uint64_t MachObjectWriter::getSymbolAddress(const MCSymbolData* SD,
const MCAsmLayout &Layout) const {
const MCSymbol &S = SD->getSymbol();
// If this is a variable, then recursively evaluate now.
if (S.isVariable()) {
MCValue Target;
if (!S.getVariableValue()->EvaluateAsRelocatable(Target, Layout))
report_fatal_error("unable to evaluate offset for variable '" +
S.getName() + "'");
// Verify that any used symbols are defined.
if (Target.getSymA() && Target.getSymA()->getSymbol().isUndefined())
report_fatal_error("unable to evaluate offset to undefined symbol '" +
Target.getSymA()->getSymbol().getName() + "'");
if (Target.getSymB() && Target.getSymB()->getSymbol().isUndefined())
report_fatal_error("unable to evaluate offset to undefined symbol '" +
Target.getSymB()->getSymbol().getName() + "'");
uint64_t Address = Target.getConstant();
if (Target.getSymA())
Address += getSymbolAddress(&Layout.getAssembler().getSymbolData(
Target.getSymA()->getSymbol()), Layout);
if (Target.getSymB())
Address += getSymbolAddress(&Layout.getAssembler().getSymbolData(
Target.getSymB()->getSymbol()), Layout);
return Address;
}
return getSectionAddress(SD->getFragment()->getParent()) +
Layout.getSymbolOffset(SD);
}
// Each LOH is composed by, in this order (each field is encoded using ULEB128):
// - Its kind.
// - Its number of arguments (let say N).
// - Its arg1.
// - ...
// - Its argN.
// <arg1> to <argN> are absolute addresses in the object file, i.e.,
// relative addresses from the beginning of the object file.
void MCLOHDirective::Emit_impl(raw_ostream &OutStream,
const MachObjectWriter &ObjWriter,
const MCAsmLayout &Layout) const {
const MCAssembler &Asm = Layout.getAssembler();
encodeULEB128(Kind, OutStream);
encodeULEB128(Args.size(), OutStream);
for (LOHArgs::const_iterator It = Args.begin(), EndIt = Args.end();
It != EndIt; ++It)
encodeULEB128(ObjWriter.getSymbolAddress(&Asm.getSymbolData(**It), Layout),
OutStream);
}
uint64_t ELFObjectWriter::SymbolValue(const MCSymbol &Sym,
const MCAsmLayout &Layout) {
if (Sym.isCommon() && Sym.isExternal())
return Sym.getCommonAlignment();
uint64_t Res;
if (!Layout.getSymbolOffset(Sym, Res))
return 0;
if (Layout.getAssembler().isThumbFunc(&Sym))
Res |= 1;
return Res;
}
static bool getSymbolOffsetImpl(const MCAsmLayout &Layout,
const MCSymbolData *SD, bool ReportError,
uint64_t &Val) {
const MCSymbol &S = SD->getSymbol();
if (!S.isVariable())
return getLabelOffset(Layout, *SD, ReportError, Val);
// If SD is a variable, evaluate it.
MCValue Target;
if (!S.getVariableValue()->EvaluateAsValue(Target, &Layout, nullptr))
report_fatal_error("unable to evaluate offset for variable '" +
S.getName() + "'");
uint64_t Offset = Target.getConstant();
const MCAssembler &Asm = Layout.getAssembler();
const MCSymbolRefExpr *A = Target.getSymA();
if (A) {
uint64_t ValA;
if (!getLabelOffset(Layout, Asm.getSymbolData(A->getSymbol()), ReportError,
ValA))
return false;
Offset += ValA;
}
const MCSymbolRefExpr *B = Target.getSymB();
if (B) {
uint64_t ValB;
if (!getLabelOffset(Layout, Asm.getSymbolData(B->getSymbol()), ReportError,
ValB))
return false;
Offset -= ValB;
}
Val = Offset;
return true;
}
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.
unsigned NumLoadCommands = 1;
uint64_t LoadCommandsSize = is64Bit() ?
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) {
++NumLoadCommands;
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 = alignTo(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();
if (NumSymbols) {
NumLoadCommands += 2;
LoadCommandsSize += (sizeof(MachO::symtab_command) +
sizeof(MachO::dysymtab_command));
}
// Add the linker option load commands sizes.
for (const auto &Option : Asm.getLinkerOptions()) {
++NumLoadCommands;
LoadCommandsSize += ComputeLinkerOptionsLoadCommandSize(Option, is64Bit());
}
// Compute the total size of the section data, as well as its file size and vm
// size.
uint64_t SectionDataStart = (is64Bit() ? sizeof(MachO::mach_header_64) :
sizeof(MachO::mach_header)) + LoadCommandsSize;
uint64_t SectionDataSize = 0;
uint64_t SectionDataFileSize = 0;
uint64_t VMSize = 0;
for (const MCSection &Sec : Asm) {
uint64_t Address = getSectionAddress(&Sec);
uint64_t Size = Layout.getSectionAddressSize(&Sec);
uint64_t FileSize = Layout.getSectionFileSize(&Sec);
FileSize += getPaddingSize(&Sec, Layout);
VMSize = std::max(VMSize, Address + Size);
if (Sec.isVirtualSection())
continue;
SectionDataSize = std::max(SectionDataSize, Address + Size);
SectionDataFileSize = std::max(SectionDataFileSize, Address + FileSize);
}
// The section data is padded to 4 bytes.
//
// FIXME: Is this machine dependent?
unsigned SectionDataPadding = OffsetToAlignment(SectionDataFileSize, 4);
SectionDataFileSize += SectionDataPadding;
// Write the prolog, starting with the header and load command...
writeHeader(MachO::MH_OBJECT, NumLoadCommands, LoadCommandsSize,
Asm.getSubsectionsViaSymbols());
uint32_t Prot =
MachO::VM_PROT_READ | MachO::VM_PROT_WRITE | MachO::VM_PROT_EXECUTE;
writeSegmentLoadCommand("", NumSections, 0, VMSize, SectionDataStart,
SectionDataSize, Prot, Prot);
// ... and then the section headers.
uint64_t RelocTableEnd = SectionDataStart + SectionDataFileSize;
for (const MCSection &Section : Asm) {
const auto &Sec = cast<MCSectionMachO>(Section);
std::vector<RelAndSymbol> &Relocs = Relocations[&Sec];
unsigned NumRelocs = Relocs.size();
uint64_t SectionStart = SectionDataStart + getSectionAddress(&Sec);
unsigned Flags = Sec.getTypeAndAttributes();
if (Sec.hasInstructions())
Flags |= MachO::S_ATTR_SOME_INSTRUCTIONS;
writeSection(Layout, Sec, getSectionAddress(&Sec), SectionStart, Flags,
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);
MachO::LoadCommandType LCType;
switch (VersionInfo.Kind) {
case MCVM_OSXVersionMin:
LCType = MachO::LC_VERSION_MIN_MACOSX;
break;
case MCVM_IOSVersionMin:
LCType = MachO::LC_VERSION_MIN_IPHONEOS;
break;
case MCVM_TvOSVersionMin:
LCType = MachO::LC_VERSION_MIN_TVOS;
break;
case MCVM_WatchOSVersionMin:
LCType = MachO::LC_VERSION_MIN_WATCHOS;
break;
}
write32(LCType);
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) {
uint64_t DataRegionsOffset = RelocTableEnd;
uint64_t DataRegionsSize = NumDataRegions * 8;
writeLinkeditLoadCommand(MachO::LC_DATA_IN_CODE, DataRegionsOffset,
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;
unsigned NumLocalSymbols = LocalSymbolData.size();
unsigned FirstExternalSymbol = FirstLocalSymbol + NumLocalSymbols;
unsigned NumExternalSymbols = ExternalSymbolData.size();
unsigned FirstUndefinedSymbol = FirstExternalSymbol + NumExternalSymbols;
unsigned NumUndefinedSymbols = UndefinedSymbolData.size();
unsigned NumIndirectSymbols = Asm.indirect_symbol_size();
unsigned NumSymTabSymbols =
NumLocalSymbols + NumExternalSymbols + NumUndefinedSymbols;
uint64_t IndirectSymbolSize = NumIndirectSymbols * 4;
uint64_t IndirectSymbolOffset = 0;
// If used, the indirect symbols are written after the section data.
if (NumIndirectSymbols)
IndirectSymbolOffset = LOHTableEnd;
// The symbol table is written after the indirect symbol data.
uint64_t SymbolTableOffset = LOHTableEnd + IndirectSymbolSize;
// The string table is written after symbol table.
uint64_t StringTableOffset =
SymbolTableOffset + NumSymTabSymbols * (is64Bit() ?
sizeof(MachO::nlist_64) :
sizeof(MachO::nlist));
writeSymtabLoadCommand(SymbolTableOffset, NumSymTabSymbols,
StringTableOffset, StringTable.data().size());
writeDysymtabLoadCommand(FirstLocalSymbol, NumLocalSymbols,
FirstExternalSymbol, NumExternalSymbols,
FirstUndefinedSymbol, NumUndefinedSymbols,
IndirectSymbolOffset, NumIndirectSymbols);
}
// Write the linker options load commands.
for (const auto &Option : Asm.getLinkerOptions())
writeLinkerOptionsLoadCommand(Option);
// Write the actual section data.
for (const MCSection &Sec : Asm) {
Asm.writeSectionData(&Sec, Layout);
uint64_t Pad = getPaddingSize(&Sec, Layout);
WriteZeros(Pad);
}
// Write the extra padding.
WriteZeros(SectionDataPadding);
// Write the relocation entries.
for (const MCSection &Sec : Asm) {
// Write the section relocation entries, in reverse order to match 'as'
// (approximately, the exact algorithm is more complicated than this).
std::vector<RelAndSymbol> &Relocs = Relocations[&Sec];
for (const RelAndSymbol &Rel : make_range(Relocs.rbegin(), Relocs.rend())) {
write32(Rel.MRE.r_word0);
write32(Rel.MRE.r_word1);
}
}
// Write out the data-in-code region payload, if there is one.
for (MCAssembler::const_data_region_iterator
it = Asm.data_region_begin(), ie = Asm.data_region_end();
it != ie; ++it) {
const DataRegionData *Data = &(*it);
uint64_t Start = getSymbolAddress(*Data->Start, Layout);
uint64_t End = getSymbolAddress(*Data->End, Layout);
DEBUG(dbgs() << "data in code region-- kind: " << Data->Kind
<< " start: " << Start << "(" << Data->Start->getName() << ")"
<< " end: " << End << "(" << Data->End->getName() << ")"
<< " size: " << End - Start
<< "\n");
write32(Start);
write16(End - Start);
write16(Data->Kind);
}
// Write out the loh commands, if there is one.
if (LOHSize) {
#ifndef NDEBUG
unsigned Start = getStream().tell();
#endif
Asm.getLOHContainer().emit(*this, Layout);
// Pad to a multiple of the pointer size.
writeBytes("", OffsetToAlignment(LOHRawSize, is64Bit() ? 8 : 4));
assert(getStream().tell() - Start == LOHSize);
}
// Write the symbol table data, if used.
if (NumSymbols) {
// Write the indirect symbol entries.
for (MCAssembler::const_indirect_symbol_iterator
it = Asm.indirect_symbol_begin(),
ie = Asm.indirect_symbol_end(); it != ie; ++it) {
// Indirect symbols in the non-lazy symbol pointer section have some
// special handling.
const MCSectionMachO &Section =
static_cast<const MCSectionMachO &>(*it->Section);
if (Section.getType() == MachO::S_NON_LAZY_SYMBOL_POINTERS) {
// If this symbol is defined and internal, mark it as such.
if (it->Symbol->isDefined() && !it->Symbol->isExternal()) {
uint32_t Flags = MachO::INDIRECT_SYMBOL_LOCAL;
if (it->Symbol->isAbsolute())
Flags |= MachO::INDIRECT_SYMBOL_ABS;
write32(Flags);
continue;
}
}
write32(it->Symbol->getIndex());
}
// FIXME: Check that offsets match computed ones.
// Write the symbol table entries.
for (auto *SymbolData :
{&LocalSymbolData, &ExternalSymbolData, &UndefinedSymbolData})
for (MachSymbolData &Entry : *SymbolData)
writeNlist(Entry, Layout);
// Write the string table.
getStream() << StringTable.data();
}
}
void MachObjectWriter::WriteObject(MCAssembler &Asm,
const MCAsmLayout &Layout) {
unsigned NumSections = Asm.size();
// The section data starts after the header, the segment load command (and
// section headers) and the symbol table.
unsigned NumLoadCommands = 1;
uint64_t LoadCommandsSize = is64Bit() ?
macho::SegmentLoadCommand64Size + NumSections * macho::Section64Size :
macho::SegmentLoadCommand32Size + NumSections * macho::Section32Size;
// Add the data-in-code load command size, if used.
unsigned NumDataRegions = Asm.getDataRegions().size();
if (NumDataRegions) {
++NumLoadCommands;
LoadCommandsSize += macho::LinkeditLoadCommandSize;
}
// Add the symbol table load command sizes, if used.
unsigned NumSymbols = LocalSymbolData.size() + ExternalSymbolData.size() +
UndefinedSymbolData.size();
if (NumSymbols) {
NumLoadCommands += 2;
LoadCommandsSize += (macho::SymtabLoadCommandSize +
macho::DysymtabLoadCommandSize);
}
// Add the linker option load commands sizes.
const std::vector<std::vector<std::string> > &LinkerOptions =
Asm.getLinkerOptions();
for (unsigned i = 0, e = LinkerOptions.size(); i != e; ++i) {
++NumLoadCommands;
LoadCommandsSize += ComputeLinkerOptionsLoadCommandSize(LinkerOptions[i],
is64Bit());
}
// Compute the total size of the section data, as well as its file size and vm
// size.
uint64_t SectionDataStart = (is64Bit() ? macho::Header64Size :
macho::Header32Size) + LoadCommandsSize;
uint64_t SectionDataSize = 0;
uint64_t SectionDataFileSize = 0;
uint64_t VMSize = 0;
for (MCAssembler::const_iterator it = Asm.begin(),
ie = Asm.end(); it != ie; ++it) {
const MCSectionData &SD = *it;
uint64_t Address = getSectionAddress(&SD);
uint64_t Size = Layout.getSectionAddressSize(&SD);
uint64_t FileSize = Layout.getSectionFileSize(&SD);
FileSize += getPaddingSize(&SD, Layout);
VMSize = std::max(VMSize, Address + Size);
if (SD.getSection().isVirtualSection())
continue;
SectionDataSize = std::max(SectionDataSize, Address + Size);
SectionDataFileSize = std::max(SectionDataFileSize, Address + FileSize);
}
// The section data is padded to 4 bytes.
//
// FIXME: Is this machine dependent?
unsigned SectionDataPadding = OffsetToAlignment(SectionDataFileSize, 4);
SectionDataFileSize += SectionDataPadding;
// Write the prolog, starting with the header and load command...
WriteHeader(NumLoadCommands, LoadCommandsSize,
Asm.getSubsectionsViaSymbols());
WriteSegmentLoadCommand(NumSections, VMSize,
SectionDataStart, SectionDataSize);
// ... and then the section headers.
uint64_t RelocTableEnd = SectionDataStart + SectionDataFileSize;
for (MCAssembler::const_iterator it = Asm.begin(),
ie = Asm.end(); it != ie; ++it) {
std::vector<macho::RelocationEntry> &Relocs = Relocations[it];
unsigned NumRelocs = Relocs.size();
uint64_t SectionStart = SectionDataStart + getSectionAddress(it);
WriteSection(Asm, Layout, *it, SectionStart, RelocTableEnd, NumRelocs);
RelocTableEnd += NumRelocs * macho::RelocationInfoSize;
}
// Write the data-in-code load command, if used.
uint64_t DataInCodeTableEnd = RelocTableEnd + NumDataRegions * 8;
if (NumDataRegions) {
uint64_t DataRegionsOffset = RelocTableEnd;
uint64_t DataRegionsSize = NumDataRegions * 8;
WriteLinkeditLoadCommand(macho::LCT_DataInCode, DataRegionsOffset,
DataRegionsSize);
}
// Write the symbol table load command, if used.
if (NumSymbols) {
unsigned FirstLocalSymbol = 0;
unsigned NumLocalSymbols = LocalSymbolData.size();
unsigned FirstExternalSymbol = FirstLocalSymbol + NumLocalSymbols;
unsigned NumExternalSymbols = ExternalSymbolData.size();
unsigned FirstUndefinedSymbol = FirstExternalSymbol + NumExternalSymbols;
unsigned NumUndefinedSymbols = UndefinedSymbolData.size();
unsigned NumIndirectSymbols = Asm.indirect_symbol_size();
unsigned NumSymTabSymbols =
NumLocalSymbols + NumExternalSymbols + NumUndefinedSymbols;
uint64_t IndirectSymbolSize = NumIndirectSymbols * 4;
uint64_t IndirectSymbolOffset = 0;
// If used, the indirect symbols are written after the section data.
if (NumIndirectSymbols)
IndirectSymbolOffset = DataInCodeTableEnd;
// The symbol table is written after the indirect symbol data.
uint64_t SymbolTableOffset = DataInCodeTableEnd + IndirectSymbolSize;
// The string table is written after symbol table.
uint64_t StringTableOffset =
SymbolTableOffset + NumSymTabSymbols * (is64Bit() ? macho::Nlist64Size :
macho::Nlist32Size);
WriteSymtabLoadCommand(SymbolTableOffset, NumSymTabSymbols,
StringTableOffset, StringTable.size());
WriteDysymtabLoadCommand(FirstLocalSymbol, NumLocalSymbols,
FirstExternalSymbol, NumExternalSymbols,
FirstUndefinedSymbol, NumUndefinedSymbols,
IndirectSymbolOffset, NumIndirectSymbols);
}
// Write the linker options load commands.
for (unsigned i = 0, e = LinkerOptions.size(); i != e; ++i) {
WriteLinkerOptionsLoadCommand(LinkerOptions[i]);
}
// Write the actual section data.
for (MCAssembler::const_iterator it = Asm.begin(),
ie = Asm.end(); it != ie; ++it) {
Asm.writeSectionData(it, Layout);
uint64_t Pad = getPaddingSize(it, Layout);
for (unsigned int i = 0; i < Pad; ++i)
Write8(0);
}
// Write the extra padding.
WriteZeros(SectionDataPadding);
// Write the relocation entries.
for (MCAssembler::const_iterator it = Asm.begin(),
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::RelocationEntry> &Relocs = Relocations[it];
for (unsigned i = 0, e = Relocs.size(); i != e; ++i) {
Write32(Relocs[e - i - 1].Word0);
Write32(Relocs[e - i - 1].Word1);
}
}
// Write out the data-in-code region payload, if there is one.
for (MCAssembler::const_data_region_iterator
it = Asm.data_region_begin(), ie = Asm.data_region_end();
it != ie; ++it) {
const DataRegionData *Data = &(*it);
uint64_t Start =
getSymbolAddress(&Layout.getAssembler().getSymbolData(*Data->Start),
Layout);
uint64_t End =
getSymbolAddress(&Layout.getAssembler().getSymbolData(*Data->End),
Layout);
DEBUG(dbgs() << "data in code region-- kind: " << Data->Kind
<< " start: " << Start << "(" << Data->Start->getName() << ")"
<< " end: " << End << "(" << Data->End->getName() << ")"
<< " size: " << End - Start
<< "\n");
Write32(Start);
Write16(End - Start);
Write16(Data->Kind);
}
// Write the symbol table data, if used.
if (NumSymbols) {
// Write the indirect symbol entries.
for (MCAssembler::const_indirect_symbol_iterator
it = Asm.indirect_symbol_begin(),
ie = Asm.indirect_symbol_end(); it != ie; ++it) {
// Indirect symbols in the non lazy symbol pointer section have some
// special handling.
const MCSectionMachO &Section =
static_cast<const MCSectionMachO&>(it->SectionData->getSection());
if (Section.getType() == MCSectionMachO::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()) {
uint32_t Flags = macho::ISF_Local;
if (it->Symbol->isAbsolute())
Flags |= macho::ISF_Absolute;
Write32(Flags);
continue;
}
}
Write32(Asm.getSymbolData(*it->Symbol).getIndex());
}
// FIXME: Check that offsets match computed ones.
// Write the symbol table entries.
for (unsigned i = 0, e = LocalSymbolData.size(); i != e; ++i)
WriteNlist(LocalSymbolData[i], Layout);
for (unsigned i = 0, e = ExternalSymbolData.size(); i != e; ++i)
WriteNlist(ExternalSymbolData[i], Layout);
for (unsigned i = 0, e = UndefinedSymbolData.size(); i != e; ++i)
WriteNlist(UndefinedSymbolData[i], Layout);
// Write the string table.
OS << StringTable.str();
}
}
bool MCExpr::EvaluateAsRelocatable(MCValue &Res,
const MCAsmLayout &Layout) const {
return EvaluateAsRelocatableImpl(Res, &Layout.getAssembler(), &Layout,
0, false);
}
bool MCExpr::EvaluateAsAbsolute(int64_t &Res,
const MCAsmLayout &Layout,
const SectionAddrMap &Addrs) const {
return EvaluateAsAbsolute(Res, &Layout.getAssembler(), &Layout, &Addrs);
}
bool MCExpr::EvaluateAsAbsolute(int64_t &Res,
const MCAsmLayout &Layout) const {
return EvaluateAsAbsolute(Res, &Layout.getAssembler(), &Layout, 0);
}
void MachObjectWriter::WriteObject(MCAssembler &Asm,
const MCAsmLayout &Layout) {
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.
unsigned NumLoadCommands = 1;
uint64_t LoadCommandsSize = is64Bit() ?
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) {
++NumLoadCommands;
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();
if (NumSymbols) {
NumLoadCommands += 2;
LoadCommandsSize += (sizeof(MachO::symtab_command) +
sizeof(MachO::dysymtab_command));
}
// Add the linker option load commands sizes.
const std::vector<std::vector<std::string> > &LinkerOptions =
Asm.getLinkerOptions();
for (unsigned i = 0, e = LinkerOptions.size(); i != e; ++i) {
++NumLoadCommands;
LoadCommandsSize += ComputeLinkerOptionsLoadCommandSize(LinkerOptions[i],
is64Bit());
}
// Compute the total size of the section data, as well as its file size and vm
// size.
uint64_t SectionDataStart = (is64Bit() ? sizeof(MachO::mach_header_64) :
sizeof(MachO::mach_header)) + LoadCommandsSize;
uint64_t SectionDataSize = 0;
uint64_t SectionDataFileSize = 0;
uint64_t VMSize = 0;
for (MCAssembler::const_iterator it = Asm.begin(),
ie = Asm.end(); it != ie; ++it) {
const MCSectionData &SD = *it;
uint64_t Address = getSectionAddress(&SD);
uint64_t Size = Layout.getSectionAddressSize(&SD);
uint64_t FileSize = Layout.getSectionFileSize(&SD);
FileSize += getPaddingSize(&SD, Layout);
VMSize = std::max(VMSize, Address + Size);
if (SD.getSection().isVirtualSection())
continue;
SectionDataSize = std::max(SectionDataSize, Address + Size);
SectionDataFileSize = std::max(SectionDataFileSize, Address + FileSize);
}
// The section data is padded to 4 bytes.
//
// FIXME: Is this machine dependent?
unsigned SectionDataPadding = OffsetToAlignment(SectionDataFileSize, 4);
SectionDataFileSize += SectionDataPadding;
// Write the prolog, starting with the header and load command...
WriteHeader(NumLoadCommands, LoadCommandsSize,
Asm.getSubsectionsViaSymbols());
WriteSegmentLoadCommand(NumSections, VMSize,
SectionDataStart, SectionDataSize);
// ... and then the section headers.
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];
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) {
uint64_t DataRegionsOffset = RelocTableEnd;
uint64_t DataRegionsSize = NumDataRegions * 8;
WriteLinkeditLoadCommand(MachO::LC_DATA_IN_CODE, DataRegionsOffset,
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;
unsigned NumLocalSymbols = LocalSymbolData.size();
unsigned FirstExternalSymbol = FirstLocalSymbol + NumLocalSymbols;
unsigned NumExternalSymbols = ExternalSymbolData.size();
unsigned FirstUndefinedSymbol = FirstExternalSymbol + NumExternalSymbols;
unsigned NumUndefinedSymbols = UndefinedSymbolData.size();
unsigned NumIndirectSymbols = Asm.indirect_symbol_size();
unsigned NumSymTabSymbols =
NumLocalSymbols + NumExternalSymbols + NumUndefinedSymbols;
uint64_t IndirectSymbolSize = NumIndirectSymbols * 4;
uint64_t IndirectSymbolOffset = 0;
// If used, the indirect symbols are written after the section data.
if (NumIndirectSymbols)
IndirectSymbolOffset = LOHTableEnd;
// The symbol table is written after the indirect symbol data.
uint64_t SymbolTableOffset = LOHTableEnd + IndirectSymbolSize;
// The string table is written after symbol table.
uint64_t StringTableOffset =
SymbolTableOffset + NumSymTabSymbols * (is64Bit() ?
sizeof(MachO::nlist_64) :
sizeof(MachO::nlist));
WriteSymtabLoadCommand(SymbolTableOffset, NumSymTabSymbols,
StringTableOffset, StringTable.data().size());
WriteDysymtabLoadCommand(FirstLocalSymbol, NumLocalSymbols,
FirstExternalSymbol, NumExternalSymbols,
FirstUndefinedSymbol, NumUndefinedSymbols,
IndirectSymbolOffset, NumIndirectSymbols);
}
// Write the linker options load commands.
for (unsigned i = 0, e = LinkerOptions.size(); i != e; ++i) {
WriteLinkerOptionsLoadCommand(LinkerOptions[i]);
}
// Write the actual section data.
for (MCAssembler::const_iterator it = Asm.begin(),
ie = Asm.end(); it != ie; ++it) {
Asm.writeSectionData(it, Layout);
uint64_t Pad = getPaddingSize(it, Layout);
for (unsigned int i = 0; i < Pad; ++i)
Write8(0);
}
// Write the extra padding.
WriteZeros(SectionDataPadding);
// Write the relocation entries.
for (MCAssembler::const_iterator it = Asm.begin(),
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];
for (unsigned i = 0, e = Relocs.size(); i != e; ++i) {
Write32(Relocs[e - i - 1].r_word0);
Write32(Relocs[e - i - 1].r_word1);
}
}
// Write out the data-in-code region payload, if there is one.
for (MCAssembler::const_data_region_iterator
it = Asm.data_region_begin(), ie = Asm.data_region_end();
it != ie; ++it) {
const DataRegionData *Data = &(*it);
uint64_t Start =
getSymbolAddress(&Layout.getAssembler().getSymbolData(*Data->Start),
Layout);
uint64_t End =
getSymbolAddress(&Layout.getAssembler().getSymbolData(*Data->End),
Layout);
DEBUG(dbgs() << "data in code region-- kind: " << Data->Kind
<< " start: " << Start << "(" << Data->Start->getName() << ")"
<< " end: " << End << "(" << Data->End->getName() << ")"
<< " size: " << End - Start
<< "\n");
Write32(Start);
Write16(End - Start);
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.
for (MCAssembler::const_indirect_symbol_iterator
it = Asm.indirect_symbol_begin(),
ie = Asm.indirect_symbol_end(); it != ie; ++it) {
// Indirect symbols in the non-lazy symbol pointer section have some
// special handling.
const MCSectionMachO &Section =
static_cast<const MCSectionMachO&>(it->SectionData->getSection());
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()) {
uint32_t Flags = MachO::INDIRECT_SYMBOL_LOCAL;
if (it->Symbol->isAbsolute())
Flags |= MachO::INDIRECT_SYMBOL_ABS;
Write32(Flags);
continue;
}
}
Write32(Asm.getSymbolData(*it->Symbol).getIndex());
}
// FIXME: Check that offsets match computed ones.
// Write the symbol table entries.
for (unsigned i = 0, e = LocalSymbolData.size(); i != e; ++i)
WriteNlist(LocalSymbolData[i], Layout);
for (unsigned i = 0, e = ExternalSymbolData.size(); i != e; ++i)
WriteNlist(ExternalSymbolData[i], Layout);
for (unsigned i = 0, e = UndefinedSymbolData.size(); i != e; ++i)
WriteNlist(UndefinedSymbolData[i], Layout);
// Write the string table.
OS << StringTable.data();
}
}
void MachObjectWriter::WriteNlist(MachSymbolData &MSD,
const MCAsmLayout &Layout) {
MCSymbolData &Data = *MSD.SymbolData;
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 (IsAlias && Symbol->isUndefined())
Type = MachO::N_INDR;
else if (Symbol->isUndefined())
Type = MachO::N_UNDF;
else if (Symbol->isAbsolute())
Type = MachO::N_ABS;
else
Type = MachO::N_SECT;
// FIXME: Set STAB bits.
if (Data.isPrivateExtern())
Type |= MachO::N_PEXT;
// Set external bit.
if (Data.isExternal() || (!IsAlias && Symbol->isUndefined()))
Type |= MachO::N_EXT;
// Compute the symbol address.
if (IsAlias && Symbol->isUndefined())
Address = AliaseeInfo->StringIndex;
else if (Symbol->isDefined())
Address = getSymbolAddress(&Data, Layout);
else if (Data.isCommon()) {
// Common symbols are encoded with the size in the address
// field, and their alignment in the flags.
Address = Data.getCommonSize();
// Common alignment is packed into the 'desc' bits.
if (unsigned Align = Data.getCommonAlignment()) {
unsigned Log2Size = Log2_32(Align);
assert((1U << Log2Size) == Align && "Invalid 'common' alignment!");
if (Log2Size > 15)
report_fatal_error("invalid 'common' alignment '" +
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(SectionIndex);
// The Mach-O streamer uses the lowest 16-bits of the flags for the 'desc'
// value.
Write16(Flags);
if (is64Bit())
Write64(Address);
else
Write32(Address);
}