComputeIsPreemptible.
    sym->ltoCanOmit = objSym.canBeOmittedFromSymbolTable() &&
                      (!sym->isDefined() || sym->ltoCanOmit);
    sym->resolve(ctx, newSym);
  }
}

void BitcodeFile::parse() {
  for (std::pair<StringRef, Comdat::SelectionKind> s : obj->getComdatTable()) {
    keptComdats.push_back(
        s.second == Comdat::NoDeduplicate ||
        ctx.symtab->comdatGroups.try_emplace(CachedHashStringRef(s.first), this)
            .second);
  }

  if (numSymbols == 0) {
    numSymbols = obj->symbols().size();
    symbols = std::make_unique<Symbol *[]>(numSymbols);
  }
  // Process defined symbols first. See the comment in
  // ObjFile<ELFT>::initializeSymbols.
  for (auto [i, irSym] : llvm::enumerate(obj->symbols()))
    if (!irSym.isUndefined())
      createBitcodeSymbol(ctx, symbols[i], irSym, *this);
  for (auto [i, irSym] : llvm::enumerate(obj->symbols()))
    if (irSym.isUndefined())
      createBitcodeSymbol(ctx, symbols[i], irSym, *this);

  for (auto l : obj->getDependentLibraries())
    addDependentLibrary(ctx, l, this);
}

void BitcodeFile::parseLazy() {
  numSymbols = obj->symbols().size();
  symbols = std::make_unique<Symbol *[]>(numSymbols);
  for (auto [i, irSym] : llvm::enumerate(obj->symbols())) {
    // Symbols can be duplicated in bitcode files because of '#include' and
    // linkonce_odr. Use uniqueSaver to save symbol names for de-duplication.
    // Update objSym.Name to reference (via StringRef) the string saver's copy;
    // this way LTO can reference the same string saver's copy rather than
    // keeping copies of its own.
    irSym.Name = ctx.uniqueSaver.save(irSym.getName());
    if (!irSym.isUndefined()) {
      auto *sym = ctx.symtab->insert(irSym.getName());
      sym->resolve(ctx, LazySymbol{*this});
      symbols[i] = sym;
    }
  }
}

void BitcodeFile::postParse() {
  for (auto [i, irSym] : llvm::enumerate(obj->symbols())) {
    const Symbol &sym = *symbols[i];
    if (sym.file == this || !sym.isDefined() || irSym.isUndefined() ||
        irSym.isCommon() || irSym.isWeak())
      continue;
    int c = irSym.getComdatIndex();
    if (c != -1 && !keptComdats[c])
      continue;
    reportDuplicate(ctx, sym, this, nullptr, 0);
  }
}

void BinaryFile::parse() {
  ArrayRef<uint8_t> data = arrayRefFromStringRef(mb.getBuffer());
  auto *section =
      make<InputSection>(this, ".data", SHT_PROGBITS, SHF_ALLOC | SHF_WRITE,
                         /*addralign=*/8, /*entsize=*/0, data);
  sections.push_back(section);

  // For each input file foo that is embedded to a result as a binary
  // blob, we define _binary_foo_{start,end,size} symbols, so that
  // user programs can access blobs by name. Non-alphanumeric
  // characters in a filename are replaced with underscore.
  std::string s = "_binary_" + mb.getBufferIdentifier().str();
  for (char &c : s)
    if (!isAlnum(c))
      c = '_';

  llvm::StringSaver &ss = ctx.saver;
  ctx.symtab->addAndCheckDuplicate(
      ctx, Defined{ctx, this, ss.save(s + "_start"), STB_GLOBAL, STV_DEFAULT,
                   STT_OBJECT, 0, 0, section});
  ctx.symtab->addAndCheckDuplicate(
      ctx, Defined{ctx, this, ss.save(s + "_end"), STB_GLOBAL, STV_DEFAULT,
                   STT_OBJECT, data.size(), 0, section});
  ctx.symtab->addAndCheckDuplicate(
      ctx, Defined{ctx, this, ss.save(s + "_size"), STB_GLOBAL, STV_DEFAULT,
                   STT_OBJECT, data.size(), 0, nullptr});
}

InputFile *elf::createInternalFile(Ctx &ctx, StringRef name) {
  auto *file =
      make<InputFile>(ctx, InputFile::InternalKind, MemoryBufferRef("", name));
  // References from an internal file do not lead to --warn-backrefs
  // diagnostics.
  file->groupId = 0;
  return file;
}

std::unique_ptr<ELFFileBase> elf::createObjFile(Ctx &ctx, MemoryBufferRef mb,
                                                StringRef archiveName,
                                                bool lazy) {
  std::unique_ptr<ELFFileBase> f;
  switch (getELFKind(ctx, mb, archiveName)) {
  case ELF32LEKind:
    f = std::make_unique<ObjFile<ELF32LE>>(ctx, ELF32LEKind, mb, archiveName);
    break;
  case ELF32BEKind:
    f = std::make_unique<ObjFile<ELF32BE>>(ctx, ELF32BEKind, mb, archiveName);
    break;
  case ELF64LEKind:
    f = std::make_unique<ObjFile<ELF64LE>>(ctx, ELF64LEKind, mb, archiveName);
    break;
  case ELF64BEKind:
    f = std::make_unique<ObjFile<ELF64BE>>(ctx, ELF64BEKind, mb, archiveName);
    break;
  default:
    llvm_unreachable("getELFKind");
  }
  f->init();
  f->lazy = lazy;
  return f;
}

template <class ELFT> void ObjFile<ELFT>::parseLazy() {
  const ArrayRef<typename ELFT::Sym> eSyms = this->getELFSyms<ELFT>();
  numSymbols = eSyms.size();
  symbols = std::make_unique<Symbol *[]>(numSymbols);

  // resolve() may trigger this->extract() if an existing symbol is an undefined
  // symbol. If that happens, this function has served its purpose, and we can
  // exit from the loop early.
  auto *symtab = ctx.symtab.get();
  for (size_t i = firstGlobal, end = eSyms.size(); i != end; ++i) {
    if (eSyms[i].st_shndx == SHN_UNDEF)
      continue;
    symbols[i] = symtab->insert(CHECK2(eSyms[i].getName(stringTable), this));
    symbols[i]->resolve(ctx, LazySymbol{*this});
    if (!lazy)
      break;
  }
}

bool InputFile::shouldExtractForCommon(StringRef name) const {
  if (isa<BitcodeFile>(this))
    return isBitcodeNonCommonDef(mb, name, archiveName);

  return isNonCommonDef(ctx, mb, name, archiveName);
}

std::string elf::replaceThinLTOSuffix(Ctx &ctx, StringRef path) {
  auto [suffix, repl] = ctx.arg.thinLTOObjectSuffixReplace;
  if (path.consume_back(suffix))
    return (path + repl).str();
  return std::string(path);
}

template class elf::ObjFile<ELF32LE>;
template class elf::ObjFile<ELF32BE>;
template class elf::ObjFile<ELF64LE>;
template class elf::ObjFile<ELF64BE>;

template void SharedFile::parse<ELF32LE>();
template void SharedFile::parse<ELF32BE>();
template void SharedFile::parse<ELF64LE>();
template void SharedFile::parse<ELF64BE>();