#include <vector>
using namespace llvm;
+// 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:
case FK_Data_1: return 0;
+ case X86::reloc_pcrel_2byte:
case FK_Data_2: return 1;
case X86::reloc_pcrel_4byte:
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;
}
default:
return false;
case X86::reloc_pcrel_1byte:
+ case X86::reloc_pcrel_2byte:
case X86::reloc_pcrel_4byte:
case X86::reloc_riprel_4byte:
case X86::reloc_riprel_4byte_movq_load:
return false;
}
+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;
+ }
+
+ if (const MCSymbolRefExpr *B = Target.getSymB()) {
+ // Modified symbol references cannot be resolved.
+ if (B->getKind() != MCSymbolRefExpr::VK_None)
+ return false;
+
+ B_Base = Asm.getAtom(&Asm.getSymbolData(B->getSymbol()));
+ if (!B_Base)
+ return false;
+ }
+
+ // 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;
+
+ // Otherwise, B must be missing and A must be the base.
+ return !B_Base && BaseSymbol == A_Base;
+}
+
+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;
+
+ const MCSymbol *A = &Target.getSymA()->getSymbol();
+ if (!A->isTemporary() || !A->isInSection() ||
+ &A->getSection() != BaseSection)
+ return false;
+
+ return true;
+}
+
namespace {
class MachObjectWriterImpl {
unsigned Is64Bit : 1;
+ uint32_t CPUType;
+ uint32_t CPUSubtype;
+
public:
- MachObjectWriterImpl(MachObjectWriter *_Writer, bool _Is64Bit)
- : Writer(_Writer), OS(Writer->getStream()), Is64Bit(_Is64Bit) {
+ MachObjectWriterImpl(MachObjectWriter *_Writer, bool _Is64Bit,
+ uint32_t _CPUType, uint32_t _CPUSubtype)
+ : Writer(_Writer), OS(Writer->getStream()), Is64Bit(_Is64Bit),
+ CPUType(_CPUType), CPUSubtype(_CPUSubtype) {
}
void Write8(uint8_t Value) { Writer->Write8(Value); }
Write32(Is64Bit ? Header_Magic64 : Header_Magic32);
- // FIXME: Support cputype.
- Write32(Is64Bit ? MachO::CPUTypeX86_64 : MachO::CPUTypeI386);
- // FIXME: Support cpusubtype.
- Write32(MachO::CPUSubType_I386_ALL);
+ Write32(CPUType);
+ Write32(CPUSubtype);
+
Write32(HFT_Object);
Write32(NumLoadCommands); // Object files have a single load command, the
// segment.
} 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;
+ if (A_Base) {
+ Index = A_Base->getIndex();
+ IsExtern = 1;
+ }
+ else {
+ Index = A_SD.getFragment()->getParent()->getOrdinal() + 1;
+ IsExtern = 0;
+ }
Type = RIT_X86_64_Unsigned;
MachRelocationEntry MRE;
(Type << 28));
Relocations[Fragment->getParent()].push_back(MRE);
- Index = B_Base->getIndex();
- IsExtern = 1;
+ if (B_Base) {
+ Index = B_Base->getIndex();
+ IsExtern = 1;
+ }
+ else {
+ Index = B_SD.getFragment()->getParent()->getOrdinal() + 1;
+ IsExtern = 0;
+ }
Type = 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
!Is64Bit &&
"Should only be called with a 32-bit TLVP relocation!");
- // If this is a subtraction then we're pcrel.
unsigned Log2Size = getFixupKindLog2Size(Fixup.getKind());
uint32_t Value = Layout.getFragmentOffset(Fragment)+Fixup.getOffset();
unsigned IsPCRel = 0;
MCSymbolData *SD_A = &Asm.getSymbolData(Target.getSymA()->getSymbol());
unsigned Index = SD_A->getIndex();
- if (Target.getSymB())
+ // 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)
MachRelocationEntry MRE;
unsigned Log2Size = getFixupKindLog2Size(Fixup.getKind());
// If this is a 32-bit TLVP reloc it's handled a bit differently.
- if (Target.getSymA()->getKind() == MCSymbolRefExpr::VK_TLVP) {
+ if (Target.getSymA() &&
+ Target.getSymA()->getKind() == MCSymbolRefExpr::VK_TLVP) {
RecordTLVPRelocation(Asm, Layout, Fragment, Fixup, Target, FixedValue);
return;
}
// See <reloc.h>.
uint32_t FixupOffset = Layout.getFragmentOffset(Fragment)+Fixup.getOffset();
- uint32_t Value = 0;
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;
} else {
// Check whether we need an external or internal relocation.
if (doesSymbolRequireExternRelocation(SD)) {
// undefined. This occurs with weak definitions, for example.
if (!SD->Symbol->isUndefined())
FixedValue -= Layout.getSymbolAddress(SD);
- Value = 0;
} else {
// The index is the section ordinal (1-based).
Index = SD->getFragment()->getParent()->getOrdinal() + 1;
- Value = Layout.getSymbolAddress(SD);
}
Type = RIT_Vanilla;
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);
}
+
+ bool IsFixupFullyResolved(const MCAssembler &Asm,
+ const MCValue Target,
+ bool IsPCRel,
+ const MCFragment *DF) const {
+ // If we are using scattered symbols, determine whether this value is
+ // actually resolved; scattering may cause atoms to move.
+ if (Asm.getBackend().hasScatteredSymbols()) {
+ if (Asm.getBackend().hasReliableSymbolDifference()) {
+ // 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);
+ } else {
+ const MCSection *BaseSection = 0;
+ if (IsPCRel)
+ BaseSection = &DF->getParent()->getSection();
+
+ return isScatteredFixupFullyResolvedSimple(Asm, Target, BaseSection);
+ }
+ }
+ return true;
+ }
+
void WriteObject(const MCAssembler &Asm, const MCAsmLayout &Layout) {
unsigned NumSections = Asm.size();
MachObjectWriter::MachObjectWriter(raw_ostream &OS,
bool Is64Bit,
+ uint32_t CPUType,
+ uint32_t CPUSubtype,
bool IsLittleEndian)
: MCObjectWriter(OS, IsLittleEndian)
{
- Impl = new MachObjectWriterImpl(this, Is64Bit);
+ Impl = new MachObjectWriterImpl(this, Is64Bit, CPUType, CPUSubtype);
}
MachObjectWriter::~MachObjectWriter() {
Target, FixedValue);
}
-void MachObjectWriter::WriteObject(const MCAssembler &Asm,
+bool MachObjectWriter::IsFixupFullyResolved(const MCAssembler &Asm,
+ const MCValue Target,
+ bool IsPCRel,
+ const MCFragment *DF) const {
+ return ((MachObjectWriterImpl*) Impl)->IsFixupFullyResolved(Asm, Target,
+ IsPCRel, DF);
+}
+
+void MachObjectWriter::WriteObject(MCAssembler &Asm,
const MCAsmLayout &Layout) {
((MachObjectWriterImpl*) Impl)->WriteObject(Asm, Layout);
}