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/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/Support/Casting.h"
32 #include "llvm/Support/CommandLine.h"
33 #include "llvm/Support/Debug.h"
34 #include "llvm/Support/Endian.h"
35 #include "llvm/Support/Format.h"
36 #include "llvm/Support/GraphWriter.h"
37 #include "llvm/Support/MachO.h"
38 #include "llvm/Support/MemoryBuffer.h"
39 #include "llvm/Support/FormattedStream.h"
40 #include "llvm/Support/TargetRegistry.h"
41 #include "llvm/Support/TargetSelect.h"
42 #include "llvm/Support/raw_ostream.h"
45 #include <system_error>
52 using namespace object;
56 cl::desc("Print line information from debug info if available"));
58 static cl::opt<std::string> DSYMFile("dsym",
59 cl::desc("Use .dSYM file for debug info"));
61 static cl::opt<bool> FullLeadingAddr("full-leading-addr",
62 cl::desc("Print full leading address"));
65 PrintImmHex("print-imm-hex",
66 cl::desc("Use hex format for immediate values"));
68 static std::string ThumbTripleName;
70 static const Target *GetTarget(const MachOObjectFile *MachOObj,
71 const char **McpuDefault,
72 const Target **ThumbTarget) {
73 // Figure out the target triple.
74 if (TripleName.empty()) {
75 llvm::Triple TT("unknown-unknown-unknown");
76 llvm::Triple ThumbTriple = Triple();
77 TT = MachOObj->getArch(McpuDefault, &ThumbTriple);
78 TripleName = TT.str();
79 ThumbTripleName = ThumbTriple.str();
82 // Get the target specific parser.
84 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
85 if (TheTarget && ThumbTripleName.empty())
88 *ThumbTarget = TargetRegistry::lookupTarget(ThumbTripleName, Error);
92 errs() << "llvm-objdump: error: unable to get target for '";
96 errs() << ThumbTripleName;
97 errs() << "', see --version and --triple.\n";
101 struct SymbolSorter {
102 bool operator()(const SymbolRef &A, const SymbolRef &B) {
103 SymbolRef::Type AType, BType;
107 uint64_t AAddr, BAddr;
108 if (AType != SymbolRef::ST_Function)
112 if (BType != SymbolRef::ST_Function)
116 return AAddr < BAddr;
120 // Types for the storted data in code table that is built before disassembly
121 // and the predicate function to sort them.
122 typedef std::pair<uint64_t, DiceRef> DiceTableEntry;
123 typedef std::vector<DiceTableEntry> DiceTable;
124 typedef DiceTable::iterator dice_table_iterator;
126 // This is used to search for a data in code table entry for the PC being
127 // disassembled. The j parameter has the PC in j.first. A single data in code
128 // table entry can cover many bytes for each of its Kind's. So if the offset,
129 // aka the i.first value, of the data in code table entry plus its Length
130 // covers the PC being searched for this will return true. If not it will
132 static bool compareDiceTableEntries(const DiceTableEntry &i,
133 const DiceTableEntry &j) {
135 i.second.getLength(Length);
137 return j.first >= i.first && j.first < i.first + Length;
140 static uint64_t DumpDataInCode(const char *bytes, uint64_t Length,
141 unsigned short Kind) {
142 uint32_t Value, Size = 1;
146 case MachO::DICE_KIND_DATA:
149 DumpBytes(StringRef(bytes, 4));
150 Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0];
151 outs() << "\t.long " << Value;
153 } else if (Length >= 2) {
155 DumpBytes(StringRef(bytes, 2));
156 Value = bytes[1] << 8 | bytes[0];
157 outs() << "\t.short " << Value;
161 DumpBytes(StringRef(bytes, 2));
163 outs() << "\t.byte " << Value;
166 if (Kind == MachO::DICE_KIND_DATA)
167 outs() << "\t@ KIND_DATA\n";
169 outs() << "\t@ data in code kind = " << Kind << "\n";
171 case MachO::DICE_KIND_JUMP_TABLE8:
173 DumpBytes(StringRef(bytes, 1));
175 outs() << "\t.byte " << format("%3u", Value) << "\t@ KIND_JUMP_TABLE8\n";
178 case MachO::DICE_KIND_JUMP_TABLE16:
180 DumpBytes(StringRef(bytes, 2));
181 Value = bytes[1] << 8 | bytes[0];
182 outs() << "\t.short " << format("%5u", Value & 0xffff)
183 << "\t@ KIND_JUMP_TABLE16\n";
186 case MachO::DICE_KIND_JUMP_TABLE32:
187 case MachO::DICE_KIND_ABS_JUMP_TABLE32:
189 DumpBytes(StringRef(bytes, 4));
190 Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0];
191 outs() << "\t.long " << Value;
192 if (Kind == MachO::DICE_KIND_JUMP_TABLE32)
193 outs() << "\t@ KIND_JUMP_TABLE32\n";
195 outs() << "\t@ KIND_ABS_JUMP_TABLE32\n";
202 static void getSectionsAndSymbols(const MachO::mach_header Header,
203 MachOObjectFile *MachOObj,
204 std::vector<SectionRef> &Sections,
205 std::vector<SymbolRef> &Symbols,
206 SmallVectorImpl<uint64_t> &FoundFns,
207 uint64_t &BaseSegmentAddress) {
208 for (const SymbolRef &Symbol : MachOObj->symbols())
209 Symbols.push_back(Symbol);
211 for (const SectionRef &Section : MachOObj->sections()) {
213 Section.getName(SectName);
214 Sections.push_back(Section);
217 MachOObjectFile::LoadCommandInfo Command =
218 MachOObj->getFirstLoadCommandInfo();
219 bool BaseSegmentAddressSet = false;
220 for (unsigned i = 0;; ++i) {
221 if (Command.C.cmd == MachO::LC_FUNCTION_STARTS) {
222 // We found a function starts segment, parse the addresses for later
224 MachO::linkedit_data_command LLC =
225 MachOObj->getLinkeditDataLoadCommand(Command);
227 MachOObj->ReadULEB128s(LLC.dataoff, FoundFns);
228 } else if (Command.C.cmd == MachO::LC_SEGMENT) {
229 MachO::segment_command SLC = MachOObj->getSegmentLoadCommand(Command);
230 StringRef SegName = SLC.segname;
231 if (!BaseSegmentAddressSet && SegName != "__PAGEZERO") {
232 BaseSegmentAddressSet = true;
233 BaseSegmentAddress = SLC.vmaddr;
237 if (i == Header.ncmds - 1)
240 Command = MachOObj->getNextLoadCommandInfo(Command);
244 static void DisassembleInputMachO2(StringRef Filename,
245 MachOObjectFile *MachOOF);
247 void llvm::DisassembleInputMachO(StringRef Filename) {
248 ErrorOr<std::unique_ptr<MemoryBuffer>> BuffOrErr =
249 MemoryBuffer::getFileOrSTDIN(Filename);
250 if (std::error_code EC = BuffOrErr.getError()) {
251 errs() << "llvm-objdump: " << Filename << ": " << EC.message() << "\n";
254 std::unique_ptr<MemoryBuffer> Buff = std::move(BuffOrErr.get());
256 std::unique_ptr<MachOObjectFile> MachOOF = std::move(
257 ObjectFile::createMachOObjectFile(Buff.get()->getMemBufferRef()).get());
259 DisassembleInputMachO2(Filename, MachOOF.get());
262 typedef DenseMap<uint64_t, StringRef> SymbolAddressMap;
263 typedef std::pair<uint64_t, const char *> BindInfoEntry;
264 typedef std::vector<BindInfoEntry> BindTable;
265 typedef BindTable::iterator bind_table_iterator;
267 // The block of info used by the Symbolizer call backs.
268 struct DisassembleInfo {
272 SymbolAddressMap *AddrMap;
273 std::vector<SectionRef> *Sections;
274 const char *class_name;
275 const char *selector_name;
277 char *demangled_name;
280 BindTable *bindtable;
283 // GuessSymbolName is passed the address of what might be a symbol and a
284 // pointer to the DisassembleInfo struct. It returns the name of a symbol
285 // with that address or nullptr if no symbol is found with that address.
286 static const char *GuessSymbolName(uint64_t value,
287 struct DisassembleInfo *info) {
288 const char *SymbolName = nullptr;
289 // A DenseMap can't lookup up some values.
290 if (value != 0xffffffffffffffffULL && value != 0xfffffffffffffffeULL) {
291 StringRef name = info->AddrMap->lookup(value);
293 SymbolName = name.data();
298 // SymbolizerGetOpInfo() is the operand information call back function.
299 // This is called to get the symbolic information for operand(s) of an
300 // instruction when it is being done. This routine does this from
301 // the relocation information, symbol table, etc. That block of information
302 // is a pointer to the struct DisassembleInfo that was passed when the
303 // disassembler context was created and passed to back to here when
304 // called back by the disassembler for instruction operands that could have
305 // relocation information. The address of the instruction containing operand is
306 // at the Pc parameter. The immediate value the operand has is passed in
307 // op_info->Value and is at Offset past the start of the instruction and has a
308 // byte Size of 1, 2 or 4. The symbolc information is returned in TagBuf is the
309 // LLVMOpInfo1 struct defined in the header "llvm-c/Disassembler.h" as symbol
310 // names and addends of the symbolic expression to add for the operand. The
311 // value of TagType is currently 1 (for the LLVMOpInfo1 struct). If symbolic
312 // information is returned then this function returns 1 else it returns 0.
313 int SymbolizerGetOpInfo(void *DisInfo, uint64_t Pc, uint64_t Offset,
314 uint64_t Size, int TagType, void *TagBuf) {
315 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
316 struct LLVMOpInfo1 *op_info = (struct LLVMOpInfo1 *)TagBuf;
317 uint64_t value = op_info->Value;
319 // Make sure all fields returned are zero if we don't set them.
320 memset((void *)op_info, '\0', sizeof(struct LLVMOpInfo1));
321 op_info->Value = value;
323 // If the TagType is not the value 1 which it code knows about or if no
324 // verbose symbolic information is wanted then just return 0, indicating no
325 // information is being returned.
326 if (TagType != 1 || info->verbose == false)
329 unsigned int Arch = info->O->getArch();
330 if (Arch == Triple::x86) {
331 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
333 // First search the section's relocation entries (if any) for an entry
334 // for this section offset.
335 uint32_t sect_addr = info->S.getAddress();
336 uint32_t sect_offset = (Pc + Offset) - sect_addr;
337 bool reloc_found = false;
339 MachO::any_relocation_info RE;
340 bool isExtern = false;
342 bool r_scattered = false;
343 uint32_t r_value, pair_r_value, r_type;
344 for (const RelocationRef &Reloc : info->S.relocations()) {
345 uint64_t RelocOffset;
346 Reloc.getOffset(RelocOffset);
347 if (RelocOffset == sect_offset) {
348 Rel = Reloc.getRawDataRefImpl();
349 RE = info->O->getRelocation(Rel);
350 r_type = info->O->getAnyRelocationType(RE);
351 r_scattered = info->O->isRelocationScattered(RE);
353 r_value = info->O->getScatteredRelocationValue(RE);
354 if (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
355 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF) {
356 DataRefImpl RelNext = Rel;
357 info->O->moveRelocationNext(RelNext);
358 MachO::any_relocation_info RENext;
359 RENext = info->O->getRelocation(RelNext);
360 if (info->O->isRelocationScattered(RENext))
361 pair_r_value = info->O->getScatteredRelocationValue(RENext);
366 isExtern = info->O->getPlainRelocationExternal(RE);
368 symbol_iterator RelocSym = Reloc.getSymbol();
376 if (reloc_found && isExtern) {
378 Symbol.getName(SymName);
379 const char *name = SymName.data();
380 op_info->AddSymbol.Present = 1;
381 op_info->AddSymbol.Name = name;
382 // For i386 extern relocation entries the value in the instruction is
383 // the offset from the symbol, and value is already set in op_info->Value.
386 if (reloc_found && (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
387 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF)) {
388 const char *add = GuessSymbolName(r_value, info);
389 const char *sub = GuessSymbolName(pair_r_value, info);
390 uint32_t offset = value - (r_value - pair_r_value);
391 op_info->AddSymbol.Present = 1;
393 op_info->AddSymbol.Name = add;
395 op_info->AddSymbol.Value = r_value;
396 op_info->SubtractSymbol.Present = 1;
398 op_info->SubtractSymbol.Name = sub;
400 op_info->SubtractSymbol.Value = pair_r_value;
401 op_info->Value = offset;
405 // Second search the external relocation entries of a fully linked image
406 // (if any) for an entry that matches this segment offset.
407 // uint32_t seg_offset = (Pc + Offset);
409 } else if (Arch == Triple::x86_64) {
410 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
412 // First search the section's relocation entries (if any) for an entry
413 // for this section offset.
414 uint64_t sect_addr = info->S.getAddress();
415 uint64_t sect_offset = (Pc + Offset) - sect_addr;
416 bool reloc_found = false;
418 MachO::any_relocation_info RE;
419 bool isExtern = false;
421 for (const RelocationRef &Reloc : info->S.relocations()) {
422 uint64_t RelocOffset;
423 Reloc.getOffset(RelocOffset);
424 if (RelocOffset == sect_offset) {
425 Rel = Reloc.getRawDataRefImpl();
426 RE = info->O->getRelocation(Rel);
427 // NOTE: Scattered relocations don't exist on x86_64.
428 isExtern = info->O->getPlainRelocationExternal(RE);
430 symbol_iterator RelocSym = Reloc.getSymbol();
437 if (reloc_found && isExtern) {
438 // The Value passed in will be adjusted by the Pc if the instruction
439 // adds the Pc. But for x86_64 external relocation entries the Value
440 // is the offset from the external symbol.
441 if (info->O->getAnyRelocationPCRel(RE))
442 op_info->Value -= Pc + Offset + Size;
444 Symbol.getName(SymName);
445 const char *name = SymName.data();
446 unsigned Type = info->O->getAnyRelocationType(RE);
447 if (Type == MachO::X86_64_RELOC_SUBTRACTOR) {
448 DataRefImpl RelNext = Rel;
449 info->O->moveRelocationNext(RelNext);
450 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
451 unsigned TypeNext = info->O->getAnyRelocationType(RENext);
452 bool isExternNext = info->O->getPlainRelocationExternal(RENext);
453 unsigned SymbolNum = info->O->getPlainRelocationSymbolNum(RENext);
454 if (TypeNext == MachO::X86_64_RELOC_UNSIGNED && isExternNext) {
455 op_info->SubtractSymbol.Present = 1;
456 op_info->SubtractSymbol.Name = name;
457 symbol_iterator RelocSymNext = info->O->getSymbolByIndex(SymbolNum);
458 Symbol = *RelocSymNext;
459 StringRef SymNameNext;
460 Symbol.getName(SymNameNext);
461 name = SymNameNext.data();
464 // TODO: add the VariantKinds to op_info->VariantKind for relocation types
465 // like: X86_64_RELOC_TLV, X86_64_RELOC_GOT_LOAD and X86_64_RELOC_GOT.
466 op_info->AddSymbol.Present = 1;
467 op_info->AddSymbol.Name = name;
471 // Second search the external relocation entries of a fully linked image
472 // (if any) for an entry that matches this segment offset.
473 // uint64_t seg_offset = (Pc + Offset);
475 } else if (Arch == Triple::arm) {
476 if (Offset != 0 || (Size != 4 && Size != 2))
478 // First search the section's relocation entries (if any) for an entry
479 // for this section offset.
480 uint32_t sect_addr = info->S.getAddress();
481 uint32_t sect_offset = (Pc + Offset) - sect_addr;
482 bool reloc_found = false;
484 MachO::any_relocation_info RE;
485 bool isExtern = false;
487 bool r_scattered = false;
488 uint32_t r_value, pair_r_value, r_type, r_length, other_half;
489 for (const RelocationRef &Reloc : info->S.relocations()) {
490 uint64_t RelocOffset;
491 Reloc.getOffset(RelocOffset);
492 if (RelocOffset == sect_offset) {
493 Rel = Reloc.getRawDataRefImpl();
494 RE = info->O->getRelocation(Rel);
495 r_length = info->O->getAnyRelocationLength(RE);
496 r_scattered = info->O->isRelocationScattered(RE);
498 r_value = info->O->getScatteredRelocationValue(RE);
499 r_type = info->O->getScatteredRelocationType(RE);
501 r_type = info->O->getAnyRelocationType(RE);
502 isExtern = info->O->getPlainRelocationExternal(RE);
504 symbol_iterator RelocSym = Reloc.getSymbol();
508 if (r_type == MachO::ARM_RELOC_HALF ||
509 r_type == MachO::ARM_RELOC_SECTDIFF ||
510 r_type == MachO::ARM_RELOC_LOCAL_SECTDIFF ||
511 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
512 DataRefImpl RelNext = Rel;
513 info->O->moveRelocationNext(RelNext);
514 MachO::any_relocation_info RENext;
515 RENext = info->O->getRelocation(RelNext);
516 other_half = info->O->getAnyRelocationAddress(RENext) & 0xffff;
517 if (info->O->isRelocationScattered(RENext))
518 pair_r_value = info->O->getScatteredRelocationValue(RENext);
524 if (reloc_found && isExtern) {
526 Symbol.getName(SymName);
527 const char *name = SymName.data();
528 op_info->AddSymbol.Present = 1;
529 op_info->AddSymbol.Name = name;
532 case MachO::ARM_RELOC_HALF:
533 if ((r_length & 0x1) == 1) {
534 op_info->Value = value << 16 | other_half;
535 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
537 op_info->Value = other_half << 16 | value;
538 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
546 case MachO::ARM_RELOC_HALF:
547 if ((r_length & 0x1) == 1) {
548 op_info->Value = value << 16 | other_half;
549 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
551 op_info->Value = other_half << 16 | value;
552 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
561 // If we have a branch that is not an external relocation entry then
562 // return 0 so the code in tryAddingSymbolicOperand() can use the
563 // SymbolLookUp call back with the branch target address to look up the
564 // symbol and possiblity add an annotation for a symbol stub.
565 if (reloc_found && isExtern == 0 && (r_type == MachO::ARM_RELOC_BR24 ||
566 r_type == MachO::ARM_THUMB_RELOC_BR22))
571 if (r_type == MachO::ARM_RELOC_HALF ||
572 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
573 if ((r_length & 0x1) == 1)
574 value = value << 16 | other_half;
576 value = other_half << 16 | value;
578 if (r_scattered && (r_type != MachO::ARM_RELOC_HALF &&
579 r_type != MachO::ARM_RELOC_HALF_SECTDIFF)) {
580 offset = value - r_value;
585 if (reloc_found && r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
586 if ((r_length & 0x1) == 1)
587 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
589 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
590 const char *add = GuessSymbolName(r_value, info);
591 const char *sub = GuessSymbolName(pair_r_value, info);
592 int32_t offset = value - (r_value - pair_r_value);
593 op_info->AddSymbol.Present = 1;
595 op_info->AddSymbol.Name = add;
597 op_info->AddSymbol.Value = r_value;
598 op_info->SubtractSymbol.Present = 1;
600 op_info->SubtractSymbol.Name = sub;
602 op_info->SubtractSymbol.Value = pair_r_value;
603 op_info->Value = offset;
607 if (reloc_found == false)
610 op_info->AddSymbol.Present = 1;
611 op_info->Value = offset;
613 if (r_type == MachO::ARM_RELOC_HALF) {
614 if ((r_length & 0x1) == 1)
615 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
617 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
620 const char *add = GuessSymbolName(value, info);
621 if (add != nullptr) {
622 op_info->AddSymbol.Name = add;
625 op_info->AddSymbol.Value = value;
627 } else if (Arch == Triple::aarch64) {
628 if (Offset != 0 || Size != 4)
630 // First search the section's relocation entries (if any) for an entry
631 // for this section offset.
632 uint64_t sect_addr = info->S.getAddress();
633 uint64_t sect_offset = (Pc + Offset) - sect_addr;
634 bool reloc_found = false;
636 MachO::any_relocation_info RE;
637 bool isExtern = false;
640 for (const RelocationRef &Reloc : info->S.relocations()) {
641 uint64_t RelocOffset;
642 Reloc.getOffset(RelocOffset);
643 if (RelocOffset == sect_offset) {
644 Rel = Reloc.getRawDataRefImpl();
645 RE = info->O->getRelocation(Rel);
646 r_type = info->O->getAnyRelocationType(RE);
647 if (r_type == MachO::ARM64_RELOC_ADDEND) {
648 DataRefImpl RelNext = Rel;
649 info->O->moveRelocationNext(RelNext);
650 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
652 value = info->O->getPlainRelocationSymbolNum(RENext);
653 op_info->Value = value;
656 // NOTE: Scattered relocations don't exist on arm64.
657 isExtern = info->O->getPlainRelocationExternal(RE);
659 symbol_iterator RelocSym = Reloc.getSymbol();
666 if (reloc_found && isExtern) {
668 Symbol.getName(SymName);
669 const char *name = SymName.data();
670 op_info->AddSymbol.Present = 1;
671 op_info->AddSymbol.Name = name;
674 case MachO::ARM64_RELOC_PAGE21:
676 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGE;
678 case MachO::ARM64_RELOC_PAGEOFF12:
680 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGEOFF;
682 case MachO::ARM64_RELOC_GOT_LOAD_PAGE21:
684 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGE;
686 case MachO::ARM64_RELOC_GOT_LOAD_PAGEOFF12:
688 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGEOFF;
690 case MachO::ARM64_RELOC_TLVP_LOAD_PAGE21:
691 /* @tvlppage is not implemented in llvm-mc */
692 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVP;
694 case MachO::ARM64_RELOC_TLVP_LOAD_PAGEOFF12:
695 /* @tvlppageoff is not implemented in llvm-mc */
696 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVOFF;
699 case MachO::ARM64_RELOC_BRANCH26:
700 op_info->VariantKind = LLVMDisassembler_VariantKind_None;
711 // GuessCstringPointer is passed the address of what might be a pointer to a
712 // literal string in a cstring section. If that address is in a cstring section
713 // it returns a pointer to that string. Else it returns nullptr.
714 const char *GuessCstringPointer(uint64_t ReferenceValue,
715 struct DisassembleInfo *info) {
716 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
717 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
718 for (unsigned I = 0;; ++I) {
719 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
720 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
721 for (unsigned J = 0; J < Seg.nsects; ++J) {
722 MachO::section_64 Sec = info->O->getSection64(Load, J);
723 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
724 if (section_type == MachO::S_CSTRING_LITERALS &&
725 ReferenceValue >= Sec.addr &&
726 ReferenceValue < Sec.addr + Sec.size) {
727 uint64_t sect_offset = ReferenceValue - Sec.addr;
728 uint64_t object_offset = Sec.offset + sect_offset;
729 StringRef MachOContents = info->O->getData();
730 uint64_t object_size = MachOContents.size();
731 const char *object_addr = (const char *)MachOContents.data();
732 if (object_offset < object_size) {
733 const char *name = object_addr + object_offset;
740 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
741 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
742 for (unsigned J = 0; J < Seg.nsects; ++J) {
743 MachO::section Sec = info->O->getSection(Load, J);
744 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
745 if (section_type == MachO::S_CSTRING_LITERALS &&
746 ReferenceValue >= Sec.addr &&
747 ReferenceValue < Sec.addr + Sec.size) {
748 uint64_t sect_offset = ReferenceValue - Sec.addr;
749 uint64_t object_offset = Sec.offset + sect_offset;
750 StringRef MachOContents = info->O->getData();
751 uint64_t object_size = MachOContents.size();
752 const char *object_addr = (const char *)MachOContents.data();
753 if (object_offset < object_size) {
754 const char *name = object_addr + object_offset;
762 if (I == LoadCommandCount - 1)
765 Load = info->O->getNextLoadCommandInfo(Load);
770 // GuessIndirectSymbol returns the name of the indirect symbol for the
771 // ReferenceValue passed in or nullptr. This is used when ReferenceValue maybe
772 // an address of a symbol stub or a lazy or non-lazy pointer to associate the
773 // symbol name being referenced by the stub or pointer.
774 static const char *GuessIndirectSymbol(uint64_t ReferenceValue,
775 struct DisassembleInfo *info) {
776 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
777 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
778 MachO::dysymtab_command Dysymtab = info->O->getDysymtabLoadCommand();
779 MachO::symtab_command Symtab = info->O->getSymtabLoadCommand();
780 for (unsigned I = 0;; ++I) {
781 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
782 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
783 for (unsigned J = 0; J < Seg.nsects; ++J) {
784 MachO::section_64 Sec = info->O->getSection64(Load, J);
785 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
786 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
787 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
788 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
789 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
790 section_type == MachO::S_SYMBOL_STUBS) &&
791 ReferenceValue >= Sec.addr &&
792 ReferenceValue < Sec.addr + Sec.size) {
794 if (section_type == MachO::S_SYMBOL_STUBS)
795 stride = Sec.reserved2;
800 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
801 if (index < Dysymtab.nindirectsyms) {
802 uint32_t indirect_symbol =
803 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
804 if (indirect_symbol < Symtab.nsyms) {
805 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
806 SymbolRef Symbol = *Sym;
808 Symbol.getName(SymName);
809 const char *name = SymName.data();
815 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
816 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
817 for (unsigned J = 0; J < Seg.nsects; ++J) {
818 MachO::section Sec = info->O->getSection(Load, J);
819 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
820 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
821 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
822 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
823 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
824 section_type == MachO::S_SYMBOL_STUBS) &&
825 ReferenceValue >= Sec.addr &&
826 ReferenceValue < Sec.addr + Sec.size) {
828 if (section_type == MachO::S_SYMBOL_STUBS)
829 stride = Sec.reserved2;
834 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
835 if (index < Dysymtab.nindirectsyms) {
836 uint32_t indirect_symbol =
837 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
838 if (indirect_symbol < Symtab.nsyms) {
839 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
840 SymbolRef Symbol = *Sym;
842 Symbol.getName(SymName);
843 const char *name = SymName.data();
850 if (I == LoadCommandCount - 1)
853 Load = info->O->getNextLoadCommandInfo(Load);
858 // method_reference() is called passing it the ReferenceName that might be
859 // a reference it to an Objective-C method call. If so then it allocates and
860 // assembles a method call string with the values last seen and saved in
861 // the DisassembleInfo's class_name and selector_name fields. This is saved
862 // into the method field of the info and any previous string is free'ed.
863 // Then the class_name field in the info is set to nullptr. The method call
864 // string is set into ReferenceName and ReferenceType is set to
865 // LLVMDisassembler_ReferenceType_Out_Objc_Message. If this not a method call
866 // then both ReferenceType and ReferenceName are left unchanged.
867 static void method_reference(struct DisassembleInfo *info,
868 uint64_t *ReferenceType,
869 const char **ReferenceName) {
870 unsigned int Arch = info->O->getArch();
871 if (*ReferenceName != nullptr) {
872 if (strcmp(*ReferenceName, "_objc_msgSend") == 0) {
873 if (info->selector_name != nullptr) {
874 if (info->method != nullptr)
876 if (info->class_name != nullptr) {
877 info->method = (char *)malloc(5 + strlen(info->class_name) +
878 strlen(info->selector_name));
879 if (info->method != nullptr) {
880 strcpy(info->method, "+[");
881 strcat(info->method, info->class_name);
882 strcat(info->method, " ");
883 strcat(info->method, info->selector_name);
884 strcat(info->method, "]");
885 *ReferenceName = info->method;
886 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
889 info->method = (char *)malloc(9 + strlen(info->selector_name));
890 if (info->method != nullptr) {
891 if (Arch == Triple::x86_64)
892 strcpy(info->method, "-[%rdi ");
893 else if (Arch == Triple::aarch64)
894 strcpy(info->method, "-[x0 ");
896 strcpy(info->method, "-[r? ");
897 strcat(info->method, info->selector_name);
898 strcat(info->method, "]");
899 *ReferenceName = info->method;
900 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
903 info->class_name = nullptr;
905 } else if (strcmp(*ReferenceName, "_objc_msgSendSuper2") == 0) {
906 if (info->selector_name != nullptr) {
907 if (info->method != nullptr)
909 info->method = (char *)malloc(17 + strlen(info->selector_name));
910 if (info->method != nullptr) {
911 if (Arch == Triple::x86_64)
912 strcpy(info->method, "-[[%rdi super] ");
913 else if (Arch == Triple::aarch64)
914 strcpy(info->method, "-[[x0 super] ");
916 strcpy(info->method, "-[[r? super] ");
917 strcat(info->method, info->selector_name);
918 strcat(info->method, "]");
919 *ReferenceName = info->method;
920 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
922 info->class_name = nullptr;
928 // GuessPointerPointer() is passed the address of what might be a pointer to
929 // a reference to an Objective-C class, selector, message ref or cfstring.
930 // If so the value of the pointer is returned and one of the booleans are set
931 // to true. If not zero is returned and all the booleans are set to false.
932 static uint64_t GuessPointerPointer(uint64_t ReferenceValue,
933 struct DisassembleInfo *info,
934 bool &classref, bool &selref, bool &msgref,
940 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
941 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
942 for (unsigned I = 0;; ++I) {
943 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
944 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
945 for (unsigned J = 0; J < Seg.nsects; ++J) {
946 MachO::section_64 Sec = info->O->getSection64(Load, J);
947 if ((strncmp(Sec.sectname, "__objc_selrefs", 16) == 0 ||
948 strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
949 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0 ||
950 strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 ||
951 strncmp(Sec.sectname, "__cfstring", 16) == 0) &&
952 ReferenceValue >= Sec.addr &&
953 ReferenceValue < Sec.addr + Sec.size) {
954 uint64_t sect_offset = ReferenceValue - Sec.addr;
955 uint64_t object_offset = Sec.offset + sect_offset;
956 StringRef MachOContents = info->O->getData();
957 uint64_t object_size = MachOContents.size();
958 const char *object_addr = (const char *)MachOContents.data();
959 if (object_offset < object_size) {
960 uint64_t pointer_value;
961 memcpy(&pointer_value, object_addr + object_offset,
963 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
964 sys::swapByteOrder(pointer_value);
965 if (strncmp(Sec.sectname, "__objc_selrefs", 16) == 0)
967 else if (strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
968 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0)
970 else if (strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 &&
971 ReferenceValue + 8 < Sec.addr + Sec.size) {
973 memcpy(&pointer_value, object_addr + object_offset + 8,
975 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
976 sys::swapByteOrder(pointer_value);
977 } else if (strncmp(Sec.sectname, "__cfstring", 16) == 0)
979 return pointer_value;
986 // TODO: Look for LC_SEGMENT for 32-bit Mach-O files.
987 if (I == LoadCommandCount - 1)
990 Load = info->O->getNextLoadCommandInfo(Load);
995 // get_pointer_64 returns a pointer to the bytes in the object file at the
996 // Address from a section in the Mach-O file. And indirectly returns the
997 // offset into the section, number of bytes left in the section past the offset
998 // and which section is was being referenced. If the Address is not in a
999 // section nullptr is returned.
1000 const char *get_pointer_64(uint64_t Address, uint32_t &offset, uint32_t &left,
1001 SectionRef &S, DisassembleInfo *info) {
1005 for (unsigned SectIdx = 0; SectIdx != info->Sections->size(); SectIdx++) {
1006 uint64_t SectAddress = ((*(info->Sections))[SectIdx]).getAddress();
1007 uint64_t SectSize = ((*(info->Sections))[SectIdx]).getSize();
1008 if (Address >= SectAddress && Address < SectAddress + SectSize) {
1009 S = (*(info->Sections))[SectIdx];
1010 offset = Address - SectAddress;
1011 left = SectSize - offset;
1012 StringRef SectContents;
1013 ((*(info->Sections))[SectIdx]).getContents(SectContents);
1014 return SectContents.data() + offset;
1020 // get_symbol_64() returns the name of a symbol (or nullptr) and the address of
1021 // the symbol indirectly through n_value. Based on the relocation information
1022 // for the specified section offset in the specified section reference.
1023 const char *get_symbol_64(uint32_t sect_offset, SectionRef S,
1024 DisassembleInfo *info, uint64_t &n_value) {
1026 if (info->verbose == false)
1029 // See if there is an external relocation entry at the sect_offset.
1030 bool reloc_found = false;
1032 MachO::any_relocation_info RE;
1033 bool isExtern = false;
1035 for (const RelocationRef &Reloc : S.relocations()) {
1036 uint64_t RelocOffset;
1037 Reloc.getOffset(RelocOffset);
1038 if (RelocOffset == sect_offset) {
1039 Rel = Reloc.getRawDataRefImpl();
1040 RE = info->O->getRelocation(Rel);
1041 if (info->O->isRelocationScattered(RE))
1043 isExtern = info->O->getPlainRelocationExternal(RE);
1045 symbol_iterator RelocSym = Reloc.getSymbol();
1052 // If there is an external relocation entry for a symbol in this section
1053 // at this section_offset then use that symbol's value for the n_value
1054 // and return its name.
1055 const char *SymbolName = nullptr;
1056 if (reloc_found && isExtern) {
1057 Symbol.getAddress(n_value);
1059 Symbol.getName(name);
1060 if (!name.empty()) {
1061 SymbolName = name.data();
1066 // TODO: For fully linked images, look through the external relocation
1067 // entries off the dynamic symtab command. For these the r_offset is from the
1068 // start of the first writeable segment in the Mach-O file. So the offset
1069 // to this section from that segment is passed to this routine by the caller,
1070 // as the database_offset. Which is the difference of the section's starting
1071 // address and the first writable segment.
1073 // NOTE: need add passing the database_offset to this routine.
1075 // TODO: We did not find an external relocation entry so look up the
1076 // ReferenceValue as an address of a symbol and if found return that symbol's
1079 // NOTE: need add passing the ReferenceValue to this routine. Then that code
1080 // would simply be this:
1081 // SymbolName = GuessSymbolName(ReferenceValue, info);
1086 // These are structs in the Objective-C meta data and read to produce the
1087 // comments for disassembly. While these are part of the ABI they are no
1088 // public defintions. So the are here not in include/llvm/Support/MachO.h .
1090 // The cfstring object in a 64-bit Mach-O file.
1091 struct cfstring64_t {
1092 uint64_t isa; // class64_t * (64-bit pointer)
1093 uint64_t flags; // flag bits
1094 uint64_t characters; // char * (64-bit pointer)
1095 uint64_t length; // number of non-NULL characters in above
1098 // The class object in a 64-bit Mach-O file.
1100 uint64_t isa; // class64_t * (64-bit pointer)
1101 uint64_t superclass; // class64_t * (64-bit pointer)
1102 uint64_t cache; // Cache (64-bit pointer)
1103 uint64_t vtable; // IMP * (64-bit pointer)
1104 uint64_t data; // class_ro64_t * (64-bit pointer)
1107 struct class_ro64_t {
1109 uint32_t instanceStart;
1110 uint32_t instanceSize;
1112 uint64_t ivarLayout; // const uint8_t * (64-bit pointer)
1113 uint64_t name; // const char * (64-bit pointer)
1114 uint64_t baseMethods; // const method_list_t * (64-bit pointer)
1115 uint64_t baseProtocols; // const protocol_list_t * (64-bit pointer)
1116 uint64_t ivars; // const ivar_list_t * (64-bit pointer)
1117 uint64_t weakIvarLayout; // const uint8_t * (64-bit pointer)
1118 uint64_t baseProperties; // const struct objc_property_list (64-bit pointer)
1121 inline void swapStruct(struct cfstring64_t &cfs) {
1122 sys::swapByteOrder(cfs.isa);
1123 sys::swapByteOrder(cfs.flags);
1124 sys::swapByteOrder(cfs.characters);
1125 sys::swapByteOrder(cfs.length);
1128 inline void swapStruct(struct class64_t &c) {
1129 sys::swapByteOrder(c.isa);
1130 sys::swapByteOrder(c.superclass);
1131 sys::swapByteOrder(c.cache);
1132 sys::swapByteOrder(c.vtable);
1133 sys::swapByteOrder(c.data);
1136 inline void swapStruct(struct class_ro64_t &cro) {
1137 sys::swapByteOrder(cro.flags);
1138 sys::swapByteOrder(cro.instanceStart);
1139 sys::swapByteOrder(cro.instanceSize);
1140 sys::swapByteOrder(cro.reserved);
1141 sys::swapByteOrder(cro.ivarLayout);
1142 sys::swapByteOrder(cro.name);
1143 sys::swapByteOrder(cro.baseMethods);
1144 sys::swapByteOrder(cro.baseProtocols);
1145 sys::swapByteOrder(cro.ivars);
1146 sys::swapByteOrder(cro.weakIvarLayout);
1147 sys::swapByteOrder(cro.baseProperties);
1150 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
1151 struct DisassembleInfo *info);
1153 // get_objc2_64bit_class_name() is used for disassembly and is passed a pointer
1154 // to an Objective-C class and returns the class name. It is also passed the
1155 // address of the pointer, so when the pointer is zero as it can be in an .o
1156 // file, that is used to look for an external relocation entry with a symbol
1158 const char *get_objc2_64bit_class_name(uint64_t pointer_value,
1159 uint64_t ReferenceValue,
1160 struct DisassembleInfo *info) {
1162 uint32_t offset, left;
1165 // The pointer_value can be 0 in an object file and have a relocation
1166 // entry for the class symbol at the ReferenceValue (the address of the
1168 if (pointer_value == 0) {
1169 r = get_pointer_64(ReferenceValue, offset, left, S, info);
1170 if (r == nullptr || left < sizeof(uint64_t))
1173 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
1174 if (symbol_name == nullptr)
1176 const char *class_name = strrchr(symbol_name, '$');
1177 if (class_name != nullptr && class_name[1] == '_' && class_name[2] != '\0')
1178 return class_name + 2;
1183 // The case were the pointer_value is non-zero and points to a class defined
1184 // in this Mach-O file.
1185 r = get_pointer_64(pointer_value, offset, left, S, info);
1186 if (r == nullptr || left < sizeof(struct class64_t))
1189 memcpy(&c, r, sizeof(struct class64_t));
1190 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1194 r = get_pointer_64(c.data, offset, left, S, info);
1195 if (r == nullptr || left < sizeof(struct class_ro64_t))
1197 struct class_ro64_t cro;
1198 memcpy(&cro, r, sizeof(struct class_ro64_t));
1199 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1203 const char *name = get_pointer_64(cro.name, offset, left, S, info);
1207 // get_objc2_64bit_cfstring_name is used for disassembly and is passed a
1208 // pointer to a cfstring and returns its name or nullptr.
1209 const char *get_objc2_64bit_cfstring_name(uint64_t ReferenceValue,
1210 struct DisassembleInfo *info) {
1211 const char *r, *name;
1212 uint32_t offset, left;
1214 struct cfstring64_t cfs;
1215 uint64_t cfs_characters;
1217 r = get_pointer_64(ReferenceValue, offset, left, S, info);
1218 if (r == nullptr || left < sizeof(struct cfstring64_t))
1220 memcpy(&cfs, r, sizeof(struct cfstring64_t));
1221 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1223 if (cfs.characters == 0) {
1225 const char *symbol_name = get_symbol_64(
1226 offset + offsetof(struct cfstring64_t, characters), S, info, n_value);
1227 if (symbol_name == nullptr)
1229 cfs_characters = n_value;
1231 cfs_characters = cfs.characters;
1232 name = get_pointer_64(cfs_characters, offset, left, S, info);
1237 // get_objc2_64bit_selref() is used for disassembly and is passed a the address
1238 // of a pointer to an Objective-C selector reference when the pointer value is
1239 // zero as in a .o file and is likely to have a external relocation entry with
1240 // who's symbol's n_value is the real pointer to the selector name. If that is
1241 // the case the real pointer to the selector name is returned else 0 is
1243 uint64_t get_objc2_64bit_selref(uint64_t ReferenceValue,
1244 struct DisassembleInfo *info) {
1245 uint32_t offset, left;
1248 const char *r = get_pointer_64(ReferenceValue, offset, left, S, info);
1249 if (r == nullptr || left < sizeof(uint64_t))
1252 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
1253 if (symbol_name == nullptr)
1258 // GuessLiteralPointer returns a string which for the item in the Mach-O file
1259 // for the address passed in as ReferenceValue for printing as a comment with
1260 // the instruction and also returns the corresponding type of that item
1261 // indirectly through ReferenceType.
1263 // If ReferenceValue is an address of literal cstring then a pointer to the
1264 // cstring is returned and ReferenceType is set to
1265 // LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr .
1267 // If ReferenceValue is an address of an Objective-C CFString, Selector ref or
1268 // Class ref that name is returned and the ReferenceType is set accordingly.
1270 // Lastly, literals which are Symbol address in a literal pool are looked for
1271 // and if found the symbol name is returned and ReferenceType is set to
1272 // LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr .
1274 // If there is no item in the Mach-O file for the address passed in as
1275 // ReferenceValue nullptr is returned and ReferenceType is unchanged.
1276 const char *GuessLiteralPointer(uint64_t ReferenceValue, uint64_t ReferencePC,
1277 uint64_t *ReferenceType,
1278 struct DisassembleInfo *info) {
1279 // First see if there is an external relocation entry at the ReferencePC.
1280 uint64_t sect_addr = info->S.getAddress();
1281 uint64_t sect_offset = ReferencePC - sect_addr;
1282 bool reloc_found = false;
1284 MachO::any_relocation_info RE;
1285 bool isExtern = false;
1287 for (const RelocationRef &Reloc : info->S.relocations()) {
1288 uint64_t RelocOffset;
1289 Reloc.getOffset(RelocOffset);
1290 if (RelocOffset == sect_offset) {
1291 Rel = Reloc.getRawDataRefImpl();
1292 RE = info->O->getRelocation(Rel);
1293 if (info->O->isRelocationScattered(RE))
1295 isExtern = info->O->getPlainRelocationExternal(RE);
1297 symbol_iterator RelocSym = Reloc.getSymbol();
1304 // If there is an external relocation entry for a symbol in a section
1305 // then used that symbol's value for the value of the reference.
1306 if (reloc_found && isExtern) {
1307 if (info->O->getAnyRelocationPCRel(RE)) {
1308 unsigned Type = info->O->getAnyRelocationType(RE);
1309 if (Type == MachO::X86_64_RELOC_SIGNED) {
1310 Symbol.getAddress(ReferenceValue);
1315 // Look for literals such as Objective-C CFStrings refs, Selector refs,
1316 // Message refs and Class refs.
1317 bool classref, selref, msgref, cfstring;
1318 uint64_t pointer_value = GuessPointerPointer(ReferenceValue, info, classref,
1319 selref, msgref, cfstring);
1320 if (classref == true && pointer_value == 0) {
1321 // Note the ReferenceValue is a pointer into the __objc_classrefs section.
1322 // And the pointer_value in that section is typically zero as it will be
1323 // set by dyld as part of the "bind information".
1324 const char *name = get_dyld_bind_info_symbolname(ReferenceValue, info);
1325 if (name != nullptr) {
1326 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
1327 const char *class_name = strrchr(name, '$');
1328 if (class_name != nullptr && class_name[1] == '_' &&
1329 class_name[2] != '\0') {
1330 info->class_name = class_name + 2;
1336 if (classref == true) {
1337 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
1339 get_objc2_64bit_class_name(pointer_value, ReferenceValue, info);
1340 if (name != nullptr)
1341 info->class_name = name;
1343 name = "bad class ref";
1347 if (cfstring == true) {
1348 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_CFString_Ref;
1349 const char *name = get_objc2_64bit_cfstring_name(ReferenceValue, info);
1353 if (selref == true && pointer_value == 0)
1354 pointer_value = get_objc2_64bit_selref(ReferenceValue, info);
1356 if (pointer_value != 0)
1357 ReferenceValue = pointer_value;
1359 const char *name = GuessCstringPointer(ReferenceValue, info);
1361 if (pointer_value != 0 && selref == true) {
1362 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Selector_Ref;
1363 info->selector_name = name;
1364 } else if (pointer_value != 0 && msgref == true) {
1365 info->class_name = nullptr;
1366 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message_Ref;
1367 info->selector_name = name;
1369 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr;
1373 // Lastly look for an indirect symbol with this ReferenceValue which is in
1374 // a literal pool. If found return that symbol name.
1375 name = GuessIndirectSymbol(ReferenceValue, info);
1377 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr;
1384 // SymbolizerSymbolLookUp is the symbol lookup function passed when creating
1385 // the Symbolizer. It looks up the ReferenceValue using the info passed via the
1386 // pointer to the struct DisassembleInfo that was passed when MCSymbolizer
1387 // is created and returns the symbol name that matches the ReferenceValue or
1388 // nullptr if none. The ReferenceType is passed in for the IN type of
1389 // reference the instruction is making from the values in defined in the header
1390 // "llvm-c/Disassembler.h". On return the ReferenceType can set to a specific
1391 // Out type and the ReferenceName will also be set which is added as a comment
1392 // to the disassembled instruction.
1395 // If the symbol name is a C++ mangled name then the demangled name is
1396 // returned through ReferenceName and ReferenceType is set to
1397 // LLVMDisassembler_ReferenceType_DeMangled_Name .
1400 // When this is called to get a symbol name for a branch target then the
1401 // ReferenceType will be LLVMDisassembler_ReferenceType_In_Branch and then
1402 // SymbolValue will be looked for in the indirect symbol table to determine if
1403 // it is an address for a symbol stub. If so then the symbol name for that
1404 // stub is returned indirectly through ReferenceName and then ReferenceType is
1405 // set to LLVMDisassembler_ReferenceType_Out_SymbolStub.
1407 // When this is called with an value loaded via a PC relative load then
1408 // ReferenceType will be LLVMDisassembler_ReferenceType_In_PCrel_Load then the
1409 // SymbolValue is checked to be an address of literal pointer, symbol pointer,
1410 // or an Objective-C meta data reference. If so the output ReferenceType is
1411 // set to correspond to that as well as setting the ReferenceName.
1412 const char *SymbolizerSymbolLookUp(void *DisInfo, uint64_t ReferenceValue,
1413 uint64_t *ReferenceType,
1414 uint64_t ReferencePC,
1415 const char **ReferenceName) {
1416 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
1417 // If no verbose symbolic information is wanted then just return nullptr.
1418 if (info->verbose == false) {
1419 *ReferenceName = nullptr;
1420 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1424 const char *SymbolName = GuessSymbolName(ReferenceValue, info);
1426 if (*ReferenceType == LLVMDisassembler_ReferenceType_In_Branch) {
1427 *ReferenceName = GuessIndirectSymbol(ReferenceValue, info);
1428 if (*ReferenceName != nullptr) {
1429 method_reference(info, ReferenceType, ReferenceName);
1430 if (*ReferenceType != LLVMDisassembler_ReferenceType_Out_Objc_Message)
1431 *ReferenceType = LLVMDisassembler_ReferenceType_Out_SymbolStub;
1434 if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
1435 if (info->demangled_name != nullptr)
1436 free(info->demangled_name);
1438 info->demangled_name =
1439 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
1440 if (info->demangled_name != nullptr) {
1441 *ReferenceName = info->demangled_name;
1442 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
1444 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1447 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1448 } else if (*ReferenceType == LLVMDisassembler_ReferenceType_In_PCrel_Load) {
1450 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
1452 method_reference(info, ReferenceType, ReferenceName);
1454 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1455 // If this is arm64 and the reference is an adrp instruction save the
1456 // instruction, passed in ReferenceValue and the address of the instruction
1457 // for use later if we see and add immediate instruction.
1458 } else if (info->O->getArch() == Triple::aarch64 &&
1459 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADRP) {
1460 info->adrp_inst = ReferenceValue;
1461 info->adrp_addr = ReferencePC;
1462 SymbolName = nullptr;
1463 *ReferenceName = nullptr;
1464 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1465 // If this is arm64 and reference is an add immediate instruction and we
1467 // seen an adrp instruction just before it and the adrp's Xd register
1469 // this add's Xn register reconstruct the value being referenced and look to
1470 // see if it is a literal pointer. Note the add immediate instruction is
1471 // passed in ReferenceValue.
1472 } else if (info->O->getArch() == Triple::aarch64 &&
1473 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADDXri &&
1474 ReferencePC - 4 == info->adrp_addr &&
1475 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
1476 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
1477 uint32_t addxri_inst;
1478 uint64_t adrp_imm, addxri_imm;
1481 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
1482 if (info->adrp_inst & 0x0200000)
1483 adrp_imm |= 0xfffffffffc000000LL;
1485 addxri_inst = ReferenceValue;
1486 addxri_imm = (addxri_inst >> 10) & 0xfff;
1487 if (((addxri_inst >> 22) & 0x3) == 1)
1490 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
1491 (adrp_imm << 12) + addxri_imm;
1494 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
1495 if (*ReferenceName == nullptr)
1496 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1497 // If this is arm64 and the reference is a load register instruction and we
1498 // have seen an adrp instruction just before it and the adrp's Xd register
1499 // matches this add's Xn register reconstruct the value being referenced and
1500 // look to see if it is a literal pointer. Note the load register
1501 // instruction is passed in ReferenceValue.
1502 } else if (info->O->getArch() == Triple::aarch64 &&
1503 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXui &&
1504 ReferencePC - 4 == info->adrp_addr &&
1505 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
1506 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
1507 uint32_t ldrxui_inst;
1508 uint64_t adrp_imm, ldrxui_imm;
1511 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
1512 if (info->adrp_inst & 0x0200000)
1513 adrp_imm |= 0xfffffffffc000000LL;
1515 ldrxui_inst = ReferenceValue;
1516 ldrxui_imm = (ldrxui_inst >> 10) & 0xfff;
1518 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
1519 (adrp_imm << 12) + (ldrxui_imm << 3);
1522 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
1523 if (*ReferenceName == nullptr)
1524 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1526 // If this arm64 and is an load register (PC-relative) instruction the
1527 // ReferenceValue is the PC plus the immediate value.
1528 else if (info->O->getArch() == Triple::aarch64 &&
1529 (*ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXl ||
1530 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADR)) {
1532 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
1533 if (*ReferenceName == nullptr)
1534 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1537 else if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
1538 if (info->demangled_name != nullptr)
1539 free(info->demangled_name);
1541 info->demangled_name =
1542 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
1543 if (info->demangled_name != nullptr) {
1544 *ReferenceName = info->demangled_name;
1545 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
1550 *ReferenceName = nullptr;
1551 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1557 /// \brief Emits the comments that are stored in the CommentStream.
1558 /// Each comment in the CommentStream must end with a newline.
1559 static void emitComments(raw_svector_ostream &CommentStream,
1560 SmallString<128> &CommentsToEmit,
1561 formatted_raw_ostream &FormattedOS,
1562 const MCAsmInfo &MAI) {
1563 // Flush the stream before taking its content.
1564 CommentStream.flush();
1565 StringRef Comments = CommentsToEmit.str();
1566 // Get the default information for printing a comment.
1567 const char *CommentBegin = MAI.getCommentString();
1568 unsigned CommentColumn = MAI.getCommentColumn();
1569 bool IsFirst = true;
1570 while (!Comments.empty()) {
1572 FormattedOS << '\n';
1573 // Emit a line of comments.
1574 FormattedOS.PadToColumn(CommentColumn);
1575 size_t Position = Comments.find('\n');
1576 FormattedOS << CommentBegin << ' ' << Comments.substr(0, Position);
1577 // Move after the newline character.
1578 Comments = Comments.substr(Position + 1);
1581 FormattedOS.flush();
1583 // Tell the comment stream that the vector changed underneath it.
1584 CommentsToEmit.clear();
1585 CommentStream.resync();
1588 static void DisassembleInputMachO2(StringRef Filename,
1589 MachOObjectFile *MachOOF) {
1590 const char *McpuDefault = nullptr;
1591 const Target *ThumbTarget = nullptr;
1592 const Target *TheTarget = GetTarget(MachOOF, &McpuDefault, &ThumbTarget);
1594 // GetTarget prints out stuff.
1597 if (MCPU.empty() && McpuDefault)
1600 std::unique_ptr<const MCInstrInfo> InstrInfo(TheTarget->createMCInstrInfo());
1601 std::unique_ptr<const MCInstrInfo> ThumbInstrInfo;
1603 ThumbInstrInfo.reset(ThumbTarget->createMCInstrInfo());
1605 // Package up features to be passed to target/subtarget
1606 std::string FeaturesStr;
1607 if (MAttrs.size()) {
1608 SubtargetFeatures Features;
1609 for (unsigned i = 0; i != MAttrs.size(); ++i)
1610 Features.AddFeature(MAttrs[i]);
1611 FeaturesStr = Features.getString();
1614 // Set up disassembler.
1615 std::unique_ptr<const MCRegisterInfo> MRI(
1616 TheTarget->createMCRegInfo(TripleName));
1617 std::unique_ptr<const MCAsmInfo> AsmInfo(
1618 TheTarget->createMCAsmInfo(*MRI, TripleName));
1619 std::unique_ptr<const MCSubtargetInfo> STI(
1620 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr));
1621 MCContext Ctx(AsmInfo.get(), MRI.get(), nullptr);
1622 std::unique_ptr<MCDisassembler> DisAsm(
1623 TheTarget->createMCDisassembler(*STI, Ctx));
1624 std::unique_ptr<MCSymbolizer> Symbolizer;
1625 struct DisassembleInfo SymbolizerInfo;
1626 std::unique_ptr<MCRelocationInfo> RelInfo(
1627 TheTarget->createMCRelocationInfo(TripleName, Ctx));
1629 Symbolizer.reset(TheTarget->createMCSymbolizer(
1630 TripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
1631 &SymbolizerInfo, &Ctx, RelInfo.release()));
1632 DisAsm->setSymbolizer(std::move(Symbolizer));
1634 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
1635 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
1636 AsmPrinterVariant, *AsmInfo, *InstrInfo, *MRI, *STI));
1637 // Set the display preference for hex vs. decimal immediates.
1638 IP->setPrintImmHex(PrintImmHex);
1639 // Comment stream and backing vector.
1640 SmallString<128> CommentsToEmit;
1641 raw_svector_ostream CommentStream(CommentsToEmit);
1643 if (!AsmInfo || !STI || !DisAsm || !IP) {
1644 errs() << "error: couldn't initialize disassembler for target "
1645 << TripleName << '\n';
1649 // Set up thumb disassembler.
1650 std::unique_ptr<const MCRegisterInfo> ThumbMRI;
1651 std::unique_ptr<const MCAsmInfo> ThumbAsmInfo;
1652 std::unique_ptr<const MCSubtargetInfo> ThumbSTI;
1653 std::unique_ptr<MCDisassembler> ThumbDisAsm;
1654 std::unique_ptr<MCInstPrinter> ThumbIP;
1655 std::unique_ptr<MCContext> ThumbCtx;
1656 std::unique_ptr<MCSymbolizer> ThumbSymbolizer;
1657 struct DisassembleInfo ThumbSymbolizerInfo;
1658 std::unique_ptr<MCRelocationInfo> ThumbRelInfo;
1660 ThumbMRI.reset(ThumbTarget->createMCRegInfo(ThumbTripleName));
1662 ThumbTarget->createMCAsmInfo(*ThumbMRI, ThumbTripleName));
1664 ThumbTarget->createMCSubtargetInfo(ThumbTripleName, MCPU, FeaturesStr));
1665 ThumbCtx.reset(new MCContext(ThumbAsmInfo.get(), ThumbMRI.get(), nullptr));
1666 ThumbDisAsm.reset(ThumbTarget->createMCDisassembler(*ThumbSTI, *ThumbCtx));
1667 MCContext *PtrThumbCtx = ThumbCtx.get();
1669 ThumbTarget->createMCRelocationInfo(ThumbTripleName, *PtrThumbCtx));
1671 ThumbSymbolizer.reset(ThumbTarget->createMCSymbolizer(
1672 ThumbTripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
1673 &ThumbSymbolizerInfo, PtrThumbCtx, ThumbRelInfo.release()));
1674 ThumbDisAsm->setSymbolizer(std::move(ThumbSymbolizer));
1676 int ThumbAsmPrinterVariant = ThumbAsmInfo->getAssemblerDialect();
1677 ThumbIP.reset(ThumbTarget->createMCInstPrinter(
1678 ThumbAsmPrinterVariant, *ThumbAsmInfo, *ThumbInstrInfo, *ThumbMRI,
1680 // Set the display preference for hex vs. decimal immediates.
1681 ThumbIP->setPrintImmHex(PrintImmHex);
1684 if (ThumbTarget && (!ThumbAsmInfo || !ThumbSTI || !ThumbDisAsm || !ThumbIP)) {
1685 errs() << "error: couldn't initialize disassembler for target "
1686 << ThumbTripleName << '\n';
1690 outs() << '\n' << Filename << ":\n\n";
1692 MachO::mach_header Header = MachOOF->getHeader();
1694 // FIXME: Using the -cfg command line option, this code used to be able to
1695 // annotate relocations with the referenced symbol's name, and if this was
1696 // inside a __[cf]string section, the data it points to. This is now replaced
1697 // by the upcoming MCSymbolizer, which needs the appropriate setup done above.
1698 std::vector<SectionRef> Sections;
1699 std::vector<SymbolRef> Symbols;
1700 SmallVector<uint64_t, 8> FoundFns;
1701 uint64_t BaseSegmentAddress;
1703 getSectionsAndSymbols(Header, MachOOF, Sections, Symbols, FoundFns,
1704 BaseSegmentAddress);
1706 // Sort the symbols by address, just in case they didn't come in that way.
1707 std::sort(Symbols.begin(), Symbols.end(), SymbolSorter());
1709 // Build a data in code table that is sorted on by the address of each entry.
1710 uint64_t BaseAddress = 0;
1711 if (Header.filetype == MachO::MH_OBJECT)
1712 BaseAddress = Sections[0].getAddress();
1714 BaseAddress = BaseSegmentAddress;
1716 for (dice_iterator DI = MachOOF->begin_dices(), DE = MachOOF->end_dices();
1719 DI->getOffset(Offset);
1720 Dices.push_back(std::make_pair(BaseAddress + Offset, *DI));
1722 array_pod_sort(Dices.begin(), Dices.end());
1725 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
1727 raw_ostream &DebugOut = nulls();
1730 std::unique_ptr<DIContext> diContext;
1731 ObjectFile *DbgObj = MachOOF;
1732 // Try to find debug info and set up the DIContext for it.
1734 // A separate DSym file path was specified, parse it as a macho file,
1735 // get the sections and supply it to the section name parsing machinery.
1736 if (!DSYMFile.empty()) {
1737 ErrorOr<std::unique_ptr<MemoryBuffer>> BufOrErr =
1738 MemoryBuffer::getFileOrSTDIN(DSYMFile);
1739 if (std::error_code EC = BufOrErr.getError()) {
1740 errs() << "llvm-objdump: " << Filename << ": " << EC.message() << '\n';
1744 ObjectFile::createMachOObjectFile(BufOrErr.get()->getMemBufferRef())
1749 // Setup the DIContext
1750 diContext.reset(DIContext::getDWARFContext(*DbgObj));
1753 for (unsigned SectIdx = 0; SectIdx != Sections.size(); SectIdx++) {
1755 bool SectIsText = Sections[SectIdx].isText();
1756 if (SectIsText == false)
1760 if (Sections[SectIdx].getName(SectName) || SectName != "__text")
1761 continue; // Skip non-text sections
1763 DataRefImpl DR = Sections[SectIdx].getRawDataRefImpl();
1765 StringRef SegmentName = MachOOF->getSectionFinalSegmentName(DR);
1766 if (SegmentName != "__TEXT")
1770 Sections[SectIdx].getContents(BytesStr);
1771 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()),
1773 uint64_t SectAddress = Sections[SectIdx].getAddress();
1775 bool symbolTableWorked = false;
1777 // Parse relocations.
1778 std::vector<std::pair<uint64_t, SymbolRef>> Relocs;
1779 for (const RelocationRef &Reloc : Sections[SectIdx].relocations()) {
1780 uint64_t RelocOffset;
1781 Reloc.getOffset(RelocOffset);
1782 uint64_t SectionAddress = Sections[SectIdx].getAddress();
1783 RelocOffset -= SectionAddress;
1785 symbol_iterator RelocSym = Reloc.getSymbol();
1787 Relocs.push_back(std::make_pair(RelocOffset, *RelocSym));
1789 array_pod_sort(Relocs.begin(), Relocs.end());
1791 // Create a map of symbol addresses to symbol names for use by
1792 // the SymbolizerSymbolLookUp() routine.
1793 SymbolAddressMap AddrMap;
1794 for (const SymbolRef &Symbol : MachOOF->symbols()) {
1797 if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data ||
1798 ST == SymbolRef::ST_Other) {
1800 Symbol.getAddress(Address);
1802 Symbol.getName(SymName);
1803 AddrMap[Address] = SymName;
1806 // Set up the block of info used by the Symbolizer call backs.
1807 SymbolizerInfo.verbose = true;
1808 SymbolizerInfo.O = MachOOF;
1809 SymbolizerInfo.S = Sections[SectIdx];
1810 SymbolizerInfo.AddrMap = &AddrMap;
1811 SymbolizerInfo.Sections = &Sections;
1812 SymbolizerInfo.class_name = nullptr;
1813 SymbolizerInfo.selector_name = nullptr;
1814 SymbolizerInfo.method = nullptr;
1815 SymbolizerInfo.demangled_name = nullptr;
1816 SymbolizerInfo.bindtable = nullptr;
1817 // Same for the ThumbSymbolizer
1818 ThumbSymbolizerInfo.verbose = true;
1819 ThumbSymbolizerInfo.O = MachOOF;
1820 ThumbSymbolizerInfo.S = Sections[SectIdx];
1821 ThumbSymbolizerInfo.AddrMap = &AddrMap;
1822 ThumbSymbolizerInfo.Sections = &Sections;
1823 ThumbSymbolizerInfo.class_name = nullptr;
1824 ThumbSymbolizerInfo.selector_name = nullptr;
1825 ThumbSymbolizerInfo.method = nullptr;
1826 ThumbSymbolizerInfo.demangled_name = nullptr;
1827 ThumbSymbolizerInfo.bindtable = nullptr;
1829 // Disassemble symbol by symbol.
1830 for (unsigned SymIdx = 0; SymIdx != Symbols.size(); SymIdx++) {
1832 Symbols[SymIdx].getName(SymName);
1835 Symbols[SymIdx].getType(ST);
1836 if (ST != SymbolRef::ST_Function)
1839 // Make sure the symbol is defined in this section.
1840 bool containsSym = Sections[SectIdx].containsSymbol(Symbols[SymIdx]);
1844 // Start at the address of the symbol relative to the section's address.
1846 uint64_t SectionAddress = Sections[SectIdx].getAddress();
1847 Symbols[SymIdx].getAddress(Start);
1848 Start -= SectionAddress;
1850 // Stop disassembling either at the beginning of the next symbol or at
1851 // the end of the section.
1852 bool containsNextSym = false;
1853 uint64_t NextSym = 0;
1854 uint64_t NextSymIdx = SymIdx + 1;
1855 while (Symbols.size() > NextSymIdx) {
1856 SymbolRef::Type NextSymType;
1857 Symbols[NextSymIdx].getType(NextSymType);
1858 if (NextSymType == SymbolRef::ST_Function) {
1860 Sections[SectIdx].containsSymbol(Symbols[NextSymIdx]);
1861 Symbols[NextSymIdx].getAddress(NextSym);
1862 NextSym -= SectionAddress;
1868 uint64_t SectSize = Sections[SectIdx].getSize();
1869 uint64_t End = containsNextSym ? NextSym : SectSize;
1872 symbolTableWorked = true;
1874 DataRefImpl Symb = Symbols[SymIdx].getRawDataRefImpl();
1876 (MachOOF->getSymbolFlags(Symb) & SymbolRef::SF_Thumb) && ThumbTarget;
1878 outs() << SymName << ":\n";
1879 DILineInfo lastLine;
1880 for (uint64_t Index = Start; Index < End; Index += Size) {
1883 uint64_t PC = SectAddress + Index;
1884 if (FullLeadingAddr) {
1885 if (MachOOF->is64Bit())
1886 outs() << format("%016" PRIx64, PC);
1888 outs() << format("%08" PRIx64, PC);
1890 outs() << format("%8" PRIx64 ":", PC);
1895 // Check the data in code table here to see if this is data not an
1896 // instruction to be disassembled.
1898 Dice.push_back(std::make_pair(PC, DiceRef()));
1899 dice_table_iterator DTI =
1900 std::search(Dices.begin(), Dices.end(), Dice.begin(), Dice.end(),
1901 compareDiceTableEntries);
1902 if (DTI != Dices.end()) {
1904 DTI->second.getLength(Length);
1906 DTI->second.getKind(Kind);
1907 Size = DumpDataInCode(reinterpret_cast<const char *>(Bytes.data()) +
1910 if ((Kind == MachO::DICE_KIND_JUMP_TABLE8) &&
1911 (PC == (DTI->first + Length - 1)) && (Length & 1))
1916 SmallVector<char, 64> AnnotationsBytes;
1917 raw_svector_ostream Annotations(AnnotationsBytes);
1921 gotInst = ThumbDisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
1922 PC, DebugOut, Annotations);
1924 gotInst = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), PC,
1925 DebugOut, Annotations);
1927 if (!NoShowRawInsn) {
1928 DumpBytes(StringRef(
1929 reinterpret_cast<const char *>(Bytes.data()) + Index, Size));
1931 formatted_raw_ostream FormattedOS(outs());
1932 Annotations.flush();
1933 StringRef AnnotationsStr = Annotations.str();
1935 ThumbIP->printInst(&Inst, FormattedOS, AnnotationsStr);
1937 IP->printInst(&Inst, FormattedOS, AnnotationsStr);
1938 emitComments(CommentStream, CommentsToEmit, FormattedOS, *AsmInfo);
1940 // Print debug info.
1942 DILineInfo dli = diContext->getLineInfoForAddress(PC);
1943 // Print valid line info if it changed.
1944 if (dli != lastLine && dli.Line != 0)
1945 outs() << "\t## " << dli.FileName << ':' << dli.Line << ':'
1951 unsigned int Arch = MachOOF->getArch();
1952 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
1953 outs() << format("\t.byte 0x%02x #bad opcode\n",
1954 *(Bytes.data() + Index) & 0xff);
1955 Size = 1; // skip exactly one illegible byte and move on.
1956 } else if (Arch == Triple::aarch64) {
1957 uint32_t opcode = (*(Bytes.data() + Index) & 0xff) |
1958 (*(Bytes.data() + Index + 1) & 0xff) << 8 |
1959 (*(Bytes.data() + Index + 2) & 0xff) << 16 |
1960 (*(Bytes.data() + Index + 3) & 0xff) << 24;
1961 outs() << format("\t.long\t0x%08x\n", opcode);
1964 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
1966 Size = 1; // skip illegible bytes
1971 if (!symbolTableWorked) {
1972 // Reading the symbol table didn't work, disassemble the whole section.
1973 uint64_t SectAddress = Sections[SectIdx].getAddress();
1974 uint64_t SectSize = Sections[SectIdx].getSize();
1976 for (uint64_t Index = 0; Index < SectSize; Index += InstSize) {
1979 uint64_t PC = SectAddress + Index;
1980 if (DisAsm->getInstruction(Inst, InstSize, Bytes.slice(Index), PC,
1981 DebugOut, nulls())) {
1982 if (FullLeadingAddr) {
1983 if (MachOOF->is64Bit())
1984 outs() << format("%016" PRIx64, PC);
1986 outs() << format("%08" PRIx64, PC);
1988 outs() << format("%8" PRIx64 ":", PC);
1990 if (!NoShowRawInsn) {
1993 StringRef(reinterpret_cast<const char *>(Bytes.data()) + Index,
1996 IP->printInst(&Inst, outs(), "");
1999 unsigned int Arch = MachOOF->getArch();
2000 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
2001 outs() << format("\t.byte 0x%02x #bad opcode\n",
2002 *(Bytes.data() + Index) & 0xff);
2003 InstSize = 1; // skip exactly one illegible byte and move on.
2005 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
2007 InstSize = 1; // skip illegible bytes
2012 if (SymbolizerInfo.method != nullptr)
2013 free(SymbolizerInfo.method);
2014 if (SymbolizerInfo.demangled_name != nullptr)
2015 free(SymbolizerInfo.demangled_name);
2016 if (SymbolizerInfo.bindtable != nullptr)
2017 delete SymbolizerInfo.bindtable;
2018 if (ThumbSymbolizerInfo.method != nullptr)
2019 free(ThumbSymbolizerInfo.method);
2020 if (ThumbSymbolizerInfo.demangled_name != nullptr)
2021 free(ThumbSymbolizerInfo.demangled_name);
2022 if (ThumbSymbolizerInfo.bindtable != nullptr)
2023 delete ThumbSymbolizerInfo.bindtable;
2027 //===----------------------------------------------------------------------===//
2028 // __compact_unwind section dumping
2029 //===----------------------------------------------------------------------===//
2033 template <typename T> static uint64_t readNext(const char *&Buf) {
2034 using llvm::support::little;
2035 using llvm::support::unaligned;
2037 uint64_t Val = support::endian::read<T, little, unaligned>(Buf);
2042 struct CompactUnwindEntry {
2043 uint32_t OffsetInSection;
2045 uint64_t FunctionAddr;
2047 uint32_t CompactEncoding;
2048 uint64_t PersonalityAddr;
2051 RelocationRef FunctionReloc;
2052 RelocationRef PersonalityReloc;
2053 RelocationRef LSDAReloc;
2055 CompactUnwindEntry(StringRef Contents, unsigned Offset, bool Is64)
2056 : OffsetInSection(Offset) {
2058 read<uint64_t>(Contents.data() + Offset);
2060 read<uint32_t>(Contents.data() + Offset);
2064 template <typename UIntPtr> void read(const char *Buf) {
2065 FunctionAddr = readNext<UIntPtr>(Buf);
2066 Length = readNext<uint32_t>(Buf);
2067 CompactEncoding = readNext<uint32_t>(Buf);
2068 PersonalityAddr = readNext<UIntPtr>(Buf);
2069 LSDAAddr = readNext<UIntPtr>(Buf);
2074 /// Given a relocation from __compact_unwind, consisting of the RelocationRef
2075 /// and data being relocated, determine the best base Name and Addend to use for
2076 /// display purposes.
2078 /// 1. An Extern relocation will directly reference a symbol (and the data is
2079 /// then already an addend), so use that.
2080 /// 2. Otherwise the data is an offset in the object file's layout; try to find
2081 // a symbol before it in the same section, and use the offset from there.
2082 /// 3. Finally, if all that fails, fall back to an offset from the start of the
2083 /// referenced section.
2084 static void findUnwindRelocNameAddend(const MachOObjectFile *Obj,
2085 std::map<uint64_t, SymbolRef> &Symbols,
2086 const RelocationRef &Reloc, uint64_t Addr,
2087 StringRef &Name, uint64_t &Addend) {
2088 if (Reloc.getSymbol() != Obj->symbol_end()) {
2089 Reloc.getSymbol()->getName(Name);
2094 auto RE = Obj->getRelocation(Reloc.getRawDataRefImpl());
2095 SectionRef RelocSection = Obj->getRelocationSection(RE);
2097 uint64_t SectionAddr = RelocSection.getAddress();
2099 auto Sym = Symbols.upper_bound(Addr);
2100 if (Sym == Symbols.begin()) {
2101 // The first symbol in the object is after this reference, the best we can
2102 // do is section-relative notation.
2103 RelocSection.getName(Name);
2104 Addend = Addr - SectionAddr;
2108 // Go back one so that SymbolAddress <= Addr.
2111 section_iterator SymSection = Obj->section_end();
2112 Sym->second.getSection(SymSection);
2113 if (RelocSection == *SymSection) {
2114 // There's a valid symbol in the same section before this reference.
2115 Sym->second.getName(Name);
2116 Addend = Addr - Sym->first;
2120 // There is a symbol before this reference, but it's in a different
2121 // section. Probably not helpful to mention it, so use the section name.
2122 RelocSection.getName(Name);
2123 Addend = Addr - SectionAddr;
2126 static void printUnwindRelocDest(const MachOObjectFile *Obj,
2127 std::map<uint64_t, SymbolRef> &Symbols,
2128 const RelocationRef &Reloc, uint64_t Addr) {
2132 if (!Reloc.getObjectFile())
2135 findUnwindRelocNameAddend(Obj, Symbols, Reloc, Addr, Name, Addend);
2139 outs() << " + " << format("0x%" PRIx64, Addend);
2143 printMachOCompactUnwindSection(const MachOObjectFile *Obj,
2144 std::map<uint64_t, SymbolRef> &Symbols,
2145 const SectionRef &CompactUnwind) {
2147 assert(Obj->isLittleEndian() &&
2148 "There should not be a big-endian .o with __compact_unwind");
2150 bool Is64 = Obj->is64Bit();
2151 uint32_t PointerSize = Is64 ? sizeof(uint64_t) : sizeof(uint32_t);
2152 uint32_t EntrySize = 3 * PointerSize + 2 * sizeof(uint32_t);
2155 CompactUnwind.getContents(Contents);
2157 SmallVector<CompactUnwindEntry, 4> CompactUnwinds;
2159 // First populate the initial raw offsets, encodings and so on from the entry.
2160 for (unsigned Offset = 0; Offset < Contents.size(); Offset += EntrySize) {
2161 CompactUnwindEntry Entry(Contents.data(), Offset, Is64);
2162 CompactUnwinds.push_back(Entry);
2165 // Next we need to look at the relocations to find out what objects are
2166 // actually being referred to.
2167 for (const RelocationRef &Reloc : CompactUnwind.relocations()) {
2168 uint64_t RelocAddress;
2169 Reloc.getOffset(RelocAddress);
2171 uint32_t EntryIdx = RelocAddress / EntrySize;
2172 uint32_t OffsetInEntry = RelocAddress - EntryIdx * EntrySize;
2173 CompactUnwindEntry &Entry = CompactUnwinds[EntryIdx];
2175 if (OffsetInEntry == 0)
2176 Entry.FunctionReloc = Reloc;
2177 else if (OffsetInEntry == PointerSize + 2 * sizeof(uint32_t))
2178 Entry.PersonalityReloc = Reloc;
2179 else if (OffsetInEntry == 2 * PointerSize + 2 * sizeof(uint32_t))
2180 Entry.LSDAReloc = Reloc;
2182 llvm_unreachable("Unexpected relocation in __compact_unwind section");
2185 // Finally, we're ready to print the data we've gathered.
2186 outs() << "Contents of __compact_unwind section:\n";
2187 for (auto &Entry : CompactUnwinds) {
2188 outs() << " Entry at offset "
2189 << format("0x%" PRIx32, Entry.OffsetInSection) << ":\n";
2191 // 1. Start of the region this entry applies to.
2192 outs() << " start: " << format("0x%" PRIx64,
2193 Entry.FunctionAddr) << ' ';
2194 printUnwindRelocDest(Obj, Symbols, Entry.FunctionReloc, Entry.FunctionAddr);
2197 // 2. Length of the region this entry applies to.
2198 outs() << " length: " << format("0x%" PRIx32, Entry.Length)
2200 // 3. The 32-bit compact encoding.
2201 outs() << " compact encoding: "
2202 << format("0x%08" PRIx32, Entry.CompactEncoding) << '\n';
2204 // 4. The personality function, if present.
2205 if (Entry.PersonalityReloc.getObjectFile()) {
2206 outs() << " personality function: "
2207 << format("0x%" PRIx64, Entry.PersonalityAddr) << ' ';
2208 printUnwindRelocDest(Obj, Symbols, Entry.PersonalityReloc,
2209 Entry.PersonalityAddr);
2213 // 5. This entry's language-specific data area.
2214 if (Entry.LSDAReloc.getObjectFile()) {
2215 outs() << " LSDA: " << format("0x%" PRIx64,
2216 Entry.LSDAAddr) << ' ';
2217 printUnwindRelocDest(Obj, Symbols, Entry.LSDAReloc, Entry.LSDAAddr);
2223 //===----------------------------------------------------------------------===//
2224 // __unwind_info section dumping
2225 //===----------------------------------------------------------------------===//
2227 static void printRegularSecondLevelUnwindPage(const char *PageStart) {
2228 const char *Pos = PageStart;
2229 uint32_t Kind = readNext<uint32_t>(Pos);
2231 assert(Kind == 2 && "kind for a regular 2nd level index should be 2");
2233 uint16_t EntriesStart = readNext<uint16_t>(Pos);
2234 uint16_t NumEntries = readNext<uint16_t>(Pos);
2236 Pos = PageStart + EntriesStart;
2237 for (unsigned i = 0; i < NumEntries; ++i) {
2238 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
2239 uint32_t Encoding = readNext<uint32_t>(Pos);
2241 outs() << " [" << i << "]: "
2242 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
2244 << "encoding=" << format("0x%08" PRIx32, Encoding) << '\n';
2248 static void printCompressedSecondLevelUnwindPage(
2249 const char *PageStart, uint32_t FunctionBase,
2250 const SmallVectorImpl<uint32_t> &CommonEncodings) {
2251 const char *Pos = PageStart;
2252 uint32_t Kind = readNext<uint32_t>(Pos);
2254 assert(Kind == 3 && "kind for a compressed 2nd level index should be 3");
2256 uint16_t EntriesStart = readNext<uint16_t>(Pos);
2257 uint16_t NumEntries = readNext<uint16_t>(Pos);
2259 uint16_t EncodingsStart = readNext<uint16_t>(Pos);
2260 readNext<uint16_t>(Pos);
2261 const auto *PageEncodings = reinterpret_cast<const support::ulittle32_t *>(
2262 PageStart + EncodingsStart);
2264 Pos = PageStart + EntriesStart;
2265 for (unsigned i = 0; i < NumEntries; ++i) {
2266 uint32_t Entry = readNext<uint32_t>(Pos);
2267 uint32_t FunctionOffset = FunctionBase + (Entry & 0xffffff);
2268 uint32_t EncodingIdx = Entry >> 24;
2271 if (EncodingIdx < CommonEncodings.size())
2272 Encoding = CommonEncodings[EncodingIdx];
2274 Encoding = PageEncodings[EncodingIdx - CommonEncodings.size()];
2276 outs() << " [" << i << "]: "
2277 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
2279 << "encoding[" << EncodingIdx
2280 << "]=" << format("0x%08" PRIx32, Encoding) << '\n';
2284 static void printMachOUnwindInfoSection(const MachOObjectFile *Obj,
2285 std::map<uint64_t, SymbolRef> &Symbols,
2286 const SectionRef &UnwindInfo) {
2288 assert(Obj->isLittleEndian() &&
2289 "There should not be a big-endian .o with __unwind_info");
2291 outs() << "Contents of __unwind_info section:\n";
2294 UnwindInfo.getContents(Contents);
2295 const char *Pos = Contents.data();
2297 //===----------------------------------
2299 //===----------------------------------
2301 uint32_t Version = readNext<uint32_t>(Pos);
2302 outs() << " Version: "
2303 << format("0x%" PRIx32, Version) << '\n';
2304 assert(Version == 1 && "only understand version 1");
2306 uint32_t CommonEncodingsStart = readNext<uint32_t>(Pos);
2307 outs() << " Common encodings array section offset: "
2308 << format("0x%" PRIx32, CommonEncodingsStart) << '\n';
2309 uint32_t NumCommonEncodings = readNext<uint32_t>(Pos);
2310 outs() << " Number of common encodings in array: "
2311 << format("0x%" PRIx32, NumCommonEncodings) << '\n';
2313 uint32_t PersonalitiesStart = readNext<uint32_t>(Pos);
2314 outs() << " Personality function array section offset: "
2315 << format("0x%" PRIx32, PersonalitiesStart) << '\n';
2316 uint32_t NumPersonalities = readNext<uint32_t>(Pos);
2317 outs() << " Number of personality functions in array: "
2318 << format("0x%" PRIx32, NumPersonalities) << '\n';
2320 uint32_t IndicesStart = readNext<uint32_t>(Pos);
2321 outs() << " Index array section offset: "
2322 << format("0x%" PRIx32, IndicesStart) << '\n';
2323 uint32_t NumIndices = readNext<uint32_t>(Pos);
2324 outs() << " Number of indices in array: "
2325 << format("0x%" PRIx32, NumIndices) << '\n';
2327 //===----------------------------------
2328 // A shared list of common encodings
2329 //===----------------------------------
2331 // These occupy indices in the range [0, N] whenever an encoding is referenced
2332 // from a compressed 2nd level index table. In practice the linker only
2333 // creates ~128 of these, so that indices are available to embed encodings in
2334 // the 2nd level index.
2336 SmallVector<uint32_t, 64> CommonEncodings;
2337 outs() << " Common encodings: (count = " << NumCommonEncodings << ")\n";
2338 Pos = Contents.data() + CommonEncodingsStart;
2339 for (unsigned i = 0; i < NumCommonEncodings; ++i) {
2340 uint32_t Encoding = readNext<uint32_t>(Pos);
2341 CommonEncodings.push_back(Encoding);
2343 outs() << " encoding[" << i << "]: " << format("0x%08" PRIx32, Encoding)
2347 //===----------------------------------
2348 // Personality functions used in this executable
2349 //===----------------------------------
2351 // There should be only a handful of these (one per source language,
2352 // roughly). Particularly since they only get 2 bits in the compact encoding.
2354 outs() << " Personality functions: (count = " << NumPersonalities << ")\n";
2355 Pos = Contents.data() + PersonalitiesStart;
2356 for (unsigned i = 0; i < NumPersonalities; ++i) {
2357 uint32_t PersonalityFn = readNext<uint32_t>(Pos);
2358 outs() << " personality[" << i + 1
2359 << "]: " << format("0x%08" PRIx32, PersonalityFn) << '\n';
2362 //===----------------------------------
2363 // The level 1 index entries
2364 //===----------------------------------
2366 // These specify an approximate place to start searching for the more detailed
2367 // information, sorted by PC.
2370 uint32_t FunctionOffset;
2371 uint32_t SecondLevelPageStart;
2375 SmallVector<IndexEntry, 4> IndexEntries;
2377 outs() << " Top level indices: (count = " << NumIndices << ")\n";
2378 Pos = Contents.data() + IndicesStart;
2379 for (unsigned i = 0; i < NumIndices; ++i) {
2382 Entry.FunctionOffset = readNext<uint32_t>(Pos);
2383 Entry.SecondLevelPageStart = readNext<uint32_t>(Pos);
2384 Entry.LSDAStart = readNext<uint32_t>(Pos);
2385 IndexEntries.push_back(Entry);
2387 outs() << " [" << i << "]: "
2388 << "function offset=" << format("0x%08" PRIx32, Entry.FunctionOffset)
2390 << "2nd level page offset="
2391 << format("0x%08" PRIx32, Entry.SecondLevelPageStart) << ", "
2392 << "LSDA offset=" << format("0x%08" PRIx32, Entry.LSDAStart) << '\n';
2395 //===----------------------------------
2396 // Next come the LSDA tables
2397 //===----------------------------------
2399 // The LSDA layout is rather implicit: it's a contiguous array of entries from
2400 // the first top-level index's LSDAOffset to the last (sentinel).
2402 outs() << " LSDA descriptors:\n";
2403 Pos = Contents.data() + IndexEntries[0].LSDAStart;
2404 int NumLSDAs = (IndexEntries.back().LSDAStart - IndexEntries[0].LSDAStart) /
2405 (2 * sizeof(uint32_t));
2406 for (int i = 0; i < NumLSDAs; ++i) {
2407 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
2408 uint32_t LSDAOffset = readNext<uint32_t>(Pos);
2409 outs() << " [" << i << "]: "
2410 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
2412 << "LSDA offset=" << format("0x%08" PRIx32, LSDAOffset) << '\n';
2415 //===----------------------------------
2416 // Finally, the 2nd level indices
2417 //===----------------------------------
2419 // Generally these are 4K in size, and have 2 possible forms:
2420 // + Regular stores up to 511 entries with disparate encodings
2421 // + Compressed stores up to 1021 entries if few enough compact encoding
2423 outs() << " Second level indices:\n";
2424 for (unsigned i = 0; i < IndexEntries.size() - 1; ++i) {
2425 // The final sentinel top-level index has no associated 2nd level page
2426 if (IndexEntries[i].SecondLevelPageStart == 0)
2429 outs() << " Second level index[" << i << "]: "
2430 << "offset in section="
2431 << format("0x%08" PRIx32, IndexEntries[i].SecondLevelPageStart)
2433 << "base function offset="
2434 << format("0x%08" PRIx32, IndexEntries[i].FunctionOffset) << '\n';
2436 Pos = Contents.data() + IndexEntries[i].SecondLevelPageStart;
2437 uint32_t Kind = *reinterpret_cast<const support::ulittle32_t *>(Pos);
2439 printRegularSecondLevelUnwindPage(Pos);
2441 printCompressedSecondLevelUnwindPage(Pos, IndexEntries[i].FunctionOffset,
2444 llvm_unreachable("Do not know how to print this kind of 2nd level page");
2448 void llvm::printMachOUnwindInfo(const MachOObjectFile *Obj) {
2449 std::map<uint64_t, SymbolRef> Symbols;
2450 for (const SymbolRef &SymRef : Obj->symbols()) {
2451 // Discard any undefined or absolute symbols. They're not going to take part
2452 // in the convenience lookup for unwind info and just take up resources.
2453 section_iterator Section = Obj->section_end();
2454 SymRef.getSection(Section);
2455 if (Section == Obj->section_end())
2459 SymRef.getAddress(Addr);
2460 Symbols.insert(std::make_pair(Addr, SymRef));
2463 for (const SectionRef &Section : Obj->sections()) {
2465 Section.getName(SectName);
2466 if (SectName == "__compact_unwind")
2467 printMachOCompactUnwindSection(Obj, Symbols, Section);
2468 else if (SectName == "__unwind_info")
2469 printMachOUnwindInfoSection(Obj, Symbols, Section);
2470 else if (SectName == "__eh_frame")
2471 outs() << "llvm-objdump: warning: unhandled __eh_frame section\n";
2475 static void PrintMachHeader(uint32_t magic, uint32_t cputype,
2476 uint32_t cpusubtype, uint32_t filetype,
2477 uint32_t ncmds, uint32_t sizeofcmds, uint32_t flags,
2479 outs() << "Mach header\n";
2480 outs() << " magic cputype cpusubtype caps filetype ncmds "
2481 "sizeofcmds flags\n";
2483 if (magic == MachO::MH_MAGIC)
2484 outs() << " MH_MAGIC";
2485 else if (magic == MachO::MH_MAGIC_64)
2486 outs() << "MH_MAGIC_64";
2488 outs() << format(" 0x%08" PRIx32, magic);
2490 case MachO::CPU_TYPE_I386:
2492 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
2493 case MachO::CPU_SUBTYPE_I386_ALL:
2497 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
2501 case MachO::CPU_TYPE_X86_64:
2502 outs() << " X86_64";
2503 case MachO::CPU_SUBTYPE_X86_64_ALL:
2506 case MachO::CPU_SUBTYPE_X86_64_H:
2507 outs() << " Haswell";
2508 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
2510 case MachO::CPU_TYPE_ARM:
2512 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
2513 case MachO::CPU_SUBTYPE_ARM_ALL:
2516 case MachO::CPU_SUBTYPE_ARM_V4T:
2519 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
2522 case MachO::CPU_SUBTYPE_ARM_XSCALE:
2523 outs() << " XSCALE";
2525 case MachO::CPU_SUBTYPE_ARM_V6:
2528 case MachO::CPU_SUBTYPE_ARM_V6M:
2531 case MachO::CPU_SUBTYPE_ARM_V7:
2534 case MachO::CPU_SUBTYPE_ARM_V7EM:
2537 case MachO::CPU_SUBTYPE_ARM_V7K:
2540 case MachO::CPU_SUBTYPE_ARM_V7M:
2543 case MachO::CPU_SUBTYPE_ARM_V7S:
2547 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
2551 case MachO::CPU_TYPE_ARM64:
2553 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
2554 case MachO::CPU_SUBTYPE_ARM64_ALL:
2558 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
2562 case MachO::CPU_TYPE_POWERPC:
2564 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
2565 case MachO::CPU_SUBTYPE_POWERPC_ALL:
2569 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
2573 case MachO::CPU_TYPE_POWERPC64:
2575 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
2576 case MachO::CPU_SUBTYPE_POWERPC_ALL:
2580 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
2585 if ((cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64) {
2588 outs() << format(" 0x%02" PRIx32,
2589 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
2592 case MachO::MH_OBJECT:
2593 outs() << " OBJECT";
2595 case MachO::MH_EXECUTE:
2596 outs() << " EXECUTE";
2598 case MachO::MH_FVMLIB:
2599 outs() << " FVMLIB";
2601 case MachO::MH_CORE:
2604 case MachO::MH_PRELOAD:
2605 outs() << " PRELOAD";
2607 case MachO::MH_DYLIB:
2610 case MachO::MH_DYLIB_STUB:
2611 outs() << " DYLIB_STUB";
2613 case MachO::MH_DYLINKER:
2614 outs() << " DYLINKER";
2616 case MachO::MH_BUNDLE:
2617 outs() << " BUNDLE";
2619 case MachO::MH_DSYM:
2622 case MachO::MH_KEXT_BUNDLE:
2623 outs() << " KEXTBUNDLE";
2626 outs() << format(" %10u", filetype);
2629 outs() << format(" %5u", ncmds);
2630 outs() << format(" %10u", sizeofcmds);
2632 if (f & MachO::MH_NOUNDEFS) {
2633 outs() << " NOUNDEFS";
2634 f &= ~MachO::MH_NOUNDEFS;
2636 if (f & MachO::MH_INCRLINK) {
2637 outs() << " INCRLINK";
2638 f &= ~MachO::MH_INCRLINK;
2640 if (f & MachO::MH_DYLDLINK) {
2641 outs() << " DYLDLINK";
2642 f &= ~MachO::MH_DYLDLINK;
2644 if (f & MachO::MH_BINDATLOAD) {
2645 outs() << " BINDATLOAD";
2646 f &= ~MachO::MH_BINDATLOAD;
2648 if (f & MachO::MH_PREBOUND) {
2649 outs() << " PREBOUND";
2650 f &= ~MachO::MH_PREBOUND;
2652 if (f & MachO::MH_SPLIT_SEGS) {
2653 outs() << " SPLIT_SEGS";
2654 f &= ~MachO::MH_SPLIT_SEGS;
2656 if (f & MachO::MH_LAZY_INIT) {
2657 outs() << " LAZY_INIT";
2658 f &= ~MachO::MH_LAZY_INIT;
2660 if (f & MachO::MH_TWOLEVEL) {
2661 outs() << " TWOLEVEL";
2662 f &= ~MachO::MH_TWOLEVEL;
2664 if (f & MachO::MH_FORCE_FLAT) {
2665 outs() << " FORCE_FLAT";
2666 f &= ~MachO::MH_FORCE_FLAT;
2668 if (f & MachO::MH_NOMULTIDEFS) {
2669 outs() << " NOMULTIDEFS";
2670 f &= ~MachO::MH_NOMULTIDEFS;
2672 if (f & MachO::MH_NOFIXPREBINDING) {
2673 outs() << " NOFIXPREBINDING";
2674 f &= ~MachO::MH_NOFIXPREBINDING;
2676 if (f & MachO::MH_PREBINDABLE) {
2677 outs() << " PREBINDABLE";
2678 f &= ~MachO::MH_PREBINDABLE;
2680 if (f & MachO::MH_ALLMODSBOUND) {
2681 outs() << " ALLMODSBOUND";
2682 f &= ~MachO::MH_ALLMODSBOUND;
2684 if (f & MachO::MH_SUBSECTIONS_VIA_SYMBOLS) {
2685 outs() << " SUBSECTIONS_VIA_SYMBOLS";
2686 f &= ~MachO::MH_SUBSECTIONS_VIA_SYMBOLS;
2688 if (f & MachO::MH_CANONICAL) {
2689 outs() << " CANONICAL";
2690 f &= ~MachO::MH_CANONICAL;
2692 if (f & MachO::MH_WEAK_DEFINES) {
2693 outs() << " WEAK_DEFINES";
2694 f &= ~MachO::MH_WEAK_DEFINES;
2696 if (f & MachO::MH_BINDS_TO_WEAK) {
2697 outs() << " BINDS_TO_WEAK";
2698 f &= ~MachO::MH_BINDS_TO_WEAK;
2700 if (f & MachO::MH_ALLOW_STACK_EXECUTION) {
2701 outs() << " ALLOW_STACK_EXECUTION";
2702 f &= ~MachO::MH_ALLOW_STACK_EXECUTION;
2704 if (f & MachO::MH_DEAD_STRIPPABLE_DYLIB) {
2705 outs() << " DEAD_STRIPPABLE_DYLIB";
2706 f &= ~MachO::MH_DEAD_STRIPPABLE_DYLIB;
2708 if (f & MachO::MH_PIE) {
2710 f &= ~MachO::MH_PIE;
2712 if (f & MachO::MH_NO_REEXPORTED_DYLIBS) {
2713 outs() << " NO_REEXPORTED_DYLIBS";
2714 f &= ~MachO::MH_NO_REEXPORTED_DYLIBS;
2716 if (f & MachO::MH_HAS_TLV_DESCRIPTORS) {
2717 outs() << " MH_HAS_TLV_DESCRIPTORS";
2718 f &= ~MachO::MH_HAS_TLV_DESCRIPTORS;
2720 if (f & MachO::MH_NO_HEAP_EXECUTION) {
2721 outs() << " MH_NO_HEAP_EXECUTION";
2722 f &= ~MachO::MH_NO_HEAP_EXECUTION;
2724 if (f & MachO::MH_APP_EXTENSION_SAFE) {
2725 outs() << " APP_EXTENSION_SAFE";
2726 f &= ~MachO::MH_APP_EXTENSION_SAFE;
2728 if (f != 0 || flags == 0)
2729 outs() << format(" 0x%08" PRIx32, f);
2731 outs() << format(" 0x%08" PRIx32, magic);
2732 outs() << format(" %7d", cputype);
2733 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
2734 outs() << format(" 0x%02" PRIx32,
2735 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
2736 outs() << format(" %10u", filetype);
2737 outs() << format(" %5u", ncmds);
2738 outs() << format(" %10u", sizeofcmds);
2739 outs() << format(" 0x%08" PRIx32, flags);
2744 static void PrintSegmentCommand(uint32_t cmd, uint32_t cmdsize,
2745 StringRef SegName, uint64_t vmaddr,
2746 uint64_t vmsize, uint64_t fileoff,
2747 uint64_t filesize, uint32_t maxprot,
2748 uint32_t initprot, uint32_t nsects,
2749 uint32_t flags, uint32_t object_size,
2751 uint64_t expected_cmdsize;
2752 if (cmd == MachO::LC_SEGMENT) {
2753 outs() << " cmd LC_SEGMENT\n";
2754 expected_cmdsize = nsects;
2755 expected_cmdsize *= sizeof(struct MachO::section);
2756 expected_cmdsize += sizeof(struct MachO::segment_command);
2758 outs() << " cmd LC_SEGMENT_64\n";
2759 expected_cmdsize = nsects;
2760 expected_cmdsize *= sizeof(struct MachO::section_64);
2761 expected_cmdsize += sizeof(struct MachO::segment_command_64);
2763 outs() << " cmdsize " << cmdsize;
2764 if (cmdsize != expected_cmdsize)
2765 outs() << " Inconsistent size\n";
2768 outs() << " segname " << SegName << "\n";
2769 if (cmd == MachO::LC_SEGMENT_64) {
2770 outs() << " vmaddr " << format("0x%016" PRIx64, vmaddr) << "\n";
2771 outs() << " vmsize " << format("0x%016" PRIx64, vmsize) << "\n";
2773 outs() << " vmaddr " << format("0x%08" PRIx32, vmaddr) << "\n";
2774 outs() << " vmsize " << format("0x%08" PRIx32, vmsize) << "\n";
2776 outs() << " fileoff " << fileoff;
2777 if (fileoff > object_size)
2778 outs() << " (past end of file)\n";
2781 outs() << " filesize " << filesize;
2782 if (fileoff + filesize > object_size)
2783 outs() << " (past end of file)\n";
2788 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
2789 MachO::VM_PROT_EXECUTE)) != 0)
2790 outs() << " maxprot ?" << format("0x%08" PRIx32, maxprot) << "\n";
2792 if (maxprot & MachO::VM_PROT_READ)
2793 outs() << " maxprot r";
2795 outs() << " maxprot -";
2796 if (maxprot & MachO::VM_PROT_WRITE)
2800 if (maxprot & MachO::VM_PROT_EXECUTE)
2806 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
2807 MachO::VM_PROT_EXECUTE)) != 0)
2808 outs() << " initprot ?" << format("0x%08" PRIx32, initprot) << "\n";
2810 if (initprot & MachO::VM_PROT_READ)
2811 outs() << " initprot r";
2813 outs() << " initprot -";
2814 if (initprot & MachO::VM_PROT_WRITE)
2818 if (initprot & MachO::VM_PROT_EXECUTE)
2824 outs() << " maxprot " << format("0x%08" PRIx32, maxprot) << "\n";
2825 outs() << " initprot " << format("0x%08" PRIx32, initprot) << "\n";
2827 outs() << " nsects " << nsects << "\n";
2831 outs() << " (none)\n";
2833 if (flags & MachO::SG_HIGHVM) {
2834 outs() << " HIGHVM";
2835 flags &= ~MachO::SG_HIGHVM;
2837 if (flags & MachO::SG_FVMLIB) {
2838 outs() << " FVMLIB";
2839 flags &= ~MachO::SG_FVMLIB;
2841 if (flags & MachO::SG_NORELOC) {
2842 outs() << " NORELOC";
2843 flags &= ~MachO::SG_NORELOC;
2845 if (flags & MachO::SG_PROTECTED_VERSION_1) {
2846 outs() << " PROTECTED_VERSION_1";
2847 flags &= ~MachO::SG_PROTECTED_VERSION_1;
2850 outs() << format(" 0x%08" PRIx32, flags) << " (unknown flags)\n";
2855 outs() << " flags " << format("0x%" PRIx32, flags) << "\n";
2859 static void PrintSection(const char *sectname, const char *segname,
2860 uint64_t addr, uint64_t size, uint32_t offset,
2861 uint32_t align, uint32_t reloff, uint32_t nreloc,
2862 uint32_t flags, uint32_t reserved1, uint32_t reserved2,
2863 uint32_t cmd, const char *sg_segname,
2864 uint32_t filetype, uint32_t object_size,
2866 outs() << "Section\n";
2867 outs() << " sectname " << format("%.16s\n", sectname);
2868 outs() << " segname " << format("%.16s", segname);
2869 if (filetype != MachO::MH_OBJECT && strncmp(sg_segname, segname, 16) != 0)
2870 outs() << " (does not match segment)\n";
2873 if (cmd == MachO::LC_SEGMENT_64) {
2874 outs() << " addr " << format("0x%016" PRIx64, addr) << "\n";
2875 outs() << " size " << format("0x%016" PRIx64, size);
2877 outs() << " addr " << format("0x%08" PRIx32, addr) << "\n";
2878 outs() << " size " << format("0x%08" PRIx32, size);
2880 if ((flags & MachO::S_ZEROFILL) != 0 && offset + size > object_size)
2881 outs() << " (past end of file)\n";
2884 outs() << " offset " << offset;
2885 if (offset > object_size)
2886 outs() << " (past end of file)\n";
2889 uint32_t align_shifted = 1 << align;
2890 outs() << " align 2^" << align << " (" << align_shifted << ")\n";
2891 outs() << " reloff " << reloff;
2892 if (reloff > object_size)
2893 outs() << " (past end of file)\n";
2896 outs() << " nreloc " << nreloc;
2897 if (reloff + nreloc * sizeof(struct MachO::relocation_info) > object_size)
2898 outs() << " (past end of file)\n";
2901 uint32_t section_type = flags & MachO::SECTION_TYPE;
2904 if (section_type == MachO::S_REGULAR)
2905 outs() << " S_REGULAR\n";
2906 else if (section_type == MachO::S_ZEROFILL)
2907 outs() << " S_ZEROFILL\n";
2908 else if (section_type == MachO::S_CSTRING_LITERALS)
2909 outs() << " S_CSTRING_LITERALS\n";
2910 else if (section_type == MachO::S_4BYTE_LITERALS)
2911 outs() << " S_4BYTE_LITERALS\n";
2912 else if (section_type == MachO::S_8BYTE_LITERALS)
2913 outs() << " S_8BYTE_LITERALS\n";
2914 else if (section_type == MachO::S_16BYTE_LITERALS)
2915 outs() << " S_16BYTE_LITERALS\n";
2916 else if (section_type == MachO::S_LITERAL_POINTERS)
2917 outs() << " S_LITERAL_POINTERS\n";
2918 else if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS)
2919 outs() << " S_NON_LAZY_SYMBOL_POINTERS\n";
2920 else if (section_type == MachO::S_LAZY_SYMBOL_POINTERS)
2921 outs() << " S_LAZY_SYMBOL_POINTERS\n";
2922 else if (section_type == MachO::S_SYMBOL_STUBS)
2923 outs() << " S_SYMBOL_STUBS\n";
2924 else if (section_type == MachO::S_MOD_INIT_FUNC_POINTERS)
2925 outs() << " S_MOD_INIT_FUNC_POINTERS\n";
2926 else if (section_type == MachO::S_MOD_TERM_FUNC_POINTERS)
2927 outs() << " S_MOD_TERM_FUNC_POINTERS\n";
2928 else if (section_type == MachO::S_COALESCED)
2929 outs() << " S_COALESCED\n";
2930 else if (section_type == MachO::S_INTERPOSING)
2931 outs() << " S_INTERPOSING\n";
2932 else if (section_type == MachO::S_DTRACE_DOF)
2933 outs() << " S_DTRACE_DOF\n";
2934 else if (section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS)
2935 outs() << " S_LAZY_DYLIB_SYMBOL_POINTERS\n";
2936 else if (section_type == MachO::S_THREAD_LOCAL_REGULAR)
2937 outs() << " S_THREAD_LOCAL_REGULAR\n";
2938 else if (section_type == MachO::S_THREAD_LOCAL_ZEROFILL)
2939 outs() << " S_THREAD_LOCAL_ZEROFILL\n";
2940 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLES)
2941 outs() << " S_THREAD_LOCAL_VARIABLES\n";
2942 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
2943 outs() << " S_THREAD_LOCAL_VARIABLE_POINTERS\n";
2944 else if (section_type == MachO::S_THREAD_LOCAL_INIT_FUNCTION_POINTERS)
2945 outs() << " S_THREAD_LOCAL_INIT_FUNCTION_POINTERS\n";
2947 outs() << format("0x%08" PRIx32, section_type) << "\n";
2948 outs() << "attributes";
2949 uint32_t section_attributes = flags & MachO::SECTION_ATTRIBUTES;
2950 if (section_attributes & MachO::S_ATTR_PURE_INSTRUCTIONS)
2951 outs() << " PURE_INSTRUCTIONS";
2952 if (section_attributes & MachO::S_ATTR_NO_TOC)
2953 outs() << " NO_TOC";
2954 if (section_attributes & MachO::S_ATTR_STRIP_STATIC_SYMS)
2955 outs() << " STRIP_STATIC_SYMS";
2956 if (section_attributes & MachO::S_ATTR_NO_DEAD_STRIP)
2957 outs() << " NO_DEAD_STRIP";
2958 if (section_attributes & MachO::S_ATTR_LIVE_SUPPORT)
2959 outs() << " LIVE_SUPPORT";
2960 if (section_attributes & MachO::S_ATTR_SELF_MODIFYING_CODE)
2961 outs() << " SELF_MODIFYING_CODE";
2962 if (section_attributes & MachO::S_ATTR_DEBUG)
2964 if (section_attributes & MachO::S_ATTR_SOME_INSTRUCTIONS)
2965 outs() << " SOME_INSTRUCTIONS";
2966 if (section_attributes & MachO::S_ATTR_EXT_RELOC)
2967 outs() << " EXT_RELOC";
2968 if (section_attributes & MachO::S_ATTR_LOC_RELOC)
2969 outs() << " LOC_RELOC";
2970 if (section_attributes == 0)
2971 outs() << " (none)";
2974 outs() << " flags " << format("0x%08" PRIx32, flags) << "\n";
2975 outs() << " reserved1 " << reserved1;
2976 if (section_type == MachO::S_SYMBOL_STUBS ||
2977 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
2978 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
2979 section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
2980 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
2981 outs() << " (index into indirect symbol table)\n";
2984 outs() << " reserved2 " << reserved2;
2985 if (section_type == MachO::S_SYMBOL_STUBS)
2986 outs() << " (size of stubs)\n";
2991 static void PrintSymtabLoadCommand(MachO::symtab_command st, bool Is64Bit,
2992 uint32_t object_size) {
2993 outs() << " cmd LC_SYMTAB\n";
2994 outs() << " cmdsize " << st.cmdsize;
2995 if (st.cmdsize != sizeof(struct MachO::symtab_command))
2996 outs() << " Incorrect size\n";
2999 outs() << " symoff " << st.symoff;
3000 if (st.symoff > object_size)
3001 outs() << " (past end of file)\n";
3004 outs() << " nsyms " << st.nsyms;
3007 big_size = st.nsyms;
3008 big_size *= sizeof(struct MachO::nlist_64);
3009 big_size += st.symoff;
3010 if (big_size > object_size)
3011 outs() << " (past end of file)\n";
3015 big_size = st.nsyms;
3016 big_size *= sizeof(struct MachO::nlist);
3017 big_size += st.symoff;
3018 if (big_size > object_size)
3019 outs() << " (past end of file)\n";
3023 outs() << " stroff " << st.stroff;
3024 if (st.stroff > object_size)
3025 outs() << " (past end of file)\n";
3028 outs() << " strsize " << st.strsize;
3029 big_size = st.stroff;
3030 big_size += st.strsize;
3031 if (big_size > object_size)
3032 outs() << " (past end of file)\n";
3037 static void PrintDysymtabLoadCommand(MachO::dysymtab_command dyst,
3038 uint32_t nsyms, uint32_t object_size,
3040 outs() << " cmd LC_DYSYMTAB\n";
3041 outs() << " cmdsize " << dyst.cmdsize;
3042 if (dyst.cmdsize != sizeof(struct MachO::dysymtab_command))
3043 outs() << " Incorrect size\n";
3046 outs() << " ilocalsym " << dyst.ilocalsym;
3047 if (dyst.ilocalsym > nsyms)
3048 outs() << " (greater than the number of symbols)\n";
3051 outs() << " nlocalsym " << dyst.nlocalsym;
3053 big_size = dyst.ilocalsym;
3054 big_size += dyst.nlocalsym;
3055 if (big_size > nsyms)
3056 outs() << " (past the end of the symbol table)\n";
3059 outs() << " iextdefsym " << dyst.iextdefsym;
3060 if (dyst.iextdefsym > nsyms)
3061 outs() << " (greater than the number of symbols)\n";
3064 outs() << " nextdefsym " << dyst.nextdefsym;
3065 big_size = dyst.iextdefsym;
3066 big_size += dyst.nextdefsym;
3067 if (big_size > nsyms)
3068 outs() << " (past the end of the symbol table)\n";
3071 outs() << " iundefsym " << dyst.iundefsym;
3072 if (dyst.iundefsym > nsyms)
3073 outs() << " (greater than the number of symbols)\n";
3076 outs() << " nundefsym " << dyst.nundefsym;
3077 big_size = dyst.iundefsym;
3078 big_size += dyst.nundefsym;
3079 if (big_size > nsyms)
3080 outs() << " (past the end of the symbol table)\n";
3083 outs() << " tocoff " << dyst.tocoff;
3084 if (dyst.tocoff > object_size)
3085 outs() << " (past end of file)\n";
3088 outs() << " ntoc " << dyst.ntoc;
3089 big_size = dyst.ntoc;
3090 big_size *= sizeof(struct MachO::dylib_table_of_contents);
3091 big_size += dyst.tocoff;
3092 if (big_size > object_size)
3093 outs() << " (past end of file)\n";
3096 outs() << " modtaboff " << dyst.modtaboff;
3097 if (dyst.modtaboff > object_size)
3098 outs() << " (past end of file)\n";
3101 outs() << " nmodtab " << dyst.nmodtab;
3104 modtabend = dyst.nmodtab;
3105 modtabend *= sizeof(struct MachO::dylib_module_64);
3106 modtabend += dyst.modtaboff;
3108 modtabend = dyst.nmodtab;
3109 modtabend *= sizeof(struct MachO::dylib_module);
3110 modtabend += dyst.modtaboff;
3112 if (modtabend > object_size)
3113 outs() << " (past end of file)\n";
3116 outs() << " extrefsymoff " << dyst.extrefsymoff;
3117 if (dyst.extrefsymoff > object_size)
3118 outs() << " (past end of file)\n";
3121 outs() << " nextrefsyms " << dyst.nextrefsyms;
3122 big_size = dyst.nextrefsyms;
3123 big_size *= sizeof(struct MachO::dylib_reference);
3124 big_size += dyst.extrefsymoff;
3125 if (big_size > object_size)
3126 outs() << " (past end of file)\n";
3129 outs() << " indirectsymoff " << dyst.indirectsymoff;
3130 if (dyst.indirectsymoff > object_size)
3131 outs() << " (past end of file)\n";
3134 outs() << " nindirectsyms " << dyst.nindirectsyms;
3135 big_size = dyst.nindirectsyms;
3136 big_size *= sizeof(uint32_t);
3137 big_size += dyst.indirectsymoff;
3138 if (big_size > object_size)
3139 outs() << " (past end of file)\n";
3142 outs() << " extreloff " << dyst.extreloff;
3143 if (dyst.extreloff > object_size)
3144 outs() << " (past end of file)\n";
3147 outs() << " nextrel " << dyst.nextrel;
3148 big_size = dyst.nextrel;
3149 big_size *= sizeof(struct MachO::relocation_info);
3150 big_size += dyst.extreloff;
3151 if (big_size > object_size)
3152 outs() << " (past end of file)\n";
3155 outs() << " locreloff " << dyst.locreloff;
3156 if (dyst.locreloff > object_size)
3157 outs() << " (past end of file)\n";
3160 outs() << " nlocrel " << dyst.nlocrel;
3161 big_size = dyst.nlocrel;
3162 big_size *= sizeof(struct MachO::relocation_info);
3163 big_size += dyst.locreloff;
3164 if (big_size > object_size)
3165 outs() << " (past end of file)\n";
3170 static void PrintDyldInfoLoadCommand(MachO::dyld_info_command dc,
3171 uint32_t object_size) {
3172 if (dc.cmd == MachO::LC_DYLD_INFO)
3173 outs() << " cmd LC_DYLD_INFO\n";
3175 outs() << " cmd LC_DYLD_INFO_ONLY\n";
3176 outs() << " cmdsize " << dc.cmdsize;
3177 if (dc.cmdsize != sizeof(struct MachO::dyld_info_command))
3178 outs() << " Incorrect size\n";
3181 outs() << " rebase_off " << dc.rebase_off;
3182 if (dc.rebase_off > object_size)
3183 outs() << " (past end of file)\n";
3186 outs() << " rebase_size " << dc.rebase_size;
3188 big_size = dc.rebase_off;
3189 big_size += dc.rebase_size;
3190 if (big_size > object_size)
3191 outs() << " (past end of file)\n";
3194 outs() << " bind_off " << dc.bind_off;
3195 if (dc.bind_off > object_size)
3196 outs() << " (past end of file)\n";
3199 outs() << " bind_size " << dc.bind_size;
3200 big_size = dc.bind_off;
3201 big_size += dc.bind_size;
3202 if (big_size > object_size)
3203 outs() << " (past end of file)\n";
3206 outs() << " weak_bind_off " << dc.weak_bind_off;
3207 if (dc.weak_bind_off > object_size)
3208 outs() << " (past end of file)\n";
3211 outs() << " weak_bind_size " << dc.weak_bind_size;
3212 big_size = dc.weak_bind_off;
3213 big_size += dc.weak_bind_size;
3214 if (big_size > object_size)
3215 outs() << " (past end of file)\n";
3218 outs() << " lazy_bind_off " << dc.lazy_bind_off;
3219 if (dc.lazy_bind_off > object_size)
3220 outs() << " (past end of file)\n";
3223 outs() << " lazy_bind_size " << dc.lazy_bind_size;
3224 big_size = dc.lazy_bind_off;
3225 big_size += dc.lazy_bind_size;
3226 if (big_size > object_size)
3227 outs() << " (past end of file)\n";
3230 outs() << " export_off " << dc.export_off;
3231 if (dc.export_off > object_size)
3232 outs() << " (past end of file)\n";
3235 outs() << " export_size " << dc.export_size;
3236 big_size = dc.export_off;
3237 big_size += dc.export_size;
3238 if (big_size > object_size)
3239 outs() << " (past end of file)\n";
3244 static void PrintDyldLoadCommand(MachO::dylinker_command dyld,
3246 if (dyld.cmd == MachO::LC_ID_DYLINKER)
3247 outs() << " cmd LC_ID_DYLINKER\n";
3248 else if (dyld.cmd == MachO::LC_LOAD_DYLINKER)
3249 outs() << " cmd LC_LOAD_DYLINKER\n";
3250 else if (dyld.cmd == MachO::LC_DYLD_ENVIRONMENT)
3251 outs() << " cmd LC_DYLD_ENVIRONMENT\n";
3253 outs() << " cmd ?(" << dyld.cmd << ")\n";
3254 outs() << " cmdsize " << dyld.cmdsize;
3255 if (dyld.cmdsize < sizeof(struct MachO::dylinker_command))
3256 outs() << " Incorrect size\n";
3259 if (dyld.name >= dyld.cmdsize)
3260 outs() << " name ?(bad offset " << dyld.name << ")\n";
3262 const char *P = (const char *)(Ptr) + dyld.name;
3263 outs() << " name " << P << " (offset " << dyld.name << ")\n";
3267 static void PrintUuidLoadCommand(MachO::uuid_command uuid) {
3268 outs() << " cmd LC_UUID\n";
3269 outs() << " cmdsize " << uuid.cmdsize;
3270 if (uuid.cmdsize != sizeof(struct MachO::uuid_command))
3271 outs() << " Incorrect size\n";
3275 outs() << format("%02" PRIX32, uuid.uuid[0]);
3276 outs() << format("%02" PRIX32, uuid.uuid[1]);
3277 outs() << format("%02" PRIX32, uuid.uuid[2]);
3278 outs() << format("%02" PRIX32, uuid.uuid[3]);
3280 outs() << format("%02" PRIX32, uuid.uuid[4]);
3281 outs() << format("%02" PRIX32, uuid.uuid[5]);
3283 outs() << format("%02" PRIX32, uuid.uuid[6]);
3284 outs() << format("%02" PRIX32, uuid.uuid[7]);
3286 outs() << format("%02" PRIX32, uuid.uuid[8]);
3287 outs() << format("%02" PRIX32, uuid.uuid[9]);
3289 outs() << format("%02" PRIX32, uuid.uuid[10]);
3290 outs() << format("%02" PRIX32, uuid.uuid[11]);
3291 outs() << format("%02" PRIX32, uuid.uuid[12]);
3292 outs() << format("%02" PRIX32, uuid.uuid[13]);
3293 outs() << format("%02" PRIX32, uuid.uuid[14]);
3294 outs() << format("%02" PRIX32, uuid.uuid[15]);
3298 static void PrintVersionMinLoadCommand(MachO::version_min_command vd) {
3299 if (vd.cmd == MachO::LC_VERSION_MIN_MACOSX)
3300 outs() << " cmd LC_VERSION_MIN_MACOSX\n";
3301 else if (vd.cmd == MachO::LC_VERSION_MIN_IPHONEOS)
3302 outs() << " cmd LC_VERSION_MIN_IPHONEOS\n";
3304 outs() << " cmd " << vd.cmd << " (?)\n";
3305 outs() << " cmdsize " << vd.cmdsize;
3306 if (vd.cmdsize != sizeof(struct MachO::version_min_command))
3307 outs() << " Incorrect size\n";
3310 outs() << " version " << ((vd.version >> 16) & 0xffff) << "."
3311 << ((vd.version >> 8) & 0xff);
3312 if ((vd.version & 0xff) != 0)
3313 outs() << "." << (vd.version & 0xff);
3316 outs() << " sdk n/a\n";
3318 outs() << " sdk " << ((vd.sdk >> 16) & 0xffff) << "."
3319 << ((vd.sdk >> 8) & 0xff);
3321 if ((vd.sdk & 0xff) != 0)
3322 outs() << "." << (vd.sdk & 0xff);
3326 static void PrintSourceVersionCommand(MachO::source_version_command sd) {
3327 outs() << " cmd LC_SOURCE_VERSION\n";
3328 outs() << " cmdsize " << sd.cmdsize;
3329 if (sd.cmdsize != sizeof(struct MachO::source_version_command))
3330 outs() << " Incorrect size\n";
3333 uint64_t a = (sd.version >> 40) & 0xffffff;
3334 uint64_t b = (sd.version >> 30) & 0x3ff;
3335 uint64_t c = (sd.version >> 20) & 0x3ff;
3336 uint64_t d = (sd.version >> 10) & 0x3ff;
3337 uint64_t e = sd.version & 0x3ff;
3338 outs() << " version " << a << "." << b;
3340 outs() << "." << c << "." << d << "." << e;
3342 outs() << "." << c << "." << d;
3348 static void PrintEntryPointCommand(MachO::entry_point_command ep) {
3349 outs() << " cmd LC_MAIN\n";
3350 outs() << " cmdsize " << ep.cmdsize;
3351 if (ep.cmdsize != sizeof(struct MachO::entry_point_command))
3352 outs() << " Incorrect size\n";
3355 outs() << " entryoff " << ep.entryoff << "\n";
3356 outs() << " stacksize " << ep.stacksize << "\n";
3359 static void PrintDylibCommand(MachO::dylib_command dl, const char *Ptr) {
3360 if (dl.cmd == MachO::LC_ID_DYLIB)
3361 outs() << " cmd LC_ID_DYLIB\n";
3362 else if (dl.cmd == MachO::LC_LOAD_DYLIB)
3363 outs() << " cmd LC_LOAD_DYLIB\n";
3364 else if (dl.cmd == MachO::LC_LOAD_WEAK_DYLIB)
3365 outs() << " cmd LC_LOAD_WEAK_DYLIB\n";
3366 else if (dl.cmd == MachO::LC_REEXPORT_DYLIB)
3367 outs() << " cmd LC_REEXPORT_DYLIB\n";
3368 else if (dl.cmd == MachO::LC_LAZY_LOAD_DYLIB)
3369 outs() << " cmd LC_LAZY_LOAD_DYLIB\n";
3370 else if (dl.cmd == MachO::LC_LOAD_UPWARD_DYLIB)
3371 outs() << " cmd LC_LOAD_UPWARD_DYLIB\n";
3373 outs() << " cmd " << dl.cmd << " (unknown)\n";
3374 outs() << " cmdsize " << dl.cmdsize;
3375 if (dl.cmdsize < sizeof(struct MachO::dylib_command))
3376 outs() << " Incorrect size\n";
3379 if (dl.dylib.name < dl.cmdsize) {
3380 const char *P = (const char *)(Ptr) + dl.dylib.name;
3381 outs() << " name " << P << " (offset " << dl.dylib.name << ")\n";
3383 outs() << " name ?(bad offset " << dl.dylib.name << ")\n";
3385 outs() << " time stamp " << dl.dylib.timestamp << " ";
3386 time_t t = dl.dylib.timestamp;
3387 outs() << ctime(&t);
3388 outs() << " current version ";
3389 if (dl.dylib.current_version == 0xffffffff)
3392 outs() << ((dl.dylib.current_version >> 16) & 0xffff) << "."
3393 << ((dl.dylib.current_version >> 8) & 0xff) << "."
3394 << (dl.dylib.current_version & 0xff) << "\n";
3395 outs() << "compatibility version ";
3396 if (dl.dylib.compatibility_version == 0xffffffff)
3399 outs() << ((dl.dylib.compatibility_version >> 16) & 0xffff) << "."
3400 << ((dl.dylib.compatibility_version >> 8) & 0xff) << "."
3401 << (dl.dylib.compatibility_version & 0xff) << "\n";
3404 static void PrintLinkEditDataCommand(MachO::linkedit_data_command ld,
3405 uint32_t object_size) {
3406 if (ld.cmd == MachO::LC_CODE_SIGNATURE)
3407 outs() << " cmd LC_FUNCTION_STARTS\n";
3408 else if (ld.cmd == MachO::LC_SEGMENT_SPLIT_INFO)
3409 outs() << " cmd LC_SEGMENT_SPLIT_INFO\n";
3410 else if (ld.cmd == MachO::LC_FUNCTION_STARTS)
3411 outs() << " cmd LC_FUNCTION_STARTS\n";
3412 else if (ld.cmd == MachO::LC_DATA_IN_CODE)
3413 outs() << " cmd LC_DATA_IN_CODE\n";
3414 else if (ld.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS)
3415 outs() << " cmd LC_DYLIB_CODE_SIGN_DRS\n";
3416 else if (ld.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT)
3417 outs() << " cmd LC_LINKER_OPTIMIZATION_HINT\n";
3419 outs() << " cmd " << ld.cmd << " (?)\n";
3420 outs() << " cmdsize " << ld.cmdsize;
3421 if (ld.cmdsize != sizeof(struct MachO::linkedit_data_command))
3422 outs() << " Incorrect size\n";
3425 outs() << " dataoff " << ld.dataoff;
3426 if (ld.dataoff > object_size)
3427 outs() << " (past end of file)\n";
3430 outs() << " datasize " << ld.datasize;
3431 uint64_t big_size = ld.dataoff;
3432 big_size += ld.datasize;
3433 if (big_size > object_size)
3434 outs() << " (past end of file)\n";
3439 static void PrintLoadCommands(const MachOObjectFile *Obj, uint32_t ncmds,
3440 uint32_t filetype, uint32_t cputype,
3442 StringRef Buf = Obj->getData();
3443 MachOObjectFile::LoadCommandInfo Command = Obj->getFirstLoadCommandInfo();
3444 for (unsigned i = 0;; ++i) {
3445 outs() << "Load command " << i << "\n";
3446 if (Command.C.cmd == MachO::LC_SEGMENT) {
3447 MachO::segment_command SLC = Obj->getSegmentLoadCommand(Command);
3448 const char *sg_segname = SLC.segname;
3449 PrintSegmentCommand(SLC.cmd, SLC.cmdsize, SLC.segname, SLC.vmaddr,
3450 SLC.vmsize, SLC.fileoff, SLC.filesize, SLC.maxprot,
3451 SLC.initprot, SLC.nsects, SLC.flags, Buf.size(),
3453 for (unsigned j = 0; j < SLC.nsects; j++) {
3454 MachO::section_64 S = Obj->getSection64(Command, j);
3455 PrintSection(S.sectname, S.segname, S.addr, S.size, S.offset, S.align,
3456 S.reloff, S.nreloc, S.flags, S.reserved1, S.reserved2,
3457 SLC.cmd, sg_segname, filetype, Buf.size(), verbose);
3459 } else if (Command.C.cmd == MachO::LC_SEGMENT_64) {
3460 MachO::segment_command_64 SLC_64 = Obj->getSegment64LoadCommand(Command);
3461 const char *sg_segname = SLC_64.segname;
3462 PrintSegmentCommand(SLC_64.cmd, SLC_64.cmdsize, SLC_64.segname,
3463 SLC_64.vmaddr, SLC_64.vmsize, SLC_64.fileoff,
3464 SLC_64.filesize, SLC_64.maxprot, SLC_64.initprot,
3465 SLC_64.nsects, SLC_64.flags, Buf.size(), verbose);
3466 for (unsigned j = 0; j < SLC_64.nsects; j++) {
3467 MachO::section_64 S_64 = Obj->getSection64(Command, j);
3468 PrintSection(S_64.sectname, S_64.segname, S_64.addr, S_64.size,
3469 S_64.offset, S_64.align, S_64.reloff, S_64.nreloc,
3470 S_64.flags, S_64.reserved1, S_64.reserved2, SLC_64.cmd,
3471 sg_segname, filetype, Buf.size(), verbose);
3473 } else if (Command.C.cmd == MachO::LC_SYMTAB) {
3474 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
3475 PrintSymtabLoadCommand(Symtab, Obj->is64Bit(), Buf.size());
3476 } else if (Command.C.cmd == MachO::LC_DYSYMTAB) {
3477 MachO::dysymtab_command Dysymtab = Obj->getDysymtabLoadCommand();
3478 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
3479 PrintDysymtabLoadCommand(Dysymtab, Symtab.nsyms, Buf.size(),
3481 } else if (Command.C.cmd == MachO::LC_DYLD_INFO ||
3482 Command.C.cmd == MachO::LC_DYLD_INFO_ONLY) {
3483 MachO::dyld_info_command DyldInfo = Obj->getDyldInfoLoadCommand(Command);
3484 PrintDyldInfoLoadCommand(DyldInfo, Buf.size());
3485 } else if (Command.C.cmd == MachO::LC_LOAD_DYLINKER ||
3486 Command.C.cmd == MachO::LC_ID_DYLINKER ||
3487 Command.C.cmd == MachO::LC_DYLD_ENVIRONMENT) {
3488 MachO::dylinker_command Dyld = Obj->getDylinkerCommand(Command);
3489 PrintDyldLoadCommand(Dyld, Command.Ptr);
3490 } else if (Command.C.cmd == MachO::LC_UUID) {
3491 MachO::uuid_command Uuid = Obj->getUuidCommand(Command);
3492 PrintUuidLoadCommand(Uuid);
3493 } else if (Command.C.cmd == MachO::LC_VERSION_MIN_MACOSX) {
3494 MachO::version_min_command Vd = Obj->getVersionMinLoadCommand(Command);
3495 PrintVersionMinLoadCommand(Vd);
3496 } else if (Command.C.cmd == MachO::LC_SOURCE_VERSION) {
3497 MachO::source_version_command Sd = Obj->getSourceVersionCommand(Command);
3498 PrintSourceVersionCommand(Sd);
3499 } else if (Command.C.cmd == MachO::LC_MAIN) {
3500 MachO::entry_point_command Ep = Obj->getEntryPointCommand(Command);
3501 PrintEntryPointCommand(Ep);
3502 } else if (Command.C.cmd == MachO::LC_LOAD_DYLIB ||
3503 Command.C.cmd == MachO::LC_ID_DYLIB ||
3504 Command.C.cmd == MachO::LC_LOAD_WEAK_DYLIB ||
3505 Command.C.cmd == MachO::LC_REEXPORT_DYLIB ||
3506 Command.C.cmd == MachO::LC_LAZY_LOAD_DYLIB ||
3507 Command.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) {
3508 MachO::dylib_command Dl = Obj->getDylibIDLoadCommand(Command);
3509 PrintDylibCommand(Dl, Command.Ptr);
3510 } else if (Command.C.cmd == MachO::LC_CODE_SIGNATURE ||
3511 Command.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO ||
3512 Command.C.cmd == MachO::LC_FUNCTION_STARTS ||
3513 Command.C.cmd == MachO::LC_DATA_IN_CODE ||
3514 Command.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS ||
3515 Command.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) {
3516 MachO::linkedit_data_command Ld =
3517 Obj->getLinkeditDataLoadCommand(Command);
3518 PrintLinkEditDataCommand(Ld, Buf.size());
3520 outs() << " cmd ?(" << format("0x%08" PRIx32, Command.C.cmd)
3522 outs() << " cmdsize " << Command.C.cmdsize << "\n";
3523 // TODO: get and print the raw bytes of the load command.
3525 // TODO: print all the other kinds of load commands.
3529 Command = Obj->getNextLoadCommandInfo(Command);
3533 static void getAndPrintMachHeader(const MachOObjectFile *Obj, uint32_t &ncmds,
3534 uint32_t &filetype, uint32_t &cputype,
3536 if (Obj->is64Bit()) {
3537 MachO::mach_header_64 H_64;
3538 H_64 = Obj->getHeader64();
3539 PrintMachHeader(H_64.magic, H_64.cputype, H_64.cpusubtype, H_64.filetype,
3540 H_64.ncmds, H_64.sizeofcmds, H_64.flags, verbose);
3542 filetype = H_64.filetype;
3543 cputype = H_64.cputype;
3545 MachO::mach_header H;
3546 H = Obj->getHeader();
3547 PrintMachHeader(H.magic, H.cputype, H.cpusubtype, H.filetype, H.ncmds,
3548 H.sizeofcmds, H.flags, verbose);
3550 filetype = H.filetype;
3551 cputype = H.cputype;
3555 void llvm::printMachOFileHeader(const object::ObjectFile *Obj) {
3556 const MachOObjectFile *file = dyn_cast<const MachOObjectFile>(Obj);
3558 uint32_t filetype = 0;
3559 uint32_t cputype = 0;
3560 getAndPrintMachHeader(file, ncmds, filetype, cputype, true);
3561 PrintLoadCommands(file, ncmds, filetype, cputype, true);
3564 //===----------------------------------------------------------------------===//
3565 // export trie dumping
3566 //===----------------------------------------------------------------------===//
3568 void llvm::printMachOExportsTrie(const object::MachOObjectFile *Obj) {
3569 for (const llvm::object::ExportEntry &Entry : Obj->exports()) {
3570 uint64_t Flags = Entry.flags();
3571 bool ReExport = (Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT);
3572 bool WeakDef = (Flags & MachO::EXPORT_SYMBOL_FLAGS_WEAK_DEFINITION);
3573 bool ThreadLocal = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
3574 MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL);
3575 bool Abs = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
3576 MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE);
3577 bool Resolver = (Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER);
3579 outs() << "[re-export] ";
3581 outs() << format("0x%08llX ",
3582 Entry.address()); // FIXME:add in base address
3583 outs() << Entry.name();
3584 if (WeakDef || ThreadLocal || Resolver || Abs) {
3585 bool NeedsComma = false;
3588 outs() << "weak_def";
3594 outs() << "per-thread";
3600 outs() << "absolute";
3606 outs() << format("resolver=0x%08llX", Entry.other());
3612 StringRef DylibName = "unknown";
3613 int Ordinal = Entry.other() - 1;
3614 Obj->getLibraryShortNameByIndex(Ordinal, DylibName);
3615 if (Entry.otherName().empty())
3616 outs() << " (from " << DylibName << ")";
3618 outs() << " (" << Entry.otherName() << " from " << DylibName << ")";
3624 //===----------------------------------------------------------------------===//
3625 // rebase table dumping
3626 //===----------------------------------------------------------------------===//
3631 SegInfo(const object::MachOObjectFile *Obj);
3633 StringRef segmentName(uint32_t SegIndex);
3634 StringRef sectionName(uint32_t SegIndex, uint64_t SegOffset);
3635 uint64_t address(uint32_t SegIndex, uint64_t SegOffset);
3638 struct SectionInfo {
3641 StringRef SectionName;
3642 StringRef SegmentName;
3643 uint64_t OffsetInSegment;
3644 uint64_t SegmentStartAddress;
3645 uint32_t SegmentIndex;
3647 const SectionInfo &findSection(uint32_t SegIndex, uint64_t SegOffset);
3648 SmallVector<SectionInfo, 32> Sections;
3652 SegInfo::SegInfo(const object::MachOObjectFile *Obj) {
3653 // Build table of sections so segIndex/offset pairs can be translated.
3654 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0;
3655 StringRef CurSegName;
3656 uint64_t CurSegAddress;
3657 for (const SectionRef &Section : Obj->sections()) {
3659 if (error(Section.getName(Info.SectionName)))
3661 Info.Address = Section.getAddress();
3662 Info.Size = Section.getSize();
3664 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl());
3665 if (!Info.SegmentName.equals(CurSegName)) {
3667 CurSegName = Info.SegmentName;
3668 CurSegAddress = Info.Address;
3670 Info.SegmentIndex = CurSegIndex - 1;
3671 Info.OffsetInSegment = Info.Address - CurSegAddress;
3672 Info.SegmentStartAddress = CurSegAddress;
3673 Sections.push_back(Info);
3677 StringRef SegInfo::segmentName(uint32_t SegIndex) {
3678 for (const SectionInfo &SI : Sections) {
3679 if (SI.SegmentIndex == SegIndex)
3680 return SI.SegmentName;
3682 llvm_unreachable("invalid segIndex");
3685 const SegInfo::SectionInfo &SegInfo::findSection(uint32_t SegIndex,
3686 uint64_t OffsetInSeg) {
3687 for (const SectionInfo &SI : Sections) {
3688 if (SI.SegmentIndex != SegIndex)
3690 if (SI.OffsetInSegment > OffsetInSeg)
3692 if (OffsetInSeg >= (SI.OffsetInSegment + SI.Size))
3696 llvm_unreachable("segIndex and offset not in any section");
3699 StringRef SegInfo::sectionName(uint32_t SegIndex, uint64_t OffsetInSeg) {
3700 return findSection(SegIndex, OffsetInSeg).SectionName;
3703 uint64_t SegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) {
3704 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg);
3705 return SI.SegmentStartAddress + OffsetInSeg;
3708 void llvm::printMachORebaseTable(const object::MachOObjectFile *Obj) {
3709 // Build table of sections so names can used in final output.
3710 SegInfo sectionTable(Obj);
3712 outs() << "segment section address type\n";
3713 for (const llvm::object::MachORebaseEntry &Entry : Obj->rebaseTable()) {
3714 uint32_t SegIndex = Entry.segmentIndex();
3715 uint64_t OffsetInSeg = Entry.segmentOffset();
3716 StringRef SegmentName = sectionTable.segmentName(SegIndex);
3717 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
3718 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
3720 // Table lines look like: __DATA __nl_symbol_ptr 0x0000F00C pointer
3721 outs() << format("%-8s %-18s 0x%08" PRIX64 " %s\n",
3722 SegmentName.str().c_str(), SectionName.str().c_str(),
3723 Address, Entry.typeName().str().c_str());
3727 static StringRef ordinalName(const object::MachOObjectFile *Obj, int Ordinal) {
3728 StringRef DylibName;
3730 case MachO::BIND_SPECIAL_DYLIB_SELF:
3731 return "this-image";
3732 case MachO::BIND_SPECIAL_DYLIB_MAIN_EXECUTABLE:
3733 return "main-executable";
3734 case MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP:
3735 return "flat-namespace";
3738 std::error_code EC =
3739 Obj->getLibraryShortNameByIndex(Ordinal - 1, DylibName);
3741 return "<<bad library ordinal>>";
3745 return "<<unknown special ordinal>>";
3748 //===----------------------------------------------------------------------===//
3749 // bind table dumping
3750 //===----------------------------------------------------------------------===//
3752 void llvm::printMachOBindTable(const object::MachOObjectFile *Obj) {
3753 // Build table of sections so names can used in final output.
3754 SegInfo sectionTable(Obj);
3756 outs() << "segment section address type "
3757 "addend dylib symbol\n";
3758 for (const llvm::object::MachOBindEntry &Entry : Obj->bindTable()) {
3759 uint32_t SegIndex = Entry.segmentIndex();
3760 uint64_t OffsetInSeg = Entry.segmentOffset();
3761 StringRef SegmentName = sectionTable.segmentName(SegIndex);
3762 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
3763 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
3765 // Table lines look like:
3766 // __DATA __got 0x00012010 pointer 0 libSystem ___stack_chk_guard
3768 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_WEAK_IMPORT)
3769 Attr = " (weak_import)";
3770 outs() << left_justify(SegmentName, 8) << " "
3771 << left_justify(SectionName, 18) << " "
3772 << format_hex(Address, 10, true) << " "
3773 << left_justify(Entry.typeName(), 8) << " "
3774 << format_decimal(Entry.addend(), 8) << " "
3775 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
3776 << Entry.symbolName() << Attr << "\n";
3780 //===----------------------------------------------------------------------===//
3781 // lazy bind table dumping
3782 //===----------------------------------------------------------------------===//
3784 void llvm::printMachOLazyBindTable(const object::MachOObjectFile *Obj) {
3785 // Build table of sections so names can used in final output.
3786 SegInfo sectionTable(Obj);
3788 outs() << "segment section address "
3790 for (const llvm::object::MachOBindEntry &Entry : Obj->lazyBindTable()) {
3791 uint32_t SegIndex = Entry.segmentIndex();
3792 uint64_t OffsetInSeg = Entry.segmentOffset();
3793 StringRef SegmentName = sectionTable.segmentName(SegIndex);
3794 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
3795 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
3797 // Table lines look like:
3798 // __DATA __got 0x00012010 libSystem ___stack_chk_guard
3799 outs() << left_justify(SegmentName, 8) << " "
3800 << left_justify(SectionName, 18) << " "
3801 << format_hex(Address, 10, true) << " "
3802 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
3803 << Entry.symbolName() << "\n";
3807 //===----------------------------------------------------------------------===//
3808 // weak bind table dumping
3809 //===----------------------------------------------------------------------===//
3811 void llvm::printMachOWeakBindTable(const object::MachOObjectFile *Obj) {
3812 // Build table of sections so names can used in final output.
3813 SegInfo sectionTable(Obj);
3815 outs() << "segment section address "
3816 "type addend symbol\n";
3817 for (const llvm::object::MachOBindEntry &Entry : Obj->weakBindTable()) {
3818 // Strong symbols don't have a location to update.
3819 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) {
3820 outs() << " strong "
3821 << Entry.symbolName() << "\n";
3824 uint32_t SegIndex = Entry.segmentIndex();
3825 uint64_t OffsetInSeg = Entry.segmentOffset();
3826 StringRef SegmentName = sectionTable.segmentName(SegIndex);
3827 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
3828 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
3830 // Table lines look like:
3831 // __DATA __data 0x00001000 pointer 0 _foo
3832 outs() << left_justify(SegmentName, 8) << " "
3833 << left_justify(SectionName, 18) << " "
3834 << format_hex(Address, 10, true) << " "
3835 << left_justify(Entry.typeName(), 8) << " "
3836 << format_decimal(Entry.addend(), 8) << " " << Entry.symbolName()
3841 // get_dyld_bind_info_symbolname() is used for disassembly and passed an
3842 // address, ReferenceValue, in the Mach-O file and looks in the dyld bind
3843 // information for that address. If the address is found its binding symbol
3844 // name is returned. If not nullptr is returned.
3845 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
3846 struct DisassembleInfo *info) {
3847 if (info->bindtable == nullptr) {
3848 info->bindtable = new (BindTable);
3849 SegInfo sectionTable(info->O);
3850 for (const llvm::object::MachOBindEntry &Entry : info->O->bindTable()) {
3851 uint32_t SegIndex = Entry.segmentIndex();
3852 uint64_t OffsetInSeg = Entry.segmentOffset();
3853 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
3854 const char *SymbolName = nullptr;
3855 StringRef name = Entry.symbolName();
3857 SymbolName = name.data();
3858 info->bindtable->push_back(std::make_pair(Address, SymbolName));
3861 for (bind_table_iterator BI = info->bindtable->begin(),
3862 BE = info->bindtable->end();
3864 uint64_t Address = BI->first;
3865 if (ReferenceValue == Address) {
3866 const char *SymbolName = BI->second;