//
//===----------------------------------------------------------------------===//
-#include "llvm/MC/MachObjectWriter.h"
#include "llvm/ADT/StringMap.h"
#include "llvm/ADT/Twine.h"
#include "llvm/MC/MCAssembler.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/ErrorHandling.h"
-#include "llvm/Support/MachO.h"
#include "llvm/Target/TargetAsmBackend.h"
// FIXME: Gross.
#include <vector>
using namespace llvm;
+using namespace llvm::object;
+// FIXME: this has been copied from (or to) X86AsmBackend.cpp
static unsigned getFixupKindLog2Size(unsigned Kind) {
switch (Kind) {
- default: llvm_unreachable("invalid fixup kind!");
- case X86::reloc_pcrel_1byte:
+ // FIXME: Until ARM has it's own relocation stuff spun off, it comes
+ // through here and we don't want it to puke all over. Any reasonable
+ // values will only come when ARM relocation support gets added, at which
+ // point this will be X86 only again and the llvm_unreachable can be
+ // re-enabled.
+ default: return 0;// llvm_unreachable("invalid fixup kind!");
+ case FK_PCRel_1:
case FK_Data_1: return 0;
+ case FK_PCRel_2:
case FK_Data_2: return 1;
- case X86::reloc_pcrel_4byte:
+ case FK_PCRel_4:
case X86::reloc_riprel_4byte:
case X86::reloc_riprel_4byte_movq_load:
+ case X86::reloc_signed_4byte:
case FK_Data_4: return 2;
case FK_Data_8: return 3;
}
switch (Kind) {
default:
return false;
- case X86::reloc_pcrel_1byte:
- case X86::reloc_pcrel_4byte:
+ case FK_PCRel_1:
+ case FK_PCRel_2:
+ case FK_PCRel_4:
case X86::reloc_riprel_4byte:
case X86::reloc_riprel_4byte_movq_load:
return true;
Kind == X86::reloc_riprel_4byte_movq_load;
}
-namespace {
-
-class MachObjectWriterImpl {
- // See <mach-o/loader.h>.
- enum {
- Header_Magic32 = 0xFEEDFACE,
- Header_Magic64 = 0xFEEDFACF
- };
+static bool doesSymbolRequireExternRelocation(MCSymbolData *SD) {
+ // Undefined symbols are always extern.
+ if (SD->Symbol->isUndefined())
+ return true;
- enum {
- Header32Size = 28,
- Header64Size = 32,
- SegmentLoadCommand32Size = 56,
- SegmentLoadCommand64Size = 72,
- Section32Size = 68,
- Section64Size = 80,
- SymtabLoadCommandSize = 24,
- DysymtabLoadCommandSize = 80,
- Nlist32Size = 12,
- Nlist64Size = 16,
- RelocationInfoSize = 8
- };
+ // References to weak definitions require external relocation entries; the
+ // definition may not always be the one in the same object file.
+ if (SD->getFlags() & SF_WeakDefinition)
+ return true;
- enum HeaderFileType {
- HFT_Object = 0x1
- };
+ // Otherwise, we can use an internal relocation.
+ return false;
+}
- enum HeaderFlags {
- HF_SubsectionsViaSymbols = 0x2000
- };
+static bool isScatteredFixupFullyResolved(const MCAssembler &Asm,
+ const MCValue Target,
+ const MCSymbolData *BaseSymbol) {
+ // The effective fixup address is
+ // addr(atom(A)) + offset(A)
+ // - addr(atom(B)) - offset(B)
+ // - addr(BaseSymbol) + <fixup offset from base symbol>
+ // and the offsets are not relocatable, so the fixup is fully resolved when
+ // addr(atom(A)) - addr(atom(B)) - addr(BaseSymbol) == 0.
+ //
+ // Note that "false" is almost always conservatively correct (it means we emit
+ // a relocation which is unnecessary), except when it would force us to emit a
+ // relocation which the target cannot encode.
+
+ const MCSymbolData *A_Base = 0, *B_Base = 0;
+ if (const MCSymbolRefExpr *A = Target.getSymA()) {
+ // Modified symbol references cannot be resolved.
+ if (A->getKind() != MCSymbolRefExpr::VK_None)
+ return false;
+
+ A_Base = Asm.getAtom(&Asm.getSymbolData(A->getSymbol()));
+ if (!A_Base)
+ return false;
+ }
- enum LoadCommandType {
- LCT_Segment = 0x1,
- LCT_Symtab = 0x2,
- LCT_Dysymtab = 0xb,
- LCT_Segment64 = 0x19
- };
+ if (const MCSymbolRefExpr *B = Target.getSymB()) {
+ // Modified symbol references cannot be resolved.
+ if (B->getKind() != MCSymbolRefExpr::VK_None)
+ return false;
- // See <mach-o/nlist.h>.
- enum SymbolTypeType {
- STT_Undefined = 0x00,
- STT_Absolute = 0x02,
- STT_Section = 0x0e
- };
+ B_Base = Asm.getAtom(&Asm.getSymbolData(B->getSymbol()));
+ if (!B_Base)
+ return false;
+ }
- enum SymbolTypeFlags {
- // If any of these bits are set, then the entry is a stab entry number (see
- // <mach-o/stab.h>. Otherwise the other masks apply.
- STF_StabsEntryMask = 0xe0,
+ // If there is no base, A and B have to be the same atom for this fixup to be
+ // fully resolved.
+ if (!BaseSymbol)
+ return A_Base == B_Base;
- STF_TypeMask = 0x0e,
- STF_External = 0x01,
- STF_PrivateExtern = 0x10
- };
+ // Otherwise, B must be missing and A must be the base.
+ return !B_Base && BaseSymbol == A_Base;
+}
- /// IndirectSymbolFlags - Flags for encoding special values in the indirect
- /// symbol entry.
- enum IndirectSymbolFlags {
- ISF_Local = 0x80000000,
- ISF_Absolute = 0x40000000
- };
+static bool isScatteredFixupFullyResolvedSimple(const MCAssembler &Asm,
+ const MCValue Target,
+ const MCSection *BaseSection) {
+ // The effective fixup address is
+ // addr(atom(A)) + offset(A)
+ // - addr(atom(B)) - offset(B)
+ // - addr(<base symbol>) + <fixup offset from base symbol>
+ // and the offsets are not relocatable, so the fixup is fully resolved when
+ // addr(atom(A)) - addr(atom(B)) - addr(<base symbol>)) == 0.
+ //
+ // The simple (Darwin, except on x86_64) way of dealing with this was to
+ // assume that any reference to a temporary symbol *must* be a temporary
+ // symbol in the same atom, unless the sections differ. Therefore, any PCrel
+ // relocation to a temporary symbol (in the same section) is fully
+ // resolved. This also works in conjunction with absolutized .set, which
+ // 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.
+
+ // Non-relative fixups are only resolved if constant.
+ if (!BaseSection)
+ return Target.isAbsolute();
+
+ // Otherwise, relative fixups are only resolved if not a difference and the
+ // target is a temporary in the same section.
+ if (Target.isAbsolute() || Target.getSymB())
+ return false;
- /// RelocationFlags - Special flags for addresses.
- enum RelocationFlags {
- RF_Scattered = 0x80000000
- };
+ const MCSymbol *A = &Target.getSymA()->getSymbol();
+ if (!A->isTemporary() || !A->isInSection() ||
+ &A->getSection() != BaseSection)
+ return false;
- enum RelocationInfoType {
- RIT_Vanilla = 0,
- RIT_Pair = 1,
- RIT_Difference = 2,
- RIT_PreboundLazyPointer = 3,
- RIT_LocalDifference = 4
- };
+ return true;
+}
- /// X86_64 uses its own relocation types.
- enum RelocationInfoTypeX86_64 {
- RIT_X86_64_Unsigned = 0,
- RIT_X86_64_Signed = 1,
- RIT_X86_64_Branch = 2,
- RIT_X86_64_GOTLoad = 3,
- RIT_X86_64_GOT = 4,
- RIT_X86_64_Subtractor = 5,
- RIT_X86_64_Signed1 = 6,
- RIT_X86_64_Signed2 = 7,
- RIT_X86_64_Signed4 = 8
- };
+namespace {
+class MachObjectWriter : public MCObjectWriter {
/// MachSymbolData - Helper struct for containing some precomputed information
/// on symbols.
struct MachSymbolData {
// Support lexicographic sorting.
bool operator<(const MachSymbolData &RHS) const {
- const std::string &Name = SymbolData->getSymbol().getName();
- return Name < RHS.SymbolData->getSymbol().getName();
+ return SymbolData->getSymbol().getName() <
+ RHS.SymbolData->getSymbol().getName();
}
};
/// @name Relocation Data
/// @{
- struct MachRelocationEntry {
- uint32_t Word0;
- uint32_t Word1;
- };
-
llvm::DenseMap<const MCSectionData*,
- std::vector<MachRelocationEntry> > Relocations;
+ std::vector<macho::RelocationEntry> > Relocations;
+ llvm::DenseMap<const MCSectionData*, unsigned> IndirectSymBase;
/// @}
/// @name Symbol Table Data
/// @}
- MachObjectWriter *Writer;
-
- raw_ostream &OS;
-
unsigned Is64Bit : 1;
-public:
- MachObjectWriterImpl(MachObjectWriter *_Writer, bool _Is64Bit)
- : Writer(_Writer), OS(Writer->getStream()), Is64Bit(_Is64Bit) {
- }
+ uint32_t CPUType;
+ uint32_t CPUSubtype;
- void Write8(uint8_t Value) { Writer->Write8(Value); }
- void Write16(uint16_t Value) { Writer->Write16(Value); }
- void Write32(uint32_t Value) { Writer->Write32(Value); }
- void Write64(uint64_t Value) { Writer->Write64(Value); }
- void WriteZeros(unsigned N) { Writer->WriteZeros(N); }
- void WriteBytes(StringRef Str, unsigned ZeroFillSize = 0) {
- Writer->WriteBytes(Str, ZeroFillSize);
+public:
+ MachObjectWriter(raw_ostream &_OS,
+ bool _Is64Bit, uint32_t _CPUType, uint32_t _CPUSubtype,
+ bool _IsLittleEndian)
+ : MCObjectWriter(_OS, _IsLittleEndian),
+ Is64Bit(_Is64Bit), CPUType(_CPUType), CPUSubtype(_CPUSubtype) {
}
void WriteHeader(unsigned NumLoadCommands, unsigned LoadCommandsSize,
uint32_t Flags = 0;
if (SubsectionsViaSymbols)
- Flags |= HF_SubsectionsViaSymbols;
+ Flags |= macho::HF_SubsectionsViaSymbols;
// struct mach_header (28 bytes) or
// struct mach_header_64 (32 bytes)
uint64_t Start = OS.tell();
(void) Start;
- Write32(Is64Bit ? Header_Magic64 : Header_Magic32);
+ Write32(Is64Bit ? macho::HM_Object64 : macho::HM_Object32);
- // FIXME: Support cputype.
- Write32(Is64Bit ? MachO::CPUTypeX86_64 : MachO::CPUTypeI386);
- // FIXME: Support cpusubtype.
- Write32(MachO::CPUSubType_I386_ALL);
- Write32(HFT_Object);
- Write32(NumLoadCommands); // Object files have a single load command, the
- // segment.
+ Write32(CPUType);
+ Write32(CPUSubtype);
+
+ Write32(macho::HFT_Object);
+ Write32(NumLoadCommands);
Write32(LoadCommandsSize);
Write32(Flags);
if (Is64Bit)
Write32(0); // reserved
- assert(OS.tell() - Start == Is64Bit ? Header64Size : Header32Size);
+ assert(OS.tell() - Start == Is64Bit ?
+ macho::Header64Size : macho::Header32Size);
}
/// WriteSegmentLoadCommand - Write a segment load command.
uint64_t Start = OS.tell();
(void) Start;
- unsigned SegmentLoadCommandSize = Is64Bit ? SegmentLoadCommand64Size :
- SegmentLoadCommand32Size;
- Write32(Is64Bit ? LCT_Segment64 : LCT_Segment);
+ unsigned SegmentLoadCommandSize = Is64Bit ? macho::SegmentLoadCommand64Size:
+ macho::SegmentLoadCommand32Size;
+ Write32(Is64Bit ? macho::LCT_Segment64 : macho::LCT_Segment);
Write32(SegmentLoadCommandSize +
- NumSections * (Is64Bit ? Section64Size : Section32Size));
+ NumSections * (Is64Bit ? macho::Section64Size :
+ macho::Section32Size));
WriteBytes("", 16);
if (Is64Bit) {
uint64_t SectionSize = Layout.getSectionSize(&SD);
// The offset is unused for virtual sections.
- if (Asm.getBackend().isVirtualSection(SD.getSection())) {
+ if (SD.getSection().isVirtualSection()) {
assert(Layout.getSectionFileSize(&SD) == 0 && "Invalid file size!");
FileOffset = 0;
}
uint64_t Start = OS.tell();
(void) Start;
- // FIXME: cast<> support!
- const MCSectionMachO &Section =
- static_cast<const MCSectionMachO&>(SD.getSection());
+ const MCSectionMachO &Section = cast<MCSectionMachO>(SD.getSection());
WriteBytes(Section.getSectionName(), 16);
WriteBytes(Section.getSegmentName(), 16);
if (Is64Bit) {
Write32(NumRelocations ? RelocationsStart : 0);
Write32(NumRelocations);
Write32(Flags);
- Write32(0); // reserved1
+ Write32(IndirectSymBase.lookup(&SD)); // reserved1
Write32(Section.getStubSize()); // reserved2
if (Is64Bit)
Write32(0); // reserved3
- assert(OS.tell() - Start == Is64Bit ? Section64Size : Section32Size);
+ assert(OS.tell() - Start == Is64Bit ? macho::Section64Size :
+ macho::Section32Size);
}
void WriteSymtabLoadCommand(uint32_t SymbolOffset, uint32_t NumSymbols,
uint64_t Start = OS.tell();
(void) Start;
- Write32(LCT_Symtab);
- Write32(SymtabLoadCommandSize);
+ Write32(macho::LCT_Symtab);
+ Write32(macho::SymtabLoadCommandSize);
Write32(SymbolOffset);
Write32(NumSymbols);
Write32(StringTableOffset);
Write32(StringTableSize);
- assert(OS.tell() - Start == SymtabLoadCommandSize);
+ assert(OS.tell() - Start == macho::SymtabLoadCommandSize);
}
void WriteDysymtabLoadCommand(uint32_t FirstLocalSymbol,
uint64_t Start = OS.tell();
(void) Start;
- Write32(LCT_Dysymtab);
- Write32(DysymtabLoadCommandSize);
+ Write32(macho::LCT_Dysymtab);
+ Write32(macho::DysymtabLoadCommandSize);
Write32(FirstLocalSymbol);
Write32(NumLocalSymbols);
Write32(FirstExternalSymbol);
Write32(0); // locreloff
Write32(0); // nlocrel
- assert(OS.tell() - Start == DysymtabLoadCommandSize);
+ assert(OS.tell() - Start == macho::DysymtabLoadCommandSize);
}
void WriteNlist(MachSymbolData &MSD, const MCAsmLayout &Layout) {
//
// FIXME: Are the prebound or indirect fields possible here?
if (Symbol.isUndefined())
- Type = STT_Undefined;
+ Type = macho::STT_Undefined;
else if (Symbol.isAbsolute())
- Type = STT_Absolute;
+ Type = macho::STT_Absolute;
else
- Type = STT_Section;
+ Type = macho::STT_Section;
// FIXME: Set STAB bits.
if (Data.isPrivateExtern())
- Type |= STF_PrivateExtern;
+ Type |= macho::STF_PrivateExtern;
// Set external bit.
if (Data.isExternal() || Symbol.isUndefined())
- Type |= STF_External;
+ Type |= macho::STF_External;
// Compute the symbol address.
if (Symbol.isDefined()) {
if (Symbol.isAbsolute()) {
- llvm_unreachable("FIXME: Not yet implemented!");
+ Address = cast<MCConstantExpr>(Symbol.getVariableValue())->getValue();
} else {
Address = Layout.getSymbolAddress(&Data);
}
void RecordX86_64Relocation(const MCAssembler &Asm, const MCAsmLayout &Layout,
const MCFragment *Fragment,
- const MCAsmFixup &Fixup, MCValue Target,
+ const MCFixup &Fixup, MCValue Target,
uint64_t &FixedValue) {
- unsigned IsPCRel = isFixupKindPCRel(Fixup.Kind);
- unsigned IsRIPRel = isFixupKindRIPRel(Fixup.Kind);
- unsigned Log2Size = getFixupKindLog2Size(Fixup.Kind);
+ unsigned IsPCRel = isFixupKindPCRel(Fixup.getKind());
+ unsigned IsRIPRel = isFixupKindRIPRel(Fixup.getKind());
+ unsigned Log2Size = getFixupKindLog2Size(Fixup.getKind());
// See <reloc.h>.
- uint32_t Address = Layout.getFragmentOffset(Fragment) + Fixup.Offset;
+ uint32_t FixupOffset =
+ Layout.getFragmentOffset(Fragment) + Fixup.getOffset();
+ uint32_t FixupAddress =
+ Layout.getFragmentAddress(Fragment) + Fixup.getOffset();
int64_t Value = 0;
unsigned Index = 0;
unsigned IsExtern = 0;
// actual expression addend without the PCrel bias. However, instructions
// with data following the relocation are not accomodated for (see comment
// below regarding SIGNED{1,2,4}), so it isn't exactly that either.
- Value += 1 << Log2Size;
+ Value += 1LL << Log2Size;
}
if (Target.isAbsolute()) { // constant
// SymbolNum of 0 indicates the absolute section.
- Type = RIT_X86_64_Unsigned;
+ Type = macho::RIT_X86_64_Unsigned;
Index = 0;
// FIXME: I believe this is broken, I don't think the linker can
// yet).
if (IsPCRel) {
IsExtern = 1;
- Type = RIT_X86_64_Branch;
+ Type = macho::RIT_X86_64_Branch;
}
} else if (Target.getSymB()) { // A - B + constant
const MCSymbol *A = &Target.getSymA()->getSymbol();
MCSymbolData &A_SD = Asm.getSymbolData(*A);
- const MCSymbolData *A_Base = Asm.getAtom(Layout, &A_SD);
+ const MCSymbolData *A_Base = Asm.getAtom(&A_SD);
const MCSymbol *B = &Target.getSymB()->getSymbol();
MCSymbolData &B_SD = Asm.getSymbolData(*B);
- const MCSymbolData *B_Base = Asm.getAtom(Layout, &B_SD);
+ const MCSymbolData *B_Base = Asm.getAtom(&B_SD);
// Neither symbol can be modified.
if (Target.getSymA()->getKind() != MCSymbolRefExpr::VK_None ||
if (IsPCRel)
report_fatal_error("unsupported pc-relative relocation of difference");
- // We don't currently support any situation where one or both of the
- // symbols would require a local relocation. This is almost certainly
- // unused and may not be possible to encode correctly.
- if (!A_Base || !B_Base)
- report_fatal_error("unsupported local relocations in difference");
+ // The support for the situation where one or both of the symbols would
+ // require a local relocation is handled just like if the symbols were
+ // external. This is certainly used in the case of debug sections where
+ // the section has only temporary symbols and thus the symbols don't have
+ // base symbols. This is encoded using the section ordinal and
+ // non-extern relocation entries.
// Darwin 'as' doesn't emit correct relocations for this (it ends up with
- // a single SIGNED relocation); reject it for now.
- if (A_Base == B_Base)
+ // a single SIGNED relocation); reject it for now. Except the case where
+ // both symbols don't have a base, equal but both NULL.
+ if (A_Base == B_Base && A_Base)
report_fatal_error("unsupported relocation with identical base");
- Value += Layout.getSymbolAddress(&A_SD) - Layout.getSymbolAddress(A_Base);
- Value -= Layout.getSymbolAddress(&B_SD) - Layout.getSymbolAddress(B_Base);
+ Value += Layout.getSymbolAddress(&A_SD) -
+ (A_Base == NULL ? 0 : Layout.getSymbolAddress(A_Base));
+ Value -= Layout.getSymbolAddress(&B_SD) -
+ (B_Base == NULL ? 0 : Layout.getSymbolAddress(B_Base));
- Index = A_Base->getIndex();
- IsExtern = 1;
- Type = RIT_X86_64_Unsigned;
+ if (A_Base) {
+ Index = A_Base->getIndex();
+ IsExtern = 1;
+ }
+ else {
+ Index = A_SD.getFragment()->getParent()->getOrdinal() + 1;
+ IsExtern = 0;
+ }
+ Type = macho::RIT_X86_64_Unsigned;
- MachRelocationEntry MRE;
- MRE.Word0 = Address;
+ macho::RelocationEntry MRE;
+ MRE.Word0 = FixupOffset;
MRE.Word1 = ((Index << 0) |
(IsPCRel << 24) |
(Log2Size << 25) |
(Type << 28));
Relocations[Fragment->getParent()].push_back(MRE);
- Index = B_Base->getIndex();
- IsExtern = 1;
- Type = RIT_X86_64_Subtractor;
+ if (B_Base) {
+ Index = B_Base->getIndex();
+ IsExtern = 1;
+ }
+ else {
+ Index = B_SD.getFragment()->getParent()->getOrdinal() + 1;
+ IsExtern = 0;
+ }
+ Type = macho::RIT_X86_64_Subtractor;
} else {
const MCSymbol *Symbol = &Target.getSymA()->getSymbol();
MCSymbolData &SD = Asm.getSymbolData(*Symbol);
- const MCSymbolData *Base = Asm.getAtom(Layout, &SD);
+ const MCSymbolData *Base = Asm.getAtom(&SD);
+
+ // Relocations inside debug sections always use local relocations when
+ // possible. This seems to be done because the debugger doesn't fully
+ // understand x86_64 relocation entries, and expects to find values that
+ // have already been fixed up.
+ if (Symbol->isInSection()) {
+ const MCSectionMachO &Section = static_cast<const MCSectionMachO&>(
+ Fragment->getParent()->getSection());
+ if (Section.hasAttribute(MCSectionMachO::S_ATTR_DEBUG))
+ Base = 0;
+ }
// x86_64 almost always uses external relocations, except when there is no
// symbol to use as a base address (a local symbol with no preceeding
// Add the local offset, if needed.
if (Base != &SD)
Value += Layout.getSymbolAddress(&SD) - Layout.getSymbolAddress(Base);
- } else {
+ } else if (Symbol->isInSection()) {
// The index is the section ordinal (1-based).
Index = SD.getFragment()->getParent()->getOrdinal() + 1;
IsExtern = 0;
Value += Layout.getSymbolAddress(&SD);
if (IsPCRel)
- Value -= Address + (1 << Log2Size);
+ Value -= FixupAddress + (1 << Log2Size);
+ } else {
+ report_fatal_error("unsupported relocation of undefined symbol '" +
+ Symbol->getName() + "'");
}
MCSymbolRefExpr::VariantKind Modifier = Target.getSymA()->getKind();
// x86_64 distinguishes movq foo@GOTPCREL so that the linker can
// rewrite the movq to an leaq at link time if the symbol ends up in
// the same linkage unit.
- if (unsigned(Fixup.Kind) == X86::reloc_riprel_4byte_movq_load)
- Type = RIT_X86_64_GOTLoad;
+ if (unsigned(Fixup.getKind()) == X86::reloc_riprel_4byte_movq_load)
+ Type = macho::RIT_X86_64_GOTLoad;
else
- Type = RIT_X86_64_GOT;
- } else if (Modifier != MCSymbolRefExpr::VK_None)
+ Type = macho::RIT_X86_64_GOT;
+ } else if (Modifier == MCSymbolRefExpr::VK_TLVP) {
+ Type = macho::RIT_X86_64_TLV;
+ } else if (Modifier != MCSymbolRefExpr::VK_None) {
report_fatal_error("unsupported symbol modifier in relocation");
- else
- Type = RIT_X86_64_Signed;
+ } else {
+ Type = macho::RIT_X86_64_Signed;
+
+ // The Darwin x86_64 relocation format has a problem where it cannot
+ // encode an address (L<foo> + <constant>) which is outside the atom
+ // containing L<foo>. Generally, this shouldn't occur but it does
+ // happen when we have a RIPrel instruction with data following the
+ // relocation entry (e.g., movb $012, L0(%rip)). Even with the PCrel
+ // adjustment Darwin x86_64 uses, the offset is still negative and
+ // the linker has no way to recognize this.
+ //
+ // To work around this, Darwin uses several special relocation types
+ // to indicate the offsets. However, the specification or
+ // implementation of these seems to also be incomplete; they should
+ // adjust the addend as well based on the actual encoded instruction
+ // (the additional bias), but instead appear to just look at the
+ // final offset.
+ switch (-(Target.getConstant() + (1LL << Log2Size))) {
+ case 1: Type = macho::RIT_X86_64_Signed1; break;
+ case 2: Type = macho::RIT_X86_64_Signed2; break;
+ case 4: Type = macho::RIT_X86_64_Signed4; break;
+ }
+ }
} else {
if (Modifier != MCSymbolRefExpr::VK_None)
report_fatal_error("unsupported symbol modifier in branch "
"relocation");
- Type = RIT_X86_64_Branch;
- }
-
- // The Darwin x86_64 relocation format has a problem where it cannot
- // encode an address (L<foo> + <constant>) which is outside the atom
- // containing L<foo>. Generally, this shouldn't occur but it does happen
- // when we have a RIPrel instruction with data following the relocation
- // entry (e.g., movb $012, L0(%rip)). Even with the PCrel adjustment
- // Darwin x86_64 uses, the offset is still negative and the linker has
- // no way to recognize this.
- //
- // To work around this, Darwin uses several special relocation types to
- // indicate the offsets. However, the specification or implementation of
- // these seems to also be incomplete; they should adjust the addend as
- // well based on the actual encoded instruction (the additional bias),
- // but instead appear to just look at the final offset.
- if (IsRIPRel) {
- switch (-(Target.getConstant() + (1 << Log2Size))) {
- case 1: Type = RIT_X86_64_Signed1; break;
- case 2: Type = RIT_X86_64_Signed2; break;
- case 4: Type = RIT_X86_64_Signed4; break;
- }
+ Type = macho::RIT_X86_64_Branch;
}
} else {
if (Modifier == MCSymbolRefExpr::VK_GOT) {
- Type = RIT_X86_64_GOT;
+ Type = macho::RIT_X86_64_GOT;
} else if (Modifier == MCSymbolRefExpr::VK_GOTPCREL) {
// GOTPCREL is allowed as a modifier on non-PCrel instructions, in
// which case all we do is set the PCrel bit in the relocation entry;
// this is used with exception handling, for example. The source is
// required to include any necessary offset directly.
- Type = RIT_X86_64_GOT;
+ Type = macho::RIT_X86_64_GOT;
IsPCRel = 1;
+ } else if (Modifier == MCSymbolRefExpr::VK_TLVP) {
+ report_fatal_error("TLVP symbol modifier should have been rip-rel");
} else if (Modifier != MCSymbolRefExpr::VK_None)
report_fatal_error("unsupported symbol modifier in relocation");
else
- Type = RIT_X86_64_Unsigned;
+ Type = macho::RIT_X86_64_Unsigned;
}
}
FixedValue = Value;
// struct relocation_info (8 bytes)
- MachRelocationEntry MRE;
- MRE.Word0 = Address;
+ macho::RelocationEntry MRE;
+ MRE.Word0 = FixupOffset;
MRE.Word1 = ((Index << 0) |
(IsPCRel << 24) |
(Log2Size << 25) |
void RecordScatteredRelocation(const MCAssembler &Asm,
const MCAsmLayout &Layout,
const MCFragment *Fragment,
- const MCAsmFixup &Fixup, MCValue Target,
+ const MCFixup &Fixup, MCValue Target,
uint64_t &FixedValue) {
- uint32_t Address = Layout.getFragmentOffset(Fragment) + Fixup.Offset;
- unsigned IsPCRel = isFixupKindPCRel(Fixup.Kind);
- unsigned Log2Size = getFixupKindLog2Size(Fixup.Kind);
- unsigned Type = RIT_Vanilla;
+ uint32_t FixupOffset = Layout.getFragmentOffset(Fragment)+Fixup.getOffset();
+ unsigned IsPCRel = isFixupKindPCRel(Fixup.getKind());
+ unsigned Log2Size = getFixupKindLog2Size(Fixup.getKind());
+ unsigned Type = macho::RIT_Vanilla;
// See <reloc.h>.
const MCSymbol *A = &Target.getSymA()->getSymbol();
// Note that there is no longer any semantic difference between these two
// relocation types from the linkers point of view, this is done solely
// for pedantic compatibility with 'as'.
- Type = A_SD->isExternal() ? RIT_Difference : RIT_LocalDifference;
+ Type = A_SD->isExternal() ? macho::RIT_Difference :
+ macho::RIT_LocalDifference;
Value2 = Layout.getSymbolAddress(B_SD);
}
// Relocations are written out in reverse order, so the PAIR comes first.
- if (Type == RIT_Difference || Type == RIT_LocalDifference) {
- MachRelocationEntry MRE;
+ if (Type == macho::RIT_Difference || Type == macho::RIT_LocalDifference) {
+ macho::RelocationEntry MRE;
MRE.Word0 = ((0 << 0) |
- (RIT_Pair << 24) |
+ (macho::RIT_Pair << 24) |
(Log2Size << 28) |
(IsPCRel << 30) |
- RF_Scattered);
+ macho::RF_Scattered);
MRE.Word1 = Value2;
Relocations[Fragment->getParent()].push_back(MRE);
}
- MachRelocationEntry MRE;
- MRE.Word0 = ((Address << 0) |
- (Type << 24) |
- (Log2Size << 28) |
- (IsPCRel << 30) |
- RF_Scattered);
+ macho::RelocationEntry MRE;
+ MRE.Word0 = ((FixupOffset << 0) |
+ (Type << 24) |
+ (Log2Size << 28) |
+ (IsPCRel << 30) |
+ macho::RF_Scattered);
MRE.Word1 = Value;
Relocations[Fragment->getParent()].push_back(MRE);
}
+ void RecordTLVPRelocation(const MCAssembler &Asm,
+ const MCAsmLayout &Layout,
+ const MCFragment *Fragment,
+ const MCFixup &Fixup, MCValue Target,
+ uint64_t &FixedValue) {
+ assert(Target.getSymA()->getKind() == MCSymbolRefExpr::VK_TLVP &&
+ !Is64Bit &&
+ "Should only be called with a 32-bit TLVP relocation!");
+
+ unsigned Log2Size = getFixupKindLog2Size(Fixup.getKind());
+ uint32_t Value = Layout.getFragmentOffset(Fragment)+Fixup.getOffset();
+ unsigned IsPCRel = 0;
+
+ // Get the symbol data.
+ MCSymbolData *SD_A = &Asm.getSymbolData(Target.getSymA()->getSymbol());
+ unsigned Index = SD_A->getIndex();
+
+ // We're only going to have a second symbol in pic mode and it'll be a
+ // subtraction from the picbase. For 32-bit pic the addend is the difference
+ // between the picbase and the next address. For 32-bit static the addend
+ // is zero.
+ if (Target.getSymB()) {
+ // If this is a subtraction then we're pcrel.
+ uint32_t FixupAddress =
+ Layout.getFragmentAddress(Fragment) + Fixup.getOffset();
+ MCSymbolData *SD_B = &Asm.getSymbolData(Target.getSymB()->getSymbol());
+ IsPCRel = 1;
+ FixedValue = (FixupAddress - Layout.getSymbolAddress(SD_B) +
+ Target.getConstant());
+ FixedValue += 1ULL << Log2Size;
+ } else {
+ FixedValue = 0;
+ }
+
+ // struct relocation_info (8 bytes)
+ macho::RelocationEntry MRE;
+ MRE.Word0 = Value;
+ MRE.Word1 = ((Index << 0) |
+ (IsPCRel << 24) |
+ (Log2Size << 25) |
+ (1 << 27) | // Extern
+ (macho::RIT_TLV << 28)); // Type
+ Relocations[Fragment->getParent()].push_back(MRE);
+ }
+
void RecordRelocation(const MCAssembler &Asm, const MCAsmLayout &Layout,
- const MCFragment *Fragment, const MCAsmFixup &Fixup,
+ const MCFragment *Fragment, const MCFixup &Fixup,
MCValue Target, uint64_t &FixedValue) {
if (Is64Bit) {
RecordX86_64Relocation(Asm, Layout, Fragment, Fixup, Target, FixedValue);
return;
}
- unsigned IsPCRel = isFixupKindPCRel(Fixup.Kind);
- unsigned Log2Size = getFixupKindLog2Size(Fixup.Kind);
+ unsigned IsPCRel = isFixupKindPCRel(Fixup.getKind());
+ unsigned Log2Size = getFixupKindLog2Size(Fixup.getKind());
+
+ // If this is a 32-bit TLVP reloc it's handled a bit differently.
+ if (Target.getSymA() &&
+ Target.getSymA()->getKind() == MCSymbolRefExpr::VK_TLVP) {
+ RecordTLVPRelocation(Asm, Layout, Fragment, Fixup, Target, FixedValue);
+ return;
+ }
// If this is a difference or a defined symbol plus an offset, then we need
// a scattered relocation entry.
+ // Differences always require scattered relocations.
+ if (Target.getSymB())
+ return RecordScatteredRelocation(Asm, Layout, Fragment, Fixup,
+ Target, FixedValue);
+
+ // Get the symbol data, if any.
+ MCSymbolData *SD = 0;
+ if (Target.getSymA())
+ SD = &Asm.getSymbolData(Target.getSymA()->getSymbol());
+
+ // If this is an internal relocation with an offset, it also needs a
+ // scattered relocation entry.
uint32_t Offset = Target.getConstant();
if (IsPCRel)
Offset += 1 << Log2Size;
- if (Target.getSymB() ||
- (Target.getSymA() && !Target.getSymA()->getSymbol().isUndefined() &&
- Offset)) {
- RecordScatteredRelocation(Asm, Layout, Fragment, Fixup,Target,FixedValue);
- return;
- }
+ if (Offset && SD && !doesSymbolRequireExternRelocation(SD))
+ return RecordScatteredRelocation(Asm, Layout, Fragment, Fixup,
+ Target, FixedValue);
// See <reloc.h>.
- uint32_t Address = Layout.getFragmentOffset(Fragment) + Fixup.Offset;
- uint32_t Value = 0;
+ uint32_t FixupOffset = Layout.getFragmentOffset(Fragment)+Fixup.getOffset();
unsigned Index = 0;
unsigned IsExtern = 0;
unsigned Type = 0;
//
// FIXME: Currently, these are never generated (see code below). I cannot
// find a case where they are actually emitted.
- Type = RIT_Vanilla;
- Value = 0;
+ Type = macho::RIT_Vanilla;
} else {
- const MCSymbol *Symbol = &Target.getSymA()->getSymbol();
- MCSymbolData *SD = &Asm.getSymbolData(*Symbol);
-
- if (Symbol->isUndefined()) {
+ // Check whether we need an external or internal relocation.
+ if (doesSymbolRequireExternRelocation(SD)) {
IsExtern = 1;
Index = SD->getIndex();
- Value = 0;
+ // For external relocations, make sure to offset the fixup value to
+ // compensate for the addend of the symbol address, if it was
+ // undefined. This occurs with weak definitions, for example.
+ if (!SD->Symbol->isUndefined())
+ FixedValue -= Layout.getSymbolAddress(SD);
} else {
// The index is the section ordinal (1-based).
Index = SD->getFragment()->getParent()->getOrdinal() + 1;
- Value = Layout.getSymbolAddress(SD);
}
- Type = RIT_Vanilla;
+ Type = macho::RIT_Vanilla;
}
// struct relocation_info (8 bytes)
- MachRelocationEntry MRE;
- MRE.Word0 = Address;
+ macho::RelocationEntry MRE;
+ MRE.Word0 = FixupOffset;
MRE.Word1 = ((Index << 0) |
(IsPCRel << 24) |
(Log2Size << 25) |
// FIXME: Revisit this when the dust settles.
// Bind non lazy symbol pointers first.
+ unsigned IndirectIndex = 0;
for (MCAssembler::indirect_symbol_iterator it = Asm.indirect_symbol_begin(),
- ie = Asm.indirect_symbol_end(); it != ie; ++it) {
- // FIXME: cast<> support!
+ ie = Asm.indirect_symbol_end(); it != ie; ++it, ++IndirectIndex) {
const MCSectionMachO &Section =
- static_cast<const MCSectionMachO&>(it->SectionData->getSection());
+ cast<MCSectionMachO>(it->SectionData->getSection());
if (Section.getType() != MCSectionMachO::S_NON_LAZY_SYMBOL_POINTERS)
continue;
+ // Initialize the section indirect symbol base, if necessary.
+ if (!IndirectSymBase.count(it->SectionData))
+ IndirectSymBase[it->SectionData] = IndirectIndex;
+
Asm.getOrCreateSymbolData(*it->Symbol);
}
// Then lazy symbol pointers and symbol stubs.
+ IndirectIndex = 0;
for (MCAssembler::indirect_symbol_iterator it = Asm.indirect_symbol_begin(),
- ie = Asm.indirect_symbol_end(); it != ie; ++it) {
- // FIXME: cast<> support!
+ ie = Asm.indirect_symbol_end(); it != ie; ++it, ++IndirectIndex) {
const MCSectionMachO &Section =
- static_cast<const MCSectionMachO&>(it->SectionData->getSection());
+ cast<MCSectionMachO>(it->SectionData->getSection());
if (Section.getType() != MCSectionMachO::S_LAZY_SYMBOL_POINTERS &&
Section.getType() != MCSectionMachO::S_SYMBOL_STUBS)
continue;
+ // Initialize the section indirect symbol base, if necessary.
+ if (!IndirectSymBase.count(it->SectionData))
+ IndirectSymBase[it->SectionData] = IndirectIndex;
+
// Set the symbol type to undefined lazy, but only on construction.
//
// FIXME: Do not hardcode.
const MCSymbol &Symbol = it->getSymbol();
// Ignore non-linker visible symbols.
- if (!Asm.isSymbolLinkerVisible(it))
+ if (!Asm.isSymbolLinkerVisible(it->getSymbol()))
continue;
if (!it->isExternal() && !Symbol.isUndefined())
const MCSymbol &Symbol = it->getSymbol();
// Ignore non-linker visible symbols.
- if (!Asm.isSymbolLinkerVisible(it))
+ if (!Asm.isSymbolLinkerVisible(it->getSymbol()))
continue;
if (it->isExternal() || Symbol.isUndefined())
UndefinedSymbolData);
}
- void WriteObject(const MCAssembler &Asm, const MCAsmLayout &Layout) {
+
+ bool IsFixupFullyResolved(const MCAssembler &Asm,
+ const MCValue Target,
+ bool IsPCRel,
+ const MCFragment *DF) const {
+ // If we aren't using scattered symbols, the fixup is fully resolved.
+ if (!Asm.getBackend().hasScatteredSymbols())
+ return true;
+
+ // Otherwise, determine whether this value is actually resolved; scattering
+ // may cause atoms to move.
+
+ // Check if we are using the "simple" resolution algorithm (e.g.,
+ // i386).
+ if (!Asm.getBackend().hasReliableSymbolDifference()) {
+ const MCSection *BaseSection = 0;
+ if (IsPCRel)
+ BaseSection = &DF->getParent()->getSection();
+
+ return isScatteredFixupFullyResolvedSimple(Asm, Target, BaseSection);
+ }
+
+ // Otherwise, compute the proper answer as reliably as possible.
+
+ // If this is a PCrel relocation, find the base atom (identified by its
+ // symbol) that the fixup value is relative to.
+ const MCSymbolData *BaseSymbol = 0;
+ if (IsPCRel) {
+ BaseSymbol = DF->getAtom();
+ if (!BaseSymbol)
+ return false;
+ }
+
+ return isScatteredFixupFullyResolved(Asm, Target, BaseSymbol);
+ }
+
+ void 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 ?
- SegmentLoadCommand64Size + NumSections * Section64Size :
- SegmentLoadCommand32Size + NumSections * Section32Size;
+ macho::SegmentLoadCommand64Size + NumSections * macho::Section64Size :
+ macho::SegmentLoadCommand32Size + NumSections * macho::Section32Size;
// Add the symbol table load command sizes, if used.
unsigned NumSymbols = LocalSymbolData.size() + ExternalSymbolData.size() +
UndefinedSymbolData.size();
if (NumSymbols) {
NumLoadCommands += 2;
- LoadCommandsSize += SymtabLoadCommandSize + DysymtabLoadCommandSize;
+ LoadCommandsSize += (macho::SymtabLoadCommandSize +
+ macho::DysymtabLoadCommandSize);
}
// Compute the total size of the section data, as well as its file size and
// vm size.
- uint64_t SectionDataStart = (Is64Bit ? Header64Size : Header32Size)
- + LoadCommandsSize;
+ uint64_t SectionDataStart = (Is64Bit ? macho::Header64Size :
+ macho::Header32Size) + LoadCommandsSize;
uint64_t SectionDataSize = 0;
uint64_t SectionDataFileSize = 0;
uint64_t VMSize = 0;
VMSize = std::max(VMSize, Address + Size);
- if (Asm.getBackend().isVirtualSection(SD.getSection()))
+ if (SD.getSection().isVirtualSection())
continue;
SectionDataSize = std::max(SectionDataSize, Address + Size);
uint64_t RelocTableEnd = SectionDataStart + SectionDataFileSize;
for (MCAssembler::const_iterator it = Asm.begin(),
ie = Asm.end(); it != ie; ++it) {
- std::vector<MachRelocationEntry> &Relocs = Relocations[it];
+ std::vector<macho::RelocationEntry> &Relocs = Relocations[it];
unsigned NumRelocs = Relocs.size();
uint64_t SectionStart = SectionDataStart + Layout.getSectionAddress(it);
WriteSection(Asm, Layout, *it, SectionStart, RelocTableEnd, NumRelocs);
- RelocTableEnd += NumRelocs * RelocationInfoSize;
+ RelocTableEnd += NumRelocs * macho::RelocationInfoSize;
}
// Write the symbol table load command, if used.
// The string table is written after symbol table.
uint64_t StringTableOffset =
- SymbolTableOffset + NumSymTabSymbols * (Is64Bit ? Nlist64Size :
- Nlist32Size);
+ SymbolTableOffset + NumSymTabSymbols * (Is64Bit ? macho::Nlist64Size :
+ macho::Nlist32Size);
WriteSymtabLoadCommand(SymbolTableOffset, NumSymTabSymbols,
StringTableOffset, StringTable.size());
// Write the actual section data.
for (MCAssembler::const_iterator it = Asm.begin(),
ie = Asm.end(); it != ie; ++it)
- Asm.WriteSectionData(it, Layout, Writer);
+ Asm.WriteSectionData(it, Layout, this);
// Write the extra padding.
WriteZeros(SectionDataPadding);
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<MachRelocationEntry> &Relocs = Relocations[it];
+ 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);
// If this symbol is defined and internal, mark it as such.
if (it->Symbol->isDefined() &&
!Asm.getSymbolData(*it->Symbol).isExternal()) {
- uint32_t Flags = ISF_Local;
+ uint32_t Flags = macho::ISF_Local;
if (it->Symbol->isAbsolute())
- Flags |= ISF_Absolute;
+ Flags |= macho::ISF_Absolute;
Write32(Flags);
continue;
}
}
-MachObjectWriter::MachObjectWriter(raw_ostream &OS,
- bool Is64Bit,
- bool IsLittleEndian)
- : MCObjectWriter(OS, IsLittleEndian)
-{
- Impl = new MachObjectWriterImpl(this, Is64Bit);
-}
-
-MachObjectWriter::~MachObjectWriter() {
- delete (MachObjectWriterImpl*) Impl;
-}
-
-void MachObjectWriter::ExecutePostLayoutBinding(MCAssembler &Asm) {
- ((MachObjectWriterImpl*) Impl)->ExecutePostLayoutBinding(Asm);
-}
-
-void MachObjectWriter::RecordRelocation(const MCAssembler &Asm,
- const MCAsmLayout &Layout,
- const MCFragment *Fragment,
- const MCAsmFixup &Fixup, MCValue Target,
- uint64_t &FixedValue) {
- ((MachObjectWriterImpl*) Impl)->RecordRelocation(Asm, Layout, Fragment, Fixup,
- Target, FixedValue);
-}
-
-void MachObjectWriter::WriteObject(const MCAssembler &Asm,
- const MCAsmLayout &Layout) {
- ((MachObjectWriterImpl*) Impl)->WriteObject(Asm, Layout);
+MCObjectWriter *llvm::createMachObjectWriter(raw_ostream &OS, bool is64Bit,
+ uint32_t CPUType,
+ uint32_t CPUSubtype,
+ bool IsLittleEndian) {
+ return new MachObjectWriter(OS, is64Bit, CPUType, CPUSubtype, IsLittleEndian);
}