diff options
-rw-r--r-- | include/llvm/MC/MCMachObjectWriter.h | 192 | ||||
-rw-r--r-- | lib/MC/MachObjectWriter.cpp | 1004 | ||||
-rw-r--r-- | lib/Target/ARM/ARMMachObjectWriter.cpp | 357 | ||||
-rw-r--r-- | lib/Target/PowerPC/PPCAsmBackend.cpp | 6 | ||||
-rw-r--r-- | lib/Target/X86/X86MachObjectWriter.cpp | 522 |
5 files changed, 1096 insertions, 985 deletions
diff --git a/include/llvm/MC/MCMachObjectWriter.h b/include/llvm/MC/MCMachObjectWriter.h index ec51031d0b..9bb598f546 100644 --- a/include/llvm/MC/MCMachObjectWriter.h +++ b/include/llvm/MC/MCMachObjectWriter.h @@ -10,11 +10,20 @@ #ifndef LLVM_MC_MCMACHOBJECTWRITER_H #define LLVM_MC_MCMACHOBJECTWRITER_H +#include "llvm/ADT/DenseMap.h" +#include "llvm/ADT/OwningPtr.h" +#include "llvm/ADT/SmallString.h" +#include "llvm/MC/MCExpr.h" #include "llvm/MC/MCObjectWriter.h" +#include "llvm/Object/MachOFormat.h" #include "llvm/Support/DataTypes.h" +#include <vector> namespace llvm { +class MCSectionData; +class MachObjectWriter; + class MCMachObjectTargetWriter { const unsigned Is64Bit : 1; const uint32_t CPUType; @@ -48,8 +57,191 @@ public: } /// @} + + /// @name API + /// @{ + + virtual void RecordRelocation(MachObjectWriter *Writer, + const MCAssembler &Asm, + const MCAsmLayout &Layout, + const MCFragment *Fragment, + const MCFixup &Fixup, + MCValue Target, + uint64_t &FixedValue) = 0; + + /// @} }; +class MachObjectWriter : public MCObjectWriter { + /// MachSymbolData - Helper struct for containing some precomputed information + /// on symbols. + struct MachSymbolData { + MCSymbolData *SymbolData; + uint64_t StringIndex; + uint8_t SectionIndex; + + // Support lexicographic sorting. + bool operator<(const MachSymbolData &RHS) const; + }; + + /// The target specific Mach-O writer instance. + llvm::OwningPtr<MCMachObjectTargetWriter> TargetObjectWriter; + + /// @name Relocation Data + /// @{ + + llvm::DenseMap<const MCSectionData*, + std::vector<object::macho::RelocationEntry> > Relocations; + llvm::DenseMap<const MCSectionData*, unsigned> IndirectSymBase; + + /// @} + /// @name Symbol Table Data + /// @{ + + SmallString<256> StringTable; + std::vector<MachSymbolData> LocalSymbolData; + std::vector<MachSymbolData> ExternalSymbolData; + std::vector<MachSymbolData> UndefinedSymbolData; + + /// @} + +public: + MachObjectWriter(MCMachObjectTargetWriter *MOTW, raw_ostream &_OS, + bool _IsLittleEndian) + : MCObjectWriter(_OS, _IsLittleEndian), TargetObjectWriter(MOTW) { + } + + /// @name Utility Methods + /// @{ + + bool isFixupKindPCRel(const MCAssembler &Asm, unsigned Kind); + + SectionAddrMap SectionAddress; + + SectionAddrMap &getSectionAddressMap() { return SectionAddress; } + + uint64_t getSectionAddress(const MCSectionData* SD) const { + return SectionAddress.lookup(SD); + } + uint64_t getSymbolAddress(const MCSymbolData* SD, + const MCAsmLayout &Layout) const; + + uint64_t getFragmentAddress(const MCFragment *Fragment, + const MCAsmLayout &Layout) const; + + uint64_t getPaddingSize(const MCSectionData *SD, + const MCAsmLayout &Layout) const; + + bool doesSymbolRequireExternRelocation(const MCSymbolData *SD); + + /// @} + + /// @name Target Writer Proxy Accessors + /// @{ + + bool is64Bit() const { return TargetObjectWriter->is64Bit(); } + bool isARM() const { + uint32_t CPUType = TargetObjectWriter->getCPUType() & + ~object::mach::CTFM_ArchMask; + return CPUType == object::mach::CTM_ARM; + } + + /// @} + + void WriteHeader(unsigned NumLoadCommands, unsigned LoadCommandsSize, + bool SubsectionsViaSymbols); + + /// WriteSegmentLoadCommand - Write a segment load command. + /// + /// \arg NumSections - The number of sections in this segment. + /// \arg SectionDataSize - The total size of the sections. + void WriteSegmentLoadCommand(unsigned NumSections, + uint64_t VMSize, + uint64_t SectionDataStartOffset, + uint64_t SectionDataSize); + + void WriteSection(const MCAssembler &Asm, const MCAsmLayout &Layout, + const MCSectionData &SD, uint64_t FileOffset, + uint64_t RelocationsStart, unsigned NumRelocations); + + void WriteSymtabLoadCommand(uint32_t SymbolOffset, uint32_t NumSymbols, + uint32_t StringTableOffset, + uint32_t StringTableSize); + + void WriteDysymtabLoadCommand(uint32_t FirstLocalSymbol, + uint32_t NumLocalSymbols, + uint32_t FirstExternalSymbol, + uint32_t NumExternalSymbols, + uint32_t FirstUndefinedSymbol, + uint32_t NumUndefinedSymbols, + uint32_t IndirectSymbolOffset, + uint32_t NumIndirectSymbols); + + void WriteNlist(MachSymbolData &MSD, const MCAsmLayout &Layout); + + // FIXME: We really need to improve the relocation validation. Basically, we + // want to implement a separate computation which evaluates the relocation + // entry as the linker would, and verifies that the resultant fixup value is + // exactly what the encoder wanted. This will catch several classes of + // problems: + // + // - Relocation entry bugs, the two algorithms are unlikely to have the same + // exact bug. + // + // - Relaxation issues, where we forget to relax something. + // + // - Input errors, where something cannot be correctly encoded. 'as' allows + // these through in many cases. + + void addRelocation(const MCSectionData *SD, + object::macho::RelocationEntry &MRE) { + Relocations[SD].push_back(MRE); + } + + void RecordScatteredRelocation(const MCAssembler &Asm, + const MCAsmLayout &Layout, + const MCFragment *Fragment, + const MCFixup &Fixup, MCValue Target, + unsigned Log2Size, + uint64_t &FixedValue); + + void RecordTLVPRelocation(const MCAssembler &Asm, + const MCAsmLayout &Layout, + const MCFragment *Fragment, + const MCFixup &Fixup, MCValue Target, + uint64_t &FixedValue); + + void RecordRelocation(const MCAssembler &Asm, const MCAsmLayout &Layout, + const MCFragment *Fragment, const MCFixup &Fixup, + MCValue Target, uint64_t &FixedValue); + + void BindIndirectSymbols(MCAssembler &Asm); + + /// 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 ComputeSymbolTable(MCAssembler &Asm, SmallString<256> &StringTable, + std::vector<MachSymbolData> &LocalSymbolData, + std::vector<MachSymbolData> &ExternalSymbolData, + std::vector<MachSymbolData> &UndefinedSymbolData); + + void computeSectionAddresses(const MCAssembler &Asm, + const MCAsmLayout &Layout); + + void ExecutePostLayoutBinding(MCAssembler &Asm, const MCAsmLayout &Layout); + + virtual bool IsSymbolRefDifferenceFullyResolvedImpl(const MCAssembler &Asm, + const MCSymbolData &DataA, + const MCFragment &FB, + bool InSet, + bool IsPCRel) const; + + void WriteObject(MCAssembler &Asm, const MCAsmLayout &Layout); +}; + + /// \brief Construct a new Mach-O writer instance. /// /// This routine takes ownership of the target writer subclass. diff --git a/lib/MC/MachObjectWriter.cpp b/lib/MC/MachObjectWriter.cpp index e11b9df82b..69efe231ad 100644 --- a/lib/MC/MachObjectWriter.cpp +++ b/lib/MC/MachObjectWriter.cpp @@ -23,34 +23,12 @@ #include "llvm/Support/ErrorHandling.h" #include "llvm/Target/TargetAsmBackend.h" -// FIXME: Gross. -#include "../Target/ARM/ARMFixupKinds.h" -#include "../Target/X86/X86FixupKinds.h" - #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 FK_PCRel_1: - case FK_Data_1: return 0; - case FK_PCRel_2: - case FK_Data_2: return 1; - case FK_PCRel_4: - // FIXME: Remove these!!! - 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; - } -} - -static bool doesSymbolRequireExternRelocation(MCSymbolData *SD) { +bool MachObjectWriter:: +doesSymbolRequireExternRelocation(const MCSymbolData *SD) { // Undefined symbols are always extern. if (SD->Symbol->isUndefined()) return true; @@ -64,207 +42,24 @@ static bool doesSymbolRequireExternRelocation(MCSymbolData *SD) { return false; } -namespace { - -class MachObjectWriter : public MCObjectWriter { - /// MachSymbolData - Helper struct for containing some precomputed information - /// on symbols. - struct MachSymbolData { - MCSymbolData *SymbolData; - uint64_t StringIndex; - uint8_t SectionIndex; - - // Support lexicographic sorting. - bool operator<(const MachSymbolData &RHS) const { - return SymbolData->getSymbol().getName() < - RHS.SymbolData->getSymbol().getName(); - } - }; - - /// The target specific Mach-O writer instance. - llvm::OwningPtr<MCMachObjectTargetWriter> TargetObjectWriter; - - /// @name Relocation Data - /// @{ - - llvm::DenseMap<const MCSectionData*, - std::vector<macho::RelocationEntry> > Relocations; - llvm::DenseMap<const MCSectionData*, unsigned> IndirectSymBase; - - /// @} - /// @name Symbol Table Data - /// @{ - - SmallString<256> StringTable; - std::vector<MachSymbolData> LocalSymbolData; - std::vector<MachSymbolData> ExternalSymbolData; - std::vector<MachSymbolData> UndefinedSymbolData; - - /// @} - -private: - /// @name Utility Methods - /// @{ - - bool isFixupKindPCRel(const MCAssembler &Asm, unsigned Kind) { - const MCFixupKindInfo &FKI = Asm.getBackend().getFixupKindInfo( - (MCFixupKind) Kind); - - return FKI.Flags & MCFixupKindInfo::FKF_IsPCRel; - } - - /// @} - - SectionAddrMap SectionAddress; - uint64_t getSectionAddress(const MCSectionData* SD) const { - return SectionAddress.lookup(SD); - } - uint64_t getSymbolAddress(const MCSymbolData* SD, - const MCAsmLayout &Layout) const; - - uint64_t getFragmentAddress(const MCFragment *Fragment, - const MCAsmLayout &Layout) const { - return getSectionAddress(Fragment->getParent()) + - Layout.getFragmentOffset(Fragment); - } - - uint64_t getPaddingSize(const MCSectionData *SD, - const MCAsmLayout &Layout) const; -public: - MachObjectWriter(MCMachObjectTargetWriter *MOTW, raw_ostream &_OS, - bool _IsLittleEndian) - : MCObjectWriter(_OS, _IsLittleEndian), TargetObjectWriter(MOTW) { - } - - /// @name Target Writer Proxy Accessors - /// @{ +bool MachObjectWriter:: +MachSymbolData::operator<(const MachSymbolData &RHS) const { + return SymbolData->getSymbol().getName() < + RHS.SymbolData->getSymbol().getName(); +} - bool is64Bit() const { return TargetObjectWriter->is64Bit(); } - bool isARM() const { - uint32_t CPUType = TargetObjectWriter->getCPUType() & ~mach::CTFM_ArchMask; - return CPUType == mach::CTM_ARM; - } +bool MachObjectWriter::isFixupKindPCRel(const MCAssembler &Asm, unsigned Kind) { + const MCFixupKindInfo &FKI = Asm.getBackend().getFixupKindInfo( + (MCFixupKind) Kind); - /// @} - - void WriteHeader(unsigned NumLoadCommands, unsigned LoadCommandsSize, - bool SubsectionsViaSymbols); - - /// WriteSegmentLoadCommand - Write a segment load command. - /// - /// \arg NumSections - The number of sections in this segment. - /// \arg SectionDataSize - The total size of the sections. - void WriteSegmentLoadCommand(unsigned NumSections, - uint64_t VMSize, - uint64_t SectionDataStartOffset, - uint64_t SectionDataSize); - - void WriteSection(const MCAssembler &Asm, const MCAsmLayout &Layout, - const MCSectionData &SD, uint64_t FileOffset, - uint64_t RelocationsStart, unsigned NumRelocations); - - void WriteSymtabLoadCommand(uint32_t SymbolOffset, uint32_t NumSymbols, - uint32_t StringTableOffset, - uint32_t StringTableSize); - - void WriteDysymtabLoadCommand(uint32_t FirstLocalSymbol, - uint32_t NumLocalSymbols, - uint32_t FirstExternalSymbol, - uint32_t NumExternalSymbols, - uint32_t FirstUndefinedSymbol, - uint32_t NumUndefinedSymbols, - uint32_t IndirectSymbolOffset, - uint32_t NumIndirectSymbols); - - void WriteNlist(MachSymbolData &MSD, const MCAsmLayout &Layout); - - // FIXME: We really need to improve the relocation validation. Basically, we - // want to implement a separate computation which evaluates the relocation - // entry as the linker would, and verifies that the resultant fixup value is - // exactly what the encoder wanted. This will catch several classes of - // problems: - // - // - Relocation entry bugs, the two algorithms are unlikely to have the same - // exact bug. - // - // - Relaxation issues, where we forget to relax something. - // - // - Input errors, where something cannot be correctly encoded. 'as' allows - // these through in many cases. + return FKI.Flags & MCFixupKindInfo::FKF_IsPCRel; +} - static bool isFixupKindRIPRel(unsigned Kind) { - return Kind == X86::reloc_riprel_4byte || - Kind == X86::reloc_riprel_4byte_movq_load; - } - void RecordX86_64Relocation(const MCAssembler &Asm, const MCAsmLayout &Layout, - const MCFragment *Fragment, - const MCFixup &Fixup, MCValue Target, - uint64_t &FixedValue); - - void RecordScatteredRelocation(const MCAssembler &Asm, - const MCAsmLayout &Layout, - const MCFragment *Fragment, - const MCFixup &Fixup, MCValue Target, - unsigned Log2Size, - uint64_t &FixedValue); - - void RecordARMScatteredRelocation(const MCAssembler &Asm, - const MCAsmLayout &Layout, - const MCFragment *Fragment, - const MCFixup &Fixup, MCValue Target, - unsigned Log2Size, - uint64_t &FixedValue); - - void RecordARMMovwMovtRelocation(const MCAssembler &Asm, - const MCAsmLayout &Layout, - const MCFragment *Fragment, - const MCFixup &Fixup, MCValue Target, - uint64_t &FixedValue); - - void RecordTLVPRelocation(const MCAssembler &Asm, - const MCAsmLayout &Layout, - const MCFragment *Fragment, - const MCFixup &Fixup, MCValue Target, - uint64_t &FixedValue); - - static bool getARMFixupKindMachOInfo(unsigned Kind, unsigned &RelocType, - unsigned &Log2Size); - - void RecordARMRelocation(const MCAssembler &Asm, const MCAsmLayout &Layout, - const MCFragment *Fragment, const MCFixup &Fixup, - MCValue Target, uint64_t &FixedValue); - - void RecordRelocation(const MCAssembler &Asm, const MCAsmLayout &Layout, - const MCFragment *Fragment, const MCFixup &Fixup, - MCValue Target, uint64_t &FixedValue); - - void BindIndirectSymbols(MCAssembler &Asm); - - /// 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 ComputeSymbolTable(MCAssembler &Asm, SmallString<256> &StringTable, - std::vector<MachSymbolData> &LocalSymbolData, - std::vector<MachSymbolData> &ExternalSymbolData, - std::vector<MachSymbolData> &UndefinedSymbolData); - - void computeSectionAddresses(const MCAssembler &Asm, - const MCAsmLayout &Layout); - - void ExecutePostLayoutBinding(MCAssembler &Asm, const MCAsmLayout &Layout); - - virtual bool IsSymbolRefDifferenceFullyResolvedImpl(const MCAssembler &Asm, - const MCSymbolData &DataA, - const MCFragment &FB, - bool InSet, - bool IsPCRel) const; - - void WriteObject(MCAssembler &Asm, const MCAsmLayout &Layout); -}; - -} // end anonymous namespace +uint64_t MachObjectWriter::getFragmentAddress(const MCFragment *Fragment, + const MCAsmLayout &Layout) const { + return getSectionAddress(Fragment->getParent()) + + Layout.getFragmentOffset(Fragment); +} uint64_t MachObjectWriter::getSymbolAddress(const MCSymbolData* SD, const MCAsmLayout &Layout) const { @@ -556,775 +351,14 @@ void MachObjectWriter::WriteNlist(MachSymbolData &MSD, Write32(Address); } -void MachObjectWriter::RecordX86_64Relocation(const MCAssembler &Asm, - const MCAsmLayout &Layout, - const MCFragment *Fragment, - const MCFixup &Fixup, - MCValue Target, - uint64_t &FixedValue) { - unsigned IsPCRel = isFixupKindPCRel(Asm, Fixup.getKind()); - unsigned IsRIPRel = isFixupKindRIPRel(Fixup.getKind()); - unsigned Log2Size = getFixupKindLog2Size(Fixup.getKind()); - - // See <reloc.h>. - uint32_t FixupOffset = - Layout.getFragmentOffset(Fragment) + Fixup.getOffset(); - uint32_t FixupAddress = - getFragmentAddress(Fragment, Layout) + Fixup.getOffset(); - int64_t Value = 0; - unsigned Index = 0; - unsigned IsExtern = 0; - unsigned Type = 0; - - Value = Target.getConstant(); - - if (IsPCRel) { - // Compensate for the relocation offset, Darwin x86_64 relocations only have - // the addend and appear to have attempted to define it to be the actual - // expression addend without the PCrel bias. However, instructions with data - // following the relocation are not accommodated for (see comment below - // regarding SIGNED{1,2,4}), so it isn't exactly that either. - Value += 1LL << Log2Size; - } - - if (Target.isAbsolute()) { // constant - // SymbolNum of 0 indicates the absolute section. - Type = macho::RIT_X86_64_Unsigned; - Index = 0; - - // FIXME: I believe this is broken, I don't think the linker can understand - // it. I think it would require a local relocation, but I'm not sure if that - // would work either. The official way to get an absolute PCrel relocation - // is to use an absolute symbol (which we don't support yet). - if (IsPCRel) { - IsExtern = 1; - 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(&A_SD); - - const MCSymbol *B = &Target.getSymB()->getSymbol(); - MCSymbolData &B_SD = Asm.getSymbolData(*B); - const MCSymbolData *B_Base = Asm.getAtom(&B_SD); - - // Neither symbol can be modified. - if (Target.getSymA()->getKind() != MCSymbolRefExpr::VK_None || - Target.getSymB()->getKind() != MCSymbolRefExpr::VK_None) - report_fatal_error("unsupported relocation of modified symbol"); - - // We don't support PCrel relocations of differences. Darwin 'as' doesn't - // implement most of these correctly. - if (IsPCRel) - report_fatal_error("unsupported pc-relative relocation of 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. 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 += getSymbolAddress(&A_SD, Layout) - - (A_Base == NULL ? 0 : getSymbolAddress(A_Base, Layout)); - Value -= getSymbolAddress(&B_SD, Layout) - - (B_Base == NULL ? 0 : getSymbolAddress(B_Base, Layout)); - - 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; - - macho::RelocationEntry MRE; - MRE.Word0 = FixupOffset; - MRE.Word1 = ((Index << 0) | - (IsPCRel << 24) | - (Log2Size << 25) | - (IsExtern << 27) | - (Type << 28)); - Relocations[Fragment->getParent()].push_back(MRE); - - 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(&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 preceding - // non-local symbol). - if (Base) { - Index = Base->getIndex(); - IsExtern = 1; - - // Add the local offset, if needed. - if (Base != &SD) - Value += Layout.getSymbolOffset(&SD) - Layout.getSymbolOffset(Base); - } else if (Symbol->isInSection() && !Symbol->isVariable()) { - // The index is the section ordinal (1-based). - Index = SD.getFragment()->getParent()->getOrdinal() + 1; - IsExtern = 0; - Value += getSymbolAddress(&SD, Layout); - - if (IsPCRel) - Value -= FixupAddress + (1 << Log2Size); - } else if (Symbol->isVariable()) { - const MCExpr *Value = Symbol->getVariableValue(); - int64_t Res; - bool isAbs = Value->EvaluateAsAbsolute(Res, Layout, SectionAddress); - if (isAbs) { - FixedValue = Res; - return; - } else { - report_fatal_error("unsupported relocation of variable '" + - Symbol->getName() + "'"); - } - } else { - report_fatal_error("unsupported relocation of undefined symbol '" + - Symbol->getName() + "'"); - } - - MCSymbolRefExpr::VariantKind Modifier = Target.getSymA()->getKind(); - if (IsPCRel) { - if (IsRIPRel) { - if (Modifier == MCSymbolRefExpr::VK_GOTPCREL) { - // 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.getKind()) == X86::reloc_riprel_4byte_movq_load) - Type = macho::RIT_X86_64_GOTLoad; - else - 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 = 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 = macho::RIT_X86_64_Branch; - } - } else { - if (Modifier == MCSymbolRefExpr::VK_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 = 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 = macho::RIT_X86_64_Unsigned; - } - } - - // x86_64 always writes custom values into the fixups. - FixedValue = Value; - - // struct relocation_info (8 bytes) - macho::RelocationEntry MRE; - MRE.Word0 = FixupOffset; - MRE.Word1 = ((Index << 0) | - (IsPCRel << 24) | - (Log2Size << 25) | - (IsExtern << 27) | - (Type << 28)); - Relocations[Fragment->getParent()].push_back(MRE); -} - -void MachObjectWriter::RecordScatteredRelocation(const MCAssembler &Asm, - const MCAsmLayout &Layout, - const MCFragment *Fragment, - const MCFixup &Fixup, - MCValue Target, - unsigned Log2Size, - uint64_t &FixedValue) { - uint32_t FixupOffset = Layout.getFragmentOffset(Fragment)+Fixup.getOffset(); - unsigned IsPCRel = isFixupKindPCRel(Asm, Fixup.getKind()); - unsigned Type = macho::RIT_Vanilla; - - // See <reloc.h>. - const MCSymbol *A = &Target.getSymA()->getSymbol(); - MCSymbolData *A_SD = &Asm.getSymbolData(*A); - - if (!A_SD->getFragment()) - report_fatal_error("symbol '" + A->getName() + - "' can not be undefined in a subtraction expression"); - - uint32_t Value = getSymbolAddress(A_SD, Layout); - uint64_t SecAddr = getSectionAddress(A_SD->getFragment()->getParent()); - FixedValue += SecAddr; - uint32_t Value2 = 0; - - if (const MCSymbolRefExpr *B = Target.getSymB()) { - MCSymbolData *B_SD = &Asm.getSymbolData(B->getSymbol()); - - if (!B_SD->getFragment()) - report_fatal_error("symbol '" + B->getSymbol().getName() + - "' can not be undefined in a subtraction expression"); - - // Select the appropriate difference relocation type. - // - // 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() ? (unsigned)macho::RIT_Difference : - (unsigned)macho::RIT_Generic_LocalDifference; - Value2 = getSymbolAddress(B_SD, Layout); - FixedValue -= getSectionAddress(B_SD->getFragment()->getParent()); - } - - // Relocations are written out in reverse order, so the PAIR comes first. - if (Type == macho::RIT_Difference || - Type == macho::RIT_Generic_LocalDifference) { - macho::RelocationEntry MRE; - MRE.Word0 = ((0 << 0) | - (macho::RIT_Pair << 24) | - (Log2Size << 28) | - (IsPCRel << 30) | - macho::RF_Scattered); - MRE.Word1 = Value2; - Relocations[Fragment->getParent()].push_back(MRE); - } - - 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 MachObjectWriter::RecordARMScatteredRelocation(const MCAssembler &Asm, - const MCAsmLayout &Layout, - const MCFragment *Fragment, - const MCFixup &Fixup, - MCValue Target, - unsigned Log2Size, - uint64_t &FixedValue) { - uint32_t FixupOffset = Layout.getFragmentOffset(Fragment)+Fixup.getOffset(); - unsigned IsPCRel = isFixupKindPCRel(Asm, Fixup.getKind()); - unsigned Type = macho::RIT_Vanilla; - - // See <reloc.h>. - const MCSymbol *A = &Target.getSymA()->getSymbol(); - MCSymbolData *A_SD = &Asm.getSymbolData(*A); - - if (!A_SD->getFragment()) - report_fatal_error("symbol '" + A->getName() + - "' can not be undefined in a subtraction expression"); - - uint32_t Value = getSymbolAddress(A_SD, Layout); - uint64_t SecAddr = getSectionAddress(A_SD->getFragment()->getParent()); - FixedValue += SecAddr; - uint32_t Value2 = 0; - - if (const MCSymbolRefExpr *B = Target.getSymB()) { - MCSymbolData *B_SD = &Asm.getSymbolData(B->getSymbol()); - - if (!B_SD->getFragment()) - report_fatal_error("symbol '" + B->getSymbol().getName() + - "' can not be undefined in a subtraction expression"); - - // Select the appropriate difference relocation type. - Type = macho::RIT_Difference; - Value2 = getSymbolAddress(B_SD, Layout); - FixedValue -= getSectionAddress(B_SD->getFragment()->getParent()); - } - - // Relocations are written out in reverse order, so the PAIR comes first. - if (Type == macho::RIT_Difference || - Type == macho::RIT_Generic_LocalDifference) { - macho::RelocationEntry MRE; - MRE.Word0 = ((0 << 0) | - (macho::RIT_Pair << 24) | - (Log2Size << 28) | - (IsPCRel << 30) | - macho::RF_Scattered); - MRE.Word1 = Value2; - Relocations[Fragment->getParent()].push_back(MRE); - } - - 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 MachObjectWriter::RecordARMMovwMovtRelocation(const MCAssembler &Asm, - const MCAsmLayout &Layout, - const MCFragment *Fragment, - const MCFixup &Fixup, - MCValue Target, - uint64_t &FixedValue) { - uint32_t FixupOffset = Layout.getFragmentOffset(Fragment)+Fixup.getOffset(); - unsigned IsPCRel = isFixupKindPCRel(Asm, Fixup.getKind()); - unsigned Type = macho::RIT_ARM_Half; - - // See <reloc.h>. - const MCSymbol *A = &Target.getSymA()->getSymbol(); - MCSymbolData *A_SD = &Asm.getSymbolData(*A); - - if (!A_SD->getFragment()) - report_fatal_error("symbol '" + A->getName() + - "' can not be undefined in a subtraction expression"); - - uint32_t Value = getSymbolAddress(A_SD, Layout); - uint32_t Value2 = 0; - uint64_t SecAddr = getSectionAddress(A_SD->getFragment()->getParent()); - FixedValue += SecAddr; - - if (const MCSymbolRefExpr *B = Target.getSymB()) { - MCSymbolData *B_SD = &Asm.getSymbolData(B->getSymbol()); - - if (!B_SD->getFragment()) - report_fatal_error("symbol '" + B->getSymbol().getName() + - "' can not be undefined in a subtraction expression"); - - // Select the appropriate difference relocation type. - Type = macho::RIT_ARM_HalfDifference; - Value2 = getSymbolAddress(B_SD, Layout); - FixedValue -= getSectionAddress(B_SD->getFragment()->getParent()); - } - - // Relocations are written out in reverse order, so the PAIR comes first. - // ARM_RELOC_HALF and ARM_RELOC_HALF_SECTDIFF abuse the r_length field: - // - // For these two r_type relocations they always have a pair following them and - // the r_length bits are used differently. The encoding of the r_length is as |