diff options
Diffstat (limited to 'lib/MC')
-rw-r--r-- | lib/MC/CMakeLists.txt | 3 | ||||
-rw-r--r-- | lib/MC/MCAsmInfo.cpp | 1 | ||||
-rw-r--r-- | lib/MC/MCAsmInfoDarwin.cpp | 1 | ||||
-rw-r--r-- | lib/MC/MCAsmStreamer.cpp | 59 | ||||
-rw-r--r-- | lib/MC/MCAssembler.cpp | 673 | ||||
-rw-r--r-- | lib/MC/MCContext.cpp | 53 | ||||
-rw-r--r-- | lib/MC/MCExpr.cpp | 14 | ||||
-rw-r--r-- | lib/MC/MCInst.cpp | 2 | ||||
-rw-r--r-- | lib/MC/MCLabel.cpp | 21 | ||||
-rw-r--r-- | lib/MC/MCLoggingStreamer.cpp | 208 | ||||
-rw-r--r-- | lib/MC/MCMachOStreamer.cpp | 226 | ||||
-rw-r--r-- | lib/MC/MCNullStreamer.cpp | 9 | ||||
-rw-r--r-- | lib/MC/MCParser/AsmLexer.cpp | 12 | ||||
-rw-r--r-- | lib/MC/MCParser/AsmParser.cpp | 251 | ||||
-rw-r--r-- | lib/MC/MCSection.cpp | 27 | ||||
-rw-r--r-- | lib/MC/MCSectionCOFF.cpp | 76 | ||||
-rw-r--r-- | lib/MC/MCSectionMachO.cpp | 11 | ||||
-rw-r--r-- | lib/MC/MCStreamer.cpp | 2 | ||||
-rw-r--r-- | lib/MC/MCSymbol.cpp | 12 | ||||
-rw-r--r-- | lib/MC/MachObjectWriter.cpp | 201 |
20 files changed, 1296 insertions, 566 deletions
diff --git a/lib/MC/CMakeLists.txt b/lib/MC/CMakeLists.txt index dba0e14..5e8a3b6 100644 --- a/lib/MC/CMakeLists.txt +++ b/lib/MC/CMakeLists.txt @@ -10,10 +10,13 @@ add_llvm_library(LLVMMC MCExpr.cpp MCInst.cpp MCInstPrinter.cpp + MCLabel.cpp + MCLoggingStreamer.cpp MCMachOStreamer.cpp MCNullStreamer.cpp MCObjectWriter.cpp MCSection.cpp + MCSectionCOFF.cpp MCSectionELF.cpp MCSectionMachO.cpp MCStreamer.cpp diff --git a/lib/MC/MCAsmInfo.cpp b/lib/MC/MCAsmInfo.cpp index 2b23994..a275be2 100644 --- a/lib/MC/MCAsmInfo.cpp +++ b/lib/MC/MCAsmInfo.cpp @@ -21,6 +21,7 @@ using namespace llvm; MCAsmInfo::MCAsmInfo() { HasSubsectionsViaSymbols = false; HasMachoZeroFillDirective = false; + HasMachoTBSSDirective = false; HasStaticCtorDtorReferenceInStaticMode = false; MaxInstLength = 4; PCSymbol = "$"; diff --git a/lib/MC/MCAsmInfoDarwin.cpp b/lib/MC/MCAsmInfoDarwin.cpp index 3c31caa..0bd3b2d 100644 --- a/lib/MC/MCAsmInfoDarwin.cpp +++ b/lib/MC/MCAsmInfoDarwin.cpp @@ -35,6 +35,7 @@ MCAsmInfoDarwin::MCAsmInfoDarwin() { WeakRefDirective = "\t.weak_reference "; ZeroDirective = "\t.space\t"; // ".space N" emits N zeros. HasMachoZeroFillDirective = true; // Uses .zerofill + HasMachoTBSSDirective = true; // Uses .tbss HasStaticCtorDtorReferenceInStaticMode = true; HiddenVisibilityAttr = MCSA_PrivateExtern; diff --git a/lib/MC/MCAsmStreamer.cpp b/lib/MC/MCAsmStreamer.cpp index 2c7e1c4..57b2bcc 100644 --- a/lib/MC/MCAsmStreamer.cpp +++ b/lib/MC/MCAsmStreamer.cpp @@ -109,7 +109,10 @@ public: virtual void EmitSymbolAttribute(MCSymbol *Symbol, MCSymbolAttr Attribute); virtual void EmitSymbolDesc(MCSymbol *Symbol, unsigned DescValue); - + virtual void BeginCOFFSymbolDef(const MCSymbol *Symbol); + virtual void EmitCOFFSymbolStorageClass(int StorageClass); + virtual void EmitCOFFSymbolType(int Type); + virtual void EndCOFFSymbolDef(); virtual void EmitELFSize(MCSymbol *Symbol, const MCExpr *Value); virtual void EmitCommonSymbol(MCSymbol *Symbol, uint64_t Size, unsigned ByteAlignment); @@ -123,6 +126,9 @@ public: virtual void EmitZerofill(const MCSection *Section, MCSymbol *Symbol = 0, unsigned Size = 0, unsigned ByteAlignment = 0); + virtual void EmitTBSSSymbol (const MCSection *Section, MCSymbol *Symbol, + uint64_t Size, unsigned ByteAlignment = 0); + virtual void EmitBytes(StringRef Data, unsigned AddrSpace); virtual void EmitValue(const MCExpr *Value, unsigned Size,unsigned AddrSpace); @@ -218,6 +224,7 @@ void MCAsmStreamer::SwitchSection(const MCSection *Section) { void MCAsmStreamer::EmitLabel(MCSymbol *Symbol) { assert(Symbol->isUndefined() && "Cannot define a symbol twice!"); + assert(!Symbol->isVariable() && "Cannot emit a variable symbol!"); assert(CurSection && "Cannot emit before setting section!"); OS << *Symbol << ":"; @@ -234,16 +241,11 @@ void MCAsmStreamer::EmitAssemblerFlag(MCAssemblerFlag Flag) { } void MCAsmStreamer::EmitAssignment(MCSymbol *Symbol, const MCExpr *Value) { - // Only absolute symbols can be redefined. - assert((Symbol->isUndefined() || Symbol->isAbsolute()) && - "Cannot define a symbol twice!"); - OS << *Symbol << " = " << *Value; EmitEOL(); // FIXME: Lift context changes into super class. - // FIXME: Set associated section. - Symbol->setValue(Value); + Symbol->setVariableValue(Value); } void MCAsmStreamer::EmitSymbolAttribute(MCSymbol *Symbol, @@ -297,6 +299,26 @@ void MCAsmStreamer::EmitSymbolDesc(MCSymbol *Symbol, unsigned DescValue) { EmitEOL(); } +void MCAsmStreamer::BeginCOFFSymbolDef(const MCSymbol *Symbol) { + OS << "\t.def\t " << *Symbol << ';'; + EmitEOL(); +} + +void MCAsmStreamer::EmitCOFFSymbolStorageClass (int StorageClass) { + OS << "\t.scl\t" << StorageClass << ';'; + EmitEOL(); +} + +void MCAsmStreamer::EmitCOFFSymbolType (int Type) { + OS << "\t.type\t" << Type << ';'; + EmitEOL(); +} + +void MCAsmStreamer::EndCOFFSymbolDef() { + OS << "\t.endef"; + EmitEOL(); +} + void MCAsmStreamer::EmitELFSize(MCSymbol *Symbol, const MCExpr *Value) { assert(MAI.hasDotTypeDotSizeDirective()); OS << "\t.size\t" << *Symbol << ", " << *Value << '\n'; @@ -341,6 +363,23 @@ void MCAsmStreamer::EmitZerofill(const MCSection *Section, MCSymbol *Symbol, EmitEOL(); } +// .tbss sym, size, align +// This depends that the symbol has already been mangled from the original, +// e.g. _a. +void MCAsmStreamer::EmitTBSSSymbol(const MCSection *Section, MCSymbol *Symbol, + uint64_t Size, unsigned ByteAlignment) { + assert(Symbol != NULL && "Symbol shouldn't be NULL!"); + // Instead of using the Section we'll just use the shortcut. + // This is a mach-o specific directive and section. + OS << ".tbss " << *Symbol << ", " << Size; + + // Output align if we have it. We default to 1 so don't bother printing + // that. + if (ByteAlignment > 1) OS << ", " << Log2_32(ByteAlignment); + + EmitEOL(); +} + static inline char toOctal(int X) { return (X&7)+'0'; } static void PrintQuotedString(StringRef Data, raw_ostream &OS) { @@ -630,9 +669,11 @@ void MCAsmStreamer::EmitInstruction(const MCInst &Inst) { AddEncodingComment(Inst); // Show the MCInst if enabled. - if (ShowInst) + if (ShowInst) { Inst.dump_pretty(GetCommentOS(), &MAI, InstPrinter.get(), "\n "); - + GetCommentOS() << "\n"; + } + // If we have an AsmPrinter, use that to print, otherwise print the MCInst. if (InstPrinter) InstPrinter->printInst(&Inst, OS); diff --git a/lib/MC/MCAssembler.cpp b/lib/MC/MCAssembler.cpp index 69afcc8..5936656 100644 --- a/lib/MC/MCAssembler.cpp +++ b/lib/MC/MCAssembler.cpp @@ -47,93 +47,131 @@ STATISTIC(SectionLayouts, "Number of section layouts"); /* *** */ -void MCAsmLayout::UpdateForSlide(MCFragment *F, int SlideAmount) { - // We shouldn't have to do anything special to support negative slides, and it - // is a perfectly valid thing to do as long as other parts of the system are - // can guarantee convergence. - assert(SlideAmount >= 0 && "Negative slides not yet supported"); +MCAsmLayout::MCAsmLayout(MCAssembler &Asm) + : Assembler(Asm), LastValidFragment(0) + { + // Compute the section layout order. Virtual sections must go last. + for (MCAssembler::iterator it = Asm.begin(), ie = Asm.end(); it != ie; ++it) + if (!Asm.getBackend().isVirtualSection(it->getSection())) + SectionOrder.push_back(&*it); + for (MCAssembler::iterator it = Asm.begin(), ie = Asm.end(); it != ie; ++it) + if (Asm.getBackend().isVirtualSection(it->getSection())) + SectionOrder.push_back(&*it); +} - // Update the layout by simply recomputing the layout for the entire - // file. This is trivially correct, but very slow. - // - // FIXME-PERF: This is O(N^2), but will be eliminated once we get smarter. +bool MCAsmLayout::isSectionUpToDate(const MCSectionData *SD) const { + // The first section is always up-to-date. + unsigned Index = SD->getLayoutOrder(); + if (!Index) + return true; - // Layout the concrete sections and fragments. - MCAssembler &Asm = getAssembler(); - uint64_t Address = 0; - for (MCAssembler::iterator it = Asm.begin(), ie = Asm.end(); it != ie; ++it) { - // Skip virtual sections. - if (Asm.getBackend().isVirtualSection(it->getSection())) - continue; + // Otherwise, sections are always implicitly computed when the preceeding + // fragment is layed out. + const MCSectionData *Prev = getSectionOrder()[Index - 1]; + return isFragmentUpToDate(&(Prev->getFragmentList().back())); +} + +bool MCAsmLayout::isFragmentUpToDate(const MCFragment *F) const { + return (LastValidFragment && + F->getLayoutOrder() <= LastValidFragment->getLayoutOrder()); +} - // Layout the section fragments and its size. - Address = Asm.LayoutSection(*it, *this, Address); +void MCAsmLayout::UpdateForSlide(MCFragment *F, int SlideAmount) { + // If this fragment wasn't already up-to-date, we don't need to do anything. + if (!isFragmentUpToDate(F)) + return; + + // Otherwise, reset the last valid fragment to the predecessor of the + // invalidated fragment. + LastValidFragment = F->getPrevNode(); + if (!LastValidFragment) { + unsigned Index = F->getParent()->getLayoutOrder(); + if (Index != 0) { + MCSectionData *Prev = getSectionOrder()[Index - 1]; + LastValidFragment = &(Prev->getFragmentList().back()); + } } +} - // Layout the virtual sections. - for (MCAssembler::iterator it = Asm.begin(), ie = Asm.end(); it != ie; ++it) { - if (!Asm.getBackend().isVirtualSection(it->getSection())) - continue; +void MCAsmLayout::EnsureValid(const MCFragment *F) const { + // Advance the layout position until the fragment is up-to-date. + while (!isFragmentUpToDate(F)) { + // Advance to the next fragment. + MCFragment *Cur = LastValidFragment; + if (Cur) + Cur = Cur->getNextNode(); + if (!Cur) { + unsigned NextIndex = 0; + if (LastValidFragment) + NextIndex = LastValidFragment->getParent()->getLayoutOrder() + 1; + Cur = SectionOrder[NextIndex]->begin(); + } - // Layout the section fragments and its size. - Address = Asm.LayoutSection(*it, *this, Address); + const_cast<MCAsmLayout*>(this)->LayoutFragment(Cur); } } +void MCAsmLayout::FragmentReplaced(MCFragment *Src, MCFragment *Dst) { + if (LastValidFragment == Src) + LastValidFragment = Dst; + + Dst->Offset = Src->Offset; + Dst->EffectiveSize = Src->EffectiveSize; +} + uint64_t MCAsmLayout::getFragmentAddress(const MCFragment *F) const { assert(F->getParent() && "Missing section()!"); return getSectionAddress(F->getParent()) + getFragmentOffset(F); } uint64_t MCAsmLayout::getFragmentEffectiveSize(const MCFragment *F) const { + EnsureValid(F); assert(F->EffectiveSize != ~UINT64_C(0) && "Address not set!"); return F->EffectiveSize; } -void MCAsmLayout::setFragmentEffectiveSize(MCFragment *F, uint64_t Value) { - F->EffectiveSize = Value; -} - uint64_t MCAsmLayout::getFragmentOffset(const MCFragment *F) const { + EnsureValid(F); assert(F->Offset != ~UINT64_C(0) && "Address not set!"); return F->Offset; } -void MCAsmLayout::setFragmentOffset(MCFragment *F, uint64_t Value) { - F->Offset = Value; -} - uint64_t MCAsmLayout::getSymbolAddress(const MCSymbolData *SD) const { assert(SD->getFragment() && "Invalid getAddress() on undefined symbol!"); return getFragmentAddress(SD->getFragment()) + SD->getOffset(); } uint64_t MCAsmLayout::getSectionAddress(const MCSectionData *SD) const { + EnsureValid(SD->begin()); assert(SD->Address != ~UINT64_C(0) && "Address not set!"); return SD->Address; } -void MCAsmLayout::setSectionAddress(MCSectionData *SD, uint64_t Value) { - SD->Address = Value; -} - -uint64_t MCAsmLayout::getSectionSize(const MCSectionData *SD) const { - assert(SD->Size != ~UINT64_C(0) && "File size not set!"); - return SD->Size; -} -void MCAsmLayout::setSectionSize(MCSectionData *SD, uint64_t Value) { - SD->Size = Value; +uint64_t MCAsmLayout::getSectionAddressSize(const MCSectionData *SD) const { + // The size is the last fragment's end offset. + const MCFragment &F = SD->getFragmentList().back(); + return getFragmentOffset(&F) + getFragmentEffectiveSize(&F); } uint64_t MCAsmLayout::getSectionFileSize(const MCSectionData *SD) const { - assert(SD->FileSize != ~UINT64_C(0) && "File size not set!"); - return SD->FileSize; -} -void MCAsmLayout::setSectionFileSize(MCSectionData *SD, uint64_t Value) { - SD->FileSize = Value; + // Virtual sections have no file size. + if (getAssembler().getBackend().isVirtualSection(SD->getSection())) + return 0; + + // Otherwise, the file size is the same as the address space size. + return getSectionAddressSize(SD); } - /// @} +uint64_t MCAsmLayout::getSectionSize(const MCSectionData *SD) const { + // The logical size is the address space size minus any tail padding. + uint64_t Size = getSectionAddressSize(SD); + const MCAlignFragment *AF = + dyn_cast<MCAlignFragment>(&(SD->getFragmentList().back())); + if (AF && AF->hasOnlyAlignAddress()) + Size -= getFragmentEffectiveSize(AF); + + return Size; +} /* *** */ @@ -141,17 +179,12 @@ MCFragment::MCFragment() : Kind(FragmentType(~0)) { } MCFragment::MCFragment(FragmentType _Kind, MCSectionData *_Parent) - : Kind(_Kind), - Parent(_Parent), - EffectiveSize(~UINT64_C(0)) + : Kind(_Kind), Parent(_Parent), Atom(0), EffectiveSize(~UINT64_C(0)) { if (Parent) Parent->getFragmentList().push_back(this); } -MCFragment::~MCFragment() { -} - /* *** */ MCSectionData::MCSectionData() : Section(0) {} @@ -160,8 +193,6 @@ MCSectionData::MCSectionData(const MCSection &_Section, MCAssembler *A) : Section(&_Section), Alignment(1), Address(~UINT64_C(0)), - Size(~UINT64_C(0)), - FileSize(~UINT64_C(0)), HasInstructions(false) { if (A) @@ -195,7 +226,7 @@ MCAssembler::~MCAssembler() { } static bool isScatteredFixupFullyResolvedSimple(const MCAssembler &Asm, - const MCAsmFixup &Fixup, + const MCFixup &Fixup, const MCValue Target, const MCSection *BaseSection) { // The effective fixup address is @@ -233,7 +264,7 @@ static bool isScatteredFixupFullyResolvedSimple(const MCAssembler &Asm, static bool isScatteredFixupFullyResolved(const MCAssembler &Asm, const MCAsmLayout &Layout, - const MCAsmFixup &Fixup, + const MCFixup &Fixup, const MCValue Target, const MCSymbolData *BaseSymbol) { // The effective fixup address is @@ -291,36 +322,6 @@ bool MCAssembler::isSymbolLinkerVisible(const MCSymbolData *SD) const { SD->getFragment()->getParent()->getSection()); } -// FIXME-PERF: This routine is really slow. -const MCSymbolData *MCAssembler::getAtomForAddress(const MCAsmLayout &Layout, - const MCSectionData *Section, - uint64_t Address) const { - const MCSymbolData *Best = 0; - uint64_t BestAddress = 0; - - for (MCAssembler::const_symbol_iterator it = symbol_begin(), - ie = symbol_end(); it != ie; ++it) { - // Ignore non-linker visible symbols. - if (!isSymbolLinkerVisible(it)) - continue; - - // Ignore symbols not in the same section. - if (!it->getFragment() || it->getFragment()->getParent() != Section) - continue; - - // Otherwise, find the closest symbol preceding this address (ties are - // resolved in favor of the last defined symbol). - uint64_t SymbolAddress = Layout.getSymbolAddress(it); - if (SymbolAddress <= Address && (!Best || SymbolAddress >= BestAddress)) { - Best = it; - BestAddress = SymbolAddress; - } - } - - return Best; -} - -// FIXME-PERF: This routine is really slow. const MCSymbolData *MCAssembler::getAtom(const MCAsmLayout &Layout, const MCSymbolData *SD) const { // Linker visible symbols define atoms. @@ -331,17 +332,22 @@ const MCSymbolData *MCAssembler::getAtom(const MCAsmLayout &Layout, if (!SD->getFragment()) return 0; - // Otherwise, search by address. - return getAtomForAddress(Layout, SD->getFragment()->getParent(), - Layout.getSymbolAddress(SD)); + // Non-linker visible symbols in sections which can't be atomized have no + // defining atom. + if (!getBackend().isSectionAtomizable( + SD->getFragment()->getParent()->getSection())) + return 0; + + // Otherwise, return the atom for the containing fragment. + return SD->getFragment()->getAtom(); } bool MCAssembler::EvaluateFixup(const MCAsmLayout &Layout, - const MCAsmFixup &Fixup, const MCFragment *DF, + const MCFixup &Fixup, const MCFragment *DF, MCValue &Target, uint64_t &Value) const { ++stats::EvaluateFixup; - if (!Fixup.Value->EvaluateAsRelocatable(Target, &Layout)) + if (!Fixup.getValue()->EvaluateAsRelocatable(Target, &Layout)) report_fatal_error("expected relocatable expression"); // FIXME: How do non-scattered symbols work in ELF? I presume the linker @@ -350,8 +356,8 @@ bool MCAssembler::EvaluateFixup(const MCAsmLayout &Layout, Value = Target.getConstant(); - bool IsPCRel = - Emitter.getFixupKindInfo(Fixup.Kind).Flags & MCFixupKindInfo::FKF_IsPCRel; + bool IsPCRel = Emitter.getFixupKindInfo( + Fixup.getKind()).Flags & MCFixupKindInfo::FKF_IsPCRel; bool IsResolved = true; if (const MCSymbolRefExpr *A = Target.getSymA()) { if (A->getSymbol().isDefined()) @@ -374,8 +380,7 @@ bool MCAssembler::EvaluateFixup(const MCAsmLayout &Layout, // symbol) that the fixup value is relative to. const MCSymbolData *BaseSymbol = 0; if (IsPCRel) { - BaseSymbol = getAtomForAddress( - Layout, DF->getParent(), Layout.getFragmentAddress(DF)+Fixup.Offset); + BaseSymbol = DF->getAtom(); if (!BaseSymbol) IsResolved = false; } @@ -394,117 +399,123 @@ bool MCAssembler::EvaluateFixup(const MCAsmLayout &Layout, } if (IsPCRel) - Value -= Layout.getFragmentAddress(DF) + Fixup.Offset; + Value -= Layout.getFragmentAddress(DF) + Fixup.getOffset(); return IsResolved; } -uint64_t MCAssembler::LayoutSection(MCSectionData &SD, - MCAsmLayout &Layout, - uint64_t StartAddress) { - bool IsVirtual = getBackend().isVirtualSection(SD.getSection()); - - ++stats::SectionLayouts; - - // Align this section if necessary by adding padding bytes to the previous - // section. It is safe to adjust this out-of-band, because no symbol or - // fragment is allowed to point past the end of the section at any time. - if (uint64_t Pad = OffsetToAlignment(StartAddress, SD.getAlignment())) { - // Unless this section is virtual (where we are allowed to adjust the offset - // freely), the padding goes in the previous section. - if (!IsVirtual) { - // Find the previous non-virtual section. - iterator it = &SD; - assert(it != begin() && "Invalid initial section address!"); - for (--it; getBackend().isVirtualSection(it->getSection()); --it) ; - Layout.setSectionFileSize(&*it, Layout.getSectionFileSize(&*it) + Pad); - } +uint64_t MCAssembler::ComputeFragmentSize(MCAsmLayout &Layout, + const MCFragment &F, + uint64_t SectionAddress, + uint64_t FragmentOffset) const { + switch (F.getKind()) { + case MCFragment::FT_Data: + return cast<MCDataFragment>(F).getContents().size(); + case MCFragment::FT_Fill: + return cast<MCFillFragment>(F).getSize(); + case MCFragment::FT_Inst: + return cast<MCInstFragment>(F).getInstSize(); - StartAddress += Pad; - } + case MCFragment::FT_Align: { + const MCAlignFragment &AF = cast<MCAlignFragment>(F); - // Set the aligned section address. - Layout.setSectionAddress(&SD, StartAddress); + assert((!AF.hasOnlyAlignAddress() || !AF.getNextNode()) && + "Invalid OnlyAlignAddress bit, not the last fragment!"); - uint64_t Address = StartAddress; - for (MCSectionData::iterator it = SD.begin(), ie = SD.end(); it != ie; ++it) { - MCFragment &F = *it; + uint64_t Size = OffsetToAlignment(SectionAddress + FragmentOffset, + AF.getAlignment()); - ++stats::FragmentLayouts; + // Honor MaxBytesToEmit. + if (Size > AF.getMaxBytesToEmit()) + return 0; - uint64_t FragmentOffset = Address - StartAddress; - Layout.setFragmentOffset(&F, FragmentOffset); + return Size; + } - // Evaluate fragment size. - uint64_t EffectiveSize = 0; - switch (F.getKind()) { - case MCFragment::FT_Align: { - MCAlignFragment &AF = cast<MCAlignFragment>(F); + case MCFragment::FT_Org: { + const MCOrgFragment &OF = cast<MCOrgFragment>(F); - EffectiveSize = OffsetToAlignment(Address, AF.getAlignment()); - if (EffectiveSize > AF.getMaxBytesToEmit()) - EffectiveSize = 0; - break; - } + // FIXME: We should compute this sooner, we don't want to recurse here, and + // we would like to be more functional. + int64_t TargetLocation; + if (!OF.getOffset().EvaluateAsAbsolute(TargetLocation, &Layout)) + report_fatal_error("expected assembly-time absolute expression"); - case MCFragment::FT_Data: - EffectiveSize = cast<MCDataFragment>(F).getContents().size(); - break; + // FIXME: We need a way to communicate this error. + int64_t Offset = TargetLocation - FragmentOffset; + if (Offset < 0) + report_fatal_error("invalid .org offset '" + Twine(TargetLocation) + + "' (at offset '" + Twine(FragmentOffset) + "'"); - case MCFragment::FT_Fill: { - MCFillFragment &FF = cast<MCFillFragment>(F); - EffectiveSize = FF.getValueSize() * FF.getCount(); - break; - } + return Offset; + } + } - case MCFragment::FT_Inst: - EffectiveSize = cast<MCInstFragment>(F).getInstSize(); - break; + assert(0 && "invalid fragment kind"); + return 0; +} - case MCFragment::FT_Org: { - MCOrgFragment &OF = cast<MCOrgFragment>(F); +void MCAsmLayout::LayoutFile() { + // Initialize the first section and set the valid fragment layout point. All + // actual layout computations are done lazily. + LastValidFragment = 0; + if (!getSectionOrder().empty()) + getSectionOrder().front()->Address = 0; +} - int64_t TargetLocation; - if (!OF.getOffset().EvaluateAsAbsolute(TargetLocation, &Layout)) - report_fatal_error("expected assembly-time absolute expression"); +void MCAsmLayout::LayoutFragment(MCFragment *F) { + MCFragment *Prev = F->getPrevNode(); - // FIXME: We need a way to communicate this error. - int64_t Offset = TargetLocation - FragmentOffset; - if (Offset < 0) - report_fatal_error("invalid .org offset '" + Twine(TargetLocation) + - "' (at offset '" + Twine(FragmentOffset) + "'"); + // We should never try to recompute something which is up-to-date. + assert(!isFragmentUpToDate(F) && "Attempt to recompute up-to-date fragment!"); + // We should never try to compute the fragment layout if the section isn't + // up-to-date. + assert(isSectionUpToDate(F->getParent()) && + "Attempt to compute fragment before it's section!"); + // We should never try to compute the fragment layout if it's predecessor + // isn't up-to-date. + assert((!Prev || isFragmentUpToDate(Prev)) && + "Attempt to compute fragment before it's predecessor!"); - EffectiveSize = Offset; - break; - } + ++stats::FragmentLayouts; - case MCFragment::FT_ZeroFill: { - MCZeroFillFragment &ZFF = cast<MCZeroFillFragment>(F); + // Compute the fragment start address. + uint64_t StartAddress = F->getParent()->Address; + uint64_t Address = StartAddress; + if (Prev) + Address += Prev->Offset + Prev->EffectiveSize; + + // Compute fragment offset and size. + F->Offset = Address - StartAddress; + F->EffectiveSize = getAssembler().ComputeFragmentSize(*this, *F, StartAddress, + F->Offset); + LastValidFragment = F; + + // If this is the last fragment in a section, update the next section address. + if (!F->getNextNode()) { + unsigned NextIndex = F->getParent()->getLayoutOrder() + 1; + if (NextIndex != getSectionOrder().size()) + LayoutSection(getSectionOrder()[NextIndex]); + } +} - // Align the fragment offset; it is safe to adjust the offset freely since - // this is only in virtual sections. - // - // FIXME: We shouldn't be doing this here. - Address = RoundUpToAlignment(Address, ZFF.getAlignment()); - Layout.setFragmentOffset(&F, Address - StartAddress); +void MCAsmLayout::LayoutSection(MCSectionData *SD) { + unsigned SectionOrderIndex = SD->getLayoutOrder(); - EffectiveSize = ZFF.getSize(); - break; - } - } + ++stats::SectionLayouts; - Layout.setFragmentEffectiveSize(&F, EffectiveSize); - Address += EffectiveSize; + // Compute the section start address. + uint64_t StartAddress = 0; + if (SectionOrderIndex) { + MCSectionData *Prev = getSectionOrder()[SectionOrderIndex - 1]; + StartAddress = getSectionAddress(Prev) + getSectionAddressSize(Prev); } - // Set the section sizes. - Layout.setSectionSize(&SD, Address - StartAddress); - if (IsVirtual) - Layout.setSectionFileSize(&SD, 0); - else - Layout.setSectionFileSize(&SD, Address - StartAddress); + // Honor the section alignment requirements. + StartAddress = RoundUpToAlignment(StartAddress, SD->getAlignment()); - return Address; + // Set the section address. + SD->Address = StartAddress; } /// WriteFragmentData - Write the \arg F data to the output file. @@ -522,6 +533,8 @@ static void WriteFragmentData(const MCAssembler &Asm, const MCAsmLayout &Layout, MCAlignFragment &AF = cast<MCAlignFragment>(F); uint64_t Count = FragmentSize / AF.getValueSize(); + assert(AF.getValueSize() && "Invalid virtual align in concrete fragment!"); + // FIXME: This error shouldn't actually occur (the front end should emit // multiple .align directives to enforce the semantics it wants), but is // severe enough that we want to report it. How to handle this? @@ -535,7 +548,7 @@ static void WriteFragmentData(const MCAssembler &Asm, const MCAsmLayout &Layout, // the Count bytes. Then if that did not fill any bytes or there are any // bytes left to fill use the the Value and ValueSize to fill the rest. // If we are aligning with nops, ask that target to emit the right data. - if (AF.getEmitNops()) { + if (AF.hasEmitNops()) { if (!Asm.getBackend().WriteNopData(Count, OW)) report_fatal_error("unable to write nop sequence of " + Twine(Count) + " bytes"); @@ -565,7 +578,10 @@ static void WriteFragmentData(const MCAssembler &Asm, const MCAsmLayout &Layout, case MCFragment::FT_Fill: { MCFillFragment &FF = cast<MCFillFragment>(F); - for (uint64_t i = 0, e = FF.getCount(); i != e; ++i) { + + assert(FF.getValueSize() && "Invalid virtual align in concrete fragment!"); + + for (uint64_t i = 0, e = FF.getSize() / FF.getValueSize(); i != e; ++i) { switch (FF.getValueSize()) { default: assert(0 && "Invalid size!"); @@ -590,11 +606,6 @@ static void WriteFragmentData(const MCAssembler &Asm, const MCAsmLayout &Layout, break; } - - case MCFragment::FT_ZeroFill: { - assert(0 && "Invalid zero fill fragment in concrete section!"); - break; - } } assert(OW->getStream().tell() - Start == FragmentSize); @@ -603,12 +614,27 @@ static void WriteFragmentData(const MCAssembler &Asm, const MCAsmLayout &Layout, void MCAssembler::WriteSectionData(const MCSectionData *SD, const MCAsmLayout &Layout, MCObjectWriter *OW) const { - uint64_t SectionSize = Layout.getSectionSize(SD); - uint64_t SectionFileSize = Layout.getSectionFileSize(SD); - // Ignore virtual sections. if (getBackend().isVirtualSection(SD->getSection())) { - assert(SectionFileSize == 0 && "Invalid size for section!"); + assert(Layout.getSectionFileSize(SD) == 0 && "Invalid size for section!"); + + // Check that contents are only things legal inside a virtual section. + for (MCSectionData::const_iterator it = SD->begin(), + ie = SD->end(); it != ie; ++it) { + switch (it->getKind()) { + default: + assert(0 && "Invalid fragment in virtual section!"); + case MCFragment::FT_Align: + assert(!cast<MCAlignFragment>(it)->getValueSize() && + "Invalid align in virtual section!"); + break; + case MCFragment::FT_Fill: + assert(!cast<MCFillFragment>(it)->getValueSize() && + "Invalid fill in virtual section!"); + break; + } + } + return; } @@ -619,11 +645,7 @@ void MCAssembler::WriteSectionData(const MCSectionData *SD, ie = SD->end(); it != ie; ++it) WriteFragmentData(*this, Layout, *it, OW); - // Add section padding. - assert(SectionFileSize >= SectionSize && "Invalid section sizes!"); - OW->WriteZeros(SectionFileSize - SectionSize); - - assert(OW->getStream().tell() - Start == SectionFileSize); + assert(OW->getStream().tell() - Start == Layout.getSectionFileSize(SD)); } void MCAssembler::Finish() { @@ -631,20 +653,60 @@ void MCAssembler::Finish() { llvm::errs() << "assembler backend - pre-layout\n--\n"; dump(); }); - // Assign section and fragment ordinals, all subsequent backend code is - // responsible for updating these in place. + // Create the layout object. + MCAsmLayout Layout(*this); + + // Insert additional align fragments for concrete sections to explicitly pad + // the previous section to match their alignment requirements. This is for + // 'gas' compatibility, it shouldn't strictly be necessary. + // + // FIXME: This may be Mach-O specific. + for (unsigned i = 1, e = Layout.getSectionOrder().size(); i < e; ++i) { + MCSectionData *SD = Layout.getSectionOrder()[i]; + + // Ignore sections without alignment requirements. + unsigned Align = SD->getAlignment(); + if (Align <= 1) + continue; + + // Ignore virtual sections, they don't cause file size modifications. + if (getBackend().isVirtualSection(SD->getSection())) + continue; + + // Otherwise, create a new align fragment at the end of the previous + // section. + MCAlignFragment *AF = new MCAlignFragment(Align, 0, 1, Align, + Layout.getSectionOrder()[i - 1]); + AF->setOnlyAlignAddress(true); + } + + // Create dummy fragments and assign section ordinals. unsigned SectionIndex = 0; - unsigned FragmentIndex = 0; for (MCAssembler::iterator it = begin(), ie = end(); it != ie; ++it) { + // Create dummy fragments to eliminate any empty sections, this simplifies + // layout. + if (it->getFragmentList().empty()) { + unsigned ValueSize = 1; + if (getBackend().isVirtualSection(it->getSection())) + ValueSize = 1; + new MCFillFragment(0, 1, 0, it); + } + it->setOrdinal(SectionIndex++); + } - for (MCSectionData::iterator it2 = it->begin(), - ie2 = it->end(); it2 != ie2; ++it2) - it2->setOrdinal(FragmentIndex++); + // Assign layout order indices to sections and fragments. + unsigned FragmentIndex = 0; + for (unsigned i = 0, e = Layout.getSectionOrder().size(); i != e; ++i) { + MCSectionData *SD = Layout.getSectionOrder()[i]; + SD->setLayoutOrder(i); + + for (MCSectionData::iterator it2 = SD->begin(), + ie2 = SD->end(); it2 != ie2; ++it2) + it2->setLayoutOrder(FragmentIndex++); } // Layout until everything fits. - MCAsmLayout Layout(*this); while (LayoutOnce(Layout)) continue; @@ -678,7 +740,7 @@ void MCAssembler::Finish() { for (MCDataFragment::fixup_iterator it3 = DF->fixup_begin(), ie3 = DF->fixup_end(); it3 != ie3; ++it3) { - MCAsmFixup &Fixup = *it3; + MCFixup &Fixup = *it3; // Evaluate the fixup. MCValue Target; @@ -702,7 +764,7 @@ void MCAssembler::Finish() { stats::ObjectBytes += OS.tell() - StartOffset; } -bool MCAssembler::FixupNeedsRelaxation(const MCAsmFixup &Fixup, +bool MCAssembler::FixupNeedsRelaxation(const MCFixup &Fixup, const MCFragment *DF, const MCAsmLayout &Layout) const { if (getRelaxAll()) @@ -725,7 +787,7 @@ bool MCAssembler::FragmentNeedsRelaxation(const MCInstFragment *IF, // If this inst doesn't ever need relaxation, ignore it. This occurs when we // are intentionally pushing out inst fragments, or because we relaxed a // previous instruction to one that doesn't need relaxation. - if (!getBackend().MayNeedRelaxation(IF->getInst(), IF->getFixups())) + if (!getBackend().MayNeedRelaxation(IF->getInst())) return false; for (MCInstFragment::const_fixup_iterator it = IF->fixup_begin(), @@ -739,25 +801,8 @@ bool MCAssembler::FragmentNeedsRelaxation(const MCInstFragment *IF, bool MCAssembler::LayoutOnce(MCAsmLayout &Layout) { ++stats::RelaxationSteps; - // Layout the concrete sections and fragments. - uint64_t Address = 0; - for (iterator it = begin(), ie = end(); it != ie; ++it) { - // Skip virtual sections. - if (getBackend().isVirtualSection(it->getSection())) - continue; - - // Layout the section fragments and its size. - Address = LayoutSection(*it, Layout, Address); - } - - // Layout the virtual sections. - for (iterator it = begin(), ie = end(); it != ie; ++it) { - if (!getBackend().isVirtualSection(it->getSection())) - continue; - - // Layout the section fragments and its size. - Address = LayoutSection(*it, Layout, Address); - } + // Layout the sections in order. + Layout.LayoutFile(); // Scan for fragments that need relaxation. bool WasRelaxed = false; @@ -779,7 +824,7 @@ bool MCAssembler::LayoutOnce(MCAsmLayout &Layout) { // Relax the fragment. MCInst Relaxed; - getBackend().RelaxInstruction(IF, Relaxed); + getBackend().RelaxInstruction(IF->getInst(), Relaxed); // Encode the new instruction. // @@ -796,17 +841,12 @@ bool MCAssembler::LayoutOnce(MCAsmLayout &Layout) { IF->setInst(Relaxed); IF->getCode() = Code; IF->getFixups().clear(); - for (unsigned i = 0, e = Fixups.size(); i != e; ++i) { - MCFixup &F = Fixups[i]; - IF->getFixups().push_back(MCAsmFixup(F.getOffset(), *F.getValue(), - F.getKind())); - } + // FIXME: Eliminate copy. + for (unsigned i = 0, e = Fixups.size(); i != e; ++i) + IF->getFixups().push_back(Fixups[i]); - // Update the layout, and remember that we relaxed. If we are relaxing - // everything, we can skip this step since nothing will depend on updating - // the values. - if (!getRelaxAll()) - Layout.UpdateForSlide(IF, SlideAmount); + // Update the layout, and remember that we relaxed. + Layout.UpdateForSlide(IF, SlideAmount); WasRelaxed = true; } } @@ -838,12 +878,10 @@ void MCAssembler::FinishLayout(MCAsmLayout &Layout) { SD.getFragmentList().insert(it2, DF); // Update the data fragments layout data. - // - // FIXME: Add MCAsmLayout utility for this. DF->setParent(IF->getParent()); - DF->setOrdinal(IF->getOrdinal()); - Layout.setFragmentOffset(DF, Layout.getFragmentOffset(IF)); - Layout.setFragmentEffectiveSize(DF, Layout.getFragmentEffectiveSize(IF)); + DF->setAtom(IF->getAtom()); + DF->setLayoutOrder(IF->getLayoutOrder()); + Layout.FragmentReplaced(IF, DF); // Copy in the data and the fixups. DF->getContents().append(IF->getCode().begin(), IF->getCode().end()); @@ -861,9 +899,10 @@ void MCAssembler::FinishLayout(MCAsmLayout &Layout) { namespace llvm { -raw_ostream &operator<<(raw_ostream &OS, const MCAsmFixup &AF) { - OS << "<MCAsmFixup" << " Offset:" << AF.Offset << " Value:" << *AF.Value - << " Kind:" << AF.Kind << ">"; +raw_ostream &operator<<(raw_ostream &OS, const MCFixup &AF) { + OS << "<MCFixup" << " Offset:" << AF.getOffset() + << " Value:" << *AF.getValue() + << " Kind:" << AF.getKind() << ">"; return OS; } @@ -872,94 +911,82 @@ raw_ostream &operator<<(raw_ostream &OS, const MCAsmFixup &AF) { void MCFragment::dump() { raw_ostream &OS = llvm::errs(); - OS << "<MCFragment " << (void*) this << " Offset:" << Offset - << " EffectiveSize:" << EffectiveSize; - - OS << ">"; -} - -void MCAlignFragment::dump() { - raw_ostream &OS = llvm::errs(); + OS << "<"; + switch (getKind()) { + case MCFragment::FT_Align: OS << "MCAlignFragment"; break; + case MCFragment::FT_Data: OS << "MCDataFragment"; break; + case MCFragment::FT_Fill: OS << "MCFillFragment"; break; + case MCFragment::FT_Inst: OS << "MCInstFragment"; break; + case MCFragment::FT_Org: OS << "MCOrgFragment"; break; + } - OS << "<MCAlignFragment "; - this->MCFragment::dump(); - OS << "\n "; - OS << " Alignment:" << getAlignment() - << " Value:" << getValue() << " ValueSize:" << getValueSize() - << " MaxBytesToEmit:" << getMaxBytesToEmit() << ">"; -} + OS << "<MCFragment " << (void*) this << " LayoutOrder:" << LayoutOrder + << " Offset:" << Offset << " EffectiveSize:" << EffectiveSize << ">"; -void MCDataFragment::dump() { - raw_ostream &OS = llvm::errs(); - - OS << "<MCDataFragment "; - this->MCFragment::dump(); - OS << "\n "; - OS << " Contents:["; - for (unsigned i = 0, e = getContents().size(); i != e; ++i) { - if (i) OS << ","; - OS << hexdigit((Contents[i] >> 4) & 0xF) << hexdigit(Contents[i] & 0xF); - } - OS << "] (" << getContents().size() << " bytes)"; - - if (!getFixups().empty()) { - OS << ",\n "; - OS << " Fixups:["; - for (fixup_iterator it = fixup_begin(), ie = fixup_end(); it != ie; ++it) { - if (it != fixup_begin()) OS << ",\n "; - OS << *it; + switch (getKind()) { + case MCFragment::FT_Align: { + const MCAlignFragment *AF = cast<MCAlignFragment>(this); + if (AF->hasEmitNops()) + OS << " (emit nops)"; + if (AF->hasOnlyAlignAddress()) + OS << " (only align section)"; + OS << "\n "; + OS << " Alignment:" << AF->getAlignment() + << " Value:" << AF->getValue() << " ValueSize:" << AF->getValueSize() + << " MaxBytesToEmit:" << AF->getMaxBytesToEmit() << ">"; + break; + } + case MCFragment::FT_Data: { + const MCDataFragment *DF = cast<MCDataFragment>(this); + OS << "\n "; + OS << " Contents:["; + const SmallVectorImpl<char> &Contents = DF->getContents(); + for (unsigned i = 0, e = Contents.size(); i != e; ++i) { + if (i) OS << ","; + OS << hexdigit((Contents[i] >> 4) & 0xF) << hexdigit(Contents[i] & 0xF); } - OS << "]"; + OS << "] (" << Contents.size() << " bytes)"; + + if (!DF->getFixups().empty()) { + OS << ",\n "; + OS << " Fixups:["; + for (MCDataFragment::const_fixup_iterator it = DF->fixup_begin(), + ie = DF->fixup_end(); it != ie; ++it) { + if (it != DF->fixup_begin()) OS << ",\n "; + OS << *it; + } + OS << "]"; + } + break; + } + case MCFragment::FT_Fill: { + const MCFillFragment *FF = cast<MCFillFragment>(this); + OS << " Value:" << FF->getValue() << " ValueSize:" << FF->getValueSize() + << " Size:" << FF->getSize(); + break; + } + case MCFragment::FT_Inst: { + const MCInstFragment *IF = cast<MCInstFragment>(this); + OS << "\n "; + OS << " Inst:"; + IF->getInst().dump_pretty(OS); + break; + } + case MCFragment::FT_Org: { + const MCOrgFragment *OF = cast<MCOrgFragment>(this); + OS << "\n "; + OS << " Offset:" << OF->getOffset() << " Value:" << OF->getValue(); + break; + } } - - OS << ">"; -} - -void MCFillFragment::dump() { - raw_ostream &OS = llvm::errs(); - - OS << "<MCFillFragment "; - this->MCFragment::dump(); - OS << "\n "; - OS << " Value:" << getValue() << " ValueSize:" << getValueSize() - << " Count:" << getCount() << ">"; -} - -void MCInstFragment::dump() { - raw_ostream &OS = llvm::errs(); - - OS << "<MCInstFragment "; - this->MCFragment::dump(); - OS << "\n "; - OS << " Inst:"; - getInst().dump_pretty(OS); OS << ">"; } -void MCOrgFragment::dump() { - raw_ostream &OS = llvm::errs(); - - OS << "<MCOrgFragment "; - this->MCFragment::dump(); - OS << "\n "; - OS << " Offset:" << getOffset() << " Value:" << getValue() << ">"; -} - -void MCZeroFillFragment::dump() { - raw_ostream &OS = llvm::errs(); - - OS << "<MCZeroFillFragment "; - this->MCFragment::dump(); - OS << "\n "; - OS << " Size:" << getSize() << " Alignment:" << getAlignment() << ">"; -} - void MCSectionData::dump() { raw_ostream &OS = llvm::errs(); OS << "<MCSectionData"; OS << " Alignment:" << getAlignment() << " Address:" << Address - << " Size:" << Size << " FileSize:" << FileSize << " Fragments:[\n "; for (iterator it = begin(), ie = end(); it != ie; ++it) { if (it != begin()) OS << ",\n "; diff --git a/lib/MC/MCContext.cpp b/lib/MC/MCContext.cpp index dc757bb..53ffc94 100644 --- a/lib/MC/MCContext.cpp +++ b/lib/MC/MCContext.cpp @@ -11,18 +11,22 @@ #include "llvm/MC/MCAsmInfo.h" #include "llvm/MC/MCSectionMachO.h" #include "llvm/MC/MCSectionELF.h" +#include "llvm/MC/MCSectionCOFF.h" #include "llvm/MC/MCSymbol.h" +#include "llvm/MC/MCLabel.h" #include "llvm/ADT/SmallString.h" #include "llvm/ADT/Twine.h" using namespace llvm; typedef StringMap<const MCSectionMachO*> MachOUniqueMapTy; typedef StringMap<const MCSectionELF*> ELFUniqueMapTy; +typedef StringMap<const MCSectionCOFF*> COFFUniqueMapTy; MCContext::MCContext(const MCAsmInfo &mai) : MAI(mai), NextUniqueID(0) { MachOUniquingMap = 0; ELFUniquingMap = 0; + COFFUniquingMap = 0; } MCContext::~MCContext() { @@ -32,6 +36,7 @@ MCContext::~MCContext() { // If we have the MachO uniquing map, free it. delete (MachOUniqueMapTy*)MachOUniquingMap; delete (ELFUniqueMapTy*)ELFUniquingMap; + delete (COFFUniqueMapTy*)COFFUniquingMap; } //===----------------------------------------------------------------------===// @@ -67,6 +72,34 @@ MCSymbol *MCContext::CreateTempSymbol() { "tmp" + Twine(NextUniqueID++)); } +unsigned MCContext::NextInstance(int64_t LocalLabelVal) { + MCLabel *&Label = Instances[LocalLabelVal]; + if (!Label) + Label = new (*this) MCLabel(0); + return Label->incInstance(); +} + +unsigned MCContext::GetInstance(int64_t LocalLabelVal) { + MCLabel *&Label = Instances[LocalLabelVal]; + if (!Label) + Label = new (*this) MCLabel(0); + return Label->getInstance(); +} + +MCSymbol *MCContext::CreateDirectionalLocalSymbol(int64_t LocalLabelVal) { + return GetOrCreateSymbol(Twine(MAI.getPrivateGlobalPrefix()) + + Twine(LocalLabelVal) + + "\2" + + Twine(NextInstance(LocalLabelVal))); +} +MCSymbol *MCContext::GetDirectionalLocalSymbol(int64_t LocalLabelVal, + int bORf) { + return GetOrCreateSymbol(Twine(MAI.getPrivateGlobalPrefix()) + + Twine(LocalLabelVal) + + "\2" + + Twine(GetInstance(LocalLabelVal) + bORf)); +} + MCSymbol *MCContext::LookupSymbol(StringRef Name) const { return Symbols.lookup(Name); } @@ -122,4 +155,22 @@ getELFSection(StringRef Section, unsigned Type, unsigned Flags, return Result; } - +const MCSection *MCContext::getCOFFSection(StringRef Section, + unsigned Characteristics, + int Selection, + SectionKind Kind) { + if (COFFUniquingMap == 0) + COFFUniquingMap = new COFFUniqueMapTy(); + COFFUniqueMapTy &Map = *(COFFUniqueMapTy*)COFFUniquingMap; + + // Do the lookup, if we have a hit, return it. + StringMapEntry<const MCSectionCOFF*> &Entry = Map.GetOrCreateValue(Section); + if (Entry.getValue()) return Entry.getValue(); + + MCSectionCOFF *Result = new (*this) MCSectionCOFF(Entry.getKey(), + Characteristics, + Selection, Kind); + + Entry.setValue(Result); + return Result; +} diff --git a/lib/MC/MCExpr.cpp b/lib/MC/MCExpr.cpp index bc670ab..c000dd7 100644 --- a/lib/MC/MCExpr.cpp +++ b/lib/MC/MCExpr.cpp @@ -39,6 +39,10 @@ void MCExpr::print(raw_ostream &OS) const { const MCSymbolRefExpr &SRE = cast<MCSymbolRefExpr>(*this); const MCSymbol &Sym = SRE.getSymbol(); + if (SRE.getKind() == MCSymbolRefExpr::VK_ARM_HI16 || + SRE.getKind() == MCSymbolRefExpr::VK_ARM_LO16) + OS << MCSymbolRefExpr::getVariantKindName(SRE.getKind()); + // Parenthesize names that start with $ so that they don't look like // absolute names. if (Sym.getName()[0] == '$') @@ -46,7 +50,9 @@ void MCExpr::print(raw_ostream &OS) const { else OS << Sym; - if (SRE.getKind() != MCSymbolRefExpr::VK_None) + if (SRE.getKind() != MCSymbolRefExpr::VK_None && + SRE.getKind() != MCSymbolRefExpr::VK_ARM_HI16 && + SRE.getKind() != MCSymbolRefExpr::VK_ARM_LO16) OS << '@' << MCSymbolRefExpr::getVariantKindName(SRE.getKind()); return; @@ -169,6 +175,9 @@ StringRef MCSymbolRefExpr::getVariantKindName(VariantKind Kind) { case VK_PLT: return "PLT"; case VK_TLSGD: return "TLSGD"; case VK_TPOFF: return "TPOFF"; + case VK_ARM_HI16: return ":upper16:"; + case VK_ARM_LO16: return ":lower16:"; + case VK_TLVP: return "TLVP"; } } @@ -184,6 +193,7 @@ MCSymbolRefExpr::getVariantKindForName(StringRef Name) { .Case("PLT", VK_PLT) .Case("TLSGD", VK_TLSGD) .Case("TPOFF", VK_TPOFF) + .Case("TLVP", VK_TLVP) .Default(VK_Invalid); } @@ -249,7 +259,7 @@ bool MCExpr::EvaluateAsRelocatable(MCValue &Res, // Evaluate recursively if this is a variable. if (Sym.isVariable()) { - if (!Sym.getValue()->EvaluateAsRelocatable(Res, Layout)) + if (!Sym.getVariableValue()->EvaluateAsRelocatable(Res, Layout)) return false; // Absolutize symbol differences between defined symbols when we have a diff --git a/lib/MC/MCInst.cpp b/lib/MC/MCInst.cpp index de142dc..4cb628b 100644 --- a/lib/MC/MCInst.cpp +++ b/lib/MC/MCInst.cpp @@ -57,7 +57,7 @@ void MCInst::dump_pretty(raw_ostream &OS, const MCAsmInfo *MAI, OS << Separator; getOperand(i).print(OS, MAI); } - OS << ">\n"; + OS << ">"; } void MCInst::dump() const { diff --git a/lib/MC/MCLabel.cpp b/lib/MC/MCLabel.cpp new file mode 100644 index 0000000..9c0fc92 --- /dev/null +++ b/lib/MC/MCLabel.cpp @@ -0,0 +1,21 @@ +//===- lib/MC/MCLabel.cpp - MCLabel implementation ----------------------===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// + +#include "llvm/MC/MCLabel.h" +#include "llvm/Support/Debug.h" +#include "llvm/Support/raw_ostream.h" +using namespace llvm; + +void MCLabel::print(raw_ostream &OS) const { + OS << '"' << getInstance() << '"'; +} + +void MCLabel::dump() const { + print(dbgs()); +} diff --git a/lib/MC/MCLoggingStreamer.cpp b/lib/MC/MCLoggingStreamer.cpp new file mode 100644 index 0000000..b96040a --- /dev/null +++ b/lib/MC/MCLoggingStreamer.cpp @@ -0,0 +1,208 @@ +//===- lib/MC/MCLoggingStreamer.cpp - API Logging Streamer ----------------===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// + +#include "llvm/MC/MCStreamer.h" +#include "llvm/ADT/OwningPtr.h" +#include "llvm/ADT/Twine.h" +#include "llvm/Support/raw_ostream.h" +using namespace llvm; + +namespace { + +class MCLoggingStreamer : public MCStreamer { + llvm::OwningPtr<MCStreamer> Child; + + raw_ostream &OS; + +public: + MCLoggingStreamer(MCStreamer *_Child, raw_ostream &_OS) + : MCStreamer(_Child->getContext()), Child(_Child), OS(_OS) {} + + void LogCall(const char *Function) { + OS << Function << "\n"; + } + + void LogCall(const char *Function, const Twine &Message) { + OS << Function << ": " << Message << "\n"; + } + + virtual bool isVerboseAsm() const { return Child->isVerboseAsm(); } + + virtual bool hasRawTextSupport() const { return Child->hasRawTextSupport(); } + + virtual raw_ostream &GetCommentOS() { return Child->GetCommentOS(); } + + virtual void AddComment(const Twine &T) { + LogCall("AddComment", T); + return Child->AddComment(T); + } + + virtual void AddBlankLine() { + LogCall("AddBlankLine"); + return Child->AddBlankLine(); + } + + virtual void SwitchSection(const MCSection *Section) { + CurSection = Section; + LogCall("SwitchSection"); + return Child->SwitchSection(Section); + } + + virtual void EmitLabel(MCSymbol *Symbol) { + LogCall("EmitLabel"); + return Child->EmitLabel(Symbol); + } + + virtual void EmitAssemblerFlag(MCAssemblerFlag Flag) { + LogCall("EmitAssemblerFlag"); + return Child->EmitAssemblerFlag(Flag); + } + + virtual void EmitAssignment(MCSymbol *Symbol, const MCExpr *Value) { + LogCall("EmitAssignment"); + return Child->EmitAssignment(Symbol, Value); + } + + virtual void EmitSymbolAttribute(MCSymbol *Symbol, MCSymbolAttr Attribute) { + LogCall("EmitSymbolAttribute"); + return Child->EmitSymbolAttribute(Symbol, Attribute); + } + + virtual void EmitSymbolDesc(MCSymbol *Symbol, unsigned DescValue) { + LogCall("EmitSymbolDesc"); + return Child->EmitSymbolDesc(Symbol, DescValue); + } + + virtual void BeginCOFFSymbolDef(const MCSymbol *Symbol) { + LogCall("BeginCOFFSymbolDef"); + return Child->BeginCOFFSymbolDef(Symbol); + } + + virtual void EmitCOFFSymbolStorageClass(int StorageClass) { + LogCall("EmitCOFFSymbolStorageClass"); + return Child->EmitCOFFSymbolStorageClass(StorageClass); + } + + virtual void EmitCOFFSymbolType(int Type) { + LogCall("EmitCOFFSymbolType"); + return Child->EmitCOFFSymbolType(Type); + } + + virtual void EndCOFFSymbolDef() { + LogCall("EndCOFFSymbolDef"); + return Child->EndCOFFSymbolDef(); + } + + virtual void EmitELFSize(MCSymbol *Symbol, const MCExpr *Value) { + LogCall("EmitELFSize"); + return Child->EmitELFSize(Symbol, Value); + } + + virtual void EmitCommonSymbol(MCSymbol *Symbol, uint64_t Size, + unsigned ByteAlignment) { + LogCall("EmitCommonSymbol"); + return Child->EmitCommonSymbol(Symbol, Size, ByteAlignment); + } + + virtual void EmitLocalCommonSymbol(MCSymbol *Symbol, uint64_t Size) { + LogCall("EmitLocalCommonSymbol"); + return Child->EmitLocalCommonSymbol(Symbol, Size); + } + + virtual void EmitZerofill(const MCSection *Section, MCSymbol *Symbol = 0, + unsigned Size = 0, unsigned ByteAlignment = 0) { + LogCall("EmitZerofill"); + return Child->EmitZerofill(Section, Symbol, Size, ByteAlignment); + } + + virtual void EmitTBSSSymbol (const MCSection *Section, MCSymbol *Symbol, + uint64_t Size, unsigned ByteAlignment = 0) { + LogCall("EmitTBSSSymbol"); + return Child->EmitTBSSSymbol(Section, Symbol, Size, ByteAlignment); + } + + virtual void EmitBytes(StringRef Data, unsigned AddrSpace) { + LogCall("EmitBytes"); + return Child->EmitBytes(Data, AddrSpace); + } + + virtual void EmitValue(const MCExpr *Value, unsigned Size,unsigned AddrSpace){ + LogCall("EmitValue"); + return Child->EmitValue(Value, Size, AddrSpace); + } + + virtual void EmitIntValue(uint64_t Value, unsigned Size, unsigned AddrSpace) { + LogCall("EmitIntValue"); + return Child->EmitIntValue(Value, Size, AddrSpace); + } + + virtual void EmitGPRel32Value(const MCExpr *Value) { + LogCall("EmitGPRel32Value"); + return Child->EmitGPRel32Value(Value); + } + + virtual void EmitFill(uint64_t NumBytes, uint8_t FillValue, + unsigned AddrSpace) { + LogCall("EmitFill"); + return Child->EmitFill(NumBytes, FillValue, AddrSpace); + } + + virtual void EmitValueToAlignment(unsigned ByteAlignment, int64_t Value = 0, + unsigned ValueSize = 1, + unsigned MaxBytesToEmit = 0) { + LogCall("EmitValueToAlignment"); + return Child->EmitValueToAlignment(ByteAlignment, Value, + ValueSize, MaxBytesToEmit); + } + + virtual void EmitCodeAlignment(unsigned ByteAlignment, + unsigned MaxBytesToEmit = 0) { + LogCall("EmitCodeAlignment"); + return Child->EmitCodeAlignment(ByteAlignment, MaxBytesToEmit); + } + + virtual void EmitValueToOffset(const MCExpr *Offset, + unsigned char Value = 0) { + LogCall("EmitValueToOffset"); + return Child->EmitValueToOffset(Offset, Value); + } + + virtual void EmitFileDirective(StringRef Filename) { + LogCall("EmitFileDirective", "FileName:" + Filename); + return Child->EmitFileDirective(Filename); + } + + virtual void EmitDwarfFileDirective(unsigned FileNo, StringRef Filename) { + LogCall("EmitDwarfFileDirective", + "FileNo:" + Twine(FileNo) + " Filename:" + Filename); + return Child->EmitDwarfFileDirective(FileNo, Filename); + } + + virtual void EmitInstruction(const MCInst &Inst) { + LogCall("EmitInstruction"); + return Child->EmitInstruction(Inst); + } + + virtual void EmitRawText(StringRef String) { + LogCall("EmitRawText", "\"" + String + "\""); + return Child->EmitRawText(String); + } + + virtual void Finish() { + LogCall("Finish"); + return Child->Finish(); + } + +}; + +} // end anonymous namespace. + +MCStreamer *llvm::createLoggingStreamer(MCStreamer *Child, raw_ostream &OS) { + return new MCLoggingStreamer(Child, OS); +} diff --git a/lib/MC/MCMachOStreamer.cpp b/lib/MC/MCMachOStreamer.cpp index 120f837..27e4e98 100644 --- a/lib/MC/MCMachOStreamer.cpp +++ b/lib/MC/MCMachOStreamer.cpp @@ -16,6 +16,7 @@ #include "llvm/MC/MCInst.h" #include "llvm/MC/MCSection.h" #include "llvm/MC/MCSymbol.h" +#include "llvm/MC/MCMachOSymbolFlags.h" #include "llvm/Support/ErrorHandling.h" #include "llvm/Support/raw_ostream.h" #include "llvm/Target/TargetAsmBackend.h" @@ -25,30 +26,14 @@ using namespace llvm; namespace { class MCMachOStreamer : public MCStreamer { - /// SymbolFlags - We store the value for the 'desc' symbol field in the lowest - /// 16 bits of the implementation defined flags. - enum SymbolFlags { // See <mach-o/nlist.h>. - SF_DescFlagsMask = 0xFFFF, - - // Reference type flags. - SF_ReferenceTypeMask = 0x0007, - SF_ReferenceTypeUndefinedNonLazy = 0x0000, - SF_ReferenceTypeUndefinedLazy = 0x0001, - SF_ReferenceTypeDefined = 0x0002, - SF_ReferenceTypePrivateDefined = 0x0003, - SF_ReferenceTypePrivateUndefinedNonLazy = 0x0004, - SF_ReferenceTypePrivateUndefinedLazy = 0x0005, - - // Other 'desc' flags. - SF_NoDeadStrip = 0x0020, - SF_WeakReference = 0x0040, - SF_WeakDefinition = 0x0080 - }; private: MCAssembler Assembler; MCSectionData *CurSectionData; + /// Track the current atom for each section. + DenseMap<const MCSectionData*, MCSymbolData*> CurrentAtomMap; + private: MCFragment *getCurrentFragment() const { assert(CurSectionData && "No current section!"); @@ -64,10 +49,20 @@ private: MCDataFragment *getOrCreateDataFragment() const { MCDataFragment *F = dyn_cast_or_null<MCDataFragment>(getCurrentFragment()); if (!F) - F = new MCDataFragment(CurSectionData); + F = createDataFragment(); return F; } + /// Create a new data fragment in the current section. + MCDataFragment *createDataFragment() const { + MCDataFragment *DF = new MCDataFragment(CurSectionData); + DF->setAtom(CurrentAtomMap.lookup(CurSectionData)); + return DF; + } + + void EmitInstToFragment(const MCInst &Inst); + void EmitInstToData(const MCInst &Inst); + public: MCMachOStreamer(MCContext &Context, TargetAsmBackend &TAB, raw_ostream &_OS, MCCodeEmitter *_Emitter) @@ -114,6 +109,18 @@ public: virtual void EmitSymbolDesc(MCSymbol *Symbol, unsigned DescValue); virtual void EmitCommonSymbol(MCSymbol *Symbol, uint64_t Size, unsigned ByteAlignment); + virtual void BeginCOFFSymbolDef(const MCSymbol *Symbol) { + assert(0 && "macho doesn't support this directive"); + } + virtual void EmitCOFFSymbolStorageClass(int StorageClass) { + assert(0 && "macho doesn't support this directive"); + } + virtual void EmitCOFFSymbolType(int Type) { + assert(0 && "macho doesn't support this directive"); + } + virtual void EndCOFFSymbolDef() { + assert(0 && "macho doesn't support this directive"); + } virtual void EmitELFSize(MCSymbol *Symbol, const MCExpr *Value) { assert(0 && "macho doesn't support this directive"); } @@ -122,6 +129,8 @@ public: } virtual void EmitZerofill(const MCSection *Section, MCSymbol *Symbol = 0, unsigned Size = 0, unsigned ByteAlignment = 0); + virtual void EmitTBSSSymbol(const MCSection *Section, MCSymbol *Symbol, + uint64_t Size, unsigned ByteAlignment = 0); virtual void EmitBytes(StringRef Data, unsigned AddrSpace); virtual void EmitValue(const MCExpr *Value, unsigned Size,unsigned AddrSpace); virtual void EmitGPRel32Value(const MCExpr *Value) { @@ -134,14 +143,14 @@ public: unsigned MaxBytesToEmit = 0); virtual void EmitValueToOffset(const MCExpr *Offset, unsigned char Value = 0); - + virtual void EmitFileDirective(StringRef Filename) { - errs() << "FIXME: MCMachoStreamer:EmitFileDirective not implemented\n"; + report_fatal_error("unsupported directive: '.file'"); } virtual void EmitDwarfFileDirective(unsigned FileNo, StringRef Filename) { - errs() << "FIXME: MCMachoStreamer:EmitDwarfFileDirective not implemented\n"; + report_fatal_error("unsupported directive: '.file'"); } - + virtual void EmitInstruction(const MCInst &Inst); virtual void Finish(); @@ -152,7 +161,7 @@ public: void MCMachOStreamer::SwitchSection(const MCSection *Section) { assert(Section && "Cannot switch to a null section!"); - + // If already in this section, then this is a noop. if (Section == CurSection) return; @@ -162,20 +171,39 @@ void MCMachOStreamer::SwitchSection(const MCSection *Section) { void MCMachOStreamer::EmitLabel(MCSymbol *Symbol) { assert(Symbol->isUndefined() && "Cannot define a symbol twice!"); + assert(!Symbol->isVariable() && "Cannot emit a variable symbol!"); + assert(CurSection && "Cannot emit before setting section!"); + + MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol); + + // Update the current atom map, if necessary. + bool MustCreateFragment = false; + if (Assembler.isSymbolLinkerVisible(&SD)) { + CurrentAtomMap[CurSectionData] = &SD; + + // We have to create a new fragment, fragments cannot span atoms. + MustCreateFragment = true; + } // FIXME: This is wasteful, we don't necessarily need to create a data // fragment. Instead, we should mark the symbol as pointing into the data // fragment if it exists, otherwise we should just queue the label and set its // fragment pointer when we emit the next fragment. - MCDataFragment *F = getOrCreateDataFragment(); - MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol); + MCDataFragment *F = + MustCreateFragment ? createDataFragment() : getOrCreateDataFragment(); assert(!SD.getFragment() && "Unexpected fragment on symbol data!"); SD.setFragment(F); SD.setOffset(F->getContents().size()); - // This causes the reference type and weak reference flags to be cleared. - SD.setFlags(SD.getFlags() & ~(SF_WeakReference | SF_ReferenceTypeMask)); - + // This causes the reference type flag to be cleared. Darwin 'as' was "trying" + // to clear the weak reference and weak definition bits too, but the + // implementation was buggy. For now we just try to match 'as', for + // diffability. + // + // FIXME: Cleanup this code, these bits should be emitted based on semantic + // properties, not on the order of definition, etc. + SD.setFlags(SD.getFlags() & ~SF_ReferenceTypeMask); + Symbol->setSection(*CurSection); } @@ -190,13 +218,9 @@ void MCMachOStreamer::EmitAssemblerFlag(MCAssemblerFlag Flag) { } void MCMachOStreamer::EmitAssignment(MCSymbol *Symbol, const MCExpr *Value) { - // Only absolute symbols can be redefined. - assert((Symbol->isUndefined() || Symbol->isAbsolute()) && - "Cannot define a symbol twice!"); - // FIXME: Lift context changes into super class. - // FIXME: Set associated section. - Symbol->setValue(AddValueSymbols(Value)); + Assembler.getOrCreateSymbolData(*Symbol); + Symbol->setVariableValue(AddValueSymbols(Value)); } void MCMachOStreamer::EmitSymbolAttribute(MCSymbol *Symbol, @@ -243,6 +267,13 @@ void MCMachOStreamer::EmitSymbolAttribute(MCSymbol *Symbol, case MCSA_Global: SD.setExternal(true); + // This effectively clears the undefined lazy bit, in Darwin 'as', although + // it isn't very consistent because it implements this as part of symbol + // lookup. + // + // FIXME: Cleanup this code, these bits should be emitted based on semantic + // properties, not on the order of definition, etc. + SD.setFlags(SD.getFlags() & ~SF_ReferenceTypeUndefinedLazy); break; case MCSA_LazyReference: @@ -280,7 +311,7 @@ void MCMachOStreamer::EmitSymbolAttribute(MCSymbol *Symbol, void MCMachOStreamer::EmitSymbolDesc(MCSymbol *Symbol, unsigned DescValue) { // Encode the 'desc' value into the lowest implementation defined bits. - assert(DescValue == (DescValue & SF_DescFlagsMask) && + assert(DescValue == (DescValue & SF_DescFlagsMask) && "Invalid .desc value!"); Assembler.getOrCreateSymbolData(*Symbol).setFlags(DescValue&SF_DescFlagsMask); } @@ -309,8 +340,14 @@ void MCMachOStreamer::EmitZerofill(const MCSection *Section, MCSymbol *Symbol, MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol); - MCFragment *F = new MCZeroFillFragment(Size, ByteAlignment, &SectData); + // Emit an align fragment if necessary. + if (ByteAlignment != 1) + new MCAlignFragment(ByteAlignment, 0, 0, ByteAlignment, &SectData); + + MCFragment *F = new MCFillFragment(0, 0, Size, &SectData); SD.setFragment(F); + if (Assembler.isSymbolLinkerVisible(&SD)) + F->setAtom(&SD); Symbol->setSection(*Section); @@ -319,6 +356,14 @@ void MCMachOStreamer::EmitZerofill(const MCSection *Section, MCSymbol *Symbol, SectData.setAlignment(ByteAlignment); } +// This should always be called with the thread local bss section. Like the +// .zerofill directive this doesn't actually switch sections on us. +void MCMachOStreamer::EmitTBSSSymbol(const MCSection *Section, MCSymbol *Symbol, + uint64_t Size, unsigned ByteAlignment) { + EmitZerofill(Section, Symbol, Size, ByteAlignment); + return; +} + void MCMachOStreamer::EmitBytes(StringRef Data, unsigned AddrSpace) { getOrCreateDataFragment()->getContents().append(Data.begin(), Data.end()); } @@ -334,8 +379,9 @@ void MCMachOStreamer::EmitValue(const MCExpr *Value, unsigned Size, for (unsigned i = 0; i != Size; ++i) DF->getContents().push_back(uint8_t(AbsValue >> (i * 8))); } else { - DF->addFixup(MCAsmFixup(DF->getContents().size(), *AddValueSymbols(Value), - MCFixup::getKindForSize(Size))); + DF->addFixup(MCFixup::Create(DF->getContents().size(), + AddValueSymbols(Value), + MCFixup::getKindForSize(Size))); DF->getContents().resize(DF->getContents().size() + Size, 0); } } @@ -345,8 +391,9 @@ void MCMachOStreamer::EmitValueToAlignment(unsigned ByteAlignment, unsigned MaxBytesToEmit) { if (MaxBytesToEmit == 0) MaxBytesToEmit = ByteAlignment; - new MCAlignFragment(ByteAlignment, Value, ValueSize, MaxBytesToEmit, - false /* EmitNops */, CurSectionData); + MCFragment *F = new MCAlignFragment(ByteAlignment, Value, ValueSize, + MaxBytesToEmit, CurSectionData); + F->setAtom(CurrentAtomMap.lookup(CurSectionData)); // Update the maximum alignment on the current section if necessary. if (ByteAlignment > CurSectionData->getAlignment()) @@ -357,8 +404,10 @@ void MCMachOStreamer::EmitCodeAlignment(unsigned ByteAlignment, unsigned MaxBytesToEmit) { if (MaxBytesToEmit == 0) MaxBytesToEmit = ByteAlignment; - new MCAlignFragment(ByteAlignment, 0, 1, MaxBytesToEmit, - true /* EmitNops */, CurSectionData); + MCAlignFragment *F = new MCAlignFragment(ByteAlignment, 0, 1, MaxBytesToEmit, + CurSectionData); + F->setEmitNops(true); + F->setAtom(CurrentAtomMap.lookup(CurSectionData)); // Update the maximum alignment on the current section if necessary. if (ByteAlignment > CurSectionData->getAlignment()) @@ -367,19 +416,30 @@ void MCMachOStreamer::EmitCodeAlignment(unsigned ByteAlignment, void MCMachOStreamer::EmitValueToOffset(const MCExpr *Offset, unsigned char Value) { - new MCOrgFragment(*Offset, Value, CurSectionData); + MCFragment *F = new MCOrgFragment(*Offset, Value, CurSectionData); + F->setAtom(CurrentAtomMap.lookup(CurSectionData)); } -void MCMachOStreamer::EmitInstruction(const MCInst &Inst) { - // Scan for values. - for (unsigned i = 0; i != Inst.getNumOperands(); ++i) - if (Inst.getOperand(i).isExpr()) - AddValueSymbols(Inst.getOperand(i).getExpr()); +void MCMachOStreamer::EmitInstToFragment(const MCInst &Inst) { + MCInstFragment *IF = new MCInstFragment(Inst, CurSectionData); + IF->setAtom(CurrentAtomMap.lookup(CurSectionData)); - CurSectionData->setHasInstructions(true); + // Add the fixups and data. + // + // FIXME: Revisit this design decision when relaxation is done, we may be + // able to get away with not storing any extra data in the MCInst. + SmallVector<MCFixup, 4> Fixups; + SmallString<256> Code; + raw_svector_ostream VecOS(Code); + Assembler.getEmitter().EncodeInstruction(Inst, VecOS, Fixups); + VecOS.flush(); - // FIXME-PERF: Common case is that we don't need to relax, encode directly - // onto the data fragments buffers. + IF->getCode() = Code; + IF->getFixups() = Fixups; +} + +void MCMachOStreamer::EmitInstToData(const MCInst &Inst) { + MCDataFragment *DF = getOrCreateDataFragment(); SmallVector<MCFixup, 4> Fixups; SmallString<256> Code; @@ -387,47 +447,41 @@ void MCMachOStreamer::EmitInstruction(const MCInst &Inst) { Assembler.getEmitter().EncodeInstruction(Inst, VecOS, Fixups); VecOS.flush(); - // FIXME: Eliminate this copy. - SmallVector<MCAsmFixup, 4> AsmFixups; + // Add the fixups and data. for (unsigned i = 0, e = Fixups.size(); i != e; ++i) { - MCFixup &F = Fixups[i]; - AsmFixups.push_back(MCAsmFixup(F.getOffset(), *F.getValue(), - F.getKind())); + Fixups[i].setOffset(Fixups[i].getOffset() + DF->getContents().size()); + DF->addFixup(Fixups[i]); } + DF->getContents().append(Code.begin(), Code.end()); +} - // See if we might need to relax this instruction, if so it needs its own - // fragment. - // - // FIXME-PERF: Support target hook to do a fast path that avoids the encoder, - // when we can immediately tell that we will get something which might need - // relaxation (and compute its size). - // - // FIXME-PERF: We should also be smart about immediately relaxing instructions - // which we can already show will never possibly fit (we can also do a very - // good job of this before we do the first relaxation pass, because we have - // total knowledge about undefined symbols at that point). Even now, though, - // we can do a decent job, especially on Darwin where scattering means that we - // are going to often know that we can never fully resolve a fixup. - if (Assembler.getBackend().MayNeedRelaxation(Inst, AsmFixups)) { - MCInstFragment *IF = new MCInstFragment(Inst, CurSectionData); - - // Add the fixups and data. - // - // FIXME: Revisit this design decision when relaxation is done, we may be - // able to get away with not storing any extra data in the MCInst. - IF->getCode() = Code; - IF->getFixups() = AsmFixups; +void MCMachOStreamer::EmitInstruction(const MCInst &Inst) { + // Scan for values. + for (unsigned i = Inst.getNumOperands(); i--; ) + if (Inst.getOperand(i).isExpr()) + AddValueSymbols(Inst.getOperand(i).getExpr()); + CurSectionData->setHasInstructions(true); + + // If this instruction doesn't need relaxation, just emit it as data. + if (!Assembler.getBackend().MayNeedRelaxation(Inst)) { + EmitInstToData(Inst); return; } - // Add the fixups and data. - MCDataFragment *DF = getOrCreateDataFragment(); - for (unsigned i = 0, e = AsmFixups.size(); i != e; ++i) { - AsmFixups[i].Offset += DF->getContents().size(); - DF->addFixup(AsmFixups[i]); + // Otherwise, if we are relaxing everything, relax the instruction as much as + // possible and emit it as data. + if (Assembler.getRelaxAll()) { + MCInst Relaxed; + Assembler.getBackend().RelaxInstruction(Inst, Relaxed); + while (Assembler.getBackend().MayNeedRelaxation(Relaxed)) + Assembler.getBackend().RelaxInstruction(Relaxed, Relaxed); + EmitInstToData(Relaxed); + return; } - DF->getContents().append(Code.begin(), Code.end()); + + // Otherwise emit to a separate fragment. + EmitInstToFragment(Inst); } void MCMachOStreamer::Finish() { diff --git a/lib/MC/MCNullStreamer.cpp b/lib/MC/MCNullStreamer.cpp index 5f0c64a..5332ade 100644 --- a/lib/MC/MCNullStreamer.cpp +++ b/lib/MC/MCNullStreamer.cpp @@ -42,6 +42,12 @@ namespace { virtual void EmitSymbolAttribute(MCSymbol *Symbol, MCSymbolAttr Attribute){} virtual void EmitSymbolDesc(MCSymbol *Symbol, unsigned DescValue) {} + + virtual void BeginCOFFSymbolDef(const MCSymbol *Symbol) {} + virtual void EmitCOFFSymbolStorageClass(int StorageClass) {} + virtual void EmitCOFFSymbolType(int Type) {} + virtual void EndCOFFSymbolDef() {} + virtual void EmitELFSize(MCSymbol *Symbol, const MCExpr *Value) {} virtual void EmitCommonSymbol(MCSymbol *Symbol, uint64_t Size, unsigned ByteAlignment) {} @@ -49,7 +55,8 @@ namespace { virtual void EmitZerofill(const MCSection *Section, MCSymbol *Symbol = 0, unsigned Size = 0, unsigned ByteAlignment = 0) {} - + virtual void EmitTBSSSymbol(const MCSection *Section, MCSymbol *Symbol, + uint64_t Size, unsigned ByteAlignment) {} virtual void EmitBytes(StringRef Data, unsigned AddrSpace) {} virtual void EmitValue(const MCExpr *Value, unsigned Size, diff --git a/lib/MC/MCParser/AsmLexer.cpp b/lib/MC/MCParser/AsmLexer.cpp index 1183312..1cbe09a 100644 --- a/lib/MC/MCParser/AsmLexer.cpp +++ b/lib/MC/MCParser/AsmLexer.cpp @@ -132,11 +132,6 @@ AsmToken AsmLexer::LexLineComment() { /// Decimal integer: [1-9][0-9]* /// TODO: FP literal. AsmToken AsmLexer::LexDigit() { - if (*CurPtr == ':') - return ReturnError(TokStart, "FIXME: local label not implemented"); - if (*CurPtr == 'f' || *CurPtr == 'b') - return ReturnError(TokStart, "FIXME: directional label not implemented"); - // Decimal integer: [1-9][0-9]* if (CurPtr[-1] != '0') { while (isdigit(*CurPtr)) @@ -158,6 +153,12 @@ AsmToken AsmLexer::LexDigit() { if (*CurPtr == 'b') { ++CurPtr; + // See if we actually have "0b" as part of something like "jmp 0b\n" + if (!isdigit(CurPtr[0])) { + --CurPtr; + StringRef Result(TokStart, CurPtr - TokStart); + return AsmToken(AsmToken::Integer, Result, 0); + } const char *NumStart = CurPtr; while (CurPtr[0] == '0' || CurPtr[0] == '1') ++CurPtr; @@ -280,6 +281,7 @@ AsmToken AsmLexer::LexToken() { case '*': return AsmToken(AsmToken::Star, StringRef(TokStart, 1)); case ',': return AsmToken(AsmToken::Comma, StringRef(TokStart, 1)); case '$': return AsmToken(AsmToken::Dollar, StringRef(TokStart, 1)); + case '@': return AsmToken(AsmToken::At, StringRef(TokStart, 1)); case '=': if (*CurPtr == '=') return ++CurPtr, AsmToken(AsmToken::EqualEqual, StringRef(TokStart, 2)); diff --git a/lib/MC/MCParser/AsmParser.cpp b/lib/MC/MCParser/AsmParser.cpp index a63d2e4..4523eab 100644 --- a/lib/MC/MCParser/AsmParser.cpp +++ b/lib/MC/MCParser/AsmParser.cpp @@ -13,6 +13,7 @@ #include "llvm/MC/MCParser/AsmParser.h" #include "llvm/ADT/SmallString.h" +#include "llvm/ADT/StringSwitch.h" #include "llvm/ADT/Twine.h" #include "llvm/MC/MCContext.h" #include "llvm/MC/MCExpr.h" @@ -189,6 +190,9 @@ bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) { std::pair<StringRef, StringRef> Split = getTok().getIdentifier().split('@'); MCSymbol *Sym = CreateSymbol(Split.first); + // Mark the symbol as used in an expression. + Sym->setUsedInExpr(true); + // Lookup the symbol variant if used. MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; if (Split.first.size() != getTok().getIdentifier().size()) @@ -199,11 +203,11 @@ bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) { // If this is an absolute variable reference, substitute it now to preserve // semantics in the face of reassignment. - if (Sym->getValue() && isa<MCConstantExpr>(Sym->getValue())) { + if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) { if (Variant) return Error(EndLoc, "unexpected modified on variable reference"); - Res = Sym->getValue(); + Res = Sym->getVariableValue(); return false; } @@ -211,11 +215,28 @@ bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) { Res = MCSymbolRefExpr::Create(Sym, Variant, getContext()); return false; } - case AsmToken::Integer: - Res = MCConstantExpr::Create(getTok().getIntVal(), getContext()); + case AsmToken::Integer: { + SMLoc Loc = getTok().getLoc(); + int64_t IntVal = getTok().getIntVal(); + Res = MCConstantExpr::Create(IntVal, getContext()); EndLoc = Lexer.getLoc(); Lex(); // Eat token. + // Look for 'b' or 'f' following an Integer as a directional label + if (Lexer.getKind() == AsmToken::Identifier) { + StringRef IDVal = getTok().getString(); + if (IDVal == "f" || IDVal == "b"){ + MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal, + IDVal == "f" ? 1 : 0); + Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, + getContext()); + if(IDVal == "b" && Sym->isUndefined()) + return Error(Loc, "invalid reference to undefined symbol"); + EndLoc = Lexer.getLoc(); + Lex(); // Eat identifier. + } + } return false; + } case AsmToken::Dot: { // This is a '.' reference, which references the current PC. Emit a // temporary label to the streamer and refer to it. @@ -411,6 +432,7 @@ bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, /// ::= Label* Identifier OperandList* EndOfStatement bool AsmParser::ParseStatement() { if (Lexer.is(AsmToken::EndOfStatement)) { + Out.AddBlankLine(); Lex(); return false; } @@ -419,7 +441,25 @@ bool AsmParser::ParseStatement() { AsmToken ID = getTok(); SMLoc IDLoc = ID.getLoc(); StringRef IDVal; - if (ParseIdentifier(IDVal)) { + int64_t LocalLabelVal = -1; + // GUESS allow an integer followed by a ':' as a directional local label + if (Lexer.is(AsmToken::Integer)) { + LocalLabelVal = getTok().getIntVal(); + if (LocalLabelVal < 0) { + if (!TheCondState.Ignore) + return TokError("unexpected token at start of statement"); + IDVal = ""; + } + else { + IDVal = getTok().getString(); + Lex(); // Consume the integer token to be used as an identifier token. + if (Lexer.getKind() != AsmToken::Colon) { + if (!TheCondState.Ignore) + return TokError("unexpected token at start of statement"); + } + } + } + else if (ParseIdentifier(IDVal)) { if (!TheCondState.Ignore) return TokError("unexpected token at start of statement"); IDVal = ""; @@ -456,13 +496,25 @@ bool AsmParser::ParseStatement() { // FIXME: Diagnostics. Note the location of the definition as a label. // FIXME: This doesn't diagnose assignment to a symbol which has been // implicitly marked as external. - MCSymbol *Sym = CreateSymbol(IDVal); - if (!Sym->isUndefined()) + MCSymbol *Sym; + if (LocalLabelVal == -1) + Sym = CreateSymbol(IDVal); + else + Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal); + if (!Sym->isUndefined() || Sym->isVariable()) return Error(IDLoc, "invalid symbol redefinition"); // Emit the label. Out.EmitLabel(Sym); + // Consume any end of statement token, if present, to avoid spurious + // AddBlankLine calls(). + if (Lexer.is(AsmToken::EndOfStatement)) { + Lex(); + if (Lexer.is(AsmToken::Eof)) + return false; + } + return ParseStatement(); } @@ -620,6 +672,16 @@ bool AsmParser::ParseStatement() { return ParseDirectiveSectionSwitch("__OBJC", "__selector_strs", MCSectionMachO::S_CSTRING_LITERALS); + if (IDVal == ".tdata") + return ParseDirectiveSectionSwitch("__DATA", "__thread_data", + MCSectionMachO::S_THREAD_LOCAL_REGULAR); + if (IDVal == ".tlv") + return ParseDirectiveSectionSwitch("__DATA", "__thread_vars", + MCSectionMachO::S_THREAD_LOCAL_VARIABLES); + if (IDVal == ".thread_init_func") + return ParseDirectiveSectionSwitch("__DATA", "__thread_init", + MCSectionMachO::S_THREAD_LOCAL_INIT_FUNCTION_POINTERS); + // Assembler features if (IDVal == ".set") return ParseDirectiveSet(); @@ -686,6 +748,8 @@ bool AsmParser::ParseStatement() { return ParseDirectiveSymbolAttribute(MCSA_Protected); if (IDVal == ".reference") return ParseDirectiveSymbolAttribute(MCSA_Reference); + if (IDVal == ".type") + return ParseDirectiveELFType(); if (IDVal == ".weak") return ParseDirectiveSymbolAttribute(MCSA_Weak); if (IDVal == ".weak_definition") @@ -703,6 +767,8 @@ bool AsmParser::ParseStatement() { return ParseDirectiveDarwinSymbolDesc(); if (IDVal == ".lsym") return ParseDirectiveDarwinLsym(); + if (IDVal == ".tbss") + return ParseDirectiveDarwinTBSS(); if (IDVal == ".subsections_via_symbols") return ParseDirectiveDarwinSubsectionsViaSymbols(); @@ -729,8 +795,13 @@ bool AsmParser::ParseStatement() { return false; } + // Canonicalize the opcode to lower case. + SmallString<128> Opcode; + for (unsigned i = 0, e = IDVal.size(); i != e; ++i) + Opcode.push_back(tolower(IDVal[i])); + SmallVector<MCParsedAsmOperand*, 8> ParsedOperands; - bool HadError = getTargetParser().ParseInstruction(IDVal, IDLoc, + bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc, ParsedOperands); if (!HadError && Lexer.isNot(AsmToken::EndOfStatement)) HadError = TokError("unexpected token in argument list"); @@ -786,11 +857,13 @@ bool AsmParser::ParseAssignment(const StringRef &Name) { // // FIXME: Diagnostics. Note the location of the definition as a label. // FIXME: Diagnose assignment to protected identifier (e.g., register name). - if (!Sym->isUndefined() && !Sym->isAbsolute()) + if (Sym->isUndefined() && !Sym->isUsedInExpr()) + ; // Allow redefinitions of undefined symbols only used in directives. + else if (!Sym->isUndefined() && !Sym->isAbsolute()) return Error(EqualLoc, "redefinition of '" + Name + "'"); else if (!Sym->isVariable()) return Error(EqualLoc, "invalid assignment to '" + Name + "'"); - else if (!isa<MCConstantExpr>(Sym->getValue())) + else if (!isa<MCConstantExpr>(Sym->getVariableValue())) return Error(EqualLoc, "invalid reassignment of non-absolute variable '" + Name + "'"); } else @@ -798,6 +871,8 @@ bool AsmParser::ParseAssignment(const StringRef &Name) { // FIXME: Handle '.'. + Sym->setUsedInExpr(true); + // Do the assignment. Out.EmitAssignment(Sym, Value); @@ -1008,7 +1083,11 @@ bool AsmParser::ParseDirectiveValue(unsigned Size) { if (ParseExpression(Value)) return true; - Out.EmitValue(Value, Size, DEFAULT_ADDRSPACE); + // Special case constant expressions to match code generator. + if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) + Out.EmitIntValue(MCE->getValue(), Size, DEFAULT_ADDRSPACE); + else + Out.EmitValue(Value, Size, DEFAULT_ADDRSPACE); if (Lexer.is(AsmToken::EndOfStatement)) break; @@ -1090,8 +1169,7 @@ bool AsmParser::ParseDirectiveFill() { return TokError("invalid '.fill' size, expected 1, 2, 4, or 8"); for (uint64_t i = 0, e = NumValues; i != e; ++i) - Out.EmitValue(MCConstantExpr::Create(FillExpr, getContext()), FillSize, - DEFAULT_ADDRSPACE); + Out.EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE); return false; } @@ -1169,10 +1247,8 @@ bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) { Lex(); - if (!HasFillExpr) { - // FIXME: Sometimes fill with nop. + if (!HasFillExpr) FillExpr = 0; - } // Compute alignment in bytes. if (IsPow2) { @@ -1200,14 +1276,21 @@ bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) { } } - // FIXME: hard code the parser to use EmitCodeAlignment for text when using - // the TextAlignFillValue. - if(Out.getCurrentSection()->getKind().isText() && - Lexer.getMAI().getTextAlignFillValue() == FillExpr) + // Check whether we should use optimal code alignment for this .align + // directive. + // + // FIXME: This should be using a target hook. + bool UseCodeAlign = false; + if (const MCSectionMachO *S = dyn_cast<MCSectionMachO>( + Out.getCurrentSection())) + UseCodeAlign = S->hasAttribute(MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS); + if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) && + ValueSize == 1 && UseCodeAlign) { Out.EmitCodeAlignment(Alignment, MaxBytesToFill); - else + } else { // FIXME: Target specific behavior about how the "extra" bytes are filled. Out.EmitValueToAlignment(Alignment, FillExpr, ValueSize, MaxBytesToFill); + } return false; } @@ -1239,6 +1322,52 @@ bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) { return false; } +/// ParseDirectiveELFType +/// ::= .type identifier , @attribute +bool AsmParser::ParseDirectiveELFType() { + StringRef Name; + if (ParseIdentifier(Name)) + return TokError("expected identifier in directive"); + + // Handle the identifier as the key symbol. + MCSymbol *Sym = CreateSymbol(Name); + + if (Lexer.isNot(AsmToken::Comma)) + return TokError("unexpected token in '.type' directive"); + Lex(); + + if (Lexer.isNot(AsmToken::At)) + return TokError("expected '@' before type"); + Lex(); + + StringRef Type; + SMLoc TypeLoc; + + TypeLoc = Lexer.getLoc(); + if (ParseIdentifier(Type)) + return TokError("expected symbol type in directive"); + + MCSymbolAttr Attr = StringSwitch<MCSymbolAttr>(Type) + .Case("function", MCSA_ELF_TypeFunction) + .Case("object", MCSA_ELF_TypeObject) + .Case("tls_object", MCSA_ELF_TypeTLS) + .Case("common", MCSA_ELF_TypeCommon) + .Case("notype", MCSA_ELF_TypeNoType) + .Default(MCSA_Invalid); + + if (Attr == MCSA_Invalid) + return Error(TypeLoc, "unsupported attribute in '.type' directive"); + + if (Lexer.isNot(AsmToken::EndOfStatement)) + return TokError("unexpected token in '.type' directive"); + + Lex(); + + Out.EmitSymbolAttribute(Sym, Attr); + + return false; +} + /// ParseDirectiveDarwinSymbolDesc /// ::= .desc identifier , expression bool AsmParser::ParseDirectiveDarwinSymbolDesc() { @@ -1316,7 +1445,7 @@ bool AsmParser::ParseDirectiveComm(bool IsLocal) { return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't " "be less than zero"); - // NOTE: The alignment in the directive is a power of 2 value, the assember + // NOTE: The alignment in the directive is a power of 2 value, the assembler // may internally end up wanting an alignment in bytes. // FIXME: Diagnose overflow. if (Pow2Alignment < 0) @@ -1344,22 +1473,18 @@ bool AsmParser::ParseDirectiveComm(bool IsLocal) { /// ::= .zerofill segname , sectname [, identifier , size_expression [ /// , align_expression ]] bool AsmParser::ParseDirectiveDarwinZerofill() { - // FIXME: Handle quoted names here. - - if (Lexer.isNot(AsmToken::Identifier)) + StringRef Segment; + if (ParseIdentifier(Segment)) return TokError("expected segment name after '.zerofill' directive"); - StringRef Segment = getTok().getString(); - Lex(); if (Lexer.isNot(AsmToken::Comma)) return TokError("unexpected token in directive"); Lex(); - - if (Lexer.isNot(AsmToken::Identifier)) + + StringRef Section; + if (ParseIdentifier(Section)) return TokError("expected section name after comma in '.zerofill' " "directive"); - StringRef Section = getTok().getString(); - Lex(); // If this is the end of the line all that was wanted was to create the // the section but with no symbol. @@ -1375,13 +1500,13 @@ bool AsmParser::ParseDirectiveDarwinZerofill() { return TokError("unexpected token in directive"); Lex(); - if (Lexer.isNot(AsmToken::Identifier)) + SMLoc IDLoc = Lexer.getLoc(); + StringRef IDStr; + if (ParseIdentifier(IDStr)) return TokError("expected identifier in directive"); // handle the identifier as the key symbol. - SMLoc IDLoc = Lexer.getLoc(); - MCSymbol *Sym = CreateSymbol(getTok().getString()); - Lex(); + MCSymbol *Sym = CreateSymbol(IDStr); if (Lexer.isNot(AsmToken::Comma)) return TokError("unexpected token in directive"); @@ -1410,7 +1535,7 @@ bool AsmParser::ParseDirectiveDarwinZerofill() { return Error(SizeLoc, "invalid '.zerofill' directive size, can't be less " "than zero"); - // NOTE: The alignment in the directive is a power of 2 value, the assember + // NOTE: The alignment in the directive is a power of 2 value, the assembler // may internally end up wanting an alignment in bytes. // FIXME: Diagnose overflow. if (Pow2Alignment < 0) @@ -1431,6 +1556,60 @@ bool AsmParser::ParseDirectiveDarwinZerofill() { return false; } +/// ParseDirectiveDarwinTBSS +/// ::= .tbss identifier, size, align +bool AsmParser::ParseDirectiveDarwinTBSS() { + SMLoc IDLoc = Lexer.getLoc(); + StringRef Name; + if (ParseIdentifier(Name)) + return TokError("expected identifier in directive"); + + // Handle the identifier as the key symbol. + MCSymbol *Sym = CreateSymbol(Name); + + if (Lexer.isNot(AsmToken::Comma)) + return TokError("unexpected token in directive"); + Lex(); + + int64_t Size; + SMLoc SizeLoc = Lexer.getLoc(); + if (ParseAbsoluteExpression(Size)) + return true; + + int64_t Pow2Alignment = 0; + SMLoc Pow2AlignmentLoc; + if (Lexer.is(AsmToken::Comma)) { + Lex(); + Pow2AlignmentLoc = Lexer.getLoc(); + if (ParseAbsoluteExpression(Pow2Alignment)) + return true; + } + + if (Lexer.isNot(AsmToken::EndOfStatement)) + return TokError("unexpected token in '.tbss' directive"); + + Lex(); + + if (Size < 0) + return Error(SizeLoc, "invalid '.tbss' directive size, can't be less than" + "zero"); + + // FIXME: Diagnose overflow. + if (Pow2Alignment < 0) + return Error(Pow2AlignmentLoc, "invalid '.tbss' alignment, can't be less" + "than zero"); + + if (!Sym->isUndefined()) + return Error(IDLoc, "invalid symbol redefinition"); + + Out.EmitTBSSSymbol(Ctx.getMachOSection("__DATA", "__thread_bss", + MCSectionMachO::S_THREAD_LOCAL_ZEROFILL, + 0, SectionKind::getThreadBSS()), + Sym, Size, 1 << Pow2Alignment); + + return false; +} + /// ParseDirectiveDarwinSubsectionsViaSymbols /// ::= .subsections_via_symbols bool AsmParser::ParseDirectiveDarwinSubsectionsViaSymbols() { diff --git a/lib/MC/MCSection.cpp b/lib/MC/MCSection.cpp index f6e9636..a792d56 100644 --- a/lib/MC/MCSection.cpp +++ b/lib/MC/MCSection.cpp @@ -20,30 +20,3 @@ using namespace llvm; MCSection::~MCSection() { } -//===----------------------------------------------------------------------===// -// MCSectionCOFF -//===----------------------------------------------------------------------===// - -MCSectionCOFF *MCSectionCOFF:: -Create(StringRef Name, bool IsDirective, SectionKind K, MCContext &Ctx) { - char *NameCopy = static_cast<char*>( - Ctx.Allocate(Name.size(), /*Alignment=*/1)); - memcpy(NameCopy, Name.data(), Name.size()); - return new (Ctx) MCSectionCOFF(StringRef(NameCopy, Name.size()), - IsDirective, K); -} - -void MCSectionCOFF::PrintSwitchToSection(const MCAsmInfo &MAI, - raw_ostream &OS) const { - - if (isDirective()) { - OS << getName() << '\n'; - return; - } - OS << "\t.section\t" << getName() << ",\""; - if (getKind().isText()) - OS << 'x'; - if (getKind().isWriteable()) - OS << 'w'; - OS << "\"\n"; -} diff --git a/lib/MC/MCSectionCOFF.cpp b/lib/MC/MCSectionCOFF.cpp new file mode 100644 index 0000000..d57bb0c --- /dev/null +++ b/lib/MC/MCSectionCOFF.cpp @@ -0,0 +1,76 @@ +//===- lib/MC/MCSectionCOFF.cpp - COFF Code Section Representation --------===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// + +#include "llvm/MC/MCSectionCOFF.h" +#include "llvm/MC/MCAsmInfo.h" +#include "llvm/MC/MCContext.h" +#include "llvm/MC/MCSymbol.h" +#include "llvm/Support/raw_ostream.h" +using namespace llvm; + +MCSectionCOFF::~MCSectionCOFF() {} // anchor. + +// ShouldOmitSectionDirective - Decides whether a '.section' directive +// should be printed before the section name +bool MCSectionCOFF::ShouldOmitSectionDirective(StringRef Name, + const MCAsmInfo &MAI) const { + + // FIXME: Does .section .bss/.data/.text work everywhere?? + if (Name == ".text" || Name == ".data" || Name == ".bss") + return true; + + return false; +} + +void MCSectionCOFF::PrintSwitchToSection(const MCAsmInfo &MAI, + raw_ostream &OS) const { + + // standard sections don't require the '.section' + if (ShouldOmitSectionDirective(SectionName, MAI)) { + OS << '\t' << getSectionName() << '\n'; + return; + } + + OS << "\t.section\t" << getSectionName() << ",\""; + if (getKind().isText()) + OS << 'x'; + if (getKind().isWriteable()) + OS << 'w'; + else + OS << 'r'; + if (getCharacteristics() & MCSectionCOFF::IMAGE_SCN_MEM_DISCARDABLE) + OS << 'n'; + OS << "\"\n"; + + if (getCharacteristics() & MCSectionCOFF::IMAGE_SCN_LNK_COMDAT) { + switch (Selection) { + case IMAGE_COMDAT_SELECT_NODUPLICATES: + OS << "\t.linkonce one_only\n"; + break; + case IMAGE_COMDAT_SELECT_ANY: + OS << "\t.linkonce discard\n"; + break; + case IMAGE_COMDAT_SELECT_SAME_SIZE: + OS << "\t.linkonce same_size\n"; + break; + case IMAGE_COMDAT_SELECT_EXACT_MATCH: + OS << "\t.linkonce same_contents\n"; + break; + //NOTE: as of binutils 2.20, there is no way to specifiy select largest + // with the .linkonce directive. For now, we treat it as an invalid + // comdat selection value. + case IMAGE_COMDAT_SELECT_LARGEST: + // OS << "\t.linkonce largest\n"; + // break; + default: + assert (0 && "unsupported COFF selection type"); + break; + } + } +} diff --git a/lib/MC/MCSectionMachO.cpp b/lib/MC/MCSectionMachO.cpp index 3a18cee..ded3b20 100644 --- a/lib/MC/MCSectionMachO.cpp +++ b/lib/MC/MCSectionMachO.cpp @@ -34,7 +34,14 @@ static const struct { { "interposing", "S_INTERPOSING" }, // 0x0D { "16byte_literals", "S_16BYTE_LITERALS" }, // 0x0E { 0, /*FIXME??*/ "S_DTRACE_DOF" }, // 0x0F - { 0, /*FIXME??*/ "S_LAZY_DYLIB_SYMBOL_POINTERS" } // 0x10 + { 0, /*FIXME??*/ "S_LAZY_DYLIB_SYMBOL_POINTERS" }, // 0x10 + { "thread_local_regular", "S_THREAD_LOCAL_REGULAR" }, // 0x11 + { "thread_local_zerofill", "S_THREAD_LOCAL_ZEROFILL" }, // 0x12 + { "thread_local_variables", "S_THREAD_LOCAL_VARIABLES" }, // 0x13 + { "thread_local_variable_pointers", + "S_THREAD_LOCAL_VARIABLE_POINTERS" }, // 0x14 + { "thread_local_init_function_pointers", + "S_THREAD_LOCAL_INIT_FUNCTION_POINTERS"}, // 0x15 }; @@ -66,7 +73,7 @@ ENTRY(0 /*FIXME*/, S_ATTR_LOC_RELOC) MCSectionMachO::MCSectionMachO(StringRef Segment, StringRef Section, unsigned TAA, unsigned reserved2, SectionKind K) - : MCSection(K), TypeAndAttributes(TAA), Reserved2(reserved2) { + : MCSection(SV_MachO, K), TypeAndAttributes(TAA), Reserved2(reserved2) { assert(Segment.size() <= 16 && Section.size() <= 16 && "Segment or section string too long"); for (unsigned i = 0; i != 16; ++i) { diff --git a/lib/MC/MCStreamer.cpp b/lib/MC/MCStreamer.cpp index 4f484a2..573f2a3 100644 --- a/lib/MC/MCStreamer.cpp +++ b/lib/MC/MCStreamer.cpp @@ -48,7 +48,7 @@ void MCStreamer::EmitFill(uint64_t NumBytes, uint8_t FillValue, EmitValue(E, 1, AddrSpace); } -/// EmitRawText - If this file is backed by a assembly streamer, this dumps +/// EmitRawText - If this file is backed by an assembly streamer, this dumps /// the specified string in the output .s file. This capability is /// indicated by the hasRawTextSupport() predicate. void MCStreamer::EmitRawText(StringRef String) { diff --git a/lib/MC/MCSymbol.cpp b/lib/MC/MCSymbol.cpp index 3fb1233..07751f7 100644 --- a/lib/MC/MCSymbol.cpp +++ b/lib/MC/MCSymbol.cpp @@ -8,6 +8,7 @@ //===----------------------------------------------------------------------===// #include "llvm/MC/MCSymbol.h" +#include "llvm/MC/MCExpr.h" #include "llvm/Support/Debug.h" #include "llvm/Support/raw_ostream.h" using namespace llvm; @@ -38,6 +39,17 @@ static bool NameNeedsQuoting(StringRef Str) { return false; } +void MCSymbol::setVariableValue(const MCExpr *Value) { + assert(Value && "Invalid variable value!"); + assert((isUndefined() || (isAbsolute() && isa<MCConstantExpr>(Value))) && + "Invalid redefinition!"); + this->Value = Value; + + // Mark the variable as absolute as appropriate. + if (isa<MCConstantExpr>(Value)) + setAbsolute(); +} + void MCSymbol::print(raw_ostream &OS) const { // The name for this MCSymbol is required to be a valid target name. However, // some targets support quoting names with funny characters. If the name diff --git a/lib/MC/MachObjectWriter.cpp b/lib/MC/MachObjectWriter.cpp index a533ccf..3207e99 100644 --- a/lib/MC/MachObjectWriter.cpp +++ b/lib/MC/MachObjectWriter.cpp @@ -16,6 +16,7 @@ #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/Support/ErrorHandling.h" #include "llvm/Support/MachO.h" @@ -58,6 +59,20 @@ static bool isFixupKindRIPRel(unsigned Kind) { Kind == X86::reloc_riprel_4byte_movq_load; } +static bool doesSymbolRequireExternRelocation(MCSymbolData *SD) { + // Undefined symbols are always extern. + if (SD->Symbol->isUndefined()) + return true; + + // 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; + + // Otherwise, we can use an internal relocation. + return false; +} + namespace { class MachObjectWriterImpl { @@ -130,7 +145,8 @@ class MachObjectWriterImpl { RIT_Pair = 1, RIT_Difference = 2, RIT_PreboundLazyPointer = 3, - RIT_LocalDifference = 4 + RIT_LocalDifference = 4, + RIT_TLV = 5 }; /// X86_64 uses its own relocation types. @@ -143,7 +159,8 @@ class MachObjectWriterImpl { RIT_X86_64_Subtractor = 5, RIT_X86_64_Signed1 = 6, RIT_X86_64_Signed2 = 7, - RIT_X86_64_Signed4 = 8 + RIT_X86_64_Signed4 = 8, + RIT_X86_64_TLV = 9 }; /// MachSymbolData - Helper struct for containing some precomputed information @@ -155,8 +172,8 @@ class MachObjectWriterImpl { // 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(); } }; @@ -170,6 +187,7 @@ class MachObjectWriterImpl { llvm::DenseMap<const MCSectionData*, std::vector<MachRelocationEntry> > Relocations; + llvm::DenseMap<const MCSectionData*, unsigned> IndirectSymBase; /// @} /// @name Symbol Table Data @@ -289,9 +307,7 @@ public: 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) { @@ -312,7 +328,7 @@ public: 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 @@ -404,7 +420,7 @@ public: // 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); } @@ -456,14 +472,17 @@ public: 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; @@ -532,7 +551,7 @@ public: Type = RIT_X86_64_Unsigned; MachRelocationEntry MRE; - MRE.Word0 = Address; + MRE.Word0 = FixupOffset; MRE.Word1 = ((Index << 0) | (IsPCRel << 24) | (Log2Size << 25) | @@ -548,6 +567,17 @@ public: MCSymbolData &SD = Asm.getSymbolData(*Symbol); const MCSymbolData *Base = Asm.getAtom(Layout, &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 // non-local symbol). @@ -558,14 +588,17 @@ public: // 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(); @@ -575,14 +608,37 @@ public: // 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) + if (unsigned(Fixup.getKind()) == X86::reloc_riprel_4byte_movq_load) Type = RIT_X86_64_GOTLoad; else Type = RIT_X86_64_GOT; - } else if (Modifier != MCSymbolRefExpr::VK_None) + } else if (Modifier == MCSymbolRefExpr::VK_TLVP) { + Type = RIT_X86_64_TLV; + } else if (Modifier != MCSymbolRefExpr::VK_None) { report_fatal_error("unsupported symbol modifier in relocation"); - else + } else { Type = 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 = RIT_X86_64_Signed1; break; + case 2: Type = RIT_X86_64_Signed2; break; + case 4: Type = RIT_X86_64_Signed4; break; + } + } } else { if (Modifier != MCSymbolRefExpr::VK_None) report_fatal_error("unsupported symbol modifier in branch " @@ -590,27 +646,6 @@ public: 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() + (1LL << 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; - } - } } else { if (Modifier == MCSymbolRefExpr::VK_GOT) { Type = RIT_X86_64_GOT; @@ -621,6 +656,8 @@ public: // required to include any necessary offset directly. Type = 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 @@ -633,7 +670,7 @@ public: // struct relocation_info (8 bytes) MachRelocationEntry MRE; - MRE.Word0 = Address; + MRE.Word0 = FixupOffset; MRE.Word1 = ((Index << 0) | (IsPCRel << 24) | (Log2Size << 25) | @@ -645,11 +682,11 @@ public: 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); + uint32_t FixupOffset = Layout.getFragmentOffset(Fragment)+Fixup.getOffset(); + unsigned IsPCRel = isFixupKindPCRel(Fixup.getKind()); + unsigned Log2Size = getFixupKindLog2Size(Fixup.getKind()); unsigned Type = RIT_Vanilla; // See <reloc.h>. @@ -692,40 +729,49 @@ public: } MachRelocationEntry MRE; - MRE.Word0 = ((Address << 0) | - (Type << 24) | - (Log2Size << 28) | - (IsPCRel << 30) | + MRE.Word0 = ((FixupOffset << 0) | + (Type << 24) | + (Log2Size << 28) | + (IsPCRel << 30) | RF_Scattered); MRE.Word1 = Value; 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 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 FixupOffset = Layout.getFragmentOffset(Fragment)+Fixup.getOffset(); uint32_t Value = 0; unsigned Index = 0; unsigned IsExtern = 0; @@ -739,12 +785,15 @@ public: Type = RIT_Vanilla; Value = 0; } 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(); + // 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); Value = 0; } else { // The index is the section ordinal (1-based). @@ -757,7 +806,7 @@ public: // struct relocation_info (8 bytes) MachRelocationEntry MRE; - MRE.Word0 = Address; + MRE.Word0 = FixupOffset; MRE.Word1 = ((Index << 0) | (IsPCRel << 24) | (Log2Size << 25) | @@ -775,29 +824,37 @@ public: // 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. @@ -1111,7 +1168,7 @@ void MachObjectWriter::ExecutePostLayoutBinding(MCAssembler &Asm) { void MachObjectWriter::RecordRelocation(const MCAssembler &Asm, const MCAsmLayout &Layout, const MCFragment *Fragment, - const MCAsmFixup &Fixup, MCValue Target, + const MCFixup &Fixup, MCValue Target, uint64_t &FixedValue) { ((MachObjectWriterImpl*) Impl)->RecordRelocation(Asm, Layout, Fragment, Fixup, Target, FixedValue); |