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 DumpCstringSection(MachOObjectFile *O, const char *sect,
541 uint32_t sect_size, uint64_t sect_addr,
543 for (uint32_t i = 0; i < sect_size; i++) {
545 outs() << format("0x%016" PRIx64, sect_addr + i) << " ";
547 outs() << format("0x%08" PRIx64, sect_addr + i) << " ";
548 for ( ; i < sect_size && sect[i] != '\0'; i++) {
552 outs().write_escaped(p);
554 if (i < sect_size && sect[i] == '\0')
559 static void DumpInitTermPointerSection(MachOObjectFile *O, const char *sect,
560 uint32_t sect_size, uint64_t sect_addr,
561 SymbolAddressMap *AddrMap,
565 stride = sizeof(uint64_t);
567 stride = sizeof(uint32_t);
568 for (uint32_t i = 0; i < sect_size; i += stride) {
569 const char *SymbolName = nullptr;
571 outs() << format("0x%016" PRIx64, sect_addr + i * stride) << " ";
572 uint64_t pointer_value;
573 memcpy(&pointer_value, sect + i, stride);
574 if (O->isLittleEndian() != sys::IsLittleEndianHost)
575 sys::swapByteOrder(pointer_value);
576 outs() << format("0x%016" PRIx64, pointer_value);
578 SymbolName = GuessSymbolName(pointer_value, AddrMap);
580 outs() << format("0x%08" PRIx64, sect_addr + i * stride) << " ";
581 uint32_t pointer_value;
582 memcpy(&pointer_value, sect + i, stride);
583 if (O->isLittleEndian() != sys::IsLittleEndianHost)
584 sys::swapByteOrder(pointer_value);
585 outs() << format("0x%08" PRIx32, pointer_value);
587 SymbolName = GuessSymbolName(pointer_value, AddrMap);
590 outs() << " " << SymbolName;
595 static void DumpRawSectionContents(MachOObjectFile *O, const char *sect,
596 uint32_t size, uint64_t addr) {
597 uint32_t cputype = O->getHeader().cputype;
598 if (cputype == MachO::CPU_TYPE_I386 || cputype == MachO::CPU_TYPE_X86_64) {
600 for (uint32_t i = 0; i < size; i += j, addr += j) {
602 outs() << format("%016" PRIx64, addr) << "\t";
604 outs() << format("%08" PRIx64, sect) << "\t";
605 for (j = 0; j < 16 && i + j < size; j++) {
606 uint8_t byte_word = *(sect + i + j);
607 outs() << format("%02" PRIx32, (uint32_t)byte_word) << " ";
613 for (uint32_t i = 0; i < size; i += j, addr += j) {
615 outs() << format("%016" PRIx64, addr) << "\t";
617 outs() << format("%08" PRIx64, sect) << "\t";
618 for (j = 0; j < 4 * sizeof(int32_t) && i + j < size;
619 j += sizeof(int32_t)) {
620 if (i + j + sizeof(int32_t) < size) {
622 memcpy(&long_word, sect + i + j, sizeof(int32_t));
623 if (O->isLittleEndian() != sys::IsLittleEndianHost)
624 sys::swapByteOrder(long_word);
625 outs() << format("%08" PRIx32, long_word) << " ";
627 for (uint32_t k = 0; i + j + k < size; k++) {
628 uint8_t byte_word = *(sect + i + j);
629 outs() << format("%02" PRIx32, (uint32_t)byte_word) << " ";
638 static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF,
639 StringRef DisSegName, StringRef DisSectName);
641 static void DumpSectionContents(StringRef Filename, MachOObjectFile *O,
643 SymbolAddressMap AddrMap;
645 CreateSymbolAddressMap(O, &AddrMap);
647 for (unsigned i = 0; i < DumpSections.size(); ++i) {
648 StringRef DumpSection = DumpSections[i];
649 std::pair<StringRef, StringRef> DumpSegSectName;
650 DumpSegSectName = DumpSection.split(',');
651 StringRef DumpSegName, DumpSectName;
652 if (DumpSegSectName.second.size()) {
653 DumpSegName = DumpSegSectName.first;
654 DumpSectName = DumpSegSectName.second;
657 DumpSectName = DumpSegSectName.first;
659 for (const SectionRef &Section : O->sections()) {
661 Section.getName(SectName);
662 DataRefImpl Ref = Section.getRawDataRefImpl();
663 StringRef SegName = O->getSectionFinalSegmentName(Ref);
664 if ((DumpSegName.empty() || SegName == DumpSegName) &&
665 (SectName == DumpSectName)) {
666 outs() << "Contents of (" << SegName << "," << SectName
668 uint32_t section_flags;
670 const MachO::section_64 Sec = O->getSection64(Ref);
671 section_flags = Sec.flags;
674 const MachO::section Sec = O->getSection(Ref);
675 section_flags = Sec.flags;
677 uint32_t section_type = section_flags & MachO::SECTION_TYPE;
680 Section.getContents(BytesStr);
681 const char *sect = reinterpret_cast<const char *>(BytesStr.data());
682 uint32_t sect_size = BytesStr.size();
683 uint64_t sect_addr = Section.getAddress();
686 if ((section_flags & MachO::S_ATTR_PURE_INSTRUCTIONS) ||
687 (section_flags & MachO::S_ATTR_SOME_INSTRUCTIONS)) {
688 DisassembleMachO(Filename, O, SegName, SectName);
691 switch (section_type) {
692 case MachO::S_REGULAR:
693 DumpRawSectionContents(O, sect, sect_size, sect_addr);
695 case MachO::S_ZEROFILL:
696 outs() << "zerofill section and has no contents in the file\n";
698 case MachO::S_CSTRING_LITERALS:
699 DumpCstringSection(O, sect, sect_size, sect_addr, verbose);
701 case MachO::S_MOD_INIT_FUNC_POINTERS:
702 case MachO::S_MOD_TERM_FUNC_POINTERS:
703 DumpInitTermPointerSection(O, sect, sect_size, sect_addr, &AddrMap,
707 outs() << "Unknown section type ("
708 << format("0x%08" PRIx32, section_type) << ")\n";
709 DumpRawSectionContents(O, sect, sect_size, sect_addr);
713 if (section_type == MachO::S_ZEROFILL)
714 outs() << "zerofill section and has no contents in the file\n";
716 DumpRawSectionContents(O, sect, sect_size, sect_addr);
723 // checkMachOAndArchFlags() checks to see if the ObjectFile is a Mach-O file
724 // and if it is and there is a list of architecture flags is specified then
725 // check to make sure this Mach-O file is one of those architectures or all
726 // architectures were specified. If not then an error is generated and this
727 // routine returns false. Else it returns true.
728 static bool checkMachOAndArchFlags(ObjectFile *O, StringRef Filename) {
729 if (isa<MachOObjectFile>(O) && !ArchAll && ArchFlags.size() != 0) {
730 MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O);
731 bool ArchFound = false;
732 MachO::mach_header H;
733 MachO::mach_header_64 H_64;
735 if (MachO->is64Bit()) {
736 H_64 = MachO->MachOObjectFile::getHeader64();
737 T = MachOObjectFile::getArch(H_64.cputype, H_64.cpusubtype);
739 H = MachO->MachOObjectFile::getHeader();
740 T = MachOObjectFile::getArch(H.cputype, H.cpusubtype);
743 for (i = 0; i < ArchFlags.size(); ++i) {
744 if (ArchFlags[i] == T.getArchName())
749 errs() << "llvm-objdump: file: " + Filename + " does not contain "
750 << "architecture: " + ArchFlags[i] + "\n";
757 // ProcessMachO() is passed a single opened Mach-O file, which may be an
758 // archive member and or in a slice of a universal file. It prints the
759 // the file name and header info and then processes it according to the
760 // command line options.
761 static void ProcessMachO(StringRef Filename, MachOObjectFile *MachOOF,
762 StringRef ArchiveMemberName = StringRef(),
763 StringRef ArchitectureName = StringRef()) {
764 // If we are doing some processing here on the Mach-O file print the header
765 // info. And don't print it otherwise like in the case of printing the
766 // UniversalHeaders or ArchiveHeaders.
767 if (Disassemble || PrivateHeaders || ExportsTrie || Rebase || Bind ||
768 LazyBind || WeakBind || IndirectSymbols || DataInCode || LinkOptHints ||
769 DumpSections.size() != 0) {
771 if (!ArchiveMemberName.empty())
772 outs() << '(' << ArchiveMemberName << ')';
773 if (!ArchitectureName.empty())
774 outs() << " (architecture " << ArchitectureName << ")";
779 DisassembleMachO(Filename, MachOOF, "__TEXT", "__text");
781 PrintIndirectSymbols(MachOOF, true);
783 PrintDataInCodeTable(MachOOF, true);
785 PrintLinkOptHints(MachOOF);
787 PrintRelocations(MachOOF);
789 PrintSectionHeaders(MachOOF);
791 PrintSectionContents(MachOOF);
792 if (DumpSections.size() != 0)
793 DumpSectionContents(Filename, MachOOF, true);
795 PrintSymbolTable(MachOOF);
797 printMachOUnwindInfo(MachOOF);
799 printMachOFileHeader(MachOOF);
801 printExportsTrie(MachOOF);
803 printRebaseTable(MachOOF);
805 printBindTable(MachOOF);
807 printLazyBindTable(MachOOF);
809 printWeakBindTable(MachOOF);
812 // printUnknownCPUType() helps print_fat_headers for unknown CPU's.
813 static void printUnknownCPUType(uint32_t cputype, uint32_t cpusubtype) {
814 outs() << " cputype (" << cputype << ")\n";
815 outs() << " cpusubtype (" << cpusubtype << ")\n";
818 // printCPUType() helps print_fat_headers by printing the cputype and
819 // pusubtype (symbolically for the one's it knows about).
820 static void printCPUType(uint32_t cputype, uint32_t cpusubtype) {
822 case MachO::CPU_TYPE_I386:
823 switch (cpusubtype) {
824 case MachO::CPU_SUBTYPE_I386_ALL:
825 outs() << " cputype CPU_TYPE_I386\n";
826 outs() << " cpusubtype CPU_SUBTYPE_I386_ALL\n";
829 printUnknownCPUType(cputype, cpusubtype);
833 case MachO::CPU_TYPE_X86_64:
834 switch (cpusubtype) {
835 case MachO::CPU_SUBTYPE_X86_64_ALL:
836 outs() << " cputype CPU_TYPE_X86_64\n";
837 outs() << " cpusubtype CPU_SUBTYPE_X86_64_ALL\n";
839 case MachO::CPU_SUBTYPE_X86_64_H:
840 outs() << " cputype CPU_TYPE_X86_64\n";
841 outs() << " cpusubtype CPU_SUBTYPE_X86_64_H\n";
844 printUnknownCPUType(cputype, cpusubtype);
848 case MachO::CPU_TYPE_ARM:
849 switch (cpusubtype) {
850 case MachO::CPU_SUBTYPE_ARM_ALL:
851 outs() << " cputype CPU_TYPE_ARM\n";
852 outs() << " cpusubtype CPU_SUBTYPE_ARM_ALL\n";
854 case MachO::CPU_SUBTYPE_ARM_V4T:
855 outs() << " cputype CPU_TYPE_ARM\n";
856 outs() << " cpusubtype CPU_SUBTYPE_ARM_V4T\n";
858 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
859 outs() << " cputype CPU_TYPE_ARM\n";
860 outs() << " cpusubtype CPU_SUBTYPE_ARM_V5TEJ\n";
862 case MachO::CPU_SUBTYPE_ARM_XSCALE:
863 outs() << " cputype CPU_TYPE_ARM\n";
864 outs() << " cpusubtype CPU_SUBTYPE_ARM_XSCALE\n";
866 case MachO::CPU_SUBTYPE_ARM_V6:
867 outs() << " cputype CPU_TYPE_ARM\n";
868 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6\n";
870 case MachO::CPU_SUBTYPE_ARM_V6M:
871 outs() << " cputype CPU_TYPE_ARM\n";
872 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6M\n";
874 case MachO::CPU_SUBTYPE_ARM_V7:
875 outs() << " cputype CPU_TYPE_ARM\n";
876 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7\n";
878 case MachO::CPU_SUBTYPE_ARM_V7EM:
879 outs() << " cputype CPU_TYPE_ARM\n";
880 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7EM\n";
882 case MachO::CPU_SUBTYPE_ARM_V7K:
883 outs() << " cputype CPU_TYPE_ARM\n";
884 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7K\n";
886 case MachO::CPU_SUBTYPE_ARM_V7M:
887 outs() << " cputype CPU_TYPE_ARM\n";
888 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7M\n";
890 case MachO::CPU_SUBTYPE_ARM_V7S:
891 outs() << " cputype CPU_TYPE_ARM\n";
892 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7S\n";
895 printUnknownCPUType(cputype, cpusubtype);
899 case MachO::CPU_TYPE_ARM64:
900 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
901 case MachO::CPU_SUBTYPE_ARM64_ALL:
902 outs() << " cputype CPU_TYPE_ARM64\n";
903 outs() << " cpusubtype CPU_SUBTYPE_ARM64_ALL\n";
906 printUnknownCPUType(cputype, cpusubtype);
911 printUnknownCPUType(cputype, cpusubtype);
916 static void printMachOUniversalHeaders(const object::MachOUniversalBinary *UB,
918 outs() << "Fat headers\n";
920 outs() << "fat_magic FAT_MAGIC\n";
922 outs() << "fat_magic " << format("0x%" PRIx32, MachO::FAT_MAGIC) << "\n";
924 uint32_t nfat_arch = UB->getNumberOfObjects();
925 StringRef Buf = UB->getData();
926 uint64_t size = Buf.size();
927 uint64_t big_size = sizeof(struct MachO::fat_header) +
928 nfat_arch * sizeof(struct MachO::fat_arch);
929 outs() << "nfat_arch " << UB->getNumberOfObjects();
931 outs() << " (malformed, contains zero architecture types)\n";
932 else if (big_size > size)
933 outs() << " (malformed, architectures past end of file)\n";
937 for (uint32_t i = 0; i < nfat_arch; ++i) {
938 MachOUniversalBinary::ObjectForArch OFA(UB, i);
939 uint32_t cputype = OFA.getCPUType();
940 uint32_t cpusubtype = OFA.getCPUSubType();
941 outs() << "architecture ";
942 for (uint32_t j = 0; i != 0 && j <= i - 1; j++) {
943 MachOUniversalBinary::ObjectForArch other_OFA(UB, j);
944 uint32_t other_cputype = other_OFA.getCPUType();
945 uint32_t other_cpusubtype = other_OFA.getCPUSubType();
946 if (cputype != 0 && cpusubtype != 0 && cputype == other_cputype &&
947 (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) ==
948 (other_cpusubtype & ~MachO::CPU_SUBTYPE_MASK)) {
949 outs() << "(illegal duplicate architecture) ";
954 outs() << OFA.getArchTypeName() << "\n";
955 printCPUType(cputype, cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
958 outs() << " cputype " << cputype << "\n";
959 outs() << " cpusubtype " << (cpusubtype & ~MachO::CPU_SUBTYPE_MASK)
963 (cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64)
964 outs() << " capabilities CPU_SUBTYPE_LIB64\n";
966 outs() << " capabilities "
967 << format("0x%" PRIx32,
968 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24) << "\n";
969 outs() << " offset " << OFA.getOffset();
970 if (OFA.getOffset() > size)
971 outs() << " (past end of file)";
972 if (OFA.getOffset() % (1 << OFA.getAlign()) != 0)
973 outs() << " (not aligned on it's alignment (2^" << OFA.getAlign() << ")";
975 outs() << " size " << OFA.getSize();
976 big_size = OFA.getOffset() + OFA.getSize();
978 outs() << " (past end of file)";
980 outs() << " align 2^" << OFA.getAlign() << " (" << (1 << OFA.getAlign())
985 static void printArchiveChild(Archive::Child &C, bool verbose,
988 outs() << C.getChildOffset() << "\t";
989 sys::fs::perms Mode = C.getAccessMode();
991 // FIXME: this first dash, "-", is for (Mode & S_IFMT) == S_IFREG.
992 // But there is nothing in sys::fs::perms for S_IFMT or S_IFREG.
994 if (Mode & sys::fs::owner_read)
998 if (Mode & sys::fs::owner_write)
1002 if (Mode & sys::fs::owner_exe)
1006 if (Mode & sys::fs::group_read)
1010 if (Mode & sys::fs::group_write)
1014 if (Mode & sys::fs::group_exe)
1018 if (Mode & sys::fs::others_read)
1022 if (Mode & sys::fs::others_write)
1026 if (Mode & sys::fs::others_exe)
1031 outs() << format("0%o ", Mode);
1034 unsigned UID = C.getUID();
1035 outs() << format("%3d/", UID);
1036 unsigned GID = C.getGID();
1037 outs() << format("%-3d ", GID);
1038 uint64_t Size = C.getRawSize();
1039 outs() << format("%5" PRId64, Size) << " ";
1041 StringRef RawLastModified = C.getRawLastModified();
1044 if (RawLastModified.getAsInteger(10, Seconds))
1045 outs() << "(date: \"%s\" contains non-decimal chars) " << RawLastModified;
1047 // Since cime(3) returns a 26 character string of the form:
1048 // "Sun Sep 16 01:03:52 1973\n\0"
1049 // just print 24 characters.
1051 outs() << format("%.24s ", ctime(&t));
1054 outs() << RawLastModified << " ";
1058 ErrorOr<StringRef> NameOrErr = C.getName();
1059 if (NameOrErr.getError()) {
1060 StringRef RawName = C.getRawName();
1061 outs() << RawName << "\n";
1063 StringRef Name = NameOrErr.get();
1064 outs() << Name << "\n";
1067 StringRef RawName = C.getRawName();
1068 outs() << RawName << "\n";
1072 static void printArchiveHeaders(Archive *A, bool verbose, bool print_offset) {
1073 if (A->hasSymbolTable()) {
1074 Archive::child_iterator S = A->getSymbolTableChild();
1075 Archive::Child C = *S;
1076 printArchiveChild(C, verbose, print_offset);
1078 for (Archive::child_iterator I = A->child_begin(), E = A->child_end(); I != E;
1080 Archive::Child C = *I;
1081 printArchiveChild(C, verbose, print_offset);
1085 // ParseInputMachO() parses the named Mach-O file in Filename and handles the
1086 // -arch flags selecting just those slices as specified by them and also parses
1087 // archive files. Then for each individual Mach-O file ProcessMachO() is
1088 // called to process the file based on the command line options.
1089 void llvm::ParseInputMachO(StringRef Filename) {
1090 // Check for -arch all and verifiy the -arch flags are valid.
1091 for (unsigned i = 0; i < ArchFlags.size(); ++i) {
1092 if (ArchFlags[i] == "all") {
1095 if (!MachOObjectFile::isValidArch(ArchFlags[i])) {
1096 errs() << "llvm-objdump: Unknown architecture named '" + ArchFlags[i] +
1097 "'for the -arch option\n";
1103 // Attempt to open the binary.
1104 ErrorOr<OwningBinary<Binary>> BinaryOrErr = createBinary(Filename);
1105 if (std::error_code EC = BinaryOrErr.getError()) {
1106 errs() << "llvm-objdump: '" << Filename << "': " << EC.message() << ".\n";
1109 Binary &Bin = *BinaryOrErr.get().getBinary();
1111 if (Archive *A = dyn_cast<Archive>(&Bin)) {
1112 outs() << "Archive : " << Filename << "\n";
1114 printArchiveHeaders(A, true, false);
1115 for (Archive::child_iterator I = A->child_begin(), E = A->child_end();
1117 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = I->getAsBinary();
1118 if (ChildOrErr.getError())
1120 if (MachOObjectFile *O = dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
1121 if (!checkMachOAndArchFlags(O, Filename))
1123 ProcessMachO(Filename, O, O->getFileName());
1128 if (UniversalHeaders) {
1129 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin))
1130 printMachOUniversalHeaders(UB, true);
1132 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin)) {
1133 // If we have a list of architecture flags specified dump only those.
1134 if (!ArchAll && ArchFlags.size() != 0) {
1135 // Look for a slice in the universal binary that matches each ArchFlag.
1137 for (unsigned i = 0; i < ArchFlags.size(); ++i) {
1139 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1140 E = UB->end_objects();
1142 if (ArchFlags[i] == I->getArchTypeName()) {
1144 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr =
1145 I->getAsObjectFile();
1146 std::string ArchitectureName = "";
1147 if (ArchFlags.size() > 1)
1148 ArchitectureName = I->getArchTypeName();
1150 ObjectFile &O = *ObjOrErr.get();
1151 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
1152 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
1153 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
1154 I->getAsArchive()) {
1155 std::unique_ptr<Archive> &A = *AOrErr;
1156 outs() << "Archive : " << Filename;
1157 if (!ArchitectureName.empty())
1158 outs() << " (architecture " << ArchitectureName << ")";
1161 printArchiveHeaders(A.get(), true, false);
1162 for (Archive::child_iterator AI = A->child_begin(),
1163 AE = A->child_end();
1165 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1166 if (ChildOrErr.getError())
1168 if (MachOObjectFile *O =
1169 dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
1170 ProcessMachO(Filename, O, O->getFileName(), ArchitectureName);
1176 errs() << "llvm-objdump: file: " + Filename + " does not contain "
1177 << "architecture: " + ArchFlags[i] + "\n";
1183 // No architecture flags were specified so if this contains a slice that
1184 // matches the host architecture dump only that.
1186 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1187 E = UB->end_objects();
1189 if (MachOObjectFile::getHostArch().getArchName() ==
1190 I->getArchTypeName()) {
1191 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
1192 std::string ArchiveName;
1193 ArchiveName.clear();
1195 ObjectFile &O = *ObjOrErr.get();
1196 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
1197 ProcessMachO(Filename, MachOOF);
1198 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
1199 I->getAsArchive()) {
1200 std::unique_ptr<Archive> &A = *AOrErr;
1201 outs() << "Archive : " << Filename << "\n";
1203 printArchiveHeaders(A.get(), true, false);
1204 for (Archive::child_iterator AI = A->child_begin(),
1205 AE = A->child_end();
1207 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1208 if (ChildOrErr.getError())
1210 if (MachOObjectFile *O =
1211 dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
1212 ProcessMachO(Filename, O, O->getFileName());
1219 // Either all architectures have been specified or none have been specified
1220 // and this does not contain the host architecture so dump all the slices.
1221 bool moreThanOneArch = UB->getNumberOfObjects() > 1;
1222 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1223 E = UB->end_objects();
1225 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
1226 std::string ArchitectureName = "";
1227 if (moreThanOneArch)
1228 ArchitectureName = I->getArchTypeName();
1230 ObjectFile &Obj = *ObjOrErr.get();
1231 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&Obj))
1232 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
1233 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr = I->getAsArchive()) {
1234 std::unique_ptr<Archive> &A = *AOrErr;
1235 outs() << "Archive : " << Filename;
1236 if (!ArchitectureName.empty())
1237 outs() << " (architecture " << ArchitectureName << ")";
1240 printArchiveHeaders(A.get(), true, false);
1241 for (Archive::child_iterator AI = A->child_begin(), AE = A->child_end();
1243 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1244 if (ChildOrErr.getError())
1246 if (MachOObjectFile *O =
1247 dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
1248 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(O))
1249 ProcessMachO(Filename, MachOOF, MachOOF->getFileName(),
1257 if (ObjectFile *O = dyn_cast<ObjectFile>(&Bin)) {
1258 if (!checkMachOAndArchFlags(O, Filename))
1260 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&*O)) {
1261 ProcessMachO(Filename, MachOOF);
1263 errs() << "llvm-objdump: '" << Filename << "': "
1264 << "Object is not a Mach-O file type.\n";
1266 errs() << "llvm-objdump: '" << Filename << "': "
1267 << "Unrecognized file type.\n";
1270 typedef std::pair<uint64_t, const char *> BindInfoEntry;
1271 typedef std::vector<BindInfoEntry> BindTable;
1272 typedef BindTable::iterator bind_table_iterator;
1274 // The block of info used by the Symbolizer call backs.
1275 struct DisassembleInfo {
1279 SymbolAddressMap *AddrMap;
1280 std::vector<SectionRef> *Sections;
1281 const char *class_name;
1282 const char *selector_name;
1284 char *demangled_name;
1287 BindTable *bindtable;
1290 // SymbolizerGetOpInfo() is the operand information call back function.
1291 // This is called to get the symbolic information for operand(s) of an
1292 // instruction when it is being done. This routine does this from
1293 // the relocation information, symbol table, etc. That block of information
1294 // is a pointer to the struct DisassembleInfo that was passed when the
1295 // disassembler context was created and passed to back to here when
1296 // called back by the disassembler for instruction operands that could have
1297 // relocation information. The address of the instruction containing operand is
1298 // at the Pc parameter. The immediate value the operand has is passed in
1299 // op_info->Value and is at Offset past the start of the instruction and has a
1300 // byte Size of 1, 2 or 4. The symbolc information is returned in TagBuf is the
1301 // LLVMOpInfo1 struct defined in the header "llvm-c/Disassembler.h" as symbol
1302 // names and addends of the symbolic expression to add for the operand. The
1303 // value of TagType is currently 1 (for the LLVMOpInfo1 struct). If symbolic
1304 // information is returned then this function returns 1 else it returns 0.
1305 int SymbolizerGetOpInfo(void *DisInfo, uint64_t Pc, uint64_t Offset,
1306 uint64_t Size, int TagType, void *TagBuf) {
1307 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
1308 struct LLVMOpInfo1 *op_info = (struct LLVMOpInfo1 *)TagBuf;
1309 uint64_t value = op_info->Value;
1311 // Make sure all fields returned are zero if we don't set them.
1312 memset((void *)op_info, '\0', sizeof(struct LLVMOpInfo1));
1313 op_info->Value = value;
1315 // If the TagType is not the value 1 which it code knows about or if no
1316 // verbose symbolic information is wanted then just return 0, indicating no
1317 // information is being returned.
1318 if (TagType != 1 || info->verbose == false)
1321 unsigned int Arch = info->O->getArch();
1322 if (Arch == Triple::x86) {
1323 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
1325 // First search the section's relocation entries (if any) for an entry
1326 // for this section offset.
1327 uint32_t sect_addr = info->S.getAddress();
1328 uint32_t sect_offset = (Pc + Offset) - sect_addr;
1329 bool reloc_found = false;
1331 MachO::any_relocation_info RE;
1332 bool isExtern = false;
1334 bool r_scattered = false;
1335 uint32_t r_value, pair_r_value, r_type;
1336 for (const RelocationRef &Reloc : info->S.relocations()) {
1337 uint64_t RelocOffset;
1338 Reloc.getOffset(RelocOffset);
1339 if (RelocOffset == sect_offset) {
1340 Rel = Reloc.getRawDataRefImpl();
1341 RE = info->O->getRelocation(Rel);
1342 r_type = info->O->getAnyRelocationType(RE);
1343 r_scattered = info->O->isRelocationScattered(RE);
1345 r_value = info->O->getScatteredRelocationValue(RE);
1346 if (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
1347 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF) {
1348 DataRefImpl RelNext = Rel;
1349 info->O->moveRelocationNext(RelNext);
1350 MachO::any_relocation_info RENext;
1351 RENext = info->O->getRelocation(RelNext);
1352 if (info->O->isRelocationScattered(RENext))
1353 pair_r_value = info->O->getScatteredRelocationValue(RENext);
1358 isExtern = info->O->getPlainRelocationExternal(RE);
1360 symbol_iterator RelocSym = Reloc.getSymbol();
1368 if (reloc_found && isExtern) {
1370 Symbol.getName(SymName);
1371 const char *name = SymName.data();
1372 op_info->AddSymbol.Present = 1;
1373 op_info->AddSymbol.Name = name;
1374 // For i386 extern relocation entries the value in the instruction is
1375 // the offset from the symbol, and value is already set in op_info->Value.
1378 if (reloc_found && (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
1379 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF)) {
1380 const char *add = GuessSymbolName(r_value, info->AddrMap);
1381 const char *sub = GuessSymbolName(pair_r_value, info->AddrMap);
1382 uint32_t offset = value - (r_value - pair_r_value);
1383 op_info->AddSymbol.Present = 1;
1385 op_info->AddSymbol.Name = add;
1387 op_info->AddSymbol.Value = r_value;
1388 op_info->SubtractSymbol.Present = 1;
1390 op_info->SubtractSymbol.Name = sub;
1392 op_info->SubtractSymbol.Value = pair_r_value;
1393 op_info->Value = offset;
1397 // Second search the external relocation entries of a fully linked image
1398 // (if any) for an entry that matches this segment offset.
1399 // uint32_t seg_offset = (Pc + Offset);
1401 } else if (Arch == Triple::x86_64) {
1402 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
1404 // First search the section's relocation entries (if any) for an entry
1405 // for this section offset.
1406 uint64_t sect_addr = info->S.getAddress();
1407 uint64_t sect_offset = (Pc + Offset) - sect_addr;
1408 bool reloc_found = false;
1410 MachO::any_relocation_info RE;
1411 bool isExtern = false;
1413 for (const RelocationRef &Reloc : info->S.relocations()) {
1414 uint64_t RelocOffset;
1415 Reloc.getOffset(RelocOffset);
1416 if (RelocOffset == sect_offset) {
1417 Rel = Reloc.getRawDataRefImpl();
1418 RE = info->O->getRelocation(Rel);
1419 // NOTE: Scattered relocations don't exist on x86_64.
1420 isExtern = info->O->getPlainRelocationExternal(RE);
1422 symbol_iterator RelocSym = Reloc.getSymbol();
1429 if (reloc_found && isExtern) {
1430 // The Value passed in will be adjusted by the Pc if the instruction
1431 // adds the Pc. But for x86_64 external relocation entries the Value
1432 // is the offset from the external symbol.
1433 if (info->O->getAnyRelocationPCRel(RE))
1434 op_info->Value -= Pc + Offset + Size;
1436 Symbol.getName(SymName);
1437 const char *name = SymName.data();
1438 unsigned Type = info->O->getAnyRelocationType(RE);
1439 if (Type == MachO::X86_64_RELOC_SUBTRACTOR) {
1440 DataRefImpl RelNext = Rel;
1441 info->O->moveRelocationNext(RelNext);
1442 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
1443 unsigned TypeNext = info->O->getAnyRelocationType(RENext);
1444 bool isExternNext = info->O->getPlainRelocationExternal(RENext);
1445 unsigned SymbolNum = info->O->getPlainRelocationSymbolNum(RENext);
1446 if (TypeNext == MachO::X86_64_RELOC_UNSIGNED && isExternNext) {
1447 op_info->SubtractSymbol.Present = 1;
1448 op_info->SubtractSymbol.Name = name;
1449 symbol_iterator RelocSymNext = info->O->getSymbolByIndex(SymbolNum);
1450 Symbol = *RelocSymNext;
1451 StringRef SymNameNext;
1452 Symbol.getName(SymNameNext);
1453 name = SymNameNext.data();
1456 // TODO: add the VariantKinds to op_info->VariantKind for relocation types
1457 // like: X86_64_RELOC_TLV, X86_64_RELOC_GOT_LOAD and X86_64_RELOC_GOT.
1458 op_info->AddSymbol.Present = 1;
1459 op_info->AddSymbol.Name = name;
1463 // Second search the external relocation entries of a fully linked image
1464 // (if any) for an entry that matches this segment offset.
1465 // uint64_t seg_offset = (Pc + Offset);
1467 } else if (Arch == Triple::arm) {
1468 if (Offset != 0 || (Size != 4 && Size != 2))
1470 // First search the section's relocation entries (if any) for an entry
1471 // for this section offset.
1472 uint32_t sect_addr = info->S.getAddress();
1473 uint32_t sect_offset = (Pc + Offset) - sect_addr;
1474 bool reloc_found = false;
1476 MachO::any_relocation_info RE;
1477 bool isExtern = false;
1479 bool r_scattered = false;
1480 uint32_t r_value, pair_r_value, r_type, r_length, other_half;
1481 for (const RelocationRef &Reloc : info->S.relocations()) {
1482 uint64_t RelocOffset;
1483 Reloc.getOffset(RelocOffset);
1484 if (RelocOffset == sect_offset) {
1485 Rel = Reloc.getRawDataRefImpl();
1486 RE = info->O->getRelocation(Rel);
1487 r_length = info->O->getAnyRelocationLength(RE);
1488 r_scattered = info->O->isRelocationScattered(RE);
1490 r_value = info->O->getScatteredRelocationValue(RE);
1491 r_type = info->O->getScatteredRelocationType(RE);
1493 r_type = info->O->getAnyRelocationType(RE);
1494 isExtern = info->O->getPlainRelocationExternal(RE);
1496 symbol_iterator RelocSym = Reloc.getSymbol();
1500 if (r_type == MachO::ARM_RELOC_HALF ||
1501 r_type == MachO::ARM_RELOC_SECTDIFF ||
1502 r_type == MachO::ARM_RELOC_LOCAL_SECTDIFF ||
1503 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1504 DataRefImpl RelNext = Rel;
1505 info->O->moveRelocationNext(RelNext);
1506 MachO::any_relocation_info RENext;
1507 RENext = info->O->getRelocation(RelNext);
1508 other_half = info->O->getAnyRelocationAddress(RENext) & 0xffff;
1509 if (info->O->isRelocationScattered(RENext))
1510 pair_r_value = info->O->getScatteredRelocationValue(RENext);
1516 if (reloc_found && isExtern) {
1518 Symbol.getName(SymName);
1519 const char *name = SymName.data();
1520 op_info->AddSymbol.Present = 1;
1521 op_info->AddSymbol.Name = name;
1523 case MachO::ARM_RELOC_HALF:
1524 if ((r_length & 0x1) == 1) {
1525 op_info->Value = value << 16 | other_half;
1526 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1528 op_info->Value = other_half << 16 | value;
1529 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1538 // If we have a branch that is not an external relocation entry then
1539 // return 0 so the code in tryAddingSymbolicOperand() can use the
1540 // SymbolLookUp call back with the branch target address to look up the
1541 // symbol and possiblity add an annotation for a symbol stub.
1542 if (reloc_found && isExtern == 0 && (r_type == MachO::ARM_RELOC_BR24 ||
1543 r_type == MachO::ARM_THUMB_RELOC_BR22))
1546 uint32_t offset = 0;
1548 if (r_type == MachO::ARM_RELOC_HALF ||
1549 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1550 if ((r_length & 0x1) == 1)
1551 value = value << 16 | other_half;
1553 value = other_half << 16 | value;
1555 if (r_scattered && (r_type != MachO::ARM_RELOC_HALF &&
1556 r_type != MachO::ARM_RELOC_HALF_SECTDIFF)) {
1557 offset = value - r_value;
1562 if (reloc_found && r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1563 if ((r_length & 0x1) == 1)
1564 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1566 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1567 const char *add = GuessSymbolName(r_value, info->AddrMap);
1568 const char *sub = GuessSymbolName(pair_r_value, info->AddrMap);
1569 int32_t offset = value - (r_value - pair_r_value);
1570 op_info->AddSymbol.Present = 1;
1572 op_info->AddSymbol.Name = add;
1574 op_info->AddSymbol.Value = r_value;
1575 op_info->SubtractSymbol.Present = 1;
1577 op_info->SubtractSymbol.Name = sub;
1579 op_info->SubtractSymbol.Value = pair_r_value;
1580 op_info->Value = offset;
1584 if (reloc_found == false)
1587 op_info->AddSymbol.Present = 1;
1588 op_info->Value = offset;
1590 if (r_type == MachO::ARM_RELOC_HALF) {
1591 if ((r_length & 0x1) == 1)
1592 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1594 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1597 const char *add = GuessSymbolName(value, info->AddrMap);
1598 if (add != nullptr) {
1599 op_info->AddSymbol.Name = add;
1602 op_info->AddSymbol.Value = value;
1604 } else if (Arch == Triple::aarch64) {
1605 if (Offset != 0 || Size != 4)
1607 // First search the section's relocation entries (if any) for an entry
1608 // for this section offset.
1609 uint64_t sect_addr = info->S.getAddress();
1610 uint64_t sect_offset = (Pc + Offset) - sect_addr;
1611 bool reloc_found = false;
1613 MachO::any_relocation_info RE;
1614 bool isExtern = false;
1616 uint32_t r_type = 0;
1617 for (const RelocationRef &Reloc : info->S.relocations()) {
1618 uint64_t RelocOffset;
1619 Reloc.getOffset(RelocOffset);
1620 if (RelocOffset == sect_offset) {
1621 Rel = Reloc.getRawDataRefImpl();
1622 RE = info->O->getRelocation(Rel);
1623 r_type = info->O->getAnyRelocationType(RE);
1624 if (r_type == MachO::ARM64_RELOC_ADDEND) {
1625 DataRefImpl RelNext = Rel;
1626 info->O->moveRelocationNext(RelNext);
1627 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
1629 value = info->O->getPlainRelocationSymbolNum(RENext);
1630 op_info->Value = value;
1633 // NOTE: Scattered relocations don't exist on arm64.
1634 isExtern = info->O->getPlainRelocationExternal(RE);
1636 symbol_iterator RelocSym = Reloc.getSymbol();
1643 if (reloc_found && isExtern) {
1645 Symbol.getName(SymName);
1646 const char *name = SymName.data();
1647 op_info->AddSymbol.Present = 1;
1648 op_info->AddSymbol.Name = name;
1651 case MachO::ARM64_RELOC_PAGE21:
1653 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGE;
1655 case MachO::ARM64_RELOC_PAGEOFF12:
1657 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGEOFF;
1659 case MachO::ARM64_RELOC_GOT_LOAD_PAGE21:
1661 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGE;
1663 case MachO::ARM64_RELOC_GOT_LOAD_PAGEOFF12:
1665 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGEOFF;
1667 case MachO::ARM64_RELOC_TLVP_LOAD_PAGE21:
1668 /* @tvlppage is not implemented in llvm-mc */
1669 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVP;
1671 case MachO::ARM64_RELOC_TLVP_LOAD_PAGEOFF12:
1672 /* @tvlppageoff is not implemented in llvm-mc */
1673 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVOFF;
1676 case MachO::ARM64_RELOC_BRANCH26:
1677 op_info->VariantKind = LLVMDisassembler_VariantKind_None;
1688 // GuessCstringPointer is passed the address of what might be a pointer to a
1689 // literal string in a cstring section. If that address is in a cstring section
1690 // it returns a pointer to that string. Else it returns nullptr.
1691 const char *GuessCstringPointer(uint64_t ReferenceValue,
1692 struct DisassembleInfo *info) {
1693 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1694 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1695 for (unsigned I = 0;; ++I) {
1696 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1697 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1698 for (unsigned J = 0; J < Seg.nsects; ++J) {
1699 MachO::section_64 Sec = info->O->getSection64(Load, J);
1700 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1701 if (section_type == MachO::S_CSTRING_LITERALS &&
1702 ReferenceValue >= Sec.addr &&
1703 ReferenceValue < Sec.addr + Sec.size) {
1704 uint64_t sect_offset = ReferenceValue - Sec.addr;
1705 uint64_t object_offset = Sec.offset + sect_offset;
1706 StringRef MachOContents = info->O->getData();
1707 uint64_t object_size = MachOContents.size();
1708 const char *object_addr = (const char *)MachOContents.data();
1709 if (object_offset < object_size) {
1710 const char *name = object_addr + object_offset;
1717 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
1718 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
1719 for (unsigned J = 0; J < Seg.nsects; ++J) {
1720 MachO::section Sec = info->O->getSection(Load, J);
1721 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1722 if (section_type == MachO::S_CSTRING_LITERALS &&
1723 ReferenceValue >= Sec.addr &&
1724 ReferenceValue < Sec.addr + Sec.size) {
1725 uint64_t sect_offset = ReferenceValue - Sec.addr;
1726 uint64_t object_offset = Sec.offset + sect_offset;
1727 StringRef MachOContents = info->O->getData();
1728 uint64_t object_size = MachOContents.size();
1729 const char *object_addr = (const char *)MachOContents.data();
1730 if (object_offset < object_size) {
1731 const char *name = object_addr + object_offset;
1739 if (I == LoadCommandCount - 1)
1742 Load = info->O->getNextLoadCommandInfo(Load);
1747 // GuessIndirectSymbol returns the name of the indirect symbol for the
1748 // ReferenceValue passed in or nullptr. This is used when ReferenceValue maybe
1749 // an address of a symbol stub or a lazy or non-lazy pointer to associate the
1750 // symbol name being referenced by the stub or pointer.
1751 static const char *GuessIndirectSymbol(uint64_t ReferenceValue,
1752 struct DisassembleInfo *info) {
1753 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1754 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1755 MachO::dysymtab_command Dysymtab = info->O->getDysymtabLoadCommand();
1756 MachO::symtab_command Symtab = info->O->getSymtabLoadCommand();
1757 for (unsigned I = 0;; ++I) {
1758 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1759 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1760 for (unsigned J = 0; J < Seg.nsects; ++J) {
1761 MachO::section_64 Sec = info->O->getSection64(Load, J);
1762 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1763 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
1764 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
1765 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
1766 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
1767 section_type == MachO::S_SYMBOL_STUBS) &&
1768 ReferenceValue >= Sec.addr &&
1769 ReferenceValue < Sec.addr + Sec.size) {
1771 if (section_type == MachO::S_SYMBOL_STUBS)
1772 stride = Sec.reserved2;
1777 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
1778 if (index < Dysymtab.nindirectsyms) {
1779 uint32_t indirect_symbol =
1780 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
1781 if (indirect_symbol < Symtab.nsyms) {
1782 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
1783 SymbolRef Symbol = *Sym;
1785 Symbol.getName(SymName);
1786 const char *name = SymName.data();
1792 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
1793 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
1794 for (unsigned J = 0; J < Seg.nsects; ++J) {
1795 MachO::section Sec = info->O->getSection(Load, J);
1796 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1797 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
1798 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
1799 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
1800 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
1801 section_type == MachO::S_SYMBOL_STUBS) &&
1802 ReferenceValue >= Sec.addr &&
1803 ReferenceValue < Sec.addr + Sec.size) {
1805 if (section_type == MachO::S_SYMBOL_STUBS)
1806 stride = Sec.reserved2;
1811 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
1812 if (index < Dysymtab.nindirectsyms) {
1813 uint32_t indirect_symbol =
1814 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
1815 if (indirect_symbol < Symtab.nsyms) {
1816 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
1817 SymbolRef Symbol = *Sym;
1819 Symbol.getName(SymName);
1820 const char *name = SymName.data();
1827 if (I == LoadCommandCount - 1)
1830 Load = info->O->getNextLoadCommandInfo(Load);
1835 // method_reference() is called passing it the ReferenceName that might be
1836 // a reference it to an Objective-C method call. If so then it allocates and
1837 // assembles a method call string with the values last seen and saved in
1838 // the DisassembleInfo's class_name and selector_name fields. This is saved
1839 // into the method field of the info and any previous string is free'ed.
1840 // Then the class_name field in the info is set to nullptr. The method call
1841 // string is set into ReferenceName and ReferenceType is set to
1842 // LLVMDisassembler_ReferenceType_Out_Objc_Message. If this not a method call
1843 // then both ReferenceType and ReferenceName are left unchanged.
1844 static void method_reference(struct DisassembleInfo *info,
1845 uint64_t *ReferenceType,
1846 const char **ReferenceName) {
1847 unsigned int Arch = info->O->getArch();
1848 if (*ReferenceName != nullptr) {
1849 if (strcmp(*ReferenceName, "_objc_msgSend") == 0) {
1850 if (info->selector_name != nullptr) {
1851 if (info->method != nullptr)
1853 if (info->class_name != nullptr) {
1854 info->method = (char *)malloc(5 + strlen(info->class_name) +
1855 strlen(info->selector_name));
1856 if (info->method != nullptr) {
1857 strcpy(info->method, "+[");
1858 strcat(info->method, info->class_name);
1859 strcat(info->method, " ");
1860 strcat(info->method, info->selector_name);
1861 strcat(info->method, "]");
1862 *ReferenceName = info->method;
1863 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1866 info->method = (char *)malloc(9 + strlen(info->selector_name));
1867 if (info->method != nullptr) {
1868 if (Arch == Triple::x86_64)
1869 strcpy(info->method, "-[%rdi ");
1870 else if (Arch == Triple::aarch64)
1871 strcpy(info->method, "-[x0 ");
1873 strcpy(info->method, "-[r? ");
1874 strcat(info->method, info->selector_name);
1875 strcat(info->method, "]");
1876 *ReferenceName = info->method;
1877 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1880 info->class_name = nullptr;
1882 } else if (strcmp(*ReferenceName, "_objc_msgSendSuper2") == 0) {
1883 if (info->selector_name != nullptr) {
1884 if (info->method != nullptr)
1886 info->method = (char *)malloc(17 + strlen(info->selector_name));
1887 if (info->method != nullptr) {
1888 if (Arch == Triple::x86_64)
1889 strcpy(info->method, "-[[%rdi super] ");
1890 else if (Arch == Triple::aarch64)
1891 strcpy(info->method, "-[[x0 super] ");
1893 strcpy(info->method, "-[[r? super] ");
1894 strcat(info->method, info->selector_name);
1895 strcat(info->method, "]");
1896 *ReferenceName = info->method;
1897 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1899 info->class_name = nullptr;
1905 // GuessPointerPointer() is passed the address of what might be a pointer to
1906 // a reference to an Objective-C class, selector, message ref or cfstring.
1907 // If so the value of the pointer is returned and one of the booleans are set
1908 // to true. If not zero is returned and all the booleans are set to false.
1909 static uint64_t GuessPointerPointer(uint64_t ReferenceValue,
1910 struct DisassembleInfo *info,
1911 bool &classref, bool &selref, bool &msgref,
1917 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1918 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1919 for (unsigned I = 0;; ++I) {
1920 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1921 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1922 for (unsigned J = 0; J < Seg.nsects; ++J) {
1923 MachO::section_64 Sec = info->O->getSection64(Load, J);
1924 if ((strncmp(Sec.sectname, "__objc_selrefs", 16) == 0 ||
1925 strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
1926 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0 ||
1927 strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 ||
1928 strncmp(Sec.sectname, "__cfstring", 16) == 0) &&
1929 ReferenceValue >= Sec.addr &&
1930 ReferenceValue < Sec.addr + Sec.size) {
1931 uint64_t sect_offset = ReferenceValue - Sec.addr;
1932 uint64_t object_offset = Sec.offset + sect_offset;
1933 StringRef MachOContents = info->O->getData();
1934 uint64_t object_size = MachOContents.size();
1935 const char *object_addr = (const char *)MachOContents.data();
1936 if (object_offset < object_size) {
1937 uint64_t pointer_value;
1938 memcpy(&pointer_value, object_addr + object_offset,
1940 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1941 sys::swapByteOrder(pointer_value);
1942 if (strncmp(Sec.sectname, "__objc_selrefs", 16) == 0)
1944 else if (strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
1945 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0)
1947 else if (strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 &&
1948 ReferenceValue + 8 < Sec.addr + Sec.size) {
1950 memcpy(&pointer_value, object_addr + object_offset + 8,
1952 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1953 sys::swapByteOrder(pointer_value);
1954 } else if (strncmp(Sec.sectname, "__cfstring", 16) == 0)
1956 return pointer_value;
1963 // TODO: Look for LC_SEGMENT for 32-bit Mach-O files.
1964 if (I == LoadCommandCount - 1)
1967 Load = info->O->getNextLoadCommandInfo(Load);
1972 // get_pointer_64 returns a pointer to the bytes in the object file at the
1973 // Address from a section in the Mach-O file. And indirectly returns the
1974 // offset into the section, number of bytes left in the section past the offset
1975 // and which section is was being referenced. If the Address is not in a
1976 // section nullptr is returned.
1977 const char *get_pointer_64(uint64_t Address, uint32_t &offset, uint32_t &left,
1978 SectionRef &S, DisassembleInfo *info) {
1982 for (unsigned SectIdx = 0; SectIdx != info->Sections->size(); SectIdx++) {
1983 uint64_t SectAddress = ((*(info->Sections))[SectIdx]).getAddress();
1984 uint64_t SectSize = ((*(info->Sections))[SectIdx]).getSize();
1985 if (Address >= SectAddress && Address < SectAddress + SectSize) {
1986 S = (*(info->Sections))[SectIdx];
1987 offset = Address - SectAddress;
1988 left = SectSize - offset;
1989 StringRef SectContents;
1990 ((*(info->Sections))[SectIdx]).getContents(SectContents);
1991 return SectContents.data() + offset;
1997 // get_symbol_64() returns the name of a symbol (or nullptr) and the address of
1998 // the symbol indirectly through n_value. Based on the relocation information
1999 // for the specified section offset in the specified section reference.
2000 const char *get_symbol_64(uint32_t sect_offset, SectionRef S,
2001 DisassembleInfo *info, uint64_t &n_value) {
2003 if (info->verbose == false)
2006 // See if there is an external relocation entry at the sect_offset.
2007 bool reloc_found = false;
2009 MachO::any_relocation_info RE;
2010 bool isExtern = false;
2012 for (const RelocationRef &Reloc : S.relocations()) {
2013 uint64_t RelocOffset;
2014 Reloc.getOffset(RelocOffset);
2015 if (RelocOffset == sect_offset) {
2016 Rel = Reloc.getRawDataRefImpl();
2017 RE = info->O->getRelocation(Rel);
2018 if (info->O->isRelocationScattered(RE))
2020 isExtern = info->O->getPlainRelocationExternal(RE);
2022 symbol_iterator RelocSym = Reloc.getSymbol();
2029 // If there is an external relocation entry for a symbol in this section
2030 // at this section_offset then use that symbol's value for the n_value
2031 // and return its name.
2032 const char *SymbolName = nullptr;
2033 if (reloc_found && isExtern) {
2034 Symbol.getAddress(n_value);
2036 Symbol.getName(name);
2037 if (!name.empty()) {
2038 SymbolName = name.data();
2043 // TODO: For fully linked images, look through the external relocation
2044 // entries off the dynamic symtab command. For these the r_offset is from the
2045 // start of the first writeable segment in the Mach-O file. So the offset
2046 // to this section from that segment is passed to this routine by the caller,
2047 // as the database_offset. Which is the difference of the section's starting
2048 // address and the first writable segment.
2050 // NOTE: need add passing the database_offset to this routine.
2052 // TODO: We did not find an external relocation entry so look up the
2053 // ReferenceValue as an address of a symbol and if found return that symbol's
2056 // NOTE: need add passing the ReferenceValue to this routine. Then that code
2057 // would simply be this:
2058 // SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap);
2063 // These are structs in the Objective-C meta data and read to produce the
2064 // comments for disassembly. While these are part of the ABI they are no
2065 // public defintions. So the are here not in include/llvm/Support/MachO.h .
2067 // The cfstring object in a 64-bit Mach-O file.
2068 struct cfstring64_t {
2069 uint64_t isa; // class64_t * (64-bit pointer)
2070 uint64_t flags; // flag bits
2071 uint64_t characters; // char * (64-bit pointer)
2072 uint64_t length; // number of non-NULL characters in above
2075 // The class object in a 64-bit Mach-O file.
2077 uint64_t isa; // class64_t * (64-bit pointer)
2078 uint64_t superclass; // class64_t * (64-bit pointer)
2079 uint64_t cache; // Cache (64-bit pointer)
2080 uint64_t vtable; // IMP * (64-bit pointer)
2081 uint64_t data; // class_ro64_t * (64-bit pointer)
2084 struct class_ro64_t {
2086 uint32_t instanceStart;
2087 uint32_t instanceSize;
2089 uint64_t ivarLayout; // const uint8_t * (64-bit pointer)
2090 uint64_t name; // const char * (64-bit pointer)
2091 uint64_t baseMethods; // const method_list_t * (64-bit pointer)
2092 uint64_t baseProtocols; // const protocol_list_t * (64-bit pointer)
2093 uint64_t ivars; // const ivar_list_t * (64-bit pointer)
2094 uint64_t weakIvarLayout; // const uint8_t * (64-bit pointer)
2095 uint64_t baseProperties; // const struct objc_property_list (64-bit pointer)
2098 inline void swapStruct(struct cfstring64_t &cfs) {
2099 sys::swapByteOrder(cfs.isa);
2100 sys::swapByteOrder(cfs.flags);
2101 sys::swapByteOrder(cfs.characters);
2102 sys::swapByteOrder(cfs.length);
2105 inline void swapStruct(struct class64_t &c) {
2106 sys::swapByteOrder(c.isa);
2107 sys::swapByteOrder(c.superclass);
2108 sys::swapByteOrder(c.cache);
2109 sys::swapByteOrder(c.vtable);
2110 sys::swapByteOrder(c.data);
2113 inline void swapStruct(struct class_ro64_t &cro) {
2114 sys::swapByteOrder(cro.flags);
2115 sys::swapByteOrder(cro.instanceStart);
2116 sys::swapByteOrder(cro.instanceSize);
2117 sys::swapByteOrder(cro.reserved);
2118 sys::swapByteOrder(cro.ivarLayout);
2119 sys::swapByteOrder(cro.name);
2120 sys::swapByteOrder(cro.baseMethods);
2121 sys::swapByteOrder(cro.baseProtocols);
2122 sys::swapByteOrder(cro.ivars);
2123 sys::swapByteOrder(cro.weakIvarLayout);
2124 sys::swapByteOrder(cro.baseProperties);
2127 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
2128 struct DisassembleInfo *info);
2130 // get_objc2_64bit_class_name() is used for disassembly and is passed a pointer
2131 // to an Objective-C class and returns the class name. It is also passed the
2132 // address of the pointer, so when the pointer is zero as it can be in an .o
2133 // file, that is used to look for an external relocation entry with a symbol
2135 const char *get_objc2_64bit_class_name(uint64_t pointer_value,
2136 uint64_t ReferenceValue,
2137 struct DisassembleInfo *info) {
2139 uint32_t offset, left;
2142 // The pointer_value can be 0 in an object file and have a relocation
2143 // entry for the class symbol at the ReferenceValue (the address of the
2145 if (pointer_value == 0) {
2146 r = get_pointer_64(ReferenceValue, offset, left, S, info);
2147 if (r == nullptr || left < sizeof(uint64_t))
2150 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
2151 if (symbol_name == nullptr)
2153 const char *class_name = strrchr(symbol_name, '$');
2154 if (class_name != nullptr && class_name[1] == '_' && class_name[2] != '\0')
2155 return class_name + 2;
2160 // The case were the pointer_value is non-zero and points to a class defined
2161 // in this Mach-O file.
2162 r = get_pointer_64(pointer_value, offset, left, S, info);
2163 if (r == nullptr || left < sizeof(struct class64_t))
2166 memcpy(&c, r, sizeof(struct class64_t));
2167 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2171 r = get_pointer_64(c.data, offset, left, S, info);
2172 if (r == nullptr || left < sizeof(struct class_ro64_t))
2174 struct class_ro64_t cro;
2175 memcpy(&cro, r, sizeof(struct class_ro64_t));
2176 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2180 const char *name = get_pointer_64(cro.name, offset, left, S, info);
2184 // get_objc2_64bit_cfstring_name is used for disassembly and is passed a
2185 // pointer to a cfstring and returns its name or nullptr.
2186 const char *get_objc2_64bit_cfstring_name(uint64_t ReferenceValue,
2187 struct DisassembleInfo *info) {
2188 const char *r, *name;
2189 uint32_t offset, left;
2191 struct cfstring64_t cfs;
2192 uint64_t cfs_characters;
2194 r = get_pointer_64(ReferenceValue, offset, left, S, info);
2195 if (r == nullptr || left < sizeof(struct cfstring64_t))
2197 memcpy(&cfs, r, sizeof(struct cfstring64_t));
2198 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2200 if (cfs.characters == 0) {
2202 const char *symbol_name = get_symbol_64(
2203 offset + offsetof(struct cfstring64_t, characters), S, info, n_value);
2204 if (symbol_name == nullptr)
2206 cfs_characters = n_value;
2208 cfs_characters = cfs.characters;
2209 name = get_pointer_64(cfs_characters, offset, left, S, info);
2214 // get_objc2_64bit_selref() is used for disassembly and is passed a the address
2215 // of a pointer to an Objective-C selector reference when the pointer value is
2216 // zero as in a .o file and is likely to have a external relocation entry with
2217 // who's symbol's n_value is the real pointer to the selector name. If that is
2218 // the case the real pointer to the selector name is returned else 0 is
2220 uint64_t get_objc2_64bit_selref(uint64_t ReferenceValue,
2221 struct DisassembleInfo *info) {
2222 uint32_t offset, left;
2225 const char *r = get_pointer_64(ReferenceValue, offset, left, S, info);
2226 if (r == nullptr || left < sizeof(uint64_t))
2229 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
2230 if (symbol_name == nullptr)
2235 // GuessLiteralPointer returns a string which for the item in the Mach-O file
2236 // for the address passed in as ReferenceValue for printing as a comment with
2237 // the instruction and also returns the corresponding type of that item
2238 // indirectly through ReferenceType.
2240 // If ReferenceValue is an address of literal cstring then a pointer to the
2241 // cstring is returned and ReferenceType is set to
2242 // LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr .
2244 // If ReferenceValue is an address of an Objective-C CFString, Selector ref or
2245 // Class ref that name is returned and the ReferenceType is set accordingly.
2247 // Lastly, literals which are Symbol address in a literal pool are looked for
2248 // and if found the symbol name is returned and ReferenceType is set to
2249 // LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr .
2251 // If there is no item in the Mach-O file for the address passed in as
2252 // ReferenceValue nullptr is returned and ReferenceType is unchanged.
2253 const char *GuessLiteralPointer(uint64_t ReferenceValue, uint64_t ReferencePC,
2254 uint64_t *ReferenceType,
2255 struct DisassembleInfo *info) {
2256 // First see if there is an external relocation entry at the ReferencePC.
2257 uint64_t sect_addr = info->S.getAddress();
2258 uint64_t sect_offset = ReferencePC - sect_addr;
2259 bool reloc_found = false;
2261 MachO::any_relocation_info RE;
2262 bool isExtern = false;
2264 for (const RelocationRef &Reloc : info->S.relocations()) {
2265 uint64_t RelocOffset;
2266 Reloc.getOffset(RelocOffset);
2267 if (RelocOffset == sect_offset) {
2268 Rel = Reloc.getRawDataRefImpl();
2269 RE = info->O->getRelocation(Rel);
2270 if (info->O->isRelocationScattered(RE))
2272 isExtern = info->O->getPlainRelocationExternal(RE);
2274 symbol_iterator RelocSym = Reloc.getSymbol();
2281 // If there is an external relocation entry for a symbol in a section
2282 // then used that symbol's value for the value of the reference.
2283 if (reloc_found && isExtern) {
2284 if (info->O->getAnyRelocationPCRel(RE)) {
2285 unsigned Type = info->O->getAnyRelocationType(RE);
2286 if (Type == MachO::X86_64_RELOC_SIGNED) {
2287 Symbol.getAddress(ReferenceValue);
2292 // Look for literals such as Objective-C CFStrings refs, Selector refs,
2293 // Message refs and Class refs.
2294 bool classref, selref, msgref, cfstring;
2295 uint64_t pointer_value = GuessPointerPointer(ReferenceValue, info, classref,
2296 selref, msgref, cfstring);
2297 if (classref == true && pointer_value == 0) {
2298 // Note the ReferenceValue is a pointer into the __objc_classrefs section.
2299 // And the pointer_value in that section is typically zero as it will be
2300 // set by dyld as part of the "bind information".
2301 const char *name = get_dyld_bind_info_symbolname(ReferenceValue, info);
2302 if (name != nullptr) {
2303 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
2304 const char *class_name = strrchr(name, '$');
2305 if (class_name != nullptr && class_name[1] == '_' &&
2306 class_name[2] != '\0') {
2307 info->class_name = class_name + 2;
2313 if (classref == true) {
2314 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
2316 get_objc2_64bit_class_name(pointer_value, ReferenceValue, info);
2317 if (name != nullptr)
2318 info->class_name = name;
2320 name = "bad class ref";
2324 if (cfstring == true) {
2325 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_CFString_Ref;
2326 const char *name = get_objc2_64bit_cfstring_name(ReferenceValue, info);
2330 if (selref == true && pointer_value == 0)
2331 pointer_value = get_objc2_64bit_selref(ReferenceValue, info);
2333 if (pointer_value != 0)
2334 ReferenceValue = pointer_value;
2336 const char *name = GuessCstringPointer(ReferenceValue, info);
2338 if (pointer_value != 0 && selref == true) {
2339 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Selector_Ref;
2340 info->selector_name = name;
2341 } else if (pointer_value != 0 && msgref == true) {
2342 info->class_name = nullptr;
2343 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message_Ref;
2344 info->selector_name = name;
2346 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr;
2350 // Lastly look for an indirect symbol with this ReferenceValue which is in
2351 // a literal pool. If found return that symbol name.
2352 name = GuessIndirectSymbol(ReferenceValue, info);
2354 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr;
2361 // SymbolizerSymbolLookUp is the symbol lookup function passed when creating
2362 // the Symbolizer. It looks up the ReferenceValue using the info passed via the
2363 // pointer to the struct DisassembleInfo that was passed when MCSymbolizer
2364 // is created and returns the symbol name that matches the ReferenceValue or
2365 // nullptr if none. The ReferenceType is passed in for the IN type of
2366 // reference the instruction is making from the values in defined in the header
2367 // "llvm-c/Disassembler.h". On return the ReferenceType can set to a specific
2368 // Out type and the ReferenceName will also be set which is added as a comment
2369 // to the disassembled instruction.
2372 // If the symbol name is a C++ mangled name then the demangled name is
2373 // returned through ReferenceName and ReferenceType is set to
2374 // LLVMDisassembler_ReferenceType_DeMangled_Name .
2377 // When this is called to get a symbol name for a branch target then the
2378 // ReferenceType will be LLVMDisassembler_ReferenceType_In_Branch and then
2379 // SymbolValue will be looked for in the indirect symbol table to determine if
2380 // it is an address for a symbol stub. If so then the symbol name for that
2381 // stub is returned indirectly through ReferenceName and then ReferenceType is
2382 // set to LLVMDisassembler_ReferenceType_Out_SymbolStub.
2384 // When this is called with an value loaded via a PC relative load then
2385 // ReferenceType will be LLVMDisassembler_ReferenceType_In_PCrel_Load then the
2386 // SymbolValue is checked to be an address of literal pointer, symbol pointer,
2387 // or an Objective-C meta data reference. If so the output ReferenceType is
2388 // set to correspond to that as well as setting the ReferenceName.
2389 const char *SymbolizerSymbolLookUp(void *DisInfo, uint64_t ReferenceValue,
2390 uint64_t *ReferenceType,
2391 uint64_t ReferencePC,
2392 const char **ReferenceName) {
2393 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
2394 // If no verbose symbolic information is wanted then just return nullptr.
2395 if (info->verbose == false) {
2396 *ReferenceName = nullptr;
2397 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2401 const char *SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap);
2403 if (*ReferenceType == LLVMDisassembler_ReferenceType_In_Branch) {
2404 *ReferenceName = GuessIndirectSymbol(ReferenceValue, info);
2405 if (*ReferenceName != nullptr) {
2406 method_reference(info, ReferenceType, ReferenceName);
2407 if (*ReferenceType != LLVMDisassembler_ReferenceType_Out_Objc_Message)
2408 *ReferenceType = LLVMDisassembler_ReferenceType_Out_SymbolStub;
2411 if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
2412 if (info->demangled_name != nullptr)
2413 free(info->demangled_name);
2415 info->demangled_name =
2416 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
2417 if (info->demangled_name != nullptr) {
2418 *ReferenceName = info->demangled_name;
2419 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
2421 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2424 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2425 } else if (*ReferenceType == LLVMDisassembler_ReferenceType_In_PCrel_Load) {
2427 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2429 method_reference(info, ReferenceType, ReferenceName);
2431 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2432 // If this is arm64 and the reference is an adrp instruction save the
2433 // instruction, passed in ReferenceValue and the address of the instruction
2434 // for use later if we see and add immediate instruction.
2435 } else if (info->O->getArch() == Triple::aarch64 &&
2436 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADRP) {
2437 info->adrp_inst = ReferenceValue;
2438 info->adrp_addr = ReferencePC;
2439 SymbolName = nullptr;
2440 *ReferenceName = nullptr;
2441 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2442 // If this is arm64 and reference is an add immediate instruction and we
2444 // seen an adrp instruction just before it and the adrp's Xd register
2446 // this add's Xn register reconstruct the value being referenced and look to
2447 // see if it is a literal pointer. Note the add immediate instruction is
2448 // passed in ReferenceValue.
2449 } else if (info->O->getArch() == Triple::aarch64 &&
2450 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADDXri &&
2451 ReferencePC - 4 == info->adrp_addr &&
2452 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
2453 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
2454 uint32_t addxri_inst;
2455 uint64_t adrp_imm, addxri_imm;
2458 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
2459 if (info->adrp_inst & 0x0200000)
2460 adrp_imm |= 0xfffffffffc000000LL;
2462 addxri_inst = ReferenceValue;
2463 addxri_imm = (addxri_inst >> 10) & 0xfff;
2464 if (((addxri_inst >> 22) & 0x3) == 1)
2467 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
2468 (adrp_imm << 12) + addxri_imm;
2471 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2472 if (*ReferenceName == nullptr)
2473 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2474 // If this is arm64 and the reference is a load register instruction and we
2475 // have seen an adrp instruction just before it and the adrp's Xd register
2476 // matches this add's Xn register reconstruct the value being referenced and
2477 // look to see if it is a literal pointer. Note the load register
2478 // instruction is passed in ReferenceValue.
2479 } else if (info->O->getArch() == Triple::aarch64 &&
2480 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXui &&
2481 ReferencePC - 4 == info->adrp_addr &&
2482 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
2483 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
2484 uint32_t ldrxui_inst;
2485 uint64_t adrp_imm, ldrxui_imm;
2488 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
2489 if (info->adrp_inst & 0x0200000)
2490 adrp_imm |= 0xfffffffffc000000LL;
2492 ldrxui_inst = ReferenceValue;
2493 ldrxui_imm = (ldrxui_inst >> 10) & 0xfff;
2495 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
2496 (adrp_imm << 12) + (ldrxui_imm << 3);
2499 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2500 if (*ReferenceName == nullptr)
2501 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2503 // If this arm64 and is an load register (PC-relative) instruction the
2504 // ReferenceValue is the PC plus the immediate value.
2505 else if (info->O->getArch() == Triple::aarch64 &&
2506 (*ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXl ||
2507 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADR)) {
2509 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2510 if (*ReferenceName == nullptr)
2511 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2514 else if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
2515 if (info->demangled_name != nullptr)
2516 free(info->demangled_name);
2518 info->demangled_name =
2519 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
2520 if (info->demangled_name != nullptr) {
2521 *ReferenceName = info->demangled_name;
2522 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
2527 *ReferenceName = nullptr;
2528 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2534 /// \brief Emits the comments that are stored in the CommentStream.
2535 /// Each comment in the CommentStream must end with a newline.
2536 static void emitComments(raw_svector_ostream &CommentStream,
2537 SmallString<128> &CommentsToEmit,
2538 formatted_raw_ostream &FormattedOS,
2539 const MCAsmInfo &MAI) {
2540 // Flush the stream before taking its content.
2541 CommentStream.flush();
2542 StringRef Comments = CommentsToEmit.str();
2543 // Get the default information for printing a comment.
2544 const char *CommentBegin = MAI.getCommentString();
2545 unsigned CommentColumn = MAI.getCommentColumn();
2546 bool IsFirst = true;
2547 while (!Comments.empty()) {
2549 FormattedOS << '\n';
2550 // Emit a line of comments.
2551 FormattedOS.PadToColumn(CommentColumn);
2552 size_t Position = Comments.find('\n');
2553 FormattedOS << CommentBegin << ' ' << Comments.substr(0, Position);
2554 // Move after the newline character.
2555 Comments = Comments.substr(Position + 1);
2558 FormattedOS.flush();
2560 // Tell the comment stream that the vector changed underneath it.
2561 CommentsToEmit.clear();
2562 CommentStream.resync();
2565 static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF,
2566 StringRef DisSegName, StringRef DisSectName) {
2567 const char *McpuDefault = nullptr;
2568 const Target *ThumbTarget = nullptr;
2569 const Target *TheTarget = GetTarget(MachOOF, &McpuDefault, &ThumbTarget);
2571 // GetTarget prints out stuff.
2574 if (MCPU.empty() && McpuDefault)
2577 std::unique_ptr<const MCInstrInfo> InstrInfo(TheTarget->createMCInstrInfo());
2578 std::unique_ptr<const MCInstrInfo> ThumbInstrInfo;
2580 ThumbInstrInfo.reset(ThumbTarget->createMCInstrInfo());
2582 // Package up features to be passed to target/subtarget
2583 std::string FeaturesStr;
2584 if (MAttrs.size()) {
2585 SubtargetFeatures Features;
2586 for (unsigned i = 0; i != MAttrs.size(); ++i)
2587 Features.AddFeature(MAttrs[i]);
2588 FeaturesStr = Features.getString();
2591 // Set up disassembler.
2592 std::unique_ptr<const MCRegisterInfo> MRI(
2593 TheTarget->createMCRegInfo(TripleName));
2594 std::unique_ptr<const MCAsmInfo> AsmInfo(
2595 TheTarget->createMCAsmInfo(*MRI, TripleName));
2596 std::unique_ptr<const MCSubtargetInfo> STI(
2597 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr));
2598 MCContext Ctx(AsmInfo.get(), MRI.get(), nullptr);
2599 std::unique_ptr<MCDisassembler> DisAsm(
2600 TheTarget->createMCDisassembler(*STI, Ctx));
2601 std::unique_ptr<MCSymbolizer> Symbolizer;
2602 struct DisassembleInfo SymbolizerInfo;
2603 std::unique_ptr<MCRelocationInfo> RelInfo(
2604 TheTarget->createMCRelocationInfo(TripleName, Ctx));
2606 Symbolizer.reset(TheTarget->createMCSymbolizer(
2607 TripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
2608 &SymbolizerInfo, &Ctx, std::move(RelInfo)));
2609 DisAsm->setSymbolizer(std::move(Symbolizer));
2611 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
2612 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
2613 AsmPrinterVariant, *AsmInfo, *InstrInfo, *MRI, *STI));
2614 // Set the display preference for hex vs. decimal immediates.
2615 IP->setPrintImmHex(PrintImmHex);
2616 // Comment stream and backing vector.
2617 SmallString<128> CommentsToEmit;
2618 raw_svector_ostream CommentStream(CommentsToEmit);
2619 // FIXME: Setting the CommentStream in the InstPrinter is problematic in that
2620 // if it is done then arm64 comments for string literals don't get printed
2621 // and some constant get printed instead and not setting it causes intel
2622 // (32-bit and 64-bit) comments printed with different spacing before the
2623 // comment causing different diffs with the 'C' disassembler library API.
2624 // IP->setCommentStream(CommentStream);
2626 if (!AsmInfo || !STI || !DisAsm || !IP) {
2627 errs() << "error: couldn't initialize disassembler for target "
2628 << TripleName << '\n';
2632 // Set up thumb disassembler.
2633 std::unique_ptr<const MCRegisterInfo> ThumbMRI;
2634 std::unique_ptr<const MCAsmInfo> ThumbAsmInfo;
2635 std::unique_ptr<const MCSubtargetInfo> ThumbSTI;
2636 std::unique_ptr<MCDisassembler> ThumbDisAsm;
2637 std::unique_ptr<MCInstPrinter> ThumbIP;
2638 std::unique_ptr<MCContext> ThumbCtx;
2639 std::unique_ptr<MCSymbolizer> ThumbSymbolizer;
2640 struct DisassembleInfo ThumbSymbolizerInfo;
2641 std::unique_ptr<MCRelocationInfo> ThumbRelInfo;
2643 ThumbMRI.reset(ThumbTarget->createMCRegInfo(ThumbTripleName));
2645 ThumbTarget->createMCAsmInfo(*ThumbMRI, ThumbTripleName));
2647 ThumbTarget->createMCSubtargetInfo(ThumbTripleName, MCPU, FeaturesStr));
2648 ThumbCtx.reset(new MCContext(ThumbAsmInfo.get(), ThumbMRI.get(), nullptr));
2649 ThumbDisAsm.reset(ThumbTarget->createMCDisassembler(*ThumbSTI, *ThumbCtx));
2650 MCContext *PtrThumbCtx = ThumbCtx.get();
2652 ThumbTarget->createMCRelocationInfo(ThumbTripleName, *PtrThumbCtx));
2654 ThumbSymbolizer.reset(ThumbTarget->createMCSymbolizer(
2655 ThumbTripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
2656 &ThumbSymbolizerInfo, PtrThumbCtx, std::move(ThumbRelInfo)));
2657 ThumbDisAsm->setSymbolizer(std::move(ThumbSymbolizer));
2659 int ThumbAsmPrinterVariant = ThumbAsmInfo->getAssemblerDialect();
2660 ThumbIP.reset(ThumbTarget->createMCInstPrinter(
2661 ThumbAsmPrinterVariant, *ThumbAsmInfo, *ThumbInstrInfo, *ThumbMRI,
2663 // Set the display preference for hex vs. decimal immediates.
2664 ThumbIP->setPrintImmHex(PrintImmHex);
2667 if (ThumbTarget && (!ThumbAsmInfo || !ThumbSTI || !ThumbDisAsm || !ThumbIP)) {
2668 errs() << "error: couldn't initialize disassembler for target "
2669 << ThumbTripleName << '\n';
2673 MachO::mach_header Header = MachOOF->getHeader();
2675 // FIXME: Using the -cfg command line option, this code used to be able to
2676 // annotate relocations with the referenced symbol's name, and if this was
2677 // inside a __[cf]string section, the data it points to. This is now replaced
2678 // by the upcoming MCSymbolizer, which needs the appropriate setup done above.
2679 std::vector<SectionRef> Sections;
2680 std::vector<SymbolRef> Symbols;
2681 SmallVector<uint64_t, 8> FoundFns;
2682 uint64_t BaseSegmentAddress;
2684 getSectionsAndSymbols(Header, MachOOF, Sections, Symbols, FoundFns,
2685 BaseSegmentAddress);
2687 // Sort the symbols by address, just in case they didn't come in that way.
2688 std::sort(Symbols.begin(), Symbols.end(), SymbolSorter());
2690 // Build a data in code table that is sorted on by the address of each entry.
2691 uint64_t BaseAddress = 0;
2692 if (Header.filetype == MachO::MH_OBJECT)
2693 BaseAddress = Sections[0].getAddress();
2695 BaseAddress = BaseSegmentAddress;
2697 for (dice_iterator DI = MachOOF->begin_dices(), DE = MachOOF->end_dices();
2700 DI->getOffset(Offset);
2701 Dices.push_back(std::make_pair(BaseAddress + Offset, *DI));
2703 array_pod_sort(Dices.begin(), Dices.end());
2706 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
2708 raw_ostream &DebugOut = nulls();
2711 std::unique_ptr<DIContext> diContext;
2712 ObjectFile *DbgObj = MachOOF;
2713 // Try to find debug info and set up the DIContext for it.
2715 // A separate DSym file path was specified, parse it as a macho file,
2716 // get the sections and supply it to the section name parsing machinery.
2717 if (!DSYMFile.empty()) {
2718 ErrorOr<std::unique_ptr<MemoryBuffer>> BufOrErr =
2719 MemoryBuffer::getFileOrSTDIN(DSYMFile);
2720 if (std::error_code EC = BufOrErr.getError()) {
2721 errs() << "llvm-objdump: " << Filename << ": " << EC.message() << '\n';
2725 ObjectFile::createMachOObjectFile(BufOrErr.get()->getMemBufferRef())
2730 // Setup the DIContext
2731 diContext.reset(DIContext::getDWARFContext(*DbgObj));
2734 if (DumpSections.size() == 0)
2735 outs() << "(" << DisSegName << "," << DisSectName << ") section\n";
2737 for (unsigned SectIdx = 0; SectIdx != Sections.size(); SectIdx++) {
2739 if (Sections[SectIdx].getName(SectName) || SectName != DisSectName)
2742 DataRefImpl DR = Sections[SectIdx].getRawDataRefImpl();
2744 StringRef SegmentName = MachOOF->getSectionFinalSegmentName(DR);
2745 if (SegmentName != DisSegName)
2749 Sections[SectIdx].getContents(BytesStr);
2750 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()),
2752 uint64_t SectAddress = Sections[SectIdx].getAddress();
2754 bool symbolTableWorked = false;
2756 // Parse relocations.
2757 std::vector<std::pair<uint64_t, SymbolRef>> Relocs;
2758 for (const RelocationRef &Reloc : Sections[SectIdx].relocations()) {
2759 uint64_t RelocOffset;
2760 Reloc.getOffset(RelocOffset);
2761 uint64_t SectionAddress = Sections[SectIdx].getAddress();
2762 RelocOffset -= SectionAddress;
2764 symbol_iterator RelocSym = Reloc.getSymbol();
2766 Relocs.push_back(std::make_pair(RelocOffset, *RelocSym));
2768 array_pod_sort(Relocs.begin(), Relocs.end());
2770 // Create a map of symbol addresses to symbol names for use by
2771 // the SymbolizerSymbolLookUp() routine.
2772 SymbolAddressMap AddrMap;
2773 for (const SymbolRef &Symbol : MachOOF->symbols()) {
2776 if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data ||
2777 ST == SymbolRef::ST_Other) {
2779 Symbol.getAddress(Address);
2781 Symbol.getName(SymName);
2782 AddrMap[Address] = SymName;
2785 // Set up the block of info used by the Symbolizer call backs.
2786 SymbolizerInfo.verbose = true;
2787 SymbolizerInfo.O = MachOOF;
2788 SymbolizerInfo.S = Sections[SectIdx];
2789 SymbolizerInfo.AddrMap = &AddrMap;
2790 SymbolizerInfo.Sections = &Sections;
2791 SymbolizerInfo.class_name = nullptr;
2792 SymbolizerInfo.selector_name = nullptr;
2793 SymbolizerInfo.method = nullptr;
2794 SymbolizerInfo.demangled_name = nullptr;
2795 SymbolizerInfo.bindtable = nullptr;
2796 SymbolizerInfo.adrp_addr = 0;
2797 SymbolizerInfo.adrp_inst = 0;
2798 // Same for the ThumbSymbolizer
2799 ThumbSymbolizerInfo.verbose = true;
2800 ThumbSymbolizerInfo.O = MachOOF;
2801 ThumbSymbolizerInfo.S = Sections[SectIdx];
2802 ThumbSymbolizerInfo.AddrMap = &AddrMap;
2803 ThumbSymbolizerInfo.Sections = &Sections;
2804 ThumbSymbolizerInfo.class_name = nullptr;
2805 ThumbSymbolizerInfo.selector_name = nullptr;
2806 ThumbSymbolizerInfo.method = nullptr;
2807 ThumbSymbolizerInfo.demangled_name = nullptr;
2808 ThumbSymbolizerInfo.bindtable = nullptr;
2809 ThumbSymbolizerInfo.adrp_addr = 0;
2810 ThumbSymbolizerInfo.adrp_inst = 0;
2812 // Disassemble symbol by symbol.
2813 for (unsigned SymIdx = 0; SymIdx != Symbols.size(); SymIdx++) {
2815 Symbols[SymIdx].getName(SymName);
2818 Symbols[SymIdx].getType(ST);
2819 if (ST != SymbolRef::ST_Function)
2822 // Make sure the symbol is defined in this section.
2823 bool containsSym = Sections[SectIdx].containsSymbol(Symbols[SymIdx]);
2827 // Start at the address of the symbol relative to the section's address.
2829 uint64_t SectionAddress = Sections[SectIdx].getAddress();
2830 Symbols[SymIdx].getAddress(Start);
2831 Start -= SectionAddress;
2833 // Stop disassembling either at the beginning of the next symbol or at
2834 // the end of the section.
2835 bool containsNextSym = false;
2836 uint64_t NextSym = 0;
2837 uint64_t NextSymIdx = SymIdx + 1;
2838 while (Symbols.size() > NextSymIdx) {
2839 SymbolRef::Type NextSymType;
2840 Symbols[NextSymIdx].getType(NextSymType);
2841 if (NextSymType == SymbolRef::ST_Function) {
2843 Sections[SectIdx].containsSymbol(Symbols[NextSymIdx]);
2844 Symbols[NextSymIdx].getAddress(NextSym);
2845 NextSym -= SectionAddress;
2851 uint64_t SectSize = Sections[SectIdx].getSize();
2852 uint64_t End = containsNextSym ? NextSym : SectSize;
2855 symbolTableWorked = true;
2857 DataRefImpl Symb = Symbols[SymIdx].getRawDataRefImpl();
2859 (MachOOF->getSymbolFlags(Symb) & SymbolRef::SF_Thumb) && ThumbTarget;
2861 outs() << SymName << ":\n";
2862 DILineInfo lastLine;
2863 for (uint64_t Index = Start; Index < End; Index += Size) {
2866 uint64_t PC = SectAddress + Index;
2867 if (FullLeadingAddr) {
2868 if (MachOOF->is64Bit())
2869 outs() << format("%016" PRIx64, PC);
2871 outs() << format("%08" PRIx64, PC);
2873 outs() << format("%8" PRIx64 ":", PC);
2878 // Check the data in code table here to see if this is data not an
2879 // instruction to be disassembled.
2881 Dice.push_back(std::make_pair(PC, DiceRef()));
2882 dice_table_iterator DTI =
2883 std::search(Dices.begin(), Dices.end(), Dice.begin(), Dice.end(),
2884 compareDiceTableEntries);
2885 if (DTI != Dices.end()) {
2887 DTI->second.getLength(Length);
2889 DTI->second.getKind(Kind);
2890 Size = DumpDataInCode(reinterpret_cast<const char *>(Bytes.data()) +
2893 if ((Kind == MachO::DICE_KIND_JUMP_TABLE8) &&
2894 (PC == (DTI->first + Length - 1)) && (Length & 1))
2899 SmallVector<char, 64> AnnotationsBytes;
2900 raw_svector_ostream Annotations(AnnotationsBytes);
2904 gotInst = ThumbDisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
2905 PC, DebugOut, Annotations);
2907 gotInst = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), PC,
2908 DebugOut, Annotations);
2910 if (!NoShowRawInsn) {
2911 DumpBytes(StringRef(
2912 reinterpret_cast<const char *>(Bytes.data()) + Index, Size));
2914 formatted_raw_ostream FormattedOS(outs());
2915 Annotations.flush();
2916 StringRef AnnotationsStr = Annotations.str();
2918 ThumbIP->printInst(&Inst, FormattedOS, AnnotationsStr);
2920 IP->printInst(&Inst, FormattedOS, AnnotationsStr);
2921 emitComments(CommentStream, CommentsToEmit, FormattedOS, *AsmInfo);
2923 // Print debug info.
2925 DILineInfo dli = diContext->getLineInfoForAddress(PC);
2926 // Print valid line info if it changed.
2927 if (dli != lastLine && dli.Line != 0)
2928 outs() << "\t## " << dli.FileName << ':' << dli.Line << ':'
2934 unsigned int Arch = MachOOF->getArch();
2935 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
2936 outs() << format("\t.byte 0x%02x #bad opcode\n",
2937 *(Bytes.data() + Index) & 0xff);
2938 Size = 1; // skip exactly one illegible byte and move on.
2939 } else if (Arch == Triple::aarch64) {
2940 uint32_t opcode = (*(Bytes.data() + Index) & 0xff) |
2941 (*(Bytes.data() + Index + 1) & 0xff) << 8 |
2942 (*(Bytes.data() + Index + 2) & 0xff) << 16 |
2943 (*(Bytes.data() + Index + 3) & 0xff) << 24;
2944 outs() << format("\t.long\t0x%08x\n", opcode);
2947 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
2949 Size = 1; // skip illegible bytes
2954 if (!symbolTableWorked) {
2955 // Reading the symbol table didn't work, disassemble the whole section.
2956 uint64_t SectAddress = Sections[SectIdx].getAddress();
2957 uint64_t SectSize = Sections[SectIdx].getSize();
2959 for (uint64_t Index = 0; Index < SectSize; Index += InstSize) {
2962 uint64_t PC = SectAddress + Index;
2963 if (DisAsm->getInstruction(Inst, InstSize, Bytes.slice(Index), PC,
2964 DebugOut, nulls())) {
2965 if (FullLeadingAddr) {
2966 if (MachOOF->is64Bit())
2967 outs() << format("%016" PRIx64, PC);
2969 outs() << format("%08" PRIx64, PC);
2971 outs() << format("%8" PRIx64 ":", PC);
2973 if (!NoShowRawInsn) {
2976 StringRef(reinterpret_cast<const char *>(Bytes.data()) + Index,
2979 IP->printInst(&Inst, outs(), "");
2982 unsigned int Arch = MachOOF->getArch();
2983 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
2984 outs() << format("\t.byte 0x%02x #bad opcode\n",
2985 *(Bytes.data() + Index) & 0xff);
2986 InstSize = 1; // skip exactly one illegible byte and move on.
2988 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
2990 InstSize = 1; // skip illegible bytes
2995 // The TripleName's need to be reset if we are called again for a different
2998 ThumbTripleName = "";
3000 if (SymbolizerInfo.method != nullptr)
3001 free(SymbolizerInfo.method);
3002 if (SymbolizerInfo.demangled_name != nullptr)
3003 free(SymbolizerInfo.demangled_name);
3004 if (SymbolizerInfo.bindtable != nullptr)
3005 delete SymbolizerInfo.bindtable;
3006 if (ThumbSymbolizerInfo.method != nullptr)
3007 free(ThumbSymbolizerInfo.method);
3008 if (ThumbSymbolizerInfo.demangled_name != nullptr)
3009 free(ThumbSymbolizerInfo.demangled_name);
3010 if (ThumbSymbolizerInfo.bindtable != nullptr)
3011 delete ThumbSymbolizerInfo.bindtable;
3015 //===----------------------------------------------------------------------===//
3016 // __compact_unwind section dumping
3017 //===----------------------------------------------------------------------===//
3021 template <typename T> static uint64_t readNext(const char *&Buf) {
3022 using llvm::support::little;
3023 using llvm::support::unaligned;
3025 uint64_t Val = support::endian::read<T, little, unaligned>(Buf);
3030 struct CompactUnwindEntry {
3031 uint32_t OffsetInSection;
3033 uint64_t FunctionAddr;
3035 uint32_t CompactEncoding;
3036 uint64_t PersonalityAddr;
3039 RelocationRef FunctionReloc;
3040 RelocationRef PersonalityReloc;
3041 RelocationRef LSDAReloc;
3043 CompactUnwindEntry(StringRef Contents, unsigned Offset, bool Is64)
3044 : OffsetInSection(Offset) {
3046 read<uint64_t>(Contents.data() + Offset);
3048 read<uint32_t>(Contents.data() + Offset);
3052 template <typename UIntPtr> void read(const char *Buf) {
3053 FunctionAddr = readNext<UIntPtr>(Buf);
3054 Length = readNext<uint32_t>(Buf);
3055 CompactEncoding = readNext<uint32_t>(Buf);
3056 PersonalityAddr = readNext<UIntPtr>(Buf);
3057 LSDAAddr = readNext<UIntPtr>(Buf);
3062 /// Given a relocation from __compact_unwind, consisting of the RelocationRef
3063 /// and data being relocated, determine the best base Name and Addend to use for
3064 /// display purposes.
3066 /// 1. An Extern relocation will directly reference a symbol (and the data is
3067 /// then already an addend), so use that.
3068 /// 2. Otherwise the data is an offset in the object file's layout; try to find
3069 // a symbol before it in the same section, and use the offset from there.
3070 /// 3. Finally, if all that fails, fall back to an offset from the start of the
3071 /// referenced section.
3072 static void findUnwindRelocNameAddend(const MachOObjectFile *Obj,
3073 std::map<uint64_t, SymbolRef> &Symbols,
3074 const RelocationRef &Reloc, uint64_t Addr,
3075 StringRef &Name, uint64_t &Addend) {
3076 if (Reloc.getSymbol() != Obj->symbol_end()) {
3077 Reloc.getSymbol()->getName(Name);
3082 auto RE = Obj->getRelocation(Reloc.getRawDataRefImpl());
3083 SectionRef RelocSection = Obj->getRelocationSection(RE);
3085 uint64_t SectionAddr = RelocSection.getAddress();
3087 auto Sym = Symbols.upper_bound(Addr);
3088 if (Sym == Symbols.begin()) {
3089 // The first symbol in the object is after this reference, the best we can
3090 // do is section-relative notation.
3091 RelocSection.getName(Name);
3092 Addend = Addr - SectionAddr;
3096 // Go back one so that SymbolAddress <= Addr.
3099 section_iterator SymSection = Obj->section_end();
3100 Sym->second.getSection(SymSection);
3101 if (RelocSection == *SymSection) {
3102 // There's a valid symbol in the same section before this reference.
3103 Sym->second.getName(Name);
3104 Addend = Addr - Sym->first;
3108 // There is a symbol before this reference, but it's in a different
3109 // section. Probably not helpful to mention it, so use the section name.
3110 RelocSection.getName(Name);
3111 Addend = Addr - SectionAddr;
3114 static void printUnwindRelocDest(const MachOObjectFile *Obj,
3115 std::map<uint64_t, SymbolRef> &Symbols,
3116 const RelocationRef &Reloc, uint64_t Addr) {
3120 if (!Reloc.getObjectFile())
3123 findUnwindRelocNameAddend(Obj, Symbols, Reloc, Addr, Name, Addend);
3127 outs() << " + " << format("0x%" PRIx64, Addend);
3131 printMachOCompactUnwindSection(const MachOObjectFile *Obj,
3132 std::map<uint64_t, SymbolRef> &Symbols,
3133 const SectionRef &CompactUnwind) {
3135 assert(Obj->isLittleEndian() &&
3136 "There should not be a big-endian .o with __compact_unwind");
3138 bool Is64 = Obj->is64Bit();
3139 uint32_t PointerSize = Is64 ? sizeof(uint64_t) : sizeof(uint32_t);
3140 uint32_t EntrySize = 3 * PointerSize + 2 * sizeof(uint32_t);
3143 CompactUnwind.getContents(Contents);
3145 SmallVector<CompactUnwindEntry, 4> CompactUnwinds;
3147 // First populate the initial raw offsets, encodings and so on from the entry.
3148 for (unsigned Offset = 0; Offset < Contents.size(); Offset += EntrySize) {
3149 CompactUnwindEntry Entry(Contents.data(), Offset, Is64);
3150 CompactUnwinds.push_back(Entry);
3153 // Next we need to look at the relocations to find out what objects are
3154 // actually being referred to.
3155 for (const RelocationRef &Reloc : CompactUnwind.relocations()) {
3156 uint64_t RelocAddress;
3157 Reloc.getOffset(RelocAddress);
3159 uint32_t EntryIdx = RelocAddress / EntrySize;
3160 uint32_t OffsetInEntry = RelocAddress - EntryIdx * EntrySize;
3161 CompactUnwindEntry &Entry = CompactUnwinds[EntryIdx];
3163 if (OffsetInEntry == 0)
3164 Entry.FunctionReloc = Reloc;
3165 else if (OffsetInEntry == PointerSize + 2 * sizeof(uint32_t))
3166 Entry.PersonalityReloc = Reloc;
3167 else if (OffsetInEntry == 2 * PointerSize + 2 * sizeof(uint32_t))
3168 Entry.LSDAReloc = Reloc;
3170 llvm_unreachable("Unexpected relocation in __compact_unwind section");
3173 // Finally, we're ready to print the data we've gathered.
3174 outs() << "Contents of __compact_unwind section:\n";
3175 for (auto &Entry : CompactUnwinds) {
3176 outs() << " Entry at offset "
3177 << format("0x%" PRIx32, Entry.OffsetInSection) << ":\n";
3179 // 1. Start of the region this entry applies to.
3180 outs() << " start: " << format("0x%" PRIx64,
3181 Entry.FunctionAddr) << ' ';
3182 printUnwindRelocDest(Obj, Symbols, Entry.FunctionReloc, Entry.FunctionAddr);
3185 // 2. Length of the region this entry applies to.
3186 outs() << " length: " << format("0x%" PRIx32, Entry.Length)
3188 // 3. The 32-bit compact encoding.
3189 outs() << " compact encoding: "
3190 << format("0x%08" PRIx32, Entry.CompactEncoding) << '\n';
3192 // 4. The personality function, if present.
3193 if (Entry.PersonalityReloc.getObjectFile()) {
3194 outs() << " personality function: "
3195 << format("0x%" PRIx64, Entry.PersonalityAddr) << ' ';
3196 printUnwindRelocDest(Obj, Symbols, Entry.PersonalityReloc,
3197 Entry.PersonalityAddr);
3201 // 5. This entry's language-specific data area.
3202 if (Entry.LSDAReloc.getObjectFile()) {
3203 outs() << " LSDA: " << format("0x%" PRIx64,
3204 Entry.LSDAAddr) << ' ';
3205 printUnwindRelocDest(Obj, Symbols, Entry.LSDAReloc, Entry.LSDAAddr);
3211 //===----------------------------------------------------------------------===//
3212 // __unwind_info section dumping
3213 //===----------------------------------------------------------------------===//
3215 static void printRegularSecondLevelUnwindPage(const char *PageStart) {
3216 const char *Pos = PageStart;
3217 uint32_t Kind = readNext<uint32_t>(Pos);
3219 assert(Kind == 2 && "kind for a regular 2nd level index should be 2");
3221 uint16_t EntriesStart = readNext<uint16_t>(Pos);
3222 uint16_t NumEntries = readNext<uint16_t>(Pos);
3224 Pos = PageStart + EntriesStart;
3225 for (unsigned i = 0; i < NumEntries; ++i) {
3226 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
3227 uint32_t Encoding = readNext<uint32_t>(Pos);
3229 outs() << " [" << i << "]: "
3230 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
3232 << "encoding=" << format("0x%08" PRIx32, Encoding) << '\n';
3236 static void printCompressedSecondLevelUnwindPage(
3237 const char *PageStart, uint32_t FunctionBase,
3238 const SmallVectorImpl<uint32_t> &CommonEncodings) {
3239 const char *Pos = PageStart;
3240 uint32_t Kind = readNext<uint32_t>(Pos);
3242 assert(Kind == 3 && "kind for a compressed 2nd level index should be 3");
3244 uint16_t EntriesStart = readNext<uint16_t>(Pos);
3245 uint16_t NumEntries = readNext<uint16_t>(Pos);
3247 uint16_t EncodingsStart = readNext<uint16_t>(Pos);
3248 readNext<uint16_t>(Pos);
3249 const auto *PageEncodings = reinterpret_cast<const support::ulittle32_t *>(
3250 PageStart + EncodingsStart);
3252 Pos = PageStart + EntriesStart;
3253 for (unsigned i = 0; i < NumEntries; ++i) {
3254 uint32_t Entry = readNext<uint32_t>(Pos);
3255 uint32_t FunctionOffset = FunctionBase + (Entry & 0xffffff);
3256 uint32_t EncodingIdx = Entry >> 24;
3259 if (EncodingIdx < CommonEncodings.size())
3260 Encoding = CommonEncodings[EncodingIdx];
3262 Encoding = PageEncodings[EncodingIdx - CommonEncodings.size()];
3264 outs() << " [" << i << "]: "
3265 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
3267 << "encoding[" << EncodingIdx
3268 << "]=" << format("0x%08" PRIx32, Encoding) << '\n';
3272 static void printMachOUnwindInfoSection(const MachOObjectFile *Obj,
3273 std::map<uint64_t, SymbolRef> &Symbols,
3274 const SectionRef &UnwindInfo) {
3276 assert(Obj->isLittleEndian() &&
3277 "There should not be a big-endian .o with __unwind_info");
3279 outs() << "Contents of __unwind_info section:\n";
3282 UnwindInfo.getContents(Contents);
3283 const char *Pos = Contents.data();
3285 //===----------------------------------
3287 //===----------------------------------
3289 uint32_t Version = readNext<uint32_t>(Pos);
3290 outs() << " Version: "
3291 << format("0x%" PRIx32, Version) << '\n';
3292 assert(Version == 1 && "only understand version 1");
3294 uint32_t CommonEncodingsStart = readNext<uint32_t>(Pos);
3295 outs() << " Common encodings array section offset: "
3296 << format("0x%" PRIx32, CommonEncodingsStart) << '\n';
3297 uint32_t NumCommonEncodings = readNext<uint32_t>(Pos);
3298 outs() << " Number of common encodings in array: "
3299 << format("0x%" PRIx32, NumCommonEncodings) << '\n';
3301 uint32_t PersonalitiesStart = readNext<uint32_t>(Pos);
3302 outs() << " Personality function array section offset: "
3303 << format("0x%" PRIx32, PersonalitiesStart) << '\n';
3304 uint32_t NumPersonalities = readNext<uint32_t>(Pos);
3305 outs() << " Number of personality functions in array: "
3306 << format("0x%" PRIx32, NumPersonalities) << '\n';
3308 uint32_t IndicesStart = readNext<uint32_t>(Pos);
3309 outs() << " Index array section offset: "
3310 << format("0x%" PRIx32, IndicesStart) << '\n';
3311 uint32_t NumIndices = readNext<uint32_t>(Pos);
3312 outs() << " Number of indices in array: "
3313 << format("0x%" PRIx32, NumIndices) << '\n';
3315 //===----------------------------------
3316 // A shared list of common encodings
3317 //===----------------------------------
3319 // These occupy indices in the range [0, N] whenever an encoding is referenced
3320 // from a compressed 2nd level index table. In practice the linker only
3321 // creates ~128 of these, so that indices are available to embed encodings in
3322 // the 2nd level index.
3324 SmallVector<uint32_t, 64> CommonEncodings;
3325 outs() << " Common encodings: (count = " << NumCommonEncodings << ")\n";
3326 Pos = Contents.data() + CommonEncodingsStart;
3327 for (unsigned i = 0; i < NumCommonEncodings; ++i) {
3328 uint32_t Encoding = readNext<uint32_t>(Pos);
3329 CommonEncodings.push_back(Encoding);
3331 outs() << " encoding[" << i << "]: " << format("0x%08" PRIx32, Encoding)
3335 //===----------------------------------
3336 // Personality functions used in this executable
3337 //===----------------------------------
3339 // There should be only a handful of these (one per source language,
3340 // roughly). Particularly since they only get 2 bits in the compact encoding.
3342 outs() << " Personality functions: (count = " << NumPersonalities << ")\n";
3343 Pos = Contents.data() + PersonalitiesStart;
3344 for (unsigned i = 0; i < NumPersonalities; ++i) {
3345 uint32_t PersonalityFn = readNext<uint32_t>(Pos);
3346 outs() << " personality[" << i + 1
3347 << "]: " << format("0x%08" PRIx32, PersonalityFn) << '\n';
3350 //===----------------------------------
3351 // The level 1 index entries
3352 //===----------------------------------
3354 // These specify an approximate place to start searching for the more detailed
3355 // information, sorted by PC.
3358 uint32_t FunctionOffset;
3359 uint32_t SecondLevelPageStart;
3363 SmallVector<IndexEntry, 4> IndexEntries;
3365 outs() << " Top level indices: (count = " << NumIndices << ")\n";
3366 Pos = Contents.data() + IndicesStart;
3367 for (unsigned i = 0; i < NumIndices; ++i) {
3370 Entry.FunctionOffset = readNext<uint32_t>(Pos);
3371 Entry.SecondLevelPageStart = readNext<uint32_t>(Pos);
3372 Entry.LSDAStart = readNext<uint32_t>(Pos);
3373 IndexEntries.push_back(Entry);
3375 outs() << " [" << i << "]: "
3376 << "function offset=" << format("0x%08" PRIx32, Entry.FunctionOffset)
3378 << "2nd level page offset="
3379 << format("0x%08" PRIx32, Entry.SecondLevelPageStart) << ", "
3380 << "LSDA offset=" << format("0x%08" PRIx32, Entry.LSDAStart) << '\n';
3383 //===----------------------------------
3384 // Next come the LSDA tables
3385 //===----------------------------------
3387 // The LSDA layout is rather implicit: it's a contiguous array of entries from
3388 // the first top-level index's LSDAOffset to the last (sentinel).
3390 outs() << " LSDA descriptors:\n";
3391 Pos = Contents.data() + IndexEntries[0].LSDAStart;
3392 int NumLSDAs = (IndexEntries.back().LSDAStart - IndexEntries[0].LSDAStart) /
3393 (2 * sizeof(uint32_t));
3394 for (int i = 0; i < NumLSDAs; ++i) {
3395 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
3396 uint32_t LSDAOffset = readNext<uint32_t>(Pos);
3397 outs() << " [" << i << "]: "
3398 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
3400 << "LSDA offset=" << format("0x%08" PRIx32, LSDAOffset) << '\n';
3403 //===----------------------------------
3404 // Finally, the 2nd level indices
3405 //===----------------------------------
3407 // Generally these are 4K in size, and have 2 possible forms:
3408 // + Regular stores up to 511 entries with disparate encodings
3409 // + Compressed stores up to 1021 entries if few enough compact encoding
3411 outs() << " Second level indices:\n";
3412 for (unsigned i = 0; i < IndexEntries.size() - 1; ++i) {
3413 // The final sentinel top-level index has no associated 2nd level page
3414 if (IndexEntries[i].SecondLevelPageStart == 0)
3417 outs() << " Second level index[" << i << "]: "
3418 << "offset in section="
3419 << format("0x%08" PRIx32, IndexEntries[i].SecondLevelPageStart)
3421 << "base function offset="
3422 << format("0x%08" PRIx32, IndexEntries[i].FunctionOffset) << '\n';
3424 Pos = Contents.data() + IndexEntries[i].SecondLevelPageStart;
3425 uint32_t Kind = *reinterpret_cast<const support::ulittle32_t *>(Pos);
3427 printRegularSecondLevelUnwindPage(Pos);
3429 printCompressedSecondLevelUnwindPage(Pos, IndexEntries[i].FunctionOffset,
3432 llvm_unreachable("Do not know how to print this kind of 2nd level page");
3436 void llvm::printMachOUnwindInfo(const MachOObjectFile *Obj) {
3437 std::map<uint64_t, SymbolRef> Symbols;
3438 for (const SymbolRef &SymRef : Obj->symbols()) {
3439 // Discard any undefined or absolute symbols. They're not going to take part
3440 // in the convenience lookup for unwind info and just take up resources.
3441 section_iterator Section = Obj->section_end();
3442 SymRef.getSection(Section);
3443 if (Section == Obj->section_end())
3447 SymRef.getAddress(Addr);
3448 Symbols.insert(std::make_pair(Addr, SymRef));
3451 for (const SectionRef &Section : Obj->sections()) {
3453 Section.getName(SectName);
3454 if (SectName == "__compact_unwind")
3455 printMachOCompactUnwindSection(Obj, Symbols, Section);
3456 else if (SectName == "__unwind_info")
3457 printMachOUnwindInfoSection(Obj, Symbols, Section);
3458 else if (SectName == "__eh_frame")
3459 outs() << "llvm-objdump: warning: unhandled __eh_frame section\n";
3463 static void PrintMachHeader(uint32_t magic, uint32_t cputype,
3464 uint32_t cpusubtype, uint32_t filetype,
3465 uint32_t ncmds, uint32_t sizeofcmds, uint32_t flags,
3467 outs() << "Mach header\n";
3468 outs() << " magic cputype cpusubtype caps filetype ncmds "
3469 "sizeofcmds flags\n";
3471 if (magic == MachO::MH_MAGIC)
3472 outs() << " MH_MAGIC";
3473 else if (magic == MachO::MH_MAGIC_64)
3474 outs() << "MH_MAGIC_64";
3476 outs() << format(" 0x%08" PRIx32, magic);
3478 case MachO::CPU_TYPE_I386:
3480 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3481 case MachO::CPU_SUBTYPE_I386_ALL:
3485 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3489 case MachO::CPU_TYPE_X86_64:
3490 outs() << " X86_64";
3491 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3492 case MachO::CPU_SUBTYPE_X86_64_ALL:
3495 case MachO::CPU_SUBTYPE_X86_64_H:
3496 outs() << " Haswell";
3499 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3503 case MachO::CPU_TYPE_ARM:
3505 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3506 case MachO::CPU_SUBTYPE_ARM_ALL:
3509 case MachO::CPU_SUBTYPE_ARM_V4T:
3512 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
3515 case MachO::CPU_SUBTYPE_ARM_XSCALE:
3516 outs() << " XSCALE";
3518 case MachO::CPU_SUBTYPE_ARM_V6:
3521 case MachO::CPU_SUBTYPE_ARM_V6M:
3524 case MachO::CPU_SUBTYPE_ARM_V7:
3527 case MachO::CPU_SUBTYPE_ARM_V7EM:
3530 case MachO::CPU_SUBTYPE_ARM_V7K:
3533 case MachO::CPU_SUBTYPE_ARM_V7M:
3536 case MachO::CPU_SUBTYPE_ARM_V7S:
3540 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3544 case MachO::CPU_TYPE_ARM64:
3546 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3547 case MachO::CPU_SUBTYPE_ARM64_ALL:
3551 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3555 case MachO::CPU_TYPE_POWERPC:
3557 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3558 case MachO::CPU_SUBTYPE_POWERPC_ALL:
3562 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3566 case MachO::CPU_TYPE_POWERPC64:
3568 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3569 case MachO::CPU_SUBTYPE_POWERPC_ALL:
3573 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3578 if ((cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64) {
3581 outs() << format(" 0x%02" PRIx32,
3582 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
3585 case MachO::MH_OBJECT:
3586 outs() << " OBJECT";
3588 case MachO::MH_EXECUTE:
3589 outs() << " EXECUTE";
3591 case MachO::MH_FVMLIB:
3592 outs() << " FVMLIB";
3594 case MachO::MH_CORE:
3597 case MachO::MH_PRELOAD:
3598 outs() << " PRELOAD";
3600 case MachO::MH_DYLIB:
3603 case MachO::MH_DYLIB_STUB:
3604 outs() << " DYLIB_STUB";
3606 case MachO::MH_DYLINKER:
3607 outs() << " DYLINKER";
3609 case MachO::MH_BUNDLE:
3610 outs() << " BUNDLE";
3612 case MachO::MH_DSYM:
3615 case MachO::MH_KEXT_BUNDLE:
3616 outs() << " KEXTBUNDLE";
3619 outs() << format(" %10u", filetype);
3622 outs() << format(" %5u", ncmds);
3623 outs() << format(" %10u", sizeofcmds);
3625 if (f & MachO::MH_NOUNDEFS) {
3626 outs() << " NOUNDEFS";
3627 f &= ~MachO::MH_NOUNDEFS;
3629 if (f & MachO::MH_INCRLINK) {
3630 outs() << " INCRLINK";
3631 f &= ~MachO::MH_INCRLINK;
3633 if (f & MachO::MH_DYLDLINK) {
3634 outs() << " DYLDLINK";
3635 f &= ~MachO::MH_DYLDLINK;
3637 if (f & MachO::MH_BINDATLOAD) {
3638 outs() << " BINDATLOAD";
3639 f &= ~MachO::MH_BINDATLOAD;
3641 if (f & MachO::MH_PREBOUND) {
3642 outs() << " PREBOUND";
3643 f &= ~MachO::MH_PREBOUND;
3645 if (f & MachO::MH_SPLIT_SEGS) {
3646 outs() << " SPLIT_SEGS";
3647 f &= ~MachO::MH_SPLIT_SEGS;
3649 if (f & MachO::MH_LAZY_INIT) {
3650 outs() << " LAZY_INIT";
3651 f &= ~MachO::MH_LAZY_INIT;
3653 if (f & MachO::MH_TWOLEVEL) {
3654 outs() << " TWOLEVEL";
3655 f &= ~MachO::MH_TWOLEVEL;
3657 if (f & MachO::MH_FORCE_FLAT) {
3658 outs() << " FORCE_FLAT";
3659 f &= ~MachO::MH_FORCE_FLAT;
3661 if (f & MachO::MH_NOMULTIDEFS) {
3662 outs() << " NOMULTIDEFS";
3663 f &= ~MachO::MH_NOMULTIDEFS;
3665 if (f & MachO::MH_NOFIXPREBINDING) {
3666 outs() << " NOFIXPREBINDING";
3667 f &= ~MachO::MH_NOFIXPREBINDING;
3669 if (f & MachO::MH_PREBINDABLE) {
3670 outs() << " PREBINDABLE";
3671 f &= ~MachO::MH_PREBINDABLE;
3673 if (f & MachO::MH_ALLMODSBOUND) {
3674 outs() << " ALLMODSBOUND";
3675 f &= ~MachO::MH_ALLMODSBOUND;
3677 if (f & MachO::MH_SUBSECTIONS_VIA_SYMBOLS) {
3678 outs() << " SUBSECTIONS_VIA_SYMBOLS";
3679 f &= ~MachO::MH_SUBSECTIONS_VIA_SYMBOLS;
3681 if (f & MachO::MH_CANONICAL) {
3682 outs() << " CANONICAL";
3683 f &= ~MachO::MH_CANONICAL;
3685 if (f & MachO::MH_WEAK_DEFINES) {
3686 outs() << " WEAK_DEFINES";
3687 f &= ~MachO::MH_WEAK_DEFINES;
3689 if (f & MachO::MH_BINDS_TO_WEAK) {
3690 outs() << " BINDS_TO_WEAK";
3691 f &= ~MachO::MH_BINDS_TO_WEAK;
3693 if (f & MachO::MH_ALLOW_STACK_EXECUTION) {
3694 outs() << " ALLOW_STACK_EXECUTION";
3695 f &= ~MachO::MH_ALLOW_STACK_EXECUTION;
3697 if (f & MachO::MH_DEAD_STRIPPABLE_DYLIB) {
3698 outs() << " DEAD_STRIPPABLE_DYLIB";
3699 f &= ~MachO::MH_DEAD_STRIPPABLE_DYLIB;
3701 if (f & MachO::MH_PIE) {
3703 f &= ~MachO::MH_PIE;
3705 if (f & MachO::MH_NO_REEXPORTED_DYLIBS) {
3706 outs() << " NO_REEXPORTED_DYLIBS";
3707 f &= ~MachO::MH_NO_REEXPORTED_DYLIBS;
3709 if (f & MachO::MH_HAS_TLV_DESCRIPTORS) {
3710 outs() << " MH_HAS_TLV_DESCRIPTORS";
3711 f &= ~MachO::MH_HAS_TLV_DESCRIPTORS;
3713 if (f & MachO::MH_NO_HEAP_EXECUTION) {
3714 outs() << " MH_NO_HEAP_EXECUTION";
3715 f &= ~MachO::MH_NO_HEAP_EXECUTION;
3717 if (f & MachO::MH_APP_EXTENSION_SAFE) {
3718 outs() << " APP_EXTENSION_SAFE";
3719 f &= ~MachO::MH_APP_EXTENSION_SAFE;
3721 if (f != 0 || flags == 0)
3722 outs() << format(" 0x%08" PRIx32, f);
3724 outs() << format(" 0x%08" PRIx32, magic);
3725 outs() << format(" %7d", cputype);
3726 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3727 outs() << format(" 0x%02" PRIx32,
3728 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
3729 outs() << format(" %10u", filetype);
3730 outs() << format(" %5u", ncmds);
3731 outs() << format(" %10u", sizeofcmds);
3732 outs() << format(" 0x%08" PRIx32, flags);
3737 static void PrintSegmentCommand(uint32_t cmd, uint32_t cmdsize,
3738 StringRef SegName, uint64_t vmaddr,
3739 uint64_t vmsize, uint64_t fileoff,
3740 uint64_t filesize, uint32_t maxprot,
3741 uint32_t initprot, uint32_t nsects,
3742 uint32_t flags, uint32_t object_size,
3744 uint64_t expected_cmdsize;
3745 if (cmd == MachO::LC_SEGMENT) {
3746 outs() << " cmd LC_SEGMENT\n";
3747 expected_cmdsize = nsects;
3748 expected_cmdsize *= sizeof(struct MachO::section);
3749 expected_cmdsize += sizeof(struct MachO::segment_command);
3751 outs() << " cmd LC_SEGMENT_64\n";
3752 expected_cmdsize = nsects;
3753 expected_cmdsize *= sizeof(struct MachO::section_64);
3754 expected_cmdsize += sizeof(struct MachO::segment_command_64);
3756 outs() << " cmdsize " << cmdsize;
3757 if (cmdsize != expected_cmdsize)
3758 outs() << " Inconsistent size\n";
3761 outs() << " segname " << SegName << "\n";
3762 if (cmd == MachO::LC_SEGMENT_64) {
3763 outs() << " vmaddr " << format("0x%016" PRIx64, vmaddr) << "\n";
3764 outs() << " vmsize " << format("0x%016" PRIx64, vmsize) << "\n";
3766 outs() << " vmaddr " << format("0x%08" PRIx64, vmaddr) << "\n";
3767 outs() << " vmsize " << format("0x%08" PRIx64, vmsize) << "\n";
3769 outs() << " fileoff " << fileoff;
3770 if (fileoff > object_size)
3771 outs() << " (past end of file)\n";
3774 outs() << " filesize " << filesize;
3775 if (fileoff + filesize > object_size)
3776 outs() << " (past end of file)\n";
3781 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
3782 MachO::VM_PROT_EXECUTE)) != 0)
3783 outs() << " maxprot ?" << format("0x%08" PRIx32, maxprot) << "\n";
3785 if (maxprot & MachO::VM_PROT_READ)
3786 outs() << " maxprot r";
3788 outs() << " maxprot -";
3789 if (maxprot & MachO::VM_PROT_WRITE)
3793 if (maxprot & MachO::VM_PROT_EXECUTE)
3799 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
3800 MachO::VM_PROT_EXECUTE)) != 0)
3801 outs() << " initprot ?" << format("0x%08" PRIx32, initprot) << "\n";
3803 if (initprot & MachO::VM_PROT_READ)
3804 outs() << " initprot r";
3806 outs() << " initprot -";
3807 if (initprot & MachO::VM_PROT_WRITE)
3811 if (initprot & MachO::VM_PROT_EXECUTE)
3817 outs() << " maxprot " << format("0x%08" PRIx32, maxprot) << "\n";
3818 outs() << " initprot " << format("0x%08" PRIx32, initprot) << "\n";
3820 outs() << " nsects " << nsects << "\n";
3824 outs() << " (none)\n";
3826 if (flags & MachO::SG_HIGHVM) {
3827 outs() << " HIGHVM";
3828 flags &= ~MachO::SG_HIGHVM;
3830 if (flags & MachO::SG_FVMLIB) {
3831 outs() << " FVMLIB";
3832 flags &= ~MachO::SG_FVMLIB;
3834 if (flags & MachO::SG_NORELOC) {
3835 outs() << " NORELOC";
3836 flags &= ~MachO::SG_NORELOC;
3838 if (flags & MachO::SG_PROTECTED_VERSION_1) {
3839 outs() << " PROTECTED_VERSION_1";
3840 flags &= ~MachO::SG_PROTECTED_VERSION_1;
3843 outs() << format(" 0x%08" PRIx32, flags) << " (unknown flags)\n";
3848 outs() << " flags " << format("0x%" PRIx32, flags) << "\n";
3852 static void PrintSection(const char *sectname, const char *segname,
3853 uint64_t addr, uint64_t size, uint32_t offset,
3854 uint32_t align, uint32_t reloff, uint32_t nreloc,
3855 uint32_t flags, uint32_t reserved1, uint32_t reserved2,
3856 uint32_t cmd, const char *sg_segname,
3857 uint32_t filetype, uint32_t object_size,
3859 outs() << "Section\n";
3860 outs() << " sectname " << format("%.16s\n", sectname);
3861 outs() << " segname " << format("%.16s", segname);
3862 if (filetype != MachO::MH_OBJECT && strncmp(sg_segname, segname, 16) != 0)
3863 outs() << " (does not match segment)\n";
3866 if (cmd == MachO::LC_SEGMENT_64) {
3867 outs() << " addr " << format("0x%016" PRIx64, addr) << "\n";
3868 outs() << " size " << format("0x%016" PRIx64, size);
3870 outs() << " addr " << format("0x%08" PRIx64, addr) << "\n";
3871 outs() << " size " << format("0x%08" PRIx64, size);
3873 if ((flags & MachO::S_ZEROFILL) != 0 && offset + size > object_size)
3874 outs() << " (past end of file)\n";
3877 outs() << " offset " << offset;
3878 if (offset > object_size)
3879 outs() << " (past end of file)\n";
3882 uint32_t align_shifted = 1 << align;
3883 outs() << " align 2^" << align << " (" << align_shifted << ")\n";
3884 outs() << " reloff " << reloff;
3885 if (reloff > object_size)
3886 outs() << " (past end of file)\n";
3889 outs() << " nreloc " << nreloc;
3890 if (reloff + nreloc * sizeof(struct MachO::relocation_info) > object_size)
3891 outs() << " (past end of file)\n";
3894 uint32_t section_type = flags & MachO::SECTION_TYPE;
3897 if (section_type == MachO::S_REGULAR)
3898 outs() << " S_REGULAR\n";
3899 else if (section_type == MachO::S_ZEROFILL)
3900 outs() << " S_ZEROFILL\n";
3901 else if (section_type == MachO::S_CSTRING_LITERALS)
3902 outs() << " S_CSTRING_LITERALS\n";
3903 else if (section_type == MachO::S_4BYTE_LITERALS)
3904 outs() << " S_4BYTE_LITERALS\n";
3905 else if (section_type == MachO::S_8BYTE_LITERALS)
3906 outs() << " S_8BYTE_LITERALS\n";
3907 else if (section_type == MachO::S_16BYTE_LITERALS)
3908 outs() << " S_16BYTE_LITERALS\n";
3909 else if (section_type == MachO::S_LITERAL_POINTERS)
3910 outs() << " S_LITERAL_POINTERS\n";
3911 else if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS)
3912 outs() << " S_NON_LAZY_SYMBOL_POINTERS\n";
3913 else if (section_type == MachO::S_LAZY_SYMBOL_POINTERS)
3914 outs() << " S_LAZY_SYMBOL_POINTERS\n";
3915 else if (section_type == MachO::S_SYMBOL_STUBS)
3916 outs() << " S_SYMBOL_STUBS\n";
3917 else if (section_type == MachO::S_MOD_INIT_FUNC_POINTERS)
3918 outs() << " S_MOD_INIT_FUNC_POINTERS\n";
3919 else if (section_type == MachO::S_MOD_TERM_FUNC_POINTERS)
3920 outs() << " S_MOD_TERM_FUNC_POINTERS\n";
3921 else if (section_type == MachO::S_COALESCED)
3922 outs() << " S_COALESCED\n";
3923 else if (section_type == MachO::S_INTERPOSING)
3924 outs() << " S_INTERPOSING\n";
3925 else if (section_type == MachO::S_DTRACE_DOF)
3926 outs() << " S_DTRACE_DOF\n";
3927 else if (section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS)
3928 outs() << " S_LAZY_DYLIB_SYMBOL_POINTERS\n";
3929 else if (section_type == MachO::S_THREAD_LOCAL_REGULAR)
3930 outs() << " S_THREAD_LOCAL_REGULAR\n";
3931 else if (section_type == MachO::S_THREAD_LOCAL_ZEROFILL)
3932 outs() << " S_THREAD_LOCAL_ZEROFILL\n";
3933 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLES)
3934 outs() << " S_THREAD_LOCAL_VARIABLES\n";
3935 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
3936 outs() << " S_THREAD_LOCAL_VARIABLE_POINTERS\n";
3937 else if (section_type == MachO::S_THREAD_LOCAL_INIT_FUNCTION_POINTERS)
3938 outs() << " S_THREAD_LOCAL_INIT_FUNCTION_POINTERS\n";
3940 outs() << format("0x%08" PRIx32, section_type) << "\n";
3941 outs() << "attributes";
3942 uint32_t section_attributes = flags & MachO::SECTION_ATTRIBUTES;
3943 if (section_attributes & MachO::S_ATTR_PURE_INSTRUCTIONS)
3944 outs() << " PURE_INSTRUCTIONS";
3945 if (section_attributes & MachO::S_ATTR_NO_TOC)
3946 outs() << " NO_TOC";
3947 if (section_attributes & MachO::S_ATTR_STRIP_STATIC_SYMS)
3948 outs() << " STRIP_STATIC_SYMS";
3949 if (section_attributes & MachO::S_ATTR_NO_DEAD_STRIP)
3950 outs() << " NO_DEAD_STRIP";
3951 if (section_attributes & MachO::S_ATTR_LIVE_SUPPORT)
3952 outs() << " LIVE_SUPPORT";
3953 if (section_attributes & MachO::S_ATTR_SELF_MODIFYING_CODE)
3954 outs() << " SELF_MODIFYING_CODE";
3955 if (section_attributes & MachO::S_ATTR_DEBUG)
3957 if (section_attributes & MachO::S_ATTR_SOME_INSTRUCTIONS)
3958 outs() << " SOME_INSTRUCTIONS";
3959 if (section_attributes & MachO::S_ATTR_EXT_RELOC)
3960 outs() << " EXT_RELOC";
3961 if (section_attributes & MachO::S_ATTR_LOC_RELOC)
3962 outs() << " LOC_RELOC";
3963 if (section_attributes == 0)
3964 outs() << " (none)";
3967 outs() << " flags " << format("0x%08" PRIx32, flags) << "\n";
3968 outs() << " reserved1 " << reserved1;
3969 if (section_type == MachO::S_SYMBOL_STUBS ||
3970 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
3971 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
3972 section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
3973 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
3974 outs() << " (index into indirect symbol table)\n";
3977 outs() << " reserved2 " << reserved2;
3978 if (section_type == MachO::S_SYMBOL_STUBS)
3979 outs() << " (size of stubs)\n";
3984 static void PrintSymtabLoadCommand(MachO::symtab_command st, bool Is64Bit,
3985 uint32_t object_size) {
3986 outs() << " cmd LC_SYMTAB\n";
3987 outs() << " cmdsize " << st.cmdsize;
3988 if (st.cmdsize != sizeof(struct MachO::symtab_command))
3989 outs() << " Incorrect size\n";
3992 outs() << " symoff " << st.symoff;
3993 if (st.symoff > object_size)
3994 outs() << " (past end of file)\n";
3997 outs() << " nsyms " << st.nsyms;
4000 big_size = st.nsyms;
4001 big_size *= sizeof(struct MachO::nlist_64);
4002 big_size += st.symoff;
4003 if (big_size > object_size)
4004 outs() << " (past end of file)\n";
4008 big_size = st.nsyms;
4009 big_size *= sizeof(struct MachO::nlist);
4010 big_size += st.symoff;
4011 if (big_size > object_size)
4012 outs() << " (past end of file)\n";
4016 outs() << " stroff " << st.stroff;
4017 if (st.stroff > object_size)
4018 outs() << " (past end of file)\n";
4021 outs() << " strsize " << st.strsize;
4022 big_size = st.stroff;
4023 big_size += st.strsize;
4024 if (big_size > object_size)
4025 outs() << " (past end of file)\n";
4030 static void PrintDysymtabLoadCommand(MachO::dysymtab_command dyst,
4031 uint32_t nsyms, uint32_t object_size,
4033 outs() << " cmd LC_DYSYMTAB\n";
4034 outs() << " cmdsize " << dyst.cmdsize;
4035 if (dyst.cmdsize != sizeof(struct MachO::dysymtab_command))
4036 outs() << " Incorrect size\n";
4039 outs() << " ilocalsym " << dyst.ilocalsym;
4040 if (dyst.ilocalsym > nsyms)
4041 outs() << " (greater than the number of symbols)\n";
4044 outs() << " nlocalsym " << dyst.nlocalsym;
4046 big_size = dyst.ilocalsym;
4047 big_size += dyst.nlocalsym;
4048 if (big_size > nsyms)
4049 outs() << " (past the end of the symbol table)\n";
4052 outs() << " iextdefsym " << dyst.iextdefsym;
4053 if (dyst.iextdefsym > nsyms)
4054 outs() << " (greater than the number of symbols)\n";
4057 outs() << " nextdefsym " << dyst.nextdefsym;
4058 big_size = dyst.iextdefsym;
4059 big_size += dyst.nextdefsym;
4060 if (big_size > nsyms)
4061 outs() << " (past the end of the symbol table)\n";
4064 outs() << " iundefsym " << dyst.iundefsym;
4065 if (dyst.iundefsym > nsyms)
4066 outs() << " (greater than the number of symbols)\n";
4069 outs() << " nundefsym " << dyst.nundefsym;
4070 big_size = dyst.iundefsym;
4071 big_size += dyst.nundefsym;
4072 if (big_size > nsyms)
4073 outs() << " (past the end of the symbol table)\n";
4076 outs() << " tocoff " << dyst.tocoff;
4077 if (dyst.tocoff > object_size)
4078 outs() << " (past end of file)\n";
4081 outs() << " ntoc " << dyst.ntoc;
4082 big_size = dyst.ntoc;
4083 big_size *= sizeof(struct MachO::dylib_table_of_contents);
4084 big_size += dyst.tocoff;
4085 if (big_size > object_size)
4086 outs() << " (past end of file)\n";
4089 outs() << " modtaboff " << dyst.modtaboff;
4090 if (dyst.modtaboff > object_size)
4091 outs() << " (past end of file)\n";
4094 outs() << " nmodtab " << dyst.nmodtab;
4097 modtabend = dyst.nmodtab;
4098 modtabend *= sizeof(struct MachO::dylib_module_64);
4099 modtabend += dyst.modtaboff;
4101 modtabend = dyst.nmodtab;
4102 modtabend *= sizeof(struct MachO::dylib_module);
4103 modtabend += dyst.modtaboff;
4105 if (modtabend > object_size)
4106 outs() << " (past end of file)\n";
4109 outs() << " extrefsymoff " << dyst.extrefsymoff;
4110 if (dyst.extrefsymoff > object_size)
4111 outs() << " (past end of file)\n";
4114 outs() << " nextrefsyms " << dyst.nextrefsyms;
4115 big_size = dyst.nextrefsyms;
4116 big_size *= sizeof(struct MachO::dylib_reference);
4117 big_size += dyst.extrefsymoff;
4118 if (big_size > object_size)
4119 outs() << " (past end of file)\n";
4122 outs() << " indirectsymoff " << dyst.indirectsymoff;
4123 if (dyst.indirectsymoff > object_size)
4124 outs() << " (past end of file)\n";
4127 outs() << " nindirectsyms " << dyst.nindirectsyms;
4128 big_size = dyst.nindirectsyms;
4129 big_size *= sizeof(uint32_t);
4130 big_size += dyst.indirectsymoff;
4131 if (big_size > object_size)
4132 outs() << " (past end of file)\n";
4135 outs() << " extreloff " << dyst.extreloff;
4136 if (dyst.extreloff > object_size)
4137 outs() << " (past end of file)\n";
4140 outs() << " nextrel " << dyst.nextrel;
4141 big_size = dyst.nextrel;
4142 big_size *= sizeof(struct MachO::relocation_info);
4143 big_size += dyst.extreloff;
4144 if (big_size > object_size)
4145 outs() << " (past end of file)\n";
4148 outs() << " locreloff " << dyst.locreloff;
4149 if (dyst.locreloff > object_size)
4150 outs() << " (past end of file)\n";
4153 outs() << " nlocrel " << dyst.nlocrel;
4154 big_size = dyst.nlocrel;
4155 big_size *= sizeof(struct MachO::relocation_info);
4156 big_size += dyst.locreloff;
4157 if (big_size > object_size)
4158 outs() << " (past end of file)\n";
4163 static void PrintDyldInfoLoadCommand(MachO::dyld_info_command dc,
4164 uint32_t object_size) {
4165 if (dc.cmd == MachO::LC_DYLD_INFO)
4166 outs() << " cmd LC_DYLD_INFO\n";
4168 outs() << " cmd LC_DYLD_INFO_ONLY\n";
4169 outs() << " cmdsize " << dc.cmdsize;
4170 if (dc.cmdsize != sizeof(struct MachO::dyld_info_command))
4171 outs() << " Incorrect size\n";
4174 outs() << " rebase_off " << dc.rebase_off;
4175 if (dc.rebase_off > object_size)
4176 outs() << " (past end of file)\n";
4179 outs() << " rebase_size " << dc.rebase_size;
4181 big_size = dc.rebase_off;
4182 big_size += dc.rebase_size;
4183 if (big_size > object_size)
4184 outs() << " (past end of file)\n";
4187 outs() << " bind_off " << dc.bind_off;
4188 if (dc.bind_off > object_size)
4189 outs() << " (past end of file)\n";
4192 outs() << " bind_size " << dc.bind_size;
4193 big_size = dc.bind_off;
4194 big_size += dc.bind_size;
4195 if (big_size > object_size)
4196 outs() << " (past end of file)\n";
4199 outs() << " weak_bind_off " << dc.weak_bind_off;
4200 if (dc.weak_bind_off > object_size)
4201 outs() << " (past end of file)\n";
4204 outs() << " weak_bind_size " << dc.weak_bind_size;
4205 big_size = dc.weak_bind_off;
4206 big_size += dc.weak_bind_size;
4207 if (big_size > object_size)
4208 outs() << " (past end of file)\n";
4211 outs() << " lazy_bind_off " << dc.lazy_bind_off;
4212 if (dc.lazy_bind_off > object_size)
4213 outs() << " (past end of file)\n";
4216 outs() << " lazy_bind_size " << dc.lazy_bind_size;
4217 big_size = dc.lazy_bind_off;
4218 big_size += dc.lazy_bind_size;
4219 if (big_size > object_size)
4220 outs() << " (past end of file)\n";
4223 outs() << " export_off " << dc.export_off;
4224 if (dc.export_off > object_size)
4225 outs() << " (past end of file)\n";
4228 outs() << " export_size " << dc.export_size;
4229 big_size = dc.export_off;
4230 big_size += dc.export_size;
4231 if (big_size > object_size)
4232 outs() << " (past end of file)\n";
4237 static void PrintDyldLoadCommand(MachO::dylinker_command dyld,
4239 if (dyld.cmd == MachO::LC_ID_DYLINKER)
4240 outs() << " cmd LC_ID_DYLINKER\n";
4241 else if (dyld.cmd == MachO::LC_LOAD_DYLINKER)
4242 outs() << " cmd LC_LOAD_DYLINKER\n";
4243 else if (dyld.cmd == MachO::LC_DYLD_ENVIRONMENT)
4244 outs() << " cmd LC_DYLD_ENVIRONMENT\n";
4246 outs() << " cmd ?(" << dyld.cmd << ")\n";
4247 outs() << " cmdsize " << dyld.cmdsize;
4248 if (dyld.cmdsize < sizeof(struct MachO::dylinker_command))
4249 outs() << " Incorrect size\n";
4252 if (dyld.name >= dyld.cmdsize)
4253 outs() << " name ?(bad offset " << dyld.name << ")\n";
4255 const char *P = (const char *)(Ptr) + dyld.name;
4256 outs() << " name " << P << " (offset " << dyld.name << ")\n";
4260 static void PrintUuidLoadCommand(MachO::uuid_command uuid) {
4261 outs() << " cmd LC_UUID\n";
4262 outs() << " cmdsize " << uuid.cmdsize;
4263 if (uuid.cmdsize != sizeof(struct MachO::uuid_command))
4264 outs() << " Incorrect size\n";
4268 outs() << format("%02" PRIX32, uuid.uuid[0]);
4269 outs() << format("%02" PRIX32, uuid.uuid[1]);
4270 outs() << format("%02" PRIX32, uuid.uuid[2]);
4271 outs() << format("%02" PRIX32, uuid.uuid[3]);
4273 outs() << format("%02" PRIX32, uuid.uuid[4]);
4274 outs() << format("%02" PRIX32, uuid.uuid[5]);
4276 outs() << format("%02" PRIX32, uuid.uuid[6]);
4277 outs() << format("%02" PRIX32, uuid.uuid[7]);
4279 outs() << format("%02" PRIX32, uuid.uuid[8]);
4280 outs() << format("%02" PRIX32, uuid.uuid[9]);
4282 outs() << format("%02" PRIX32, uuid.uuid[10]);
4283 outs() << format("%02" PRIX32, uuid.uuid[11]);
4284 outs() << format("%02" PRIX32, uuid.uuid[12]);
4285 outs() << format("%02" PRIX32, uuid.uuid[13]);
4286 outs() << format("%02" PRIX32, uuid.uuid[14]);
4287 outs() << format("%02" PRIX32, uuid.uuid[15]);
4291 static void PrintRpathLoadCommand(MachO::rpath_command rpath, const char *Ptr) {
4292 outs() << " cmd LC_RPATH\n";
4293 outs() << " cmdsize " << rpath.cmdsize;
4294 if (rpath.cmdsize < sizeof(struct MachO::rpath_command))
4295 outs() << " Incorrect size\n";
4298 if (rpath.path >= rpath.cmdsize)
4299 outs() << " path ?(bad offset " << rpath.path << ")\n";
4301 const char *P = (const char *)(Ptr) + rpath.path;
4302 outs() << " path " << P << " (offset " << rpath.path << ")\n";
4306 static void PrintVersionMinLoadCommand(MachO::version_min_command vd) {
4307 if (vd.cmd == MachO::LC_VERSION_MIN_MACOSX)
4308 outs() << " cmd LC_VERSION_MIN_MACOSX\n";
4309 else if (vd.cmd == MachO::LC_VERSION_MIN_IPHONEOS)
4310 outs() << " cmd LC_VERSION_MIN_IPHONEOS\n";
4312 outs() << " cmd " << vd.cmd << " (?)\n";
4313 outs() << " cmdsize " << vd.cmdsize;
4314 if (vd.cmdsize != sizeof(struct MachO::version_min_command))
4315 outs() << " Incorrect size\n";
4318 outs() << " version " << ((vd.version >> 16) & 0xffff) << "."
4319 << ((vd.version >> 8) & 0xff);
4320 if ((vd.version & 0xff) != 0)
4321 outs() << "." << (vd.version & 0xff);
4324 outs() << " sdk n/a";
4326 outs() << " sdk " << ((vd.sdk >> 16) & 0xffff) << "."
4327 << ((vd.sdk >> 8) & 0xff);
4329 if ((vd.sdk & 0xff) != 0)
4330 outs() << "." << (vd.sdk & 0xff);
4334 static void PrintSourceVersionCommand(MachO::source_version_command sd) {
4335 outs() << " cmd LC_SOURCE_VERSION\n";
4336 outs() << " cmdsize " << sd.cmdsize;
4337 if (sd.cmdsize != sizeof(struct MachO::source_version_command))
4338 outs() << " Incorrect size\n";
4341 uint64_t a = (sd.version >> 40) & 0xffffff;
4342 uint64_t b = (sd.version >> 30) & 0x3ff;
4343 uint64_t c = (sd.version >> 20) & 0x3ff;
4344 uint64_t d = (sd.version >> 10) & 0x3ff;
4345 uint64_t e = sd.version & 0x3ff;
4346 outs() << " version " << a << "." << b;
4348 outs() << "." << c << "." << d << "." << e;
4350 outs() << "." << c << "." << d;
4356 static void PrintEntryPointCommand(MachO::entry_point_command ep) {
4357 outs() << " cmd LC_MAIN\n";
4358 outs() << " cmdsize " << ep.cmdsize;
4359 if (ep.cmdsize != sizeof(struct MachO::entry_point_command))
4360 outs() << " Incorrect size\n";
4363 outs() << " entryoff " << ep.entryoff << "\n";
4364 outs() << " stacksize " << ep.stacksize << "\n";
4367 static void PrintEncryptionInfoCommand(MachO::encryption_info_command ec,
4368 uint32_t object_size) {
4369 outs() << " cmd LC_ENCRYPTION_INFO\n";
4370 outs() << " cmdsize " << ec.cmdsize;
4371 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command))
4372 outs() << " Incorrect size\n";
4375 outs() << " cryptoff " << ec.cryptoff;
4376 if (ec.cryptoff > object_size)
4377 outs() << " (past end of file)\n";
4380 outs() << " cryptsize " << ec.cryptsize;
4381 if (ec.cryptsize > object_size)
4382 outs() << " (past end of file)\n";
4385 outs() << " cryptid " << ec.cryptid << "\n";
4388 static void PrintEncryptionInfoCommand64(MachO::encryption_info_command_64 ec,
4389 uint32_t object_size) {
4390 outs() << " cmd LC_ENCRYPTION_INFO_64\n";
4391 outs() << " cmdsize " << ec.cmdsize;
4392 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command_64))
4393 outs() << " Incorrect size\n";
4396 outs() << " cryptoff " << ec.cryptoff;
4397 if (ec.cryptoff > object_size)
4398 outs() << " (past end of file)\n";
4401 outs() << " cryptsize " << ec.cryptsize;
4402 if (ec.cryptsize > object_size)
4403 outs() << " (past end of file)\n";
4406 outs() << " cryptid " << ec.cryptid << "\n";
4407 outs() << " pad " << ec.pad << "\n";
4410 static void PrintLinkerOptionCommand(MachO::linker_option_command lo,
4412 outs() << " cmd LC_LINKER_OPTION\n";
4413 outs() << " cmdsize " << lo.cmdsize;
4414 if (lo.cmdsize < sizeof(struct MachO::linker_option_command))
4415 outs() << " Incorrect size\n";
4418 outs() << " count " << lo.count << "\n";
4419 const char *string = Ptr + sizeof(struct MachO::linker_option_command);
4420 uint32_t left = lo.cmdsize - sizeof(struct MachO::linker_option_command);
4423 while (*string == '\0' && left > 0) {
4429 outs() << " string #" << i << " " << format("%.*s\n", left, string);
4430 uint32_t NullPos = StringRef(string, left).find('\0');
4431 uint32_t len = std::min(NullPos, left) + 1;
4437 outs() << " count " << lo.count << " does not match number of strings "
4441 static void PrintSubFrameworkCommand(MachO::sub_framework_command sub,
4443 outs() << " cmd LC_SUB_FRAMEWORK\n";
4444 outs() << " cmdsize " << sub.cmdsize;
4445 if (sub.cmdsize < sizeof(struct MachO::sub_framework_command))
4446 outs() << " Incorrect size\n";
4449 if (sub.umbrella < sub.cmdsize) {
4450 const char *P = Ptr + sub.umbrella;
4451 outs() << " umbrella " << P << " (offset " << sub.umbrella << ")\n";
4453 outs() << " umbrella ?(bad offset " << sub.umbrella << ")\n";
4457 static void PrintSubUmbrellaCommand(MachO::sub_umbrella_command sub,
4459 outs() << " cmd LC_SUB_UMBRELLA\n";
4460 outs() << " cmdsize " << sub.cmdsize;
4461 if (sub.cmdsize < sizeof(struct MachO::sub_umbrella_command))
4462 outs() << " Incorrect size\n";
4465 if (sub.sub_umbrella < sub.cmdsize) {
4466 const char *P = Ptr + sub.sub_umbrella;
4467 outs() << " sub_umbrella " << P << " (offset " << sub.sub_umbrella << ")\n";
4469 outs() << " sub_umbrella ?(bad offset " << sub.sub_umbrella << ")\n";
4473 static void PrintSubLibraryCommand(MachO::sub_library_command sub,
4475 outs() << " cmd LC_SUB_LIBRARY\n";
4476 outs() << " cmdsize " << sub.cmdsize;
4477 if (sub.cmdsize < sizeof(struct MachO::sub_library_command))
4478 outs() << " Incorrect size\n";
4481 if (sub.sub_library < sub.cmdsize) {
4482 const char *P = Ptr + sub.sub_library;
4483 outs() << " sub_library " << P << " (offset " << sub.sub_library << ")\n";
4485 outs() << " sub_library ?(bad offset " << sub.sub_library << ")\n";
4489 static void PrintSubClientCommand(MachO::sub_client_command sub,
4491 outs() << " cmd LC_SUB_CLIENT\n";
4492 outs() << " cmdsize " << sub.cmdsize;
4493 if (sub.cmdsize < sizeof(struct MachO::sub_client_command))
4494 outs() << " Incorrect size\n";
4497 if (sub.client < sub.cmdsize) {
4498 const char *P = Ptr + sub.client;
4499 outs() << " client " << P << " (offset " << sub.client << ")\n";
4501 outs() << " client ?(bad offset " << sub.client << ")\n";
4505 static void PrintRoutinesCommand(MachO::routines_command r) {
4506 outs() << " cmd LC_ROUTINES\n";
4507 outs() << " cmdsize " << r.cmdsize;
4508 if (r.cmdsize != sizeof(struct MachO::routines_command))
4509 outs() << " Incorrect size\n";
4512 outs() << " init_address " << format("0x%08" PRIx32, r.init_address) << "\n";
4513 outs() << " init_module " << r.init_module << "\n";
4514 outs() << " reserved1 " << r.reserved1 << "\n";
4515 outs() << " reserved2 " << r.reserved2 << "\n";
4516 outs() << " reserved3 " << r.reserved3 << "\n";
4517 outs() << " reserved4 " << r.reserved4 << "\n";
4518 outs() << " reserved5 " << r.reserved5 << "\n";
4519 outs() << " reserved6 " << r.reserved6 << "\n";
4522 static void PrintRoutinesCommand64(MachO::routines_command_64 r) {
4523 outs() << " cmd LC_ROUTINES_64\n";
4524 outs() << " cmdsize " << r.cmdsize;
4525 if (r.cmdsize != sizeof(struct MachO::routines_command_64))
4526 outs() << " Incorrect size\n";
4529 outs() << " init_address " << format("0x%016" PRIx64, r.init_address) << "\n";
4530 outs() << " init_module " << r.init_module << "\n";
4531 outs() << " reserved1 " << r.reserved1 << "\n";
4532 outs() << " reserved2 " << r.reserved2 << "\n";
4533 outs() << " reserved3 " << r.reserved3 << "\n";
4534 outs() << " reserved4 " << r.reserved4 << "\n";
4535 outs() << " reserved5 " << r.reserved5 << "\n";
4536 outs() << " reserved6 " << r.reserved6 << "\n";
4539 static void Print_x86_thread_state64_t(MachO::x86_thread_state64_t &cpu64) {
4540 outs() << " rax " << format("0x%016" PRIx64, cpu64.rax);
4541 outs() << " rbx " << format("0x%016" PRIx64, cpu64.rbx);
4542 outs() << " rcx " << format("0x%016" PRIx64, cpu64.rcx) << "\n";
4543 outs() << " rdx " << format("0x%016" PRIx64, cpu64.rdx);
4544 outs() << " rdi " << format("0x%016" PRIx64, cpu64.rdi);
4545 outs() << " rsi " << format("0x%016" PRIx64, cpu64.rsi) << "\n";
4546 outs() << " rbp " << format("0x%016" PRIx64, cpu64.rbp);
4547 outs() << " rsp " << format("0x%016" PRIx64, cpu64.rsp);
4548 outs() << " r8 " << format("0x%016" PRIx64, cpu64.r8) << "\n";
4549 outs() << " r9 " << format("0x%016" PRIx64, cpu64.r9);
4550 outs() << " r10 " << format("0x%016" PRIx64, cpu64.r10);
4551 outs() << " r11 " << format("0x%016" PRIx64, cpu64.r11) << "\n";
4552 outs() << " r12 " << format("0x%016" PRIx64, cpu64.r12);
4553 outs() << " r13 " << format("0x%016" PRIx64, cpu64.r13);
4554 outs() << " r14 " << format("0x%016" PRIx64, cpu64.r14) << "\n";
4555 outs() << " r15 " << format("0x%016" PRIx64, cpu64.r15);
4556 outs() << " rip " << format("0x%016" PRIx64, cpu64.rip) << "\n";
4557 outs() << "rflags " << format("0x%016" PRIx64, cpu64.rflags);
4558 outs() << " cs " << format("0x%016" PRIx64, cpu64.cs);
4559 outs() << " fs " << format("0x%016" PRIx64, cpu64.fs) << "\n";
4560 outs() << " gs " << format("0x%016" PRIx64, cpu64.gs) << "\n";
4563 static void Print_mmst_reg(MachO::mmst_reg_t &r) {
4565 outs() << "\t mmst_reg ";
4566 for (f = 0; f < 10; f++)
4567 outs() << format("%02" PRIx32, (r.mmst_reg[f] & 0xff)) << " ";
4569 outs() << "\t mmst_rsrv ";
4570 for (f = 0; f < 6; f++)
4571 outs() << format("%02" PRIx32, (r.mmst_rsrv[f] & 0xff)) << " ";
4575 static void Print_xmm_reg(MachO::xmm_reg_t &r) {
4577 outs() << "\t xmm_reg ";
4578 for (f = 0; f < 16; f++)
4579 outs() << format("%02" PRIx32, (r.xmm_reg[f] & 0xff)) << " ";
4583 static void Print_x86_float_state_t(MachO::x86_float_state64_t &fpu) {
4584 outs() << "\t fpu_reserved[0] " << fpu.fpu_reserved[0];
4585 outs() << " fpu_reserved[1] " << fpu.fpu_reserved[1] << "\n";
4586 outs() << "\t control: invalid " << fpu.fpu_fcw.invalid;
4587 outs() << " denorm " << fpu.fpu_fcw.denorm;
4588 outs() << " zdiv " << fpu.fpu_fcw.zdiv;
4589 outs() << " ovrfl " << fpu.fpu_fcw.ovrfl;
4590 outs() << " undfl " << fpu.fpu_fcw.undfl;
4591 outs() << " precis " << fpu.fpu_fcw.precis << "\n";
4592 outs() << "\t\t pc ";
4593 if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_24B)
4594 outs() << "FP_PREC_24B ";
4595 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_53B)
4596 outs() << "FP_PREC_53B ";
4597 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_64B)
4598 outs() << "FP_PREC_64B ";
4600 outs() << fpu.fpu_fcw.pc << " ";
4602 if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_NEAR)
4603 outs() << "FP_RND_NEAR ";
4604 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_DOWN)
4605 outs() << "FP_RND_DOWN ";
4606 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_UP)
4607 outs() << "FP_RND_UP ";
4608 else if (fpu.fpu_fcw.rc == MachO::x86_FP_CHOP)
4609 outs() << "FP_CHOP ";
4611 outs() << "\t status: invalid " << fpu.fpu_fsw.invalid;
4612 outs() << " denorm " << fpu.fpu_fsw.denorm;
4613 outs() << " zdiv " << fpu.fpu_fsw.zdiv;
4614 outs() << " ovrfl " << fpu.fpu_fsw.ovrfl;
4615 outs() << " undfl " << fpu.fpu_fsw.undfl;
4616 outs() << " precis " << fpu.fpu_fsw.precis;
4617 outs() << " stkflt " << fpu.fpu_fsw.stkflt << "\n";
4618 outs() << "\t errsumm " << fpu.fpu_fsw.errsumm;
4619 outs() << " c0 " << fpu.fpu_fsw.c0;
4620 outs() << " c1 " << fpu.fpu_fsw.c1;
4621 outs() << " c2 " << fpu.fpu_fsw.c2;
4622 outs() << " tos " << fpu.fpu_fsw.tos;
4623 outs() << " c3 " << fpu.fpu_fsw.c3;
4624 outs() << " busy " << fpu.fpu_fsw.busy << "\n";
4625 outs() << "\t fpu_ftw " << format("0x%02" PRIx32, fpu.fpu_ftw);
4626 outs() << " fpu_rsrv1 " << format("0x%02" PRIx32, fpu.fpu_rsrv1);
4627 outs() << " fpu_fop " << format("0x%04" PRIx32, fpu.fpu_fop);
4628 outs() << " fpu_ip " << format("0x%08" PRIx32, fpu.fpu_ip) << "\n";
4629 outs() << "\t fpu_cs " << format("0x%04" PRIx32, fpu.fpu_cs);
4630 outs() << " fpu_rsrv2 " << format("0x%04" PRIx32, fpu.fpu_rsrv2);
4631 outs() << " fpu_dp " << format("0x%08" PRIx32, fpu.fpu_dp);
4632 outs() << " fpu_ds " << format("0x%04" PRIx32, fpu.fpu_ds) << "\n";
4633 outs() << "\t fpu_rsrv3 " << format("0x%04" PRIx32, fpu.fpu_rsrv3);
4634 outs() << " fpu_mxcsr " << format("0x%08" PRIx32, fpu.fpu_mxcsr);
4635 outs() << " fpu_mxcsrmask " << format("0x%08" PRIx32, fpu.fpu_mxcsrmask);
4637 outs() << "\t fpu_stmm0:\n";
4638 Print_mmst_reg(fpu.fpu_stmm0);
4639 outs() << "\t fpu_stmm1:\n";
4640 Print_mmst_reg(fpu.fpu_stmm1);
4641 outs() << "\t fpu_stmm2:\n";
4642 Print_mmst_reg(fpu.fpu_stmm2);
4643 outs() << "\t fpu_stmm3:\n";
4644 Print_mmst_reg(fpu.fpu_stmm3);
4645 outs() << "\t fpu_stmm4:\n";
4646 Print_mmst_reg(fpu.fpu_stmm4);
4647 outs() << "\t fpu_stmm5:\n";
4648 Print_mmst_reg(fpu.fpu_stmm5);
4649 outs() << "\t fpu_stmm6:\n";
4650 Print_mmst_reg(fpu.fpu_stmm6);
4651 outs() << "\t fpu_stmm7:\n";
4652 Print_mmst_reg(fpu.fpu_stmm7);
4653 outs() << "\t fpu_xmm0:\n";
4654 Print_xmm_reg(fpu.fpu_xmm0);
4655 outs() << "\t fpu_xmm1:\n";
4656 Print_xmm_reg(fpu.fpu_xmm1);
4657 outs() << "\t fpu_xmm2:\n";
4658 Print_xmm_reg(fpu.fpu_xmm2);
4659 outs() << "\t fpu_xmm3:\n";
4660 Print_xmm_reg(fpu.fpu_xmm3);
4661 outs() << "\t fpu_xmm4:\n";
4662 Print_xmm_reg(fpu.fpu_xmm4);
4663 outs() << "\t fpu_xmm5:\n";
4664 Print_xmm_reg(fpu.fpu_xmm5);
4665 outs() << "\t fpu_xmm6:\n";
4666 Print_xmm_reg(fpu.fpu_xmm6);
4667 outs() << "\t fpu_xmm7:\n";
4668 Print_xmm_reg(fpu.fpu_xmm7);
4669 outs() << "\t fpu_xmm8:\n";
4670 Print_xmm_reg(fpu.fpu_xmm8);
4671 outs() << "\t fpu_xmm9:\n";
4672 Print_xmm_reg(fpu.fpu_xmm9);
4673 outs() << "\t fpu_xmm10:\n";
4674 Print_xmm_reg(fpu.fpu_xmm10);
4675 outs() << "\t fpu_xmm11:\n";
4676 Print_xmm_reg(fpu.fpu_xmm11);
4677 outs() << "\t fpu_xmm12:\n";
4678 Print_xmm_reg(fpu.fpu_xmm12);
4679 outs() << "\t fpu_xmm13:\n";
4680 Print_xmm_reg(fpu.fpu_xmm13);
4681 outs() << "\t fpu_xmm14:\n";
4682 Print_xmm_reg(fpu.fpu_xmm14);
4683 outs() << "\t fpu_xmm15:\n";
4684 Print_xmm_reg(fpu.fpu_xmm15);
4685 outs() << "\t fpu_rsrv4:\n";
4686 for (uint32_t f = 0; f < 6; f++) {
4688 for (uint32_t g = 0; g < 16; g++)
4689 outs() << format("%02" PRIx32, fpu.fpu_rsrv4[f * g]) << " ";
4692 outs() << "\t fpu_reserved1 " << format("0x%08" PRIx32, fpu.fpu_reserved1);
4696 static void Print_x86_exception_state_t(MachO::x86_exception_state64_t &exc64) {
4697 outs() << "\t trapno " << format("0x%08" PRIx32, exc64.trapno);
4698 outs() << " err " << format("0x%08" PRIx32, exc64.err);
4699 outs() << " faultvaddr " << format("0x%016" PRIx64, exc64.faultvaddr) << "\n";
4702 static void PrintThreadCommand(MachO::thread_command t, const char *Ptr,
4703 bool isLittleEndian, uint32_t cputype) {
4704 if (t.cmd == MachO::LC_THREAD)
4705 outs() << " cmd LC_THREAD\n";
4706 else if (t.cmd == MachO::LC_UNIXTHREAD)
4707 outs() << " cmd LC_UNIXTHREAD\n";
4709 outs() << " cmd " << t.cmd << " (unknown)\n";
4710 outs() << " cmdsize " << t.cmdsize;
4711 if (t.cmdsize < sizeof(struct MachO::thread_command) + 2 * sizeof(uint32_t))
4712 outs() << " Incorrect size\n";
4716 const char *begin = Ptr + sizeof(struct MachO::thread_command);
4717 const char *end = Ptr + t.cmdsize;
4718 uint32_t flavor, count, left;
4719 if (cputype == MachO::CPU_TYPE_X86_64) {
4720 while (begin < end) {
4721 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4722 memcpy((char *)&flavor, begin, sizeof(uint32_t));
4723 begin += sizeof(uint32_t);
4728 if (isLittleEndian != sys::IsLittleEndianHost)
4729 sys::swapByteOrder(flavor);
4730 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4731 memcpy((char *)&count, begin, sizeof(uint32_t));
4732 begin += sizeof(uint32_t);
4737 if (isLittleEndian != sys::IsLittleEndianHost)
4738 sys::swapByteOrder(count);
4739 if (flavor == MachO::x86_THREAD_STATE64) {
4740 outs() << " flavor x86_THREAD_STATE64\n";
4741 if (count == MachO::x86_THREAD_STATE64_COUNT)
4742 outs() << " count x86_THREAD_STATE64_COUNT\n";
4744 outs() << " count " << count
4745 << " (not x86_THREAD_STATE64_COUNT)\n";
4746 MachO::x86_thread_state64_t cpu64;
4748 if (left >= sizeof(MachO::x86_thread_state64_t)) {
4749 memcpy(&cpu64, begin, sizeof(MachO::x86_thread_state64_t));
4750 begin += sizeof(MachO::x86_thread_state64_t);
4752 memset(&cpu64, '\0', sizeof(MachO::x86_thread_state64_t));
4753 memcpy(&cpu64, begin, left);
4756 if (isLittleEndian != sys::IsLittleEndianHost)
4758 Print_x86_thread_state64_t(cpu64);
4759 } else if (flavor == MachO::x86_THREAD_STATE) {
4760 outs() << " flavor x86_THREAD_STATE\n";
4761 if (count == MachO::x86_THREAD_STATE_COUNT)
4762 outs() << " count x86_THREAD_STATE_COUNT\n";
4764 outs() << " count " << count
4765 << " (not x86_THREAD_STATE_COUNT)\n";
4766 struct MachO::x86_thread_state_t ts;
4768 if (left >= sizeof(MachO::x86_thread_state_t)) {
4769 memcpy(&ts, begin, sizeof(MachO::x86_thread_state_t));
4770 begin += sizeof(MachO::x86_thread_state_t);
4772 memset(&ts, '\0', sizeof(MachO::x86_thread_state_t));
4773 memcpy(&ts, begin, left);
4776 if (isLittleEndian != sys::IsLittleEndianHost)
4778 if (ts.tsh.flavor == MachO::x86_THREAD_STATE64) {
4779 outs() << "\t tsh.flavor x86_THREAD_STATE64 ";
4780 if (ts.tsh.count == MachO::x86_THREAD_STATE64_COUNT)
4781 outs() << "tsh.count x86_THREAD_STATE64_COUNT\n";
4783 outs() << "tsh.count " << ts.tsh.count
4784 << " (not x86_THREAD_STATE64_COUNT\n";
4785 Print_x86_thread_state64_t(ts.uts.ts64);
4787 outs() << "\t tsh.flavor " << ts.tsh.flavor << " tsh.count "
4788 << ts.tsh.count << "\n";
4790 } else if (flavor == MachO::x86_FLOAT_STATE) {
4791 outs() << " flavor x86_FLOAT_STATE\n";
4792 if (count == MachO::x86_FLOAT_STATE_COUNT)
4793 outs() << " count x86_FLOAT_STATE_COUNT\n";
4795 outs() << " count " << count << " (not x86_FLOAT_STATE_COUNT)\n";
4796 struct MachO::x86_float_state_t fs;
4798 if (left >= sizeof(MachO::x86_float_state_t)) {
4799 memcpy(&fs, begin, sizeof(MachO::x86_float_state_t));
4800 begin += sizeof(MachO::x86_float_state_t);
4802 memset(&fs, '\0', sizeof(MachO::x86_float_state_t));
4803 memcpy(&fs, begin, left);
4806 if (isLittleEndian != sys::IsLittleEndianHost)
4808 if (fs.fsh.flavor == MachO::x86_FLOAT_STATE64) {
4809 outs() << "\t fsh.flavor x86_FLOAT_STATE64 ";
4810 if (fs.fsh.count == MachO::x86_FLOAT_STATE64_COUNT)
4811 outs() << "fsh.count x86_FLOAT_STATE64_COUNT\n";
4813 outs() << "fsh.count " << fs.fsh.count
4814 << " (not x86_FLOAT_STATE64_COUNT\n";
4815 Print_x86_float_state_t(fs.ufs.fs64);
4817 outs() << "\t fsh.flavor " << fs.fsh.flavor << " fsh.count "
4818 << fs.fsh.count << "\n";
4820 } else if (flavor == MachO::x86_EXCEPTION_STATE) {
4821 outs() << " flavor x86_EXCEPTION_STATE\n";
4822 if (count == MachO::x86_EXCEPTION_STATE_COUNT)
4823 outs() << " count x86_EXCEPTION_STATE_COUNT\n";
4825 outs() << " count " << count
4826 << " (not x86_EXCEPTION_STATE_COUNT)\n";
4827 struct MachO::x86_exception_state_t es;
4829 if (left >= sizeof(MachO::x86_exception_state_t)) {
4830 memcpy(&es, begin, sizeof(MachO::x86_exception_state_t));
4831 begin += sizeof(MachO::x86_exception_state_t);
4833 memset(&es, '\0', sizeof(MachO::x86_exception_state_t));
4834 memcpy(&es, begin, left);
4837 if (isLittleEndian != sys::IsLittleEndianHost)
4839 if (es.esh.flavor == MachO::x86_EXCEPTION_STATE64) {
4840 outs() << "\t esh.flavor x86_EXCEPTION_STATE64\n";
4841 if (es.esh.count == MachO::x86_EXCEPTION_STATE64_COUNT)
4842 outs() << "\t esh.count x86_EXCEPTION_STATE64_COUNT\n";
4844 outs() << "\t esh.count " << es.esh.count
4845 << " (not x86_EXCEPTION_STATE64_COUNT\n";
4846 Print_x86_exception_state_t(es.ues.es64);
4848 outs() << "\t esh.flavor " << es.esh.flavor << " esh.count "
4849 << es.esh.count << "\n";
4852 outs() << " flavor " << flavor << " (unknown)\n";
4853 outs() << " count " << count << "\n";
4854 outs() << " state (unknown)\n";
4855 begin += count * sizeof(uint32_t);
4859 while (begin < end) {
4860 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4861 memcpy((char *)&flavor, begin, sizeof(uint32_t));
4862 begin += sizeof(uint32_t);
4867 if (isLittleEndian != sys::IsLittleEndianHost)
4868 sys::swapByteOrder(flavor);
4869 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4870 memcpy((char *)&count, begin, sizeof(uint32_t));
4871 begin += sizeof(uint32_t);
4876 if (isLittleEndian != sys::IsLittleEndianHost)
4877 sys::swapByteOrder(count);
4878 outs() << " flavor " << flavor << "\n";
4879 outs() << " count " << count << "\n";
4880 outs() << " state (Unknown cputype/cpusubtype)\n";
4881 begin += count * sizeof(uint32_t);
4886 static void PrintDylibCommand(MachO::dylib_command dl, const char *Ptr) {
4887 if (dl.cmd == MachO::LC_ID_DYLIB)
4888 outs() << " cmd LC_ID_DYLIB\n";
4889 else if (dl.cmd == MachO::LC_LOAD_DYLIB)
4890 outs() << " cmd LC_LOAD_DYLIB\n";
4891 else if (dl.cmd == MachO::LC_LOAD_WEAK_DYLIB)
4892 outs() << " cmd LC_LOAD_WEAK_DYLIB\n";
4893 else if (dl.cmd == MachO::LC_REEXPORT_DYLIB)
4894 outs() << " cmd LC_REEXPORT_DYLIB\n";
4895 else if (dl.cmd == MachO::LC_LAZY_LOAD_DYLIB)
4896 outs() << " cmd LC_LAZY_LOAD_DYLIB\n";
4897 else if (dl.cmd == MachO::LC_LOAD_UPWARD_DYLIB)
4898 outs() << " cmd LC_LOAD_UPWARD_DYLIB\n";
4900 outs() << " cmd " << dl.cmd << " (unknown)\n";
4901 outs() << " cmdsize " << dl.cmdsize;
4902 if (dl.cmdsize < sizeof(struct MachO::dylib_command))
4903 outs() << " Incorrect size\n";
4906 if (dl.dylib.name < dl.cmdsize) {
4907 const char *P = (const char *)(Ptr) + dl.dylib.name;
4908 outs() << " name " << P << " (offset " << dl.dylib.name << ")\n";
4910 outs() << " name ?(bad offset " << dl.dylib.name << ")\n";
4912 outs() << " time stamp " << dl.dylib.timestamp << " ";
4913 time_t t = dl.dylib.timestamp;
4914 outs() << ctime(&t);
4915 outs() << " current version ";
4916 if (dl.dylib.current_version == 0xffffffff)
4919 outs() << ((dl.dylib.current_version >> 16) & 0xffff) << "."
4920 << ((dl.dylib.current_version >> 8) & 0xff) << "."
4921 << (dl.dylib.current_version & 0xff) << "\n";
4922 outs() << "compatibility version ";
4923 if (dl.dylib.compatibility_version == 0xffffffff)
4926 outs() << ((dl.dylib.compatibility_version >> 16) & 0xffff) << "."
4927 << ((dl.dylib.compatibility_version >> 8) & 0xff) << "."
4928 << (dl.dylib.compatibility_version & 0xff) << "\n";
4931 static void PrintLinkEditDataCommand(MachO::linkedit_data_command ld,
4932 uint32_t object_size) {
4933 if (ld.cmd == MachO::LC_CODE_SIGNATURE)
4934 outs() << " cmd LC_FUNCTION_STARTS\n";
4935 else if (ld.cmd == MachO::LC_SEGMENT_SPLIT_INFO)
4936 outs() << " cmd LC_SEGMENT_SPLIT_INFO\n";
4937 else if (ld.cmd == MachO::LC_FUNCTION_STARTS)
4938 outs() << " cmd LC_FUNCTION_STARTS\n";
4939 else if (ld.cmd == MachO::LC_DATA_IN_CODE)
4940 outs() << " cmd LC_DATA_IN_CODE\n";
4941 else if (ld.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS)
4942 outs() << " cmd LC_DYLIB_CODE_SIGN_DRS\n";
4943 else if (ld.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT)
4944 outs() << " cmd LC_LINKER_OPTIMIZATION_HINT\n";
4946 outs() << " cmd " << ld.cmd << " (?)\n";
4947 outs() << " cmdsize " << ld.cmdsize;
4948 if (ld.cmdsize != sizeof(struct MachO::linkedit_data_command))
4949 outs() << " Incorrect size\n";
4952 outs() << " dataoff " << ld.dataoff;
4953 if (ld.dataoff > object_size)
4954 outs() << " (past end of file)\n";
4957 outs() << " datasize " << ld.datasize;
4958 uint64_t big_size = ld.dataoff;
4959 big_size += ld.datasize;
4960 if (big_size > object_size)
4961 outs() << " (past end of file)\n";
4966 static void PrintLoadCommands(const MachOObjectFile *Obj, uint32_t ncmds,
4967 uint32_t filetype, uint32_t cputype,
4971 StringRef Buf = Obj->getData();
4972 MachOObjectFile::LoadCommandInfo Command = Obj->getFirstLoadCommandInfo();
4973 for (unsigned i = 0;; ++i) {
4974 outs() << "Load command " << i << "\n";
4975 if (Command.C.cmd == MachO::LC_SEGMENT) {
4976 MachO::segment_command SLC = Obj->getSegmentLoadCommand(Command);
4977 const char *sg_segname = SLC.segname;
4978 PrintSegmentCommand(SLC.cmd, SLC.cmdsize, SLC.segname, SLC.vmaddr,
4979 SLC.vmsize, SLC.fileoff, SLC.filesize, SLC.maxprot,
4980 SLC.initprot, SLC.nsects, SLC.flags, Buf.size(),
4982 for (unsigned j = 0; j < SLC.nsects; j++) {
4983 MachO::section S = Obj->getSection(Command, j);
4984 PrintSection(S.sectname, S.segname, S.addr, S.size, S.offset, S.align,
4985 S.reloff, S.nreloc, S.flags, S.reserved1, S.reserved2,
4986 SLC.cmd, sg_segname, filetype, Buf.size(), verbose);
4988 } else if (Command.C.cmd == MachO::LC_SEGMENT_64) {
4989 MachO::segment_command_64 SLC_64 = Obj->getSegment64LoadCommand(Command);
4990 const char *sg_segname = SLC_64.segname;
4991 PrintSegmentCommand(SLC_64.cmd, SLC_64.cmdsize, SLC_64.segname,
4992 SLC_64.vmaddr, SLC_64.vmsize, SLC_64.fileoff,
4993 SLC_64.filesize, SLC_64.maxprot, SLC_64.initprot,
4994 SLC_64.nsects, SLC_64.flags, Buf.size(), verbose);
4995 for (unsigned j = 0; j < SLC_64.nsects; j++) {
4996 MachO::section_64 S_64 = Obj->getSection64(Command, j);
4997 PrintSection(S_64.sectname, S_64.segname, S_64.addr, S_64.size,
4998 S_64.offset, S_64.align, S_64.reloff, S_64.nreloc,
4999 S_64.flags, S_64.reserved1, S_64.reserved2, SLC_64.cmd,
5000 sg_segname, filetype, Buf.size(), verbose);
5002 } else if (Command.C.cmd == MachO::LC_SYMTAB) {
5003 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
5004 PrintSymtabLoadCommand(Symtab, Obj->is64Bit(), Buf.size());
5005 } else if (Command.C.cmd == MachO::LC_DYSYMTAB) {
5006 MachO::dysymtab_command Dysymtab = Obj->getDysymtabLoadCommand();
5007 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
5008 PrintDysymtabLoadCommand(Dysymtab, Symtab.nsyms, Buf.size(),
5010 } else if (Command.C.cmd == MachO::LC_DYLD_INFO ||
5011 Command.C.cmd == MachO::LC_DYLD_INFO_ONLY) {
5012 MachO::dyld_info_command DyldInfo = Obj->getDyldInfoLoadCommand(Command);
5013 PrintDyldInfoLoadCommand(DyldInfo, Buf.size());
5014 } else if (Command.C.cmd == MachO::LC_LOAD_DYLINKER ||
5015 Command.C.cmd == MachO::LC_ID_DYLINKER ||
5016 Command.C.cmd == MachO::LC_DYLD_ENVIRONMENT) {
5017 MachO::dylinker_command Dyld = Obj->getDylinkerCommand(Command);
5018 PrintDyldLoadCommand(Dyld, Command.Ptr);
5019 } else if (Command.C.cmd == MachO::LC_UUID) {
5020 MachO::uuid_command Uuid = Obj->getUuidCommand(Command);
5021 PrintUuidLoadCommand(Uuid);
5022 } else if (Command.C.cmd == MachO::LC_RPATH) {
5023 MachO::rpath_command Rpath = Obj->getRpathCommand(Command);
5024 PrintRpathLoadCommand(Rpath, Command.Ptr);
5025 } else if (Command.C.cmd == MachO::LC_VERSION_MIN_MACOSX ||
5026 Command.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS) {
5027 MachO::version_min_command Vd = Obj->getVersionMinLoadCommand(Command);
5028 PrintVersionMinLoadCommand(Vd);
5029 } else if (Command.C.cmd == MachO::LC_SOURCE_VERSION) {
5030 MachO::source_version_command Sd = Obj->getSourceVersionCommand(Command);
5031 PrintSourceVersionCommand(Sd);
5032 } else if (Command.C.cmd == MachO::LC_MAIN) {
5033 MachO::entry_point_command Ep = Obj->getEntryPointCommand(Command);
5034 PrintEntryPointCommand(Ep);
5035 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO) {
5036 MachO::encryption_info_command Ei =
5037 Obj->getEncryptionInfoCommand(Command);
5038 PrintEncryptionInfoCommand(Ei, Buf.size());
5039 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO_64) {
5040 MachO::encryption_info_command_64 Ei =
5041 Obj->getEncryptionInfoCommand64(Command);
5042 PrintEncryptionInfoCommand64(Ei, Buf.size());
5043 } else if (Command.C.cmd == MachO::LC_LINKER_OPTION) {
5044 MachO::linker_option_command Lo =
5045 Obj->getLinkerOptionLoadCommand(Command);
5046 PrintLinkerOptionCommand(Lo, Command.Ptr);
5047 } else if (Command.C.cmd == MachO::LC_SUB_FRAMEWORK) {
5048 MachO::sub_framework_command Sf = Obj->getSubFrameworkCommand(Command);
5049 PrintSubFrameworkCommand(Sf, Command.Ptr);
5050 } else if (Command.C.cmd == MachO::LC_SUB_UMBRELLA) {
5051 MachO::sub_umbrella_command Sf = Obj->getSubUmbrellaCommand(Command);
5052 PrintSubUmbrellaCommand(Sf, Command.Ptr);
5053 } else if (Command.C.cmd == MachO::LC_SUB_LIBRARY) {
5054 MachO::sub_library_command Sl = Obj->getSubLibraryCommand(Command);
5055 PrintSubLibraryCommand(Sl, Command.Ptr);
5056 } else if (Command.C.cmd == MachO::LC_SUB_CLIENT) {
5057 MachO::sub_client_command Sc = Obj->getSubClientCommand(Command);
5058 PrintSubClientCommand(Sc, Command.Ptr);
5059 } else if (Command.C.cmd == MachO::LC_ROUTINES) {
5060 MachO::routines_command Rc = Obj->getRoutinesCommand(Command);
5061 PrintRoutinesCommand(Rc);
5062 } else if (Command.C.cmd == MachO::LC_ROUTINES_64) {
5063 MachO::routines_command_64 Rc = Obj->getRoutinesCommand64(Command);
5064 PrintRoutinesCommand64(Rc);
5065 } else if (Command.C.cmd == MachO::LC_THREAD ||
5066 Command.C.cmd == MachO::LC_UNIXTHREAD) {
5067 MachO::thread_command Tc = Obj->getThreadCommand(Command);
5068 PrintThreadCommand(Tc, Command.Ptr, Obj->isLittleEndian(), cputype);
5069 } else if (Command.C.cmd == MachO::LC_LOAD_DYLIB ||
5070 Command.C.cmd == MachO::LC_ID_DYLIB ||
5071 Command.C.cmd == MachO::LC_LOAD_WEAK_DYLIB ||
5072 Command.C.cmd == MachO::LC_REEXPORT_DYLIB ||
5073 Command.C.cmd == MachO::LC_LAZY_LOAD_DYLIB ||
5074 Command.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) {
5075 MachO::dylib_command Dl = Obj->getDylibIDLoadCommand(Command);
5076 PrintDylibCommand(Dl, Command.Ptr);
5077 } else if (Command.C.cmd == MachO::LC_CODE_SIGNATURE ||
5078 Command.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO ||
5079 Command.C.cmd == MachO::LC_FUNCTION_STARTS ||
5080 Command.C.cmd == MachO::LC_DATA_IN_CODE ||
5081 Command.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS ||
5082 Command.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) {
5083 MachO::linkedit_data_command Ld =
5084 Obj->getLinkeditDataLoadCommand(Command);
5085 PrintLinkEditDataCommand(Ld, Buf.size());
5087 outs() << " cmd ?(" << format("0x%08" PRIx32, Command.C.cmd)
5089 outs() << " cmdsize " << Command.C.cmdsize << "\n";
5090 // TODO: get and print the raw bytes of the load command.
5092 // TODO: print all the other kinds of load commands.
5096 Command = Obj->getNextLoadCommandInfo(Command);
5100 static void getAndPrintMachHeader(const MachOObjectFile *Obj, uint32_t &ncmds,
5101 uint32_t &filetype, uint32_t &cputype,
5103 if (Obj->is64Bit()) {
5104 MachO::mach_header_64 H_64;
5105 H_64 = Obj->getHeader64();
5106 PrintMachHeader(H_64.magic, H_64.cputype, H_64.cpusubtype, H_64.filetype,
5107 H_64.ncmds, H_64.sizeofcmds, H_64.flags, verbose);
5109 filetype = H_64.filetype;
5110 cputype = H_64.cputype;
5112 MachO::mach_header H;
5113 H = Obj->getHeader();
5114 PrintMachHeader(H.magic, H.cputype, H.cpusubtype, H.filetype, H.ncmds,
5115 H.sizeofcmds, H.flags, verbose);
5117 filetype = H.filetype;
5118 cputype = H.cputype;
5122 void llvm::printMachOFileHeader(const object::ObjectFile *Obj) {
5123 const MachOObjectFile *file = dyn_cast<const MachOObjectFile>(Obj);
5125 uint32_t filetype = 0;
5126 uint32_t cputype = 0;
5127 getAndPrintMachHeader(file, ncmds, filetype, cputype, true);
5128 PrintLoadCommands(file, ncmds, filetype, cputype, true);
5131 //===----------------------------------------------------------------------===//
5132 // export trie dumping
5133 //===----------------------------------------------------------------------===//
5135 void llvm::printMachOExportsTrie(const object::MachOObjectFile *Obj) {
5136 for (const llvm::object::ExportEntry &Entry : Obj->exports()) {
5137 uint64_t Flags = Entry.flags();
5138 bool ReExport = (Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT);
5139 bool WeakDef = (Flags & MachO::EXPORT_SYMBOL_FLAGS_WEAK_DEFINITION);
5140 bool ThreadLocal = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
5141 MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL);
5142 bool Abs = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
5143 MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE);
5144 bool Resolver = (Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER);
5146 outs() << "[re-export] ";
5148 outs() << format("0x%08llX ",
5149 Entry.address()); // FIXME:add in base address
5150 outs() << Entry.name();
5151 if (WeakDef || ThreadLocal || Resolver || Abs) {
5152 bool NeedsComma = false;
5155 outs() << "weak_def";
5161 outs() << "per-thread";
5167 outs() << "absolute";
5173 outs() << format("resolver=0x%08llX", Entry.other());
5179 StringRef DylibName = "unknown";
5180 int Ordinal = Entry.other() - 1;
5181 Obj->getLibraryShortNameByIndex(Ordinal, DylibName);
5182 if (Entry.otherName().empty())
5183 outs() << " (from " << DylibName << ")";
5185 outs() << " (" << Entry.otherName() << " from " << DylibName << ")";
5191 //===----------------------------------------------------------------------===//
5192 // rebase table dumping
5193 //===----------------------------------------------------------------------===//
5198 SegInfo(const object::MachOObjectFile *Obj);
5200 StringRef segmentName(uint32_t SegIndex);
5201 StringRef sectionName(uint32_t SegIndex, uint64_t SegOffset);
5202 uint64_t address(uint32_t SegIndex, uint64_t SegOffset);
5205 struct SectionInfo {
5208 StringRef SectionName;
5209 StringRef SegmentName;
5210 uint64_t OffsetInSegment;
5211 uint64_t SegmentStartAddress;
5212 uint32_t SegmentIndex;
5214 const SectionInfo &findSection(uint32_t SegIndex, uint64_t SegOffset);
5215 SmallVector<SectionInfo, 32> Sections;
5219 SegInfo::SegInfo(const object::MachOObjectFile *Obj) {
5220 // Build table of sections so segIndex/offset pairs can be translated.
5221 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0;
5222 StringRef CurSegName;
5223 uint64_t CurSegAddress;
5224 for (const SectionRef &Section : Obj->sections()) {
5226 if (error(Section.getName(Info.SectionName)))
5228 Info.Address = Section.getAddress();
5229 Info.Size = Section.getSize();
5231 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl());
5232 if (!Info.SegmentName.equals(CurSegName)) {
5234 CurSegName = Info.SegmentName;
5235 CurSegAddress = Info.Address;
5237 Info.SegmentIndex = CurSegIndex - 1;
5238 Info.OffsetInSegment = Info.Address - CurSegAddress;
5239 Info.SegmentStartAddress = CurSegAddress;
5240 Sections.push_back(Info);
5244 StringRef SegInfo::segmentName(uint32_t SegIndex) {
5245 for (const SectionInfo &SI : Sections) {
5246 if (SI.SegmentIndex == SegIndex)
5247 return SI.SegmentName;
5249 llvm_unreachable("invalid segIndex");
5252 const SegInfo::SectionInfo &SegInfo::findSection(uint32_t SegIndex,
5253 uint64_t OffsetInSeg) {
5254 for (const SectionInfo &SI : Sections) {
5255 if (SI.SegmentIndex != SegIndex)
5257 if (SI.OffsetInSegment > OffsetInSeg)
5259 if (OffsetInSeg >= (SI.OffsetInSegment + SI.Size))
5263 llvm_unreachable("segIndex and offset not in any section");
5266 StringRef SegInfo::sectionName(uint32_t SegIndex, uint64_t OffsetInSeg) {
5267 return findSection(SegIndex, OffsetInSeg).SectionName;
5270 uint64_t SegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) {
5271 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg);
5272 return SI.SegmentStartAddress + OffsetInSeg;
5275 void llvm::printMachORebaseTable(const object::MachOObjectFile *Obj) {
5276 // Build table of sections so names can used in final output.
5277 SegInfo sectionTable(Obj);
5279 outs() << "segment section address type\n";
5280 for (const llvm::object::MachORebaseEntry &Entry : Obj->rebaseTable()) {
5281 uint32_t SegIndex = Entry.segmentIndex();
5282 uint64_t OffsetInSeg = Entry.segmentOffset();
5283 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5284 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5285 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5287 // Table lines look like: __DATA __nl_symbol_ptr 0x0000F00C pointer
5288 outs() << format("%-8s %-18s 0x%08" PRIX64 " %s\n",
5289 SegmentName.str().c_str(), SectionName.str().c_str(),
5290 Address, Entry.typeName().str().c_str());
5294 static StringRef ordinalName(const object::MachOObjectFile *Obj, int Ordinal) {
5295 StringRef DylibName;
5297 case MachO::BIND_SPECIAL_DYLIB_SELF:
5298 return "this-image";
5299 case MachO::BIND_SPECIAL_DYLIB_MAIN_EXECUTABLE:
5300 return "main-executable";
5301 case MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP:
5302 return "flat-namespace";
5305 std::error_code EC =
5306 Obj->getLibraryShortNameByIndex(Ordinal - 1, DylibName);
5308 return "<<bad library ordinal>>";
5312 return "<<unknown special ordinal>>";
5315 //===----------------------------------------------------------------------===//
5316 // bind table dumping
5317 //===----------------------------------------------------------------------===//
5319 void llvm::printMachOBindTable(const object::MachOObjectFile *Obj) {
5320 // Build table of sections so names can used in final output.
5321 SegInfo sectionTable(Obj);
5323 outs() << "segment section address type "
5324 "addend dylib symbol\n";
5325 for (const llvm::object::MachOBindEntry &Entry : Obj->bindTable()) {
5326 uint32_t SegIndex = Entry.segmentIndex();
5327 uint64_t OffsetInSeg = Entry.segmentOffset();
5328 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5329 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5330 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5332 // Table lines look like:
5333 // __DATA __got 0x00012010 pointer 0 libSystem ___stack_chk_guard
5335 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_WEAK_IMPORT)
5336 Attr = " (weak_import)";
5337 outs() << left_justify(SegmentName, 8) << " "
5338 << left_justify(SectionName, 18) << " "
5339 << format_hex(Address, 10, true) << " "
5340 << left_justify(Entry.typeName(), 8) << " "
5341 << format_decimal(Entry.addend(), 8) << " "
5342 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
5343 << Entry.symbolName() << Attr << "\n";
5347 //===----------------------------------------------------------------------===//
5348 // lazy bind table dumping
5349 //===----------------------------------------------------------------------===//
5351 void llvm::printMachOLazyBindTable(const object::MachOObjectFile *Obj) {
5352 // Build table of sections so names can used in final output.
5353 SegInfo sectionTable(Obj);
5355 outs() << "segment section address "
5357 for (const llvm::object::MachOBindEntry &Entry : Obj->lazyBindTable()) {
5358 uint32_t SegIndex = Entry.segmentIndex();
5359 uint64_t OffsetInSeg = Entry.segmentOffset();
5360 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5361 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5362 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5364 // Table lines look like:
5365 // __DATA __got 0x00012010 libSystem ___stack_chk_guard
5366 outs() << left_justify(SegmentName, 8) << " "
5367 << left_justify(SectionName, 18) << " "
5368 << format_hex(Address, 10, true) << " "
5369 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
5370 << Entry.symbolName() << "\n";
5374 //===----------------------------------------------------------------------===//
5375 // weak bind table dumping
5376 //===----------------------------------------------------------------------===//
5378 void llvm::printMachOWeakBindTable(const object::MachOObjectFile *Obj) {
5379 // Build table of sections so names can used in final output.
5380 SegInfo sectionTable(Obj);
5382 outs() << "segment section address "
5383 "type addend symbol\n";
5384 for (const llvm::object::MachOBindEntry &Entry : Obj->weakBindTable()) {
5385 // Strong symbols don't have a location to update.
5386 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) {
5387 outs() << " strong "
5388 << Entry.symbolName() << "\n";
5391 uint32_t SegIndex = Entry.segmentIndex();
5392 uint64_t OffsetInSeg = Entry.segmentOffset();
5393 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5394 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5395 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5397 // Table lines look like:
5398 // __DATA __data 0x00001000 pointer 0 _foo
5399 outs() << left_justify(SegmentName, 8) << " "
5400 << left_justify(SectionName, 18) << " "
5401 << format_hex(Address, 10, true) << " "
5402 << left_justify(Entry.typeName(), 8) << " "
5403 << format_decimal(Entry.addend(), 8) << " " << Entry.symbolName()
5408 // get_dyld_bind_info_symbolname() is used for disassembly and passed an
5409 // address, ReferenceValue, in the Mach-O file and looks in the dyld bind
5410 // information for that address. If the address is found its binding symbol
5411 // name is returned. If not nullptr is returned.
5412 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
5413 struct DisassembleInfo *info) {
5414 if (info->bindtable == nullptr) {
5415 info->bindtable = new (BindTable);
5416 SegInfo sectionTable(info->O);
5417 for (const llvm::object::MachOBindEntry &Entry : info->O->bindTable()) {
5418 uint32_t SegIndex = Entry.segmentIndex();
5419 uint64_t OffsetInSeg = Entry.segmentOffset();
5420 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5421 const char *SymbolName = nullptr;
5422 StringRef name = Entry.symbolName();
5424 SymbolName = name.data();
5425 info->bindtable->push_back(std::make_pair(Address, SymbolName));
5428 for (bind_table_iterator BI = info->bindtable->begin(),
5429 BE = info->bindtable->end();
5431 uint64_t Address = BI->first;
5432 if (ReferenceValue == Address) {
5433 const char *SymbolName = BI->second;