//===----------------------------------------------------------------------===//
#include "llvm/MC/MCMachObjectWriter.h"
-#include "llvm/ADT/OwningPtr.h"
#include "llvm/ADT/StringMap.h"
#include "llvm/ADT/Twine.h"
-#include "llvm/MC/MCAssembler.h"
#include "llvm/MC/MCAsmBackend.h"
#include "llvm/MC/MCAsmLayout.h"
+#include "llvm/MC/MCAssembler.h"
#include "llvm/MC/MCExpr.h"
+#include "llvm/MC/MCFixupKindInfo.h"
+#include "llvm/MC/MCMachOSymbolFlags.h"
#include "llvm/MC/MCObjectWriter.h"
#include "llvm/MC/MCSectionMachO.h"
#include "llvm/MC/MCSymbol.h"
-#include "llvm/MC/MCMachOSymbolFlags.h"
#include "llvm/MC/MCValue.h"
#include "llvm/Object/MachOFormat.h"
+#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
-
#include <vector>
using namespace llvm;
using namespace llvm::object;
+void MachObjectWriter::reset() {
+ Relocations.clear();
+ IndirectSymBase.clear();
+ StringTable.clear();
+ LocalSymbolData.clear();
+ ExternalSymbolData.clear();
+ UndefinedSymbolData.clear();
+ MCObjectWriter::reset();
+}
+
bool MachObjectWriter::
doesSymbolRequireExternRelocation(const MCSymbolData *SD) {
// Undefined symbols are always extern.
// If this is a variable, then recursively evaluate now.
if (S.isVariable()) {
+ if (const MCConstantExpr *C =
+ dyn_cast<const MCConstantExpr>(S.getVariableValue()))
+ return C->getValue();
+
+
MCValue Target;
if (!S.getVariableValue()->EvaluateAsRelocatable(Target, Layout))
report_fatal_error("unable to evaluate offset for variable '" +
/// WriteSegmentLoadCommand - Write a segment load command.
///
-/// \arg NumSections - The number of sections in this segment.
-/// \arg SectionDataSize - The total size of the sections.
+/// \param NumSections The number of sections in this segment.
+/// \param SectionDataSize The total size of the sections.
void MachObjectWriter::WriteSegmentLoadCommand(unsigned NumSections,
uint64_t VMSize,
uint64_t SectionDataStartOffset,
// Compute the symbol address.
if (Symbol.isDefined()) {
- if (Symbol.isAbsolute()) {
- Address = cast<MCConstantExpr>(Symbol.getVariableValue())->getValue();
- } else {
- Address = getSymbolAddress(&Data, Layout);
- }
+ 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.
Write32(Address);
}
+void MachObjectWriter::WriteLinkeditLoadCommand(uint32_t Type,
+ uint32_t DataOffset,
+ uint32_t DataSize) {
+ uint64_t Start = OS.tell();
+ (void) Start;
+
+ Write32(Type);
+ Write32(macho::LinkeditLoadCommandSize);
+ Write32(DataOffset);
+ Write32(DataSize);
+
+ assert(OS.tell() - Start == macho::LinkeditLoadCommandSize);
+}
+
+static unsigned ComputeLinkerOptionsLoadCommandSize(
+ const std::vector<std::string> &Options, bool is64Bit)
+{
+ unsigned Size = sizeof(macho::LinkerOptionsLoadCommand);
+ for (unsigned i = 0, e = Options.size(); i != e; ++i)
+ Size += Options[i].size() + 1;
+ return RoundUpToAlignment(Size, is64Bit ? 8 : 4);
+}
+
+void MachObjectWriter::WriteLinkerOptionsLoadCommand(
+ const std::vector<std::string> &Options)
+{
+ unsigned Size = ComputeLinkerOptionsLoadCommandSize(Options, is64Bit());
+ uint64_t Start = OS.tell();
+ (void) Start;
+
+ Write32(macho::LCT_LinkerOptions);
+ Write32(Size);
+ Write32(Options.size());
+ uint64_t BytesWritten = sizeof(macho::LinkerOptionsLoadCommand);
+ for (unsigned i = 0, e = Options.size(); i != e; ++i) {
+ // Write each string, including the null byte.
+ const std::string &Option = Options[i];
+ WriteBytes(Option.c_str(), Option.size() + 1);
+ BytesWritten += Option.size() + 1;
+ }
+
+ // Pad to a multiple of the pointer size.
+ WriteBytes("", OffsetToAlignment(BytesWritten, is64Bit() ? 8 : 4));
+
+ assert(OS.tell() - Start == Size);
+}
+
+
void MachObjectWriter::RecordRelocation(const MCAssembler &Asm,
const MCAsmLayout &Layout,
const MCFragment *Fragment,
continue;
// Initialize the section indirect symbol base, if necessary.
- if (!IndirectSymBase.count(it->SectionData))
- IndirectSymBase[it->SectionData] = IndirectIndex;
+ IndirectSymBase.insert(std::make_pair(it->SectionData, IndirectIndex));
Asm.getOrCreateSymbolData(*it->Symbol);
}
continue;
// Initialize the section indirect symbol base, if necessary.
- if (!IndirectSymBase.count(it->SectionData))
- IndirectSymBase[it->SectionData] = IndirectIndex;
+ IndirectSymBase.insert(std::make_pair(it->SectionData, IndirectIndex));
// Set the symbol type to undefined lazy, but only on construction.
//
}
}
+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;
+ if (!SD.getSymbol().isVariable())
+ continue;
+
+ // Is the variable is a symbol difference (SA - SB + C) expression,
+ // and neither symbol is external, mark the variable as absolute.
+ const MCExpr *Expr = SD.getSymbol().getVariableValue();
+ MCValue Value;
+ if (Expr->EvaluateAsRelocatable(Value, Layout)) {
+ if (Value.getSymA() && Value.getSymB())
+ const_cast<MCSymbol*>(&SD.getSymbol())->setAbsolute();
+ }
+ }
+}
+
void MachObjectWriter::ExecutePostLayoutBinding(MCAssembler &Asm,
const MCAsmLayout &Layout) {
computeSectionAddresses(Asm, Layout);
// Create symbol data for any indirect symbols.
BindIndirectSymbols(Asm);
+ // 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);
// requires the compiler to use .set to absolutize the differences between
// symbols which the compiler knows to be assembly time constants, so we
// don't need to worry about considering symbol differences fully resolved.
+ //
+ // If the file isn't using sub-sections-via-symbols, we can make the
+ // same assumptions about any symbol that we normally make about
+ // assembler locals.
if (!Asm.getBackend().hasReliableSymbolDifference()) {
- if (!SA.isTemporary() || !SA.isInSection() || &SecA != &SecB)
+ if (!SA.isInSection() || &SecA != &SecB ||
+ (!SA.isTemporary() &&
+ FB.getAtom() != Asm.getSymbolData(SA).getFragment()->getAtom() &&
+ Asm.getSubsectionsViaSymbols()))
return false;
return true;
}
}
void MachObjectWriter::WriteObject(MCAssembler &Asm,
- const MCAsmLayout &Layout) {
+ const MCAsmLayout &Layout) {
unsigned NumSections = Asm.size();
// The section data starts after the header, the segment load command (and
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();
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 :
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;
// If used, the indirect symbols are written after the section data.
if (NumIndirectSymbols)
- IndirectSymbolOffset = RelocTableEnd;
+ IndirectSymbolOffset = DataInCodeTableEnd;
// The symbol table is written after the indirect symbol data.
- uint64_t SymbolTableOffset = RelocTableEnd + IndirectSymbolSize;
+ uint64_t SymbolTableOffset = DataInCodeTableEnd + IndirectSymbolSize;
// The string table is written after symbol table.
uint64_t StringTableOffset =
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);
+ Asm.writeSectionData(it, Layout);
uint64_t Pad = getPaddingSize(it, Layout);
for (unsigned int i = 0; i < Pad; ++i)
}
}
+ // 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.