1 //===-- MachODump.cpp - Object file dumping utility for llvm --------------===//
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 implements the MachO-specific dumper for llvm-objdump.
12 //===----------------------------------------------------------------------===//
14 #include "llvm-objdump.h"
15 #include "llvm-c/Disassembler.h"
16 #include "llvm/ADT/STLExtras.h"
17 #include "llvm/ADT/StringExtras.h"
18 #include "llvm/ADT/Triple.h"
19 #include "llvm/Config/config.h"
20 #include "llvm/DebugInfo/DWARF/DIContext.h"
21 #include "llvm/MC/MCAsmInfo.h"
22 #include "llvm/MC/MCContext.h"
23 #include "llvm/MC/MCDisassembler.h"
24 #include "llvm/MC/MCInst.h"
25 #include "llvm/MC/MCInstPrinter.h"
26 #include "llvm/MC/MCInstrDesc.h"
27 #include "llvm/MC/MCInstrInfo.h"
28 #include "llvm/MC/MCRegisterInfo.h"
29 #include "llvm/MC/MCSubtargetInfo.h"
30 #include "llvm/Object/MachO.h"
31 #include "llvm/Object/MachOUniversal.h"
32 #include "llvm/Support/Casting.h"
33 #include "llvm/Support/CommandLine.h"
34 #include "llvm/Support/Debug.h"
35 #include "llvm/Support/Endian.h"
36 #include "llvm/Support/Format.h"
37 #include "llvm/Support/FormattedStream.h"
38 #include "llvm/Support/GraphWriter.h"
39 #include "llvm/Support/LEB128.h"
40 #include "llvm/Support/MachO.h"
41 #include "llvm/Support/MemoryBuffer.h"
42 #include "llvm/Support/TargetRegistry.h"
43 #include "llvm/Support/TargetSelect.h"
44 #include "llvm/Support/raw_ostream.h"
47 #include <system_error>
54 using namespace object;
58 cl::desc("Print line information from debug info if available"));
60 static cl::opt<std::string> DSYMFile("dsym",
61 cl::desc("Use .dSYM file for debug info"));
63 static cl::opt<bool> FullLeadingAddr("full-leading-addr",
64 cl::desc("Print full leading address"));
67 PrintImmHex("print-imm-hex",
68 cl::desc("Use hex format for immediate values"));
70 cl::opt<bool> llvm::UniversalHeaders("universal-headers",
71 cl::desc("Print Mach-O universal headers "
72 "(requires -macho)"));
75 llvm::ArchiveHeaders("archive-headers",
76 cl::desc("Print archive headers for Mach-O archives "
77 "(requires -macho)"));
80 llvm::IndirectSymbols("indirect-symbols",
81 cl::desc("Print indirect symbol table for Mach-O "
82 "objects (requires -macho)"));
85 llvm::DataInCode("data-in-code",
86 cl::desc("Print the data in code table for Mach-O objects "
87 "(requires -macho)"));
90 llvm::LinkOptHints("link-opt-hints",
91 cl::desc("Print the linker optimization hints for "
92 "Mach-O objects (requires -macho)"));
95 llvm::DumpSections("section",
96 cl::desc("Prints the specified segment,section for "
97 "Mach-O objects (requires -macho)"));
99 static cl::list<std::string>
100 ArchFlags("arch", cl::desc("architecture(s) from a Mach-O file to dump"),
102 bool ArchAll = false;
104 static std::string ThumbTripleName;
106 static const Target *GetTarget(const MachOObjectFile *MachOObj,
107 const char **McpuDefault,
108 const Target **ThumbTarget) {
109 // Figure out the target triple.
110 if (TripleName.empty()) {
111 llvm::Triple TT("unknown-unknown-unknown");
112 llvm::Triple ThumbTriple = Triple();
113 TT = MachOObj->getArch(McpuDefault, &ThumbTriple);
114 TripleName = TT.str();
115 ThumbTripleName = ThumbTriple.str();
118 // Get the target specific parser.
120 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
121 if (TheTarget && ThumbTripleName.empty())
124 *ThumbTarget = TargetRegistry::lookupTarget(ThumbTripleName, Error);
128 errs() << "llvm-objdump: error: unable to get target for '";
130 errs() << TripleName;
132 errs() << ThumbTripleName;
133 errs() << "', see --version and --triple.\n";
137 struct SymbolSorter {
138 bool operator()(const SymbolRef &A, const SymbolRef &B) {
139 SymbolRef::Type AType, BType;
143 uint64_t AAddr, BAddr;
144 if (AType != SymbolRef::ST_Function)
148 if (BType != SymbolRef::ST_Function)
152 return AAddr < BAddr;
156 // Types for the storted data in code table that is built before disassembly
157 // and the predicate function to sort them.
158 typedef std::pair<uint64_t, DiceRef> DiceTableEntry;
159 typedef std::vector<DiceTableEntry> DiceTable;
160 typedef DiceTable::iterator dice_table_iterator;
162 // This is used to search for a data in code table entry for the PC being
163 // disassembled. The j parameter has the PC in j.first. A single data in code
164 // table entry can cover many bytes for each of its Kind's. So if the offset,
165 // aka the i.first value, of the data in code table entry plus its Length
166 // covers the PC being searched for this will return true. If not it will
168 static bool compareDiceTableEntries(const DiceTableEntry &i,
169 const DiceTableEntry &j) {
171 i.second.getLength(Length);
173 return j.first >= i.first && j.first < i.first + Length;
176 static uint64_t DumpDataInCode(const char *bytes, uint64_t Length,
177 unsigned short Kind) {
178 uint32_t Value, Size = 1;
182 case MachO::DICE_KIND_DATA:
185 DumpBytes(StringRef(bytes, 4));
186 Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0];
187 outs() << "\t.long " << Value;
189 } else if (Length >= 2) {
191 DumpBytes(StringRef(bytes, 2));
192 Value = bytes[1] << 8 | bytes[0];
193 outs() << "\t.short " << Value;
197 DumpBytes(StringRef(bytes, 2));
199 outs() << "\t.byte " << Value;
202 if (Kind == MachO::DICE_KIND_DATA)
203 outs() << "\t@ KIND_DATA\n";
205 outs() << "\t@ data in code kind = " << Kind << "\n";
207 case MachO::DICE_KIND_JUMP_TABLE8:
209 DumpBytes(StringRef(bytes, 1));
211 outs() << "\t.byte " << format("%3u", Value) << "\t@ KIND_JUMP_TABLE8\n";
214 case MachO::DICE_KIND_JUMP_TABLE16:
216 DumpBytes(StringRef(bytes, 2));
217 Value = bytes[1] << 8 | bytes[0];
218 outs() << "\t.short " << format("%5u", Value & 0xffff)
219 << "\t@ KIND_JUMP_TABLE16\n";
222 case MachO::DICE_KIND_JUMP_TABLE32:
223 case MachO::DICE_KIND_ABS_JUMP_TABLE32:
225 DumpBytes(StringRef(bytes, 4));
226 Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0];
227 outs() << "\t.long " << Value;
228 if (Kind == MachO::DICE_KIND_JUMP_TABLE32)
229 outs() << "\t@ KIND_JUMP_TABLE32\n";
231 outs() << "\t@ KIND_ABS_JUMP_TABLE32\n";
238 static void getSectionsAndSymbols(const MachO::mach_header Header,
239 MachOObjectFile *MachOObj,
240 std::vector<SectionRef> &Sections,
241 std::vector<SymbolRef> &Symbols,
242 SmallVectorImpl<uint64_t> &FoundFns,
243 uint64_t &BaseSegmentAddress) {
244 for (const SymbolRef &Symbol : MachOObj->symbols()) {
246 Symbol.getName(SymName);
247 if (!SymName.startswith("ltmp"))
248 Symbols.push_back(Symbol);
251 for (const SectionRef &Section : MachOObj->sections()) {
253 Section.getName(SectName);
254 Sections.push_back(Section);
257 MachOObjectFile::LoadCommandInfo Command =
258 MachOObj->getFirstLoadCommandInfo();
259 bool BaseSegmentAddressSet = false;
260 for (unsigned i = 0;; ++i) {
261 if (Command.C.cmd == MachO::LC_FUNCTION_STARTS) {
262 // We found a function starts segment, parse the addresses for later
264 MachO::linkedit_data_command LLC =
265 MachOObj->getLinkeditDataLoadCommand(Command);
267 MachOObj->ReadULEB128s(LLC.dataoff, FoundFns);
268 } else if (Command.C.cmd == MachO::LC_SEGMENT) {
269 MachO::segment_command SLC = MachOObj->getSegmentLoadCommand(Command);
270 StringRef SegName = SLC.segname;
271 if (!BaseSegmentAddressSet && SegName != "__PAGEZERO") {
272 BaseSegmentAddressSet = true;
273 BaseSegmentAddress = SLC.vmaddr;
277 if (i == Header.ncmds - 1)
280 Command = MachOObj->getNextLoadCommandInfo(Command);
284 static void PrintIndirectSymbolTable(MachOObjectFile *O, bool verbose,
285 uint32_t n, uint32_t count,
286 uint32_t stride, uint64_t addr) {
287 MachO::dysymtab_command Dysymtab = O->getDysymtabLoadCommand();
288 uint32_t nindirectsyms = Dysymtab.nindirectsyms;
289 if (n > nindirectsyms)
290 outs() << " (entries start past the end of the indirect symbol "
291 "table) (reserved1 field greater than the table size)";
292 else if (n + count > nindirectsyms)
293 outs() << " (entries extends past the end of the indirect symbol "
296 uint32_t cputype = O->getHeader().cputype;
297 if (cputype & MachO::CPU_ARCH_ABI64)
298 outs() << "address index";
300 outs() << "address index";
305 for (uint32_t j = 0; j < count && n + j < nindirectsyms; j++) {
306 if (cputype & MachO::CPU_ARCH_ABI64)
307 outs() << format("0x%016" PRIx64, addr + j * stride) << " ";
309 outs() << format("0x%08" PRIx32, addr + j * stride) << " ";
310 MachO::dysymtab_command Dysymtab = O->getDysymtabLoadCommand();
311 uint32_t indirect_symbol = O->getIndirectSymbolTableEntry(Dysymtab, n + j);
312 if (indirect_symbol == MachO::INDIRECT_SYMBOL_LOCAL) {
316 if (indirect_symbol ==
317 (MachO::INDIRECT_SYMBOL_LOCAL | MachO::INDIRECT_SYMBOL_ABS)) {
318 outs() << "LOCAL ABSOLUTE\n";
321 if (indirect_symbol == MachO::INDIRECT_SYMBOL_ABS) {
322 outs() << "ABSOLUTE\n";
325 outs() << format("%5u ", indirect_symbol);
326 MachO::symtab_command Symtab = O->getSymtabLoadCommand();
327 if (indirect_symbol < Symtab.nsyms) {
328 symbol_iterator Sym = O->getSymbolByIndex(indirect_symbol);
329 SymbolRef Symbol = *Sym;
331 Symbol.getName(SymName);
340 static void PrintIndirectSymbols(MachOObjectFile *O, bool verbose) {
341 uint32_t LoadCommandCount = O->getHeader().ncmds;
342 MachOObjectFile::LoadCommandInfo Load = O->getFirstLoadCommandInfo();
343 for (unsigned I = 0;; ++I) {
344 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
345 MachO::segment_command_64 Seg = O->getSegment64LoadCommand(Load);
346 for (unsigned J = 0; J < Seg.nsects; ++J) {
347 MachO::section_64 Sec = O->getSection64(Load, J);
348 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
349 if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
350 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
351 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
352 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
353 section_type == MachO::S_SYMBOL_STUBS) {
355 if (section_type == MachO::S_SYMBOL_STUBS)
356 stride = Sec.reserved2;
360 outs() << "Can't print indirect symbols for (" << Sec.segname << ","
361 << Sec.sectname << ") "
362 << "(size of stubs in reserved2 field is zero)\n";
365 uint32_t count = Sec.size / stride;
366 outs() << "Indirect symbols for (" << Sec.segname << ","
367 << Sec.sectname << ") " << count << " entries";
368 uint32_t n = Sec.reserved1;
369 PrintIndirectSymbolTable(O, verbose, n, count, stride, Sec.addr);
372 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
373 MachO::segment_command Seg = O->getSegmentLoadCommand(Load);
374 for (unsigned J = 0; J < Seg.nsects; ++J) {
375 MachO::section Sec = O->getSection(Load, J);
376 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
377 if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
378 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
379 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
380 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
381 section_type == MachO::S_SYMBOL_STUBS) {
383 if (section_type == MachO::S_SYMBOL_STUBS)
384 stride = Sec.reserved2;
388 outs() << "Can't print indirect symbols for (" << Sec.segname << ","
389 << Sec.sectname << ") "
390 << "(size of stubs in reserved2 field is zero)\n";
393 uint32_t count = Sec.size / stride;
394 outs() << "Indirect symbols for (" << Sec.segname << ","
395 << Sec.sectname << ") " << count << " entries";
396 uint32_t n = Sec.reserved1;
397 PrintIndirectSymbolTable(O, verbose, n, count, stride, Sec.addr);
401 if (I == LoadCommandCount - 1)
404 Load = O->getNextLoadCommandInfo(Load);
408 static void PrintDataInCodeTable(MachOObjectFile *O, bool verbose) {
409 MachO::linkedit_data_command DIC = O->getDataInCodeLoadCommand();
410 uint32_t nentries = DIC.datasize / sizeof(struct MachO::data_in_code_entry);
411 outs() << "Data in code table (" << nentries << " entries)\n";
412 outs() << "offset length kind\n";
413 for (dice_iterator DI = O->begin_dices(), DE = O->end_dices(); DI != DE;
416 DI->getOffset(Offset);
417 outs() << format("0x%08" PRIx32, Offset) << " ";
419 DI->getLength(Length);
420 outs() << format("%6u", Length) << " ";
425 case MachO::DICE_KIND_DATA:
428 case MachO::DICE_KIND_JUMP_TABLE8:
429 outs() << "JUMP_TABLE8";
431 case MachO::DICE_KIND_JUMP_TABLE16:
432 outs() << "JUMP_TABLE16";
434 case MachO::DICE_KIND_JUMP_TABLE32:
435 outs() << "JUMP_TABLE32";
437 case MachO::DICE_KIND_ABS_JUMP_TABLE32:
438 outs() << "ABS_JUMP_TABLE32";
441 outs() << format("0x%04" PRIx32, Kind);
445 outs() << format("0x%04" PRIx32, Kind);
450 static void PrintLinkOptHints(MachOObjectFile *O) {
451 MachO::linkedit_data_command LohLC = O->getLinkOptHintsLoadCommand();
452 const char *loh = O->getData().substr(LohLC.dataoff, 1).data();
453 uint32_t nloh = LohLC.datasize;
454 outs() << "Linker optimiztion hints (" << nloh << " total bytes)\n";
455 for (uint32_t i = 0; i < nloh;) {
457 uint64_t identifier = decodeULEB128((const uint8_t *)(loh + i), &n);
459 outs() << " identifier " << identifier << " ";
462 switch (identifier) {
464 outs() << "AdrpAdrp\n";
467 outs() << "AdrpLdr\n";
470 outs() << "AdrpAddLdr\n";
473 outs() << "AdrpLdrGotLdr\n";
476 outs() << "AdrpAddStr\n";
479 outs() << "AdrpLdrGotStr\n";
482 outs() << "AdrpAdd\n";
485 outs() << "AdrpLdrGot\n";
488 outs() << "Unknown identifier value\n";
491 uint64_t narguments = decodeULEB128((const uint8_t *)(loh + i), &n);
493 outs() << " narguments " << narguments << "\n";
497 for (uint32_t j = 0; j < narguments; j++) {
498 uint64_t value = decodeULEB128((const uint8_t *)(loh + i), &n);
500 outs() << "\tvalue " << format("0x%" PRIx64, value) << "\n";
507 typedef DenseMap<uint64_t, StringRef> SymbolAddressMap;
509 static void CreateSymbolAddressMap(MachOObjectFile *O,
510 SymbolAddressMap *AddrMap) {
511 // Create a map of symbol addresses to symbol names.
512 for (const SymbolRef &Symbol : O->symbols()) {
515 if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data ||
516 ST == SymbolRef::ST_Other) {
518 Symbol.getAddress(Address);
520 Symbol.getName(SymName);
521 (*AddrMap)[Address] = SymName;
526 // GuessSymbolName is passed the address of what might be a symbol and a
527 // pointer to the SymbolAddressMap. It returns the name of a symbol
528 // with that address or nullptr if no symbol is found with that address.
529 static const char *GuessSymbolName(uint64_t value, SymbolAddressMap *AddrMap) {
530 const char *SymbolName = nullptr;
531 // A DenseMap can't lookup up some values.
532 if (value != 0xffffffffffffffffULL && value != 0xfffffffffffffffeULL) {
533 StringRef name = AddrMap->lookup(value);
535 SymbolName = name.data();
540 static void DumpInitTermPointerSection(MachOObjectFile *O, const char *sect,
541 uint32_t sect_size, uint64_t sect_addr,
542 SymbolAddressMap *AddrMap,
546 stride = sizeof(uint64_t);
548 stride = sizeof(uint32_t);
549 for (uint32_t i = 0; i < sect_size; i += stride) {
550 const char *SymbolName = nullptr;
552 outs() << format("0x%016" PRIx64, sect_addr + i * stride) << " ";
553 uint64_t pointer_value;
554 memcpy(&pointer_value, sect + i, stride);
555 if (O->isLittleEndian() != sys::IsLittleEndianHost)
556 sys::swapByteOrder(pointer_value);
557 outs() << format("0x%016" PRIx64, pointer_value);
559 SymbolName = GuessSymbolName(pointer_value, AddrMap);
561 outs() << format("0x%08" PRIx64, sect_addr + i * stride) << " ";
562 uint32_t pointer_value;
563 memcpy(&pointer_value, sect + i, stride);
564 if (O->isLittleEndian() != sys::IsLittleEndianHost)
565 sys::swapByteOrder(pointer_value);
566 outs() << format("0x%08" PRIx32, pointer_value);
568 SymbolName = GuessSymbolName(pointer_value, AddrMap);
571 outs() << " " << SymbolName;
576 static void DumpRawSectionContents(MachOObjectFile *O, const char *sect,
577 uint32_t size, uint64_t addr) {
578 uint32_t cputype = O->getHeader().cputype;
579 if (cputype == MachO::CPU_TYPE_I386 || cputype == MachO::CPU_TYPE_X86_64) {
581 for (uint32_t i = 0; i < size; i += j, addr += j) {
583 outs() << format("%016" PRIx64, addr) << "\t";
585 outs() << format("%08" PRIx64, sect) << "\t";
586 for (j = 0; j < 16 && i + j < size; j++) {
587 uint8_t byte_word = *(sect + i + j);
588 outs() << format("%02" PRIx32, (uint32_t)byte_word) << " ";
594 for (uint32_t i = 0; i < size; i += j, addr += j) {
596 outs() << format("%016" PRIx64, addr) << "\t";
598 outs() << format("%08" PRIx64, sect) << "\t";
599 for (j = 0; j < 4 * sizeof(int32_t) && i + j < size;
600 j += sizeof(int32_t)) {
601 if (i + j + sizeof(int32_t) < size) {
603 memcpy(&long_word, sect + i + j, sizeof(int32_t));
604 if (O->isLittleEndian() != sys::IsLittleEndianHost)
605 sys::swapByteOrder(long_word);
606 outs() << format("%08" PRIx32, long_word) << " ";
608 for (uint32_t k = 0; i + j + k < size; k++) {
609 uint8_t byte_word = *(sect + i + j);
610 outs() << format("%02" PRIx32, (uint32_t)byte_word) << " ";
619 static void DumpSectionContents(MachOObjectFile *O, bool verbose) {
620 SymbolAddressMap AddrMap;
622 CreateSymbolAddressMap(O, &AddrMap);
624 for (unsigned i = 0; i < DumpSections.size(); ++i) {
625 StringRef DumpSection = DumpSections[i];
626 std::pair<StringRef, StringRef> DumpSegSectName;
627 DumpSegSectName = DumpSection.split(',');
628 StringRef DumpSegName, DumpSectName;
629 if (DumpSegSectName.second.size()) {
630 DumpSegName = DumpSegSectName.first;
631 DumpSectName = DumpSegSectName.second;
634 DumpSectName = DumpSegSectName.first;
636 for (const SectionRef &Section : O->sections()) {
638 Section.getName(SectName);
639 DataRefImpl Ref = Section.getRawDataRefImpl();
640 StringRef SegName = O->getSectionFinalSegmentName(Ref);
641 if ((DumpSegName.empty() || SegName == DumpSegName) &&
642 (SectName == DumpSectName)) {
643 outs() << "Contents of (" << SegName << "," << SectName
645 uint32_t section_type;
647 const MachO::section_64 Sec = O->getSection64(Ref);
648 section_type = Sec.flags & MachO::SECTION_TYPE;
651 const MachO::section Sec = O->getSection(Ref);
652 section_type = Sec.flags & MachO::SECTION_TYPE;
656 Section.getContents(BytesStr);
657 const char *sect = reinterpret_cast<const char *>(BytesStr.data());
658 uint32_t sect_size = BytesStr.size();
659 uint64_t sect_addr = Section.getAddress();
662 switch (section_type) {
663 case MachO::S_REGULAR:
664 DumpRawSectionContents(O, sect, sect_size, sect_addr);
666 case MachO::S_ZEROFILL:
667 outs() << "zerofill section and has no contents in the file\n";
669 case MachO::S_MOD_INIT_FUNC_POINTERS:
670 case MachO::S_MOD_TERM_FUNC_POINTERS:
671 DumpInitTermPointerSection(O, sect, sect_size, sect_addr, &AddrMap,
675 outs() << "Unknown section type ("
676 << format("0x%08" PRIx32, section_type) << ")\n";
677 DumpRawSectionContents(O, sect, sect_size, sect_addr);
681 if (section_type == MachO::S_ZEROFILL)
682 outs() << "zerofill section and has no contents in the file\n";
684 DumpRawSectionContents(O, sect, sect_size, sect_addr);
691 // checkMachOAndArchFlags() checks to see if the ObjectFile is a Mach-O file
692 // and if it is and there is a list of architecture flags is specified then
693 // check to make sure this Mach-O file is one of those architectures or all
694 // architectures were specified. If not then an error is generated and this
695 // routine returns false. Else it returns true.
696 static bool checkMachOAndArchFlags(ObjectFile *O, StringRef Filename) {
697 if (isa<MachOObjectFile>(O) && !ArchAll && ArchFlags.size() != 0) {
698 MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O);
699 bool ArchFound = false;
700 MachO::mach_header H;
701 MachO::mach_header_64 H_64;
703 if (MachO->is64Bit()) {
704 H_64 = MachO->MachOObjectFile::getHeader64();
705 T = MachOObjectFile::getArch(H_64.cputype, H_64.cpusubtype);
707 H = MachO->MachOObjectFile::getHeader();
708 T = MachOObjectFile::getArch(H.cputype, H.cpusubtype);
711 for (i = 0; i < ArchFlags.size(); ++i) {
712 if (ArchFlags[i] == T.getArchName())
717 errs() << "llvm-objdump: file: " + Filename + " does not contain "
718 << "architecture: " + ArchFlags[i] + "\n";
725 static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF);
727 // ProcessMachO() is passed a single opened Mach-O file, which may be an
728 // archive member and or in a slice of a universal file. It prints the
729 // the file name and header info and then processes it according to the
730 // command line options.
731 static void ProcessMachO(StringRef Filename, MachOObjectFile *MachOOF,
732 StringRef ArchiveMemberName = StringRef(),
733 StringRef ArchitectureName = StringRef()) {
734 // If we are doing some processing here on the Mach-O file print the header
735 // info. And don't print it otherwise like in the case of printing the
736 // UniversalHeaders or ArchiveHeaders.
737 if (Disassemble || PrivateHeaders || ExportsTrie || Rebase || Bind ||
738 LazyBind || WeakBind || IndirectSymbols || DataInCode || LinkOptHints ||
739 DumpSections.size() != 0) {
741 if (!ArchiveMemberName.empty())
742 outs() << '(' << ArchiveMemberName << ')';
743 if (!ArchitectureName.empty())
744 outs() << " (architecture " << ArchitectureName << ")";
749 DisassembleMachO(Filename, MachOOF);
751 PrintIndirectSymbols(MachOOF, true);
753 PrintDataInCodeTable(MachOOF, true);
755 PrintLinkOptHints(MachOOF);
757 PrintRelocations(MachOOF);
759 PrintSectionHeaders(MachOOF);
761 PrintSectionContents(MachOOF);
762 if (DumpSections.size() != 0)
763 DumpSectionContents(MachOOF, true);
765 PrintSymbolTable(MachOOF);
767 printMachOUnwindInfo(MachOOF);
769 printMachOFileHeader(MachOOF);
771 printExportsTrie(MachOOF);
773 printRebaseTable(MachOOF);
775 printBindTable(MachOOF);
777 printLazyBindTable(MachOOF);
779 printWeakBindTable(MachOOF);
782 // printUnknownCPUType() helps print_fat_headers for unknown CPU's.
783 static void printUnknownCPUType(uint32_t cputype, uint32_t cpusubtype) {
784 outs() << " cputype (" << cputype << ")\n";
785 outs() << " cpusubtype (" << cpusubtype << ")\n";
788 // printCPUType() helps print_fat_headers by printing the cputype and
789 // pusubtype (symbolically for the one's it knows about).
790 static void printCPUType(uint32_t cputype, uint32_t cpusubtype) {
792 case MachO::CPU_TYPE_I386:
793 switch (cpusubtype) {
794 case MachO::CPU_SUBTYPE_I386_ALL:
795 outs() << " cputype CPU_TYPE_I386\n";
796 outs() << " cpusubtype CPU_SUBTYPE_I386_ALL\n";
799 printUnknownCPUType(cputype, cpusubtype);
803 case MachO::CPU_TYPE_X86_64:
804 switch (cpusubtype) {
805 case MachO::CPU_SUBTYPE_X86_64_ALL:
806 outs() << " cputype CPU_TYPE_X86_64\n";
807 outs() << " cpusubtype CPU_SUBTYPE_X86_64_ALL\n";
809 case MachO::CPU_SUBTYPE_X86_64_H:
810 outs() << " cputype CPU_TYPE_X86_64\n";
811 outs() << " cpusubtype CPU_SUBTYPE_X86_64_H\n";
814 printUnknownCPUType(cputype, cpusubtype);
818 case MachO::CPU_TYPE_ARM:
819 switch (cpusubtype) {
820 case MachO::CPU_SUBTYPE_ARM_ALL:
821 outs() << " cputype CPU_TYPE_ARM\n";
822 outs() << " cpusubtype CPU_SUBTYPE_ARM_ALL\n";
824 case MachO::CPU_SUBTYPE_ARM_V4T:
825 outs() << " cputype CPU_TYPE_ARM\n";
826 outs() << " cpusubtype CPU_SUBTYPE_ARM_V4T\n";
828 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
829 outs() << " cputype CPU_TYPE_ARM\n";
830 outs() << " cpusubtype CPU_SUBTYPE_ARM_V5TEJ\n";
832 case MachO::CPU_SUBTYPE_ARM_XSCALE:
833 outs() << " cputype CPU_TYPE_ARM\n";
834 outs() << " cpusubtype CPU_SUBTYPE_ARM_XSCALE\n";
836 case MachO::CPU_SUBTYPE_ARM_V6:
837 outs() << " cputype CPU_TYPE_ARM\n";
838 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6\n";
840 case MachO::CPU_SUBTYPE_ARM_V6M:
841 outs() << " cputype CPU_TYPE_ARM\n";
842 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6M\n";
844 case MachO::CPU_SUBTYPE_ARM_V7:
845 outs() << " cputype CPU_TYPE_ARM\n";
846 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7\n";
848 case MachO::CPU_SUBTYPE_ARM_V7EM:
849 outs() << " cputype CPU_TYPE_ARM\n";
850 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7EM\n";
852 case MachO::CPU_SUBTYPE_ARM_V7K:
853 outs() << " cputype CPU_TYPE_ARM\n";
854 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7K\n";
856 case MachO::CPU_SUBTYPE_ARM_V7M:
857 outs() << " cputype CPU_TYPE_ARM\n";
858 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7M\n";
860 case MachO::CPU_SUBTYPE_ARM_V7S:
861 outs() << " cputype CPU_TYPE_ARM\n";
862 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7S\n";
865 printUnknownCPUType(cputype, cpusubtype);
869 case MachO::CPU_TYPE_ARM64:
870 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
871 case MachO::CPU_SUBTYPE_ARM64_ALL:
872 outs() << " cputype CPU_TYPE_ARM64\n";
873 outs() << " cpusubtype CPU_SUBTYPE_ARM64_ALL\n";
876 printUnknownCPUType(cputype, cpusubtype);
881 printUnknownCPUType(cputype, cpusubtype);
886 static void printMachOUniversalHeaders(const object::MachOUniversalBinary *UB,
888 outs() << "Fat headers\n";
890 outs() << "fat_magic FAT_MAGIC\n";
892 outs() << "fat_magic " << format("0x%" PRIx32, MachO::FAT_MAGIC) << "\n";
894 uint32_t nfat_arch = UB->getNumberOfObjects();
895 StringRef Buf = UB->getData();
896 uint64_t size = Buf.size();
897 uint64_t big_size = sizeof(struct MachO::fat_header) +
898 nfat_arch * sizeof(struct MachO::fat_arch);
899 outs() << "nfat_arch " << UB->getNumberOfObjects();
901 outs() << " (malformed, contains zero architecture types)\n";
902 else if (big_size > size)
903 outs() << " (malformed, architectures past end of file)\n";
907 for (uint32_t i = 0; i < nfat_arch; ++i) {
908 MachOUniversalBinary::ObjectForArch OFA(UB, i);
909 uint32_t cputype = OFA.getCPUType();
910 uint32_t cpusubtype = OFA.getCPUSubType();
911 outs() << "architecture ";
912 for (uint32_t j = 0; i != 0 && j <= i - 1; j++) {
913 MachOUniversalBinary::ObjectForArch other_OFA(UB, j);
914 uint32_t other_cputype = other_OFA.getCPUType();
915 uint32_t other_cpusubtype = other_OFA.getCPUSubType();
916 if (cputype != 0 && cpusubtype != 0 && cputype == other_cputype &&
917 (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) ==
918 (other_cpusubtype & ~MachO::CPU_SUBTYPE_MASK)) {
919 outs() << "(illegal duplicate architecture) ";
924 outs() << OFA.getArchTypeName() << "\n";
925 printCPUType(cputype, cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
928 outs() << " cputype " << cputype << "\n";
929 outs() << " cpusubtype " << (cpusubtype & ~MachO::CPU_SUBTYPE_MASK)
933 (cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64)
934 outs() << " capabilities CPU_SUBTYPE_LIB64\n";
936 outs() << " capabilities "
937 << format("0x%" PRIx32,
938 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24) << "\n";
939 outs() << " offset " << OFA.getOffset();
940 if (OFA.getOffset() > size)
941 outs() << " (past end of file)";
942 if (OFA.getOffset() % (1 << OFA.getAlign()) != 0)
943 outs() << " (not aligned on it's alignment (2^" << OFA.getAlign() << ")";
945 outs() << " size " << OFA.getSize();
946 big_size = OFA.getOffset() + OFA.getSize();
948 outs() << " (past end of file)";
950 outs() << " align 2^" << OFA.getAlign() << " (" << (1 << OFA.getAlign())
955 static void printArchiveChild(Archive::Child &C, bool verbose,
958 outs() << C.getChildOffset() << "\t";
959 sys::fs::perms Mode = C.getAccessMode();
961 // FIXME: this first dash, "-", is for (Mode & S_IFMT) == S_IFREG.
962 // But there is nothing in sys::fs::perms for S_IFMT or S_IFREG.
964 if (Mode & sys::fs::owner_read)
968 if (Mode & sys::fs::owner_write)
972 if (Mode & sys::fs::owner_exe)
976 if (Mode & sys::fs::group_read)
980 if (Mode & sys::fs::group_write)
984 if (Mode & sys::fs::group_exe)
988 if (Mode & sys::fs::others_read)
992 if (Mode & sys::fs::others_write)
996 if (Mode & sys::fs::others_exe)
1001 outs() << format("0%o ", Mode);
1004 unsigned UID = C.getUID();
1005 outs() << format("%3d/", UID);
1006 unsigned GID = C.getGID();
1007 outs() << format("%-3d ", GID);
1008 uint64_t Size = C.getRawSize();
1009 outs() << format("%5" PRId64, Size) << " ";
1011 StringRef RawLastModified = C.getRawLastModified();
1014 if (RawLastModified.getAsInteger(10, Seconds))
1015 outs() << "(date: \"%s\" contains non-decimal chars) " << RawLastModified;
1017 // Since cime(3) returns a 26 character string of the form:
1018 // "Sun Sep 16 01:03:52 1973\n\0"
1019 // just print 24 characters.
1021 outs() << format("%.24s ", ctime(&t));
1024 outs() << RawLastModified << " ";
1028 ErrorOr<StringRef> NameOrErr = C.getName();
1029 if (NameOrErr.getError()) {
1030 StringRef RawName = C.getRawName();
1031 outs() << RawName << "\n";
1033 StringRef Name = NameOrErr.get();
1034 outs() << Name << "\n";
1037 StringRef RawName = C.getRawName();
1038 outs() << RawName << "\n";
1042 static void printArchiveHeaders(Archive *A, bool verbose, bool print_offset) {
1043 if (A->hasSymbolTable()) {
1044 Archive::child_iterator S = A->getSymbolTableChild();
1045 Archive::Child C = *S;
1046 printArchiveChild(C, verbose, print_offset);
1048 for (Archive::child_iterator I = A->child_begin(), E = A->child_end(); I != E;
1050 Archive::Child C = *I;
1051 printArchiveChild(C, verbose, print_offset);
1055 // ParseInputMachO() parses the named Mach-O file in Filename and handles the
1056 // -arch flags selecting just those slices as specified by them and also parses
1057 // archive files. Then for each individual Mach-O file ProcessMachO() is
1058 // called to process the file based on the command line options.
1059 void llvm::ParseInputMachO(StringRef Filename) {
1060 // Check for -arch all and verifiy the -arch flags are valid.
1061 for (unsigned i = 0; i < ArchFlags.size(); ++i) {
1062 if (ArchFlags[i] == "all") {
1065 if (!MachOObjectFile::isValidArch(ArchFlags[i])) {
1066 errs() << "llvm-objdump: Unknown architecture named '" + ArchFlags[i] +
1067 "'for the -arch option\n";
1073 // Attempt to open the binary.
1074 ErrorOr<OwningBinary<Binary>> BinaryOrErr = createBinary(Filename);
1075 if (std::error_code EC = BinaryOrErr.getError()) {
1076 errs() << "llvm-objdump: '" << Filename << "': " << EC.message() << ".\n";
1079 Binary &Bin = *BinaryOrErr.get().getBinary();
1081 if (Archive *A = dyn_cast<Archive>(&Bin)) {
1082 outs() << "Archive : " << Filename << "\n";
1084 printArchiveHeaders(A, true, false);
1085 for (Archive::child_iterator I = A->child_begin(), E = A->child_end();
1087 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = I->getAsBinary();
1088 if (ChildOrErr.getError())
1090 if (MachOObjectFile *O = dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
1091 if (!checkMachOAndArchFlags(O, Filename))
1093 ProcessMachO(Filename, O, O->getFileName());
1098 if (UniversalHeaders) {
1099 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin))
1100 printMachOUniversalHeaders(UB, true);
1102 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin)) {
1103 // If we have a list of architecture flags specified dump only those.
1104 if (!ArchAll && ArchFlags.size() != 0) {
1105 // Look for a slice in the universal binary that matches each ArchFlag.
1107 for (unsigned i = 0; i < ArchFlags.size(); ++i) {
1109 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1110 E = UB->end_objects();
1112 if (ArchFlags[i] == I->getArchTypeName()) {
1114 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr =
1115 I->getAsObjectFile();
1116 std::string ArchitectureName = "";
1117 if (ArchFlags.size() > 1)
1118 ArchitectureName = I->getArchTypeName();
1120 ObjectFile &O = *ObjOrErr.get();
1121 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
1122 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
1123 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
1124 I->getAsArchive()) {
1125 std::unique_ptr<Archive> &A = *AOrErr;
1126 outs() << "Archive : " << Filename;
1127 if (!ArchitectureName.empty())
1128 outs() << " (architecture " << ArchitectureName << ")";
1131 printArchiveHeaders(A.get(), true, false);
1132 for (Archive::child_iterator AI = A->child_begin(),
1133 AE = A->child_end();
1135 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1136 if (ChildOrErr.getError())
1138 if (MachOObjectFile *O =
1139 dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
1140 ProcessMachO(Filename, O, O->getFileName(), ArchitectureName);
1146 errs() << "llvm-objdump: file: " + Filename + " does not contain "
1147 << "architecture: " + ArchFlags[i] + "\n";
1153 // No architecture flags were specified so if this contains a slice that
1154 // matches the host architecture dump only that.
1156 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1157 E = UB->end_objects();
1159 if (MachOObjectFile::getHostArch().getArchName() ==
1160 I->getArchTypeName()) {
1161 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
1162 std::string ArchiveName;
1163 ArchiveName.clear();
1165 ObjectFile &O = *ObjOrErr.get();
1166 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
1167 ProcessMachO(Filename, MachOOF);
1168 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
1169 I->getAsArchive()) {
1170 std::unique_ptr<Archive> &A = *AOrErr;
1171 outs() << "Archive : " << Filename << "\n";
1173 printArchiveHeaders(A.get(), true, false);
1174 for (Archive::child_iterator AI = A->child_begin(),
1175 AE = A->child_end();
1177 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1178 if (ChildOrErr.getError())
1180 if (MachOObjectFile *O =
1181 dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
1182 ProcessMachO(Filename, O, O->getFileName());
1189 // Either all architectures have been specified or none have been specified
1190 // and this does not contain the host architecture so dump all the slices.
1191 bool moreThanOneArch = UB->getNumberOfObjects() > 1;
1192 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1193 E = UB->end_objects();
1195 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
1196 std::string ArchitectureName = "";
1197 if (moreThanOneArch)
1198 ArchitectureName = I->getArchTypeName();
1200 ObjectFile &Obj = *ObjOrErr.get();
1201 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&Obj))
1202 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
1203 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr = I->getAsArchive()) {
1204 std::unique_ptr<Archive> &A = *AOrErr;
1205 outs() << "Archive : " << Filename;
1206 if (!ArchitectureName.empty())
1207 outs() << " (architecture " << ArchitectureName << ")";
1210 printArchiveHeaders(A.get(), true, false);
1211 for (Archive::child_iterator AI = A->child_begin(), AE = A->child_end();
1213 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1214 if (ChildOrErr.getError())
1216 if (MachOObjectFile *O =
1217 dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
1218 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(O))
1219 ProcessMachO(Filename, MachOOF, MachOOF->getFileName(),
1227 if (ObjectFile *O = dyn_cast<ObjectFile>(&Bin)) {
1228 if (!checkMachOAndArchFlags(O, Filename))
1230 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&*O)) {
1231 ProcessMachO(Filename, MachOOF);
1233 errs() << "llvm-objdump: '" << Filename << "': "
1234 << "Object is not a Mach-O file type.\n";
1236 errs() << "llvm-objdump: '" << Filename << "': "
1237 << "Unrecognized file type.\n";
1240 typedef std::pair<uint64_t, const char *> BindInfoEntry;
1241 typedef std::vector<BindInfoEntry> BindTable;
1242 typedef BindTable::iterator bind_table_iterator;
1244 // The block of info used by the Symbolizer call backs.
1245 struct DisassembleInfo {
1249 SymbolAddressMap *AddrMap;
1250 std::vector<SectionRef> *Sections;
1251 const char *class_name;
1252 const char *selector_name;
1254 char *demangled_name;
1257 BindTable *bindtable;
1260 // SymbolizerGetOpInfo() is the operand information call back function.
1261 // This is called to get the symbolic information for operand(s) of an
1262 // instruction when it is being done. This routine does this from
1263 // the relocation information, symbol table, etc. That block of information
1264 // is a pointer to the struct DisassembleInfo that was passed when the
1265 // disassembler context was created and passed to back to here when
1266 // called back by the disassembler for instruction operands that could have
1267 // relocation information. The address of the instruction containing operand is
1268 // at the Pc parameter. The immediate value the operand has is passed in
1269 // op_info->Value and is at Offset past the start of the instruction and has a
1270 // byte Size of 1, 2 or 4. The symbolc information is returned in TagBuf is the
1271 // LLVMOpInfo1 struct defined in the header "llvm-c/Disassembler.h" as symbol
1272 // names and addends of the symbolic expression to add for the operand. The
1273 // value of TagType is currently 1 (for the LLVMOpInfo1 struct). If symbolic
1274 // information is returned then this function returns 1 else it returns 0.
1275 int SymbolizerGetOpInfo(void *DisInfo, uint64_t Pc, uint64_t Offset,
1276 uint64_t Size, int TagType, void *TagBuf) {
1277 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
1278 struct LLVMOpInfo1 *op_info = (struct LLVMOpInfo1 *)TagBuf;
1279 uint64_t value = op_info->Value;
1281 // Make sure all fields returned are zero if we don't set them.
1282 memset((void *)op_info, '\0', sizeof(struct LLVMOpInfo1));
1283 op_info->Value = value;
1285 // If the TagType is not the value 1 which it code knows about or if no
1286 // verbose symbolic information is wanted then just return 0, indicating no
1287 // information is being returned.
1288 if (TagType != 1 || info->verbose == false)
1291 unsigned int Arch = info->O->getArch();
1292 if (Arch == Triple::x86) {
1293 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
1295 // First search the section's relocation entries (if any) for an entry
1296 // for this section offset.
1297 uint32_t sect_addr = info->S.getAddress();
1298 uint32_t sect_offset = (Pc + Offset) - sect_addr;
1299 bool reloc_found = false;
1301 MachO::any_relocation_info RE;
1302 bool isExtern = false;
1304 bool r_scattered = false;
1305 uint32_t r_value, pair_r_value, r_type;
1306 for (const RelocationRef &Reloc : info->S.relocations()) {
1307 uint64_t RelocOffset;
1308 Reloc.getOffset(RelocOffset);
1309 if (RelocOffset == sect_offset) {
1310 Rel = Reloc.getRawDataRefImpl();
1311 RE = info->O->getRelocation(Rel);
1312 r_type = info->O->getAnyRelocationType(RE);
1313 r_scattered = info->O->isRelocationScattered(RE);
1315 r_value = info->O->getScatteredRelocationValue(RE);
1316 if (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
1317 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF) {
1318 DataRefImpl RelNext = Rel;
1319 info->O->moveRelocationNext(RelNext);
1320 MachO::any_relocation_info RENext;
1321 RENext = info->O->getRelocation(RelNext);
1322 if (info->O->isRelocationScattered(RENext))
1323 pair_r_value = info->O->getScatteredRelocationValue(RENext);
1328 isExtern = info->O->getPlainRelocationExternal(RE);
1330 symbol_iterator RelocSym = Reloc.getSymbol();
1338 if (reloc_found && isExtern) {
1340 Symbol.getName(SymName);
1341 const char *name = SymName.data();
1342 op_info->AddSymbol.Present = 1;
1343 op_info->AddSymbol.Name = name;
1344 // For i386 extern relocation entries the value in the instruction is
1345 // the offset from the symbol, and value is already set in op_info->Value.
1348 if (reloc_found && (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
1349 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF)) {
1350 const char *add = GuessSymbolName(r_value, info->AddrMap);
1351 const char *sub = GuessSymbolName(pair_r_value, info->AddrMap);
1352 uint32_t offset = value - (r_value - pair_r_value);
1353 op_info->AddSymbol.Present = 1;
1355 op_info->AddSymbol.Name = add;
1357 op_info->AddSymbol.Value = r_value;
1358 op_info->SubtractSymbol.Present = 1;
1360 op_info->SubtractSymbol.Name = sub;
1362 op_info->SubtractSymbol.Value = pair_r_value;
1363 op_info->Value = offset;
1367 // Second search the external relocation entries of a fully linked image
1368 // (if any) for an entry that matches this segment offset.
1369 // uint32_t seg_offset = (Pc + Offset);
1371 } else if (Arch == Triple::x86_64) {
1372 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
1374 // First search the section's relocation entries (if any) for an entry
1375 // for this section offset.
1376 uint64_t sect_addr = info->S.getAddress();
1377 uint64_t sect_offset = (Pc + Offset) - sect_addr;
1378 bool reloc_found = false;
1380 MachO::any_relocation_info RE;
1381 bool isExtern = false;
1383 for (const RelocationRef &Reloc : info->S.relocations()) {
1384 uint64_t RelocOffset;
1385 Reloc.getOffset(RelocOffset);
1386 if (RelocOffset == sect_offset) {
1387 Rel = Reloc.getRawDataRefImpl();
1388 RE = info->O->getRelocation(Rel);
1389 // NOTE: Scattered relocations don't exist on x86_64.
1390 isExtern = info->O->getPlainRelocationExternal(RE);
1392 symbol_iterator RelocSym = Reloc.getSymbol();
1399 if (reloc_found && isExtern) {
1400 // The Value passed in will be adjusted by the Pc if the instruction
1401 // adds the Pc. But for x86_64 external relocation entries the Value
1402 // is the offset from the external symbol.
1403 if (info->O->getAnyRelocationPCRel(RE))
1404 op_info->Value -= Pc + Offset + Size;
1406 Symbol.getName(SymName);
1407 const char *name = SymName.data();
1408 unsigned Type = info->O->getAnyRelocationType(RE);
1409 if (Type == MachO::X86_64_RELOC_SUBTRACTOR) {
1410 DataRefImpl RelNext = Rel;
1411 info->O->moveRelocationNext(RelNext);
1412 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
1413 unsigned TypeNext = info->O->getAnyRelocationType(RENext);
1414 bool isExternNext = info->O->getPlainRelocationExternal(RENext);
1415 unsigned SymbolNum = info->O->getPlainRelocationSymbolNum(RENext);
1416 if (TypeNext == MachO::X86_64_RELOC_UNSIGNED && isExternNext) {
1417 op_info->SubtractSymbol.Present = 1;
1418 op_info->SubtractSymbol.Name = name;
1419 symbol_iterator RelocSymNext = info->O->getSymbolByIndex(SymbolNum);
1420 Symbol = *RelocSymNext;
1421 StringRef SymNameNext;
1422 Symbol.getName(SymNameNext);
1423 name = SymNameNext.data();
1426 // TODO: add the VariantKinds to op_info->VariantKind for relocation types
1427 // like: X86_64_RELOC_TLV, X86_64_RELOC_GOT_LOAD and X86_64_RELOC_GOT.
1428 op_info->AddSymbol.Present = 1;
1429 op_info->AddSymbol.Name = name;
1433 // Second search the external relocation entries of a fully linked image
1434 // (if any) for an entry that matches this segment offset.
1435 // uint64_t seg_offset = (Pc + Offset);
1437 } else if (Arch == Triple::arm) {
1438 if (Offset != 0 || (Size != 4 && Size != 2))
1440 // First search the section's relocation entries (if any) for an entry
1441 // for this section offset.
1442 uint32_t sect_addr = info->S.getAddress();
1443 uint32_t sect_offset = (Pc + Offset) - sect_addr;
1444 bool reloc_found = false;
1446 MachO::any_relocation_info RE;
1447 bool isExtern = false;
1449 bool r_scattered = false;
1450 uint32_t r_value, pair_r_value, r_type, r_length, other_half;
1451 for (const RelocationRef &Reloc : info->S.relocations()) {
1452 uint64_t RelocOffset;
1453 Reloc.getOffset(RelocOffset);
1454 if (RelocOffset == sect_offset) {
1455 Rel = Reloc.getRawDataRefImpl();
1456 RE = info->O->getRelocation(Rel);
1457 r_length = info->O->getAnyRelocationLength(RE);
1458 r_scattered = info->O->isRelocationScattered(RE);
1460 r_value = info->O->getScatteredRelocationValue(RE);
1461 r_type = info->O->getScatteredRelocationType(RE);
1463 r_type = info->O->getAnyRelocationType(RE);
1464 isExtern = info->O->getPlainRelocationExternal(RE);
1466 symbol_iterator RelocSym = Reloc.getSymbol();
1470 if (r_type == MachO::ARM_RELOC_HALF ||
1471 r_type == MachO::ARM_RELOC_SECTDIFF ||
1472 r_type == MachO::ARM_RELOC_LOCAL_SECTDIFF ||
1473 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1474 DataRefImpl RelNext = Rel;
1475 info->O->moveRelocationNext(RelNext);
1476 MachO::any_relocation_info RENext;
1477 RENext = info->O->getRelocation(RelNext);
1478 other_half = info->O->getAnyRelocationAddress(RENext) & 0xffff;
1479 if (info->O->isRelocationScattered(RENext))
1480 pair_r_value = info->O->getScatteredRelocationValue(RENext);
1486 if (reloc_found && isExtern) {
1488 Symbol.getName(SymName);
1489 const char *name = SymName.data();
1490 op_info->AddSymbol.Present = 1;
1491 op_info->AddSymbol.Name = name;
1494 case MachO::ARM_RELOC_HALF:
1495 if ((r_length & 0x1) == 1) {
1496 op_info->Value = value << 16 | other_half;
1497 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1499 op_info->Value = other_half << 16 | value;
1500 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1508 case MachO::ARM_RELOC_HALF:
1509 if ((r_length & 0x1) == 1) {
1510 op_info->Value = value << 16 | other_half;
1511 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1513 op_info->Value = other_half << 16 | value;
1514 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1523 // If we have a branch that is not an external relocation entry then
1524 // return 0 so the code in tryAddingSymbolicOperand() can use the
1525 // SymbolLookUp call back with the branch target address to look up the
1526 // symbol and possiblity add an annotation for a symbol stub.
1527 if (reloc_found && isExtern == 0 && (r_type == MachO::ARM_RELOC_BR24 ||
1528 r_type == MachO::ARM_THUMB_RELOC_BR22))
1531 uint32_t offset = 0;
1533 if (r_type == MachO::ARM_RELOC_HALF ||
1534 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1535 if ((r_length & 0x1) == 1)
1536 value = value << 16 | other_half;
1538 value = other_half << 16 | value;
1540 if (r_scattered && (r_type != MachO::ARM_RELOC_HALF &&
1541 r_type != MachO::ARM_RELOC_HALF_SECTDIFF)) {
1542 offset = value - r_value;
1547 if (reloc_found && r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1548 if ((r_length & 0x1) == 1)
1549 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1551 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1552 const char *add = GuessSymbolName(r_value, info->AddrMap);
1553 const char *sub = GuessSymbolName(pair_r_value, info->AddrMap);
1554 int32_t offset = value - (r_value - pair_r_value);
1555 op_info->AddSymbol.Present = 1;
1557 op_info->AddSymbol.Name = add;
1559 op_info->AddSymbol.Value = r_value;
1560 op_info->SubtractSymbol.Present = 1;
1562 op_info->SubtractSymbol.Name = sub;
1564 op_info->SubtractSymbol.Value = pair_r_value;
1565 op_info->Value = offset;
1569 if (reloc_found == false)
1572 op_info->AddSymbol.Present = 1;
1573 op_info->Value = offset;
1575 if (r_type == MachO::ARM_RELOC_HALF) {
1576 if ((r_length & 0x1) == 1)
1577 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1579 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1582 const char *add = GuessSymbolName(value, info->AddrMap);
1583 if (add != nullptr) {
1584 op_info->AddSymbol.Name = add;
1587 op_info->AddSymbol.Value = value;
1589 } else if (Arch == Triple::aarch64) {
1590 if (Offset != 0 || Size != 4)
1592 // First search the section's relocation entries (if any) for an entry
1593 // for this section offset.
1594 uint64_t sect_addr = info->S.getAddress();
1595 uint64_t sect_offset = (Pc + Offset) - sect_addr;
1596 bool reloc_found = false;
1598 MachO::any_relocation_info RE;
1599 bool isExtern = false;
1601 uint32_t r_type = 0;
1602 for (const RelocationRef &Reloc : info->S.relocations()) {
1603 uint64_t RelocOffset;
1604 Reloc.getOffset(RelocOffset);
1605 if (RelocOffset == sect_offset) {
1606 Rel = Reloc.getRawDataRefImpl();
1607 RE = info->O->getRelocation(Rel);
1608 r_type = info->O->getAnyRelocationType(RE);
1609 if (r_type == MachO::ARM64_RELOC_ADDEND) {
1610 DataRefImpl RelNext = Rel;
1611 info->O->moveRelocationNext(RelNext);
1612 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
1614 value = info->O->getPlainRelocationSymbolNum(RENext);
1615 op_info->Value = value;
1618 // NOTE: Scattered relocations don't exist on arm64.
1619 isExtern = info->O->getPlainRelocationExternal(RE);
1621 symbol_iterator RelocSym = Reloc.getSymbol();
1628 if (reloc_found && isExtern) {
1630 Symbol.getName(SymName);
1631 const char *name = SymName.data();
1632 op_info->AddSymbol.Present = 1;
1633 op_info->AddSymbol.Name = name;
1636 case MachO::ARM64_RELOC_PAGE21:
1638 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGE;
1640 case MachO::ARM64_RELOC_PAGEOFF12:
1642 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGEOFF;
1644 case MachO::ARM64_RELOC_GOT_LOAD_PAGE21:
1646 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGE;
1648 case MachO::ARM64_RELOC_GOT_LOAD_PAGEOFF12:
1650 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGEOFF;
1652 case MachO::ARM64_RELOC_TLVP_LOAD_PAGE21:
1653 /* @tvlppage is not implemented in llvm-mc */
1654 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVP;
1656 case MachO::ARM64_RELOC_TLVP_LOAD_PAGEOFF12:
1657 /* @tvlppageoff is not implemented in llvm-mc */
1658 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVOFF;
1661 case MachO::ARM64_RELOC_BRANCH26:
1662 op_info->VariantKind = LLVMDisassembler_VariantKind_None;
1673 // GuessCstringPointer is passed the address of what might be a pointer to a
1674 // literal string in a cstring section. If that address is in a cstring section
1675 // it returns a pointer to that string. Else it returns nullptr.
1676 const char *GuessCstringPointer(uint64_t ReferenceValue,
1677 struct DisassembleInfo *info) {
1678 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1679 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1680 for (unsigned I = 0;; ++I) {
1681 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1682 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1683 for (unsigned J = 0; J < Seg.nsects; ++J) {
1684 MachO::section_64 Sec = info->O->getSection64(Load, J);
1685 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1686 if (section_type == MachO::S_CSTRING_LITERALS &&
1687 ReferenceValue >= Sec.addr &&
1688 ReferenceValue < Sec.addr + Sec.size) {
1689 uint64_t sect_offset = ReferenceValue - Sec.addr;
1690 uint64_t object_offset = Sec.offset + sect_offset;
1691 StringRef MachOContents = info->O->getData();
1692 uint64_t object_size = MachOContents.size();
1693 const char *object_addr = (const char *)MachOContents.data();
1694 if (object_offset < object_size) {
1695 const char *name = object_addr + object_offset;
1702 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
1703 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
1704 for (unsigned J = 0; J < Seg.nsects; ++J) {
1705 MachO::section Sec = info->O->getSection(Load, J);
1706 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1707 if (section_type == MachO::S_CSTRING_LITERALS &&
1708 ReferenceValue >= Sec.addr &&
1709 ReferenceValue < Sec.addr + Sec.size) {
1710 uint64_t sect_offset = ReferenceValue - Sec.addr;
1711 uint64_t object_offset = Sec.offset + sect_offset;
1712 StringRef MachOContents = info->O->getData();
1713 uint64_t object_size = MachOContents.size();
1714 const char *object_addr = (const char *)MachOContents.data();
1715 if (object_offset < object_size) {
1716 const char *name = object_addr + object_offset;
1724 if (I == LoadCommandCount - 1)
1727 Load = info->O->getNextLoadCommandInfo(Load);
1732 // GuessIndirectSymbol returns the name of the indirect symbol for the
1733 // ReferenceValue passed in or nullptr. This is used when ReferenceValue maybe
1734 // an address of a symbol stub or a lazy or non-lazy pointer to associate the
1735 // symbol name being referenced by the stub or pointer.
1736 static const char *GuessIndirectSymbol(uint64_t ReferenceValue,
1737 struct DisassembleInfo *info) {
1738 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1739 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1740 MachO::dysymtab_command Dysymtab = info->O->getDysymtabLoadCommand();
1741 MachO::symtab_command Symtab = info->O->getSymtabLoadCommand();
1742 for (unsigned I = 0;; ++I) {
1743 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1744 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1745 for (unsigned J = 0; J < Seg.nsects; ++J) {
1746 MachO::section_64 Sec = info->O->getSection64(Load, J);
1747 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1748 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
1749 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
1750 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
1751 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
1752 section_type == MachO::S_SYMBOL_STUBS) &&
1753 ReferenceValue >= Sec.addr &&
1754 ReferenceValue < Sec.addr + Sec.size) {
1756 if (section_type == MachO::S_SYMBOL_STUBS)
1757 stride = Sec.reserved2;
1762 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
1763 if (index < Dysymtab.nindirectsyms) {
1764 uint32_t indirect_symbol =
1765 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
1766 if (indirect_symbol < Symtab.nsyms) {
1767 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
1768 SymbolRef Symbol = *Sym;
1770 Symbol.getName(SymName);
1771 const char *name = SymName.data();
1777 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
1778 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
1779 for (unsigned J = 0; J < Seg.nsects; ++J) {
1780 MachO::section Sec = info->O->getSection(Load, J);
1781 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1782 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
1783 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
1784 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
1785 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
1786 section_type == MachO::S_SYMBOL_STUBS) &&
1787 ReferenceValue >= Sec.addr &&
1788 ReferenceValue < Sec.addr + Sec.size) {
1790 if (section_type == MachO::S_SYMBOL_STUBS)
1791 stride = Sec.reserved2;
1796 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
1797 if (index < Dysymtab.nindirectsyms) {
1798 uint32_t indirect_symbol =
1799 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
1800 if (indirect_symbol < Symtab.nsyms) {
1801 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
1802 SymbolRef Symbol = *Sym;
1804 Symbol.getName(SymName);
1805 const char *name = SymName.data();
1812 if (I == LoadCommandCount - 1)
1815 Load = info->O->getNextLoadCommandInfo(Load);
1820 // method_reference() is called passing it the ReferenceName that might be
1821 // a reference it to an Objective-C method call. If so then it allocates and
1822 // assembles a method call string with the values last seen and saved in
1823 // the DisassembleInfo's class_name and selector_name fields. This is saved
1824 // into the method field of the info and any previous string is free'ed.
1825 // Then the class_name field in the info is set to nullptr. The method call
1826 // string is set into ReferenceName and ReferenceType is set to
1827 // LLVMDisassembler_ReferenceType_Out_Objc_Message. If this not a method call
1828 // then both ReferenceType and ReferenceName are left unchanged.
1829 static void method_reference(struct DisassembleInfo *info,
1830 uint64_t *ReferenceType,
1831 const char **ReferenceName) {
1832 unsigned int Arch = info->O->getArch();
1833 if (*ReferenceName != nullptr) {
1834 if (strcmp(*ReferenceName, "_objc_msgSend") == 0) {
1835 if (info->selector_name != nullptr) {
1836 if (info->method != nullptr)
1838 if (info->class_name != nullptr) {
1839 info->method = (char *)malloc(5 + strlen(info->class_name) +
1840 strlen(info->selector_name));
1841 if (info->method != nullptr) {
1842 strcpy(info->method, "+[");
1843 strcat(info->method, info->class_name);
1844 strcat(info->method, " ");
1845 strcat(info->method, info->selector_name);
1846 strcat(info->method, "]");
1847 *ReferenceName = info->method;
1848 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1851 info->method = (char *)malloc(9 + strlen(info->selector_name));
1852 if (info->method != nullptr) {
1853 if (Arch == Triple::x86_64)
1854 strcpy(info->method, "-[%rdi ");
1855 else if (Arch == Triple::aarch64)
1856 strcpy(info->method, "-[x0 ");
1858 strcpy(info->method, "-[r? ");
1859 strcat(info->method, info->selector_name);
1860 strcat(info->method, "]");
1861 *ReferenceName = info->method;
1862 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1865 info->class_name = nullptr;
1867 } else if (strcmp(*ReferenceName, "_objc_msgSendSuper2") == 0) {
1868 if (info->selector_name != nullptr) {
1869 if (info->method != nullptr)
1871 info->method = (char *)malloc(17 + strlen(info->selector_name));
1872 if (info->method != nullptr) {
1873 if (Arch == Triple::x86_64)
1874 strcpy(info->method, "-[[%rdi super] ");
1875 else if (Arch == Triple::aarch64)
1876 strcpy(info->method, "-[[x0 super] ");
1878 strcpy(info->method, "-[[r? super] ");
1879 strcat(info->method, info->selector_name);
1880 strcat(info->method, "]");
1881 *ReferenceName = info->method;
1882 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1884 info->class_name = nullptr;
1890 // GuessPointerPointer() is passed the address of what might be a pointer to
1891 // a reference to an Objective-C class, selector, message ref or cfstring.
1892 // If so the value of the pointer is returned and one of the booleans are set
1893 // to true. If not zero is returned and all the booleans are set to false.
1894 static uint64_t GuessPointerPointer(uint64_t ReferenceValue,
1895 struct DisassembleInfo *info,
1896 bool &classref, bool &selref, bool &msgref,
1902 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1903 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1904 for (unsigned I = 0;; ++I) {
1905 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1906 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1907 for (unsigned J = 0; J < Seg.nsects; ++J) {
1908 MachO::section_64 Sec = info->O->getSection64(Load, J);
1909 if ((strncmp(Sec.sectname, "__objc_selrefs", 16) == 0 ||
1910 strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
1911 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0 ||
1912 strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 ||
1913 strncmp(Sec.sectname, "__cfstring", 16) == 0) &&
1914 ReferenceValue >= Sec.addr &&
1915 ReferenceValue < Sec.addr + Sec.size) {
1916 uint64_t sect_offset = ReferenceValue - Sec.addr;
1917 uint64_t object_offset = Sec.offset + sect_offset;
1918 StringRef MachOContents = info->O->getData();
1919 uint64_t object_size = MachOContents.size();
1920 const char *object_addr = (const char *)MachOContents.data();
1921 if (object_offset < object_size) {
1922 uint64_t pointer_value;
1923 memcpy(&pointer_value, object_addr + object_offset,
1925 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1926 sys::swapByteOrder(pointer_value);
1927 if (strncmp(Sec.sectname, "__objc_selrefs", 16) == 0)
1929 else if (strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
1930 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0)
1932 else if (strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 &&
1933 ReferenceValue + 8 < Sec.addr + Sec.size) {
1935 memcpy(&pointer_value, object_addr + object_offset + 8,
1937 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1938 sys::swapByteOrder(pointer_value);
1939 } else if (strncmp(Sec.sectname, "__cfstring", 16) == 0)
1941 return pointer_value;
1948 // TODO: Look for LC_SEGMENT for 32-bit Mach-O files.
1949 if (I == LoadCommandCount - 1)
1952 Load = info->O->getNextLoadCommandInfo(Load);
1957 // get_pointer_64 returns a pointer to the bytes in the object file at the
1958 // Address from a section in the Mach-O file. And indirectly returns the
1959 // offset into the section, number of bytes left in the section past the offset
1960 // and which section is was being referenced. If the Address is not in a
1961 // section nullptr is returned.
1962 const char *get_pointer_64(uint64_t Address, uint32_t &offset, uint32_t &left,
1963 SectionRef &S, DisassembleInfo *info) {
1967 for (unsigned SectIdx = 0; SectIdx != info->Sections->size(); SectIdx++) {
1968 uint64_t SectAddress = ((*(info->Sections))[SectIdx]).getAddress();
1969 uint64_t SectSize = ((*(info->Sections))[SectIdx]).getSize();
1970 if (Address >= SectAddress && Address < SectAddress + SectSize) {
1971 S = (*(info->Sections))[SectIdx];
1972 offset = Address - SectAddress;
1973 left = SectSize - offset;
1974 StringRef SectContents;
1975 ((*(info->Sections))[SectIdx]).getContents(SectContents);
1976 return SectContents.data() + offset;
1982 // get_symbol_64() returns the name of a symbol (or nullptr) and the address of
1983 // the symbol indirectly through n_value. Based on the relocation information
1984 // for the specified section offset in the specified section reference.
1985 const char *get_symbol_64(uint32_t sect_offset, SectionRef S,
1986 DisassembleInfo *info, uint64_t &n_value) {
1988 if (info->verbose == false)
1991 // See if there is an external relocation entry at the sect_offset.
1992 bool reloc_found = false;
1994 MachO::any_relocation_info RE;
1995 bool isExtern = false;
1997 for (const RelocationRef &Reloc : S.relocations()) {
1998 uint64_t RelocOffset;
1999 Reloc.getOffset(RelocOffset);
2000 if (RelocOffset == sect_offset) {
2001 Rel = Reloc.getRawDataRefImpl();
2002 RE = info->O->getRelocation(Rel);
2003 if (info->O->isRelocationScattered(RE))
2005 isExtern = info->O->getPlainRelocationExternal(RE);
2007 symbol_iterator RelocSym = Reloc.getSymbol();
2014 // If there is an external relocation entry for a symbol in this section
2015 // at this section_offset then use that symbol's value for the n_value
2016 // and return its name.
2017 const char *SymbolName = nullptr;
2018 if (reloc_found && isExtern) {
2019 Symbol.getAddress(n_value);
2021 Symbol.getName(name);
2022 if (!name.empty()) {
2023 SymbolName = name.data();
2028 // TODO: For fully linked images, look through the external relocation
2029 // entries off the dynamic symtab command. For these the r_offset is from the
2030 // start of the first writeable segment in the Mach-O file. So the offset
2031 // to this section from that segment is passed to this routine by the caller,
2032 // as the database_offset. Which is the difference of the section's starting
2033 // address and the first writable segment.
2035 // NOTE: need add passing the database_offset to this routine.
2037 // TODO: We did not find an external relocation entry so look up the
2038 // ReferenceValue as an address of a symbol and if found return that symbol's
2041 // NOTE: need add passing the ReferenceValue to this routine. Then that code
2042 // would simply be this:
2043 // SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap);
2048 // These are structs in the Objective-C meta data and read to produce the
2049 // comments for disassembly. While these are part of the ABI they are no
2050 // public defintions. So the are here not in include/llvm/Support/MachO.h .
2052 // The cfstring object in a 64-bit Mach-O file.
2053 struct cfstring64_t {
2054 uint64_t isa; // class64_t * (64-bit pointer)
2055 uint64_t flags; // flag bits
2056 uint64_t characters; // char * (64-bit pointer)
2057 uint64_t length; // number of non-NULL characters in above
2060 // The class object in a 64-bit Mach-O file.
2062 uint64_t isa; // class64_t * (64-bit pointer)
2063 uint64_t superclass; // class64_t * (64-bit pointer)
2064 uint64_t cache; // Cache (64-bit pointer)
2065 uint64_t vtable; // IMP * (64-bit pointer)
2066 uint64_t data; // class_ro64_t * (64-bit pointer)
2069 struct class_ro64_t {
2071 uint32_t instanceStart;
2072 uint32_t instanceSize;
2074 uint64_t ivarLayout; // const uint8_t * (64-bit pointer)
2075 uint64_t name; // const char * (64-bit pointer)
2076 uint64_t baseMethods; // const method_list_t * (64-bit pointer)
2077 uint64_t baseProtocols; // const protocol_list_t * (64-bit pointer)
2078 uint64_t ivars; // const ivar_list_t * (64-bit pointer)
2079 uint64_t weakIvarLayout; // const uint8_t * (64-bit pointer)
2080 uint64_t baseProperties; // const struct objc_property_list (64-bit pointer)
2083 inline void swapStruct(struct cfstring64_t &cfs) {
2084 sys::swapByteOrder(cfs.isa);
2085 sys::swapByteOrder(cfs.flags);
2086 sys::swapByteOrder(cfs.characters);
2087 sys::swapByteOrder(cfs.length);
2090 inline void swapStruct(struct class64_t &c) {
2091 sys::swapByteOrder(c.isa);
2092 sys::swapByteOrder(c.superclass);
2093 sys::swapByteOrder(c.cache);
2094 sys::swapByteOrder(c.vtable);
2095 sys::swapByteOrder(c.data);
2098 inline void swapStruct(struct class_ro64_t &cro) {
2099 sys::swapByteOrder(cro.flags);
2100 sys::swapByteOrder(cro.instanceStart);
2101 sys::swapByteOrder(cro.instanceSize);
2102 sys::swapByteOrder(cro.reserved);
2103 sys::swapByteOrder(cro.ivarLayout);
2104 sys::swapByteOrder(cro.name);
2105 sys::swapByteOrder(cro.baseMethods);
2106 sys::swapByteOrder(cro.baseProtocols);
2107 sys::swapByteOrder(cro.ivars);
2108 sys::swapByteOrder(cro.weakIvarLayout);
2109 sys::swapByteOrder(cro.baseProperties);
2112 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
2113 struct DisassembleInfo *info);
2115 // get_objc2_64bit_class_name() is used for disassembly and is passed a pointer
2116 // to an Objective-C class and returns the class name. It is also passed the
2117 // address of the pointer, so when the pointer is zero as it can be in an .o
2118 // file, that is used to look for an external relocation entry with a symbol
2120 const char *get_objc2_64bit_class_name(uint64_t pointer_value,
2121 uint64_t ReferenceValue,
2122 struct DisassembleInfo *info) {
2124 uint32_t offset, left;
2127 // The pointer_value can be 0 in an object file and have a relocation
2128 // entry for the class symbol at the ReferenceValue (the address of the
2130 if (pointer_value == 0) {
2131 r = get_pointer_64(ReferenceValue, offset, left, S, info);
2132 if (r == nullptr || left < sizeof(uint64_t))
2135 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
2136 if (symbol_name == nullptr)
2138 const char *class_name = strrchr(symbol_name, '$');
2139 if (class_name != nullptr && class_name[1] == '_' && class_name[2] != '\0')
2140 return class_name + 2;
2145 // The case were the pointer_value is non-zero and points to a class defined
2146 // in this Mach-O file.
2147 r = get_pointer_64(pointer_value, offset, left, S, info);
2148 if (r == nullptr || left < sizeof(struct class64_t))
2151 memcpy(&c, r, sizeof(struct class64_t));
2152 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2156 r = get_pointer_64(c.data, offset, left, S, info);
2157 if (r == nullptr || left < sizeof(struct class_ro64_t))
2159 struct class_ro64_t cro;
2160 memcpy(&cro, r, sizeof(struct class_ro64_t));
2161 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2165 const char *name = get_pointer_64(cro.name, offset, left, S, info);
2169 // get_objc2_64bit_cfstring_name is used for disassembly and is passed a
2170 // pointer to a cfstring and returns its name or nullptr.
2171 const char *get_objc2_64bit_cfstring_name(uint64_t ReferenceValue,
2172 struct DisassembleInfo *info) {
2173 const char *r, *name;
2174 uint32_t offset, left;
2176 struct cfstring64_t cfs;
2177 uint64_t cfs_characters;
2179 r = get_pointer_64(ReferenceValue, offset, left, S, info);
2180 if (r == nullptr || left < sizeof(struct cfstring64_t))
2182 memcpy(&cfs, r, sizeof(struct cfstring64_t));
2183 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2185 if (cfs.characters == 0) {
2187 const char *symbol_name = get_symbol_64(
2188 offset + offsetof(struct cfstring64_t, characters), S, info, n_value);
2189 if (symbol_name == nullptr)
2191 cfs_characters = n_value;
2193 cfs_characters = cfs.characters;
2194 name = get_pointer_64(cfs_characters, offset, left, S, info);
2199 // get_objc2_64bit_selref() is used for disassembly and is passed a the address
2200 // of a pointer to an Objective-C selector reference when the pointer value is
2201 // zero as in a .o file and is likely to have a external relocation entry with
2202 // who's symbol's n_value is the real pointer to the selector name. If that is
2203 // the case the real pointer to the selector name is returned else 0 is
2205 uint64_t get_objc2_64bit_selref(uint64_t ReferenceValue,
2206 struct DisassembleInfo *info) {
2207 uint32_t offset, left;
2210 const char *r = get_pointer_64(ReferenceValue, offset, left, S, info);
2211 if (r == nullptr || left < sizeof(uint64_t))
2214 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
2215 if (symbol_name == nullptr)
2220 // GuessLiteralPointer returns a string which for the item in the Mach-O file
2221 // for the address passed in as ReferenceValue for printing as a comment with
2222 // the instruction and also returns the corresponding type of that item
2223 // indirectly through ReferenceType.
2225 // If ReferenceValue is an address of literal cstring then a pointer to the
2226 // cstring is returned and ReferenceType is set to
2227 // LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr .
2229 // If ReferenceValue is an address of an Objective-C CFString, Selector ref or
2230 // Class ref that name is returned and the ReferenceType is set accordingly.
2232 // Lastly, literals which are Symbol address in a literal pool are looked for
2233 // and if found the symbol name is returned and ReferenceType is set to
2234 // LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr .
2236 // If there is no item in the Mach-O file for the address passed in as
2237 // ReferenceValue nullptr is returned and ReferenceType is unchanged.
2238 const char *GuessLiteralPointer(uint64_t ReferenceValue, uint64_t ReferencePC,
2239 uint64_t *ReferenceType,
2240 struct DisassembleInfo *info) {
2241 // First see if there is an external relocation entry at the ReferencePC.
2242 uint64_t sect_addr = info->S.getAddress();
2243 uint64_t sect_offset = ReferencePC - sect_addr;
2244 bool reloc_found = false;
2246 MachO::any_relocation_info RE;
2247 bool isExtern = false;
2249 for (const RelocationRef &Reloc : info->S.relocations()) {
2250 uint64_t RelocOffset;
2251 Reloc.getOffset(RelocOffset);
2252 if (RelocOffset == sect_offset) {
2253 Rel = Reloc.getRawDataRefImpl();
2254 RE = info->O->getRelocation(Rel);
2255 if (info->O->isRelocationScattered(RE))
2257 isExtern = info->O->getPlainRelocationExternal(RE);
2259 symbol_iterator RelocSym = Reloc.getSymbol();
2266 // If there is an external relocation entry for a symbol in a section
2267 // then used that symbol's value for the value of the reference.
2268 if (reloc_found && isExtern) {
2269 if (info->O->getAnyRelocationPCRel(RE)) {
2270 unsigned Type = info->O->getAnyRelocationType(RE);
2271 if (Type == MachO::X86_64_RELOC_SIGNED) {
2272 Symbol.getAddress(ReferenceValue);
2277 // Look for literals such as Objective-C CFStrings refs, Selector refs,
2278 // Message refs and Class refs.
2279 bool classref, selref, msgref, cfstring;
2280 uint64_t pointer_value = GuessPointerPointer(ReferenceValue, info, classref,
2281 selref, msgref, cfstring);
2282 if (classref == true && pointer_value == 0) {
2283 // Note the ReferenceValue is a pointer into the __objc_classrefs section.
2284 // And the pointer_value in that section is typically zero as it will be
2285 // set by dyld as part of the "bind information".
2286 const char *name = get_dyld_bind_info_symbolname(ReferenceValue, info);
2287 if (name != nullptr) {
2288 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
2289 const char *class_name = strrchr(name, '$');
2290 if (class_name != nullptr && class_name[1] == '_' &&
2291 class_name[2] != '\0') {
2292 info->class_name = class_name + 2;
2298 if (classref == true) {
2299 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
2301 get_objc2_64bit_class_name(pointer_value, ReferenceValue, info);
2302 if (name != nullptr)
2303 info->class_name = name;
2305 name = "bad class ref";
2309 if (cfstring == true) {
2310 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_CFString_Ref;
2311 const char *name = get_objc2_64bit_cfstring_name(ReferenceValue, info);
2315 if (selref == true && pointer_value == 0)
2316 pointer_value = get_objc2_64bit_selref(ReferenceValue, info);
2318 if (pointer_value != 0)
2319 ReferenceValue = pointer_value;
2321 const char *name = GuessCstringPointer(ReferenceValue, info);
2323 if (pointer_value != 0 && selref == true) {
2324 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Selector_Ref;
2325 info->selector_name = name;
2326 } else if (pointer_value != 0 && msgref == true) {
2327 info->class_name = nullptr;
2328 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message_Ref;
2329 info->selector_name = name;
2331 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr;
2335 // Lastly look for an indirect symbol with this ReferenceValue which is in
2336 // a literal pool. If found return that symbol name.
2337 name = GuessIndirectSymbol(ReferenceValue, info);
2339 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr;
2346 // SymbolizerSymbolLookUp is the symbol lookup function passed when creating
2347 // the Symbolizer. It looks up the ReferenceValue using the info passed via the
2348 // pointer to the struct DisassembleInfo that was passed when MCSymbolizer
2349 // is created and returns the symbol name that matches the ReferenceValue or
2350 // nullptr if none. The ReferenceType is passed in for the IN type of
2351 // reference the instruction is making from the values in defined in the header
2352 // "llvm-c/Disassembler.h". On return the ReferenceType can set to a specific
2353 // Out type and the ReferenceName will also be set which is added as a comment
2354 // to the disassembled instruction.
2357 // If the symbol name is a C++ mangled name then the demangled name is
2358 // returned through ReferenceName and ReferenceType is set to
2359 // LLVMDisassembler_ReferenceType_DeMangled_Name .
2362 // When this is called to get a symbol name for a branch target then the
2363 // ReferenceType will be LLVMDisassembler_ReferenceType_In_Branch and then
2364 // SymbolValue will be looked for in the indirect symbol table to determine if
2365 // it is an address for a symbol stub. If so then the symbol name for that
2366 // stub is returned indirectly through ReferenceName and then ReferenceType is
2367 // set to LLVMDisassembler_ReferenceType_Out_SymbolStub.
2369 // When this is called with an value loaded via a PC relative load then
2370 // ReferenceType will be LLVMDisassembler_ReferenceType_In_PCrel_Load then the
2371 // SymbolValue is checked to be an address of literal pointer, symbol pointer,
2372 // or an Objective-C meta data reference. If so the output ReferenceType is
2373 // set to correspond to that as well as setting the ReferenceName.
2374 const char *SymbolizerSymbolLookUp(void *DisInfo, uint64_t ReferenceValue,
2375 uint64_t *ReferenceType,
2376 uint64_t ReferencePC,
2377 const char **ReferenceName) {
2378 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
2379 // If no verbose symbolic information is wanted then just return nullptr.
2380 if (info->verbose == false) {
2381 *ReferenceName = nullptr;
2382 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2386 const char *SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap);
2388 if (*ReferenceType == LLVMDisassembler_ReferenceType_In_Branch) {
2389 *ReferenceName = GuessIndirectSymbol(ReferenceValue, info);
2390 if (*ReferenceName != nullptr) {
2391 method_reference(info, ReferenceType, ReferenceName);
2392 if (*ReferenceType != LLVMDisassembler_ReferenceType_Out_Objc_Message)
2393 *ReferenceType = LLVMDisassembler_ReferenceType_Out_SymbolStub;
2396 if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
2397 if (info->demangled_name != nullptr)
2398 free(info->demangled_name);
2400 info->demangled_name =
2401 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
2402 if (info->demangled_name != nullptr) {
2403 *ReferenceName = info->demangled_name;
2404 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
2406 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2409 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2410 } else if (*ReferenceType == LLVMDisassembler_ReferenceType_In_PCrel_Load) {
2412 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2414 method_reference(info, ReferenceType, ReferenceName);
2416 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2417 // If this is arm64 and the reference is an adrp instruction save the
2418 // instruction, passed in ReferenceValue and the address of the instruction
2419 // for use later if we see and add immediate instruction.
2420 } else if (info->O->getArch() == Triple::aarch64 &&
2421 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADRP) {
2422 info->adrp_inst = ReferenceValue;
2423 info->adrp_addr = ReferencePC;
2424 SymbolName = nullptr;
2425 *ReferenceName = nullptr;
2426 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2427 // If this is arm64 and reference is an add immediate instruction and we
2429 // seen an adrp instruction just before it and the adrp's Xd register
2431 // this add's Xn register reconstruct the value being referenced and look to
2432 // see if it is a literal pointer. Note the add immediate instruction is
2433 // passed in ReferenceValue.
2434 } else if (info->O->getArch() == Triple::aarch64 &&
2435 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADDXri &&
2436 ReferencePC - 4 == info->adrp_addr &&
2437 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
2438 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
2439 uint32_t addxri_inst;
2440 uint64_t adrp_imm, addxri_imm;
2443 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
2444 if (info->adrp_inst & 0x0200000)
2445 adrp_imm |= 0xfffffffffc000000LL;
2447 addxri_inst = ReferenceValue;
2448 addxri_imm = (addxri_inst >> 10) & 0xfff;
2449 if (((addxri_inst >> 22) & 0x3) == 1)
2452 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
2453 (adrp_imm << 12) + addxri_imm;
2456 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2457 if (*ReferenceName == nullptr)
2458 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2459 // If this is arm64 and the reference is a load register instruction and we
2460 // have seen an adrp instruction just before it and the adrp's Xd register
2461 // matches this add's Xn register reconstruct the value being referenced and
2462 // look to see if it is a literal pointer. Note the load register
2463 // instruction is passed in ReferenceValue.
2464 } else if (info->O->getArch() == Triple::aarch64 &&
2465 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXui &&
2466 ReferencePC - 4 == info->adrp_addr &&
2467 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
2468 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
2469 uint32_t ldrxui_inst;
2470 uint64_t adrp_imm, ldrxui_imm;
2473 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
2474 if (info->adrp_inst & 0x0200000)
2475 adrp_imm |= 0xfffffffffc000000LL;
2477 ldrxui_inst = ReferenceValue;
2478 ldrxui_imm = (ldrxui_inst >> 10) & 0xfff;
2480 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
2481 (adrp_imm << 12) + (ldrxui_imm << 3);
2484 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2485 if (*ReferenceName == nullptr)
2486 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2488 // If this arm64 and is an load register (PC-relative) instruction the
2489 // ReferenceValue is the PC plus the immediate value.
2490 else if (info->O->getArch() == Triple::aarch64 &&
2491 (*ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXl ||
2492 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADR)) {
2494 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2495 if (*ReferenceName == nullptr)
2496 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2499 else if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
2500 if (info->demangled_name != nullptr)
2501 free(info->demangled_name);
2503 info->demangled_name =
2504 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
2505 if (info->demangled_name != nullptr) {
2506 *ReferenceName = info->demangled_name;
2507 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
2512 *ReferenceName = nullptr;
2513 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2519 /// \brief Emits the comments that are stored in the CommentStream.
2520 /// Each comment in the CommentStream must end with a newline.
2521 static void emitComments(raw_svector_ostream &CommentStream,
2522 SmallString<128> &CommentsToEmit,
2523 formatted_raw_ostream &FormattedOS,
2524 const MCAsmInfo &MAI) {
2525 // Flush the stream before taking its content.
2526 CommentStream.flush();
2527 StringRef Comments = CommentsToEmit.str();
2528 // Get the default information for printing a comment.
2529 const char *CommentBegin = MAI.getCommentString();
2530 unsigned CommentColumn = MAI.getCommentColumn();
2531 bool IsFirst = true;
2532 while (!Comments.empty()) {
2534 FormattedOS << '\n';
2535 // Emit a line of comments.
2536 FormattedOS.PadToColumn(CommentColumn);
2537 size_t Position = Comments.find('\n');
2538 FormattedOS << CommentBegin << ' ' << Comments.substr(0, Position);
2539 // Move after the newline character.
2540 Comments = Comments.substr(Position + 1);
2543 FormattedOS.flush();
2545 // Tell the comment stream that the vector changed underneath it.
2546 CommentsToEmit.clear();
2547 CommentStream.resync();
2550 static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF) {
2551 const char *McpuDefault = nullptr;
2552 const Target *ThumbTarget = nullptr;
2553 const Target *TheTarget = GetTarget(MachOOF, &McpuDefault, &ThumbTarget);
2555 // GetTarget prints out stuff.
2558 if (MCPU.empty() && McpuDefault)
2561 std::unique_ptr<const MCInstrInfo> InstrInfo(TheTarget->createMCInstrInfo());
2562 std::unique_ptr<const MCInstrInfo> ThumbInstrInfo;
2564 ThumbInstrInfo.reset(ThumbTarget->createMCInstrInfo());
2566 // Package up features to be passed to target/subtarget
2567 std::string FeaturesStr;
2568 if (MAttrs.size()) {
2569 SubtargetFeatures Features;
2570 for (unsigned i = 0; i != MAttrs.size(); ++i)
2571 Features.AddFeature(MAttrs[i]);
2572 FeaturesStr = Features.getString();
2575 // Set up disassembler.
2576 std::unique_ptr<const MCRegisterInfo> MRI(
2577 TheTarget->createMCRegInfo(TripleName));
2578 std::unique_ptr<const MCAsmInfo> AsmInfo(
2579 TheTarget->createMCAsmInfo(*MRI, TripleName));
2580 std::unique_ptr<const MCSubtargetInfo> STI(
2581 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr));
2582 MCContext Ctx(AsmInfo.get(), MRI.get(), nullptr);
2583 std::unique_ptr<MCDisassembler> DisAsm(
2584 TheTarget->createMCDisassembler(*STI, Ctx));
2585 std::unique_ptr<MCSymbolizer> Symbolizer;
2586 struct DisassembleInfo SymbolizerInfo;
2587 std::unique_ptr<MCRelocationInfo> RelInfo(
2588 TheTarget->createMCRelocationInfo(TripleName, Ctx));
2590 Symbolizer.reset(TheTarget->createMCSymbolizer(
2591 TripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
2592 &SymbolizerInfo, &Ctx, std::move(RelInfo)));
2593 DisAsm->setSymbolizer(std::move(Symbolizer));
2595 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
2596 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
2597 AsmPrinterVariant, *AsmInfo, *InstrInfo, *MRI, *STI));
2598 // Set the display preference for hex vs. decimal immediates.
2599 IP->setPrintImmHex(PrintImmHex);
2600 // Comment stream and backing vector.
2601 SmallString<128> CommentsToEmit;
2602 raw_svector_ostream CommentStream(CommentsToEmit);
2603 // FIXME: Setting the CommentStream in the InstPrinter is problematic in that
2604 // if it is done then arm64 comments for string literals don't get printed
2605 // and some constant get printed instead and not setting it causes intel
2606 // (32-bit and 64-bit) comments printed with different spacing before the
2607 // comment causing different diffs with the 'C' disassembler library API.
2608 // IP->setCommentStream(CommentStream);
2610 if (!AsmInfo || !STI || !DisAsm || !IP) {
2611 errs() << "error: couldn't initialize disassembler for target "
2612 << TripleName << '\n';
2616 // Set up thumb disassembler.
2617 std::unique_ptr<const MCRegisterInfo> ThumbMRI;
2618 std::unique_ptr<const MCAsmInfo> ThumbAsmInfo;
2619 std::unique_ptr<const MCSubtargetInfo> ThumbSTI;
2620 std::unique_ptr<MCDisassembler> ThumbDisAsm;
2621 std::unique_ptr<MCInstPrinter> ThumbIP;
2622 std::unique_ptr<MCContext> ThumbCtx;
2623 std::unique_ptr<MCSymbolizer> ThumbSymbolizer;
2624 struct DisassembleInfo ThumbSymbolizerInfo;
2625 std::unique_ptr<MCRelocationInfo> ThumbRelInfo;
2627 ThumbMRI.reset(ThumbTarget->createMCRegInfo(ThumbTripleName));
2629 ThumbTarget->createMCAsmInfo(*ThumbMRI, ThumbTripleName));
2631 ThumbTarget->createMCSubtargetInfo(ThumbTripleName, MCPU, FeaturesStr));
2632 ThumbCtx.reset(new MCContext(ThumbAsmInfo.get(), ThumbMRI.get(), nullptr));
2633 ThumbDisAsm.reset(ThumbTarget->createMCDisassembler(*ThumbSTI, *ThumbCtx));
2634 MCContext *PtrThumbCtx = ThumbCtx.get();
2636 ThumbTarget->createMCRelocationInfo(ThumbTripleName, *PtrThumbCtx));
2638 ThumbSymbolizer.reset(ThumbTarget->createMCSymbolizer(
2639 ThumbTripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
2640 &ThumbSymbolizerInfo, PtrThumbCtx, std::move(ThumbRelInfo)));
2641 ThumbDisAsm->setSymbolizer(std::move(ThumbSymbolizer));
2643 int ThumbAsmPrinterVariant = ThumbAsmInfo->getAssemblerDialect();
2644 ThumbIP.reset(ThumbTarget->createMCInstPrinter(
2645 ThumbAsmPrinterVariant, *ThumbAsmInfo, *ThumbInstrInfo, *ThumbMRI,
2647 // Set the display preference for hex vs. decimal immediates.
2648 ThumbIP->setPrintImmHex(PrintImmHex);
2651 if (ThumbTarget && (!ThumbAsmInfo || !ThumbSTI || !ThumbDisAsm || !ThumbIP)) {
2652 errs() << "error: couldn't initialize disassembler for target "
2653 << ThumbTripleName << '\n';
2657 MachO::mach_header Header = MachOOF->getHeader();
2659 // FIXME: Using the -cfg command line option, this code used to be able to
2660 // annotate relocations with the referenced symbol's name, and if this was
2661 // inside a __[cf]string section, the data it points to. This is now replaced
2662 // by the upcoming MCSymbolizer, which needs the appropriate setup done above.
2663 std::vector<SectionRef> Sections;
2664 std::vector<SymbolRef> Symbols;
2665 SmallVector<uint64_t, 8> FoundFns;
2666 uint64_t BaseSegmentAddress;
2668 getSectionsAndSymbols(Header, MachOOF, Sections, Symbols, FoundFns,
2669 BaseSegmentAddress);
2671 // Sort the symbols by address, just in case they didn't come in that way.
2672 std::sort(Symbols.begin(), Symbols.end(), SymbolSorter());
2674 // Build a data in code table that is sorted on by the address of each entry.
2675 uint64_t BaseAddress = 0;
2676 if (Header.filetype == MachO::MH_OBJECT)
2677 BaseAddress = Sections[0].getAddress();
2679 BaseAddress = BaseSegmentAddress;
2681 for (dice_iterator DI = MachOOF->begin_dices(), DE = MachOOF->end_dices();
2684 DI->getOffset(Offset);
2685 Dices.push_back(std::make_pair(BaseAddress + Offset, *DI));
2687 array_pod_sort(Dices.begin(), Dices.end());
2690 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
2692 raw_ostream &DebugOut = nulls();
2695 std::unique_ptr<DIContext> diContext;
2696 ObjectFile *DbgObj = MachOOF;
2697 // Try to find debug info and set up the DIContext for it.
2699 // A separate DSym file path was specified, parse it as a macho file,
2700 // get the sections and supply it to the section name parsing machinery.
2701 if (!DSYMFile.empty()) {
2702 ErrorOr<std::unique_ptr<MemoryBuffer>> BufOrErr =
2703 MemoryBuffer::getFileOrSTDIN(DSYMFile);
2704 if (std::error_code EC = BufOrErr.getError()) {
2705 errs() << "llvm-objdump: " << Filename << ": " << EC.message() << '\n';
2709 ObjectFile::createMachOObjectFile(BufOrErr.get()->getMemBufferRef())
2714 // Setup the DIContext
2715 diContext.reset(DIContext::getDWARFContext(*DbgObj));
2718 // TODO: For now this only disassembles the (__TEXT,__text) section (see the
2719 // checks in the code below at the top of this loop). It should allow a
2720 // darwin otool(1) like -s option to disassemble any named segment & section
2721 // that is marked as containing instructions with the attributes
2722 // S_ATTR_PURE_INSTRUCTIONS or S_ATTR_SOME_INSTRUCTIONS in the flags field of
2723 // the section structure.
2724 outs() << "(__TEXT,__text) section\n";
2726 for (unsigned SectIdx = 0; SectIdx != Sections.size(); SectIdx++) {
2728 bool SectIsText = Sections[SectIdx].isText();
2729 if (SectIsText == false)
2733 if (Sections[SectIdx].getName(SectName) || SectName != "__text")
2734 continue; // Skip non-text sections
2736 DataRefImpl DR = Sections[SectIdx].getRawDataRefImpl();
2738 StringRef SegmentName = MachOOF->getSectionFinalSegmentName(DR);
2739 if (SegmentName != "__TEXT")
2743 Sections[SectIdx].getContents(BytesStr);
2744 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()),
2746 uint64_t SectAddress = Sections[SectIdx].getAddress();
2748 bool symbolTableWorked = false;
2750 // Parse relocations.
2751 std::vector<std::pair<uint64_t, SymbolRef>> Relocs;
2752 for (const RelocationRef &Reloc : Sections[SectIdx].relocations()) {
2753 uint64_t RelocOffset;
2754 Reloc.getOffset(RelocOffset);
2755 uint64_t SectionAddress = Sections[SectIdx].getAddress();
2756 RelocOffset -= SectionAddress;
2758 symbol_iterator RelocSym = Reloc.getSymbol();
2760 Relocs.push_back(std::make_pair(RelocOffset, *RelocSym));
2762 array_pod_sort(Relocs.begin(), Relocs.end());
2764 // Create a map of symbol addresses to symbol names for use by
2765 // the SymbolizerSymbolLookUp() routine.
2766 SymbolAddressMap AddrMap;
2767 for (const SymbolRef &Symbol : MachOOF->symbols()) {
2770 if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data ||
2771 ST == SymbolRef::ST_Other) {
2773 Symbol.getAddress(Address);
2775 Symbol.getName(SymName);
2776 AddrMap[Address] = SymName;
2779 // Set up the block of info used by the Symbolizer call backs.
2780 SymbolizerInfo.verbose = true;
2781 SymbolizerInfo.O = MachOOF;
2782 SymbolizerInfo.S = Sections[SectIdx];
2783 SymbolizerInfo.AddrMap = &AddrMap;
2784 SymbolizerInfo.Sections = &Sections;
2785 SymbolizerInfo.class_name = nullptr;
2786 SymbolizerInfo.selector_name = nullptr;
2787 SymbolizerInfo.method = nullptr;
2788 SymbolizerInfo.demangled_name = nullptr;
2789 SymbolizerInfo.bindtable = nullptr;
2790 SymbolizerInfo.adrp_addr = 0;
2791 SymbolizerInfo.adrp_inst = 0;
2792 // Same for the ThumbSymbolizer
2793 ThumbSymbolizerInfo.verbose = true;
2794 ThumbSymbolizerInfo.O = MachOOF;
2795 ThumbSymbolizerInfo.S = Sections[SectIdx];
2796 ThumbSymbolizerInfo.AddrMap = &AddrMap;
2797 ThumbSymbolizerInfo.Sections = &Sections;
2798 ThumbSymbolizerInfo.class_name = nullptr;
2799 ThumbSymbolizerInfo.selector_name = nullptr;
2800 ThumbSymbolizerInfo.method = nullptr;
2801 ThumbSymbolizerInfo.demangled_name = nullptr;
2802 ThumbSymbolizerInfo.bindtable = nullptr;
2803 ThumbSymbolizerInfo.adrp_addr = 0;
2804 ThumbSymbolizerInfo.adrp_inst = 0;
2806 // Disassemble symbol by symbol.
2807 for (unsigned SymIdx = 0; SymIdx != Symbols.size(); SymIdx++) {
2809 Symbols[SymIdx].getName(SymName);
2812 Symbols[SymIdx].getType(ST);
2813 if (ST != SymbolRef::ST_Function)
2816 // Make sure the symbol is defined in this section.
2817 bool containsSym = Sections[SectIdx].containsSymbol(Symbols[SymIdx]);
2821 // Start at the address of the symbol relative to the section's address.
2823 uint64_t SectionAddress = Sections[SectIdx].getAddress();
2824 Symbols[SymIdx].getAddress(Start);
2825 Start -= SectionAddress;
2827 // Stop disassembling either at the beginning of the next symbol or at
2828 // the end of the section.
2829 bool containsNextSym = false;
2830 uint64_t NextSym = 0;
2831 uint64_t NextSymIdx = SymIdx + 1;
2832 while (Symbols.size() > NextSymIdx) {
2833 SymbolRef::Type NextSymType;
2834 Symbols[NextSymIdx].getType(NextSymType);
2835 if (NextSymType == SymbolRef::ST_Function) {
2837 Sections[SectIdx].containsSymbol(Symbols[NextSymIdx]);
2838 Symbols[NextSymIdx].getAddress(NextSym);
2839 NextSym -= SectionAddress;
2845 uint64_t SectSize = Sections[SectIdx].getSize();
2846 uint64_t End = containsNextSym ? NextSym : SectSize;
2849 symbolTableWorked = true;
2851 DataRefImpl Symb = Symbols[SymIdx].getRawDataRefImpl();
2853 (MachOOF->getSymbolFlags(Symb) & SymbolRef::SF_Thumb) && ThumbTarget;
2855 outs() << SymName << ":\n";
2856 DILineInfo lastLine;
2857 for (uint64_t Index = Start; Index < End; Index += Size) {
2860 uint64_t PC = SectAddress + Index;
2861 if (FullLeadingAddr) {
2862 if (MachOOF->is64Bit())
2863 outs() << format("%016" PRIx64, PC);
2865 outs() << format("%08" PRIx64, PC);
2867 outs() << format("%8" PRIx64 ":", PC);
2872 // Check the data in code table here to see if this is data not an
2873 // instruction to be disassembled.
2875 Dice.push_back(std::make_pair(PC, DiceRef()));
2876 dice_table_iterator DTI =
2877 std::search(Dices.begin(), Dices.end(), Dice.begin(), Dice.end(),
2878 compareDiceTableEntries);
2879 if (DTI != Dices.end()) {
2881 DTI->second.getLength(Length);
2883 DTI->second.getKind(Kind);
2884 Size = DumpDataInCode(reinterpret_cast<const char *>(Bytes.data()) +
2887 if ((Kind == MachO::DICE_KIND_JUMP_TABLE8) &&
2888 (PC == (DTI->first + Length - 1)) && (Length & 1))
2893 SmallVector<char, 64> AnnotationsBytes;
2894 raw_svector_ostream Annotations(AnnotationsBytes);
2898 gotInst = ThumbDisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
2899 PC, DebugOut, Annotations);
2901 gotInst = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), PC,
2902 DebugOut, Annotations);
2904 if (!NoShowRawInsn) {
2905 DumpBytes(StringRef(
2906 reinterpret_cast<const char *>(Bytes.data()) + Index, Size));
2908 formatted_raw_ostream FormattedOS(outs());
2909 Annotations.flush();
2910 StringRef AnnotationsStr = Annotations.str();
2912 ThumbIP->printInst(&Inst, FormattedOS, AnnotationsStr);
2914 IP->printInst(&Inst, FormattedOS, AnnotationsStr);
2915 emitComments(CommentStream, CommentsToEmit, FormattedOS, *AsmInfo);
2917 // Print debug info.
2919 DILineInfo dli = diContext->getLineInfoForAddress(PC);
2920 // Print valid line info if it changed.
2921 if (dli != lastLine && dli.Line != 0)
2922 outs() << "\t## " << dli.FileName << ':' << dli.Line << ':'
2928 unsigned int Arch = MachOOF->getArch();
2929 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
2930 outs() << format("\t.byte 0x%02x #bad opcode\n",
2931 *(Bytes.data() + Index) & 0xff);
2932 Size = 1; // skip exactly one illegible byte and move on.
2933 } else if (Arch == Triple::aarch64) {
2934 uint32_t opcode = (*(Bytes.data() + Index) & 0xff) |
2935 (*(Bytes.data() + Index + 1) & 0xff) << 8 |
2936 (*(Bytes.data() + Index + 2) & 0xff) << 16 |
2937 (*(Bytes.data() + Index + 3) & 0xff) << 24;
2938 outs() << format("\t.long\t0x%08x\n", opcode);
2941 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
2943 Size = 1; // skip illegible bytes
2948 if (!symbolTableWorked) {
2949 // Reading the symbol table didn't work, disassemble the whole section.
2950 uint64_t SectAddress = Sections[SectIdx].getAddress();
2951 uint64_t SectSize = Sections[SectIdx].getSize();
2953 for (uint64_t Index = 0; Index < SectSize; Index += InstSize) {
2956 uint64_t PC = SectAddress + Index;
2957 if (DisAsm->getInstruction(Inst, InstSize, Bytes.slice(Index), PC,
2958 DebugOut, nulls())) {
2959 if (FullLeadingAddr) {
2960 if (MachOOF->is64Bit())
2961 outs() << format("%016" PRIx64, PC);
2963 outs() << format("%08" PRIx64, PC);
2965 outs() << format("%8" PRIx64 ":", PC);
2967 if (!NoShowRawInsn) {
2970 StringRef(reinterpret_cast<const char *>(Bytes.data()) + Index,
2973 IP->printInst(&Inst, outs(), "");
2976 unsigned int Arch = MachOOF->getArch();
2977 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
2978 outs() << format("\t.byte 0x%02x #bad opcode\n",
2979 *(Bytes.data() + Index) & 0xff);
2980 InstSize = 1; // skip exactly one illegible byte and move on.
2982 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
2984 InstSize = 1; // skip illegible bytes
2989 // The TripleName's need to be reset if we are called again for a different
2992 ThumbTripleName = "";
2994 if (SymbolizerInfo.method != nullptr)
2995 free(SymbolizerInfo.method);
2996 if (SymbolizerInfo.demangled_name != nullptr)
2997 free(SymbolizerInfo.demangled_name);
2998 if (SymbolizerInfo.bindtable != nullptr)
2999 delete SymbolizerInfo.bindtable;
3000 if (ThumbSymbolizerInfo.method != nullptr)
3001 free(ThumbSymbolizerInfo.method);
3002 if (ThumbSymbolizerInfo.demangled_name != nullptr)
3003 free(ThumbSymbolizerInfo.demangled_name);
3004 if (ThumbSymbolizerInfo.bindtable != nullptr)
3005 delete ThumbSymbolizerInfo.bindtable;
3009 //===----------------------------------------------------------------------===//
3010 // __compact_unwind section dumping
3011 //===----------------------------------------------------------------------===//
3015 template <typename T> static uint64_t readNext(const char *&Buf) {
3016 using llvm::support::little;
3017 using llvm::support::unaligned;
3019 uint64_t Val = support::endian::read<T, little, unaligned>(Buf);
3024 struct CompactUnwindEntry {
3025 uint32_t OffsetInSection;
3027 uint64_t FunctionAddr;
3029 uint32_t CompactEncoding;
3030 uint64_t PersonalityAddr;
3033 RelocationRef FunctionReloc;
3034 RelocationRef PersonalityReloc;
3035 RelocationRef LSDAReloc;
3037 CompactUnwindEntry(StringRef Contents, unsigned Offset, bool Is64)
3038 : OffsetInSection(Offset) {
3040 read<uint64_t>(Contents.data() + Offset);
3042 read<uint32_t>(Contents.data() + Offset);
3046 template <typename UIntPtr> void read(const char *Buf) {
3047 FunctionAddr = readNext<UIntPtr>(Buf);
3048 Length = readNext<uint32_t>(Buf);
3049 CompactEncoding = readNext<uint32_t>(Buf);
3050 PersonalityAddr = readNext<UIntPtr>(Buf);
3051 LSDAAddr = readNext<UIntPtr>(Buf);
3056 /// Given a relocation from __compact_unwind, consisting of the RelocationRef
3057 /// and data being relocated, determine the best base Name and Addend to use for
3058 /// display purposes.
3060 /// 1. An Extern relocation will directly reference a symbol (and the data is
3061 /// then already an addend), so use that.
3062 /// 2. Otherwise the data is an offset in the object file's layout; try to find
3063 // a symbol before it in the same section, and use the offset from there.
3064 /// 3. Finally, if all that fails, fall back to an offset from the start of the
3065 /// referenced section.
3066 static void findUnwindRelocNameAddend(const MachOObjectFile *Obj,
3067 std::map<uint64_t, SymbolRef> &Symbols,
3068 const RelocationRef &Reloc, uint64_t Addr,
3069 StringRef &Name, uint64_t &Addend) {
3070 if (Reloc.getSymbol() != Obj->symbol_end()) {
3071 Reloc.getSymbol()->getName(Name);
3076 auto RE = Obj->getRelocation(Reloc.getRawDataRefImpl());
3077 SectionRef RelocSection = Obj->getRelocationSection(RE);
3079 uint64_t SectionAddr = RelocSection.getAddress();
3081 auto Sym = Symbols.upper_bound(Addr);
3082 if (Sym == Symbols.begin()) {
3083 // The first symbol in the object is after this reference, the best we can
3084 // do is section-relative notation.
3085 RelocSection.getName(Name);
3086 Addend = Addr - SectionAddr;
3090 // Go back one so that SymbolAddress <= Addr.
3093 section_iterator SymSection = Obj->section_end();
3094 Sym->second.getSection(SymSection);
3095 if (RelocSection == *SymSection) {
3096 // There's a valid symbol in the same section before this reference.
3097 Sym->second.getName(Name);
3098 Addend = Addr - Sym->first;
3102 // There is a symbol before this reference, but it's in a different
3103 // section. Probably not helpful to mention it, so use the section name.
3104 RelocSection.getName(Name);
3105 Addend = Addr - SectionAddr;
3108 static void printUnwindRelocDest(const MachOObjectFile *Obj,
3109 std::map<uint64_t, SymbolRef> &Symbols,
3110 const RelocationRef &Reloc, uint64_t Addr) {
3114 if (!Reloc.getObjectFile())
3117 findUnwindRelocNameAddend(Obj, Symbols, Reloc, Addr, Name, Addend);
3121 outs() << " + " << format("0x%" PRIx64, Addend);
3125 printMachOCompactUnwindSection(const MachOObjectFile *Obj,
3126 std::map<uint64_t, SymbolRef> &Symbols,
3127 const SectionRef &CompactUnwind) {
3129 assert(Obj->isLittleEndian() &&
3130 "There should not be a big-endian .o with __compact_unwind");
3132 bool Is64 = Obj->is64Bit();
3133 uint32_t PointerSize = Is64 ? sizeof(uint64_t) : sizeof(uint32_t);
3134 uint32_t EntrySize = 3 * PointerSize + 2 * sizeof(uint32_t);
3137 CompactUnwind.getContents(Contents);
3139 SmallVector<CompactUnwindEntry, 4> CompactUnwinds;
3141 // First populate the initial raw offsets, encodings and so on from the entry.
3142 for (unsigned Offset = 0; Offset < Contents.size(); Offset += EntrySize) {
3143 CompactUnwindEntry Entry(Contents.data(), Offset, Is64);
3144 CompactUnwinds.push_back(Entry);
3147 // Next we need to look at the relocations to find out what objects are
3148 // actually being referred to.
3149 for (const RelocationRef &Reloc : CompactUnwind.relocations()) {
3150 uint64_t RelocAddress;
3151 Reloc.getOffset(RelocAddress);
3153 uint32_t EntryIdx = RelocAddress / EntrySize;
3154 uint32_t OffsetInEntry = RelocAddress - EntryIdx * EntrySize;
3155 CompactUnwindEntry &Entry = CompactUnwinds[EntryIdx];
3157 if (OffsetInEntry == 0)
3158 Entry.FunctionReloc = Reloc;
3159 else if (OffsetInEntry == PointerSize + 2 * sizeof(uint32_t))
3160 Entry.PersonalityReloc = Reloc;
3161 else if (OffsetInEntry == 2 * PointerSize + 2 * sizeof(uint32_t))
3162 Entry.LSDAReloc = Reloc;
3164 llvm_unreachable("Unexpected relocation in __compact_unwind section");
3167 // Finally, we're ready to print the data we've gathered.
3168 outs() << "Contents of __compact_unwind section:\n";
3169 for (auto &Entry : CompactUnwinds) {
3170 outs() << " Entry at offset "
3171 << format("0x%" PRIx32, Entry.OffsetInSection) << ":\n";
3173 // 1. Start of the region this entry applies to.
3174 outs() << " start: " << format("0x%" PRIx64,
3175 Entry.FunctionAddr) << ' ';
3176 printUnwindRelocDest(Obj, Symbols, Entry.FunctionReloc, Entry.FunctionAddr);
3179 // 2. Length of the region this entry applies to.
3180 outs() << " length: " << format("0x%" PRIx32, Entry.Length)
3182 // 3. The 32-bit compact encoding.
3183 outs() << " compact encoding: "
3184 << format("0x%08" PRIx32, Entry.CompactEncoding) << '\n';
3186 // 4. The personality function, if present.
3187 if (Entry.PersonalityReloc.getObjectFile()) {
3188 outs() << " personality function: "
3189 << format("0x%" PRIx64, Entry.PersonalityAddr) << ' ';
3190 printUnwindRelocDest(Obj, Symbols, Entry.PersonalityReloc,
3191 Entry.PersonalityAddr);
3195 // 5. This entry's language-specific data area.
3196 if (Entry.LSDAReloc.getObjectFile()) {
3197 outs() << " LSDA: " << format("0x%" PRIx64,
3198 Entry.LSDAAddr) << ' ';
3199 printUnwindRelocDest(Obj, Symbols, Entry.LSDAReloc, Entry.LSDAAddr);
3205 //===----------------------------------------------------------------------===//
3206 // __unwind_info section dumping
3207 //===----------------------------------------------------------------------===//
3209 static void printRegularSecondLevelUnwindPage(const char *PageStart) {
3210 const char *Pos = PageStart;
3211 uint32_t Kind = readNext<uint32_t>(Pos);
3213 assert(Kind == 2 && "kind for a regular 2nd level index should be 2");
3215 uint16_t EntriesStart = readNext<uint16_t>(Pos);
3216 uint16_t NumEntries = readNext<uint16_t>(Pos);
3218 Pos = PageStart + EntriesStart;
3219 for (unsigned i = 0; i < NumEntries; ++i) {
3220 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
3221 uint32_t Encoding = readNext<uint32_t>(Pos);
3223 outs() << " [" << i << "]: "
3224 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
3226 << "encoding=" << format("0x%08" PRIx32, Encoding) << '\n';
3230 static void printCompressedSecondLevelUnwindPage(
3231 const char *PageStart, uint32_t FunctionBase,
3232 const SmallVectorImpl<uint32_t> &CommonEncodings) {
3233 const char *Pos = PageStart;
3234 uint32_t Kind = readNext<uint32_t>(Pos);
3236 assert(Kind == 3 && "kind for a compressed 2nd level index should be 3");
3238 uint16_t EntriesStart = readNext<uint16_t>(Pos);
3239 uint16_t NumEntries = readNext<uint16_t>(Pos);
3241 uint16_t EncodingsStart = readNext<uint16_t>(Pos);
3242 readNext<uint16_t>(Pos);
3243 const auto *PageEncodings = reinterpret_cast<const support::ulittle32_t *>(
3244 PageStart + EncodingsStart);
3246 Pos = PageStart + EntriesStart;
3247 for (unsigned i = 0; i < NumEntries; ++i) {
3248 uint32_t Entry = readNext<uint32_t>(Pos);
3249 uint32_t FunctionOffset = FunctionBase + (Entry & 0xffffff);
3250 uint32_t EncodingIdx = Entry >> 24;
3253 if (EncodingIdx < CommonEncodings.size())
3254 Encoding = CommonEncodings[EncodingIdx];
3256 Encoding = PageEncodings[EncodingIdx - CommonEncodings.size()];
3258 outs() << " [" << i << "]: "
3259 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
3261 << "encoding[" << EncodingIdx
3262 << "]=" << format("0x%08" PRIx32, Encoding) << '\n';
3266 static void printMachOUnwindInfoSection(const MachOObjectFile *Obj,
3267 std::map<uint64_t, SymbolRef> &Symbols,
3268 const SectionRef &UnwindInfo) {
3270 assert(Obj->isLittleEndian() &&
3271 "There should not be a big-endian .o with __unwind_info");
3273 outs() << "Contents of __unwind_info section:\n";
3276 UnwindInfo.getContents(Contents);
3277 const char *Pos = Contents.data();
3279 //===----------------------------------
3281 //===----------------------------------
3283 uint32_t Version = readNext<uint32_t>(Pos);
3284 outs() << " Version: "
3285 << format("0x%" PRIx32, Version) << '\n';
3286 assert(Version == 1 && "only understand version 1");
3288 uint32_t CommonEncodingsStart = readNext<uint32_t>(Pos);
3289 outs() << " Common encodings array section offset: "
3290 << format("0x%" PRIx32, CommonEncodingsStart) << '\n';
3291 uint32_t NumCommonEncodings = readNext<uint32_t>(Pos);
3292 outs() << " Number of common encodings in array: "
3293 << format("0x%" PRIx32, NumCommonEncodings) << '\n';
3295 uint32_t PersonalitiesStart = readNext<uint32_t>(Pos);
3296 outs() << " Personality function array section offset: "
3297 << format("0x%" PRIx32, PersonalitiesStart) << '\n';
3298 uint32_t NumPersonalities = readNext<uint32_t>(Pos);
3299 outs() << " Number of personality functions in array: "
3300 << format("0x%" PRIx32, NumPersonalities) << '\n';
3302 uint32_t IndicesStart = readNext<uint32_t>(Pos);
3303 outs() << " Index array section offset: "
3304 << format("0x%" PRIx32, IndicesStart) << '\n';
3305 uint32_t NumIndices = readNext<uint32_t>(Pos);
3306 outs() << " Number of indices in array: "
3307 << format("0x%" PRIx32, NumIndices) << '\n';
3309 //===----------------------------------
3310 // A shared list of common encodings
3311 //===----------------------------------
3313 // These occupy indices in the range [0, N] whenever an encoding is referenced
3314 // from a compressed 2nd level index table. In practice the linker only
3315 // creates ~128 of these, so that indices are available to embed encodings in
3316 // the 2nd level index.
3318 SmallVector<uint32_t, 64> CommonEncodings;
3319 outs() << " Common encodings: (count = " << NumCommonEncodings << ")\n";
3320 Pos = Contents.data() + CommonEncodingsStart;
3321 for (unsigned i = 0; i < NumCommonEncodings; ++i) {
3322 uint32_t Encoding = readNext<uint32_t>(Pos);
3323 CommonEncodings.push_back(Encoding);
3325 outs() << " encoding[" << i << "]: " << format("0x%08" PRIx32, Encoding)
3329 //===----------------------------------
3330 // Personality functions used in this executable
3331 //===----------------------------------
3333 // There should be only a handful of these (one per source language,
3334 // roughly). Particularly since they only get 2 bits in the compact encoding.
3336 outs() << " Personality functions: (count = " << NumPersonalities << ")\n";
3337 Pos = Contents.data() + PersonalitiesStart;
3338 for (unsigned i = 0; i < NumPersonalities; ++i) {
3339 uint32_t PersonalityFn = readNext<uint32_t>(Pos);
3340 outs() << " personality[" << i + 1
3341 << "]: " << format("0x%08" PRIx32, PersonalityFn) << '\n';
3344 //===----------------------------------
3345 // The level 1 index entries
3346 //===----------------------------------
3348 // These specify an approximate place to start searching for the more detailed
3349 // information, sorted by PC.
3352 uint32_t FunctionOffset;
3353 uint32_t SecondLevelPageStart;
3357 SmallVector<IndexEntry, 4> IndexEntries;
3359 outs() << " Top level indices: (count = " << NumIndices << ")\n";
3360 Pos = Contents.data() + IndicesStart;
3361 for (unsigned i = 0; i < NumIndices; ++i) {
3364 Entry.FunctionOffset = readNext<uint32_t>(Pos);
3365 Entry.SecondLevelPageStart = readNext<uint32_t>(Pos);
3366 Entry.LSDAStart = readNext<uint32_t>(Pos);
3367 IndexEntries.push_back(Entry);
3369 outs() << " [" << i << "]: "
3370 << "function offset=" << format("0x%08" PRIx32, Entry.FunctionOffset)
3372 << "2nd level page offset="
3373 << format("0x%08" PRIx32, Entry.SecondLevelPageStart) << ", "
3374 << "LSDA offset=" << format("0x%08" PRIx32, Entry.LSDAStart) << '\n';
3377 //===----------------------------------
3378 // Next come the LSDA tables
3379 //===----------------------------------
3381 // The LSDA layout is rather implicit: it's a contiguous array of entries from
3382 // the first top-level index's LSDAOffset to the last (sentinel).
3384 outs() << " LSDA descriptors:\n";
3385 Pos = Contents.data() + IndexEntries[0].LSDAStart;
3386 int NumLSDAs = (IndexEntries.back().LSDAStart - IndexEntries[0].LSDAStart) /
3387 (2 * sizeof(uint32_t));
3388 for (int i = 0; i < NumLSDAs; ++i) {
3389 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
3390 uint32_t LSDAOffset = readNext<uint32_t>(Pos);
3391 outs() << " [" << i << "]: "
3392 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
3394 << "LSDA offset=" << format("0x%08" PRIx32, LSDAOffset) << '\n';
3397 //===----------------------------------
3398 // Finally, the 2nd level indices
3399 //===----------------------------------
3401 // Generally these are 4K in size, and have 2 possible forms:
3402 // + Regular stores up to 511 entries with disparate encodings
3403 // + Compressed stores up to 1021 entries if few enough compact encoding
3405 outs() << " Second level indices:\n";
3406 for (unsigned i = 0; i < IndexEntries.size() - 1; ++i) {
3407 // The final sentinel top-level index has no associated 2nd level page
3408 if (IndexEntries[i].SecondLevelPageStart == 0)
3411 outs() << " Second level index[" << i << "]: "
3412 << "offset in section="
3413 << format("0x%08" PRIx32, IndexEntries[i].SecondLevelPageStart)
3415 << "base function offset="
3416 << format("0x%08" PRIx32, IndexEntries[i].FunctionOffset) << '\n';
3418 Pos = Contents.data() + IndexEntries[i].SecondLevelPageStart;
3419 uint32_t Kind = *reinterpret_cast<const support::ulittle32_t *>(Pos);
3421 printRegularSecondLevelUnwindPage(Pos);
3423 printCompressedSecondLevelUnwindPage(Pos, IndexEntries[i].FunctionOffset,
3426 llvm_unreachable("Do not know how to print this kind of 2nd level page");
3430 void llvm::printMachOUnwindInfo(const MachOObjectFile *Obj) {
3431 std::map<uint64_t, SymbolRef> Symbols;
3432 for (const SymbolRef &SymRef : Obj->symbols()) {
3433 // Discard any undefined or absolute symbols. They're not going to take part
3434 // in the convenience lookup for unwind info and just take up resources.
3435 section_iterator Section = Obj->section_end();
3436 SymRef.getSection(Section);
3437 if (Section == Obj->section_end())
3441 SymRef.getAddress(Addr);
3442 Symbols.insert(std::make_pair(Addr, SymRef));
3445 for (const SectionRef &Section : Obj->sections()) {
3447 Section.getName(SectName);
3448 if (SectName == "__compact_unwind")
3449 printMachOCompactUnwindSection(Obj, Symbols, Section);
3450 else if (SectName == "__unwind_info")
3451 printMachOUnwindInfoSection(Obj, Symbols, Section);
3452 else if (SectName == "__eh_frame")
3453 outs() << "llvm-objdump: warning: unhandled __eh_frame section\n";
3457 static void PrintMachHeader(uint32_t magic, uint32_t cputype,
3458 uint32_t cpusubtype, uint32_t filetype,
3459 uint32_t ncmds, uint32_t sizeofcmds, uint32_t flags,
3461 outs() << "Mach header\n";
3462 outs() << " magic cputype cpusubtype caps filetype ncmds "
3463 "sizeofcmds flags\n";
3465 if (magic == MachO::MH_MAGIC)
3466 outs() << " MH_MAGIC";
3467 else if (magic == MachO::MH_MAGIC_64)
3468 outs() << "MH_MAGIC_64";
3470 outs() << format(" 0x%08" PRIx32, magic);
3472 case MachO::CPU_TYPE_I386:
3474 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3475 case MachO::CPU_SUBTYPE_I386_ALL:
3479 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3483 case MachO::CPU_TYPE_X86_64:
3484 outs() << " X86_64";
3485 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3486 case MachO::CPU_SUBTYPE_X86_64_ALL:
3489 case MachO::CPU_SUBTYPE_X86_64_H:
3490 outs() << " Haswell";
3493 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3497 case MachO::CPU_TYPE_ARM:
3499 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3500 case MachO::CPU_SUBTYPE_ARM_ALL:
3503 case MachO::CPU_SUBTYPE_ARM_V4T:
3506 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
3509 case MachO::CPU_SUBTYPE_ARM_XSCALE:
3510 outs() << " XSCALE";
3512 case MachO::CPU_SUBTYPE_ARM_V6:
3515 case MachO::CPU_SUBTYPE_ARM_V6M:
3518 case MachO::CPU_SUBTYPE_ARM_V7:
3521 case MachO::CPU_SUBTYPE_ARM_V7EM:
3524 case MachO::CPU_SUBTYPE_ARM_V7K:
3527 case MachO::CPU_SUBTYPE_ARM_V7M:
3530 case MachO::CPU_SUBTYPE_ARM_V7S:
3534 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3538 case MachO::CPU_TYPE_ARM64:
3540 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3541 case MachO::CPU_SUBTYPE_ARM64_ALL:
3545 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3549 case MachO::CPU_TYPE_POWERPC:
3551 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3552 case MachO::CPU_SUBTYPE_POWERPC_ALL:
3556 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3560 case MachO::CPU_TYPE_POWERPC64:
3562 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3563 case MachO::CPU_SUBTYPE_POWERPC_ALL:
3567 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3572 if ((cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64) {
3575 outs() << format(" 0x%02" PRIx32,
3576 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
3579 case MachO::MH_OBJECT:
3580 outs() << " OBJECT";
3582 case MachO::MH_EXECUTE:
3583 outs() << " EXECUTE";
3585 case MachO::MH_FVMLIB:
3586 outs() << " FVMLIB";
3588 case MachO::MH_CORE:
3591 case MachO::MH_PRELOAD:
3592 outs() << " PRELOAD";
3594 case MachO::MH_DYLIB:
3597 case MachO::MH_DYLIB_STUB:
3598 outs() << " DYLIB_STUB";
3600 case MachO::MH_DYLINKER:
3601 outs() << " DYLINKER";
3603 case MachO::MH_BUNDLE:
3604 outs() << " BUNDLE";
3606 case MachO::MH_DSYM:
3609 case MachO::MH_KEXT_BUNDLE:
3610 outs() << " KEXTBUNDLE";
3613 outs() << format(" %10u", filetype);
3616 outs() << format(" %5u", ncmds);
3617 outs() << format(" %10u", sizeofcmds);
3619 if (f & MachO::MH_NOUNDEFS) {
3620 outs() << " NOUNDEFS";
3621 f &= ~MachO::MH_NOUNDEFS;
3623 if (f & MachO::MH_INCRLINK) {
3624 outs() << " INCRLINK";
3625 f &= ~MachO::MH_INCRLINK;
3627 if (f & MachO::MH_DYLDLINK) {
3628 outs() << " DYLDLINK";
3629 f &= ~MachO::MH_DYLDLINK;
3631 if (f & MachO::MH_BINDATLOAD) {
3632 outs() << " BINDATLOAD";
3633 f &= ~MachO::MH_BINDATLOAD;
3635 if (f & MachO::MH_PREBOUND) {
3636 outs() << " PREBOUND";
3637 f &= ~MachO::MH_PREBOUND;
3639 if (f & MachO::MH_SPLIT_SEGS) {
3640 outs() << " SPLIT_SEGS";
3641 f &= ~MachO::MH_SPLIT_SEGS;
3643 if (f & MachO::MH_LAZY_INIT) {
3644 outs() << " LAZY_INIT";
3645 f &= ~MachO::MH_LAZY_INIT;
3647 if (f & MachO::MH_TWOLEVEL) {
3648 outs() << " TWOLEVEL";
3649 f &= ~MachO::MH_TWOLEVEL;
3651 if (f & MachO::MH_FORCE_FLAT) {
3652 outs() << " FORCE_FLAT";
3653 f &= ~MachO::MH_FORCE_FLAT;
3655 if (f & MachO::MH_NOMULTIDEFS) {
3656 outs() << " NOMULTIDEFS";
3657 f &= ~MachO::MH_NOMULTIDEFS;
3659 if (f & MachO::MH_NOFIXPREBINDING) {
3660 outs() << " NOFIXPREBINDING";
3661 f &= ~MachO::MH_NOFIXPREBINDING;
3663 if (f & MachO::MH_PREBINDABLE) {
3664 outs() << " PREBINDABLE";
3665 f &= ~MachO::MH_PREBINDABLE;
3667 if (f & MachO::MH_ALLMODSBOUND) {
3668 outs() << " ALLMODSBOUND";
3669 f &= ~MachO::MH_ALLMODSBOUND;
3671 if (f & MachO::MH_SUBSECTIONS_VIA_SYMBOLS) {
3672 outs() << " SUBSECTIONS_VIA_SYMBOLS";
3673 f &= ~MachO::MH_SUBSECTIONS_VIA_SYMBOLS;
3675 if (f & MachO::MH_CANONICAL) {
3676 outs() << " CANONICAL";
3677 f &= ~MachO::MH_CANONICAL;
3679 if (f & MachO::MH_WEAK_DEFINES) {
3680 outs() << " WEAK_DEFINES";
3681 f &= ~MachO::MH_WEAK_DEFINES;
3683 if (f & MachO::MH_BINDS_TO_WEAK) {
3684 outs() << " BINDS_TO_WEAK";
3685 f &= ~MachO::MH_BINDS_TO_WEAK;
3687 if (f & MachO::MH_ALLOW_STACK_EXECUTION) {
3688 outs() << " ALLOW_STACK_EXECUTION";
3689 f &= ~MachO::MH_ALLOW_STACK_EXECUTION;
3691 if (f & MachO::MH_DEAD_STRIPPABLE_DYLIB) {
3692 outs() << " DEAD_STRIPPABLE_DYLIB";
3693 f &= ~MachO::MH_DEAD_STRIPPABLE_DYLIB;
3695 if (f & MachO::MH_PIE) {
3697 f &= ~MachO::MH_PIE;
3699 if (f & MachO::MH_NO_REEXPORTED_DYLIBS) {
3700 outs() << " NO_REEXPORTED_DYLIBS";
3701 f &= ~MachO::MH_NO_REEXPORTED_DYLIBS;
3703 if (f & MachO::MH_HAS_TLV_DESCRIPTORS) {
3704 outs() << " MH_HAS_TLV_DESCRIPTORS";
3705 f &= ~MachO::MH_HAS_TLV_DESCRIPTORS;
3707 if (f & MachO::MH_NO_HEAP_EXECUTION) {
3708 outs() << " MH_NO_HEAP_EXECUTION";
3709 f &= ~MachO::MH_NO_HEAP_EXECUTION;
3711 if (f & MachO::MH_APP_EXTENSION_SAFE) {
3712 outs() << " APP_EXTENSION_SAFE";
3713 f &= ~MachO::MH_APP_EXTENSION_SAFE;
3715 if (f != 0 || flags == 0)
3716 outs() << format(" 0x%08" PRIx32, f);
3718 outs() << format(" 0x%08" PRIx32, magic);
3719 outs() << format(" %7d", cputype);
3720 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3721 outs() << format(" 0x%02" PRIx32,
3722 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
3723 outs() << format(" %10u", filetype);
3724 outs() << format(" %5u", ncmds);
3725 outs() << format(" %10u", sizeofcmds);
3726 outs() << format(" 0x%08" PRIx32, flags);
3731 static void PrintSegmentCommand(uint32_t cmd, uint32_t cmdsize,
3732 StringRef SegName, uint64_t vmaddr,
3733 uint64_t vmsize, uint64_t fileoff,
3734 uint64_t filesize, uint32_t maxprot,
3735 uint32_t initprot, uint32_t nsects,
3736 uint32_t flags, uint32_t object_size,
3738 uint64_t expected_cmdsize;
3739 if (cmd == MachO::LC_SEGMENT) {
3740 outs() << " cmd LC_SEGMENT\n";
3741 expected_cmdsize = nsects;
3742 expected_cmdsize *= sizeof(struct MachO::section);
3743 expected_cmdsize += sizeof(struct MachO::segment_command);
3745 outs() << " cmd LC_SEGMENT_64\n";
3746 expected_cmdsize = nsects;
3747 expected_cmdsize *= sizeof(struct MachO::section_64);
3748 expected_cmdsize += sizeof(struct MachO::segment_command_64);
3750 outs() << " cmdsize " << cmdsize;
3751 if (cmdsize != expected_cmdsize)
3752 outs() << " Inconsistent size\n";
3755 outs() << " segname " << SegName << "\n";
3756 if (cmd == MachO::LC_SEGMENT_64) {
3757 outs() << " vmaddr " << format("0x%016" PRIx64, vmaddr) << "\n";
3758 outs() << " vmsize " << format("0x%016" PRIx64, vmsize) << "\n";
3760 outs() << " vmaddr " << format("0x%08" PRIx64, vmaddr) << "\n";
3761 outs() << " vmsize " << format("0x%08" PRIx64, vmsize) << "\n";
3763 outs() << " fileoff " << fileoff;
3764 if (fileoff > object_size)
3765 outs() << " (past end of file)\n";
3768 outs() << " filesize " << filesize;
3769 if (fileoff + filesize > object_size)
3770 outs() << " (past end of file)\n";
3775 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
3776 MachO::VM_PROT_EXECUTE)) != 0)
3777 outs() << " maxprot ?" << format("0x%08" PRIx32, maxprot) << "\n";
3779 if (maxprot & MachO::VM_PROT_READ)
3780 outs() << " maxprot r";
3782 outs() << " maxprot -";
3783 if (maxprot & MachO::VM_PROT_WRITE)
3787 if (maxprot & MachO::VM_PROT_EXECUTE)
3793 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
3794 MachO::VM_PROT_EXECUTE)) != 0)
3795 outs() << " initprot ?" << format("0x%08" PRIx32, initprot) << "\n";
3797 if (initprot & MachO::VM_PROT_READ)
3798 outs() << " initprot r";
3800 outs() << " initprot -";
3801 if (initprot & MachO::VM_PROT_WRITE)
3805 if (initprot & MachO::VM_PROT_EXECUTE)
3811 outs() << " maxprot " << format("0x%08" PRIx32, maxprot) << "\n";
3812 outs() << " initprot " << format("0x%08" PRIx32, initprot) << "\n";
3814 outs() << " nsects " << nsects << "\n";
3818 outs() << " (none)\n";
3820 if (flags & MachO::SG_HIGHVM) {
3821 outs() << " HIGHVM";
3822 flags &= ~MachO::SG_HIGHVM;
3824 if (flags & MachO::SG_FVMLIB) {
3825 outs() << " FVMLIB";
3826 flags &= ~MachO::SG_FVMLIB;
3828 if (flags & MachO::SG_NORELOC) {
3829 outs() << " NORELOC";
3830 flags &= ~MachO::SG_NORELOC;
3832 if (flags & MachO::SG_PROTECTED_VERSION_1) {
3833 outs() << " PROTECTED_VERSION_1";
3834 flags &= ~MachO::SG_PROTECTED_VERSION_1;
3837 outs() << format(" 0x%08" PRIx32, flags) << " (unknown flags)\n";
3842 outs() << " flags " << format("0x%" PRIx32, flags) << "\n";
3846 static void PrintSection(const char *sectname, const char *segname,
3847 uint64_t addr, uint64_t size, uint32_t offset,
3848 uint32_t align, uint32_t reloff, uint32_t nreloc,
3849 uint32_t flags, uint32_t reserved1, uint32_t reserved2,
3850 uint32_t cmd, const char *sg_segname,
3851 uint32_t filetype, uint32_t object_size,
3853 outs() << "Section\n";
3854 outs() << " sectname " << format("%.16s\n", sectname);
3855 outs() << " segname " << format("%.16s", segname);
3856 if (filetype != MachO::MH_OBJECT && strncmp(sg_segname, segname, 16) != 0)
3857 outs() << " (does not match segment)\n";
3860 if (cmd == MachO::LC_SEGMENT_64) {
3861 outs() << " addr " << format("0x%016" PRIx64, addr) << "\n";
3862 outs() << " size " << format("0x%016" PRIx64, size);
3864 outs() << " addr " << format("0x%08" PRIx64, addr) << "\n";
3865 outs() << " size " << format("0x%08" PRIx64, size);
3867 if ((flags & MachO::S_ZEROFILL) != 0 && offset + size > object_size)
3868 outs() << " (past end of file)\n";
3871 outs() << " offset " << offset;
3872 if (offset > object_size)
3873 outs() << " (past end of file)\n";
3876 uint32_t align_shifted = 1 << align;
3877 outs() << " align 2^" << align << " (" << align_shifted << ")\n";
3878 outs() << " reloff " << reloff;
3879 if (reloff > object_size)
3880 outs() << " (past end of file)\n";
3883 outs() << " nreloc " << nreloc;
3884 if (reloff + nreloc * sizeof(struct MachO::relocation_info) > object_size)
3885 outs() << " (past end of file)\n";
3888 uint32_t section_type = flags & MachO::SECTION_TYPE;
3891 if (section_type == MachO::S_REGULAR)
3892 outs() << " S_REGULAR\n";
3893 else if (section_type == MachO::S_ZEROFILL)
3894 outs() << " S_ZEROFILL\n";
3895 else if (section_type == MachO::S_CSTRING_LITERALS)
3896 outs() << " S_CSTRING_LITERALS\n";
3897 else if (section_type == MachO::S_4BYTE_LITERALS)
3898 outs() << " S_4BYTE_LITERALS\n";
3899 else if (section_type == MachO::S_8BYTE_LITERALS)
3900 outs() << " S_8BYTE_LITERALS\n";
3901 else if (section_type == MachO::S_16BYTE_LITERALS)
3902 outs() << " S_16BYTE_LITERALS\n";
3903 else if (section_type == MachO::S_LITERAL_POINTERS)
3904 outs() << " S_LITERAL_POINTERS\n";
3905 else if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS)
3906 outs() << " S_NON_LAZY_SYMBOL_POINTERS\n";
3907 else if (section_type == MachO::S_LAZY_SYMBOL_POINTERS)
3908 outs() << " S_LAZY_SYMBOL_POINTERS\n";
3909 else if (section_type == MachO::S_SYMBOL_STUBS)
3910 outs() << " S_SYMBOL_STUBS\n";
3911 else if (section_type == MachO::S_MOD_INIT_FUNC_POINTERS)
3912 outs() << " S_MOD_INIT_FUNC_POINTERS\n";
3913 else if (section_type == MachO::S_MOD_TERM_FUNC_POINTERS)
3914 outs() << " S_MOD_TERM_FUNC_POINTERS\n";
3915 else if (section_type == MachO::S_COALESCED)
3916 outs() << " S_COALESCED\n";
3917 else if (section_type == MachO::S_INTERPOSING)
3918 outs() << " S_INTERPOSING\n";
3919 else if (section_type == MachO::S_DTRACE_DOF)
3920 outs() << " S_DTRACE_DOF\n";
3921 else if (section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS)
3922 outs() << " S_LAZY_DYLIB_SYMBOL_POINTERS\n";
3923 else if (section_type == MachO::S_THREAD_LOCAL_REGULAR)
3924 outs() << " S_THREAD_LOCAL_REGULAR\n";
3925 else if (section_type == MachO::S_THREAD_LOCAL_ZEROFILL)
3926 outs() << " S_THREAD_LOCAL_ZEROFILL\n";
3927 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLES)
3928 outs() << " S_THREAD_LOCAL_VARIABLES\n";
3929 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
3930 outs() << " S_THREAD_LOCAL_VARIABLE_POINTERS\n";
3931 else if (section_type == MachO::S_THREAD_LOCAL_INIT_FUNCTION_POINTERS)
3932 outs() << " S_THREAD_LOCAL_INIT_FUNCTION_POINTERS\n";
3934 outs() << format("0x%08" PRIx32, section_type) << "\n";
3935 outs() << "attributes";
3936 uint32_t section_attributes = flags & MachO::SECTION_ATTRIBUTES;
3937 if (section_attributes & MachO::S_ATTR_PURE_INSTRUCTIONS)
3938 outs() << " PURE_INSTRUCTIONS";
3939 if (section_attributes & MachO::S_ATTR_NO_TOC)
3940 outs() << " NO_TOC";
3941 if (section_attributes & MachO::S_ATTR_STRIP_STATIC_SYMS)
3942 outs() << " STRIP_STATIC_SYMS";
3943 if (section_attributes & MachO::S_ATTR_NO_DEAD_STRIP)
3944 outs() << " NO_DEAD_STRIP";
3945 if (section_attributes & MachO::S_ATTR_LIVE_SUPPORT)
3946 outs() << " LIVE_SUPPORT";
3947 if (section_attributes & MachO::S_ATTR_SELF_MODIFYING_CODE)
3948 outs() << " SELF_MODIFYING_CODE";
3949 if (section_attributes & MachO::S_ATTR_DEBUG)
3951 if (section_attributes & MachO::S_ATTR_SOME_INSTRUCTIONS)
3952 outs() << " SOME_INSTRUCTIONS";
3953 if (section_attributes & MachO::S_ATTR_EXT_RELOC)
3954 outs() << " EXT_RELOC";
3955 if (section_attributes & MachO::S_ATTR_LOC_RELOC)
3956 outs() << " LOC_RELOC";
3957 if (section_attributes == 0)
3958 outs() << " (none)";
3961 outs() << " flags " << format("0x%08" PRIx32, flags) << "\n";
3962 outs() << " reserved1 " << reserved1;
3963 if (section_type == MachO::S_SYMBOL_STUBS ||
3964 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
3965 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
3966 section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
3967 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
3968 outs() << " (index into indirect symbol table)\n";
3971 outs() << " reserved2 " << reserved2;
3972 if (section_type == MachO::S_SYMBOL_STUBS)
3973 outs() << " (size of stubs)\n";
3978 static void PrintSymtabLoadCommand(MachO::symtab_command st, bool Is64Bit,
3979 uint32_t object_size) {
3980 outs() << " cmd LC_SYMTAB\n";
3981 outs() << " cmdsize " << st.cmdsize;
3982 if (st.cmdsize != sizeof(struct MachO::symtab_command))
3983 outs() << " Incorrect size\n";
3986 outs() << " symoff " << st.symoff;
3987 if (st.symoff > object_size)
3988 outs() << " (past end of file)\n";
3991 outs() << " nsyms " << st.nsyms;
3994 big_size = st.nsyms;
3995 big_size *= sizeof(struct MachO::nlist_64);
3996 big_size += st.symoff;
3997 if (big_size > object_size)
3998 outs() << " (past end of file)\n";
4002 big_size = st.nsyms;
4003 big_size *= sizeof(struct MachO::nlist);
4004 big_size += st.symoff;
4005 if (big_size > object_size)
4006 outs() << " (past end of file)\n";
4010 outs() << " stroff " << st.stroff;
4011 if (st.stroff > object_size)
4012 outs() << " (past end of file)\n";
4015 outs() << " strsize " << st.strsize;
4016 big_size = st.stroff;
4017 big_size += st.strsize;
4018 if (big_size > object_size)
4019 outs() << " (past end of file)\n";
4024 static void PrintDysymtabLoadCommand(MachO::dysymtab_command dyst,
4025 uint32_t nsyms, uint32_t object_size,
4027 outs() << " cmd LC_DYSYMTAB\n";
4028 outs() << " cmdsize " << dyst.cmdsize;
4029 if (dyst.cmdsize != sizeof(struct MachO::dysymtab_command))
4030 outs() << " Incorrect size\n";
4033 outs() << " ilocalsym " << dyst.ilocalsym;
4034 if (dyst.ilocalsym > nsyms)
4035 outs() << " (greater than the number of symbols)\n";
4038 outs() << " nlocalsym " << dyst.nlocalsym;
4040 big_size = dyst.ilocalsym;
4041 big_size += dyst.nlocalsym;
4042 if (big_size > nsyms)
4043 outs() << " (past the end of the symbol table)\n";
4046 outs() << " iextdefsym " << dyst.iextdefsym;
4047 if (dyst.iextdefsym > nsyms)
4048 outs() << " (greater than the number of symbols)\n";
4051 outs() << " nextdefsym " << dyst.nextdefsym;
4052 big_size = dyst.iextdefsym;
4053 big_size += dyst.nextdefsym;
4054 if (big_size > nsyms)
4055 outs() << " (past the end of the symbol table)\n";
4058 outs() << " iundefsym " << dyst.iundefsym;
4059 if (dyst.iundefsym > nsyms)
4060 outs() << " (greater than the number of symbols)\n";
4063 outs() << " nundefsym " << dyst.nundefsym;
4064 big_size = dyst.iundefsym;
4065 big_size += dyst.nundefsym;
4066 if (big_size > nsyms)
4067 outs() << " (past the end of the symbol table)\n";
4070 outs() << " tocoff " << dyst.tocoff;
4071 if (dyst.tocoff > object_size)
4072 outs() << " (past end of file)\n";
4075 outs() << " ntoc " << dyst.ntoc;
4076 big_size = dyst.ntoc;
4077 big_size *= sizeof(struct MachO::dylib_table_of_contents);
4078 big_size += dyst.tocoff;
4079 if (big_size > object_size)
4080 outs() << " (past end of file)\n";
4083 outs() << " modtaboff " << dyst.modtaboff;
4084 if (dyst.modtaboff > object_size)
4085 outs() << " (past end of file)\n";
4088 outs() << " nmodtab " << dyst.nmodtab;
4091 modtabend = dyst.nmodtab;
4092 modtabend *= sizeof(struct MachO::dylib_module_64);
4093 modtabend += dyst.modtaboff;
4095 modtabend = dyst.nmodtab;
4096 modtabend *= sizeof(struct MachO::dylib_module);
4097 modtabend += dyst.modtaboff;
4099 if (modtabend > object_size)
4100 outs() << " (past end of file)\n";
4103 outs() << " extrefsymoff " << dyst.extrefsymoff;
4104 if (dyst.extrefsymoff > object_size)
4105 outs() << " (past end of file)\n";
4108 outs() << " nextrefsyms " << dyst.nextrefsyms;
4109 big_size = dyst.nextrefsyms;
4110 big_size *= sizeof(struct MachO::dylib_reference);
4111 big_size += dyst.extrefsymoff;
4112 if (big_size > object_size)
4113 outs() << " (past end of file)\n";
4116 outs() << " indirectsymoff " << dyst.indirectsymoff;
4117 if (dyst.indirectsymoff > object_size)
4118 outs() << " (past end of file)\n";
4121 outs() << " nindirectsyms " << dyst.nindirectsyms;
4122 big_size = dyst.nindirectsyms;
4123 big_size *= sizeof(uint32_t);
4124 big_size += dyst.indirectsymoff;
4125 if (big_size > object_size)
4126 outs() << " (past end of file)\n";
4129 outs() << " extreloff " << dyst.extreloff;
4130 if (dyst.extreloff > object_size)
4131 outs() << " (past end of file)\n";
4134 outs() << " nextrel " << dyst.nextrel;
4135 big_size = dyst.nextrel;
4136 big_size *= sizeof(struct MachO::relocation_info);
4137 big_size += dyst.extreloff;
4138 if (big_size > object_size)
4139 outs() << " (past end of file)\n";
4142 outs() << " locreloff " << dyst.locreloff;
4143 if (dyst.locreloff > object_size)
4144 outs() << " (past end of file)\n";
4147 outs() << " nlocrel " << dyst.nlocrel;
4148 big_size = dyst.nlocrel;
4149 big_size *= sizeof(struct MachO::relocation_info);
4150 big_size += dyst.locreloff;
4151 if (big_size > object_size)
4152 outs() << " (past end of file)\n";
4157 static void PrintDyldInfoLoadCommand(MachO::dyld_info_command dc,
4158 uint32_t object_size) {
4159 if (dc.cmd == MachO::LC_DYLD_INFO)
4160 outs() << " cmd LC_DYLD_INFO\n";
4162 outs() << " cmd LC_DYLD_INFO_ONLY\n";
4163 outs() << " cmdsize " << dc.cmdsize;
4164 if (dc.cmdsize != sizeof(struct MachO::dyld_info_command))
4165 outs() << " Incorrect size\n";
4168 outs() << " rebase_off " << dc.rebase_off;
4169 if (dc.rebase_off > object_size)
4170 outs() << " (past end of file)\n";
4173 outs() << " rebase_size " << dc.rebase_size;
4175 big_size = dc.rebase_off;
4176 big_size += dc.rebase_size;
4177 if (big_size > object_size)
4178 outs() << " (past end of file)\n";
4181 outs() << " bind_off " << dc.bind_off;
4182 if (dc.bind_off > object_size)
4183 outs() << " (past end of file)\n";
4186 outs() << " bind_size " << dc.bind_size;
4187 big_size = dc.bind_off;
4188 big_size += dc.bind_size;
4189 if (big_size > object_size)
4190 outs() << " (past end of file)\n";
4193 outs() << " weak_bind_off " << dc.weak_bind_off;
4194 if (dc.weak_bind_off > object_size)
4195 outs() << " (past end of file)\n";
4198 outs() << " weak_bind_size " << dc.weak_bind_size;
4199 big_size = dc.weak_bind_off;
4200 big_size += dc.weak_bind_size;
4201 if (big_size > object_size)
4202 outs() << " (past end of file)\n";
4205 outs() << " lazy_bind_off " << dc.lazy_bind_off;
4206 if (dc.lazy_bind_off > object_size)
4207 outs() << " (past end of file)\n";
4210 outs() << " lazy_bind_size " << dc.lazy_bind_size;
4211 big_size = dc.lazy_bind_off;
4212 big_size += dc.lazy_bind_size;
4213 if (big_size > object_size)
4214 outs() << " (past end of file)\n";
4217 outs() << " export_off " << dc.export_off;
4218 if (dc.export_off > object_size)
4219 outs() << " (past end of file)\n";
4222 outs() << " export_size " << dc.export_size;
4223 big_size = dc.export_off;
4224 big_size += dc.export_size;
4225 if (big_size > object_size)
4226 outs() << " (past end of file)\n";
4231 static void PrintDyldLoadCommand(MachO::dylinker_command dyld,
4233 if (dyld.cmd == MachO::LC_ID_DYLINKER)
4234 outs() << " cmd LC_ID_DYLINKER\n";
4235 else if (dyld.cmd == MachO::LC_LOAD_DYLINKER)
4236 outs() << " cmd LC_LOAD_DYLINKER\n";
4237 else if (dyld.cmd == MachO::LC_DYLD_ENVIRONMENT)
4238 outs() << " cmd LC_DYLD_ENVIRONMENT\n";
4240 outs() << " cmd ?(" << dyld.cmd << ")\n";
4241 outs() << " cmdsize " << dyld.cmdsize;
4242 if (dyld.cmdsize < sizeof(struct MachO::dylinker_command))
4243 outs() << " Incorrect size\n";
4246 if (dyld.name >= dyld.cmdsize)
4247 outs() << " name ?(bad offset " << dyld.name << ")\n";
4249 const char *P = (const char *)(Ptr) + dyld.name;
4250 outs() << " name " << P << " (offset " << dyld.name << ")\n";
4254 static void PrintUuidLoadCommand(MachO::uuid_command uuid) {
4255 outs() << " cmd LC_UUID\n";
4256 outs() << " cmdsize " << uuid.cmdsize;
4257 if (uuid.cmdsize != sizeof(struct MachO::uuid_command))
4258 outs() << " Incorrect size\n";
4262 outs() << format("%02" PRIX32, uuid.uuid[0]);
4263 outs() << format("%02" PRIX32, uuid.uuid[1]);
4264 outs() << format("%02" PRIX32, uuid.uuid[2]);
4265 outs() << format("%02" PRIX32, uuid.uuid[3]);
4267 outs() << format("%02" PRIX32, uuid.uuid[4]);
4268 outs() << format("%02" PRIX32, uuid.uuid[5]);
4270 outs() << format("%02" PRIX32, uuid.uuid[6]);
4271 outs() << format("%02" PRIX32, uuid.uuid[7]);
4273 outs() << format("%02" PRIX32, uuid.uuid[8]);
4274 outs() << format("%02" PRIX32, uuid.uuid[9]);
4276 outs() << format("%02" PRIX32, uuid.uuid[10]);
4277 outs() << format("%02" PRIX32, uuid.uuid[11]);
4278 outs() << format("%02" PRIX32, uuid.uuid[12]);
4279 outs() << format("%02" PRIX32, uuid.uuid[13]);
4280 outs() << format("%02" PRIX32, uuid.uuid[14]);
4281 outs() << format("%02" PRIX32, uuid.uuid[15]);
4285 static void PrintRpathLoadCommand(MachO::rpath_command rpath, const char *Ptr) {
4286 outs() << " cmd LC_RPATH\n";
4287 outs() << " cmdsize " << rpath.cmdsize;
4288 if (rpath.cmdsize < sizeof(struct MachO::rpath_command))
4289 outs() << " Incorrect size\n";
4292 if (rpath.path >= rpath.cmdsize)
4293 outs() << " path ?(bad offset " << rpath.path << ")\n";
4295 const char *P = (const char *)(Ptr) + rpath.path;
4296 outs() << " path " << P << " (offset " << rpath.path << ")\n";
4300 static void PrintVersionMinLoadCommand(MachO::version_min_command vd) {
4301 if (vd.cmd == MachO::LC_VERSION_MIN_MACOSX)
4302 outs() << " cmd LC_VERSION_MIN_MACOSX\n";
4303 else if (vd.cmd == MachO::LC_VERSION_MIN_IPHONEOS)
4304 outs() << " cmd LC_VERSION_MIN_IPHONEOS\n";
4306 outs() << " cmd " << vd.cmd << " (?)\n";
4307 outs() << " cmdsize " << vd.cmdsize;
4308 if (vd.cmdsize != sizeof(struct MachO::version_min_command))
4309 outs() << " Incorrect size\n";
4312 outs() << " version " << ((vd.version >> 16) & 0xffff) << "."
4313 << ((vd.version >> 8) & 0xff);
4314 if ((vd.version & 0xff) != 0)
4315 outs() << "." << (vd.version & 0xff);
4318 outs() << " sdk n/a";
4320 outs() << " sdk " << ((vd.sdk >> 16) & 0xffff) << "."
4321 << ((vd.sdk >> 8) & 0xff);
4323 if ((vd.sdk & 0xff) != 0)
4324 outs() << "." << (vd.sdk & 0xff);
4328 static void PrintSourceVersionCommand(MachO::source_version_command sd) {
4329 outs() << " cmd LC_SOURCE_VERSION\n";
4330 outs() << " cmdsize " << sd.cmdsize;
4331 if (sd.cmdsize != sizeof(struct MachO::source_version_command))
4332 outs() << " Incorrect size\n";
4335 uint64_t a = (sd.version >> 40) & 0xffffff;
4336 uint64_t b = (sd.version >> 30) & 0x3ff;
4337 uint64_t c = (sd.version >> 20) & 0x3ff;
4338 uint64_t d = (sd.version >> 10) & 0x3ff;
4339 uint64_t e = sd.version & 0x3ff;
4340 outs() << " version " << a << "." << b;
4342 outs() << "." << c << "." << d << "." << e;
4344 outs() << "." << c << "." << d;
4350 static void PrintEntryPointCommand(MachO::entry_point_command ep) {
4351 outs() << " cmd LC_MAIN\n";
4352 outs() << " cmdsize " << ep.cmdsize;
4353 if (ep.cmdsize != sizeof(struct MachO::entry_point_command))
4354 outs() << " Incorrect size\n";
4357 outs() << " entryoff " << ep.entryoff << "\n";
4358 outs() << " stacksize " << ep.stacksize << "\n";
4361 static void PrintEncryptionInfoCommand(MachO::encryption_info_command ec,
4362 uint32_t object_size) {
4363 outs() << " cmd LC_ENCRYPTION_INFO\n";
4364 outs() << " cmdsize " << ec.cmdsize;
4365 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command))
4366 outs() << " Incorrect size\n";
4369 outs() << " cryptoff " << ec.cryptoff;
4370 if (ec.cryptoff > object_size)
4371 outs() << " (past end of file)\n";
4374 outs() << " cryptsize " << ec.cryptsize;
4375 if (ec.cryptsize > object_size)
4376 outs() << " (past end of file)\n";
4379 outs() << " cryptid " << ec.cryptid << "\n";
4382 static void PrintEncryptionInfoCommand64(MachO::encryption_info_command_64 ec,
4383 uint32_t object_size) {
4384 outs() << " cmd LC_ENCRYPTION_INFO_64\n";
4385 outs() << " cmdsize " << ec.cmdsize;
4386 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command_64))
4387 outs() << " Incorrect size\n";
4390 outs() << " cryptoff " << ec.cryptoff;
4391 if (ec.cryptoff > object_size)
4392 outs() << " (past end of file)\n";
4395 outs() << " cryptsize " << ec.cryptsize;
4396 if (ec.cryptsize > object_size)
4397 outs() << " (past end of file)\n";
4400 outs() << " cryptid " << ec.cryptid << "\n";
4401 outs() << " pad " << ec.pad << "\n";
4404 static void PrintLinkerOptionCommand(MachO::linker_option_command lo,
4406 outs() << " cmd LC_LINKER_OPTION\n";
4407 outs() << " cmdsize " << lo.cmdsize;
4408 if (lo.cmdsize < sizeof(struct MachO::linker_option_command))
4409 outs() << " Incorrect size\n";
4412 outs() << " count " << lo.count << "\n";
4413 const char *string = Ptr + sizeof(struct MachO::linker_option_command);
4414 uint32_t left = lo.cmdsize - sizeof(struct MachO::linker_option_command);
4417 while (*string == '\0' && left > 0) {
4423 outs() << " string #" << i << " " << format("%.*s\n", left, string);
4424 uint32_t NullPos = StringRef(string, left).find('\0');
4425 uint32_t len = std::min(NullPos, left) + 1;
4431 outs() << " count " << lo.count << " does not match number of strings "
4435 static void PrintSubFrameworkCommand(MachO::sub_framework_command sub,
4437 outs() << " cmd LC_SUB_FRAMEWORK\n";
4438 outs() << " cmdsize " << sub.cmdsize;
4439 if (sub.cmdsize < sizeof(struct MachO::sub_framework_command))
4440 outs() << " Incorrect size\n";
4443 if (sub.umbrella < sub.cmdsize) {
4444 const char *P = Ptr + sub.umbrella;
4445 outs() << " umbrella " << P << " (offset " << sub.umbrella << ")\n";
4447 outs() << " umbrella ?(bad offset " << sub.umbrella << ")\n";
4451 static void PrintSubUmbrellaCommand(MachO::sub_umbrella_command sub,
4453 outs() << " cmd LC_SUB_UMBRELLA\n";
4454 outs() << " cmdsize " << sub.cmdsize;
4455 if (sub.cmdsize < sizeof(struct MachO::sub_umbrella_command))
4456 outs() << " Incorrect size\n";
4459 if (sub.sub_umbrella < sub.cmdsize) {
4460 const char *P = Ptr + sub.sub_umbrella;
4461 outs() << " sub_umbrella " << P << " (offset " << sub.sub_umbrella << ")\n";
4463 outs() << " sub_umbrella ?(bad offset " << sub.sub_umbrella << ")\n";
4467 static void PrintSubLibraryCommand(MachO::sub_library_command sub,
4469 outs() << " cmd LC_SUB_LIBRARY\n";
4470 outs() << " cmdsize " << sub.cmdsize;
4471 if (sub.cmdsize < sizeof(struct MachO::sub_library_command))
4472 outs() << " Incorrect size\n";
4475 if (sub.sub_library < sub.cmdsize) {
4476 const char *P = Ptr + sub.sub_library;
4477 outs() << " sub_library " << P << " (offset " << sub.sub_library << ")\n";
4479 outs() << " sub_library ?(bad offset " << sub.sub_library << ")\n";
4483 static void PrintSubClientCommand(MachO::sub_client_command sub,
4485 outs() << " cmd LC_SUB_CLIENT\n";
4486 outs() << " cmdsize " << sub.cmdsize;
4487 if (sub.cmdsize < sizeof(struct MachO::sub_client_command))
4488 outs() << " Incorrect size\n";
4491 if (sub.client < sub.cmdsize) {
4492 const char *P = Ptr + sub.client;
4493 outs() << " client " << P << " (offset " << sub.client << ")\n";
4495 outs() << " client ?(bad offset " << sub.client << ")\n";
4499 static void PrintRoutinesCommand(MachO::routines_command r) {
4500 outs() << " cmd LC_ROUTINES\n";
4501 outs() << " cmdsize " << r.cmdsize;
4502 if (r.cmdsize != sizeof(struct MachO::routines_command))
4503 outs() << " Incorrect size\n";
4506 outs() << " init_address " << format("0x%08" PRIx32, r.init_address) << "\n";
4507 outs() << " init_module " << r.init_module << "\n";
4508 outs() << " reserved1 " << r.reserved1 << "\n";
4509 outs() << " reserved2 " << r.reserved2 << "\n";
4510 outs() << " reserved3 " << r.reserved3 << "\n";
4511 outs() << " reserved4 " << r.reserved4 << "\n";
4512 outs() << " reserved5 " << r.reserved5 << "\n";
4513 outs() << " reserved6 " << r.reserved6 << "\n";
4516 static void PrintRoutinesCommand64(MachO::routines_command_64 r) {
4517 outs() << " cmd LC_ROUTINES_64\n";
4518 outs() << " cmdsize " << r.cmdsize;
4519 if (r.cmdsize != sizeof(struct MachO::routines_command_64))
4520 outs() << " Incorrect size\n";
4523 outs() << " init_address " << format("0x%016" PRIx64, r.init_address) << "\n";
4524 outs() << " init_module " << r.init_module << "\n";
4525 outs() << " reserved1 " << r.reserved1 << "\n";
4526 outs() << " reserved2 " << r.reserved2 << "\n";
4527 outs() << " reserved3 " << r.reserved3 << "\n";
4528 outs() << " reserved4 " << r.reserved4 << "\n";
4529 outs() << " reserved5 " << r.reserved5 << "\n";
4530 outs() << " reserved6 " << r.reserved6 << "\n";
4533 static void Print_x86_thread_state64_t(MachO::x86_thread_state64_t &cpu64) {
4534 outs() << " rax " << format("0x%016" PRIx64, cpu64.rax);
4535 outs() << " rbx " << format("0x%016" PRIx64, cpu64.rbx);
4536 outs() << " rcx " << format("0x%016" PRIx64, cpu64.rcx) << "\n";
4537 outs() << " rdx " << format("0x%016" PRIx64, cpu64.rdx);
4538 outs() << " rdi " << format("0x%016" PRIx64, cpu64.rdi);
4539 outs() << " rsi " << format("0x%016" PRIx64, cpu64.rsi) << "\n";
4540 outs() << " rbp " << format("0x%016" PRIx64, cpu64.rbp);
4541 outs() << " rsp " << format("0x%016" PRIx64, cpu64.rsp);
4542 outs() << " r8 " << format("0x%016" PRIx64, cpu64.r8) << "\n";
4543 outs() << " r9 " << format("0x%016" PRIx64, cpu64.r9);
4544 outs() << " r10 " << format("0x%016" PRIx64, cpu64.r10);
4545 outs() << " r11 " << format("0x%016" PRIx64, cpu64.r11) << "\n";
4546 outs() << " r12 " << format("0x%016" PRIx64, cpu64.r12);
4547 outs() << " r13 " << format("0x%016" PRIx64, cpu64.r13);
4548 outs() << " r14 " << format("0x%016" PRIx64, cpu64.r14) << "\n";
4549 outs() << " r15 " << format("0x%016" PRIx64, cpu64.r15);
4550 outs() << " rip " << format("0x%016" PRIx64, cpu64.rip) << "\n";
4551 outs() << "rflags " << format("0x%016" PRIx64, cpu64.rflags);
4552 outs() << " cs " << format("0x%016" PRIx64, cpu64.cs);
4553 outs() << " fs " << format("0x%016" PRIx64, cpu64.fs) << "\n";
4554 outs() << " gs " << format("0x%016" PRIx64, cpu64.gs) << "\n";
4557 static void Print_mmst_reg(MachO::mmst_reg_t &r) {
4559 outs() << "\t mmst_reg ";
4560 for (f = 0; f < 10; f++)
4561 outs() << format("%02" PRIx32, (r.mmst_reg[f] & 0xff)) << " ";
4563 outs() << "\t mmst_rsrv ";
4564 for (f = 0; f < 6; f++)
4565 outs() << format("%02" PRIx32, (r.mmst_rsrv[f] & 0xff)) << " ";
4569 static void Print_xmm_reg(MachO::xmm_reg_t &r) {
4571 outs() << "\t xmm_reg ";
4572 for (f = 0; f < 16; f++)
4573 outs() << format("%02" PRIx32, (r.xmm_reg[f] & 0xff)) << " ";
4577 static void Print_x86_float_state_t(MachO::x86_float_state64_t &fpu) {
4578 outs() << "\t fpu_reserved[0] " << fpu.fpu_reserved[0];
4579 outs() << " fpu_reserved[1] " << fpu.fpu_reserved[1] << "\n";
4580 outs() << "\t control: invalid " << fpu.fpu_fcw.invalid;
4581 outs() << " denorm " << fpu.fpu_fcw.denorm;
4582 outs() << " zdiv " << fpu.fpu_fcw.zdiv;
4583 outs() << " ovrfl " << fpu.fpu_fcw.ovrfl;
4584 outs() << " undfl " << fpu.fpu_fcw.undfl;
4585 outs() << " precis " << fpu.fpu_fcw.precis << "\n";
4586 outs() << "\t\t pc ";
4587 if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_24B)
4588 outs() << "FP_PREC_24B ";
4589 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_53B)
4590 outs() << "FP_PREC_53B ";
4591 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_64B)
4592 outs() << "FP_PREC_64B ";
4594 outs() << fpu.fpu_fcw.pc << " ";
4596 if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_NEAR)
4597 outs() << "FP_RND_NEAR ";
4598 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_DOWN)
4599 outs() << "FP_RND_DOWN ";
4600 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_UP)
4601 outs() << "FP_RND_UP ";
4602 else if (fpu.fpu_fcw.rc == MachO::x86_FP_CHOP)
4603 outs() << "FP_CHOP ";
4605 outs() << "\t status: invalid " << fpu.fpu_fsw.invalid;
4606 outs() << " denorm " << fpu.fpu_fsw.denorm;
4607 outs() << " zdiv " << fpu.fpu_fsw.zdiv;
4608 outs() << " ovrfl " << fpu.fpu_fsw.ovrfl;
4609 outs() << " undfl " << fpu.fpu_fsw.undfl;
4610 outs() << " precis " << fpu.fpu_fsw.precis;
4611 outs() << " stkflt " << fpu.fpu_fsw.stkflt << "\n";
4612 outs() << "\t errsumm " << fpu.fpu_fsw.errsumm;
4613 outs() << " c0 " << fpu.fpu_fsw.c0;
4614 outs() << " c1 " << fpu.fpu_fsw.c1;
4615 outs() << " c2 " << fpu.fpu_fsw.c2;
4616 outs() << " tos " << fpu.fpu_fsw.tos;
4617 outs() << " c3 " << fpu.fpu_fsw.c3;
4618 outs() << " busy " << fpu.fpu_fsw.busy << "\n";
4619 outs() << "\t fpu_ftw " << format("0x%02" PRIx32, fpu.fpu_ftw);
4620 outs() << " fpu_rsrv1 " << format("0x%02" PRIx32, fpu.fpu_rsrv1);
4621 outs() << " fpu_fop " << format("0x%04" PRIx32, fpu.fpu_fop);
4622 outs() << " fpu_ip " << format("0x%08" PRIx32, fpu.fpu_ip) << "\n";
4623 outs() << "\t fpu_cs " << format("0x%04" PRIx32, fpu.fpu_cs);
4624 outs() << " fpu_rsrv2 " << format("0x%04" PRIx32, fpu.fpu_rsrv2);
4625 outs() << " fpu_dp " << format("0x%08" PRIx32, fpu.fpu_dp);
4626 outs() << " fpu_ds " << format("0x%04" PRIx32, fpu.fpu_ds) << "\n";
4627 outs() << "\t fpu_rsrv3 " << format("0x%04" PRIx32, fpu.fpu_rsrv3);
4628 outs() << " fpu_mxcsr " << format("0x%08" PRIx32, fpu.fpu_mxcsr);
4629 outs() << " fpu_mxcsrmask " << format("0x%08" PRIx32, fpu.fpu_mxcsrmask);
4631 outs() << "\t fpu_stmm0:\n";
4632 Print_mmst_reg(fpu.fpu_stmm0);
4633 outs() << "\t fpu_stmm1:\n";
4634 Print_mmst_reg(fpu.fpu_stmm1);
4635 outs() << "\t fpu_stmm2:\n";
4636 Print_mmst_reg(fpu.fpu_stmm2);
4637 outs() << "\t fpu_stmm3:\n";
4638 Print_mmst_reg(fpu.fpu_stmm3);
4639 outs() << "\t fpu_stmm4:\n";
4640 Print_mmst_reg(fpu.fpu_stmm4);
4641 outs() << "\t fpu_stmm5:\n";
4642 Print_mmst_reg(fpu.fpu_stmm5);
4643 outs() << "\t fpu_stmm6:\n";
4644 Print_mmst_reg(fpu.fpu_stmm6);
4645 outs() << "\t fpu_stmm7:\n";
4646 Print_mmst_reg(fpu.fpu_stmm7);
4647 outs() << "\t fpu_xmm0:\n";
4648 Print_xmm_reg(fpu.fpu_xmm0);
4649 outs() << "\t fpu_xmm1:\n";
4650 Print_xmm_reg(fpu.fpu_xmm1);
4651 outs() << "\t fpu_xmm2:\n";
4652 Print_xmm_reg(fpu.fpu_xmm2);
4653 outs() << "\t fpu_xmm3:\n";
4654 Print_xmm_reg(fpu.fpu_xmm3);
4655 outs() << "\t fpu_xmm4:\n";
4656 Print_xmm_reg(fpu.fpu_xmm4);
4657 outs() << "\t fpu_xmm5:\n";
4658 Print_xmm_reg(fpu.fpu_xmm5);
4659 outs() << "\t fpu_xmm6:\n";
4660 Print_xmm_reg(fpu.fpu_xmm6);
4661 outs() << "\t fpu_xmm7:\n";
4662 Print_xmm_reg(fpu.fpu_xmm7);
4663 outs() << "\t fpu_xmm8:\n";
4664 Print_xmm_reg(fpu.fpu_xmm8);
4665 outs() << "\t fpu_xmm9:\n";
4666 Print_xmm_reg(fpu.fpu_xmm9);
4667 outs() << "\t fpu_xmm10:\n";
4668 Print_xmm_reg(fpu.fpu_xmm10);
4669 outs() << "\t fpu_xmm11:\n";
4670 Print_xmm_reg(fpu.fpu_xmm11);
4671 outs() << "\t fpu_xmm12:\n";
4672 Print_xmm_reg(fpu.fpu_xmm12);
4673 outs() << "\t fpu_xmm13:\n";
4674 Print_xmm_reg(fpu.fpu_xmm13);
4675 outs() << "\t fpu_xmm14:\n";
4676 Print_xmm_reg(fpu.fpu_xmm14);
4677 outs() << "\t fpu_xmm15:\n";
4678 Print_xmm_reg(fpu.fpu_xmm15);
4679 outs() << "\t fpu_rsrv4:\n";
4680 for (uint32_t f = 0; f < 6; f++) {
4682 for (uint32_t g = 0; g < 16; g++)
4683 outs() << format("%02" PRIx32, fpu.fpu_rsrv4[f * g]) << " ";
4686 outs() << "\t fpu_reserved1 " << format("0x%08" PRIx32, fpu.fpu_reserved1);
4690 static void Print_x86_exception_state_t(MachO::x86_exception_state64_t &exc64) {
4691 outs() << "\t trapno " << format("0x%08" PRIx32, exc64.trapno);
4692 outs() << " err " << format("0x%08" PRIx32, exc64.err);
4693 outs() << " faultvaddr " << format("0x%016" PRIx64, exc64.faultvaddr) << "\n";
4696 static void PrintThreadCommand(MachO::thread_command t, const char *Ptr,
4697 bool isLittleEndian, uint32_t cputype) {
4698 if (t.cmd == MachO::LC_THREAD)
4699 outs() << " cmd LC_THREAD\n";
4700 else if (t.cmd == MachO::LC_UNIXTHREAD)
4701 outs() << " cmd LC_UNIXTHREAD\n";
4703 outs() << " cmd " << t.cmd << " (unknown)\n";
4704 outs() << " cmdsize " << t.cmdsize;
4705 if (t.cmdsize < sizeof(struct MachO::thread_command) + 2 * sizeof(uint32_t))
4706 outs() << " Incorrect size\n";
4710 const char *begin = Ptr + sizeof(struct MachO::thread_command);
4711 const char *end = Ptr + t.cmdsize;
4712 uint32_t flavor, count, left;
4713 if (cputype == MachO::CPU_TYPE_X86_64) {
4714 while (begin < end) {
4715 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4716 memcpy((char *)&flavor, begin, sizeof(uint32_t));
4717 begin += sizeof(uint32_t);
4722 if (isLittleEndian != sys::IsLittleEndianHost)
4723 sys::swapByteOrder(flavor);
4724 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4725 memcpy((char *)&count, begin, sizeof(uint32_t));
4726 begin += sizeof(uint32_t);
4731 if (isLittleEndian != sys::IsLittleEndianHost)
4732 sys::swapByteOrder(count);
4733 if (flavor == MachO::x86_THREAD_STATE64) {
4734 outs() << " flavor x86_THREAD_STATE64\n";
4735 if (count == MachO::x86_THREAD_STATE64_COUNT)
4736 outs() << " count x86_THREAD_STATE64_COUNT\n";
4738 outs() << " count " << count
4739 << " (not x86_THREAD_STATE64_COUNT)\n";
4740 MachO::x86_thread_state64_t cpu64;
4742 if (left >= sizeof(MachO::x86_thread_state64_t)) {
4743 memcpy(&cpu64, begin, sizeof(MachO::x86_thread_state64_t));
4744 begin += sizeof(MachO::x86_thread_state64_t);
4746 memset(&cpu64, '\0', sizeof(MachO::x86_thread_state64_t));
4747 memcpy(&cpu64, begin, left);
4750 if (isLittleEndian != sys::IsLittleEndianHost)
4752 Print_x86_thread_state64_t(cpu64);
4753 } else if (flavor == MachO::x86_THREAD_STATE) {
4754 outs() << " flavor x86_THREAD_STATE\n";
4755 if (count == MachO::x86_THREAD_STATE_COUNT)
4756 outs() << " count x86_THREAD_STATE_COUNT\n";
4758 outs() << " count " << count
4759 << " (not x86_THREAD_STATE_COUNT)\n";
4760 struct MachO::x86_thread_state_t ts;
4762 if (left >= sizeof(MachO::x86_thread_state_t)) {
4763 memcpy(&ts, begin, sizeof(MachO::x86_thread_state_t));
4764 begin += sizeof(MachO::x86_thread_state_t);
4766 memset(&ts, '\0', sizeof(MachO::x86_thread_state_t));
4767 memcpy(&ts, begin, left);
4770 if (isLittleEndian != sys::IsLittleEndianHost)
4772 if (ts.tsh.flavor == MachO::x86_THREAD_STATE64) {
4773 outs() << "\t tsh.flavor x86_THREAD_STATE64 ";
4774 if (ts.tsh.count == MachO::x86_THREAD_STATE64_COUNT)
4775 outs() << "tsh.count x86_THREAD_STATE64_COUNT\n";
4777 outs() << "tsh.count " << ts.tsh.count
4778 << " (not x86_THREAD_STATE64_COUNT\n";
4779 Print_x86_thread_state64_t(ts.uts.ts64);
4781 outs() << "\t tsh.flavor " << ts.tsh.flavor << " tsh.count "
4782 << ts.tsh.count << "\n";
4784 } else if (flavor == MachO::x86_FLOAT_STATE) {
4785 outs() << " flavor x86_FLOAT_STATE\n";
4786 if (count == MachO::x86_FLOAT_STATE_COUNT)
4787 outs() << " count x86_FLOAT_STATE_COUNT\n";
4789 outs() << " count " << count << " (not x86_FLOAT_STATE_COUNT)\n";
4790 struct MachO::x86_float_state_t fs;
4792 if (left >= sizeof(MachO::x86_float_state_t)) {
4793 memcpy(&fs, begin, sizeof(MachO::x86_float_state_t));
4794 begin += sizeof(MachO::x86_float_state_t);
4796 memset(&fs, '\0', sizeof(MachO::x86_float_state_t));
4797 memcpy(&fs, begin, left);
4800 if (isLittleEndian != sys::IsLittleEndianHost)
4802 if (fs.fsh.flavor == MachO::x86_FLOAT_STATE64) {
4803 outs() << "\t fsh.flavor x86_FLOAT_STATE64 ";
4804 if (fs.fsh.count == MachO::x86_FLOAT_STATE64_COUNT)
4805 outs() << "fsh.count x86_FLOAT_STATE64_COUNT\n";
4807 outs() << "fsh.count " << fs.fsh.count
4808 << " (not x86_FLOAT_STATE64_COUNT\n";
4809 Print_x86_float_state_t(fs.ufs.fs64);
4811 outs() << "\t fsh.flavor " << fs.fsh.flavor << " fsh.count "
4812 << fs.fsh.count << "\n";
4814 } else if (flavor == MachO::x86_EXCEPTION_STATE) {
4815 outs() << " flavor x86_EXCEPTION_STATE\n";
4816 if (count == MachO::x86_EXCEPTION_STATE_COUNT)
4817 outs() << " count x86_EXCEPTION_STATE_COUNT\n";
4819 outs() << " count " << count
4820 << " (not x86_EXCEPTION_STATE_COUNT)\n";
4821 struct MachO::x86_exception_state_t es;
4823 if (left >= sizeof(MachO::x86_exception_state_t)) {
4824 memcpy(&es, begin, sizeof(MachO::x86_exception_state_t));
4825 begin += sizeof(MachO::x86_exception_state_t);
4827 memset(&es, '\0', sizeof(MachO::x86_exception_state_t));
4828 memcpy(&es, begin, left);
4831 if (isLittleEndian != sys::IsLittleEndianHost)
4833 if (es.esh.flavor == MachO::x86_EXCEPTION_STATE64) {
4834 outs() << "\t esh.flavor x86_EXCEPTION_STATE64\n";
4835 if (es.esh.count == MachO::x86_EXCEPTION_STATE64_COUNT)
4836 outs() << "\t esh.count x86_EXCEPTION_STATE64_COUNT\n";
4838 outs() << "\t esh.count " << es.esh.count
4839 << " (not x86_EXCEPTION_STATE64_COUNT\n";
4840 Print_x86_exception_state_t(es.ues.es64);
4842 outs() << "\t esh.flavor " << es.esh.flavor << " esh.count "
4843 << es.esh.count << "\n";
4846 outs() << " flavor " << flavor << " (unknown)\n";
4847 outs() << " count " << count << "\n";
4848 outs() << " state (unknown)\n";
4849 begin += count * sizeof(uint32_t);
4853 while (begin < end) {
4854 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4855 memcpy((char *)&flavor, begin, sizeof(uint32_t));
4856 begin += sizeof(uint32_t);
4861 if (isLittleEndian != sys::IsLittleEndianHost)
4862 sys::swapByteOrder(flavor);
4863 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4864 memcpy((char *)&count, begin, sizeof(uint32_t));
4865 begin += sizeof(uint32_t);
4870 if (isLittleEndian != sys::IsLittleEndianHost)
4871 sys::swapByteOrder(count);
4872 outs() << " flavor " << flavor << "\n";
4873 outs() << " count " << count << "\n";
4874 outs() << " state (Unknown cputype/cpusubtype)\n";
4875 begin += count * sizeof(uint32_t);
4880 static void PrintDylibCommand(MachO::dylib_command dl, const char *Ptr) {
4881 if (dl.cmd == MachO::LC_ID_DYLIB)
4882 outs() << " cmd LC_ID_DYLIB\n";
4883 else if (dl.cmd == MachO::LC_LOAD_DYLIB)
4884 outs() << " cmd LC_LOAD_DYLIB\n";
4885 else if (dl.cmd == MachO::LC_LOAD_WEAK_DYLIB)
4886 outs() << " cmd LC_LOAD_WEAK_DYLIB\n";
4887 else if (dl.cmd == MachO::LC_REEXPORT_DYLIB)
4888 outs() << " cmd LC_REEXPORT_DYLIB\n";
4889 else if (dl.cmd == MachO::LC_LAZY_LOAD_DYLIB)
4890 outs() << " cmd LC_LAZY_LOAD_DYLIB\n";
4891 else if (dl.cmd == MachO::LC_LOAD_UPWARD_DYLIB)
4892 outs() << " cmd LC_LOAD_UPWARD_DYLIB\n";
4894 outs() << " cmd " << dl.cmd << " (unknown)\n";
4895 outs() << " cmdsize " << dl.cmdsize;
4896 if (dl.cmdsize < sizeof(struct MachO::dylib_command))
4897 outs() << " Incorrect size\n";
4900 if (dl.dylib.name < dl.cmdsize) {
4901 const char *P = (const char *)(Ptr) + dl.dylib.name;
4902 outs() << " name " << P << " (offset " << dl.dylib.name << ")\n";
4904 outs() << " name ?(bad offset " << dl.dylib.name << ")\n";
4906 outs() << " time stamp " << dl.dylib.timestamp << " ";
4907 time_t t = dl.dylib.timestamp;
4908 outs() << ctime(&t);
4909 outs() << " current version ";
4910 if (dl.dylib.current_version == 0xffffffff)
4913 outs() << ((dl.dylib.current_version >> 16) & 0xffff) << "."
4914 << ((dl.dylib.current_version >> 8) & 0xff) << "."
4915 << (dl.dylib.current_version & 0xff) << "\n";
4916 outs() << "compatibility version ";
4917 if (dl.dylib.compatibility_version == 0xffffffff)
4920 outs() << ((dl.dylib.compatibility_version >> 16) & 0xffff) << "."
4921 << ((dl.dylib.compatibility_version >> 8) & 0xff) << "."
4922 << (dl.dylib.compatibility_version & 0xff) << "\n";
4925 static void PrintLinkEditDataCommand(MachO::linkedit_data_command ld,
4926 uint32_t object_size) {
4927 if (ld.cmd == MachO::LC_CODE_SIGNATURE)
4928 outs() << " cmd LC_FUNCTION_STARTS\n";
4929 else if (ld.cmd == MachO::LC_SEGMENT_SPLIT_INFO)
4930 outs() << " cmd LC_SEGMENT_SPLIT_INFO\n";
4931 else if (ld.cmd == MachO::LC_FUNCTION_STARTS)
4932 outs() << " cmd LC_FUNCTION_STARTS\n";
4933 else if (ld.cmd == MachO::LC_DATA_IN_CODE)
4934 outs() << " cmd LC_DATA_IN_CODE\n";
4935 else if (ld.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS)
4936 outs() << " cmd LC_DYLIB_CODE_SIGN_DRS\n";
4937 else if (ld.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT)
4938 outs() << " cmd LC_LINKER_OPTIMIZATION_HINT\n";
4940 outs() << " cmd " << ld.cmd << " (?)\n";
4941 outs() << " cmdsize " << ld.cmdsize;
4942 if (ld.cmdsize != sizeof(struct MachO::linkedit_data_command))
4943 outs() << " Incorrect size\n";
4946 outs() << " dataoff " << ld.dataoff;
4947 if (ld.dataoff > object_size)
4948 outs() << " (past end of file)\n";
4951 outs() << " datasize " << ld.datasize;
4952 uint64_t big_size = ld.dataoff;
4953 big_size += ld.datasize;
4954 if (big_size > object_size)
4955 outs() << " (past end of file)\n";
4960 static void PrintLoadCommands(const MachOObjectFile *Obj, uint32_t ncmds,
4961 uint32_t filetype, uint32_t cputype,
4965 StringRef Buf = Obj->getData();
4966 MachOObjectFile::LoadCommandInfo Command = Obj->getFirstLoadCommandInfo();
4967 for (unsigned i = 0;; ++i) {
4968 outs() << "Load command " << i << "\n";
4969 if (Command.C.cmd == MachO::LC_SEGMENT) {
4970 MachO::segment_command SLC = Obj->getSegmentLoadCommand(Command);
4971 const char *sg_segname = SLC.segname;
4972 PrintSegmentCommand(SLC.cmd, SLC.cmdsize, SLC.segname, SLC.vmaddr,
4973 SLC.vmsize, SLC.fileoff, SLC.filesize, SLC.maxprot,
4974 SLC.initprot, SLC.nsects, SLC.flags, Buf.size(),
4976 for (unsigned j = 0; j < SLC.nsects; j++) {
4977 MachO::section S = Obj->getSection(Command, j);
4978 PrintSection(S.sectname, S.segname, S.addr, S.size, S.offset, S.align,
4979 S.reloff, S.nreloc, S.flags, S.reserved1, S.reserved2,
4980 SLC.cmd, sg_segname, filetype, Buf.size(), verbose);
4982 } else if (Command.C.cmd == MachO::LC_SEGMENT_64) {
4983 MachO::segment_command_64 SLC_64 = Obj->getSegment64LoadCommand(Command);
4984 const char *sg_segname = SLC_64.segname;
4985 PrintSegmentCommand(SLC_64.cmd, SLC_64.cmdsize, SLC_64.segname,
4986 SLC_64.vmaddr, SLC_64.vmsize, SLC_64.fileoff,
4987 SLC_64.filesize, SLC_64.maxprot, SLC_64.initprot,
4988 SLC_64.nsects, SLC_64.flags, Buf.size(), verbose);
4989 for (unsigned j = 0; j < SLC_64.nsects; j++) {
4990 MachO::section_64 S_64 = Obj->getSection64(Command, j);
4991 PrintSection(S_64.sectname, S_64.segname, S_64.addr, S_64.size,
4992 S_64.offset, S_64.align, S_64.reloff, S_64.nreloc,
4993 S_64.flags, S_64.reserved1, S_64.reserved2, SLC_64.cmd,
4994 sg_segname, filetype, Buf.size(), verbose);
4996 } else if (Command.C.cmd == MachO::LC_SYMTAB) {
4997 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
4998 PrintSymtabLoadCommand(Symtab, Obj->is64Bit(), Buf.size());
4999 } else if (Command.C.cmd == MachO::LC_DYSYMTAB) {
5000 MachO::dysymtab_command Dysymtab = Obj->getDysymtabLoadCommand();
5001 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
5002 PrintDysymtabLoadCommand(Dysymtab, Symtab.nsyms, Buf.size(),
5004 } else if (Command.C.cmd == MachO::LC_DYLD_INFO ||
5005 Command.C.cmd == MachO::LC_DYLD_INFO_ONLY) {
5006 MachO::dyld_info_command DyldInfo = Obj->getDyldInfoLoadCommand(Command);
5007 PrintDyldInfoLoadCommand(DyldInfo, Buf.size());
5008 } else if (Command.C.cmd == MachO::LC_LOAD_DYLINKER ||
5009 Command.C.cmd == MachO::LC_ID_DYLINKER ||
5010 Command.C.cmd == MachO::LC_DYLD_ENVIRONMENT) {
5011 MachO::dylinker_command Dyld = Obj->getDylinkerCommand(Command);
5012 PrintDyldLoadCommand(Dyld, Command.Ptr);
5013 } else if (Command.C.cmd == MachO::LC_UUID) {
5014 MachO::uuid_command Uuid = Obj->getUuidCommand(Command);
5015 PrintUuidLoadCommand(Uuid);
5016 } else if (Command.C.cmd == MachO::LC_RPATH) {
5017 MachO::rpath_command Rpath = Obj->getRpathCommand(Command);
5018 PrintRpathLoadCommand(Rpath, Command.Ptr);
5019 } else if (Command.C.cmd == MachO::LC_VERSION_MIN_MACOSX ||
5020 Command.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS) {
5021 MachO::version_min_command Vd = Obj->getVersionMinLoadCommand(Command);
5022 PrintVersionMinLoadCommand(Vd);
5023 } else if (Command.C.cmd == MachO::LC_SOURCE_VERSION) {
5024 MachO::source_version_command Sd = Obj->getSourceVersionCommand(Command);
5025 PrintSourceVersionCommand(Sd);
5026 } else if (Command.C.cmd == MachO::LC_MAIN) {
5027 MachO::entry_point_command Ep = Obj->getEntryPointCommand(Command);
5028 PrintEntryPointCommand(Ep);
5029 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO) {
5030 MachO::encryption_info_command Ei =
5031 Obj->getEncryptionInfoCommand(Command);
5032 PrintEncryptionInfoCommand(Ei, Buf.size());
5033 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO_64) {
5034 MachO::encryption_info_command_64 Ei =
5035 Obj->getEncryptionInfoCommand64(Command);
5036 PrintEncryptionInfoCommand64(Ei, Buf.size());
5037 } else if (Command.C.cmd == MachO::LC_LINKER_OPTION) {
5038 MachO::linker_option_command Lo =
5039 Obj->getLinkerOptionLoadCommand(Command);
5040 PrintLinkerOptionCommand(Lo, Command.Ptr);
5041 } else if (Command.C.cmd == MachO::LC_SUB_FRAMEWORK) {
5042 MachO::sub_framework_command Sf = Obj->getSubFrameworkCommand(Command);
5043 PrintSubFrameworkCommand(Sf, Command.Ptr);
5044 } else if (Command.C.cmd == MachO::LC_SUB_UMBRELLA) {
5045 MachO::sub_umbrella_command Sf = Obj->getSubUmbrellaCommand(Command);
5046 PrintSubUmbrellaCommand(Sf, Command.Ptr);
5047 } else if (Command.C.cmd == MachO::LC_SUB_LIBRARY) {
5048 MachO::sub_library_command Sl = Obj->getSubLibraryCommand(Command);
5049 PrintSubLibraryCommand(Sl, Command.Ptr);
5050 } else if (Command.C.cmd == MachO::LC_SUB_CLIENT) {
5051 MachO::sub_client_command Sc = Obj->getSubClientCommand(Command);
5052 PrintSubClientCommand(Sc, Command.Ptr);
5053 } else if (Command.C.cmd == MachO::LC_ROUTINES) {
5054 MachO::routines_command Rc = Obj->getRoutinesCommand(Command);
5055 PrintRoutinesCommand(Rc);
5056 } else if (Command.C.cmd == MachO::LC_ROUTINES_64) {
5057 MachO::routines_command_64 Rc = Obj->getRoutinesCommand64(Command);
5058 PrintRoutinesCommand64(Rc);
5059 } else if (Command.C.cmd == MachO::LC_THREAD ||
5060 Command.C.cmd == MachO::LC_UNIXTHREAD) {
5061 MachO::thread_command Tc = Obj->getThreadCommand(Command);
5062 PrintThreadCommand(Tc, Command.Ptr, Obj->isLittleEndian(), cputype);
5063 } else if (Command.C.cmd == MachO::LC_LOAD_DYLIB ||
5064 Command.C.cmd == MachO::LC_ID_DYLIB ||
5065 Command.C.cmd == MachO::LC_LOAD_WEAK_DYLIB ||
5066 Command.C.cmd == MachO::LC_REEXPORT_DYLIB ||
5067 Command.C.cmd == MachO::LC_LAZY_LOAD_DYLIB ||
5068 Command.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) {
5069 MachO::dylib_command Dl = Obj->getDylibIDLoadCommand(Command);
5070 PrintDylibCommand(Dl, Command.Ptr);
5071 } else if (Command.C.cmd == MachO::LC_CODE_SIGNATURE ||
5072 Command.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO ||
5073 Command.C.cmd == MachO::LC_FUNCTION_STARTS ||
5074 Command.C.cmd == MachO::LC_DATA_IN_CODE ||
5075 Command.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS ||
5076 Command.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) {
5077 MachO::linkedit_data_command Ld =
5078 Obj->getLinkeditDataLoadCommand(Command);
5079 PrintLinkEditDataCommand(Ld, Buf.size());
5081 outs() << " cmd ?(" << format("0x%08" PRIx32, Command.C.cmd)
5083 outs() << " cmdsize " << Command.C.cmdsize << "\n";
5084 // TODO: get and print the raw bytes of the load command.
5086 // TODO: print all the other kinds of load commands.
5090 Command = Obj->getNextLoadCommandInfo(Command);
5094 static void getAndPrintMachHeader(const MachOObjectFile *Obj, uint32_t &ncmds,
5095 uint32_t &filetype, uint32_t &cputype,
5097 if (Obj->is64Bit()) {
5098 MachO::mach_header_64 H_64;
5099 H_64 = Obj->getHeader64();
5100 PrintMachHeader(H_64.magic, H_64.cputype, H_64.cpusubtype, H_64.filetype,
5101 H_64.ncmds, H_64.sizeofcmds, H_64.flags, verbose);
5103 filetype = H_64.filetype;
5104 cputype = H_64.cputype;
5106 MachO::mach_header H;
5107 H = Obj->getHeader();
5108 PrintMachHeader(H.magic, H.cputype, H.cpusubtype, H.filetype, H.ncmds,
5109 H.sizeofcmds, H.flags, verbose);
5111 filetype = H.filetype;
5112 cputype = H.cputype;
5116 void llvm::printMachOFileHeader(const object::ObjectFile *Obj) {
5117 const MachOObjectFile *file = dyn_cast<const MachOObjectFile>(Obj);
5119 uint32_t filetype = 0;
5120 uint32_t cputype = 0;
5121 getAndPrintMachHeader(file, ncmds, filetype, cputype, true);
5122 PrintLoadCommands(file, ncmds, filetype, cputype, true);
5125 //===----------------------------------------------------------------------===//
5126 // export trie dumping
5127 //===----------------------------------------------------------------------===//
5129 void llvm::printMachOExportsTrie(const object::MachOObjectFile *Obj) {
5130 for (const llvm::object::ExportEntry &Entry : Obj->exports()) {
5131 uint64_t Flags = Entry.flags();
5132 bool ReExport = (Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT);
5133 bool WeakDef = (Flags & MachO::EXPORT_SYMBOL_FLAGS_WEAK_DEFINITION);
5134 bool ThreadLocal = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
5135 MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL);
5136 bool Abs = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
5137 MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE);
5138 bool Resolver = (Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER);
5140 outs() << "[re-export] ";
5142 outs() << format("0x%08llX ",
5143 Entry.address()); // FIXME:add in base address
5144 outs() << Entry.name();
5145 if (WeakDef || ThreadLocal || Resolver || Abs) {
5146 bool NeedsComma = false;
5149 outs() << "weak_def";
5155 outs() << "per-thread";
5161 outs() << "absolute";
5167 outs() << format("resolver=0x%08llX", Entry.other());
5173 StringRef DylibName = "unknown";
5174 int Ordinal = Entry.other() - 1;
5175 Obj->getLibraryShortNameByIndex(Ordinal, DylibName);
5176 if (Entry.otherName().empty())
5177 outs() << " (from " << DylibName << ")";
5179 outs() << " (" << Entry.otherName() << " from " << DylibName << ")";
5185 //===----------------------------------------------------------------------===//
5186 // rebase table dumping
5187 //===----------------------------------------------------------------------===//
5192 SegInfo(const object::MachOObjectFile *Obj);
5194 StringRef segmentName(uint32_t SegIndex);
5195 StringRef sectionName(uint32_t SegIndex, uint64_t SegOffset);
5196 uint64_t address(uint32_t SegIndex, uint64_t SegOffset);
5199 struct SectionInfo {
5202 StringRef SectionName;
5203 StringRef SegmentName;
5204 uint64_t OffsetInSegment;
5205 uint64_t SegmentStartAddress;
5206 uint32_t SegmentIndex;
5208 const SectionInfo &findSection(uint32_t SegIndex, uint64_t SegOffset);
5209 SmallVector<SectionInfo, 32> Sections;
5213 SegInfo::SegInfo(const object::MachOObjectFile *Obj) {
5214 // Build table of sections so segIndex/offset pairs can be translated.
5215 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0;
5216 StringRef CurSegName;
5217 uint64_t CurSegAddress;
5218 for (const SectionRef &Section : Obj->sections()) {
5220 if (error(Section.getName(Info.SectionName)))
5222 Info.Address = Section.getAddress();
5223 Info.Size = Section.getSize();
5225 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl());
5226 if (!Info.SegmentName.equals(CurSegName)) {
5228 CurSegName = Info.SegmentName;
5229 CurSegAddress = Info.Address;
5231 Info.SegmentIndex = CurSegIndex - 1;
5232 Info.OffsetInSegment = Info.Address - CurSegAddress;
5233 Info.SegmentStartAddress = CurSegAddress;
5234 Sections.push_back(Info);
5238 StringRef SegInfo::segmentName(uint32_t SegIndex) {
5239 for (const SectionInfo &SI : Sections) {
5240 if (SI.SegmentIndex == SegIndex)
5241 return SI.SegmentName;
5243 llvm_unreachable("invalid segIndex");
5246 const SegInfo::SectionInfo &SegInfo::findSection(uint32_t SegIndex,
5247 uint64_t OffsetInSeg) {
5248 for (const SectionInfo &SI : Sections) {
5249 if (SI.SegmentIndex != SegIndex)
5251 if (SI.OffsetInSegment > OffsetInSeg)
5253 if (OffsetInSeg >= (SI.OffsetInSegment + SI.Size))
5257 llvm_unreachable("segIndex and offset not in any section");
5260 StringRef SegInfo::sectionName(uint32_t SegIndex, uint64_t OffsetInSeg) {
5261 return findSection(SegIndex, OffsetInSeg).SectionName;
5264 uint64_t SegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) {
5265 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg);
5266 return SI.SegmentStartAddress + OffsetInSeg;
5269 void llvm::printMachORebaseTable(const object::MachOObjectFile *Obj) {
5270 // Build table of sections so names can used in final output.
5271 SegInfo sectionTable(Obj);
5273 outs() << "segment section address type\n";
5274 for (const llvm::object::MachORebaseEntry &Entry : Obj->rebaseTable()) {
5275 uint32_t SegIndex = Entry.segmentIndex();
5276 uint64_t OffsetInSeg = Entry.segmentOffset();
5277 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5278 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5279 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5281 // Table lines look like: __DATA __nl_symbol_ptr 0x0000F00C pointer
5282 outs() << format("%-8s %-18s 0x%08" PRIX64 " %s\n",
5283 SegmentName.str().c_str(), SectionName.str().c_str(),
5284 Address, Entry.typeName().str().c_str());
5288 static StringRef ordinalName(const object::MachOObjectFile *Obj, int Ordinal) {
5289 StringRef DylibName;
5291 case MachO::BIND_SPECIAL_DYLIB_SELF:
5292 return "this-image";
5293 case MachO::BIND_SPECIAL_DYLIB_MAIN_EXECUTABLE:
5294 return "main-executable";
5295 case MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP:
5296 return "flat-namespace";
5299 std::error_code EC =
5300 Obj->getLibraryShortNameByIndex(Ordinal - 1, DylibName);
5302 return "<<bad library ordinal>>";
5306 return "<<unknown special ordinal>>";
5309 //===----------------------------------------------------------------------===//
5310 // bind table dumping
5311 //===----------------------------------------------------------------------===//
5313 void llvm::printMachOBindTable(const object::MachOObjectFile *Obj) {
5314 // Build table of sections so names can used in final output.
5315 SegInfo sectionTable(Obj);
5317 outs() << "segment section address type "
5318 "addend dylib symbol\n";
5319 for (const llvm::object::MachOBindEntry &Entry : Obj->bindTable()) {
5320 uint32_t SegIndex = Entry.segmentIndex();
5321 uint64_t OffsetInSeg = Entry.segmentOffset();
5322 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5323 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5324 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5326 // Table lines look like:
5327 // __DATA __got 0x00012010 pointer 0 libSystem ___stack_chk_guard
5329 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_WEAK_IMPORT)
5330 Attr = " (weak_import)";
5331 outs() << left_justify(SegmentName, 8) << " "
5332 << left_justify(SectionName, 18) << " "
5333 << format_hex(Address, 10, true) << " "
5334 << left_justify(Entry.typeName(), 8) << " "
5335 << format_decimal(Entry.addend(), 8) << " "
5336 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
5337 << Entry.symbolName() << Attr << "\n";
5341 //===----------------------------------------------------------------------===//
5342 // lazy bind table dumping
5343 //===----------------------------------------------------------------------===//
5345 void llvm::printMachOLazyBindTable(const object::MachOObjectFile *Obj) {
5346 // Build table of sections so names can used in final output.
5347 SegInfo sectionTable(Obj);
5349 outs() << "segment section address "
5351 for (const llvm::object::MachOBindEntry &Entry : Obj->lazyBindTable()) {
5352 uint32_t SegIndex = Entry.segmentIndex();
5353 uint64_t OffsetInSeg = Entry.segmentOffset();
5354 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5355 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5356 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5358 // Table lines look like:
5359 // __DATA __got 0x00012010 libSystem ___stack_chk_guard
5360 outs() << left_justify(SegmentName, 8) << " "
5361 << left_justify(SectionName, 18) << " "
5362 << format_hex(Address, 10, true) << " "
5363 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
5364 << Entry.symbolName() << "\n";
5368 //===----------------------------------------------------------------------===//
5369 // weak bind table dumping
5370 //===----------------------------------------------------------------------===//
5372 void llvm::printMachOWeakBindTable(const object::MachOObjectFile *Obj) {
5373 // Build table of sections so names can used in final output.
5374 SegInfo sectionTable(Obj);
5376 outs() << "segment section address "
5377 "type addend symbol\n";
5378 for (const llvm::object::MachOBindEntry &Entry : Obj->weakBindTable()) {
5379 // Strong symbols don't have a location to update.
5380 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) {
5381 outs() << " strong "
5382 << Entry.symbolName() << "\n";
5385 uint32_t SegIndex = Entry.segmentIndex();
5386 uint64_t OffsetInSeg = Entry.segmentOffset();
5387 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5388 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5389 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5391 // Table lines look like:
5392 // __DATA __data 0x00001000 pointer 0 _foo
5393 outs() << left_justify(SegmentName, 8) << " "
5394 << left_justify(SectionName, 18) << " "
5395 << format_hex(Address, 10, true) << " "
5396 << left_justify(Entry.typeName(), 8) << " "
5397 << format_decimal(Entry.addend(), 8) << " " << Entry.symbolName()
5402 // get_dyld_bind_info_symbolname() is used for disassembly and passed an
5403 // address, ReferenceValue, in the Mach-O file and looks in the dyld bind
5404 // information for that address. If the address is found its binding symbol
5405 // name is returned. If not nullptr is returned.
5406 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
5407 struct DisassembleInfo *info) {
5408 if (info->bindtable == nullptr) {
5409 info->bindtable = new (BindTable);
5410 SegInfo sectionTable(info->O);
5411 for (const llvm::object::MachOBindEntry &Entry : info->O->bindTable()) {
5412 uint32_t SegIndex = Entry.segmentIndex();
5413 uint64_t OffsetInSeg = Entry.segmentOffset();
5414 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5415 const char *SymbolName = nullptr;
5416 StringRef name = Entry.symbolName();
5418 SymbolName = name.data();
5419 info->bindtable->push_back(std::make_pair(Address, SymbolName));
5422 for (bind_table_iterator BI = info->bindtable->begin(),
5423 BE = info->bindtable->end();
5425 uint64_t Address = BI->first;
5426 if (ReferenceValue == Address) {
5427 const char *SymbolName = BI->second;