1 //===- MachOObjectFile.cpp - Mach-O object file binding ---------*- C++ -*-===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file defines the MachOObjectFile class, which binds the MachOObject
11 // class to the generic ObjectFile wrapper.
13 //===----------------------------------------------------------------------===//
15 #include "llvm/Object/MachO.h"
16 #include "llvm/ADT/Triple.h"
17 #include "llvm/Object/MachOFormat.h"
18 #include "llvm/Support/Format.h"
19 #include "llvm/Support/MemoryBuffer.h"
25 using namespace object;
30 MachOObjectFile::MachOObjectFile(MemoryBuffer *Object, MachOObject *MOO,
32 : ObjectFile(Binary::ID_MachO, Object, ec),
34 RegisteredStringTable(std::numeric_limits<uint32_t>::max()) {
36 moveToNextSection(DRI);
37 uint32_t LoadCommandCount = MachOObj->getHeader().NumLoadCommands;
38 while (DRI.d.a < LoadCommandCount) {
39 Sections.push_back(DRI);
41 moveToNextSection(DRI);
46 ObjectFile *ObjectFile::createMachOObjectFile(MemoryBuffer *Buffer) {
49 MachOObject *MachOObj = MachOObject::LoadFromBuffer(Buffer, &Err);
52 // MachOObject takes ownership of the Buffer we passed to it, and
53 // MachOObjectFile does, too, so we need to make sure they don't get the
54 // same object. A MemoryBuffer is cheap (it's just a reference to memory,
55 // not a copy of the memory itself), so just make a new copy here for
56 // the MachOObjectFile.
57 MemoryBuffer *NewBuffer =
58 MemoryBuffer::getMemBuffer(Buffer->getBuffer(),
59 Buffer->getBufferIdentifier(), false);
60 return new MachOObjectFile(NewBuffer, MachOObj, ec);
63 /*===-- Symbols -----------------------------------------------------------===*/
65 void MachOObjectFile::moveToNextSymbol(DataRefImpl &DRI) const {
66 uint32_t LoadCommandCount = MachOObj->getHeader().NumLoadCommands;
67 while (DRI.d.a < LoadCommandCount) {
68 LoadCommandInfo LCI = MachOObj->getLoadCommandInfo(DRI.d.a);
69 if (LCI.Command.Type == macho::LCT_Symtab) {
70 InMemoryStruct<macho::SymtabLoadCommand> SymtabLoadCmd;
71 MachOObj->ReadSymtabLoadCommand(LCI, SymtabLoadCmd);
72 if (DRI.d.b < SymtabLoadCmd->NumSymbolTableEntries)
81 void MachOObjectFile::getSymbolTableEntry(DataRefImpl DRI,
82 InMemoryStruct<macho::SymbolTableEntry> &Res) const {
83 InMemoryStruct<macho::SymtabLoadCommand> SymtabLoadCmd;
84 LoadCommandInfo LCI = MachOObj->getLoadCommandInfo(DRI.d.a);
85 MachOObj->ReadSymtabLoadCommand(LCI, SymtabLoadCmd);
87 if (RegisteredStringTable != DRI.d.a) {
88 MachOObj->RegisterStringTable(*SymtabLoadCmd);
89 RegisteredStringTable = DRI.d.a;
92 MachOObj->ReadSymbolTableEntry(SymtabLoadCmd->SymbolTableOffset, DRI.d.b,
96 void MachOObjectFile::getSymbol64TableEntry(DataRefImpl DRI,
97 InMemoryStruct<macho::Symbol64TableEntry> &Res) const {
98 InMemoryStruct<macho::SymtabLoadCommand> SymtabLoadCmd;
99 LoadCommandInfo LCI = MachOObj->getLoadCommandInfo(DRI.d.a);
100 MachOObj->ReadSymtabLoadCommand(LCI, SymtabLoadCmd);
102 if (RegisteredStringTable != DRI.d.a) {
103 MachOObj->RegisterStringTable(*SymtabLoadCmd);
104 RegisteredStringTable = DRI.d.a;
107 MachOObj->ReadSymbol64TableEntry(SymtabLoadCmd->SymbolTableOffset, DRI.d.b,
112 error_code MachOObjectFile::getSymbolNext(DataRefImpl DRI,
113 SymbolRef &Result) const {
115 moveToNextSymbol(DRI);
116 Result = SymbolRef(DRI, this);
117 return object_error::success;
120 error_code MachOObjectFile::getSymbolName(DataRefImpl DRI,
121 StringRef &Result) const {
122 if (MachOObj->is64Bit()) {
123 InMemoryStruct<macho::Symbol64TableEntry> Entry;
124 getSymbol64TableEntry(DRI, Entry);
125 Result = MachOObj->getStringAtIndex(Entry->StringIndex);
127 InMemoryStruct<macho::SymbolTableEntry> Entry;
128 getSymbolTableEntry(DRI, Entry);
129 Result = MachOObj->getStringAtIndex(Entry->StringIndex);
131 return object_error::success;
134 error_code MachOObjectFile::getSymbolFileOffset(DataRefImpl DRI,
135 uint64_t &Result) const {
136 if (MachOObj->is64Bit()) {
137 InMemoryStruct<macho::Symbol64TableEntry> Entry;
138 getSymbol64TableEntry(DRI, Entry);
139 Result = Entry->Value;
140 if (Entry->SectionIndex) {
141 const MachOFormat::Section64 *Section =
142 getSection64(Sections[Entry->SectionIndex-1]);
143 Result += Section->Offset - Section->Address;
146 InMemoryStruct<macho::SymbolTableEntry> Entry;
147 getSymbolTableEntry(DRI, Entry);
148 Result = Entry->Value;
149 if (Entry->SectionIndex) {
150 const MachOFormat::Section *Section =
151 getSection(Sections[Entry->SectionIndex-1]);
152 Result += Section->Offset - Section->Address;
156 return object_error::success;
159 error_code MachOObjectFile::getSymbolAddress(DataRefImpl DRI,
160 uint64_t &Result) const {
161 if (MachOObj->is64Bit()) {
162 InMemoryStruct<macho::Symbol64TableEntry> Entry;
163 getSymbol64TableEntry(DRI, Entry);
164 Result = Entry->Value;
166 InMemoryStruct<macho::SymbolTableEntry> Entry;
167 getSymbolTableEntry(DRI, Entry);
168 Result = Entry->Value;
170 return object_error::success;
173 error_code MachOObjectFile::getSymbolSize(DataRefImpl DRI,
174 uint64_t &Result) const {
175 uint32_t LoadCommandCount = MachOObj->getHeader().NumLoadCommands;
176 uint64_t BeginOffset;
177 uint64_t EndOffset = 0;
178 uint8_t SectionIndex;
179 if (MachOObj->is64Bit()) {
180 InMemoryStruct<macho::Symbol64TableEntry> Entry;
181 getSymbol64TableEntry(DRI, Entry);
182 BeginOffset = Entry->Value;
183 SectionIndex = Entry->SectionIndex;
185 uint32_t flags = SymbolRef::SF_None;
186 getSymbolFlags(DRI, flags);
187 if (flags & SymbolRef::SF_Common)
188 Result = Entry->Value;
190 Result = UnknownAddressOrSize;
191 return object_error::success;
193 // Unfortunately symbols are unsorted so we need to touch all
194 // symbols from load command
196 uint32_t Command = DRI.d.a;
197 while (Command == DRI.d.a) {
198 moveToNextSymbol(DRI);
199 if (DRI.d.a < LoadCommandCount) {
200 getSymbol64TableEntry(DRI, Entry);
201 if (Entry->SectionIndex == SectionIndex && Entry->Value > BeginOffset)
202 if (!EndOffset || Entry->Value < EndOffset)
203 EndOffset = Entry->Value;
208 InMemoryStruct<macho::SymbolTableEntry> Entry;
209 getSymbolTableEntry(DRI, Entry);
210 BeginOffset = Entry->Value;
211 SectionIndex = Entry->SectionIndex;
213 uint32_t flags = SymbolRef::SF_None;
214 getSymbolFlags(DRI, flags);
215 if (flags & SymbolRef::SF_Common)
216 Result = Entry->Value;
218 Result = UnknownAddressOrSize;
219 return object_error::success;
221 // Unfortunately symbols are unsorted so we need to touch all
222 // symbols from load command
224 uint32_t Command = DRI.d.a;
225 while (Command == DRI.d.a) {
226 moveToNextSymbol(DRI);
227 if (DRI.d.a < LoadCommandCount) {
228 getSymbolTableEntry(DRI, Entry);
229 if (Entry->SectionIndex == SectionIndex && Entry->Value > BeginOffset)
230 if (!EndOffset || Entry->Value < EndOffset)
231 EndOffset = Entry->Value;
238 getSectionSize(Sections[SectionIndex-1], Size);
239 getSectionAddress(Sections[SectionIndex-1], EndOffset);
242 Result = EndOffset - BeginOffset;
243 return object_error::success;
246 error_code MachOObjectFile::getSymbolNMTypeChar(DataRefImpl DRI,
247 char &Result) const {
249 if (MachOObj->is64Bit()) {
250 InMemoryStruct<macho::Symbol64TableEntry> Entry;
251 getSymbol64TableEntry(DRI, Entry);
253 Flags = Entry->Flags;
255 InMemoryStruct<macho::SymbolTableEntry> Entry;
256 getSymbolTableEntry(DRI, Entry);
258 Flags = Entry->Flags;
262 switch (Type & macho::STF_TypeMask) {
263 case macho::STT_Undefined:
266 case macho::STT_Absolute:
267 case macho::STT_Section:
275 if (Flags & (macho::STF_External | macho::STF_PrivateExtern))
276 Char = toupper(static_cast<unsigned char>(Char));
278 return object_error::success;
281 error_code MachOObjectFile::getSymbolFlags(DataRefImpl DRI,
282 uint32_t &Result) const {
285 if (MachOObj->is64Bit()) {
286 InMemoryStruct<macho::Symbol64TableEntry> Entry;
287 getSymbol64TableEntry(DRI, Entry);
288 MachOFlags = Entry->Flags;
289 MachOType = Entry->Type;
291 InMemoryStruct<macho::SymbolTableEntry> Entry;
292 getSymbolTableEntry(DRI, Entry);
293 MachOFlags = Entry->Flags;
294 MachOType = Entry->Type;
297 // TODO: Correctly set SF_ThreadLocal
298 Result = SymbolRef::SF_None;
300 if ((MachOType & MachO::NlistMaskType) == MachO::NListTypeUndefined)
301 Result |= SymbolRef::SF_Undefined;
303 if (MachOFlags & macho::STF_StabsEntryMask)
304 Result |= SymbolRef::SF_FormatSpecific;
306 if (MachOType & MachO::NlistMaskExternal) {
307 Result |= SymbolRef::SF_Global;
308 if ((MachOType & MachO::NlistMaskType) == MachO::NListTypeUndefined)
309 Result |= SymbolRef::SF_Common;
312 if (MachOFlags & (MachO::NListDescWeakRef | MachO::NListDescWeakDef))
313 Result |= SymbolRef::SF_Weak;
315 if ((MachOType & MachO::NlistMaskType) == MachO::NListTypeAbsolute)
316 Result |= SymbolRef::SF_Absolute;
318 return object_error::success;
321 error_code MachOObjectFile::getSymbolSection(DataRefImpl Symb,
322 section_iterator &Res) const {
324 if (MachOObj->is64Bit()) {
325 InMemoryStruct<macho::Symbol64TableEntry> Entry;
326 getSymbol64TableEntry(Symb, Entry);
327 index = Entry->SectionIndex;
329 InMemoryStruct<macho::SymbolTableEntry> Entry;
330 getSymbolTableEntry(Symb, Entry);
331 index = Entry->SectionIndex;
335 Res = end_sections();
337 Res = section_iterator(SectionRef(Sections[index-1], this));
339 return object_error::success;
342 error_code MachOObjectFile::getSymbolType(DataRefImpl Symb,
343 SymbolRef::Type &Res) const {
345 if (MachOObj->is64Bit()) {
346 InMemoryStruct<macho::Symbol64TableEntry> Entry;
347 getSymbol64TableEntry(Symb, Entry);
348 n_type = Entry->Type;
350 InMemoryStruct<macho::SymbolTableEntry> Entry;
351 getSymbolTableEntry(Symb, Entry);
352 n_type = Entry->Type;
354 Res = SymbolRef::ST_Other;
356 // If this is a STAB debugging symbol, we can do nothing more.
357 if (n_type & MachO::NlistMaskStab) {
358 Res = SymbolRef::ST_Debug;
359 return object_error::success;
362 switch (n_type & MachO::NlistMaskType) {
363 case MachO::NListTypeUndefined :
364 Res = SymbolRef::ST_Unknown;
366 case MachO::NListTypeSection :
367 Res = SymbolRef::ST_Function;
370 return object_error::success;
373 error_code MachOObjectFile::getSymbolValue(DataRefImpl Symb,
374 uint64_t &Val) const {
375 report_fatal_error("getSymbolValue unimplemented in MachOObjectFile");
378 symbol_iterator MachOObjectFile::begin_symbols() const {
379 // DRI.d.a = segment number; DRI.d.b = symbol index.
381 moveToNextSymbol(DRI);
382 return symbol_iterator(SymbolRef(DRI, this));
385 symbol_iterator MachOObjectFile::end_symbols() const {
387 DRI.d.a = MachOObj->getHeader().NumLoadCommands;
388 return symbol_iterator(SymbolRef(DRI, this));
391 symbol_iterator MachOObjectFile::begin_dynamic_symbols() const {
393 report_fatal_error("Dynamic symbols unimplemented in MachOObjectFile");
396 symbol_iterator MachOObjectFile::end_dynamic_symbols() const {
398 report_fatal_error("Dynamic symbols unimplemented in MachOObjectFile");
401 library_iterator MachOObjectFile::begin_libraries_needed() const {
403 report_fatal_error("Needed libraries unimplemented in MachOObjectFile");
406 library_iterator MachOObjectFile::end_libraries_needed() const {
408 report_fatal_error("Needed libraries unimplemented in MachOObjectFile");
411 StringRef MachOObjectFile::getLoadName() const {
413 report_fatal_error("get_load_name() unimplemented in MachOObjectFile");
416 /*===-- Sections ----------------------------------------------------------===*/
418 void MachOObjectFile::moveToNextSection(DataRefImpl &DRI) const {
419 uint32_t LoadCommandCount = MachOObj->getHeader().NumLoadCommands;
420 while (DRI.d.a < LoadCommandCount) {
421 LoadCommandInfo LCI = MachOObj->getLoadCommandInfo(DRI.d.a);
422 if (LCI.Command.Type == macho::LCT_Segment) {
423 InMemoryStruct<macho::SegmentLoadCommand> SegmentLoadCmd;
424 MachOObj->ReadSegmentLoadCommand(LCI, SegmentLoadCmd);
425 if (DRI.d.b < SegmentLoadCmd->NumSections)
427 } else if (LCI.Command.Type == macho::LCT_Segment64) {
428 InMemoryStruct<macho::Segment64LoadCommand> Segment64LoadCmd;
429 MachOObj->ReadSegment64LoadCommand(LCI, Segment64LoadCmd);
430 if (DRI.d.b < Segment64LoadCmd->NumSections)
439 error_code MachOObjectFile::getSectionNext(DataRefImpl DRI,
440 SectionRef &Result) const {
442 moveToNextSection(DRI);
443 Result = SectionRef(DRI, this);
444 return object_error::success;
447 static bool is64BitLoadCommand(const MachOObject *MachOObj, DataRefImpl DRI) {
448 LoadCommandInfo LCI = MachOObj->getLoadCommandInfo(DRI.d.a);
449 if (LCI.Command.Type == macho::LCT_Segment64)
451 assert(LCI.Command.Type == macho::LCT_Segment && "Unexpected Type.");
455 const MachOFormat::Section *MachOObjectFile::getSection(DataRefImpl DRI) const {
456 assert(!is64BitLoadCommand(MachOObj.get(), DRI));
457 LoadCommandInfo LCI = MachOObj->getLoadCommandInfo(DRI.d.a);
458 unsigned SectionOffset = LCI.Offset + sizeof(macho::SegmentLoadCommand) +
459 DRI.d.b * sizeof(MachOFormat::Section);
460 StringRef Data = MachOObj->getData(SectionOffset, sizeof(MachOFormat::Section));
461 return reinterpret_cast<const MachOFormat::Section*>(Data.data());
464 std::size_t MachOObjectFile::getSectionIndex(DataRefImpl Sec) const {
465 SectionList::const_iterator loc =
466 std::find(Sections.begin(), Sections.end(), Sec);
467 assert(loc != Sections.end() && "Sec is not a valid section!");
468 return std::distance(Sections.begin(), loc);
471 const MachOFormat::Section64 *
472 MachOObjectFile::getSection64(DataRefImpl DRI) const {
473 assert(is64BitLoadCommand(MachOObj.get(), DRI));
474 LoadCommandInfo LCI = MachOObj->getLoadCommandInfo(DRI.d.a);
475 unsigned SectionOffset = LCI.Offset + sizeof(macho::Segment64LoadCommand) +
476 DRI.d.b * sizeof(MachOFormat::Section64);
477 StringRef Data = MachOObj->getData(SectionOffset, sizeof(MachOFormat::Section64));
478 return reinterpret_cast<const MachOFormat::Section64*>(Data.data());
481 static StringRef parseSegmentOrSectionName(const char *P) {
485 // Not null terminated, so this is a 16 char string.
486 return StringRef(P, 16);
489 ArrayRef<char> MachOObjectFile::getSectionRawName(DataRefImpl DRI) const {
490 if (is64BitLoadCommand(MachOObj.get(), DRI)) {
491 const MachOFormat::Section64 *sec = getSection64(DRI);
492 return ArrayRef<char>(sec->Name);
494 const MachOFormat::Section *sec = getSection(DRI);
495 return ArrayRef<char>(sec->Name);
499 error_code MachOObjectFile::getSectionName(DataRefImpl DRI,
500 StringRef &Result) const {
501 ArrayRef<char> Raw = getSectionRawName(DRI);
502 Result = parseSegmentOrSectionName(Raw.data());
503 return object_error::success;
507 MachOObjectFile::getSectionRawFinalSegmentName(DataRefImpl Sec) const {
508 if (is64BitLoadCommand(MachOObj.get(), Sec)) {
509 const MachOFormat::Section64 *sec = getSection64(Sec);
510 return ArrayRef<char>(sec->SegmentName, 16);
512 const MachOFormat::Section *sec = getSection(Sec);
513 return ArrayRef<char>(sec->SegmentName);
517 StringRef MachOObjectFile::getSectionFinalSegmentName(DataRefImpl DRI) const {
518 ArrayRef<char> Raw = getSectionRawFinalSegmentName(DRI);
519 return parseSegmentOrSectionName(Raw.data());
522 error_code MachOObjectFile::getSectionAddress(DataRefImpl DRI,
523 uint64_t &Result) const {
524 if (is64BitLoadCommand(MachOObj.get(), DRI)) {
525 const MachOFormat::Section64 *Sect = getSection64(DRI);
526 Result = Sect->Address;
528 const MachOFormat::Section *Sect = getSection(DRI);
529 Result = Sect->Address;
531 return object_error::success;
534 error_code MachOObjectFile::getSectionSize(DataRefImpl DRI,
535 uint64_t &Result) const {
536 if (is64BitLoadCommand(MachOObj.get(), DRI)) {
537 const MachOFormat::Section64 *Sect = getSection64(DRI);
540 const MachOFormat::Section *Sect = getSection(DRI);
543 return object_error::success;
546 error_code MachOObjectFile::getSectionContents(DataRefImpl DRI,
547 StringRef &Result) const {
548 if (is64BitLoadCommand(MachOObj.get(), DRI)) {
549 const MachOFormat::Section64 *Sect = getSection64(DRI);
550 Result = MachOObj->getData(Sect->Offset, Sect->Size);
552 const MachOFormat::Section *Sect = getSection(DRI);
553 Result = MachOObj->getData(Sect->Offset, Sect->Size);
555 return object_error::success;
558 error_code MachOObjectFile::getSectionAlignment(DataRefImpl DRI,
559 uint64_t &Result) const {
560 if (is64BitLoadCommand(MachOObj.get(), DRI)) {
561 const MachOFormat::Section64 *Sect = getSection64(DRI);
562 Result = uint64_t(1) << Sect->Align;
564 const MachOFormat::Section *Sect = getSection(DRI);
565 Result = uint64_t(1) << Sect->Align;
567 return object_error::success;
570 error_code MachOObjectFile::isSectionText(DataRefImpl DRI,
571 bool &Result) const {
572 if (is64BitLoadCommand(MachOObj.get(), DRI)) {
573 const MachOFormat::Section64 *Sect = getSection64(DRI);
574 Result = Sect->Flags & macho::SF_PureInstructions;
576 const MachOFormat::Section *Sect = getSection(DRI);
577 Result = Sect->Flags & macho::SF_PureInstructions;
579 return object_error::success;
582 error_code MachOObjectFile::isSectionData(DataRefImpl DRI,
583 bool &Result) const {
584 // FIXME: Unimplemented.
586 return object_error::success;
589 error_code MachOObjectFile::isSectionBSS(DataRefImpl DRI,
590 bool &Result) const {
591 // FIXME: Unimplemented.
593 return object_error::success;
596 error_code MachOObjectFile::isSectionRequiredForExecution(DataRefImpl Sec,
597 bool &Result) const {
598 // FIXME: Unimplemented.
600 return object_error::success;
603 error_code MachOObjectFile::isSectionVirtual(DataRefImpl Sec,
604 bool &Result) const {
605 // FIXME: Unimplemented.
607 return object_error::success;
610 error_code MachOObjectFile::isSectionZeroInit(DataRefImpl DRI,
611 bool &Result) const {
612 if (MachOObj->is64Bit()) {
613 const MachOFormat::Section64 *Sect = getSection64(DRI);
614 unsigned SectionType = Sect->Flags & MachO::SectionFlagMaskSectionType;
615 Result = (SectionType == MachO::SectionTypeZeroFill ||
616 SectionType == MachO::SectionTypeZeroFillLarge);
618 const MachOFormat::Section *Sect = getSection(DRI);
619 unsigned SectionType = Sect->Flags & MachO::SectionFlagMaskSectionType;
620 Result = (SectionType == MachO::SectionTypeZeroFill ||
621 SectionType == MachO::SectionTypeZeroFillLarge);
624 return object_error::success;
627 error_code MachOObjectFile::isSectionReadOnlyData(DataRefImpl Sec,
628 bool &Result) const {
629 // Consider using the code from isSectionText to look for __const sections.
630 // Alternately, emit S_ATTR_PURE_INSTRUCTIONS and/or S_ATTR_SOME_INSTRUCTIONS
631 // to use section attributes to distinguish code from data.
633 // FIXME: Unimplemented.
635 return object_error::success;
638 error_code MachOObjectFile::sectionContainsSymbol(DataRefImpl Sec,
640 bool &Result) const {
642 getSymbolType(Symb, ST);
643 if (ST == SymbolRef::ST_Unknown) {
645 return object_error::success;
648 uint64_t SectBegin, SectEnd;
649 getSectionAddress(Sec, SectBegin);
650 getSectionSize(Sec, SectEnd);
651 SectEnd += SectBegin;
653 if (MachOObj->is64Bit()) {
654 InMemoryStruct<macho::Symbol64TableEntry> Entry;
655 getSymbol64TableEntry(Symb, Entry);
656 uint64_t SymAddr= Entry->Value;
657 Result = (SymAddr >= SectBegin) && (SymAddr < SectEnd);
659 InMemoryStruct<macho::SymbolTableEntry> Entry;
660 getSymbolTableEntry(Symb, Entry);
661 uint64_t SymAddr= Entry->Value;
662 Result = (SymAddr >= SectBegin) && (SymAddr < SectEnd);
665 return object_error::success;
668 relocation_iterator MachOObjectFile::getSectionRelBegin(DataRefImpl Sec) const {
670 ret.d.b = getSectionIndex(Sec);
671 return relocation_iterator(RelocationRef(ret, this));
673 relocation_iterator MachOObjectFile::getSectionRelEnd(DataRefImpl Sec) const {
675 if (is64BitLoadCommand(MachOObj.get(), Sec)) {
676 const MachOFormat::Section64 *Sect = getSection64(Sec);
677 last_reloc = Sect->NumRelocationTableEntries;
679 const MachOFormat::Section *Sect = getSection(Sec);
680 last_reloc = Sect->NumRelocationTableEntries;
683 ret.d.a = last_reloc;
684 ret.d.b = getSectionIndex(Sec);
685 return relocation_iterator(RelocationRef(ret, this));
688 section_iterator MachOObjectFile::begin_sections() const {
690 moveToNextSection(DRI);
691 return section_iterator(SectionRef(DRI, this));
694 section_iterator MachOObjectFile::end_sections() const {
696 DRI.d.a = MachOObj->getHeader().NumLoadCommands;
697 return section_iterator(SectionRef(DRI, this));
700 /*===-- Relocations -------------------------------------------------------===*/
702 void MachOObjectFile::
703 getRelocation(DataRefImpl Rel,
704 InMemoryStruct<macho::RelocationEntry> &Res) const {
706 if (MachOObj->is64Bit()) {
707 const MachOFormat::Section64 *Sect = getSection64(Sections[Rel.d.b]);
708 relOffset = Sect->RelocationTableOffset;
710 const MachOFormat::Section *Sect = getSection(Sections[Rel.d.b]);
711 relOffset = Sect->RelocationTableOffset;
713 MachOObj->ReadRelocationEntry(relOffset, Rel.d.a, Res);
715 error_code MachOObjectFile::getRelocationNext(DataRefImpl Rel,
716 RelocationRef &Res) const {
718 Res = RelocationRef(Rel, this);
719 return object_error::success;
721 error_code MachOObjectFile::getRelocationAddress(DataRefImpl Rel,
722 uint64_t &Res) const {
723 const uint8_t* sectAddress = 0;
724 if (MachOObj->is64Bit()) {
725 const MachOFormat::Section64 *Sect = getSection64(Sections[Rel.d.b]);
726 sectAddress += Sect->Address;
728 const MachOFormat::Section *Sect = getSection(Sections[Rel.d.b]);
729 sectAddress += Sect->Address;
731 InMemoryStruct<macho::RelocationEntry> RE;
732 getRelocation(Rel, RE);
734 unsigned Arch = getArch();
735 bool isScattered = (Arch != Triple::x86_64) &&
736 (RE->Word0 & macho::RF_Scattered);
737 uint64_t RelAddr = 0;
739 RelAddr = RE->Word0 & 0xFFFFFF;
743 Res = reinterpret_cast<uintptr_t>(sectAddress + RelAddr);
744 return object_error::success;
746 error_code MachOObjectFile::getRelocationOffset(DataRefImpl Rel,
747 uint64_t &Res) const {
748 InMemoryStruct<macho::RelocationEntry> RE;
749 getRelocation(Rel, RE);
751 unsigned Arch = getArch();
752 bool isScattered = (Arch != Triple::x86_64) &&
753 (RE->Word0 & macho::RF_Scattered);
755 Res = RE->Word0 & 0xFFFFFF;
758 return object_error::success;
760 error_code MachOObjectFile::getRelocationSymbol(DataRefImpl Rel,
761 SymbolRef &Res) const {
762 InMemoryStruct<macho::RelocationEntry> RE;
763 getRelocation(Rel, RE);
764 uint32_t SymbolIdx = RE->Word1 & 0xffffff;
765 bool isExtern = (RE->Word1 >> 27) & 1;
768 moveToNextSymbol(Sym);
770 for (unsigned i = 0; i < SymbolIdx; i++) {
772 moveToNextSymbol(Sym);
773 assert(Sym.d.a < MachOObj->getHeader().NumLoadCommands &&
774 "Relocation symbol index out of range!");
777 Res = SymbolRef(Sym, this);
778 return object_error::success;
780 error_code MachOObjectFile::getRelocationType(DataRefImpl Rel,
781 uint64_t &Res) const {
782 InMemoryStruct<macho::RelocationEntry> RE;
783 getRelocation(Rel, RE);
787 return object_error::success;
789 error_code MachOObjectFile::getRelocationTypeName(DataRefImpl Rel,
790 SmallVectorImpl<char> &Result) const {
791 // TODO: Support scattered relocations.
793 InMemoryStruct<macho::RelocationEntry> RE;
794 getRelocation(Rel, RE);
796 unsigned Arch = getArch();
797 bool isScattered = (Arch != Triple::x86_64) &&
798 (RE->Word0 & macho::RF_Scattered);
802 r_type = (RE->Word0 >> 24) & 0xF;
804 r_type = (RE->Word1 >> 28) & 0xF;
808 static const char *const Table[] = {
809 "GENERIC_RELOC_VANILLA",
810 "GENERIC_RELOC_PAIR",
811 "GENERIC_RELOC_SECTDIFF",
812 "GENERIC_RELOC_PB_LA_PTR",
813 "GENERIC_RELOC_LOCAL_SECTDIFF",
814 "GENERIC_RELOC_TLV" };
822 case Triple::x86_64: {
823 static const char *const Table[] = {
824 "X86_64_RELOC_UNSIGNED",
825 "X86_64_RELOC_SIGNED",
826 "X86_64_RELOC_BRANCH",
827 "X86_64_RELOC_GOT_LOAD",
829 "X86_64_RELOC_SUBTRACTOR",
830 "X86_64_RELOC_SIGNED_1",
831 "X86_64_RELOC_SIGNED_2",
832 "X86_64_RELOC_SIGNED_4",
833 "X86_64_RELOC_TLV" };
842 static const char *const Table[] = {
845 "ARM_RELOC_SECTDIFF",
846 "ARM_RELOC_LOCAL_SECTDIFF",
847 "ARM_RELOC_PB_LA_PTR",
849 "ARM_THUMB_RELOC_BR22",
850 "ARM_THUMB_32BIT_BRANCH",
852 "ARM_RELOC_HALF_SECTDIFF" };
861 static const char *const Table[] = {
870 "PPC_RELOC_SECTDIFF",
871 "PPC_RELOC_PB_LA_PTR",
872 "PPC_RELOC_HI16_SECTDIFF",
873 "PPC_RELOC_LO16_SECTDIFF",
874 "PPC_RELOC_HA16_SECTDIFF",
876 "PPC_RELOC_LO14_SECTDIFF",
877 "PPC_RELOC_LOCAL_SECTDIFF" };
882 case Triple::UnknownArch:
886 Result.append(res.begin(), res.end());
887 return object_error::success;
889 error_code MachOObjectFile::getRelocationAdditionalInfo(DataRefImpl Rel,
890 int64_t &Res) const {
891 InMemoryStruct<macho::RelocationEntry> RE;
892 getRelocation(Rel, RE);
893 bool isExtern = (RE->Word1 >> 27) & 1;
896 const uint8_t* sectAddress = base();
897 if (MachOObj->is64Bit()) {
898 const MachOFormat::Section64 *Sect = getSection64(Sections[Rel.d.b]);
899 sectAddress += Sect->Offset;
901 const MachOFormat::Section *Sect = getSection(Sections[Rel.d.b]);
902 sectAddress += Sect->Offset;
904 Res = reinterpret_cast<uintptr_t>(sectAddress);
906 return object_error::success;
909 // Helper to advance a section or symbol iterator multiple increments at a time.
911 error_code advance(T &it, size_t Val) {
920 void advanceTo(T &it, size_t Val) {
921 if (error_code ec = advance(it, Val))
922 report_fatal_error(ec.message());
925 void MachOObjectFile::printRelocationTargetName(
926 InMemoryStruct<macho::RelocationEntry>& RE,
927 raw_string_ostream &fmt) const {
928 unsigned Arch = getArch();
929 bool isScattered = (Arch != Triple::x86_64) &&
930 (RE->Word0 & macho::RF_Scattered);
932 // Target of a scattered relocation is an address. In the interest of
933 // generating pretty output, scan through the symbol table looking for a
934 // symbol that aligns with that address. If we find one, print it.
935 // Otherwise, we just print the hex address of the target.
937 uint32_t Val = RE->Word1;
940 for (symbol_iterator SI = begin_symbols(), SE = end_symbols(); SI != SE;
942 if (ec) report_fatal_error(ec.message());
947 if ((ec = SI->getAddress(Addr)))
948 report_fatal_error(ec.message());
949 if (Addr != Val) continue;
950 if ((ec = SI->getName(Name)))
951 report_fatal_error(ec.message());
956 // If we couldn't find a symbol that this relocation refers to, try
957 // to find a section beginning instead.
958 for (section_iterator SI = begin_sections(), SE = end_sections(); SI != SE;
960 if (ec) report_fatal_error(ec.message());
965 if ((ec = SI->getAddress(Addr)))
966 report_fatal_error(ec.message());
967 if (Addr != Val) continue;
968 if ((ec = SI->getName(Name)))
969 report_fatal_error(ec.message());
974 fmt << format("0x%x", Val);
979 bool isExtern = (RE->Word1 >> 27) & 1;
980 uint32_t Val = RE->Word1 & 0xFFFFFF;
983 symbol_iterator SI = begin_symbols();
987 section_iterator SI = begin_sections();
995 error_code MachOObjectFile::getRelocationValueString(DataRefImpl Rel,
996 SmallVectorImpl<char> &Result) const {
997 InMemoryStruct<macho::RelocationEntry> RE;
998 getRelocation(Rel, RE);
1000 unsigned Arch = getArch();
1001 bool isScattered = (Arch != Triple::x86_64) &&
1002 (RE->Word0 & macho::RF_Scattered);
1005 raw_string_ostream fmt(fmtbuf);
1009 Type = (RE->Word0 >> 24) & 0xF;
1011 Type = (RE->Word1 >> 28) & 0xF;
1015 isPCRel = ((RE->Word0 >> 30) & 1);
1017 isPCRel = ((RE->Word1 >> 24) & 1);
1019 // Determine any addends that should be displayed with the relocation.
1020 // These require decoding the relocation type, which is triple-specific.
1022 // X86_64 has entirely custom relocation types.
1023 if (Arch == Triple::x86_64) {
1024 bool isPCRel = ((RE->Word1 >> 24) & 1);
1027 case macho::RIT_X86_64_GOTLoad: // X86_64_RELOC_GOT_LOAD
1028 case macho::RIT_X86_64_GOT: { // X86_64_RELOC_GOT
1029 printRelocationTargetName(RE, fmt);
1031 if (isPCRel) fmt << "PCREL";
1034 case macho::RIT_X86_64_Subtractor: { // X86_64_RELOC_SUBTRACTOR
1035 InMemoryStruct<macho::RelocationEntry> RENext;
1036 DataRefImpl RelNext = Rel;
1038 getRelocation(RelNext, RENext);
1040 // X86_64_SUBTRACTOR must be followed by a relocation of type
1041 // X86_64_RELOC_UNSIGNED.
1042 // NOTE: Scattered relocations don't exist on x86_64.
1043 unsigned RType = (RENext->Word1 >> 28) & 0xF;
1045 report_fatal_error("Expected X86_64_RELOC_UNSIGNED after "
1046 "X86_64_RELOC_SUBTRACTOR.");
1048 // The X86_64_RELOC_UNSIGNED contains the minuend symbol,
1049 // X86_64_SUBTRACTOR contains to the subtrahend.
1050 printRelocationTargetName(RENext, fmt);
1052 printRelocationTargetName(RE, fmt);
1055 case macho::RIT_X86_64_TLV:
1056 printRelocationTargetName(RE, fmt);
1058 if (isPCRel) fmt << "P";
1060 case macho::RIT_X86_64_Signed1: // X86_64_RELOC_SIGNED1
1061 printRelocationTargetName(RE, fmt);
1064 case macho::RIT_X86_64_Signed2: // X86_64_RELOC_SIGNED2
1065 printRelocationTargetName(RE, fmt);
1068 case macho::RIT_X86_64_Signed4: // X86_64_RELOC_SIGNED4
1069 printRelocationTargetName(RE, fmt);
1073 printRelocationTargetName(RE, fmt);
1076 // X86 and ARM share some relocation types in common.
1077 } else if (Arch == Triple::x86 || Arch == Triple::arm) {
1078 // Generic relocation types...
1080 case macho::RIT_Pair: // GENERIC_RELOC_PAIR - prints no info
1081 return object_error::success;
1082 case macho::RIT_Difference: { // GENERIC_RELOC_SECTDIFF
1083 InMemoryStruct<macho::RelocationEntry> RENext;
1084 DataRefImpl RelNext = Rel;
1086 getRelocation(RelNext, RENext);
1088 // X86 sect diff's must be followed by a relocation of type
1089 // GENERIC_RELOC_PAIR.
1090 bool isNextScattered = (Arch != Triple::x86_64) &&
1091 (RENext->Word0 & macho::RF_Scattered);
1093 if (isNextScattered)
1094 RType = (RENext->Word0 >> 24) & 0xF;
1096 RType = (RENext->Word1 >> 28) & 0xF;
1098 report_fatal_error("Expected GENERIC_RELOC_PAIR after "
1099 "GENERIC_RELOC_SECTDIFF.");
1101 printRelocationTargetName(RE, fmt);
1103 printRelocationTargetName(RENext, fmt);
1108 if (Arch == Triple::x86) {
1109 // All X86 relocations that need special printing were already
1110 // handled in the generic code.
1112 case macho::RIT_Generic_LocalDifference:{// GENERIC_RELOC_LOCAL_SECTDIFF
1113 InMemoryStruct<macho::RelocationEntry> RENext;
1114 DataRefImpl RelNext = Rel;
1116 getRelocation(RelNext, RENext);
1118 // X86 sect diff's must be followed by a relocation of type
1119 // GENERIC_RELOC_PAIR.
1120 bool isNextScattered = (Arch != Triple::x86_64) &&
1121 (RENext->Word0 & macho::RF_Scattered);
1123 if (isNextScattered)
1124 RType = (RENext->Word0 >> 24) & 0xF;
1126 RType = (RENext->Word1 >> 28) & 0xF;
1128 report_fatal_error("Expected GENERIC_RELOC_PAIR after "
1129 "GENERIC_RELOC_LOCAL_SECTDIFF.");
1131 printRelocationTargetName(RE, fmt);
1133 printRelocationTargetName(RENext, fmt);
1136 case macho::RIT_Generic_TLV: {
1137 printRelocationTargetName(RE, fmt);
1139 if (isPCRel) fmt << "P";
1143 printRelocationTargetName(RE, fmt);
1145 } else { // ARM-specific relocations
1147 case macho::RIT_ARM_Half: // ARM_RELOC_HALF
1148 case macho::RIT_ARM_HalfDifference: { // ARM_RELOC_HALF_SECTDIFF
1149 // Half relocations steal a bit from the length field to encode
1150 // whether this is an upper16 or a lower16 relocation.
1153 isUpper = (RE->Word0 >> 28) & 1;
1155 isUpper = (RE->Word1 >> 25) & 1;
1158 fmt << ":upper16:(";
1160 fmt << ":lower16:(";
1161 printRelocationTargetName(RE, fmt);
1163 InMemoryStruct<macho::RelocationEntry> RENext;
1164 DataRefImpl RelNext = Rel;
1166 getRelocation(RelNext, RENext);
1168 // ARM half relocs must be followed by a relocation of type
1170 bool isNextScattered = (Arch != Triple::x86_64) &&
1171 (RENext->Word0 & macho::RF_Scattered);
1173 if (isNextScattered)
1174 RType = (RENext->Word0 >> 24) & 0xF;
1176 RType = (RENext->Word1 >> 28) & 0xF;
1179 report_fatal_error("Expected ARM_RELOC_PAIR after "
1180 "GENERIC_RELOC_HALF");
1182 // NOTE: The half of the target virtual address is stashed in the
1183 // address field of the secondary relocation, but we can't reverse
1184 // engineer the constant offset from it without decoding the movw/movt
1185 // instruction to find the other half in its immediate field.
1187 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the
1188 // symbol/section pointer of the follow-on relocation.
1189 if (Type == macho::RIT_ARM_HalfDifference) {
1191 printRelocationTargetName(RENext, fmt);
1198 printRelocationTargetName(RE, fmt);
1203 printRelocationTargetName(RE, fmt);
1206 Result.append(fmtbuf.begin(), fmtbuf.end());
1207 return object_error::success;
1210 error_code MachOObjectFile::getRelocationHidden(DataRefImpl Rel,
1211 bool &Result) const {
1212 InMemoryStruct<macho::RelocationEntry> RE;
1213 getRelocation(Rel, RE);
1215 unsigned Arch = getArch();
1216 bool isScattered = (Arch != Triple::x86_64) &&
1217 (RE->Word0 & macho::RF_Scattered);
1220 Type = (RE->Word0 >> 24) & 0xF;
1222 Type = (RE->Word1 >> 28) & 0xF;
1226 // On arches that use the generic relocations, GENERIC_RELOC_PAIR
1227 // is always hidden.
1228 if (Arch == Triple::x86 || Arch == Triple::arm) {
1229 if (Type == macho::RIT_Pair) Result = true;
1230 } else if (Arch == Triple::x86_64) {
1231 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows
1232 // an X864_64_RELOC_SUBTRACTOR.
1233 if (Type == macho::RIT_X86_64_Unsigned && Rel.d.a > 0) {
1234 DataRefImpl RelPrev = Rel;
1236 InMemoryStruct<macho::RelocationEntry> REPrev;
1237 getRelocation(RelPrev, REPrev);
1239 unsigned PrevType = (REPrev->Word1 >> 28) & 0xF;
1241 if (PrevType == macho::RIT_X86_64_Subtractor) Result = true;
1245 return object_error::success;
1248 error_code MachOObjectFile::getLibraryNext(DataRefImpl LibData,
1249 LibraryRef &Res) const {
1250 report_fatal_error("Needed libraries unimplemented in MachOObjectFile");
1253 error_code MachOObjectFile::getLibraryPath(DataRefImpl LibData,
1254 StringRef &Res) const {
1255 report_fatal_error("Needed libraries unimplemented in MachOObjectFile");
1259 /*===-- Miscellaneous -----------------------------------------------------===*/
1261 uint8_t MachOObjectFile::getBytesInAddress() const {
1262 return MachOObj->is64Bit() ? 8 : 4;
1265 StringRef MachOObjectFile::getFileFormatName() const {
1266 if (!MachOObj->is64Bit()) {
1267 switch (MachOObj->getHeader().CPUType) {
1268 case llvm::MachO::CPUTypeI386:
1269 return "Mach-O 32-bit i386";
1270 case llvm::MachO::CPUTypeARM:
1271 return "Mach-O arm";
1272 case llvm::MachO::CPUTypePowerPC:
1273 return "Mach-O 32-bit ppc";
1275 assert((MachOObj->getHeader().CPUType & llvm::MachO::CPUArchABI64) == 0 &&
1276 "64-bit object file when we're not 64-bit?");
1277 return "Mach-O 32-bit unknown";
1281 // Make sure the cpu type has the correct mask.
1282 assert((MachOObj->getHeader().CPUType & llvm::MachO::CPUArchABI64)
1283 == llvm::MachO::CPUArchABI64 &&
1284 "32-bit object file when we're 64-bit?");
1286 switch (MachOObj->getHeader().CPUType) {
1287 case llvm::MachO::CPUTypeX86_64:
1288 return "Mach-O 64-bit x86-64";
1289 case llvm::MachO::CPUTypePowerPC64:
1290 return "Mach-O 64-bit ppc64";
1292 return "Mach-O 64-bit unknown";
1296 unsigned MachOObjectFile::getArch() const {
1297 switch (MachOObj->getHeader().CPUType) {
1298 case llvm::MachO::CPUTypeI386:
1300 case llvm::MachO::CPUTypeX86_64:
1301 return Triple::x86_64;
1302 case llvm::MachO::CPUTypeARM:
1304 case llvm::MachO::CPUTypePowerPC:
1306 case llvm::MachO::CPUTypePowerPC64:
1307 return Triple::ppc64;
1309 return Triple::UnknownArch;
1313 } // end namespace object
1314 } // end namespace llvm