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 DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF,
620 StringRef DisSegName, StringRef DisSectName);
622 static void DumpSectionContents(StringRef Filename, MachOObjectFile *O,
624 SymbolAddressMap AddrMap;
626 CreateSymbolAddressMap(O, &AddrMap);
628 for (unsigned i = 0; i < DumpSections.size(); ++i) {
629 StringRef DumpSection = DumpSections[i];
630 std::pair<StringRef, StringRef> DumpSegSectName;
631 DumpSegSectName = DumpSection.split(',');
632 StringRef DumpSegName, DumpSectName;
633 if (DumpSegSectName.second.size()) {
634 DumpSegName = DumpSegSectName.first;
635 DumpSectName = DumpSegSectName.second;
638 DumpSectName = DumpSegSectName.first;
640 for (const SectionRef &Section : O->sections()) {
642 Section.getName(SectName);
643 DataRefImpl Ref = Section.getRawDataRefImpl();
644 StringRef SegName = O->getSectionFinalSegmentName(Ref);
645 if ((DumpSegName.empty() || SegName == DumpSegName) &&
646 (SectName == DumpSectName)) {
647 outs() << "Contents of (" << SegName << "," << SectName
649 uint32_t section_flags;
651 const MachO::section_64 Sec = O->getSection64(Ref);
652 section_flags = Sec.flags;
655 const MachO::section Sec = O->getSection(Ref);
656 section_flags = Sec.flags;
658 uint32_t section_type = section_flags & MachO::SECTION_TYPE;
661 Section.getContents(BytesStr);
662 const char *sect = reinterpret_cast<const char *>(BytesStr.data());
663 uint32_t sect_size = BytesStr.size();
664 uint64_t sect_addr = Section.getAddress();
667 if ((section_flags & MachO::S_ATTR_PURE_INSTRUCTIONS) ||
668 (section_flags & MachO::S_ATTR_SOME_INSTRUCTIONS)) {
669 DisassembleMachO(Filename, O, SegName, SectName);
672 switch (section_type) {
673 case MachO::S_REGULAR:
674 DumpRawSectionContents(O, sect, sect_size, sect_addr);
676 case MachO::S_ZEROFILL:
677 outs() << "zerofill section and has no contents in the file\n";
679 case MachO::S_MOD_INIT_FUNC_POINTERS:
680 case MachO::S_MOD_TERM_FUNC_POINTERS:
681 DumpInitTermPointerSection(O, sect, sect_size, sect_addr, &AddrMap,
685 outs() << "Unknown section type ("
686 << format("0x%08" PRIx32, section_type) << ")\n";
687 DumpRawSectionContents(O, sect, sect_size, sect_addr);
691 if (section_type == MachO::S_ZEROFILL)
692 outs() << "zerofill section and has no contents in the file\n";
694 DumpRawSectionContents(O, sect, sect_size, sect_addr);
701 // checkMachOAndArchFlags() checks to see if the ObjectFile is a Mach-O file
702 // and if it is and there is a list of architecture flags is specified then
703 // check to make sure this Mach-O file is one of those architectures or all
704 // architectures were specified. If not then an error is generated and this
705 // routine returns false. Else it returns true.
706 static bool checkMachOAndArchFlags(ObjectFile *O, StringRef Filename) {
707 if (isa<MachOObjectFile>(O) && !ArchAll && ArchFlags.size() != 0) {
708 MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O);
709 bool ArchFound = false;
710 MachO::mach_header H;
711 MachO::mach_header_64 H_64;
713 if (MachO->is64Bit()) {
714 H_64 = MachO->MachOObjectFile::getHeader64();
715 T = MachOObjectFile::getArch(H_64.cputype, H_64.cpusubtype);
717 H = MachO->MachOObjectFile::getHeader();
718 T = MachOObjectFile::getArch(H.cputype, H.cpusubtype);
721 for (i = 0; i < ArchFlags.size(); ++i) {
722 if (ArchFlags[i] == T.getArchName())
727 errs() << "llvm-objdump: file: " + Filename + " does not contain "
728 << "architecture: " + ArchFlags[i] + "\n";
735 // ProcessMachO() is passed a single opened Mach-O file, which may be an
736 // archive member and or in a slice of a universal file. It prints the
737 // the file name and header info and then processes it according to the
738 // command line options.
739 static void ProcessMachO(StringRef Filename, MachOObjectFile *MachOOF,
740 StringRef ArchiveMemberName = StringRef(),
741 StringRef ArchitectureName = StringRef()) {
742 // If we are doing some processing here on the Mach-O file print the header
743 // info. And don't print it otherwise like in the case of printing the
744 // UniversalHeaders or ArchiveHeaders.
745 if (Disassemble || PrivateHeaders || ExportsTrie || Rebase || Bind ||
746 LazyBind || WeakBind || IndirectSymbols || DataInCode || LinkOptHints ||
747 DumpSections.size() != 0) {
749 if (!ArchiveMemberName.empty())
750 outs() << '(' << ArchiveMemberName << ')';
751 if (!ArchitectureName.empty())
752 outs() << " (architecture " << ArchitectureName << ")";
757 DisassembleMachO(Filename, MachOOF, "__TEXT", "__text");
759 PrintIndirectSymbols(MachOOF, true);
761 PrintDataInCodeTable(MachOOF, true);
763 PrintLinkOptHints(MachOOF);
765 PrintRelocations(MachOOF);
767 PrintSectionHeaders(MachOOF);
769 PrintSectionContents(MachOOF);
770 if (DumpSections.size() != 0)
771 DumpSectionContents(Filename, MachOOF, true);
773 PrintSymbolTable(MachOOF);
775 printMachOUnwindInfo(MachOOF);
777 printMachOFileHeader(MachOOF);
779 printExportsTrie(MachOOF);
781 printRebaseTable(MachOOF);
783 printBindTable(MachOOF);
785 printLazyBindTable(MachOOF);
787 printWeakBindTable(MachOOF);
790 // printUnknownCPUType() helps print_fat_headers for unknown CPU's.
791 static void printUnknownCPUType(uint32_t cputype, uint32_t cpusubtype) {
792 outs() << " cputype (" << cputype << ")\n";
793 outs() << " cpusubtype (" << cpusubtype << ")\n";
796 // printCPUType() helps print_fat_headers by printing the cputype and
797 // pusubtype (symbolically for the one's it knows about).
798 static void printCPUType(uint32_t cputype, uint32_t cpusubtype) {
800 case MachO::CPU_TYPE_I386:
801 switch (cpusubtype) {
802 case MachO::CPU_SUBTYPE_I386_ALL:
803 outs() << " cputype CPU_TYPE_I386\n";
804 outs() << " cpusubtype CPU_SUBTYPE_I386_ALL\n";
807 printUnknownCPUType(cputype, cpusubtype);
811 case MachO::CPU_TYPE_X86_64:
812 switch (cpusubtype) {
813 case MachO::CPU_SUBTYPE_X86_64_ALL:
814 outs() << " cputype CPU_TYPE_X86_64\n";
815 outs() << " cpusubtype CPU_SUBTYPE_X86_64_ALL\n";
817 case MachO::CPU_SUBTYPE_X86_64_H:
818 outs() << " cputype CPU_TYPE_X86_64\n";
819 outs() << " cpusubtype CPU_SUBTYPE_X86_64_H\n";
822 printUnknownCPUType(cputype, cpusubtype);
826 case MachO::CPU_TYPE_ARM:
827 switch (cpusubtype) {
828 case MachO::CPU_SUBTYPE_ARM_ALL:
829 outs() << " cputype CPU_TYPE_ARM\n";
830 outs() << " cpusubtype CPU_SUBTYPE_ARM_ALL\n";
832 case MachO::CPU_SUBTYPE_ARM_V4T:
833 outs() << " cputype CPU_TYPE_ARM\n";
834 outs() << " cpusubtype CPU_SUBTYPE_ARM_V4T\n";
836 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
837 outs() << " cputype CPU_TYPE_ARM\n";
838 outs() << " cpusubtype CPU_SUBTYPE_ARM_V5TEJ\n";
840 case MachO::CPU_SUBTYPE_ARM_XSCALE:
841 outs() << " cputype CPU_TYPE_ARM\n";
842 outs() << " cpusubtype CPU_SUBTYPE_ARM_XSCALE\n";
844 case MachO::CPU_SUBTYPE_ARM_V6:
845 outs() << " cputype CPU_TYPE_ARM\n";
846 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6\n";
848 case MachO::CPU_SUBTYPE_ARM_V6M:
849 outs() << " cputype CPU_TYPE_ARM\n";
850 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6M\n";
852 case MachO::CPU_SUBTYPE_ARM_V7:
853 outs() << " cputype CPU_TYPE_ARM\n";
854 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7\n";
856 case MachO::CPU_SUBTYPE_ARM_V7EM:
857 outs() << " cputype CPU_TYPE_ARM\n";
858 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7EM\n";
860 case MachO::CPU_SUBTYPE_ARM_V7K:
861 outs() << " cputype CPU_TYPE_ARM\n";
862 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7K\n";
864 case MachO::CPU_SUBTYPE_ARM_V7M:
865 outs() << " cputype CPU_TYPE_ARM\n";
866 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7M\n";
868 case MachO::CPU_SUBTYPE_ARM_V7S:
869 outs() << " cputype CPU_TYPE_ARM\n";
870 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7S\n";
873 printUnknownCPUType(cputype, cpusubtype);
877 case MachO::CPU_TYPE_ARM64:
878 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
879 case MachO::CPU_SUBTYPE_ARM64_ALL:
880 outs() << " cputype CPU_TYPE_ARM64\n";
881 outs() << " cpusubtype CPU_SUBTYPE_ARM64_ALL\n";
884 printUnknownCPUType(cputype, cpusubtype);
889 printUnknownCPUType(cputype, cpusubtype);
894 static void printMachOUniversalHeaders(const object::MachOUniversalBinary *UB,
896 outs() << "Fat headers\n";
898 outs() << "fat_magic FAT_MAGIC\n";
900 outs() << "fat_magic " << format("0x%" PRIx32, MachO::FAT_MAGIC) << "\n";
902 uint32_t nfat_arch = UB->getNumberOfObjects();
903 StringRef Buf = UB->getData();
904 uint64_t size = Buf.size();
905 uint64_t big_size = sizeof(struct MachO::fat_header) +
906 nfat_arch * sizeof(struct MachO::fat_arch);
907 outs() << "nfat_arch " << UB->getNumberOfObjects();
909 outs() << " (malformed, contains zero architecture types)\n";
910 else if (big_size > size)
911 outs() << " (malformed, architectures past end of file)\n";
915 for (uint32_t i = 0; i < nfat_arch; ++i) {
916 MachOUniversalBinary::ObjectForArch OFA(UB, i);
917 uint32_t cputype = OFA.getCPUType();
918 uint32_t cpusubtype = OFA.getCPUSubType();
919 outs() << "architecture ";
920 for (uint32_t j = 0; i != 0 && j <= i - 1; j++) {
921 MachOUniversalBinary::ObjectForArch other_OFA(UB, j);
922 uint32_t other_cputype = other_OFA.getCPUType();
923 uint32_t other_cpusubtype = other_OFA.getCPUSubType();
924 if (cputype != 0 && cpusubtype != 0 && cputype == other_cputype &&
925 (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) ==
926 (other_cpusubtype & ~MachO::CPU_SUBTYPE_MASK)) {
927 outs() << "(illegal duplicate architecture) ";
932 outs() << OFA.getArchTypeName() << "\n";
933 printCPUType(cputype, cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
936 outs() << " cputype " << cputype << "\n";
937 outs() << " cpusubtype " << (cpusubtype & ~MachO::CPU_SUBTYPE_MASK)
941 (cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64)
942 outs() << " capabilities CPU_SUBTYPE_LIB64\n";
944 outs() << " capabilities "
945 << format("0x%" PRIx32,
946 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24) << "\n";
947 outs() << " offset " << OFA.getOffset();
948 if (OFA.getOffset() > size)
949 outs() << " (past end of file)";
950 if (OFA.getOffset() % (1 << OFA.getAlign()) != 0)
951 outs() << " (not aligned on it's alignment (2^" << OFA.getAlign() << ")";
953 outs() << " size " << OFA.getSize();
954 big_size = OFA.getOffset() + OFA.getSize();
956 outs() << " (past end of file)";
958 outs() << " align 2^" << OFA.getAlign() << " (" << (1 << OFA.getAlign())
963 static void printArchiveChild(Archive::Child &C, bool verbose,
966 outs() << C.getChildOffset() << "\t";
967 sys::fs::perms Mode = C.getAccessMode();
969 // FIXME: this first dash, "-", is for (Mode & S_IFMT) == S_IFREG.
970 // But there is nothing in sys::fs::perms for S_IFMT or S_IFREG.
972 if (Mode & sys::fs::owner_read)
976 if (Mode & sys::fs::owner_write)
980 if (Mode & sys::fs::owner_exe)
984 if (Mode & sys::fs::group_read)
988 if (Mode & sys::fs::group_write)
992 if (Mode & sys::fs::group_exe)
996 if (Mode & sys::fs::others_read)
1000 if (Mode & sys::fs::others_write)
1004 if (Mode & sys::fs::others_exe)
1009 outs() << format("0%o ", Mode);
1012 unsigned UID = C.getUID();
1013 outs() << format("%3d/", UID);
1014 unsigned GID = C.getGID();
1015 outs() << format("%-3d ", GID);
1016 uint64_t Size = C.getRawSize();
1017 outs() << format("%5" PRId64, Size) << " ";
1019 StringRef RawLastModified = C.getRawLastModified();
1022 if (RawLastModified.getAsInteger(10, Seconds))
1023 outs() << "(date: \"%s\" contains non-decimal chars) " << RawLastModified;
1025 // Since cime(3) returns a 26 character string of the form:
1026 // "Sun Sep 16 01:03:52 1973\n\0"
1027 // just print 24 characters.
1029 outs() << format("%.24s ", ctime(&t));
1032 outs() << RawLastModified << " ";
1036 ErrorOr<StringRef> NameOrErr = C.getName();
1037 if (NameOrErr.getError()) {
1038 StringRef RawName = C.getRawName();
1039 outs() << RawName << "\n";
1041 StringRef Name = NameOrErr.get();
1042 outs() << Name << "\n";
1045 StringRef RawName = C.getRawName();
1046 outs() << RawName << "\n";
1050 static void printArchiveHeaders(Archive *A, bool verbose, bool print_offset) {
1051 if (A->hasSymbolTable()) {
1052 Archive::child_iterator S = A->getSymbolTableChild();
1053 Archive::Child C = *S;
1054 printArchiveChild(C, verbose, print_offset);
1056 for (Archive::child_iterator I = A->child_begin(), E = A->child_end(); I != E;
1058 Archive::Child C = *I;
1059 printArchiveChild(C, verbose, print_offset);
1063 // ParseInputMachO() parses the named Mach-O file in Filename and handles the
1064 // -arch flags selecting just those slices as specified by them and also parses
1065 // archive files. Then for each individual Mach-O file ProcessMachO() is
1066 // called to process the file based on the command line options.
1067 void llvm::ParseInputMachO(StringRef Filename) {
1068 // Check for -arch all and verifiy the -arch flags are valid.
1069 for (unsigned i = 0; i < ArchFlags.size(); ++i) {
1070 if (ArchFlags[i] == "all") {
1073 if (!MachOObjectFile::isValidArch(ArchFlags[i])) {
1074 errs() << "llvm-objdump: Unknown architecture named '" + ArchFlags[i] +
1075 "'for the -arch option\n";
1081 // Attempt to open the binary.
1082 ErrorOr<OwningBinary<Binary>> BinaryOrErr = createBinary(Filename);
1083 if (std::error_code EC = BinaryOrErr.getError()) {
1084 errs() << "llvm-objdump: '" << Filename << "': " << EC.message() << ".\n";
1087 Binary &Bin = *BinaryOrErr.get().getBinary();
1089 if (Archive *A = dyn_cast<Archive>(&Bin)) {
1090 outs() << "Archive : " << Filename << "\n";
1092 printArchiveHeaders(A, true, false);
1093 for (Archive::child_iterator I = A->child_begin(), E = A->child_end();
1095 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = I->getAsBinary();
1096 if (ChildOrErr.getError())
1098 if (MachOObjectFile *O = dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
1099 if (!checkMachOAndArchFlags(O, Filename))
1101 ProcessMachO(Filename, O, O->getFileName());
1106 if (UniversalHeaders) {
1107 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin))
1108 printMachOUniversalHeaders(UB, true);
1110 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin)) {
1111 // If we have a list of architecture flags specified dump only those.
1112 if (!ArchAll && ArchFlags.size() != 0) {
1113 // Look for a slice in the universal binary that matches each ArchFlag.
1115 for (unsigned i = 0; i < ArchFlags.size(); ++i) {
1117 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1118 E = UB->end_objects();
1120 if (ArchFlags[i] == I->getArchTypeName()) {
1122 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr =
1123 I->getAsObjectFile();
1124 std::string ArchitectureName = "";
1125 if (ArchFlags.size() > 1)
1126 ArchitectureName = I->getArchTypeName();
1128 ObjectFile &O = *ObjOrErr.get();
1129 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
1130 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
1131 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
1132 I->getAsArchive()) {
1133 std::unique_ptr<Archive> &A = *AOrErr;
1134 outs() << "Archive : " << Filename;
1135 if (!ArchitectureName.empty())
1136 outs() << " (architecture " << ArchitectureName << ")";
1139 printArchiveHeaders(A.get(), true, false);
1140 for (Archive::child_iterator AI = A->child_begin(),
1141 AE = A->child_end();
1143 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1144 if (ChildOrErr.getError())
1146 if (MachOObjectFile *O =
1147 dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
1148 ProcessMachO(Filename, O, O->getFileName(), ArchitectureName);
1154 errs() << "llvm-objdump: file: " + Filename + " does not contain "
1155 << "architecture: " + ArchFlags[i] + "\n";
1161 // No architecture flags were specified so if this contains a slice that
1162 // matches the host architecture dump only that.
1164 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1165 E = UB->end_objects();
1167 if (MachOObjectFile::getHostArch().getArchName() ==
1168 I->getArchTypeName()) {
1169 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
1170 std::string ArchiveName;
1171 ArchiveName.clear();
1173 ObjectFile &O = *ObjOrErr.get();
1174 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
1175 ProcessMachO(Filename, MachOOF);
1176 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
1177 I->getAsArchive()) {
1178 std::unique_ptr<Archive> &A = *AOrErr;
1179 outs() << "Archive : " << Filename << "\n";
1181 printArchiveHeaders(A.get(), true, false);
1182 for (Archive::child_iterator AI = A->child_begin(),
1183 AE = A->child_end();
1185 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1186 if (ChildOrErr.getError())
1188 if (MachOObjectFile *O =
1189 dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
1190 ProcessMachO(Filename, O, O->getFileName());
1197 // Either all architectures have been specified or none have been specified
1198 // and this does not contain the host architecture so dump all the slices.
1199 bool moreThanOneArch = UB->getNumberOfObjects() > 1;
1200 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1201 E = UB->end_objects();
1203 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
1204 std::string ArchitectureName = "";
1205 if (moreThanOneArch)
1206 ArchitectureName = I->getArchTypeName();
1208 ObjectFile &Obj = *ObjOrErr.get();
1209 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&Obj))
1210 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
1211 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr = I->getAsArchive()) {
1212 std::unique_ptr<Archive> &A = *AOrErr;
1213 outs() << "Archive : " << Filename;
1214 if (!ArchitectureName.empty())
1215 outs() << " (architecture " << ArchitectureName << ")";
1218 printArchiveHeaders(A.get(), true, false);
1219 for (Archive::child_iterator AI = A->child_begin(), AE = A->child_end();
1221 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1222 if (ChildOrErr.getError())
1224 if (MachOObjectFile *O =
1225 dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
1226 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(O))
1227 ProcessMachO(Filename, MachOOF, MachOOF->getFileName(),
1235 if (ObjectFile *O = dyn_cast<ObjectFile>(&Bin)) {
1236 if (!checkMachOAndArchFlags(O, Filename))
1238 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&*O)) {
1239 ProcessMachO(Filename, MachOOF);
1241 errs() << "llvm-objdump: '" << Filename << "': "
1242 << "Object is not a Mach-O file type.\n";
1244 errs() << "llvm-objdump: '" << Filename << "': "
1245 << "Unrecognized file type.\n";
1248 typedef std::pair<uint64_t, const char *> BindInfoEntry;
1249 typedef std::vector<BindInfoEntry> BindTable;
1250 typedef BindTable::iterator bind_table_iterator;
1252 // The block of info used by the Symbolizer call backs.
1253 struct DisassembleInfo {
1257 SymbolAddressMap *AddrMap;
1258 std::vector<SectionRef> *Sections;
1259 const char *class_name;
1260 const char *selector_name;
1262 char *demangled_name;
1265 BindTable *bindtable;
1268 // SymbolizerGetOpInfo() is the operand information call back function.
1269 // This is called to get the symbolic information for operand(s) of an
1270 // instruction when it is being done. This routine does this from
1271 // the relocation information, symbol table, etc. That block of information
1272 // is a pointer to the struct DisassembleInfo that was passed when the
1273 // disassembler context was created and passed to back to here when
1274 // called back by the disassembler for instruction operands that could have
1275 // relocation information. The address of the instruction containing operand is
1276 // at the Pc parameter. The immediate value the operand has is passed in
1277 // op_info->Value and is at Offset past the start of the instruction and has a
1278 // byte Size of 1, 2 or 4. The symbolc information is returned in TagBuf is the
1279 // LLVMOpInfo1 struct defined in the header "llvm-c/Disassembler.h" as symbol
1280 // names and addends of the symbolic expression to add for the operand. The
1281 // value of TagType is currently 1 (for the LLVMOpInfo1 struct). If symbolic
1282 // information is returned then this function returns 1 else it returns 0.
1283 int SymbolizerGetOpInfo(void *DisInfo, uint64_t Pc, uint64_t Offset,
1284 uint64_t Size, int TagType, void *TagBuf) {
1285 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
1286 struct LLVMOpInfo1 *op_info = (struct LLVMOpInfo1 *)TagBuf;
1287 uint64_t value = op_info->Value;
1289 // Make sure all fields returned are zero if we don't set them.
1290 memset((void *)op_info, '\0', sizeof(struct LLVMOpInfo1));
1291 op_info->Value = value;
1293 // If the TagType is not the value 1 which it code knows about or if no
1294 // verbose symbolic information is wanted then just return 0, indicating no
1295 // information is being returned.
1296 if (TagType != 1 || info->verbose == false)
1299 unsigned int Arch = info->O->getArch();
1300 if (Arch == Triple::x86) {
1301 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
1303 // First search the section's relocation entries (if any) for an entry
1304 // for this section offset.
1305 uint32_t sect_addr = info->S.getAddress();
1306 uint32_t sect_offset = (Pc + Offset) - sect_addr;
1307 bool reloc_found = false;
1309 MachO::any_relocation_info RE;
1310 bool isExtern = false;
1312 bool r_scattered = false;
1313 uint32_t r_value, pair_r_value, r_type;
1314 for (const RelocationRef &Reloc : info->S.relocations()) {
1315 uint64_t RelocOffset;
1316 Reloc.getOffset(RelocOffset);
1317 if (RelocOffset == sect_offset) {
1318 Rel = Reloc.getRawDataRefImpl();
1319 RE = info->O->getRelocation(Rel);
1320 r_type = info->O->getAnyRelocationType(RE);
1321 r_scattered = info->O->isRelocationScattered(RE);
1323 r_value = info->O->getScatteredRelocationValue(RE);
1324 if (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
1325 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF) {
1326 DataRefImpl RelNext = Rel;
1327 info->O->moveRelocationNext(RelNext);
1328 MachO::any_relocation_info RENext;
1329 RENext = info->O->getRelocation(RelNext);
1330 if (info->O->isRelocationScattered(RENext))
1331 pair_r_value = info->O->getScatteredRelocationValue(RENext);
1336 isExtern = info->O->getPlainRelocationExternal(RE);
1338 symbol_iterator RelocSym = Reloc.getSymbol();
1346 if (reloc_found && isExtern) {
1348 Symbol.getName(SymName);
1349 const char *name = SymName.data();
1350 op_info->AddSymbol.Present = 1;
1351 op_info->AddSymbol.Name = name;
1352 // For i386 extern relocation entries the value in the instruction is
1353 // the offset from the symbol, and value is already set in op_info->Value.
1356 if (reloc_found && (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
1357 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF)) {
1358 const char *add = GuessSymbolName(r_value, info->AddrMap);
1359 const char *sub = GuessSymbolName(pair_r_value, info->AddrMap);
1360 uint32_t offset = value - (r_value - pair_r_value);
1361 op_info->AddSymbol.Present = 1;
1363 op_info->AddSymbol.Name = add;
1365 op_info->AddSymbol.Value = r_value;
1366 op_info->SubtractSymbol.Present = 1;
1368 op_info->SubtractSymbol.Name = sub;
1370 op_info->SubtractSymbol.Value = pair_r_value;
1371 op_info->Value = offset;
1375 // Second search the external relocation entries of a fully linked image
1376 // (if any) for an entry that matches this segment offset.
1377 // uint32_t seg_offset = (Pc + Offset);
1379 } else if (Arch == Triple::x86_64) {
1380 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
1382 // First search the section's relocation entries (if any) for an entry
1383 // for this section offset.
1384 uint64_t sect_addr = info->S.getAddress();
1385 uint64_t sect_offset = (Pc + Offset) - sect_addr;
1386 bool reloc_found = false;
1388 MachO::any_relocation_info RE;
1389 bool isExtern = false;
1391 for (const RelocationRef &Reloc : info->S.relocations()) {
1392 uint64_t RelocOffset;
1393 Reloc.getOffset(RelocOffset);
1394 if (RelocOffset == sect_offset) {
1395 Rel = Reloc.getRawDataRefImpl();
1396 RE = info->O->getRelocation(Rel);
1397 // NOTE: Scattered relocations don't exist on x86_64.
1398 isExtern = info->O->getPlainRelocationExternal(RE);
1400 symbol_iterator RelocSym = Reloc.getSymbol();
1407 if (reloc_found && isExtern) {
1408 // The Value passed in will be adjusted by the Pc if the instruction
1409 // adds the Pc. But for x86_64 external relocation entries the Value
1410 // is the offset from the external symbol.
1411 if (info->O->getAnyRelocationPCRel(RE))
1412 op_info->Value -= Pc + Offset + Size;
1414 Symbol.getName(SymName);
1415 const char *name = SymName.data();
1416 unsigned Type = info->O->getAnyRelocationType(RE);
1417 if (Type == MachO::X86_64_RELOC_SUBTRACTOR) {
1418 DataRefImpl RelNext = Rel;
1419 info->O->moveRelocationNext(RelNext);
1420 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
1421 unsigned TypeNext = info->O->getAnyRelocationType(RENext);
1422 bool isExternNext = info->O->getPlainRelocationExternal(RENext);
1423 unsigned SymbolNum = info->O->getPlainRelocationSymbolNum(RENext);
1424 if (TypeNext == MachO::X86_64_RELOC_UNSIGNED && isExternNext) {
1425 op_info->SubtractSymbol.Present = 1;
1426 op_info->SubtractSymbol.Name = name;
1427 symbol_iterator RelocSymNext = info->O->getSymbolByIndex(SymbolNum);
1428 Symbol = *RelocSymNext;
1429 StringRef SymNameNext;
1430 Symbol.getName(SymNameNext);
1431 name = SymNameNext.data();
1434 // TODO: add the VariantKinds to op_info->VariantKind for relocation types
1435 // like: X86_64_RELOC_TLV, X86_64_RELOC_GOT_LOAD and X86_64_RELOC_GOT.
1436 op_info->AddSymbol.Present = 1;
1437 op_info->AddSymbol.Name = name;
1441 // Second search the external relocation entries of a fully linked image
1442 // (if any) for an entry that matches this segment offset.
1443 // uint64_t seg_offset = (Pc + Offset);
1445 } else if (Arch == Triple::arm) {
1446 if (Offset != 0 || (Size != 4 && Size != 2))
1448 // First search the section's relocation entries (if any) for an entry
1449 // for this section offset.
1450 uint32_t sect_addr = info->S.getAddress();
1451 uint32_t sect_offset = (Pc + Offset) - sect_addr;
1452 bool reloc_found = false;
1454 MachO::any_relocation_info RE;
1455 bool isExtern = false;
1457 bool r_scattered = false;
1458 uint32_t r_value, pair_r_value, r_type, r_length, other_half;
1459 for (const RelocationRef &Reloc : info->S.relocations()) {
1460 uint64_t RelocOffset;
1461 Reloc.getOffset(RelocOffset);
1462 if (RelocOffset == sect_offset) {
1463 Rel = Reloc.getRawDataRefImpl();
1464 RE = info->O->getRelocation(Rel);
1465 r_length = info->O->getAnyRelocationLength(RE);
1466 r_scattered = info->O->isRelocationScattered(RE);
1468 r_value = info->O->getScatteredRelocationValue(RE);
1469 r_type = info->O->getScatteredRelocationType(RE);
1471 r_type = info->O->getAnyRelocationType(RE);
1472 isExtern = info->O->getPlainRelocationExternal(RE);
1474 symbol_iterator RelocSym = Reloc.getSymbol();
1478 if (r_type == MachO::ARM_RELOC_HALF ||
1479 r_type == MachO::ARM_RELOC_SECTDIFF ||
1480 r_type == MachO::ARM_RELOC_LOCAL_SECTDIFF ||
1481 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1482 DataRefImpl RelNext = Rel;
1483 info->O->moveRelocationNext(RelNext);
1484 MachO::any_relocation_info RENext;
1485 RENext = info->O->getRelocation(RelNext);
1486 other_half = info->O->getAnyRelocationAddress(RENext) & 0xffff;
1487 if (info->O->isRelocationScattered(RENext))
1488 pair_r_value = info->O->getScatteredRelocationValue(RENext);
1494 if (reloc_found && isExtern) {
1496 Symbol.getName(SymName);
1497 const char *name = SymName.data();
1498 op_info->AddSymbol.Present = 1;
1499 op_info->AddSymbol.Name = name;
1502 case MachO::ARM_RELOC_HALF:
1503 if ((r_length & 0x1) == 1) {
1504 op_info->Value = value << 16 | other_half;
1505 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1507 op_info->Value = other_half << 16 | value;
1508 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1516 case MachO::ARM_RELOC_HALF:
1517 if ((r_length & 0x1) == 1) {
1518 op_info->Value = value << 16 | other_half;
1519 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1521 op_info->Value = other_half << 16 | value;
1522 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1531 // If we have a branch that is not an external relocation entry then
1532 // return 0 so the code in tryAddingSymbolicOperand() can use the
1533 // SymbolLookUp call back with the branch target address to look up the
1534 // symbol and possiblity add an annotation for a symbol stub.
1535 if (reloc_found && isExtern == 0 && (r_type == MachO::ARM_RELOC_BR24 ||
1536 r_type == MachO::ARM_THUMB_RELOC_BR22))
1539 uint32_t offset = 0;
1541 if (r_type == MachO::ARM_RELOC_HALF ||
1542 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1543 if ((r_length & 0x1) == 1)
1544 value = value << 16 | other_half;
1546 value = other_half << 16 | value;
1548 if (r_scattered && (r_type != MachO::ARM_RELOC_HALF &&
1549 r_type != MachO::ARM_RELOC_HALF_SECTDIFF)) {
1550 offset = value - r_value;
1555 if (reloc_found && r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1556 if ((r_length & 0x1) == 1)
1557 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1559 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1560 const char *add = GuessSymbolName(r_value, info->AddrMap);
1561 const char *sub = GuessSymbolName(pair_r_value, info->AddrMap);
1562 int32_t offset = value - (r_value - pair_r_value);
1563 op_info->AddSymbol.Present = 1;
1565 op_info->AddSymbol.Name = add;
1567 op_info->AddSymbol.Value = r_value;
1568 op_info->SubtractSymbol.Present = 1;
1570 op_info->SubtractSymbol.Name = sub;
1572 op_info->SubtractSymbol.Value = pair_r_value;
1573 op_info->Value = offset;
1577 if (reloc_found == false)
1580 op_info->AddSymbol.Present = 1;
1581 op_info->Value = offset;
1583 if (r_type == MachO::ARM_RELOC_HALF) {
1584 if ((r_length & 0x1) == 1)
1585 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1587 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1590 const char *add = GuessSymbolName(value, info->AddrMap);
1591 if (add != nullptr) {
1592 op_info->AddSymbol.Name = add;
1595 op_info->AddSymbol.Value = value;
1597 } else if (Arch == Triple::aarch64) {
1598 if (Offset != 0 || Size != 4)
1600 // First search the section's relocation entries (if any) for an entry
1601 // for this section offset.
1602 uint64_t sect_addr = info->S.getAddress();
1603 uint64_t sect_offset = (Pc + Offset) - sect_addr;
1604 bool reloc_found = false;
1606 MachO::any_relocation_info RE;
1607 bool isExtern = false;
1609 uint32_t r_type = 0;
1610 for (const RelocationRef &Reloc : info->S.relocations()) {
1611 uint64_t RelocOffset;
1612 Reloc.getOffset(RelocOffset);
1613 if (RelocOffset == sect_offset) {
1614 Rel = Reloc.getRawDataRefImpl();
1615 RE = info->O->getRelocation(Rel);
1616 r_type = info->O->getAnyRelocationType(RE);
1617 if (r_type == MachO::ARM64_RELOC_ADDEND) {
1618 DataRefImpl RelNext = Rel;
1619 info->O->moveRelocationNext(RelNext);
1620 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
1622 value = info->O->getPlainRelocationSymbolNum(RENext);
1623 op_info->Value = value;
1626 // NOTE: Scattered relocations don't exist on arm64.
1627 isExtern = info->O->getPlainRelocationExternal(RE);
1629 symbol_iterator RelocSym = Reloc.getSymbol();
1636 if (reloc_found && isExtern) {
1638 Symbol.getName(SymName);
1639 const char *name = SymName.data();
1640 op_info->AddSymbol.Present = 1;
1641 op_info->AddSymbol.Name = name;
1644 case MachO::ARM64_RELOC_PAGE21:
1646 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGE;
1648 case MachO::ARM64_RELOC_PAGEOFF12:
1650 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGEOFF;
1652 case MachO::ARM64_RELOC_GOT_LOAD_PAGE21:
1654 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGE;
1656 case MachO::ARM64_RELOC_GOT_LOAD_PAGEOFF12:
1658 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGEOFF;
1660 case MachO::ARM64_RELOC_TLVP_LOAD_PAGE21:
1661 /* @tvlppage is not implemented in llvm-mc */
1662 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVP;
1664 case MachO::ARM64_RELOC_TLVP_LOAD_PAGEOFF12:
1665 /* @tvlppageoff is not implemented in llvm-mc */
1666 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVOFF;
1669 case MachO::ARM64_RELOC_BRANCH26:
1670 op_info->VariantKind = LLVMDisassembler_VariantKind_None;
1681 // GuessCstringPointer is passed the address of what might be a pointer to a
1682 // literal string in a cstring section. If that address is in a cstring section
1683 // it returns a pointer to that string. Else it returns nullptr.
1684 const char *GuessCstringPointer(uint64_t ReferenceValue,
1685 struct DisassembleInfo *info) {
1686 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1687 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1688 for (unsigned I = 0;; ++I) {
1689 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1690 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1691 for (unsigned J = 0; J < Seg.nsects; ++J) {
1692 MachO::section_64 Sec = info->O->getSection64(Load, J);
1693 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1694 if (section_type == MachO::S_CSTRING_LITERALS &&
1695 ReferenceValue >= Sec.addr &&
1696 ReferenceValue < Sec.addr + Sec.size) {
1697 uint64_t sect_offset = ReferenceValue - Sec.addr;
1698 uint64_t object_offset = Sec.offset + sect_offset;
1699 StringRef MachOContents = info->O->getData();
1700 uint64_t object_size = MachOContents.size();
1701 const char *object_addr = (const char *)MachOContents.data();
1702 if (object_offset < object_size) {
1703 const char *name = object_addr + object_offset;
1710 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
1711 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
1712 for (unsigned J = 0; J < Seg.nsects; ++J) {
1713 MachO::section Sec = info->O->getSection(Load, J);
1714 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1715 if (section_type == MachO::S_CSTRING_LITERALS &&
1716 ReferenceValue >= Sec.addr &&
1717 ReferenceValue < Sec.addr + Sec.size) {
1718 uint64_t sect_offset = ReferenceValue - Sec.addr;
1719 uint64_t object_offset = Sec.offset + sect_offset;
1720 StringRef MachOContents = info->O->getData();
1721 uint64_t object_size = MachOContents.size();
1722 const char *object_addr = (const char *)MachOContents.data();
1723 if (object_offset < object_size) {
1724 const char *name = object_addr + object_offset;
1732 if (I == LoadCommandCount - 1)
1735 Load = info->O->getNextLoadCommandInfo(Load);
1740 // GuessIndirectSymbol returns the name of the indirect symbol for the
1741 // ReferenceValue passed in or nullptr. This is used when ReferenceValue maybe
1742 // an address of a symbol stub or a lazy or non-lazy pointer to associate the
1743 // symbol name being referenced by the stub or pointer.
1744 static const char *GuessIndirectSymbol(uint64_t ReferenceValue,
1745 struct DisassembleInfo *info) {
1746 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1747 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1748 MachO::dysymtab_command Dysymtab = info->O->getDysymtabLoadCommand();
1749 MachO::symtab_command Symtab = info->O->getSymtabLoadCommand();
1750 for (unsigned I = 0;; ++I) {
1751 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1752 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1753 for (unsigned J = 0; J < Seg.nsects; ++J) {
1754 MachO::section_64 Sec = info->O->getSection64(Load, J);
1755 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1756 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
1757 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
1758 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
1759 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
1760 section_type == MachO::S_SYMBOL_STUBS) &&
1761 ReferenceValue >= Sec.addr &&
1762 ReferenceValue < Sec.addr + Sec.size) {
1764 if (section_type == MachO::S_SYMBOL_STUBS)
1765 stride = Sec.reserved2;
1770 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
1771 if (index < Dysymtab.nindirectsyms) {
1772 uint32_t indirect_symbol =
1773 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
1774 if (indirect_symbol < Symtab.nsyms) {
1775 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
1776 SymbolRef Symbol = *Sym;
1778 Symbol.getName(SymName);
1779 const char *name = SymName.data();
1785 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
1786 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
1787 for (unsigned J = 0; J < Seg.nsects; ++J) {
1788 MachO::section Sec = info->O->getSection(Load, J);
1789 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1790 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
1791 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
1792 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
1793 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
1794 section_type == MachO::S_SYMBOL_STUBS) &&
1795 ReferenceValue >= Sec.addr &&
1796 ReferenceValue < Sec.addr + Sec.size) {
1798 if (section_type == MachO::S_SYMBOL_STUBS)
1799 stride = Sec.reserved2;
1804 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
1805 if (index < Dysymtab.nindirectsyms) {
1806 uint32_t indirect_symbol =
1807 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
1808 if (indirect_symbol < Symtab.nsyms) {
1809 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
1810 SymbolRef Symbol = *Sym;
1812 Symbol.getName(SymName);
1813 const char *name = SymName.data();
1820 if (I == LoadCommandCount - 1)
1823 Load = info->O->getNextLoadCommandInfo(Load);
1828 // method_reference() is called passing it the ReferenceName that might be
1829 // a reference it to an Objective-C method call. If so then it allocates and
1830 // assembles a method call string with the values last seen and saved in
1831 // the DisassembleInfo's class_name and selector_name fields. This is saved
1832 // into the method field of the info and any previous string is free'ed.
1833 // Then the class_name field in the info is set to nullptr. The method call
1834 // string is set into ReferenceName and ReferenceType is set to
1835 // LLVMDisassembler_ReferenceType_Out_Objc_Message. If this not a method call
1836 // then both ReferenceType and ReferenceName are left unchanged.
1837 static void method_reference(struct DisassembleInfo *info,
1838 uint64_t *ReferenceType,
1839 const char **ReferenceName) {
1840 unsigned int Arch = info->O->getArch();
1841 if (*ReferenceName != nullptr) {
1842 if (strcmp(*ReferenceName, "_objc_msgSend") == 0) {
1843 if (info->selector_name != nullptr) {
1844 if (info->method != nullptr)
1846 if (info->class_name != nullptr) {
1847 info->method = (char *)malloc(5 + strlen(info->class_name) +
1848 strlen(info->selector_name));
1849 if (info->method != nullptr) {
1850 strcpy(info->method, "+[");
1851 strcat(info->method, info->class_name);
1852 strcat(info->method, " ");
1853 strcat(info->method, info->selector_name);
1854 strcat(info->method, "]");
1855 *ReferenceName = info->method;
1856 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1859 info->method = (char *)malloc(9 + strlen(info->selector_name));
1860 if (info->method != nullptr) {
1861 if (Arch == Triple::x86_64)
1862 strcpy(info->method, "-[%rdi ");
1863 else if (Arch == Triple::aarch64)
1864 strcpy(info->method, "-[x0 ");
1866 strcpy(info->method, "-[r? ");
1867 strcat(info->method, info->selector_name);
1868 strcat(info->method, "]");
1869 *ReferenceName = info->method;
1870 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1873 info->class_name = nullptr;
1875 } else if (strcmp(*ReferenceName, "_objc_msgSendSuper2") == 0) {
1876 if (info->selector_name != nullptr) {
1877 if (info->method != nullptr)
1879 info->method = (char *)malloc(17 + strlen(info->selector_name));
1880 if (info->method != nullptr) {
1881 if (Arch == Triple::x86_64)
1882 strcpy(info->method, "-[[%rdi super] ");
1883 else if (Arch == Triple::aarch64)
1884 strcpy(info->method, "-[[x0 super] ");
1886 strcpy(info->method, "-[[r? super] ");
1887 strcat(info->method, info->selector_name);
1888 strcat(info->method, "]");
1889 *ReferenceName = info->method;
1890 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1892 info->class_name = nullptr;
1898 // GuessPointerPointer() is passed the address of what might be a pointer to
1899 // a reference to an Objective-C class, selector, message ref or cfstring.
1900 // If so the value of the pointer is returned and one of the booleans are set
1901 // to true. If not zero is returned and all the booleans are set to false.
1902 static uint64_t GuessPointerPointer(uint64_t ReferenceValue,
1903 struct DisassembleInfo *info,
1904 bool &classref, bool &selref, bool &msgref,
1910 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1911 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1912 for (unsigned I = 0;; ++I) {
1913 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1914 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1915 for (unsigned J = 0; J < Seg.nsects; ++J) {
1916 MachO::section_64 Sec = info->O->getSection64(Load, J);
1917 if ((strncmp(Sec.sectname, "__objc_selrefs", 16) == 0 ||
1918 strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
1919 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0 ||
1920 strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 ||
1921 strncmp(Sec.sectname, "__cfstring", 16) == 0) &&
1922 ReferenceValue >= Sec.addr &&
1923 ReferenceValue < Sec.addr + Sec.size) {
1924 uint64_t sect_offset = ReferenceValue - Sec.addr;
1925 uint64_t object_offset = Sec.offset + sect_offset;
1926 StringRef MachOContents = info->O->getData();
1927 uint64_t object_size = MachOContents.size();
1928 const char *object_addr = (const char *)MachOContents.data();
1929 if (object_offset < object_size) {
1930 uint64_t pointer_value;
1931 memcpy(&pointer_value, object_addr + object_offset,
1933 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1934 sys::swapByteOrder(pointer_value);
1935 if (strncmp(Sec.sectname, "__objc_selrefs", 16) == 0)
1937 else if (strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
1938 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0)
1940 else if (strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 &&
1941 ReferenceValue + 8 < Sec.addr + Sec.size) {
1943 memcpy(&pointer_value, object_addr + object_offset + 8,
1945 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1946 sys::swapByteOrder(pointer_value);
1947 } else if (strncmp(Sec.sectname, "__cfstring", 16) == 0)
1949 return pointer_value;
1956 // TODO: Look for LC_SEGMENT for 32-bit Mach-O files.
1957 if (I == LoadCommandCount - 1)
1960 Load = info->O->getNextLoadCommandInfo(Load);
1965 // get_pointer_64 returns a pointer to the bytes in the object file at the
1966 // Address from a section in the Mach-O file. And indirectly returns the
1967 // offset into the section, number of bytes left in the section past the offset
1968 // and which section is was being referenced. If the Address is not in a
1969 // section nullptr is returned.
1970 const char *get_pointer_64(uint64_t Address, uint32_t &offset, uint32_t &left,
1971 SectionRef &S, DisassembleInfo *info) {
1975 for (unsigned SectIdx = 0; SectIdx != info->Sections->size(); SectIdx++) {
1976 uint64_t SectAddress = ((*(info->Sections))[SectIdx]).getAddress();
1977 uint64_t SectSize = ((*(info->Sections))[SectIdx]).getSize();
1978 if (Address >= SectAddress && Address < SectAddress + SectSize) {
1979 S = (*(info->Sections))[SectIdx];
1980 offset = Address - SectAddress;
1981 left = SectSize - offset;
1982 StringRef SectContents;
1983 ((*(info->Sections))[SectIdx]).getContents(SectContents);
1984 return SectContents.data() + offset;
1990 // get_symbol_64() returns the name of a symbol (or nullptr) and the address of
1991 // the symbol indirectly through n_value. Based on the relocation information
1992 // for the specified section offset in the specified section reference.
1993 const char *get_symbol_64(uint32_t sect_offset, SectionRef S,
1994 DisassembleInfo *info, uint64_t &n_value) {
1996 if (info->verbose == false)
1999 // See if there is an external relocation entry at the sect_offset.
2000 bool reloc_found = false;
2002 MachO::any_relocation_info RE;
2003 bool isExtern = false;
2005 for (const RelocationRef &Reloc : S.relocations()) {
2006 uint64_t RelocOffset;
2007 Reloc.getOffset(RelocOffset);
2008 if (RelocOffset == sect_offset) {
2009 Rel = Reloc.getRawDataRefImpl();
2010 RE = info->O->getRelocation(Rel);
2011 if (info->O->isRelocationScattered(RE))
2013 isExtern = info->O->getPlainRelocationExternal(RE);
2015 symbol_iterator RelocSym = Reloc.getSymbol();
2022 // If there is an external relocation entry for a symbol in this section
2023 // at this section_offset then use that symbol's value for the n_value
2024 // and return its name.
2025 const char *SymbolName = nullptr;
2026 if (reloc_found && isExtern) {
2027 Symbol.getAddress(n_value);
2029 Symbol.getName(name);
2030 if (!name.empty()) {
2031 SymbolName = name.data();
2036 // TODO: For fully linked images, look through the external relocation
2037 // entries off the dynamic symtab command. For these the r_offset is from the
2038 // start of the first writeable segment in the Mach-O file. So the offset
2039 // to this section from that segment is passed to this routine by the caller,
2040 // as the database_offset. Which is the difference of the section's starting
2041 // address and the first writable segment.
2043 // NOTE: need add passing the database_offset to this routine.
2045 // TODO: We did not find an external relocation entry so look up the
2046 // ReferenceValue as an address of a symbol and if found return that symbol's
2049 // NOTE: need add passing the ReferenceValue to this routine. Then that code
2050 // would simply be this:
2051 // SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap);
2056 // These are structs in the Objective-C meta data and read to produce the
2057 // comments for disassembly. While these are part of the ABI they are no
2058 // public defintions. So the are here not in include/llvm/Support/MachO.h .
2060 // The cfstring object in a 64-bit Mach-O file.
2061 struct cfstring64_t {
2062 uint64_t isa; // class64_t * (64-bit pointer)
2063 uint64_t flags; // flag bits
2064 uint64_t characters; // char * (64-bit pointer)
2065 uint64_t length; // number of non-NULL characters in above
2068 // The class object in a 64-bit Mach-O file.
2070 uint64_t isa; // class64_t * (64-bit pointer)
2071 uint64_t superclass; // class64_t * (64-bit pointer)
2072 uint64_t cache; // Cache (64-bit pointer)
2073 uint64_t vtable; // IMP * (64-bit pointer)
2074 uint64_t data; // class_ro64_t * (64-bit pointer)
2077 struct class_ro64_t {
2079 uint32_t instanceStart;
2080 uint32_t instanceSize;
2082 uint64_t ivarLayout; // const uint8_t * (64-bit pointer)
2083 uint64_t name; // const char * (64-bit pointer)
2084 uint64_t baseMethods; // const method_list_t * (64-bit pointer)
2085 uint64_t baseProtocols; // const protocol_list_t * (64-bit pointer)
2086 uint64_t ivars; // const ivar_list_t * (64-bit pointer)
2087 uint64_t weakIvarLayout; // const uint8_t * (64-bit pointer)
2088 uint64_t baseProperties; // const struct objc_property_list (64-bit pointer)
2091 inline void swapStruct(struct cfstring64_t &cfs) {
2092 sys::swapByteOrder(cfs.isa);
2093 sys::swapByteOrder(cfs.flags);
2094 sys::swapByteOrder(cfs.characters);
2095 sys::swapByteOrder(cfs.length);
2098 inline void swapStruct(struct class64_t &c) {
2099 sys::swapByteOrder(c.isa);
2100 sys::swapByteOrder(c.superclass);
2101 sys::swapByteOrder(c.cache);
2102 sys::swapByteOrder(c.vtable);
2103 sys::swapByteOrder(c.data);
2106 inline void swapStruct(struct class_ro64_t &cro) {
2107 sys::swapByteOrder(cro.flags);
2108 sys::swapByteOrder(cro.instanceStart);
2109 sys::swapByteOrder(cro.instanceSize);
2110 sys::swapByteOrder(cro.reserved);
2111 sys::swapByteOrder(cro.ivarLayout);
2112 sys::swapByteOrder(cro.name);
2113 sys::swapByteOrder(cro.baseMethods);
2114 sys::swapByteOrder(cro.baseProtocols);
2115 sys::swapByteOrder(cro.ivars);
2116 sys::swapByteOrder(cro.weakIvarLayout);
2117 sys::swapByteOrder(cro.baseProperties);
2120 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
2121 struct DisassembleInfo *info);
2123 // get_objc2_64bit_class_name() is used for disassembly and is passed a pointer
2124 // to an Objective-C class and returns the class name. It is also passed the
2125 // address of the pointer, so when the pointer is zero as it can be in an .o
2126 // file, that is used to look for an external relocation entry with a symbol
2128 const char *get_objc2_64bit_class_name(uint64_t pointer_value,
2129 uint64_t ReferenceValue,
2130 struct DisassembleInfo *info) {
2132 uint32_t offset, left;
2135 // The pointer_value can be 0 in an object file and have a relocation
2136 // entry for the class symbol at the ReferenceValue (the address of the
2138 if (pointer_value == 0) {
2139 r = get_pointer_64(ReferenceValue, offset, left, S, info);
2140 if (r == nullptr || left < sizeof(uint64_t))
2143 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
2144 if (symbol_name == nullptr)
2146 const char *class_name = strrchr(symbol_name, '$');
2147 if (class_name != nullptr && class_name[1] == '_' && class_name[2] != '\0')
2148 return class_name + 2;
2153 // The case were the pointer_value is non-zero and points to a class defined
2154 // in this Mach-O file.
2155 r = get_pointer_64(pointer_value, offset, left, S, info);
2156 if (r == nullptr || left < sizeof(struct class64_t))
2159 memcpy(&c, r, sizeof(struct class64_t));
2160 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2164 r = get_pointer_64(c.data, offset, left, S, info);
2165 if (r == nullptr || left < sizeof(struct class_ro64_t))
2167 struct class_ro64_t cro;
2168 memcpy(&cro, r, sizeof(struct class_ro64_t));
2169 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2173 const char *name = get_pointer_64(cro.name, offset, left, S, info);
2177 // get_objc2_64bit_cfstring_name is used for disassembly and is passed a
2178 // pointer to a cfstring and returns its name or nullptr.
2179 const char *get_objc2_64bit_cfstring_name(uint64_t ReferenceValue,
2180 struct DisassembleInfo *info) {
2181 const char *r, *name;
2182 uint32_t offset, left;
2184 struct cfstring64_t cfs;
2185 uint64_t cfs_characters;
2187 r = get_pointer_64(ReferenceValue, offset, left, S, info);
2188 if (r == nullptr || left < sizeof(struct cfstring64_t))
2190 memcpy(&cfs, r, sizeof(struct cfstring64_t));
2191 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2193 if (cfs.characters == 0) {
2195 const char *symbol_name = get_symbol_64(
2196 offset + offsetof(struct cfstring64_t, characters), S, info, n_value);
2197 if (symbol_name == nullptr)
2199 cfs_characters = n_value;
2201 cfs_characters = cfs.characters;
2202 name = get_pointer_64(cfs_characters, offset, left, S, info);
2207 // get_objc2_64bit_selref() is used for disassembly and is passed a the address
2208 // of a pointer to an Objective-C selector reference when the pointer value is
2209 // zero as in a .o file and is likely to have a external relocation entry with
2210 // who's symbol's n_value is the real pointer to the selector name. If that is
2211 // the case the real pointer to the selector name is returned else 0 is
2213 uint64_t get_objc2_64bit_selref(uint64_t ReferenceValue,
2214 struct DisassembleInfo *info) {
2215 uint32_t offset, left;
2218 const char *r = get_pointer_64(ReferenceValue, offset, left, S, info);
2219 if (r == nullptr || left < sizeof(uint64_t))
2222 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
2223 if (symbol_name == nullptr)
2228 // GuessLiteralPointer returns a string which for the item in the Mach-O file
2229 // for the address passed in as ReferenceValue for printing as a comment with
2230 // the instruction and also returns the corresponding type of that item
2231 // indirectly through ReferenceType.
2233 // If ReferenceValue is an address of literal cstring then a pointer to the
2234 // cstring is returned and ReferenceType is set to
2235 // LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr .
2237 // If ReferenceValue is an address of an Objective-C CFString, Selector ref or
2238 // Class ref that name is returned and the ReferenceType is set accordingly.
2240 // Lastly, literals which are Symbol address in a literal pool are looked for
2241 // and if found the symbol name is returned and ReferenceType is set to
2242 // LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr .
2244 // If there is no item in the Mach-O file for the address passed in as
2245 // ReferenceValue nullptr is returned and ReferenceType is unchanged.
2246 const char *GuessLiteralPointer(uint64_t ReferenceValue, uint64_t ReferencePC,
2247 uint64_t *ReferenceType,
2248 struct DisassembleInfo *info) {
2249 // First see if there is an external relocation entry at the ReferencePC.
2250 uint64_t sect_addr = info->S.getAddress();
2251 uint64_t sect_offset = ReferencePC - sect_addr;
2252 bool reloc_found = false;
2254 MachO::any_relocation_info RE;
2255 bool isExtern = false;
2257 for (const RelocationRef &Reloc : info->S.relocations()) {
2258 uint64_t RelocOffset;
2259 Reloc.getOffset(RelocOffset);
2260 if (RelocOffset == sect_offset) {
2261 Rel = Reloc.getRawDataRefImpl();
2262 RE = info->O->getRelocation(Rel);
2263 if (info->O->isRelocationScattered(RE))
2265 isExtern = info->O->getPlainRelocationExternal(RE);
2267 symbol_iterator RelocSym = Reloc.getSymbol();
2274 // If there is an external relocation entry for a symbol in a section
2275 // then used that symbol's value for the value of the reference.
2276 if (reloc_found && isExtern) {
2277 if (info->O->getAnyRelocationPCRel(RE)) {
2278 unsigned Type = info->O->getAnyRelocationType(RE);
2279 if (Type == MachO::X86_64_RELOC_SIGNED) {
2280 Symbol.getAddress(ReferenceValue);
2285 // Look for literals such as Objective-C CFStrings refs, Selector refs,
2286 // Message refs and Class refs.
2287 bool classref, selref, msgref, cfstring;
2288 uint64_t pointer_value = GuessPointerPointer(ReferenceValue, info, classref,
2289 selref, msgref, cfstring);
2290 if (classref == true && pointer_value == 0) {
2291 // Note the ReferenceValue is a pointer into the __objc_classrefs section.
2292 // And the pointer_value in that section is typically zero as it will be
2293 // set by dyld as part of the "bind information".
2294 const char *name = get_dyld_bind_info_symbolname(ReferenceValue, info);
2295 if (name != nullptr) {
2296 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
2297 const char *class_name = strrchr(name, '$');
2298 if (class_name != nullptr && class_name[1] == '_' &&
2299 class_name[2] != '\0') {
2300 info->class_name = class_name + 2;
2306 if (classref == true) {
2307 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
2309 get_objc2_64bit_class_name(pointer_value, ReferenceValue, info);
2310 if (name != nullptr)
2311 info->class_name = name;
2313 name = "bad class ref";
2317 if (cfstring == true) {
2318 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_CFString_Ref;
2319 const char *name = get_objc2_64bit_cfstring_name(ReferenceValue, info);
2323 if (selref == true && pointer_value == 0)
2324 pointer_value = get_objc2_64bit_selref(ReferenceValue, info);
2326 if (pointer_value != 0)
2327 ReferenceValue = pointer_value;
2329 const char *name = GuessCstringPointer(ReferenceValue, info);
2331 if (pointer_value != 0 && selref == true) {
2332 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Selector_Ref;
2333 info->selector_name = name;
2334 } else if (pointer_value != 0 && msgref == true) {
2335 info->class_name = nullptr;
2336 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message_Ref;
2337 info->selector_name = name;
2339 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr;
2343 // Lastly look for an indirect symbol with this ReferenceValue which is in
2344 // a literal pool. If found return that symbol name.
2345 name = GuessIndirectSymbol(ReferenceValue, info);
2347 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr;
2354 // SymbolizerSymbolLookUp is the symbol lookup function passed when creating
2355 // the Symbolizer. It looks up the ReferenceValue using the info passed via the
2356 // pointer to the struct DisassembleInfo that was passed when MCSymbolizer
2357 // is created and returns the symbol name that matches the ReferenceValue or
2358 // nullptr if none. The ReferenceType is passed in for the IN type of
2359 // reference the instruction is making from the values in defined in the header
2360 // "llvm-c/Disassembler.h". On return the ReferenceType can set to a specific
2361 // Out type and the ReferenceName will also be set which is added as a comment
2362 // to the disassembled instruction.
2365 // If the symbol name is a C++ mangled name then the demangled name is
2366 // returned through ReferenceName and ReferenceType is set to
2367 // LLVMDisassembler_ReferenceType_DeMangled_Name .
2370 // When this is called to get a symbol name for a branch target then the
2371 // ReferenceType will be LLVMDisassembler_ReferenceType_In_Branch and then
2372 // SymbolValue will be looked for in the indirect symbol table to determine if
2373 // it is an address for a symbol stub. If so then the symbol name for that
2374 // stub is returned indirectly through ReferenceName and then ReferenceType is
2375 // set to LLVMDisassembler_ReferenceType_Out_SymbolStub.
2377 // When this is called with an value loaded via a PC relative load then
2378 // ReferenceType will be LLVMDisassembler_ReferenceType_In_PCrel_Load then the
2379 // SymbolValue is checked to be an address of literal pointer, symbol pointer,
2380 // or an Objective-C meta data reference. If so the output ReferenceType is
2381 // set to correspond to that as well as setting the ReferenceName.
2382 const char *SymbolizerSymbolLookUp(void *DisInfo, uint64_t ReferenceValue,
2383 uint64_t *ReferenceType,
2384 uint64_t ReferencePC,
2385 const char **ReferenceName) {
2386 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
2387 // If no verbose symbolic information is wanted then just return nullptr.
2388 if (info->verbose == false) {
2389 *ReferenceName = nullptr;
2390 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2394 const char *SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap);
2396 if (*ReferenceType == LLVMDisassembler_ReferenceType_In_Branch) {
2397 *ReferenceName = GuessIndirectSymbol(ReferenceValue, info);
2398 if (*ReferenceName != nullptr) {
2399 method_reference(info, ReferenceType, ReferenceName);
2400 if (*ReferenceType != LLVMDisassembler_ReferenceType_Out_Objc_Message)
2401 *ReferenceType = LLVMDisassembler_ReferenceType_Out_SymbolStub;
2404 if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
2405 if (info->demangled_name != nullptr)
2406 free(info->demangled_name);
2408 info->demangled_name =
2409 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
2410 if (info->demangled_name != nullptr) {
2411 *ReferenceName = info->demangled_name;
2412 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
2414 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2417 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2418 } else if (*ReferenceType == LLVMDisassembler_ReferenceType_In_PCrel_Load) {
2420 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2422 method_reference(info, ReferenceType, ReferenceName);
2424 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2425 // If this is arm64 and the reference is an adrp instruction save the
2426 // instruction, passed in ReferenceValue and the address of the instruction
2427 // for use later if we see and add immediate instruction.
2428 } else if (info->O->getArch() == Triple::aarch64 &&
2429 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADRP) {
2430 info->adrp_inst = ReferenceValue;
2431 info->adrp_addr = ReferencePC;
2432 SymbolName = nullptr;
2433 *ReferenceName = nullptr;
2434 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2435 // If this is arm64 and reference is an add immediate instruction and we
2437 // seen an adrp instruction just before it and the adrp's Xd register
2439 // this add's Xn register reconstruct the value being referenced and look to
2440 // see if it is a literal pointer. Note the add immediate instruction is
2441 // passed in ReferenceValue.
2442 } else if (info->O->getArch() == Triple::aarch64 &&
2443 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADDXri &&
2444 ReferencePC - 4 == info->adrp_addr &&
2445 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
2446 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
2447 uint32_t addxri_inst;
2448 uint64_t adrp_imm, addxri_imm;
2451 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
2452 if (info->adrp_inst & 0x0200000)
2453 adrp_imm |= 0xfffffffffc000000LL;
2455 addxri_inst = ReferenceValue;
2456 addxri_imm = (addxri_inst >> 10) & 0xfff;
2457 if (((addxri_inst >> 22) & 0x3) == 1)
2460 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
2461 (adrp_imm << 12) + addxri_imm;
2464 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2465 if (*ReferenceName == nullptr)
2466 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2467 // If this is arm64 and the reference is a load register instruction and we
2468 // have seen an adrp instruction just before it and the adrp's Xd register
2469 // matches this add's Xn register reconstruct the value being referenced and
2470 // look to see if it is a literal pointer. Note the load register
2471 // instruction is passed in ReferenceValue.
2472 } else if (info->O->getArch() == Triple::aarch64 &&
2473 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXui &&
2474 ReferencePC - 4 == info->adrp_addr &&
2475 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
2476 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
2477 uint32_t ldrxui_inst;
2478 uint64_t adrp_imm, ldrxui_imm;
2481 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
2482 if (info->adrp_inst & 0x0200000)
2483 adrp_imm |= 0xfffffffffc000000LL;
2485 ldrxui_inst = ReferenceValue;
2486 ldrxui_imm = (ldrxui_inst >> 10) & 0xfff;
2488 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
2489 (adrp_imm << 12) + (ldrxui_imm << 3);
2492 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2493 if (*ReferenceName == nullptr)
2494 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2496 // If this arm64 and is an load register (PC-relative) instruction the
2497 // ReferenceValue is the PC plus the immediate value.
2498 else if (info->O->getArch() == Triple::aarch64 &&
2499 (*ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXl ||
2500 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADR)) {
2502 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2503 if (*ReferenceName == nullptr)
2504 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2507 else if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
2508 if (info->demangled_name != nullptr)
2509 free(info->demangled_name);
2511 info->demangled_name =
2512 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
2513 if (info->demangled_name != nullptr) {
2514 *ReferenceName = info->demangled_name;
2515 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
2520 *ReferenceName = nullptr;
2521 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2527 /// \brief Emits the comments that are stored in the CommentStream.
2528 /// Each comment in the CommentStream must end with a newline.
2529 static void emitComments(raw_svector_ostream &CommentStream,
2530 SmallString<128> &CommentsToEmit,
2531 formatted_raw_ostream &FormattedOS,
2532 const MCAsmInfo &MAI) {
2533 // Flush the stream before taking its content.
2534 CommentStream.flush();
2535 StringRef Comments = CommentsToEmit.str();
2536 // Get the default information for printing a comment.
2537 const char *CommentBegin = MAI.getCommentString();
2538 unsigned CommentColumn = MAI.getCommentColumn();
2539 bool IsFirst = true;
2540 while (!Comments.empty()) {
2542 FormattedOS << '\n';
2543 // Emit a line of comments.
2544 FormattedOS.PadToColumn(CommentColumn);
2545 size_t Position = Comments.find('\n');
2546 FormattedOS << CommentBegin << ' ' << Comments.substr(0, Position);
2547 // Move after the newline character.
2548 Comments = Comments.substr(Position + 1);
2551 FormattedOS.flush();
2553 // Tell the comment stream that the vector changed underneath it.
2554 CommentsToEmit.clear();
2555 CommentStream.resync();
2558 static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF,
2559 StringRef DisSegName, StringRef DisSectName) {
2560 const char *McpuDefault = nullptr;
2561 const Target *ThumbTarget = nullptr;
2562 const Target *TheTarget = GetTarget(MachOOF, &McpuDefault, &ThumbTarget);
2564 // GetTarget prints out stuff.
2567 if (MCPU.empty() && McpuDefault)
2570 std::unique_ptr<const MCInstrInfo> InstrInfo(TheTarget->createMCInstrInfo());
2571 std::unique_ptr<const MCInstrInfo> ThumbInstrInfo;
2573 ThumbInstrInfo.reset(ThumbTarget->createMCInstrInfo());
2575 // Package up features to be passed to target/subtarget
2576 std::string FeaturesStr;
2577 if (MAttrs.size()) {
2578 SubtargetFeatures Features;
2579 for (unsigned i = 0; i != MAttrs.size(); ++i)
2580 Features.AddFeature(MAttrs[i]);
2581 FeaturesStr = Features.getString();
2584 // Set up disassembler.
2585 std::unique_ptr<const MCRegisterInfo> MRI(
2586 TheTarget->createMCRegInfo(TripleName));
2587 std::unique_ptr<const MCAsmInfo> AsmInfo(
2588 TheTarget->createMCAsmInfo(*MRI, TripleName));
2589 std::unique_ptr<const MCSubtargetInfo> STI(
2590 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr));
2591 MCContext Ctx(AsmInfo.get(), MRI.get(), nullptr);
2592 std::unique_ptr<MCDisassembler> DisAsm(
2593 TheTarget->createMCDisassembler(*STI, Ctx));
2594 std::unique_ptr<MCSymbolizer> Symbolizer;
2595 struct DisassembleInfo SymbolizerInfo;
2596 std::unique_ptr<MCRelocationInfo> RelInfo(
2597 TheTarget->createMCRelocationInfo(TripleName, Ctx));
2599 Symbolizer.reset(TheTarget->createMCSymbolizer(
2600 TripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
2601 &SymbolizerInfo, &Ctx, std::move(RelInfo)));
2602 DisAsm->setSymbolizer(std::move(Symbolizer));
2604 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
2605 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
2606 AsmPrinterVariant, *AsmInfo, *InstrInfo, *MRI, *STI));
2607 // Set the display preference for hex vs. decimal immediates.
2608 IP->setPrintImmHex(PrintImmHex);
2609 // Comment stream and backing vector.
2610 SmallString<128> CommentsToEmit;
2611 raw_svector_ostream CommentStream(CommentsToEmit);
2612 // FIXME: Setting the CommentStream in the InstPrinter is problematic in that
2613 // if it is done then arm64 comments for string literals don't get printed
2614 // and some constant get printed instead and not setting it causes intel
2615 // (32-bit and 64-bit) comments printed with different spacing before the
2616 // comment causing different diffs with the 'C' disassembler library API.
2617 // IP->setCommentStream(CommentStream);
2619 if (!AsmInfo || !STI || !DisAsm || !IP) {
2620 errs() << "error: couldn't initialize disassembler for target "
2621 << TripleName << '\n';
2625 // Set up thumb disassembler.
2626 std::unique_ptr<const MCRegisterInfo> ThumbMRI;
2627 std::unique_ptr<const MCAsmInfo> ThumbAsmInfo;
2628 std::unique_ptr<const MCSubtargetInfo> ThumbSTI;
2629 std::unique_ptr<MCDisassembler> ThumbDisAsm;
2630 std::unique_ptr<MCInstPrinter> ThumbIP;
2631 std::unique_ptr<MCContext> ThumbCtx;
2632 std::unique_ptr<MCSymbolizer> ThumbSymbolizer;
2633 struct DisassembleInfo ThumbSymbolizerInfo;
2634 std::unique_ptr<MCRelocationInfo> ThumbRelInfo;
2636 ThumbMRI.reset(ThumbTarget->createMCRegInfo(ThumbTripleName));
2638 ThumbTarget->createMCAsmInfo(*ThumbMRI, ThumbTripleName));
2640 ThumbTarget->createMCSubtargetInfo(ThumbTripleName, MCPU, FeaturesStr));
2641 ThumbCtx.reset(new MCContext(ThumbAsmInfo.get(), ThumbMRI.get(), nullptr));
2642 ThumbDisAsm.reset(ThumbTarget->createMCDisassembler(*ThumbSTI, *ThumbCtx));
2643 MCContext *PtrThumbCtx = ThumbCtx.get();
2645 ThumbTarget->createMCRelocationInfo(ThumbTripleName, *PtrThumbCtx));
2647 ThumbSymbolizer.reset(ThumbTarget->createMCSymbolizer(
2648 ThumbTripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
2649 &ThumbSymbolizerInfo, PtrThumbCtx, std::move(ThumbRelInfo)));
2650 ThumbDisAsm->setSymbolizer(std::move(ThumbSymbolizer));
2652 int ThumbAsmPrinterVariant = ThumbAsmInfo->getAssemblerDialect();
2653 ThumbIP.reset(ThumbTarget->createMCInstPrinter(
2654 ThumbAsmPrinterVariant, *ThumbAsmInfo, *ThumbInstrInfo, *ThumbMRI,
2656 // Set the display preference for hex vs. decimal immediates.
2657 ThumbIP->setPrintImmHex(PrintImmHex);
2660 if (ThumbTarget && (!ThumbAsmInfo || !ThumbSTI || !ThumbDisAsm || !ThumbIP)) {
2661 errs() << "error: couldn't initialize disassembler for target "
2662 << ThumbTripleName << '\n';
2666 MachO::mach_header Header = MachOOF->getHeader();
2668 // FIXME: Using the -cfg command line option, this code used to be able to
2669 // annotate relocations with the referenced symbol's name, and if this was
2670 // inside a __[cf]string section, the data it points to. This is now replaced
2671 // by the upcoming MCSymbolizer, which needs the appropriate setup done above.
2672 std::vector<SectionRef> Sections;
2673 std::vector<SymbolRef> Symbols;
2674 SmallVector<uint64_t, 8> FoundFns;
2675 uint64_t BaseSegmentAddress;
2677 getSectionsAndSymbols(Header, MachOOF, Sections, Symbols, FoundFns,
2678 BaseSegmentAddress);
2680 // Sort the symbols by address, just in case they didn't come in that way.
2681 std::sort(Symbols.begin(), Symbols.end(), SymbolSorter());
2683 // Build a data in code table that is sorted on by the address of each entry.
2684 uint64_t BaseAddress = 0;
2685 if (Header.filetype == MachO::MH_OBJECT)
2686 BaseAddress = Sections[0].getAddress();
2688 BaseAddress = BaseSegmentAddress;
2690 for (dice_iterator DI = MachOOF->begin_dices(), DE = MachOOF->end_dices();
2693 DI->getOffset(Offset);
2694 Dices.push_back(std::make_pair(BaseAddress + Offset, *DI));
2696 array_pod_sort(Dices.begin(), Dices.end());
2699 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
2701 raw_ostream &DebugOut = nulls();
2704 std::unique_ptr<DIContext> diContext;
2705 ObjectFile *DbgObj = MachOOF;
2706 // Try to find debug info and set up the DIContext for it.
2708 // A separate DSym file path was specified, parse it as a macho file,
2709 // get the sections and supply it to the section name parsing machinery.
2710 if (!DSYMFile.empty()) {
2711 ErrorOr<std::unique_ptr<MemoryBuffer>> BufOrErr =
2712 MemoryBuffer::getFileOrSTDIN(DSYMFile);
2713 if (std::error_code EC = BufOrErr.getError()) {
2714 errs() << "llvm-objdump: " << Filename << ": " << EC.message() << '\n';
2718 ObjectFile::createMachOObjectFile(BufOrErr.get()->getMemBufferRef())
2723 // Setup the DIContext
2724 diContext.reset(DIContext::getDWARFContext(*DbgObj));
2727 if (DumpSections.size() == 0)
2728 outs() << "(" << DisSegName << "," << DisSectName << ") section\n";
2730 for (unsigned SectIdx = 0; SectIdx != Sections.size(); SectIdx++) {
2732 if (Sections[SectIdx].getName(SectName) || SectName != DisSectName)
2735 DataRefImpl DR = Sections[SectIdx].getRawDataRefImpl();
2737 StringRef SegmentName = MachOOF->getSectionFinalSegmentName(DR);
2738 if (SegmentName != DisSegName)
2742 Sections[SectIdx].getContents(BytesStr);
2743 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()),
2745 uint64_t SectAddress = Sections[SectIdx].getAddress();
2747 bool symbolTableWorked = false;
2749 // Parse relocations.
2750 std::vector<std::pair<uint64_t, SymbolRef>> Relocs;
2751 for (const RelocationRef &Reloc : Sections[SectIdx].relocations()) {
2752 uint64_t RelocOffset;
2753 Reloc.getOffset(RelocOffset);
2754 uint64_t SectionAddress = Sections[SectIdx].getAddress();
2755 RelocOffset -= SectionAddress;
2757 symbol_iterator RelocSym = Reloc.getSymbol();
2759 Relocs.push_back(std::make_pair(RelocOffset, *RelocSym));
2761 array_pod_sort(Relocs.begin(), Relocs.end());
2763 // Create a map of symbol addresses to symbol names for use by
2764 // the SymbolizerSymbolLookUp() routine.
2765 SymbolAddressMap AddrMap;
2766 for (const SymbolRef &Symbol : MachOOF->symbols()) {
2769 if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data ||
2770 ST == SymbolRef::ST_Other) {
2772 Symbol.getAddress(Address);
2774 Symbol.getName(SymName);
2775 AddrMap[Address] = SymName;
2778 // Set up the block of info used by the Symbolizer call backs.
2779 SymbolizerInfo.verbose = true;
2780 SymbolizerInfo.O = MachOOF;
2781 SymbolizerInfo.S = Sections[SectIdx];
2782 SymbolizerInfo.AddrMap = &AddrMap;
2783 SymbolizerInfo.Sections = &Sections;
2784 SymbolizerInfo.class_name = nullptr;
2785 SymbolizerInfo.selector_name = nullptr;
2786 SymbolizerInfo.method = nullptr;
2787 SymbolizerInfo.demangled_name = nullptr;
2788 SymbolizerInfo.bindtable = nullptr;
2789 SymbolizerInfo.adrp_addr = 0;
2790 SymbolizerInfo.adrp_inst = 0;
2791 // Same for the ThumbSymbolizer
2792 ThumbSymbolizerInfo.verbose = true;
2793 ThumbSymbolizerInfo.O = MachOOF;
2794 ThumbSymbolizerInfo.S = Sections[SectIdx];
2795 ThumbSymbolizerInfo.AddrMap = &AddrMap;
2796 ThumbSymbolizerInfo.Sections = &Sections;
2797 ThumbSymbolizerInfo.class_name = nullptr;
2798 ThumbSymbolizerInfo.selector_name = nullptr;
2799 ThumbSymbolizerInfo.method = nullptr;
2800 ThumbSymbolizerInfo.demangled_name = nullptr;
2801 ThumbSymbolizerInfo.bindtable = nullptr;
2802 ThumbSymbolizerInfo.adrp_addr = 0;
2803 ThumbSymbolizerInfo.adrp_inst = 0;
2805 // Disassemble symbol by symbol.
2806 for (unsigned SymIdx = 0; SymIdx != Symbols.size(); SymIdx++) {
2808 Symbols[SymIdx].getName(SymName);
2811 Symbols[SymIdx].getType(ST);
2812 if (ST != SymbolRef::ST_Function)
2815 // Make sure the symbol is defined in this section.
2816 bool containsSym = Sections[SectIdx].containsSymbol(Symbols[SymIdx]);
2820 // Start at the address of the symbol relative to the section's address.
2822 uint64_t SectionAddress = Sections[SectIdx].getAddress();
2823 Symbols[SymIdx].getAddress(Start);
2824 Start -= SectionAddress;
2826 // Stop disassembling either at the beginning of the next symbol or at
2827 // the end of the section.
2828 bool containsNextSym = false;
2829 uint64_t NextSym = 0;
2830 uint64_t NextSymIdx = SymIdx + 1;
2831 while (Symbols.size() > NextSymIdx) {
2832 SymbolRef::Type NextSymType;
2833 Symbols[NextSymIdx].getType(NextSymType);
2834 if (NextSymType == SymbolRef::ST_Function) {
2836 Sections[SectIdx].containsSymbol(Symbols[NextSymIdx]);
2837 Symbols[NextSymIdx].getAddress(NextSym);
2838 NextSym -= SectionAddress;
2844 uint64_t SectSize = Sections[SectIdx].getSize();
2845 uint64_t End = containsNextSym ? NextSym : SectSize;
2848 symbolTableWorked = true;
2850 DataRefImpl Symb = Symbols[SymIdx].getRawDataRefImpl();
2852 (MachOOF->getSymbolFlags(Symb) & SymbolRef::SF_Thumb) && ThumbTarget;
2854 outs() << SymName << ":\n";
2855 DILineInfo lastLine;
2856 for (uint64_t Index = Start; Index < End; Index += Size) {
2859 uint64_t PC = SectAddress + Index;
2860 if (FullLeadingAddr) {
2861 if (MachOOF->is64Bit())
2862 outs() << format("%016" PRIx64, PC);
2864 outs() << format("%08" PRIx64, PC);
2866 outs() << format("%8" PRIx64 ":", PC);
2871 // Check the data in code table here to see if this is data not an
2872 // instruction to be disassembled.
2874 Dice.push_back(std::make_pair(PC, DiceRef()));
2875 dice_table_iterator DTI =
2876 std::search(Dices.begin(), Dices.end(), Dice.begin(), Dice.end(),
2877 compareDiceTableEntries);
2878 if (DTI != Dices.end()) {
2880 DTI->second.getLength(Length);
2882 DTI->second.getKind(Kind);
2883 Size = DumpDataInCode(reinterpret_cast<const char *>(Bytes.data()) +
2886 if ((Kind == MachO::DICE_KIND_JUMP_TABLE8) &&
2887 (PC == (DTI->first + Length - 1)) && (Length & 1))
2892 SmallVector<char, 64> AnnotationsBytes;
2893 raw_svector_ostream Annotations(AnnotationsBytes);
2897 gotInst = ThumbDisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
2898 PC, DebugOut, Annotations);
2900 gotInst = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), PC,
2901 DebugOut, Annotations);
2903 if (!NoShowRawInsn) {
2904 DumpBytes(StringRef(
2905 reinterpret_cast<const char *>(Bytes.data()) + Index, Size));
2907 formatted_raw_ostream FormattedOS(outs());
2908 Annotations.flush();
2909 StringRef AnnotationsStr = Annotations.str();
2911 ThumbIP->printInst(&Inst, FormattedOS, AnnotationsStr);
2913 IP->printInst(&Inst, FormattedOS, AnnotationsStr);
2914 emitComments(CommentStream, CommentsToEmit, FormattedOS, *AsmInfo);
2916 // Print debug info.
2918 DILineInfo dli = diContext->getLineInfoForAddress(PC);
2919 // Print valid line info if it changed.
2920 if (dli != lastLine && dli.Line != 0)
2921 outs() << "\t## " << dli.FileName << ':' << dli.Line << ':'
2927 unsigned int Arch = MachOOF->getArch();
2928 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
2929 outs() << format("\t.byte 0x%02x #bad opcode\n",
2930 *(Bytes.data() + Index) & 0xff);
2931 Size = 1; // skip exactly one illegible byte and move on.
2932 } else if (Arch == Triple::aarch64) {
2933 uint32_t opcode = (*(Bytes.data() + Index) & 0xff) |
2934 (*(Bytes.data() + Index + 1) & 0xff) << 8 |
2935 (*(Bytes.data() + Index + 2) & 0xff) << 16 |
2936 (*(Bytes.data() + Index + 3) & 0xff) << 24;
2937 outs() << format("\t.long\t0x%08x\n", opcode);
2940 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
2942 Size = 1; // skip illegible bytes
2947 if (!symbolTableWorked) {
2948 // Reading the symbol table didn't work, disassemble the whole section.
2949 uint64_t SectAddress = Sections[SectIdx].getAddress();
2950 uint64_t SectSize = Sections[SectIdx].getSize();
2952 for (uint64_t Index = 0; Index < SectSize; Index += InstSize) {
2955 uint64_t PC = SectAddress + Index;
2956 if (DisAsm->getInstruction(Inst, InstSize, Bytes.slice(Index), PC,
2957 DebugOut, nulls())) {
2958 if (FullLeadingAddr) {
2959 if (MachOOF->is64Bit())
2960 outs() << format("%016" PRIx64, PC);
2962 outs() << format("%08" PRIx64, PC);
2964 outs() << format("%8" PRIx64 ":", PC);
2966 if (!NoShowRawInsn) {
2969 StringRef(reinterpret_cast<const char *>(Bytes.data()) + Index,
2972 IP->printInst(&Inst, outs(), "");
2975 unsigned int Arch = MachOOF->getArch();
2976 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
2977 outs() << format("\t.byte 0x%02x #bad opcode\n",
2978 *(Bytes.data() + Index) & 0xff);
2979 InstSize = 1; // skip exactly one illegible byte and move on.
2981 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
2983 InstSize = 1; // skip illegible bytes
2988 // The TripleName's need to be reset if we are called again for a different
2991 ThumbTripleName = "";
2993 if (SymbolizerInfo.method != nullptr)
2994 free(SymbolizerInfo.method);
2995 if (SymbolizerInfo.demangled_name != nullptr)
2996 free(SymbolizerInfo.demangled_name);
2997 if (SymbolizerInfo.bindtable != nullptr)
2998 delete SymbolizerInfo.bindtable;
2999 if (ThumbSymbolizerInfo.method != nullptr)
3000 free(ThumbSymbolizerInfo.method);
3001 if (ThumbSymbolizerInfo.demangled_name != nullptr)
3002 free(ThumbSymbolizerInfo.demangled_name);
3003 if (ThumbSymbolizerInfo.bindtable != nullptr)
3004 delete ThumbSymbolizerInfo.bindtable;
3008 //===----------------------------------------------------------------------===//
3009 // __compact_unwind section dumping
3010 //===----------------------------------------------------------------------===//
3014 template <typename T> static uint64_t readNext(const char *&Buf) {
3015 using llvm::support::little;
3016 using llvm::support::unaligned;
3018 uint64_t Val = support::endian::read<T, little, unaligned>(Buf);
3023 struct CompactUnwindEntry {
3024 uint32_t OffsetInSection;
3026 uint64_t FunctionAddr;
3028 uint32_t CompactEncoding;
3029 uint64_t PersonalityAddr;
3032 RelocationRef FunctionReloc;
3033 RelocationRef PersonalityReloc;
3034 RelocationRef LSDAReloc;
3036 CompactUnwindEntry(StringRef Contents, unsigned Offset, bool Is64)
3037 : OffsetInSection(Offset) {
3039 read<uint64_t>(Contents.data() + Offset);
3041 read<uint32_t>(Contents.data() + Offset);
3045 template <typename UIntPtr> void read(const char *Buf) {
3046 FunctionAddr = readNext<UIntPtr>(Buf);
3047 Length = readNext<uint32_t>(Buf);
3048 CompactEncoding = readNext<uint32_t>(Buf);
3049 PersonalityAddr = readNext<UIntPtr>(Buf);
3050 LSDAAddr = readNext<UIntPtr>(Buf);
3055 /// Given a relocation from __compact_unwind, consisting of the RelocationRef
3056 /// and data being relocated, determine the best base Name and Addend to use for
3057 /// display purposes.
3059 /// 1. An Extern relocation will directly reference a symbol (and the data is
3060 /// then already an addend), so use that.
3061 /// 2. Otherwise the data is an offset in the object file's layout; try to find
3062 // a symbol before it in the same section, and use the offset from there.
3063 /// 3. Finally, if all that fails, fall back to an offset from the start of the
3064 /// referenced section.
3065 static void findUnwindRelocNameAddend(const MachOObjectFile *Obj,
3066 std::map<uint64_t, SymbolRef> &Symbols,
3067 const RelocationRef &Reloc, uint64_t Addr,
3068 StringRef &Name, uint64_t &Addend) {
3069 if (Reloc.getSymbol() != Obj->symbol_end()) {
3070 Reloc.getSymbol()->getName(Name);
3075 auto RE = Obj->getRelocation(Reloc.getRawDataRefImpl());
3076 SectionRef RelocSection = Obj->getRelocationSection(RE);
3078 uint64_t SectionAddr = RelocSection.getAddress();
3080 auto Sym = Symbols.upper_bound(Addr);
3081 if (Sym == Symbols.begin()) {
3082 // The first symbol in the object is after this reference, the best we can
3083 // do is section-relative notation.
3084 RelocSection.getName(Name);
3085 Addend = Addr - SectionAddr;
3089 // Go back one so that SymbolAddress <= Addr.
3092 section_iterator SymSection = Obj->section_end();
3093 Sym->second.getSection(SymSection);
3094 if (RelocSection == *SymSection) {
3095 // There's a valid symbol in the same section before this reference.
3096 Sym->second.getName(Name);
3097 Addend = Addr - Sym->first;
3101 // There is a symbol before this reference, but it's in a different
3102 // section. Probably not helpful to mention it, so use the section name.
3103 RelocSection.getName(Name);
3104 Addend = Addr - SectionAddr;
3107 static void printUnwindRelocDest(const MachOObjectFile *Obj,
3108 std::map<uint64_t, SymbolRef> &Symbols,
3109 const RelocationRef &Reloc, uint64_t Addr) {
3113 if (!Reloc.getObjectFile())
3116 findUnwindRelocNameAddend(Obj, Symbols, Reloc, Addr, Name, Addend);
3120 outs() << " + " << format("0x%" PRIx64, Addend);
3124 printMachOCompactUnwindSection(const MachOObjectFile *Obj,
3125 std::map<uint64_t, SymbolRef> &Symbols,
3126 const SectionRef &CompactUnwind) {
3128 assert(Obj->isLittleEndian() &&
3129 "There should not be a big-endian .o with __compact_unwind");
3131 bool Is64 = Obj->is64Bit();
3132 uint32_t PointerSize = Is64 ? sizeof(uint64_t) : sizeof(uint32_t);
3133 uint32_t EntrySize = 3 * PointerSize + 2 * sizeof(uint32_t);
3136 CompactUnwind.getContents(Contents);
3138 SmallVector<CompactUnwindEntry, 4> CompactUnwinds;
3140 // First populate the initial raw offsets, encodings and so on from the entry.
3141 for (unsigned Offset = 0; Offset < Contents.size(); Offset += EntrySize) {
3142 CompactUnwindEntry Entry(Contents.data(), Offset, Is64);
3143 CompactUnwinds.push_back(Entry);
3146 // Next we need to look at the relocations to find out what objects are
3147 // actually being referred to.
3148 for (const RelocationRef &Reloc : CompactUnwind.relocations()) {
3149 uint64_t RelocAddress;
3150 Reloc.getOffset(RelocAddress);
3152 uint32_t EntryIdx = RelocAddress / EntrySize;
3153 uint32_t OffsetInEntry = RelocAddress - EntryIdx * EntrySize;
3154 CompactUnwindEntry &Entry = CompactUnwinds[EntryIdx];
3156 if (OffsetInEntry == 0)
3157 Entry.FunctionReloc = Reloc;
3158 else if (OffsetInEntry == PointerSize + 2 * sizeof(uint32_t))
3159 Entry.PersonalityReloc = Reloc;
3160 else if (OffsetInEntry == 2 * PointerSize + 2 * sizeof(uint32_t))
3161 Entry.LSDAReloc = Reloc;
3163 llvm_unreachable("Unexpected relocation in __compact_unwind section");
3166 // Finally, we're ready to print the data we've gathered.
3167 outs() << "Contents of __compact_unwind section:\n";
3168 for (auto &Entry : CompactUnwinds) {
3169 outs() << " Entry at offset "
3170 << format("0x%" PRIx32, Entry.OffsetInSection) << ":\n";
3172 // 1. Start of the region this entry applies to.
3173 outs() << " start: " << format("0x%" PRIx64,
3174 Entry.FunctionAddr) << ' ';
3175 printUnwindRelocDest(Obj, Symbols, Entry.FunctionReloc, Entry.FunctionAddr);
3178 // 2. Length of the region this entry applies to.
3179 outs() << " length: " << format("0x%" PRIx32, Entry.Length)
3181 // 3. The 32-bit compact encoding.
3182 outs() << " compact encoding: "
3183 << format("0x%08" PRIx32, Entry.CompactEncoding) << '\n';
3185 // 4. The personality function, if present.
3186 if (Entry.PersonalityReloc.getObjectFile()) {
3187 outs() << " personality function: "
3188 << format("0x%" PRIx64, Entry.PersonalityAddr) << ' ';
3189 printUnwindRelocDest(Obj, Symbols, Entry.PersonalityReloc,
3190 Entry.PersonalityAddr);
3194 // 5. This entry's language-specific data area.
3195 if (Entry.LSDAReloc.getObjectFile()) {
3196 outs() << " LSDA: " << format("0x%" PRIx64,
3197 Entry.LSDAAddr) << ' ';
3198 printUnwindRelocDest(Obj, Symbols, Entry.LSDAReloc, Entry.LSDAAddr);
3204 //===----------------------------------------------------------------------===//
3205 // __unwind_info section dumping
3206 //===----------------------------------------------------------------------===//
3208 static void printRegularSecondLevelUnwindPage(const char *PageStart) {
3209 const char *Pos = PageStart;
3210 uint32_t Kind = readNext<uint32_t>(Pos);
3212 assert(Kind == 2 && "kind for a regular 2nd level index should be 2");
3214 uint16_t EntriesStart = readNext<uint16_t>(Pos);
3215 uint16_t NumEntries = readNext<uint16_t>(Pos);
3217 Pos = PageStart + EntriesStart;
3218 for (unsigned i = 0; i < NumEntries; ++i) {
3219 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
3220 uint32_t Encoding = readNext<uint32_t>(Pos);
3222 outs() << " [" << i << "]: "
3223 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
3225 << "encoding=" << format("0x%08" PRIx32, Encoding) << '\n';
3229 static void printCompressedSecondLevelUnwindPage(
3230 const char *PageStart, uint32_t FunctionBase,
3231 const SmallVectorImpl<uint32_t> &CommonEncodings) {
3232 const char *Pos = PageStart;
3233 uint32_t Kind = readNext<uint32_t>(Pos);
3235 assert(Kind == 3 && "kind for a compressed 2nd level index should be 3");
3237 uint16_t EntriesStart = readNext<uint16_t>(Pos);
3238 uint16_t NumEntries = readNext<uint16_t>(Pos);
3240 uint16_t EncodingsStart = readNext<uint16_t>(Pos);
3241 readNext<uint16_t>(Pos);
3242 const auto *PageEncodings = reinterpret_cast<const support::ulittle32_t *>(
3243 PageStart + EncodingsStart);
3245 Pos = PageStart + EntriesStart;
3246 for (unsigned i = 0; i < NumEntries; ++i) {
3247 uint32_t Entry = readNext<uint32_t>(Pos);
3248 uint32_t FunctionOffset = FunctionBase + (Entry & 0xffffff);
3249 uint32_t EncodingIdx = Entry >> 24;
3252 if (EncodingIdx < CommonEncodings.size())
3253 Encoding = CommonEncodings[EncodingIdx];
3255 Encoding = PageEncodings[EncodingIdx - CommonEncodings.size()];
3257 outs() << " [" << i << "]: "
3258 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
3260 << "encoding[" << EncodingIdx
3261 << "]=" << format("0x%08" PRIx32, Encoding) << '\n';
3265 static void printMachOUnwindInfoSection(const MachOObjectFile *Obj,
3266 std::map<uint64_t, SymbolRef> &Symbols,
3267 const SectionRef &UnwindInfo) {
3269 assert(Obj->isLittleEndian() &&
3270 "There should not be a big-endian .o with __unwind_info");
3272 outs() << "Contents of __unwind_info section:\n";
3275 UnwindInfo.getContents(Contents);
3276 const char *Pos = Contents.data();
3278 //===----------------------------------
3280 //===----------------------------------
3282 uint32_t Version = readNext<uint32_t>(Pos);
3283 outs() << " Version: "
3284 << format("0x%" PRIx32, Version) << '\n';
3285 assert(Version == 1 && "only understand version 1");
3287 uint32_t CommonEncodingsStart = readNext<uint32_t>(Pos);
3288 outs() << " Common encodings array section offset: "
3289 << format("0x%" PRIx32, CommonEncodingsStart) << '\n';
3290 uint32_t NumCommonEncodings = readNext<uint32_t>(Pos);
3291 outs() << " Number of common encodings in array: "
3292 << format("0x%" PRIx32, NumCommonEncodings) << '\n';
3294 uint32_t PersonalitiesStart = readNext<uint32_t>(Pos);
3295 outs() << " Personality function array section offset: "
3296 << format("0x%" PRIx32, PersonalitiesStart) << '\n';
3297 uint32_t NumPersonalities = readNext<uint32_t>(Pos);
3298 outs() << " Number of personality functions in array: "
3299 << format("0x%" PRIx32, NumPersonalities) << '\n';
3301 uint32_t IndicesStart = readNext<uint32_t>(Pos);
3302 outs() << " Index array section offset: "
3303 << format("0x%" PRIx32, IndicesStart) << '\n';
3304 uint32_t NumIndices = readNext<uint32_t>(Pos);
3305 outs() << " Number of indices in array: "
3306 << format("0x%" PRIx32, NumIndices) << '\n';
3308 //===----------------------------------
3309 // A shared list of common encodings
3310 //===----------------------------------
3312 // These occupy indices in the range [0, N] whenever an encoding is referenced
3313 // from a compressed 2nd level index table. In practice the linker only
3314 // creates ~128 of these, so that indices are available to embed encodings in
3315 // the 2nd level index.
3317 SmallVector<uint32_t, 64> CommonEncodings;
3318 outs() << " Common encodings: (count = " << NumCommonEncodings << ")\n";
3319 Pos = Contents.data() + CommonEncodingsStart;
3320 for (unsigned i = 0; i < NumCommonEncodings; ++i) {
3321 uint32_t Encoding = readNext<uint32_t>(Pos);
3322 CommonEncodings.push_back(Encoding);
3324 outs() << " encoding[" << i << "]: " << format("0x%08" PRIx32, Encoding)
3328 //===----------------------------------
3329 // Personality functions used in this executable
3330 //===----------------------------------
3332 // There should be only a handful of these (one per source language,
3333 // roughly). Particularly since they only get 2 bits in the compact encoding.
3335 outs() << " Personality functions: (count = " << NumPersonalities << ")\n";
3336 Pos = Contents.data() + PersonalitiesStart;
3337 for (unsigned i = 0; i < NumPersonalities; ++i) {
3338 uint32_t PersonalityFn = readNext<uint32_t>(Pos);
3339 outs() << " personality[" << i + 1
3340 << "]: " << format("0x%08" PRIx32, PersonalityFn) << '\n';
3343 //===----------------------------------
3344 // The level 1 index entries
3345 //===----------------------------------
3347 // These specify an approximate place to start searching for the more detailed
3348 // information, sorted by PC.
3351 uint32_t FunctionOffset;
3352 uint32_t SecondLevelPageStart;
3356 SmallVector<IndexEntry, 4> IndexEntries;
3358 outs() << " Top level indices: (count = " << NumIndices << ")\n";
3359 Pos = Contents.data() + IndicesStart;
3360 for (unsigned i = 0; i < NumIndices; ++i) {
3363 Entry.FunctionOffset = readNext<uint32_t>(Pos);
3364 Entry.SecondLevelPageStart = readNext<uint32_t>(Pos);
3365 Entry.LSDAStart = readNext<uint32_t>(Pos);
3366 IndexEntries.push_back(Entry);
3368 outs() << " [" << i << "]: "
3369 << "function offset=" << format("0x%08" PRIx32, Entry.FunctionOffset)
3371 << "2nd level page offset="
3372 << format("0x%08" PRIx32, Entry.SecondLevelPageStart) << ", "
3373 << "LSDA offset=" << format("0x%08" PRIx32, Entry.LSDAStart) << '\n';
3376 //===----------------------------------
3377 // Next come the LSDA tables
3378 //===----------------------------------
3380 // The LSDA layout is rather implicit: it's a contiguous array of entries from
3381 // the first top-level index's LSDAOffset to the last (sentinel).
3383 outs() << " LSDA descriptors:\n";
3384 Pos = Contents.data() + IndexEntries[0].LSDAStart;
3385 int NumLSDAs = (IndexEntries.back().LSDAStart - IndexEntries[0].LSDAStart) /
3386 (2 * sizeof(uint32_t));
3387 for (int i = 0; i < NumLSDAs; ++i) {
3388 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
3389 uint32_t LSDAOffset = readNext<uint32_t>(Pos);
3390 outs() << " [" << i << "]: "
3391 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
3393 << "LSDA offset=" << format("0x%08" PRIx32, LSDAOffset) << '\n';
3396 //===----------------------------------
3397 // Finally, the 2nd level indices
3398 //===----------------------------------
3400 // Generally these are 4K in size, and have 2 possible forms:
3401 // + Regular stores up to 511 entries with disparate encodings
3402 // + Compressed stores up to 1021 entries if few enough compact encoding
3404 outs() << " Second level indices:\n";
3405 for (unsigned i = 0; i < IndexEntries.size() - 1; ++i) {
3406 // The final sentinel top-level index has no associated 2nd level page
3407 if (IndexEntries[i].SecondLevelPageStart == 0)
3410 outs() << " Second level index[" << i << "]: "
3411 << "offset in section="
3412 << format("0x%08" PRIx32, IndexEntries[i].SecondLevelPageStart)
3414 << "base function offset="
3415 << format("0x%08" PRIx32, IndexEntries[i].FunctionOffset) << '\n';
3417 Pos = Contents.data() + IndexEntries[i].SecondLevelPageStart;
3418 uint32_t Kind = *reinterpret_cast<const support::ulittle32_t *>(Pos);
3420 printRegularSecondLevelUnwindPage(Pos);
3422 printCompressedSecondLevelUnwindPage(Pos, IndexEntries[i].FunctionOffset,
3425 llvm_unreachable("Do not know how to print this kind of 2nd level page");
3429 void llvm::printMachOUnwindInfo(const MachOObjectFile *Obj) {
3430 std::map<uint64_t, SymbolRef> Symbols;
3431 for (const SymbolRef &SymRef : Obj->symbols()) {
3432 // Discard any undefined or absolute symbols. They're not going to take part
3433 // in the convenience lookup for unwind info and just take up resources.
3434 section_iterator Section = Obj->section_end();
3435 SymRef.getSection(Section);
3436 if (Section == Obj->section_end())
3440 SymRef.getAddress(Addr);
3441 Symbols.insert(std::make_pair(Addr, SymRef));
3444 for (const SectionRef &Section : Obj->sections()) {
3446 Section.getName(SectName);
3447 if (SectName == "__compact_unwind")
3448 printMachOCompactUnwindSection(Obj, Symbols, Section);
3449 else if (SectName == "__unwind_info")
3450 printMachOUnwindInfoSection(Obj, Symbols, Section);
3451 else if (SectName == "__eh_frame")
3452 outs() << "llvm-objdump: warning: unhandled __eh_frame section\n";
3456 static void PrintMachHeader(uint32_t magic, uint32_t cputype,
3457 uint32_t cpusubtype, uint32_t filetype,
3458 uint32_t ncmds, uint32_t sizeofcmds, uint32_t flags,
3460 outs() << "Mach header\n";
3461 outs() << " magic cputype cpusubtype caps filetype ncmds "
3462 "sizeofcmds flags\n";
3464 if (magic == MachO::MH_MAGIC)
3465 outs() << " MH_MAGIC";
3466 else if (magic == MachO::MH_MAGIC_64)
3467 outs() << "MH_MAGIC_64";
3469 outs() << format(" 0x%08" PRIx32, magic);
3471 case MachO::CPU_TYPE_I386:
3473 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3474 case MachO::CPU_SUBTYPE_I386_ALL:
3478 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3482 case MachO::CPU_TYPE_X86_64:
3483 outs() << " X86_64";
3484 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3485 case MachO::CPU_SUBTYPE_X86_64_ALL:
3488 case MachO::CPU_SUBTYPE_X86_64_H:
3489 outs() << " Haswell";
3492 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3496 case MachO::CPU_TYPE_ARM:
3498 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3499 case MachO::CPU_SUBTYPE_ARM_ALL:
3502 case MachO::CPU_SUBTYPE_ARM_V4T:
3505 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
3508 case MachO::CPU_SUBTYPE_ARM_XSCALE:
3509 outs() << " XSCALE";
3511 case MachO::CPU_SUBTYPE_ARM_V6:
3514 case MachO::CPU_SUBTYPE_ARM_V6M:
3517 case MachO::CPU_SUBTYPE_ARM_V7:
3520 case MachO::CPU_SUBTYPE_ARM_V7EM:
3523 case MachO::CPU_SUBTYPE_ARM_V7K:
3526 case MachO::CPU_SUBTYPE_ARM_V7M:
3529 case MachO::CPU_SUBTYPE_ARM_V7S:
3533 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3537 case MachO::CPU_TYPE_ARM64:
3539 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3540 case MachO::CPU_SUBTYPE_ARM64_ALL:
3544 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3548 case MachO::CPU_TYPE_POWERPC:
3550 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3551 case MachO::CPU_SUBTYPE_POWERPC_ALL:
3555 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3559 case MachO::CPU_TYPE_POWERPC64:
3561 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3562 case MachO::CPU_SUBTYPE_POWERPC_ALL:
3566 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3571 if ((cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64) {
3574 outs() << format(" 0x%02" PRIx32,
3575 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
3578 case MachO::MH_OBJECT:
3579 outs() << " OBJECT";
3581 case MachO::MH_EXECUTE:
3582 outs() << " EXECUTE";
3584 case MachO::MH_FVMLIB:
3585 outs() << " FVMLIB";
3587 case MachO::MH_CORE:
3590 case MachO::MH_PRELOAD:
3591 outs() << " PRELOAD";
3593 case MachO::MH_DYLIB:
3596 case MachO::MH_DYLIB_STUB:
3597 outs() << " DYLIB_STUB";
3599 case MachO::MH_DYLINKER:
3600 outs() << " DYLINKER";
3602 case MachO::MH_BUNDLE:
3603 outs() << " BUNDLE";
3605 case MachO::MH_DSYM:
3608 case MachO::MH_KEXT_BUNDLE:
3609 outs() << " KEXTBUNDLE";
3612 outs() << format(" %10u", filetype);
3615 outs() << format(" %5u", ncmds);
3616 outs() << format(" %10u", sizeofcmds);
3618 if (f & MachO::MH_NOUNDEFS) {
3619 outs() << " NOUNDEFS";
3620 f &= ~MachO::MH_NOUNDEFS;
3622 if (f & MachO::MH_INCRLINK) {
3623 outs() << " INCRLINK";
3624 f &= ~MachO::MH_INCRLINK;
3626 if (f & MachO::MH_DYLDLINK) {
3627 outs() << " DYLDLINK";
3628 f &= ~MachO::MH_DYLDLINK;
3630 if (f & MachO::MH_BINDATLOAD) {
3631 outs() << " BINDATLOAD";
3632 f &= ~MachO::MH_BINDATLOAD;
3634 if (f & MachO::MH_PREBOUND) {
3635 outs() << " PREBOUND";
3636 f &= ~MachO::MH_PREBOUND;
3638 if (f & MachO::MH_SPLIT_SEGS) {
3639 outs() << " SPLIT_SEGS";
3640 f &= ~MachO::MH_SPLIT_SEGS;
3642 if (f & MachO::MH_LAZY_INIT) {
3643 outs() << " LAZY_INIT";
3644 f &= ~MachO::MH_LAZY_INIT;
3646 if (f & MachO::MH_TWOLEVEL) {
3647 outs() << " TWOLEVEL";
3648 f &= ~MachO::MH_TWOLEVEL;
3650 if (f & MachO::MH_FORCE_FLAT) {
3651 outs() << " FORCE_FLAT";
3652 f &= ~MachO::MH_FORCE_FLAT;
3654 if (f & MachO::MH_NOMULTIDEFS) {
3655 outs() << " NOMULTIDEFS";
3656 f &= ~MachO::MH_NOMULTIDEFS;
3658 if (f & MachO::MH_NOFIXPREBINDING) {
3659 outs() << " NOFIXPREBINDING";
3660 f &= ~MachO::MH_NOFIXPREBINDING;
3662 if (f & MachO::MH_PREBINDABLE) {
3663 outs() << " PREBINDABLE";
3664 f &= ~MachO::MH_PREBINDABLE;
3666 if (f & MachO::MH_ALLMODSBOUND) {
3667 outs() << " ALLMODSBOUND";
3668 f &= ~MachO::MH_ALLMODSBOUND;
3670 if (f & MachO::MH_SUBSECTIONS_VIA_SYMBOLS) {
3671 outs() << " SUBSECTIONS_VIA_SYMBOLS";
3672 f &= ~MachO::MH_SUBSECTIONS_VIA_SYMBOLS;
3674 if (f & MachO::MH_CANONICAL) {
3675 outs() << " CANONICAL";
3676 f &= ~MachO::MH_CANONICAL;
3678 if (f & MachO::MH_WEAK_DEFINES) {
3679 outs() << " WEAK_DEFINES";
3680 f &= ~MachO::MH_WEAK_DEFINES;
3682 if (f & MachO::MH_BINDS_TO_WEAK) {
3683 outs() << " BINDS_TO_WEAK";
3684 f &= ~MachO::MH_BINDS_TO_WEAK;
3686 if (f & MachO::MH_ALLOW_STACK_EXECUTION) {
3687 outs() << " ALLOW_STACK_EXECUTION";
3688 f &= ~MachO::MH_ALLOW_STACK_EXECUTION;
3690 if (f & MachO::MH_DEAD_STRIPPABLE_DYLIB) {
3691 outs() << " DEAD_STRIPPABLE_DYLIB";
3692 f &= ~MachO::MH_DEAD_STRIPPABLE_DYLIB;
3694 if (f & MachO::MH_PIE) {
3696 f &= ~MachO::MH_PIE;
3698 if (f & MachO::MH_NO_REEXPORTED_DYLIBS) {
3699 outs() << " NO_REEXPORTED_DYLIBS";
3700 f &= ~MachO::MH_NO_REEXPORTED_DYLIBS;
3702 if (f & MachO::MH_HAS_TLV_DESCRIPTORS) {
3703 outs() << " MH_HAS_TLV_DESCRIPTORS";
3704 f &= ~MachO::MH_HAS_TLV_DESCRIPTORS;
3706 if (f & MachO::MH_NO_HEAP_EXECUTION) {
3707 outs() << " MH_NO_HEAP_EXECUTION";
3708 f &= ~MachO::MH_NO_HEAP_EXECUTION;
3710 if (f & MachO::MH_APP_EXTENSION_SAFE) {
3711 outs() << " APP_EXTENSION_SAFE";
3712 f &= ~MachO::MH_APP_EXTENSION_SAFE;
3714 if (f != 0 || flags == 0)
3715 outs() << format(" 0x%08" PRIx32, f);
3717 outs() << format(" 0x%08" PRIx32, magic);
3718 outs() << format(" %7d", cputype);
3719 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3720 outs() << format(" 0x%02" PRIx32,
3721 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
3722 outs() << format(" %10u", filetype);
3723 outs() << format(" %5u", ncmds);
3724 outs() << format(" %10u", sizeofcmds);
3725 outs() << format(" 0x%08" PRIx32, flags);
3730 static void PrintSegmentCommand(uint32_t cmd, uint32_t cmdsize,
3731 StringRef SegName, uint64_t vmaddr,
3732 uint64_t vmsize, uint64_t fileoff,
3733 uint64_t filesize, uint32_t maxprot,
3734 uint32_t initprot, uint32_t nsects,
3735 uint32_t flags, uint32_t object_size,
3737 uint64_t expected_cmdsize;
3738 if (cmd == MachO::LC_SEGMENT) {
3739 outs() << " cmd LC_SEGMENT\n";
3740 expected_cmdsize = nsects;
3741 expected_cmdsize *= sizeof(struct MachO::section);
3742 expected_cmdsize += sizeof(struct MachO::segment_command);
3744 outs() << " cmd LC_SEGMENT_64\n";
3745 expected_cmdsize = nsects;
3746 expected_cmdsize *= sizeof(struct MachO::section_64);
3747 expected_cmdsize += sizeof(struct MachO::segment_command_64);
3749 outs() << " cmdsize " << cmdsize;
3750 if (cmdsize != expected_cmdsize)
3751 outs() << " Inconsistent size\n";
3754 outs() << " segname " << SegName << "\n";
3755 if (cmd == MachO::LC_SEGMENT_64) {
3756 outs() << " vmaddr " << format("0x%016" PRIx64, vmaddr) << "\n";
3757 outs() << " vmsize " << format("0x%016" PRIx64, vmsize) << "\n";
3759 outs() << " vmaddr " << format("0x%08" PRIx64, vmaddr) << "\n";
3760 outs() << " vmsize " << format("0x%08" PRIx64, vmsize) << "\n";
3762 outs() << " fileoff " << fileoff;
3763 if (fileoff > object_size)
3764 outs() << " (past end of file)\n";
3767 outs() << " filesize " << filesize;
3768 if (fileoff + filesize > object_size)
3769 outs() << " (past end of file)\n";
3774 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
3775 MachO::VM_PROT_EXECUTE)) != 0)
3776 outs() << " maxprot ?" << format("0x%08" PRIx32, maxprot) << "\n";
3778 if (maxprot & MachO::VM_PROT_READ)
3779 outs() << " maxprot r";
3781 outs() << " maxprot -";
3782 if (maxprot & MachO::VM_PROT_WRITE)
3786 if (maxprot & MachO::VM_PROT_EXECUTE)
3792 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
3793 MachO::VM_PROT_EXECUTE)) != 0)
3794 outs() << " initprot ?" << format("0x%08" PRIx32, initprot) << "\n";
3796 if (initprot & MachO::VM_PROT_READ)
3797 outs() << " initprot r";
3799 outs() << " initprot -";
3800 if (initprot & MachO::VM_PROT_WRITE)
3804 if (initprot & MachO::VM_PROT_EXECUTE)
3810 outs() << " maxprot " << format("0x%08" PRIx32, maxprot) << "\n";
3811 outs() << " initprot " << format("0x%08" PRIx32, initprot) << "\n";
3813 outs() << " nsects " << nsects << "\n";
3817 outs() << " (none)\n";
3819 if (flags & MachO::SG_HIGHVM) {
3820 outs() << " HIGHVM";
3821 flags &= ~MachO::SG_HIGHVM;
3823 if (flags & MachO::SG_FVMLIB) {
3824 outs() << " FVMLIB";
3825 flags &= ~MachO::SG_FVMLIB;
3827 if (flags & MachO::SG_NORELOC) {
3828 outs() << " NORELOC";
3829 flags &= ~MachO::SG_NORELOC;
3831 if (flags & MachO::SG_PROTECTED_VERSION_1) {
3832 outs() << " PROTECTED_VERSION_1";
3833 flags &= ~MachO::SG_PROTECTED_VERSION_1;
3836 outs() << format(" 0x%08" PRIx32, flags) << " (unknown flags)\n";
3841 outs() << " flags " << format("0x%" PRIx32, flags) << "\n";
3845 static void PrintSection(const char *sectname, const char *segname,
3846 uint64_t addr, uint64_t size, uint32_t offset,
3847 uint32_t align, uint32_t reloff, uint32_t nreloc,
3848 uint32_t flags, uint32_t reserved1, uint32_t reserved2,
3849 uint32_t cmd, const char *sg_segname,
3850 uint32_t filetype, uint32_t object_size,
3852 outs() << "Section\n";
3853 outs() << " sectname " << format("%.16s\n", sectname);
3854 outs() << " segname " << format("%.16s", segname);
3855 if (filetype != MachO::MH_OBJECT && strncmp(sg_segname, segname, 16) != 0)
3856 outs() << " (does not match segment)\n";
3859 if (cmd == MachO::LC_SEGMENT_64) {
3860 outs() << " addr " << format("0x%016" PRIx64, addr) << "\n";
3861 outs() << " size " << format("0x%016" PRIx64, size);
3863 outs() << " addr " << format("0x%08" PRIx64, addr) << "\n";
3864 outs() << " size " << format("0x%08" PRIx64, size);
3866 if ((flags & MachO::S_ZEROFILL) != 0 && offset + size > object_size)
3867 outs() << " (past end of file)\n";
3870 outs() << " offset " << offset;
3871 if (offset > object_size)
3872 outs() << " (past end of file)\n";
3875 uint32_t align_shifted = 1 << align;
3876 outs() << " align 2^" << align << " (" << align_shifted << ")\n";
3877 outs() << " reloff " << reloff;
3878 if (reloff > object_size)
3879 outs() << " (past end of file)\n";
3882 outs() << " nreloc " << nreloc;
3883 if (reloff + nreloc * sizeof(struct MachO::relocation_info) > object_size)
3884 outs() << " (past end of file)\n";
3887 uint32_t section_type = flags & MachO::SECTION_TYPE;
3890 if (section_type == MachO::S_REGULAR)
3891 outs() << " S_REGULAR\n";
3892 else if (section_type == MachO::S_ZEROFILL)
3893 outs() << " S_ZEROFILL\n";
3894 else if (section_type == MachO::S_CSTRING_LITERALS)
3895 outs() << " S_CSTRING_LITERALS\n";
3896 else if (section_type == MachO::S_4BYTE_LITERALS)
3897 outs() << " S_4BYTE_LITERALS\n";
3898 else if (section_type == MachO::S_8BYTE_LITERALS)
3899 outs() << " S_8BYTE_LITERALS\n";
3900 else if (section_type == MachO::S_16BYTE_LITERALS)
3901 outs() << " S_16BYTE_LITERALS\n";
3902 else if (section_type == MachO::S_LITERAL_POINTERS)
3903 outs() << " S_LITERAL_POINTERS\n";
3904 else if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS)
3905 outs() << " S_NON_LAZY_SYMBOL_POINTERS\n";
3906 else if (section_type == MachO::S_LAZY_SYMBOL_POINTERS)
3907 outs() << " S_LAZY_SYMBOL_POINTERS\n";
3908 else if (section_type == MachO::S_SYMBOL_STUBS)
3909 outs() << " S_SYMBOL_STUBS\n";
3910 else if (section_type == MachO::S_MOD_INIT_FUNC_POINTERS)
3911 outs() << " S_MOD_INIT_FUNC_POINTERS\n";
3912 else if (section_type == MachO::S_MOD_TERM_FUNC_POINTERS)
3913 outs() << " S_MOD_TERM_FUNC_POINTERS\n";
3914 else if (section_type == MachO::S_COALESCED)
3915 outs() << " S_COALESCED\n";
3916 else if (section_type == MachO::S_INTERPOSING)
3917 outs() << " S_INTERPOSING\n";
3918 else if (section_type == MachO::S_DTRACE_DOF)
3919 outs() << " S_DTRACE_DOF\n";
3920 else if (section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS)
3921 outs() << " S_LAZY_DYLIB_SYMBOL_POINTERS\n";
3922 else if (section_type == MachO::S_THREAD_LOCAL_REGULAR)
3923 outs() << " S_THREAD_LOCAL_REGULAR\n";
3924 else if (section_type == MachO::S_THREAD_LOCAL_ZEROFILL)
3925 outs() << " S_THREAD_LOCAL_ZEROFILL\n";
3926 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLES)
3927 outs() << " S_THREAD_LOCAL_VARIABLES\n";
3928 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
3929 outs() << " S_THREAD_LOCAL_VARIABLE_POINTERS\n";
3930 else if (section_type == MachO::S_THREAD_LOCAL_INIT_FUNCTION_POINTERS)
3931 outs() << " S_THREAD_LOCAL_INIT_FUNCTION_POINTERS\n";
3933 outs() << format("0x%08" PRIx32, section_type) << "\n";
3934 outs() << "attributes";
3935 uint32_t section_attributes = flags & MachO::SECTION_ATTRIBUTES;
3936 if (section_attributes & MachO::S_ATTR_PURE_INSTRUCTIONS)
3937 outs() << " PURE_INSTRUCTIONS";
3938 if (section_attributes & MachO::S_ATTR_NO_TOC)
3939 outs() << " NO_TOC";
3940 if (section_attributes & MachO::S_ATTR_STRIP_STATIC_SYMS)
3941 outs() << " STRIP_STATIC_SYMS";
3942 if (section_attributes & MachO::S_ATTR_NO_DEAD_STRIP)
3943 outs() << " NO_DEAD_STRIP";
3944 if (section_attributes & MachO::S_ATTR_LIVE_SUPPORT)
3945 outs() << " LIVE_SUPPORT";
3946 if (section_attributes & MachO::S_ATTR_SELF_MODIFYING_CODE)
3947 outs() << " SELF_MODIFYING_CODE";
3948 if (section_attributes & MachO::S_ATTR_DEBUG)
3950 if (section_attributes & MachO::S_ATTR_SOME_INSTRUCTIONS)
3951 outs() << " SOME_INSTRUCTIONS";
3952 if (section_attributes & MachO::S_ATTR_EXT_RELOC)
3953 outs() << " EXT_RELOC";
3954 if (section_attributes & MachO::S_ATTR_LOC_RELOC)
3955 outs() << " LOC_RELOC";
3956 if (section_attributes == 0)
3957 outs() << " (none)";
3960 outs() << " flags " << format("0x%08" PRIx32, flags) << "\n";
3961 outs() << " reserved1 " << reserved1;
3962 if (section_type == MachO::S_SYMBOL_STUBS ||
3963 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
3964 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
3965 section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
3966 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
3967 outs() << " (index into indirect symbol table)\n";
3970 outs() << " reserved2 " << reserved2;
3971 if (section_type == MachO::S_SYMBOL_STUBS)
3972 outs() << " (size of stubs)\n";
3977 static void PrintSymtabLoadCommand(MachO::symtab_command st, bool Is64Bit,
3978 uint32_t object_size) {
3979 outs() << " cmd LC_SYMTAB\n";
3980 outs() << " cmdsize " << st.cmdsize;
3981 if (st.cmdsize != sizeof(struct MachO::symtab_command))
3982 outs() << " Incorrect size\n";
3985 outs() << " symoff " << st.symoff;
3986 if (st.symoff > object_size)
3987 outs() << " (past end of file)\n";
3990 outs() << " nsyms " << st.nsyms;
3993 big_size = st.nsyms;
3994 big_size *= sizeof(struct MachO::nlist_64);
3995 big_size += st.symoff;
3996 if (big_size > object_size)
3997 outs() << " (past end of file)\n";
4001 big_size = st.nsyms;
4002 big_size *= sizeof(struct MachO::nlist);
4003 big_size += st.symoff;
4004 if (big_size > object_size)
4005 outs() << " (past end of file)\n";
4009 outs() << " stroff " << st.stroff;
4010 if (st.stroff > object_size)
4011 outs() << " (past end of file)\n";
4014 outs() << " strsize " << st.strsize;
4015 big_size = st.stroff;
4016 big_size += st.strsize;
4017 if (big_size > object_size)
4018 outs() << " (past end of file)\n";
4023 static void PrintDysymtabLoadCommand(MachO::dysymtab_command dyst,
4024 uint32_t nsyms, uint32_t object_size,
4026 outs() << " cmd LC_DYSYMTAB\n";
4027 outs() << " cmdsize " << dyst.cmdsize;
4028 if (dyst.cmdsize != sizeof(struct MachO::dysymtab_command))
4029 outs() << " Incorrect size\n";
4032 outs() << " ilocalsym " << dyst.ilocalsym;
4033 if (dyst.ilocalsym > nsyms)
4034 outs() << " (greater than the number of symbols)\n";
4037 outs() << " nlocalsym " << dyst.nlocalsym;
4039 big_size = dyst.ilocalsym;
4040 big_size += dyst.nlocalsym;
4041 if (big_size > nsyms)
4042 outs() << " (past the end of the symbol table)\n";
4045 outs() << " iextdefsym " << dyst.iextdefsym;
4046 if (dyst.iextdefsym > nsyms)
4047 outs() << " (greater than the number of symbols)\n";
4050 outs() << " nextdefsym " << dyst.nextdefsym;
4051 big_size = dyst.iextdefsym;
4052 big_size += dyst.nextdefsym;
4053 if (big_size > nsyms)
4054 outs() << " (past the end of the symbol table)\n";
4057 outs() << " iundefsym " << dyst.iundefsym;
4058 if (dyst.iundefsym > nsyms)
4059 outs() << " (greater than the number of symbols)\n";
4062 outs() << " nundefsym " << dyst.nundefsym;
4063 big_size = dyst.iundefsym;
4064 big_size += dyst.nundefsym;
4065 if (big_size > nsyms)
4066 outs() << " (past the end of the symbol table)\n";
4069 outs() << " tocoff " << dyst.tocoff;
4070 if (dyst.tocoff > object_size)
4071 outs() << " (past end of file)\n";
4074 outs() << " ntoc " << dyst.ntoc;
4075 big_size = dyst.ntoc;
4076 big_size *= sizeof(struct MachO::dylib_table_of_contents);
4077 big_size += dyst.tocoff;
4078 if (big_size > object_size)
4079 outs() << " (past end of file)\n";
4082 outs() << " modtaboff " << dyst.modtaboff;
4083 if (dyst.modtaboff > object_size)
4084 outs() << " (past end of file)\n";
4087 outs() << " nmodtab " << dyst.nmodtab;
4090 modtabend = dyst.nmodtab;
4091 modtabend *= sizeof(struct MachO::dylib_module_64);
4092 modtabend += dyst.modtaboff;
4094 modtabend = dyst.nmodtab;
4095 modtabend *= sizeof(struct MachO::dylib_module);
4096 modtabend += dyst.modtaboff;
4098 if (modtabend > object_size)
4099 outs() << " (past end of file)\n";
4102 outs() << " extrefsymoff " << dyst.extrefsymoff;
4103 if (dyst.extrefsymoff > object_size)
4104 outs() << " (past end of file)\n";
4107 outs() << " nextrefsyms " << dyst.nextrefsyms;
4108 big_size = dyst.nextrefsyms;
4109 big_size *= sizeof(struct MachO::dylib_reference);
4110 big_size += dyst.extrefsymoff;
4111 if (big_size > object_size)
4112 outs() << " (past end of file)\n";
4115 outs() << " indirectsymoff " << dyst.indirectsymoff;
4116 if (dyst.indirectsymoff > object_size)
4117 outs() << " (past end of file)\n";
4120 outs() << " nindirectsyms " << dyst.nindirectsyms;
4121 big_size = dyst.nindirectsyms;
4122 big_size *= sizeof(uint32_t);
4123 big_size += dyst.indirectsymoff;
4124 if (big_size > object_size)
4125 outs() << " (past end of file)\n";
4128 outs() << " extreloff " << dyst.extreloff;
4129 if (dyst.extreloff > object_size)
4130 outs() << " (past end of file)\n";
4133 outs() << " nextrel " << dyst.nextrel;
4134 big_size = dyst.nextrel;
4135 big_size *= sizeof(struct MachO::relocation_info);
4136 big_size += dyst.extreloff;
4137 if (big_size > object_size)
4138 outs() << " (past end of file)\n";
4141 outs() << " locreloff " << dyst.locreloff;
4142 if (dyst.locreloff > object_size)
4143 outs() << " (past end of file)\n";
4146 outs() << " nlocrel " << dyst.nlocrel;
4147 big_size = dyst.nlocrel;
4148 big_size *= sizeof(struct MachO::relocation_info);
4149 big_size += dyst.locreloff;
4150 if (big_size > object_size)
4151 outs() << " (past end of file)\n";
4156 static void PrintDyldInfoLoadCommand(MachO::dyld_info_command dc,
4157 uint32_t object_size) {
4158 if (dc.cmd == MachO::LC_DYLD_INFO)
4159 outs() << " cmd LC_DYLD_INFO\n";
4161 outs() << " cmd LC_DYLD_INFO_ONLY\n";
4162 outs() << " cmdsize " << dc.cmdsize;
4163 if (dc.cmdsize != sizeof(struct MachO::dyld_info_command))
4164 outs() << " Incorrect size\n";
4167 outs() << " rebase_off " << dc.rebase_off;
4168 if (dc.rebase_off > object_size)
4169 outs() << " (past end of file)\n";
4172 outs() << " rebase_size " << dc.rebase_size;
4174 big_size = dc.rebase_off;
4175 big_size += dc.rebase_size;
4176 if (big_size > object_size)
4177 outs() << " (past end of file)\n";
4180 outs() << " bind_off " << dc.bind_off;
4181 if (dc.bind_off > object_size)
4182 outs() << " (past end of file)\n";
4185 outs() << " bind_size " << dc.bind_size;
4186 big_size = dc.bind_off;
4187 big_size += dc.bind_size;
4188 if (big_size > object_size)
4189 outs() << " (past end of file)\n";
4192 outs() << " weak_bind_off " << dc.weak_bind_off;
4193 if (dc.weak_bind_off > object_size)
4194 outs() << " (past end of file)\n";
4197 outs() << " weak_bind_size " << dc.weak_bind_size;
4198 big_size = dc.weak_bind_off;
4199 big_size += dc.weak_bind_size;
4200 if (big_size > object_size)
4201 outs() << " (past end of file)\n";
4204 outs() << " lazy_bind_off " << dc.lazy_bind_off;
4205 if (dc.lazy_bind_off > object_size)
4206 outs() << " (past end of file)\n";
4209 outs() << " lazy_bind_size " << dc.lazy_bind_size;
4210 big_size = dc.lazy_bind_off;
4211 big_size += dc.lazy_bind_size;
4212 if (big_size > object_size)
4213 outs() << " (past end of file)\n";
4216 outs() << " export_off " << dc.export_off;
4217 if (dc.export_off > object_size)
4218 outs() << " (past end of file)\n";
4221 outs() << " export_size " << dc.export_size;
4222 big_size = dc.export_off;
4223 big_size += dc.export_size;
4224 if (big_size > object_size)
4225 outs() << " (past end of file)\n";
4230 static void PrintDyldLoadCommand(MachO::dylinker_command dyld,
4232 if (dyld.cmd == MachO::LC_ID_DYLINKER)
4233 outs() << " cmd LC_ID_DYLINKER\n";
4234 else if (dyld.cmd == MachO::LC_LOAD_DYLINKER)
4235 outs() << " cmd LC_LOAD_DYLINKER\n";
4236 else if (dyld.cmd == MachO::LC_DYLD_ENVIRONMENT)
4237 outs() << " cmd LC_DYLD_ENVIRONMENT\n";
4239 outs() << " cmd ?(" << dyld.cmd << ")\n";
4240 outs() << " cmdsize " << dyld.cmdsize;
4241 if (dyld.cmdsize < sizeof(struct MachO::dylinker_command))
4242 outs() << " Incorrect size\n";
4245 if (dyld.name >= dyld.cmdsize)
4246 outs() << " name ?(bad offset " << dyld.name << ")\n";
4248 const char *P = (const char *)(Ptr) + dyld.name;
4249 outs() << " name " << P << " (offset " << dyld.name << ")\n";
4253 static void PrintUuidLoadCommand(MachO::uuid_command uuid) {
4254 outs() << " cmd LC_UUID\n";
4255 outs() << " cmdsize " << uuid.cmdsize;
4256 if (uuid.cmdsize != sizeof(struct MachO::uuid_command))
4257 outs() << " Incorrect size\n";
4261 outs() << format("%02" PRIX32, uuid.uuid[0]);
4262 outs() << format("%02" PRIX32, uuid.uuid[1]);
4263 outs() << format("%02" PRIX32, uuid.uuid[2]);
4264 outs() << format("%02" PRIX32, uuid.uuid[3]);
4266 outs() << format("%02" PRIX32, uuid.uuid[4]);
4267 outs() << format("%02" PRIX32, uuid.uuid[5]);
4269 outs() << format("%02" PRIX32, uuid.uuid[6]);
4270 outs() << format("%02" PRIX32, uuid.uuid[7]);
4272 outs() << format("%02" PRIX32, uuid.uuid[8]);
4273 outs() << format("%02" PRIX32, uuid.uuid[9]);
4275 outs() << format("%02" PRIX32, uuid.uuid[10]);
4276 outs() << format("%02" PRIX32, uuid.uuid[11]);
4277 outs() << format("%02" PRIX32, uuid.uuid[12]);
4278 outs() << format("%02" PRIX32, uuid.uuid[13]);
4279 outs() << format("%02" PRIX32, uuid.uuid[14]);
4280 outs() << format("%02" PRIX32, uuid.uuid[15]);
4284 static void PrintRpathLoadCommand(MachO::rpath_command rpath, const char *Ptr) {
4285 outs() << " cmd LC_RPATH\n";
4286 outs() << " cmdsize " << rpath.cmdsize;
4287 if (rpath.cmdsize < sizeof(struct MachO::rpath_command))
4288 outs() << " Incorrect size\n";
4291 if (rpath.path >= rpath.cmdsize)
4292 outs() << " path ?(bad offset " << rpath.path << ")\n";
4294 const char *P = (const char *)(Ptr) + rpath.path;
4295 outs() << " path " << P << " (offset " << rpath.path << ")\n";
4299 static void PrintVersionMinLoadCommand(MachO::version_min_command vd) {
4300 if (vd.cmd == MachO::LC_VERSION_MIN_MACOSX)
4301 outs() << " cmd LC_VERSION_MIN_MACOSX\n";
4302 else if (vd.cmd == MachO::LC_VERSION_MIN_IPHONEOS)
4303 outs() << " cmd LC_VERSION_MIN_IPHONEOS\n";
4305 outs() << " cmd " << vd.cmd << " (?)\n";
4306 outs() << " cmdsize " << vd.cmdsize;
4307 if (vd.cmdsize != sizeof(struct MachO::version_min_command))
4308 outs() << " Incorrect size\n";
4311 outs() << " version " << ((vd.version >> 16) & 0xffff) << "."
4312 << ((vd.version >> 8) & 0xff);
4313 if ((vd.version & 0xff) != 0)
4314 outs() << "." << (vd.version & 0xff);
4317 outs() << " sdk n/a";
4319 outs() << " sdk " << ((vd.sdk >> 16) & 0xffff) << "."
4320 << ((vd.sdk >> 8) & 0xff);
4322 if ((vd.sdk & 0xff) != 0)
4323 outs() << "." << (vd.sdk & 0xff);
4327 static void PrintSourceVersionCommand(MachO::source_version_command sd) {
4328 outs() << " cmd LC_SOURCE_VERSION\n";
4329 outs() << " cmdsize " << sd.cmdsize;
4330 if (sd.cmdsize != sizeof(struct MachO::source_version_command))
4331 outs() << " Incorrect size\n";
4334 uint64_t a = (sd.version >> 40) & 0xffffff;
4335 uint64_t b = (sd.version >> 30) & 0x3ff;
4336 uint64_t c = (sd.version >> 20) & 0x3ff;
4337 uint64_t d = (sd.version >> 10) & 0x3ff;
4338 uint64_t e = sd.version & 0x3ff;
4339 outs() << " version " << a << "." << b;
4341 outs() << "." << c << "." << d << "." << e;
4343 outs() << "." << c << "." << d;
4349 static void PrintEntryPointCommand(MachO::entry_point_command ep) {
4350 outs() << " cmd LC_MAIN\n";
4351 outs() << " cmdsize " << ep.cmdsize;
4352 if (ep.cmdsize != sizeof(struct MachO::entry_point_command))
4353 outs() << " Incorrect size\n";
4356 outs() << " entryoff " << ep.entryoff << "\n";
4357 outs() << " stacksize " << ep.stacksize << "\n";
4360 static void PrintEncryptionInfoCommand(MachO::encryption_info_command ec,
4361 uint32_t object_size) {
4362 outs() << " cmd LC_ENCRYPTION_INFO\n";
4363 outs() << " cmdsize " << ec.cmdsize;
4364 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command))
4365 outs() << " Incorrect size\n";
4368 outs() << " cryptoff " << ec.cryptoff;
4369 if (ec.cryptoff > object_size)
4370 outs() << " (past end of file)\n";
4373 outs() << " cryptsize " << ec.cryptsize;
4374 if (ec.cryptsize > object_size)
4375 outs() << " (past end of file)\n";
4378 outs() << " cryptid " << ec.cryptid << "\n";
4381 static void PrintEncryptionInfoCommand64(MachO::encryption_info_command_64 ec,
4382 uint32_t object_size) {
4383 outs() << " cmd LC_ENCRYPTION_INFO_64\n";
4384 outs() << " cmdsize " << ec.cmdsize;
4385 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command_64))
4386 outs() << " Incorrect size\n";
4389 outs() << " cryptoff " << ec.cryptoff;
4390 if (ec.cryptoff > object_size)
4391 outs() << " (past end of file)\n";
4394 outs() << " cryptsize " << ec.cryptsize;
4395 if (ec.cryptsize > object_size)
4396 outs() << " (past end of file)\n";
4399 outs() << " cryptid " << ec.cryptid << "\n";
4400 outs() << " pad " << ec.pad << "\n";
4403 static void PrintLinkerOptionCommand(MachO::linker_option_command lo,
4405 outs() << " cmd LC_LINKER_OPTION\n";
4406 outs() << " cmdsize " << lo.cmdsize;
4407 if (lo.cmdsize < sizeof(struct MachO::linker_option_command))
4408 outs() << " Incorrect size\n";
4411 outs() << " count " << lo.count << "\n";
4412 const char *string = Ptr + sizeof(struct MachO::linker_option_command);
4413 uint32_t left = lo.cmdsize - sizeof(struct MachO::linker_option_command);
4416 while (*string == '\0' && left > 0) {
4422 outs() << " string #" << i << " " << format("%.*s\n", left, string);
4423 uint32_t NullPos = StringRef(string, left).find('\0');
4424 uint32_t len = std::min(NullPos, left) + 1;
4430 outs() << " count " << lo.count << " does not match number of strings "
4434 static void PrintSubFrameworkCommand(MachO::sub_framework_command sub,
4436 outs() << " cmd LC_SUB_FRAMEWORK\n";
4437 outs() << " cmdsize " << sub.cmdsize;
4438 if (sub.cmdsize < sizeof(struct MachO::sub_framework_command))
4439 outs() << " Incorrect size\n";
4442 if (sub.umbrella < sub.cmdsize) {
4443 const char *P = Ptr + sub.umbrella;
4444 outs() << " umbrella " << P << " (offset " << sub.umbrella << ")\n";
4446 outs() << " umbrella ?(bad offset " << sub.umbrella << ")\n";
4450 static void PrintSubUmbrellaCommand(MachO::sub_umbrella_command sub,
4452 outs() << " cmd LC_SUB_UMBRELLA\n";
4453 outs() << " cmdsize " << sub.cmdsize;
4454 if (sub.cmdsize < sizeof(struct MachO::sub_umbrella_command))
4455 outs() << " Incorrect size\n";
4458 if (sub.sub_umbrella < sub.cmdsize) {
4459 const char *P = Ptr + sub.sub_umbrella;
4460 outs() << " sub_umbrella " << P << " (offset " << sub.sub_umbrella << ")\n";
4462 outs() << " sub_umbrella ?(bad offset " << sub.sub_umbrella << ")\n";
4466 static void PrintSubLibraryCommand(MachO::sub_library_command sub,
4468 outs() << " cmd LC_SUB_LIBRARY\n";
4469 outs() << " cmdsize " << sub.cmdsize;
4470 if (sub.cmdsize < sizeof(struct MachO::sub_library_command))
4471 outs() << " Incorrect size\n";
4474 if (sub.sub_library < sub.cmdsize) {
4475 const char *P = Ptr + sub.sub_library;
4476 outs() << " sub_library " << P << " (offset " << sub.sub_library << ")\n";
4478 outs() << " sub_library ?(bad offset " << sub.sub_library << ")\n";
4482 static void PrintSubClientCommand(MachO::sub_client_command sub,
4484 outs() << " cmd LC_SUB_CLIENT\n";
4485 outs() << " cmdsize " << sub.cmdsize;
4486 if (sub.cmdsize < sizeof(struct MachO::sub_client_command))
4487 outs() << " Incorrect size\n";
4490 if (sub.client < sub.cmdsize) {
4491 const char *P = Ptr + sub.client;
4492 outs() << " client " << P << " (offset " << sub.client << ")\n";
4494 outs() << " client ?(bad offset " << sub.client << ")\n";
4498 static void PrintRoutinesCommand(MachO::routines_command r) {
4499 outs() << " cmd LC_ROUTINES\n";
4500 outs() << " cmdsize " << r.cmdsize;
4501 if (r.cmdsize != sizeof(struct MachO::routines_command))
4502 outs() << " Incorrect size\n";
4505 outs() << " init_address " << format("0x%08" PRIx32, r.init_address) << "\n";
4506 outs() << " init_module " << r.init_module << "\n";
4507 outs() << " reserved1 " << r.reserved1 << "\n";
4508 outs() << " reserved2 " << r.reserved2 << "\n";
4509 outs() << " reserved3 " << r.reserved3 << "\n";
4510 outs() << " reserved4 " << r.reserved4 << "\n";
4511 outs() << " reserved5 " << r.reserved5 << "\n";
4512 outs() << " reserved6 " << r.reserved6 << "\n";
4515 static void PrintRoutinesCommand64(MachO::routines_command_64 r) {
4516 outs() << " cmd LC_ROUTINES_64\n";
4517 outs() << " cmdsize " << r.cmdsize;
4518 if (r.cmdsize != sizeof(struct MachO::routines_command_64))
4519 outs() << " Incorrect size\n";
4522 outs() << " init_address " << format("0x%016" PRIx64, r.init_address) << "\n";
4523 outs() << " init_module " << r.init_module << "\n";
4524 outs() << " reserved1 " << r.reserved1 << "\n";
4525 outs() << " reserved2 " << r.reserved2 << "\n";
4526 outs() << " reserved3 " << r.reserved3 << "\n";
4527 outs() << " reserved4 " << r.reserved4 << "\n";
4528 outs() << " reserved5 " << r.reserved5 << "\n";
4529 outs() << " reserved6 " << r.reserved6 << "\n";
4532 static void Print_x86_thread_state64_t(MachO::x86_thread_state64_t &cpu64) {
4533 outs() << " rax " << format("0x%016" PRIx64, cpu64.rax);
4534 outs() << " rbx " << format("0x%016" PRIx64, cpu64.rbx);
4535 outs() << " rcx " << format("0x%016" PRIx64, cpu64.rcx) << "\n";
4536 outs() << " rdx " << format("0x%016" PRIx64, cpu64.rdx);
4537 outs() << " rdi " << format("0x%016" PRIx64, cpu64.rdi);
4538 outs() << " rsi " << format("0x%016" PRIx64, cpu64.rsi) << "\n";
4539 outs() << " rbp " << format("0x%016" PRIx64, cpu64.rbp);
4540 outs() << " rsp " << format("0x%016" PRIx64, cpu64.rsp);
4541 outs() << " r8 " << format("0x%016" PRIx64, cpu64.r8) << "\n";
4542 outs() << " r9 " << format("0x%016" PRIx64, cpu64.r9);
4543 outs() << " r10 " << format("0x%016" PRIx64, cpu64.r10);
4544 outs() << " r11 " << format("0x%016" PRIx64, cpu64.r11) << "\n";
4545 outs() << " r12 " << format("0x%016" PRIx64, cpu64.r12);
4546 outs() << " r13 " << format("0x%016" PRIx64, cpu64.r13);
4547 outs() << " r14 " << format("0x%016" PRIx64, cpu64.r14) << "\n";
4548 outs() << " r15 " << format("0x%016" PRIx64, cpu64.r15);
4549 outs() << " rip " << format("0x%016" PRIx64, cpu64.rip) << "\n";
4550 outs() << "rflags " << format("0x%016" PRIx64, cpu64.rflags);
4551 outs() << " cs " << format("0x%016" PRIx64, cpu64.cs);
4552 outs() << " fs " << format("0x%016" PRIx64, cpu64.fs) << "\n";
4553 outs() << " gs " << format("0x%016" PRIx64, cpu64.gs) << "\n";
4556 static void Print_mmst_reg(MachO::mmst_reg_t &r) {
4558 outs() << "\t mmst_reg ";
4559 for (f = 0; f < 10; f++)
4560 outs() << format("%02" PRIx32, (r.mmst_reg[f] & 0xff)) << " ";
4562 outs() << "\t mmst_rsrv ";
4563 for (f = 0; f < 6; f++)
4564 outs() << format("%02" PRIx32, (r.mmst_rsrv[f] & 0xff)) << " ";
4568 static void Print_xmm_reg(MachO::xmm_reg_t &r) {
4570 outs() << "\t xmm_reg ";
4571 for (f = 0; f < 16; f++)
4572 outs() << format("%02" PRIx32, (r.xmm_reg[f] & 0xff)) << " ";
4576 static void Print_x86_float_state_t(MachO::x86_float_state64_t &fpu) {
4577 outs() << "\t fpu_reserved[0] " << fpu.fpu_reserved[0];
4578 outs() << " fpu_reserved[1] " << fpu.fpu_reserved[1] << "\n";
4579 outs() << "\t control: invalid " << fpu.fpu_fcw.invalid;
4580 outs() << " denorm " << fpu.fpu_fcw.denorm;
4581 outs() << " zdiv " << fpu.fpu_fcw.zdiv;
4582 outs() << " ovrfl " << fpu.fpu_fcw.ovrfl;
4583 outs() << " undfl " << fpu.fpu_fcw.undfl;
4584 outs() << " precis " << fpu.fpu_fcw.precis << "\n";
4585 outs() << "\t\t pc ";
4586 if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_24B)
4587 outs() << "FP_PREC_24B ";
4588 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_53B)
4589 outs() << "FP_PREC_53B ";
4590 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_64B)
4591 outs() << "FP_PREC_64B ";
4593 outs() << fpu.fpu_fcw.pc << " ";
4595 if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_NEAR)
4596 outs() << "FP_RND_NEAR ";
4597 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_DOWN)
4598 outs() << "FP_RND_DOWN ";
4599 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_UP)
4600 outs() << "FP_RND_UP ";
4601 else if (fpu.fpu_fcw.rc == MachO::x86_FP_CHOP)
4602 outs() << "FP_CHOP ";
4604 outs() << "\t status: invalid " << fpu.fpu_fsw.invalid;
4605 outs() << " denorm " << fpu.fpu_fsw.denorm;
4606 outs() << " zdiv " << fpu.fpu_fsw.zdiv;
4607 outs() << " ovrfl " << fpu.fpu_fsw.ovrfl;
4608 outs() << " undfl " << fpu.fpu_fsw.undfl;
4609 outs() << " precis " << fpu.fpu_fsw.precis;
4610 outs() << " stkflt " << fpu.fpu_fsw.stkflt << "\n";
4611 outs() << "\t errsumm " << fpu.fpu_fsw.errsumm;
4612 outs() << " c0 " << fpu.fpu_fsw.c0;
4613 outs() << " c1 " << fpu.fpu_fsw.c1;
4614 outs() << " c2 " << fpu.fpu_fsw.c2;
4615 outs() << " tos " << fpu.fpu_fsw.tos;
4616 outs() << " c3 " << fpu.fpu_fsw.c3;
4617 outs() << " busy " << fpu.fpu_fsw.busy << "\n";
4618 outs() << "\t fpu_ftw " << format("0x%02" PRIx32, fpu.fpu_ftw);
4619 outs() << " fpu_rsrv1 " << format("0x%02" PRIx32, fpu.fpu_rsrv1);
4620 outs() << " fpu_fop " << format("0x%04" PRIx32, fpu.fpu_fop);
4621 outs() << " fpu_ip " << format("0x%08" PRIx32, fpu.fpu_ip) << "\n";
4622 outs() << "\t fpu_cs " << format("0x%04" PRIx32, fpu.fpu_cs);
4623 outs() << " fpu_rsrv2 " << format("0x%04" PRIx32, fpu.fpu_rsrv2);
4624 outs() << " fpu_dp " << format("0x%08" PRIx32, fpu.fpu_dp);
4625 outs() << " fpu_ds " << format("0x%04" PRIx32, fpu.fpu_ds) << "\n";
4626 outs() << "\t fpu_rsrv3 " << format("0x%04" PRIx32, fpu.fpu_rsrv3);
4627 outs() << " fpu_mxcsr " << format("0x%08" PRIx32, fpu.fpu_mxcsr);
4628 outs() << " fpu_mxcsrmask " << format("0x%08" PRIx32, fpu.fpu_mxcsrmask);
4630 outs() << "\t fpu_stmm0:\n";
4631 Print_mmst_reg(fpu.fpu_stmm0);
4632 outs() << "\t fpu_stmm1:\n";
4633 Print_mmst_reg(fpu.fpu_stmm1);
4634 outs() << "\t fpu_stmm2:\n";
4635 Print_mmst_reg(fpu.fpu_stmm2);
4636 outs() << "\t fpu_stmm3:\n";
4637 Print_mmst_reg(fpu.fpu_stmm3);
4638 outs() << "\t fpu_stmm4:\n";
4639 Print_mmst_reg(fpu.fpu_stmm4);
4640 outs() << "\t fpu_stmm5:\n";
4641 Print_mmst_reg(fpu.fpu_stmm5);
4642 outs() << "\t fpu_stmm6:\n";
4643 Print_mmst_reg(fpu.fpu_stmm6);
4644 outs() << "\t fpu_stmm7:\n";
4645 Print_mmst_reg(fpu.fpu_stmm7);
4646 outs() << "\t fpu_xmm0:\n";
4647 Print_xmm_reg(fpu.fpu_xmm0);
4648 outs() << "\t fpu_xmm1:\n";
4649 Print_xmm_reg(fpu.fpu_xmm1);
4650 outs() << "\t fpu_xmm2:\n";
4651 Print_xmm_reg(fpu.fpu_xmm2);
4652 outs() << "\t fpu_xmm3:\n";
4653 Print_xmm_reg(fpu.fpu_xmm3);
4654 outs() << "\t fpu_xmm4:\n";
4655 Print_xmm_reg(fpu.fpu_xmm4);
4656 outs() << "\t fpu_xmm5:\n";
4657 Print_xmm_reg(fpu.fpu_xmm5);
4658 outs() << "\t fpu_xmm6:\n";
4659 Print_xmm_reg(fpu.fpu_xmm6);
4660 outs() << "\t fpu_xmm7:\n";
4661 Print_xmm_reg(fpu.fpu_xmm7);
4662 outs() << "\t fpu_xmm8:\n";
4663 Print_xmm_reg(fpu.fpu_xmm8);
4664 outs() << "\t fpu_xmm9:\n";
4665 Print_xmm_reg(fpu.fpu_xmm9);
4666 outs() << "\t fpu_xmm10:\n";
4667 Print_xmm_reg(fpu.fpu_xmm10);
4668 outs() << "\t fpu_xmm11:\n";
4669 Print_xmm_reg(fpu.fpu_xmm11);
4670 outs() << "\t fpu_xmm12:\n";
4671 Print_xmm_reg(fpu.fpu_xmm12);
4672 outs() << "\t fpu_xmm13:\n";
4673 Print_xmm_reg(fpu.fpu_xmm13);
4674 outs() << "\t fpu_xmm14:\n";
4675 Print_xmm_reg(fpu.fpu_xmm14);
4676 outs() << "\t fpu_xmm15:\n";
4677 Print_xmm_reg(fpu.fpu_xmm15);
4678 outs() << "\t fpu_rsrv4:\n";
4679 for (uint32_t f = 0; f < 6; f++) {
4681 for (uint32_t g = 0; g < 16; g++)
4682 outs() << format("%02" PRIx32, fpu.fpu_rsrv4[f * g]) << " ";
4685 outs() << "\t fpu_reserved1 " << format("0x%08" PRIx32, fpu.fpu_reserved1);
4689 static void Print_x86_exception_state_t(MachO::x86_exception_state64_t &exc64) {
4690 outs() << "\t trapno " << format("0x%08" PRIx32, exc64.trapno);
4691 outs() << " err " << format("0x%08" PRIx32, exc64.err);
4692 outs() << " faultvaddr " << format("0x%016" PRIx64, exc64.faultvaddr) << "\n";
4695 static void PrintThreadCommand(MachO::thread_command t, const char *Ptr,
4696 bool isLittleEndian, uint32_t cputype) {
4697 if (t.cmd == MachO::LC_THREAD)
4698 outs() << " cmd LC_THREAD\n";
4699 else if (t.cmd == MachO::LC_UNIXTHREAD)
4700 outs() << " cmd LC_UNIXTHREAD\n";
4702 outs() << " cmd " << t.cmd << " (unknown)\n";
4703 outs() << " cmdsize " << t.cmdsize;
4704 if (t.cmdsize < sizeof(struct MachO::thread_command) + 2 * sizeof(uint32_t))
4705 outs() << " Incorrect size\n";
4709 const char *begin = Ptr + sizeof(struct MachO::thread_command);
4710 const char *end = Ptr + t.cmdsize;
4711 uint32_t flavor, count, left;
4712 if (cputype == MachO::CPU_TYPE_X86_64) {
4713 while (begin < end) {
4714 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4715 memcpy((char *)&flavor, begin, sizeof(uint32_t));
4716 begin += sizeof(uint32_t);
4721 if (isLittleEndian != sys::IsLittleEndianHost)
4722 sys::swapByteOrder(flavor);
4723 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4724 memcpy((char *)&count, begin, sizeof(uint32_t));
4725 begin += sizeof(uint32_t);
4730 if (isLittleEndian != sys::IsLittleEndianHost)
4731 sys::swapByteOrder(count);
4732 if (flavor == MachO::x86_THREAD_STATE64) {
4733 outs() << " flavor x86_THREAD_STATE64\n";
4734 if (count == MachO::x86_THREAD_STATE64_COUNT)
4735 outs() << " count x86_THREAD_STATE64_COUNT\n";
4737 outs() << " count " << count
4738 << " (not x86_THREAD_STATE64_COUNT)\n";
4739 MachO::x86_thread_state64_t cpu64;
4741 if (left >= sizeof(MachO::x86_thread_state64_t)) {
4742 memcpy(&cpu64, begin, sizeof(MachO::x86_thread_state64_t));
4743 begin += sizeof(MachO::x86_thread_state64_t);
4745 memset(&cpu64, '\0', sizeof(MachO::x86_thread_state64_t));
4746 memcpy(&cpu64, begin, left);
4749 if (isLittleEndian != sys::IsLittleEndianHost)
4751 Print_x86_thread_state64_t(cpu64);
4752 } else if (flavor == MachO::x86_THREAD_STATE) {
4753 outs() << " flavor x86_THREAD_STATE\n";
4754 if (count == MachO::x86_THREAD_STATE_COUNT)
4755 outs() << " count x86_THREAD_STATE_COUNT\n";
4757 outs() << " count " << count
4758 << " (not x86_THREAD_STATE_COUNT)\n";
4759 struct MachO::x86_thread_state_t ts;
4761 if (left >= sizeof(MachO::x86_thread_state_t)) {
4762 memcpy(&ts, begin, sizeof(MachO::x86_thread_state_t));
4763 begin += sizeof(MachO::x86_thread_state_t);
4765 memset(&ts, '\0', sizeof(MachO::x86_thread_state_t));
4766 memcpy(&ts, begin, left);
4769 if (isLittleEndian != sys::IsLittleEndianHost)
4771 if (ts.tsh.flavor == MachO::x86_THREAD_STATE64) {
4772 outs() << "\t tsh.flavor x86_THREAD_STATE64 ";
4773 if (ts.tsh.count == MachO::x86_THREAD_STATE64_COUNT)
4774 outs() << "tsh.count x86_THREAD_STATE64_COUNT\n";
4776 outs() << "tsh.count " << ts.tsh.count
4777 << " (not x86_THREAD_STATE64_COUNT\n";
4778 Print_x86_thread_state64_t(ts.uts.ts64);
4780 outs() << "\t tsh.flavor " << ts.tsh.flavor << " tsh.count "
4781 << ts.tsh.count << "\n";
4783 } else if (flavor == MachO::x86_FLOAT_STATE) {
4784 outs() << " flavor x86_FLOAT_STATE\n";
4785 if (count == MachO::x86_FLOAT_STATE_COUNT)
4786 outs() << " count x86_FLOAT_STATE_COUNT\n";
4788 outs() << " count " << count << " (not x86_FLOAT_STATE_COUNT)\n";
4789 struct MachO::x86_float_state_t fs;
4791 if (left >= sizeof(MachO::x86_float_state_t)) {
4792 memcpy(&fs, begin, sizeof(MachO::x86_float_state_t));
4793 begin += sizeof(MachO::x86_float_state_t);
4795 memset(&fs, '\0', sizeof(MachO::x86_float_state_t));
4796 memcpy(&fs, begin, left);
4799 if (isLittleEndian != sys::IsLittleEndianHost)
4801 if (fs.fsh.flavor == MachO::x86_FLOAT_STATE64) {
4802 outs() << "\t fsh.flavor x86_FLOAT_STATE64 ";
4803 if (fs.fsh.count == MachO::x86_FLOAT_STATE64_COUNT)
4804 outs() << "fsh.count x86_FLOAT_STATE64_COUNT\n";
4806 outs() << "fsh.count " << fs.fsh.count
4807 << " (not x86_FLOAT_STATE64_COUNT\n";
4808 Print_x86_float_state_t(fs.ufs.fs64);
4810 outs() << "\t fsh.flavor " << fs.fsh.flavor << " fsh.count "
4811 << fs.fsh.count << "\n";
4813 } else if (flavor == MachO::x86_EXCEPTION_STATE) {
4814 outs() << " flavor x86_EXCEPTION_STATE\n";
4815 if (count == MachO::x86_EXCEPTION_STATE_COUNT)
4816 outs() << " count x86_EXCEPTION_STATE_COUNT\n";
4818 outs() << " count " << count
4819 << " (not x86_EXCEPTION_STATE_COUNT)\n";
4820 struct MachO::x86_exception_state_t es;
4822 if (left >= sizeof(MachO::x86_exception_state_t)) {
4823 memcpy(&es, begin, sizeof(MachO::x86_exception_state_t));
4824 begin += sizeof(MachO::x86_exception_state_t);
4826 memset(&es, '\0', sizeof(MachO::x86_exception_state_t));
4827 memcpy(&es, begin, left);
4830 if (isLittleEndian != sys::IsLittleEndianHost)
4832 if (es.esh.flavor == MachO::x86_EXCEPTION_STATE64) {
4833 outs() << "\t esh.flavor x86_EXCEPTION_STATE64\n";
4834 if (es.esh.count == MachO::x86_EXCEPTION_STATE64_COUNT)
4835 outs() << "\t esh.count x86_EXCEPTION_STATE64_COUNT\n";
4837 outs() << "\t esh.count " << es.esh.count
4838 << " (not x86_EXCEPTION_STATE64_COUNT\n";
4839 Print_x86_exception_state_t(es.ues.es64);
4841 outs() << "\t esh.flavor " << es.esh.flavor << " esh.count "
4842 << es.esh.count << "\n";
4845 outs() << " flavor " << flavor << " (unknown)\n";
4846 outs() << " count " << count << "\n";
4847 outs() << " state (unknown)\n";
4848 begin += count * sizeof(uint32_t);
4852 while (begin < end) {
4853 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4854 memcpy((char *)&flavor, begin, sizeof(uint32_t));
4855 begin += sizeof(uint32_t);
4860 if (isLittleEndian != sys::IsLittleEndianHost)
4861 sys::swapByteOrder(flavor);
4862 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4863 memcpy((char *)&count, begin, sizeof(uint32_t));
4864 begin += sizeof(uint32_t);
4869 if (isLittleEndian != sys::IsLittleEndianHost)
4870 sys::swapByteOrder(count);
4871 outs() << " flavor " << flavor << "\n";
4872 outs() << " count " << count << "\n";
4873 outs() << " state (Unknown cputype/cpusubtype)\n";
4874 begin += count * sizeof(uint32_t);
4879 static void PrintDylibCommand(MachO::dylib_command dl, const char *Ptr) {
4880 if (dl.cmd == MachO::LC_ID_DYLIB)
4881 outs() << " cmd LC_ID_DYLIB\n";
4882 else if (dl.cmd == MachO::LC_LOAD_DYLIB)
4883 outs() << " cmd LC_LOAD_DYLIB\n";
4884 else if (dl.cmd == MachO::LC_LOAD_WEAK_DYLIB)
4885 outs() << " cmd LC_LOAD_WEAK_DYLIB\n";
4886 else if (dl.cmd == MachO::LC_REEXPORT_DYLIB)
4887 outs() << " cmd LC_REEXPORT_DYLIB\n";
4888 else if (dl.cmd == MachO::LC_LAZY_LOAD_DYLIB)
4889 outs() << " cmd LC_LAZY_LOAD_DYLIB\n";
4890 else if (dl.cmd == MachO::LC_LOAD_UPWARD_DYLIB)
4891 outs() << " cmd LC_LOAD_UPWARD_DYLIB\n";
4893 outs() << " cmd " << dl.cmd << " (unknown)\n";
4894 outs() << " cmdsize " << dl.cmdsize;
4895 if (dl.cmdsize < sizeof(struct MachO::dylib_command))
4896 outs() << " Incorrect size\n";
4899 if (dl.dylib.name < dl.cmdsize) {
4900 const char *P = (const char *)(Ptr) + dl.dylib.name;
4901 outs() << " name " << P << " (offset " << dl.dylib.name << ")\n";
4903 outs() << " name ?(bad offset " << dl.dylib.name << ")\n";
4905 outs() << " time stamp " << dl.dylib.timestamp << " ";
4906 time_t t = dl.dylib.timestamp;
4907 outs() << ctime(&t);
4908 outs() << " current version ";
4909 if (dl.dylib.current_version == 0xffffffff)
4912 outs() << ((dl.dylib.current_version >> 16) & 0xffff) << "."
4913 << ((dl.dylib.current_version >> 8) & 0xff) << "."
4914 << (dl.dylib.current_version & 0xff) << "\n";
4915 outs() << "compatibility version ";
4916 if (dl.dylib.compatibility_version == 0xffffffff)
4919 outs() << ((dl.dylib.compatibility_version >> 16) & 0xffff) << "."
4920 << ((dl.dylib.compatibility_version >> 8) & 0xff) << "."
4921 << (dl.dylib.compatibility_version & 0xff) << "\n";
4924 static void PrintLinkEditDataCommand(MachO::linkedit_data_command ld,
4925 uint32_t object_size) {
4926 if (ld.cmd == MachO::LC_CODE_SIGNATURE)
4927 outs() << " cmd LC_FUNCTION_STARTS\n";
4928 else if (ld.cmd == MachO::LC_SEGMENT_SPLIT_INFO)
4929 outs() << " cmd LC_SEGMENT_SPLIT_INFO\n";
4930 else if (ld.cmd == MachO::LC_FUNCTION_STARTS)
4931 outs() << " cmd LC_FUNCTION_STARTS\n";
4932 else if (ld.cmd == MachO::LC_DATA_IN_CODE)
4933 outs() << " cmd LC_DATA_IN_CODE\n";
4934 else if (ld.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS)
4935 outs() << " cmd LC_DYLIB_CODE_SIGN_DRS\n";
4936 else if (ld.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT)
4937 outs() << " cmd LC_LINKER_OPTIMIZATION_HINT\n";
4939 outs() << " cmd " << ld.cmd << " (?)\n";
4940 outs() << " cmdsize " << ld.cmdsize;
4941 if (ld.cmdsize != sizeof(struct MachO::linkedit_data_command))
4942 outs() << " Incorrect size\n";
4945 outs() << " dataoff " << ld.dataoff;
4946 if (ld.dataoff > object_size)
4947 outs() << " (past end of file)\n";
4950 outs() << " datasize " << ld.datasize;
4951 uint64_t big_size = ld.dataoff;
4952 big_size += ld.datasize;
4953 if (big_size > object_size)
4954 outs() << " (past end of file)\n";
4959 static void PrintLoadCommands(const MachOObjectFile *Obj, uint32_t ncmds,
4960 uint32_t filetype, uint32_t cputype,
4964 StringRef Buf = Obj->getData();
4965 MachOObjectFile::LoadCommandInfo Command = Obj->getFirstLoadCommandInfo();
4966 for (unsigned i = 0;; ++i) {
4967 outs() << "Load command " << i << "\n";
4968 if (Command.C.cmd == MachO::LC_SEGMENT) {
4969 MachO::segment_command SLC = Obj->getSegmentLoadCommand(Command);
4970 const char *sg_segname = SLC.segname;
4971 PrintSegmentCommand(SLC.cmd, SLC.cmdsize, SLC.segname, SLC.vmaddr,
4972 SLC.vmsize, SLC.fileoff, SLC.filesize, SLC.maxprot,
4973 SLC.initprot, SLC.nsects, SLC.flags, Buf.size(),
4975 for (unsigned j = 0; j < SLC.nsects; j++) {
4976 MachO::section S = Obj->getSection(Command, j);
4977 PrintSection(S.sectname, S.segname, S.addr, S.size, S.offset, S.align,
4978 S.reloff, S.nreloc, S.flags, S.reserved1, S.reserved2,
4979 SLC.cmd, sg_segname, filetype, Buf.size(), verbose);
4981 } else if (Command.C.cmd == MachO::LC_SEGMENT_64) {
4982 MachO::segment_command_64 SLC_64 = Obj->getSegment64LoadCommand(Command);
4983 const char *sg_segname = SLC_64.segname;
4984 PrintSegmentCommand(SLC_64.cmd, SLC_64.cmdsize, SLC_64.segname,
4985 SLC_64.vmaddr, SLC_64.vmsize, SLC_64.fileoff,
4986 SLC_64.filesize, SLC_64.maxprot, SLC_64.initprot,
4987 SLC_64.nsects, SLC_64.flags, Buf.size(), verbose);
4988 for (unsigned j = 0; j < SLC_64.nsects; j++) {
4989 MachO::section_64 S_64 = Obj->getSection64(Command, j);
4990 PrintSection(S_64.sectname, S_64.segname, S_64.addr, S_64.size,
4991 S_64.offset, S_64.align, S_64.reloff, S_64.nreloc,
4992 S_64.flags, S_64.reserved1, S_64.reserved2, SLC_64.cmd,
4993 sg_segname, filetype, Buf.size(), verbose);
4995 } else if (Command.C.cmd == MachO::LC_SYMTAB) {
4996 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
4997 PrintSymtabLoadCommand(Symtab, Obj->is64Bit(), Buf.size());
4998 } else if (Command.C.cmd == MachO::LC_DYSYMTAB) {
4999 MachO::dysymtab_command Dysymtab = Obj->getDysymtabLoadCommand();
5000 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
5001 PrintDysymtabLoadCommand(Dysymtab, Symtab.nsyms, Buf.size(),
5003 } else if (Command.C.cmd == MachO::LC_DYLD_INFO ||
5004 Command.C.cmd == MachO::LC_DYLD_INFO_ONLY) {
5005 MachO::dyld_info_command DyldInfo = Obj->getDyldInfoLoadCommand(Command);
5006 PrintDyldInfoLoadCommand(DyldInfo, Buf.size());
5007 } else if (Command.C.cmd == MachO::LC_LOAD_DYLINKER ||
5008 Command.C.cmd == MachO::LC_ID_DYLINKER ||
5009 Command.C.cmd == MachO::LC_DYLD_ENVIRONMENT) {
5010 MachO::dylinker_command Dyld = Obj->getDylinkerCommand(Command);
5011 PrintDyldLoadCommand(Dyld, Command.Ptr);
5012 } else if (Command.C.cmd == MachO::LC_UUID) {
5013 MachO::uuid_command Uuid = Obj->getUuidCommand(Command);
5014 PrintUuidLoadCommand(Uuid);
5015 } else if (Command.C.cmd == MachO::LC_RPATH) {
5016 MachO::rpath_command Rpath = Obj->getRpathCommand(Command);
5017 PrintRpathLoadCommand(Rpath, Command.Ptr);
5018 } else if (Command.C.cmd == MachO::LC_VERSION_MIN_MACOSX ||
5019 Command.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS) {
5020 MachO::version_min_command Vd = Obj->getVersionMinLoadCommand(Command);
5021 PrintVersionMinLoadCommand(Vd);
5022 } else if (Command.C.cmd == MachO::LC_SOURCE_VERSION) {
5023 MachO::source_version_command Sd = Obj->getSourceVersionCommand(Command);
5024 PrintSourceVersionCommand(Sd);
5025 } else if (Command.C.cmd == MachO::LC_MAIN) {
5026 MachO::entry_point_command Ep = Obj->getEntryPointCommand(Command);
5027 PrintEntryPointCommand(Ep);
5028 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO) {
5029 MachO::encryption_info_command Ei =
5030 Obj->getEncryptionInfoCommand(Command);
5031 PrintEncryptionInfoCommand(Ei, Buf.size());
5032 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO_64) {
5033 MachO::encryption_info_command_64 Ei =
5034 Obj->getEncryptionInfoCommand64(Command);
5035 PrintEncryptionInfoCommand64(Ei, Buf.size());
5036 } else if (Command.C.cmd == MachO::LC_LINKER_OPTION) {
5037 MachO::linker_option_command Lo =
5038 Obj->getLinkerOptionLoadCommand(Command);
5039 PrintLinkerOptionCommand(Lo, Command.Ptr);
5040 } else if (Command.C.cmd == MachO::LC_SUB_FRAMEWORK) {
5041 MachO::sub_framework_command Sf = Obj->getSubFrameworkCommand(Command);
5042 PrintSubFrameworkCommand(Sf, Command.Ptr);
5043 } else if (Command.C.cmd == MachO::LC_SUB_UMBRELLA) {
5044 MachO::sub_umbrella_command Sf = Obj->getSubUmbrellaCommand(Command);
5045 PrintSubUmbrellaCommand(Sf, Command.Ptr);
5046 } else if (Command.C.cmd == MachO::LC_SUB_LIBRARY) {
5047 MachO::sub_library_command Sl = Obj->getSubLibraryCommand(Command);
5048 PrintSubLibraryCommand(Sl, Command.Ptr);
5049 } else if (Command.C.cmd == MachO::LC_SUB_CLIENT) {
5050 MachO::sub_client_command Sc = Obj->getSubClientCommand(Command);
5051 PrintSubClientCommand(Sc, Command.Ptr);
5052 } else if (Command.C.cmd == MachO::LC_ROUTINES) {
5053 MachO::routines_command Rc = Obj->getRoutinesCommand(Command);
5054 PrintRoutinesCommand(Rc);
5055 } else if (Command.C.cmd == MachO::LC_ROUTINES_64) {
5056 MachO::routines_command_64 Rc = Obj->getRoutinesCommand64(Command);
5057 PrintRoutinesCommand64(Rc);
5058 } else if (Command.C.cmd == MachO::LC_THREAD ||
5059 Command.C.cmd == MachO::LC_UNIXTHREAD) {
5060 MachO::thread_command Tc = Obj->getThreadCommand(Command);
5061 PrintThreadCommand(Tc, Command.Ptr, Obj->isLittleEndian(), cputype);
5062 } else if (Command.C.cmd == MachO::LC_LOAD_DYLIB ||
5063 Command.C.cmd == MachO::LC_ID_DYLIB ||
5064 Command.C.cmd == MachO::LC_LOAD_WEAK_DYLIB ||
5065 Command.C.cmd == MachO::LC_REEXPORT_DYLIB ||
5066 Command.C.cmd == MachO::LC_LAZY_LOAD_DYLIB ||
5067 Command.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) {
5068 MachO::dylib_command Dl = Obj->getDylibIDLoadCommand(Command);
5069 PrintDylibCommand(Dl, Command.Ptr);
5070 } else if (Command.C.cmd == MachO::LC_CODE_SIGNATURE ||
5071 Command.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO ||
5072 Command.C.cmd == MachO::LC_FUNCTION_STARTS ||
5073 Command.C.cmd == MachO::LC_DATA_IN_CODE ||
5074 Command.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS ||
5075 Command.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) {
5076 MachO::linkedit_data_command Ld =
5077 Obj->getLinkeditDataLoadCommand(Command);
5078 PrintLinkEditDataCommand(Ld, Buf.size());
5080 outs() << " cmd ?(" << format("0x%08" PRIx32, Command.C.cmd)
5082 outs() << " cmdsize " << Command.C.cmdsize << "\n";
5083 // TODO: get and print the raw bytes of the load command.
5085 // TODO: print all the other kinds of load commands.
5089 Command = Obj->getNextLoadCommandInfo(Command);
5093 static void getAndPrintMachHeader(const MachOObjectFile *Obj, uint32_t &ncmds,
5094 uint32_t &filetype, uint32_t &cputype,
5096 if (Obj->is64Bit()) {
5097 MachO::mach_header_64 H_64;
5098 H_64 = Obj->getHeader64();
5099 PrintMachHeader(H_64.magic, H_64.cputype, H_64.cpusubtype, H_64.filetype,
5100 H_64.ncmds, H_64.sizeofcmds, H_64.flags, verbose);
5102 filetype = H_64.filetype;
5103 cputype = H_64.cputype;
5105 MachO::mach_header H;
5106 H = Obj->getHeader();
5107 PrintMachHeader(H.magic, H.cputype, H.cpusubtype, H.filetype, H.ncmds,
5108 H.sizeofcmds, H.flags, verbose);
5110 filetype = H.filetype;
5111 cputype = H.cputype;
5115 void llvm::printMachOFileHeader(const object::ObjectFile *Obj) {
5116 const MachOObjectFile *file = dyn_cast<const MachOObjectFile>(Obj);
5118 uint32_t filetype = 0;
5119 uint32_t cputype = 0;
5120 getAndPrintMachHeader(file, ncmds, filetype, cputype, true);
5121 PrintLoadCommands(file, ncmds, filetype, cputype, true);
5124 //===----------------------------------------------------------------------===//
5125 // export trie dumping
5126 //===----------------------------------------------------------------------===//
5128 void llvm::printMachOExportsTrie(const object::MachOObjectFile *Obj) {
5129 for (const llvm::object::ExportEntry &Entry : Obj->exports()) {
5130 uint64_t Flags = Entry.flags();
5131 bool ReExport = (Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT);
5132 bool WeakDef = (Flags & MachO::EXPORT_SYMBOL_FLAGS_WEAK_DEFINITION);
5133 bool ThreadLocal = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
5134 MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL);
5135 bool Abs = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
5136 MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE);
5137 bool Resolver = (Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER);
5139 outs() << "[re-export] ";
5141 outs() << format("0x%08llX ",
5142 Entry.address()); // FIXME:add in base address
5143 outs() << Entry.name();
5144 if (WeakDef || ThreadLocal || Resolver || Abs) {
5145 bool NeedsComma = false;
5148 outs() << "weak_def";
5154 outs() << "per-thread";
5160 outs() << "absolute";
5166 outs() << format("resolver=0x%08llX", Entry.other());
5172 StringRef DylibName = "unknown";
5173 int Ordinal = Entry.other() - 1;
5174 Obj->getLibraryShortNameByIndex(Ordinal, DylibName);
5175 if (Entry.otherName().empty())
5176 outs() << " (from " << DylibName << ")";
5178 outs() << " (" << Entry.otherName() << " from " << DylibName << ")";
5184 //===----------------------------------------------------------------------===//
5185 // rebase table dumping
5186 //===----------------------------------------------------------------------===//
5191 SegInfo(const object::MachOObjectFile *Obj);
5193 StringRef segmentName(uint32_t SegIndex);
5194 StringRef sectionName(uint32_t SegIndex, uint64_t SegOffset);
5195 uint64_t address(uint32_t SegIndex, uint64_t SegOffset);
5198 struct SectionInfo {
5201 StringRef SectionName;
5202 StringRef SegmentName;
5203 uint64_t OffsetInSegment;
5204 uint64_t SegmentStartAddress;
5205 uint32_t SegmentIndex;
5207 const SectionInfo &findSection(uint32_t SegIndex, uint64_t SegOffset);
5208 SmallVector<SectionInfo, 32> Sections;
5212 SegInfo::SegInfo(const object::MachOObjectFile *Obj) {
5213 // Build table of sections so segIndex/offset pairs can be translated.
5214 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0;
5215 StringRef CurSegName;
5216 uint64_t CurSegAddress;
5217 for (const SectionRef &Section : Obj->sections()) {
5219 if (error(Section.getName(Info.SectionName)))
5221 Info.Address = Section.getAddress();
5222 Info.Size = Section.getSize();
5224 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl());
5225 if (!Info.SegmentName.equals(CurSegName)) {
5227 CurSegName = Info.SegmentName;
5228 CurSegAddress = Info.Address;
5230 Info.SegmentIndex = CurSegIndex - 1;
5231 Info.OffsetInSegment = Info.Address - CurSegAddress;
5232 Info.SegmentStartAddress = CurSegAddress;
5233 Sections.push_back(Info);
5237 StringRef SegInfo::segmentName(uint32_t SegIndex) {
5238 for (const SectionInfo &SI : Sections) {
5239 if (SI.SegmentIndex == SegIndex)
5240 return SI.SegmentName;
5242 llvm_unreachable("invalid segIndex");
5245 const SegInfo::SectionInfo &SegInfo::findSection(uint32_t SegIndex,
5246 uint64_t OffsetInSeg) {
5247 for (const SectionInfo &SI : Sections) {
5248 if (SI.SegmentIndex != SegIndex)
5250 if (SI.OffsetInSegment > OffsetInSeg)
5252 if (OffsetInSeg >= (SI.OffsetInSegment + SI.Size))
5256 llvm_unreachable("segIndex and offset not in any section");
5259 StringRef SegInfo::sectionName(uint32_t SegIndex, uint64_t OffsetInSeg) {
5260 return findSection(SegIndex, OffsetInSeg).SectionName;
5263 uint64_t SegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) {
5264 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg);
5265 return SI.SegmentStartAddress + OffsetInSeg;
5268 void llvm::printMachORebaseTable(const object::MachOObjectFile *Obj) {
5269 // Build table of sections so names can used in final output.
5270 SegInfo sectionTable(Obj);
5272 outs() << "segment section address type\n";
5273 for (const llvm::object::MachORebaseEntry &Entry : Obj->rebaseTable()) {
5274 uint32_t SegIndex = Entry.segmentIndex();
5275 uint64_t OffsetInSeg = Entry.segmentOffset();
5276 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5277 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5278 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5280 // Table lines look like: __DATA __nl_symbol_ptr 0x0000F00C pointer
5281 outs() << format("%-8s %-18s 0x%08" PRIX64 " %s\n",
5282 SegmentName.str().c_str(), SectionName.str().c_str(),
5283 Address, Entry.typeName().str().c_str());
5287 static StringRef ordinalName(const object::MachOObjectFile *Obj, int Ordinal) {
5288 StringRef DylibName;
5290 case MachO::BIND_SPECIAL_DYLIB_SELF:
5291 return "this-image";
5292 case MachO::BIND_SPECIAL_DYLIB_MAIN_EXECUTABLE:
5293 return "main-executable";
5294 case MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP:
5295 return "flat-namespace";
5298 std::error_code EC =
5299 Obj->getLibraryShortNameByIndex(Ordinal - 1, DylibName);
5301 return "<<bad library ordinal>>";
5305 return "<<unknown special ordinal>>";
5308 //===----------------------------------------------------------------------===//
5309 // bind table dumping
5310 //===----------------------------------------------------------------------===//
5312 void llvm::printMachOBindTable(const object::MachOObjectFile *Obj) {
5313 // Build table of sections so names can used in final output.
5314 SegInfo sectionTable(Obj);
5316 outs() << "segment section address type "
5317 "addend dylib symbol\n";
5318 for (const llvm::object::MachOBindEntry &Entry : Obj->bindTable()) {
5319 uint32_t SegIndex = Entry.segmentIndex();
5320 uint64_t OffsetInSeg = Entry.segmentOffset();
5321 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5322 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5323 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5325 // Table lines look like:
5326 // __DATA __got 0x00012010 pointer 0 libSystem ___stack_chk_guard
5328 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_WEAK_IMPORT)
5329 Attr = " (weak_import)";
5330 outs() << left_justify(SegmentName, 8) << " "
5331 << left_justify(SectionName, 18) << " "
5332 << format_hex(Address, 10, true) << " "
5333 << left_justify(Entry.typeName(), 8) << " "
5334 << format_decimal(Entry.addend(), 8) << " "
5335 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
5336 << Entry.symbolName() << Attr << "\n";
5340 //===----------------------------------------------------------------------===//
5341 // lazy bind table dumping
5342 //===----------------------------------------------------------------------===//
5344 void llvm::printMachOLazyBindTable(const object::MachOObjectFile *Obj) {
5345 // Build table of sections so names can used in final output.
5346 SegInfo sectionTable(Obj);
5348 outs() << "segment section address "
5350 for (const llvm::object::MachOBindEntry &Entry : Obj->lazyBindTable()) {
5351 uint32_t SegIndex = Entry.segmentIndex();
5352 uint64_t OffsetInSeg = Entry.segmentOffset();
5353 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5354 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5355 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5357 // Table lines look like:
5358 // __DATA __got 0x00012010 libSystem ___stack_chk_guard
5359 outs() << left_justify(SegmentName, 8) << " "
5360 << left_justify(SectionName, 18) << " "
5361 << format_hex(Address, 10, true) << " "
5362 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
5363 << Entry.symbolName() << "\n";
5367 //===----------------------------------------------------------------------===//
5368 // weak bind table dumping
5369 //===----------------------------------------------------------------------===//
5371 void llvm::printMachOWeakBindTable(const object::MachOObjectFile *Obj) {
5372 // Build table of sections so names can used in final output.
5373 SegInfo sectionTable(Obj);
5375 outs() << "segment section address "
5376 "type addend symbol\n";
5377 for (const llvm::object::MachOBindEntry &Entry : Obj->weakBindTable()) {
5378 // Strong symbols don't have a location to update.
5379 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) {
5380 outs() << " strong "
5381 << Entry.symbolName() << "\n";
5384 uint32_t SegIndex = Entry.segmentIndex();
5385 uint64_t OffsetInSeg = Entry.segmentOffset();
5386 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5387 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5388 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5390 // Table lines look like:
5391 // __DATA __data 0x00001000 pointer 0 _foo
5392 outs() << left_justify(SegmentName, 8) << " "
5393 << left_justify(SectionName, 18) << " "
5394 << format_hex(Address, 10, true) << " "
5395 << left_justify(Entry.typeName(), 8) << " "
5396 << format_decimal(Entry.addend(), 8) << " " << Entry.symbolName()
5401 // get_dyld_bind_info_symbolname() is used for disassembly and passed an
5402 // address, ReferenceValue, in the Mach-O file and looks in the dyld bind
5403 // information for that address. If the address is found its binding symbol
5404 // name is returned. If not nullptr is returned.
5405 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
5406 struct DisassembleInfo *info) {
5407 if (info->bindtable == nullptr) {
5408 info->bindtable = new (BindTable);
5409 SegInfo sectionTable(info->O);
5410 for (const llvm::object::MachOBindEntry &Entry : info->O->bindTable()) {
5411 uint32_t SegIndex = Entry.segmentIndex();
5412 uint64_t OffsetInSeg = Entry.segmentOffset();
5413 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5414 const char *SymbolName = nullptr;
5415 StringRef name = Entry.symbolName();
5417 SymbolName = name.data();
5418 info->bindtable->push_back(std::make_pair(Address, SymbolName));
5421 for (bind_table_iterator BI = info->bindtable->begin(),
5422 BE = info->bindtable->end();
5424 uint64_t Address = BI->first;
5425 if (ReferenceValue == Address) {
5426 const char *SymbolName = BI->second;