#include <vector>
using namespace llvm;
+#define DEBUG_TYPE "mc"
+
void MachObjectWriter::reset() {
Relocations.clear();
IndirectSymBase.clear();
bool MachObjectWriter::
doesSymbolRequireExternRelocation(const MCSymbolData *SD) {
// Undefined symbols are always extern.
- if (SD->Symbol->isUndefined())
+ if (SD->getSymbol().isUndefined())
return true;
// References to weak definitions require external relocation entries; the
MCValue Target;
- if (!S.getVariableValue()->EvaluateAsRelocatable(Target, Layout))
+ if (!S.getVariableValue()->EvaluateAsRelocatable(Target, &Layout, nullptr))
report_fatal_error("unable to evaluate offset for variable '" +
S.getName() + "'");
unsigned Flags = Section.getTypeAndAttributes();
if (SD.hasInstructions())
- Flags |= MCSectionMachO::S_ATTR_SOME_INSTRUCTIONS;
+ Flags |= MachO::S_ATTR_SOME_INSTRUCTIONS;
assert(isPowerOf2_32(SD.getAlignment()) && "Invalid alignment!");
Write32(Log2_32(SD.getAlignment()));
assert(OS.tell() - Start == sizeof(MachO::dysymtab_command));
}
+MachObjectWriter::MachSymbolData *
+MachObjectWriter::findSymbolData(const MCSymbol &Sym) {
+ for (auto &Entry : LocalSymbolData)
+ if (&Entry.SymbolData->getSymbol() == &Sym)
+ return &Entry;
+
+ for (auto &Entry : ExternalSymbolData)
+ if (&Entry.SymbolData->getSymbol() == &Sym)
+ return &Entry;
+
+ for (auto &Entry : UndefinedSymbolData)
+ if (&Entry.SymbolData->getSymbol() == &Sym)
+ return &Entry;
+
+ return nullptr;
+}
+
void MachObjectWriter::WriteNlist(MachSymbolData &MSD,
const MCAsmLayout &Layout) {
MCSymbolData &Data = *MSD.SymbolData;
- const MCSymbol &Symbol = Data.getSymbol();
+ const MCSymbol *Symbol = &Data.getSymbol();
+ const MCSymbol *AliasedSymbol = &Symbol->AliasedSymbol();
+ uint8_t SectionIndex = MSD.SectionIndex;
uint8_t Type = 0;
uint16_t Flags = Data.getFlags();
uint64_t Address = 0;
+ bool IsAlias = Symbol != AliasedSymbol;
+
+ MachSymbolData *AliaseeInfo;
+ if (IsAlias) {
+ AliaseeInfo = findSymbolData(*AliasedSymbol);
+ if (AliaseeInfo)
+ SectionIndex = AliaseeInfo->SectionIndex;
+ Symbol = AliasedSymbol;
+ }
// Set the N_TYPE bits. See <mach-o/nlist.h>.
//
// FIXME: Are the prebound or indirect fields possible here?
- if (Symbol.isUndefined())
+ if (IsAlias && Symbol->isUndefined())
+ Type = MachO::N_INDR;
+ else if (Symbol->isUndefined())
Type = MachO::N_UNDF;
- else if (Symbol.isAbsolute())
+ else if (Symbol->isAbsolute())
Type = MachO::N_ABS;
else
Type = MachO::N_SECT;
Type |= MachO::N_PEXT;
// Set external bit.
- if (Data.isExternal() || Symbol.isUndefined())
+ if (Data.isExternal() || (!IsAlias && Symbol->isUndefined()))
Type |= MachO::N_EXT;
// Compute the symbol address.
- if (Symbol.isDefined()) {
+ if (IsAlias && Symbol->isUndefined())
+ Address = AliaseeInfo->StringIndex;
+ else if (Symbol->isDefined())
Address = getSymbolAddress(&Data, Layout);
- } else if (Data.isCommon()) {
+ else if (Data.isCommon()) {
// Common symbols are encoded with the size in the address
// field, and their alignment in the flags.
Address = Data.getCommonSize();
assert((1U << Log2Size) == Align && "Invalid 'common' alignment!");
if (Log2Size > 15)
report_fatal_error("invalid 'common' alignment '" +
- Twine(Align) + "' for '" + Symbol.getName() + "'",
+ Twine(Align) + "' for '" + Symbol->getName() + "'",
false);
// FIXME: Keep this mask with the SymbolFlags enumeration.
Flags = (Flags & 0xF0FF) | (Log2Size << 8);
}
}
+ if (Layout.getAssembler().isThumbFunc(Symbol))
+ Flags |= SF_ThumbFunc;
+
// struct nlist (12 bytes)
Write32(MSD.StringIndex);
Write8(Type);
- Write8(MSD.SectionIndex);
+ Write8(SectionIndex);
// The Mach-O streamer uses the lowest 16-bits of the flags for the 'desc'
// value.
static unsigned ComputeLinkerOptionsLoadCommandSize(
const std::vector<std::string> &Options, bool is64Bit)
{
- unsigned Size = sizeof(MachO::linker_options_command);
+ unsigned Size = sizeof(MachO::linker_option_command);
for (unsigned i = 0, e = Options.size(); i != e; ++i)
Size += Options[i].size() + 1;
return RoundUpToAlignment(Size, is64Bit ? 8 : 4);
uint64_t Start = OS.tell();
(void) Start;
- Write32(MachO::LC_LINKER_OPTIONS);
+ Write32(MachO::LC_LINKER_OPTION);
Write32(Size);
Write32(Options.size());
- uint64_t BytesWritten = sizeof(MachO::linker_options_command);
+ uint64_t BytesWritten = sizeof(MachO::linker_option_command);
for (unsigned i = 0, e = Options.size(); i != e; ++i) {
// Write each string, including the null byte.
const std::string &Option = Options[i];
assert(OS.tell() - Start == Size);
}
-
-void MachObjectWriter::RecordRelocation(const MCAssembler &Asm,
+void MachObjectWriter::RecordRelocation(MCAssembler &Asm,
const MCAsmLayout &Layout,
const MCFragment *Fragment,
- const MCFixup &Fixup,
- MCValue Target,
- uint64_t &FixedValue) {
+ const MCFixup &Fixup, MCValue Target,
+ bool &IsPCRel, uint64_t &FixedValue) {
TargetObjectWriter->RecordRelocation(this, Asm, Layout, Fragment, Fixup,
Target, FixedValue);
}
const MCSectionMachO &Section =
cast<MCSectionMachO>(it->SectionData->getSection());
- if (Section.getType() != MCSectionMachO::S_NON_LAZY_SYMBOL_POINTERS &&
- Section.getType() != MCSectionMachO::S_LAZY_SYMBOL_POINTERS &&
- Section.getType() != MCSectionMachO::S_SYMBOL_STUBS) {
+ if (Section.getType() != MachO::S_NON_LAZY_SYMBOL_POINTERS &&
+ Section.getType() != MachO::S_LAZY_SYMBOL_POINTERS &&
+ Section.getType() != MachO::S_SYMBOL_STUBS) {
MCSymbol &Symbol = *it->Symbol;
report_fatal_error("indirect symbol '" + Symbol.getName() +
"' not in a symbol pointer or stub section");
const MCSectionMachO &Section =
cast<MCSectionMachO>(it->SectionData->getSection());
- if (Section.getType() != MCSectionMachO::S_NON_LAZY_SYMBOL_POINTERS)
+ if (Section.getType() != MachO::S_NON_LAZY_SYMBOL_POINTERS)
continue;
// Initialize the section indirect symbol base, if necessary.
const MCSectionMachO &Section =
cast<MCSectionMachO>(it->SectionData->getSection());
- if (Section.getType() != MCSectionMachO::S_LAZY_SYMBOL_POINTERS &&
- Section.getType() != MCSectionMachO::S_SYMBOL_STUBS)
+ if (Section.getType() != MachO::S_LAZY_SYMBOL_POINTERS &&
+ Section.getType() != MachO::S_SYMBOL_STUBS)
continue;
// Initialize the section indirect symbol base, if necessary.
}
/// ComputeSymbolTable - Compute the symbol table data
-///
-/// \param StringTable [out] - The string table data.
-/// \param StringIndexMap [out] - Map from symbol names to offsets in the
-/// string table.
-void MachObjectWriter::
-ComputeSymbolTable(MCAssembler &Asm, SmallString<256> &StringTable,
- std::vector<MachSymbolData> &LocalSymbolData,
- std::vector<MachSymbolData> &ExternalSymbolData,
- std::vector<MachSymbolData> &UndefinedSymbolData) {
+void MachObjectWriter::ComputeSymbolTable(
+ MCAssembler &Asm, std::vector<MachSymbolData> &LocalSymbolData,
+ std::vector<MachSymbolData> &ExternalSymbolData,
+ std::vector<MachSymbolData> &UndefinedSymbolData) {
// Build section lookup table.
DenseMap<const MCSection*, uint8_t> SectionIndexMap;
unsigned Index = 1;
SectionIndexMap[&it->getSection()] = Index;
assert(Index <= 256 && "Too many sections!");
- // Index 0 is always the empty string.
- StringMap<uint64_t> StringIndexMap;
- StringTable += '\x00';
+ // Build the string table.
+ for (MCSymbolData &SD : Asm.symbols()) {
+ const MCSymbol &Symbol = SD.getSymbol();
+ if (!Asm.isSymbolLinkerVisible(Symbol))
+ continue;
- // Build the symbol arrays and the string table, but only for non-local
- // symbols.
+ StringTable.add(Symbol.getName());
+ }
+ StringTable.finalize(StringTableBuilder::MachO);
+
+ // Build the symbol arrays but only for non-local symbols.
//
- // The particular order that we collect the symbols and create the string
- // table, then sort the symbols is chosen to match 'as'. Even though it
- // doesn't matter for correctness, this is important for letting us diff .o
- // files.
- for (MCAssembler::symbol_iterator it = Asm.symbol_begin(),
- ie = Asm.symbol_end(); it != ie; ++it) {
- const MCSymbol &Symbol = it->getSymbol();
+ // The particular order that we collect and then sort the symbols is chosen to
+ // match 'as'. Even though it doesn't matter for correctness, this is
+ // important for letting us diff .o files.
+ for (MCSymbolData &SD : Asm.symbols()) {
+ const MCSymbol &Symbol = SD.getSymbol();
// Ignore non-linker visible symbols.
- if (!Asm.isSymbolLinkerVisible(it->getSymbol()))
+ if (!Asm.isSymbolLinkerVisible(Symbol))
continue;
- if (!it->isExternal() && !Symbol.isUndefined())
+ if (!SD.isExternal() && !Symbol.isUndefined())
continue;
- uint64_t &Entry = StringIndexMap[Symbol.getName()];
- if (!Entry) {
- Entry = StringTable.size();
- StringTable += Symbol.getName();
- StringTable += '\x00';
- }
-
MachSymbolData MSD;
- MSD.SymbolData = it;
- MSD.StringIndex = Entry;
+ MSD.SymbolData = &SD;
+ MSD.StringIndex = StringTable.getOffset(Symbol.getName());
if (Symbol.isUndefined()) {
MSD.SectionIndex = 0;
}
// Now add the data for local symbols.
- for (MCAssembler::symbol_iterator it = Asm.symbol_begin(),
- ie = Asm.symbol_end(); it != ie; ++it) {
- const MCSymbol &Symbol = it->getSymbol();
+ for (MCSymbolData &SD : Asm.symbols()) {
+ const MCSymbol &Symbol = SD.getSymbol();
// Ignore non-linker visible symbols.
- if (!Asm.isSymbolLinkerVisible(it->getSymbol()))
+ if (!Asm.isSymbolLinkerVisible(Symbol))
continue;
- if (it->isExternal() || Symbol.isUndefined())
+ if (SD.isExternal() || Symbol.isUndefined())
continue;
- uint64_t &Entry = StringIndexMap[Symbol.getName()];
- if (!Entry) {
- Entry = StringTable.size();
- StringTable += Symbol.getName();
- StringTable += '\x00';
- }
-
MachSymbolData MSD;
- MSD.SymbolData = it;
- MSD.StringIndex = Entry;
+ MSD.SymbolData = &SD;
+ MSD.StringIndex = StringTable.getOffset(Symbol.getName());
if (Symbol.isAbsolute()) {
MSD.SectionIndex = 0;
for (unsigned i = 0, e = UndefinedSymbolData.size(); i != e; ++i)
UndefinedSymbolData[i].SymbolData->setIndex(Index++);
- // The string table is padded to a multiple of 4.
- while (StringTable.size() % 4)
- StringTable += '\x00';
+ for (const MCSectionData &SD : Asm) {
+ std::vector<RelAndSymbol> &Relocs = Relocations[&SD];
+ for (RelAndSymbol &Rel : Relocs) {
+ if (!Rel.Sym)
+ continue;
+
+ // Set the Index and the IsExtern bit.
+ unsigned Index = Rel.Sym->getIndex();
+ assert(isInt<24>(Index));
+ if (IsLittleEndian)
+ Rel.MRE.r_word1 = (Rel.MRE.r_word1 & (-1 << 24)) | Index | (1 << 27);
+ else
+ Rel.MRE.r_word1 = (Rel.MRE.r_word1 & 0xff) | Index << 8 | (1 << 4);
+ }
+ }
}
void MachObjectWriter::computeSectionAddresses(const MCAssembler &Asm,
void MachObjectWriter::markAbsoluteVariableSymbols(MCAssembler &Asm,
const MCAsmLayout &Layout) {
- for (MCAssembler::symbol_iterator i = Asm.symbol_begin(),
- e = Asm.symbol_end();
- i != e; ++i) {
- MCSymbolData &SD = *i;
+ for (MCSymbolData &SD : Asm.symbols()) {
if (!SD.getSymbol().isVariable())
continue;
// and neither symbol is external, mark the variable as absolute.
const MCExpr *Expr = SD.getSymbol().getVariableValue();
MCValue Value;
- if (Expr->EvaluateAsRelocatable(Value, Layout)) {
+ if (Expr->EvaluateAsRelocatable(Value, &Layout, nullptr)) {
if (Value.getSymA() && Value.getSymB())
const_cast<MCSymbol*>(&SD.getSymbol())->setAbsolute();
}
// Mark symbol difference expressions in variables (from .set or = directives)
// as absolute.
markAbsoluteVariableSymbols(Asm, Layout);
-
- // Compute symbol table information and bind symbol indices.
- ComputeSymbolTable(Asm, StringTable, LocalSymbolData, ExternalSymbolData,
- UndefinedSymbolData);
}
bool MachObjectWriter::
// - addr(atom(B)) - offset(B)
// and the offsets are not relocatable, so the fixup is fully resolved when
// addr(atom(A)) - addr(atom(B)) == 0.
- const MCSymbolData *A_Base = 0, *B_Base = 0;
+ const MCSymbolData *A_Base = nullptr, *B_Base = nullptr;
const MCSymbol &SA = DataA.getSymbol().AliasedSymbol();
const MCSection &SecA = SA.getSection();
// same assumptions about any symbol that we normally make about
// assembler locals.
- if (!Asm.getBackend().hasReliableSymbolDifference()) {
+ bool hasReliableSymbolDifference = isX86_64();
+ if (!hasReliableSymbolDifference) {
if (!SA.isInSection() || &SecA != &SecB ||
(!SA.isTemporary() &&
FB.getAtom() != Asm.getSymbolData(SA).getFragment()->getAtom() &&
return false;
}
+ // If they are not in the same section, we can't compute the diff.
+ if (&SecA != &SecB)
+ return false;
+
const MCFragment *FA = Asm.getSymbolData(SA).getFragment();
// Bail if the symbol has no fragment.
return false;
A_Base = FA->getAtom();
- if (!A_Base)
- return false;
-
B_Base = FB.getAtom();
- if (!B_Base)
- return false;
// If the atoms are the same, they are guaranteed to have the same address.
if (A_Base == B_Base)
void MachObjectWriter::WriteObject(MCAssembler &Asm,
const MCAsmLayout &Layout) {
+ // Compute symbol table information and bind symbol indices.
+ ComputeSymbolTable(Asm, LocalSymbolData, ExternalSymbolData,
+ UndefinedSymbolData);
+
unsigned NumSections = Asm.size();
+ const MCAssembler::VersionMinInfoType &VersionInfo =
+ Layout.getAssembler().getVersionMinInfo();
// The section data starts after the header, the segment load command (and
// section headers) and the symbol table.
sizeof(MachO::segment_command_64) + NumSections * sizeof(MachO::section_64):
sizeof(MachO::segment_command) + NumSections * sizeof(MachO::section);
+ // Add the deployment target version info load command size, if used.
+ if (VersionInfo.Major != 0) {
+ ++NumLoadCommands;
+ LoadCommandsSize += sizeof(MachO::version_min_command);
+ }
+
// Add the data-in-code load command size, if used.
unsigned NumDataRegions = Asm.getDataRegions().size();
if (NumDataRegions) {
LoadCommandsSize += sizeof(MachO::linkedit_data_command);
}
+ // Add the loh load command size, if used.
+ uint64_t LOHRawSize = Asm.getLOHContainer().getEmitSize(*this, Layout);
+ uint64_t LOHSize = RoundUpToAlignment(LOHRawSize, is64Bit() ? 8 : 4);
+ if (LOHSize) {
+ ++NumLoadCommands;
+ LoadCommandsSize += sizeof(MachO::linkedit_data_command);
+ }
+
// Add the symbol table load command sizes, if used.
unsigned NumSymbols = LocalSymbolData.size() + ExternalSymbolData.size() +
UndefinedSymbolData.size();
uint64_t RelocTableEnd = SectionDataStart + SectionDataFileSize;
for (MCAssembler::const_iterator it = Asm.begin(),
ie = Asm.end(); it != ie; ++it) {
- std::vector<MachO::any_relocation_info> &Relocs = Relocations[it];
+ std::vector<RelAndSymbol> &Relocs = Relocations[it];
unsigned NumRelocs = Relocs.size();
uint64_t SectionStart = SectionDataStart + getSectionAddress(it);
WriteSection(Asm, Layout, *it, SectionStart, RelocTableEnd, NumRelocs);
RelocTableEnd += NumRelocs * sizeof(MachO::any_relocation_info);
}
+ // Write out the deployment target information, if it's available.
+ if (VersionInfo.Major != 0) {
+ assert(VersionInfo.Update < 256 && "unencodable update target version");
+ assert(VersionInfo.Minor < 256 && "unencodable minor target version");
+ assert(VersionInfo.Major < 65536 && "unencodable major target version");
+ uint32_t EncodedVersion = VersionInfo.Update | (VersionInfo.Minor << 8) |
+ (VersionInfo.Major << 16);
+ Write32(VersionInfo.Kind == MCVM_OSXVersionMin ? MachO::LC_VERSION_MIN_MACOSX :
+ MachO::LC_VERSION_MIN_IPHONEOS);
+ Write32(sizeof(MachO::version_min_command));
+ Write32(EncodedVersion);
+ Write32(0); // reserved.
+ }
+
// Write the data-in-code load command, if used.
uint64_t DataInCodeTableEnd = RelocTableEnd + NumDataRegions * 8;
if (NumDataRegions) {
DataRegionsSize);
}
+ // Write the loh load command, if used.
+ uint64_t LOHTableEnd = DataInCodeTableEnd + LOHSize;
+ if (LOHSize)
+ WriteLinkeditLoadCommand(MachO::LC_LINKER_OPTIMIZATION_HINT,
+ DataInCodeTableEnd, LOHSize);
+
// Write the symbol table load command, if used.
if (NumSymbols) {
unsigned FirstLocalSymbol = 0;
// If used, the indirect symbols are written after the section data.
if (NumIndirectSymbols)
- IndirectSymbolOffset = DataInCodeTableEnd;
+ IndirectSymbolOffset = LOHTableEnd;
// The symbol table is written after the indirect symbol data.
- uint64_t SymbolTableOffset = DataInCodeTableEnd + IndirectSymbolSize;
+ uint64_t SymbolTableOffset = LOHTableEnd + IndirectSymbolSize;
// The string table is written after symbol table.
uint64_t StringTableOffset =
sizeof(MachO::nlist_64) :
sizeof(MachO::nlist));
WriteSymtabLoadCommand(SymbolTableOffset, NumSymTabSymbols,
- StringTableOffset, StringTable.size());
+ StringTableOffset, StringTable.data().size());
WriteDysymtabLoadCommand(FirstLocalSymbol, NumLocalSymbols,
FirstExternalSymbol, NumExternalSymbols,
ie = Asm.end(); it != ie; ++it) {
// Write the section relocation entries, in reverse order to match 'as'
// (approximately, the exact algorithm is more complicated than this).
- std::vector<MachO::any_relocation_info> &Relocs = Relocations[it];
+ std::vector<RelAndSymbol> &Relocs = Relocations[it];
for (unsigned i = 0, e = Relocs.size(); i != e; ++i) {
- Write32(Relocs[e - i - 1].r_word0);
- Write32(Relocs[e - i - 1].r_word1);
+ Write32(Relocs[e - i - 1].MRE.r_word0);
+ Write32(Relocs[e - i - 1].MRE.r_word1);
}
}
Write16(Data->Kind);
}
+ // Write out the loh commands, if there is one.
+ if (LOHSize) {
+#ifndef NDEBUG
+ unsigned Start = OS.tell();
+#endif
+ Asm.getLOHContainer().Emit(*this, Layout);
+ // Pad to a multiple of the pointer size.
+ WriteBytes("", OffsetToAlignment(LOHRawSize, is64Bit() ? 8 : 4));
+ assert(OS.tell() - Start == LOHSize);
+ }
+
// Write the symbol table data, if used.
if (NumSymbols) {
// Write the indirect symbol entries.
// special handling.
const MCSectionMachO &Section =
static_cast<const MCSectionMachO&>(it->SectionData->getSection());
- if (Section.getType() == MCSectionMachO::S_NON_LAZY_SYMBOL_POINTERS) {
+ if (Section.getType() == MachO::S_NON_LAZY_SYMBOL_POINTERS) {
// If this symbol is defined and internal, mark it as such.
if (it->Symbol->isDefined() &&
!Asm.getSymbolData(*it->Symbol).isExternal()) {
WriteNlist(UndefinedSymbolData[i], Layout);
// Write the string table.
- OS << StringTable.str();
+ OS << StringTable.data();
}
}
projects / oota-llvm.git / blobdiff