1 //===-- llvm-objdump.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 program is a utility that works like binutils "objdump", that is, it
11 // dumps out a plethora of information about an object file depending on the
14 // The flags and output of this program should be near identical to those of
17 //===----------------------------------------------------------------------===//
19 #include "llvm-objdump.h"
20 #include "llvm/ADT/Optional.h"
21 #include "llvm/ADT/STLExtras.h"
22 #include "llvm/ADT/StringExtras.h"
23 #include "llvm/ADT/Triple.h"
24 #include "llvm/CodeGen/FaultMaps.h"
25 #include "llvm/MC/MCAsmInfo.h"
26 #include "llvm/MC/MCContext.h"
27 #include "llvm/MC/MCDisassembler.h"
28 #include "llvm/MC/MCInst.h"
29 #include "llvm/MC/MCInstPrinter.h"
30 #include "llvm/MC/MCInstrAnalysis.h"
31 #include "llvm/MC/MCInstrInfo.h"
32 #include "llvm/MC/MCObjectFileInfo.h"
33 #include "llvm/MC/MCRegisterInfo.h"
34 #include "llvm/MC/MCRelocationInfo.h"
35 #include "llvm/MC/MCSubtargetInfo.h"
36 #include "llvm/Object/Archive.h"
37 #include "llvm/Object/ELFObjectFile.h"
38 #include "llvm/Object/COFF.h"
39 #include "llvm/Object/MachO.h"
40 #include "llvm/Object/ObjectFile.h"
41 #include "llvm/Support/Casting.h"
42 #include "llvm/Support/CommandLine.h"
43 #include "llvm/Support/Debug.h"
44 #include "llvm/Support/Errc.h"
45 #include "llvm/Support/FileSystem.h"
46 #include "llvm/Support/Format.h"
47 #include "llvm/Support/GraphWriter.h"
48 #include "llvm/Support/Host.h"
49 #include "llvm/Support/ManagedStatic.h"
50 #include "llvm/Support/MemoryBuffer.h"
51 #include "llvm/Support/PrettyStackTrace.h"
52 #include "llvm/Support/Signals.h"
53 #include "llvm/Support/SourceMgr.h"
54 #include "llvm/Support/TargetRegistry.h"
55 #include "llvm/Support/TargetSelect.h"
56 #include "llvm/Support/raw_ostream.h"
60 #include <system_error>
63 using namespace object;
65 static cl::list<std::string>
66 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore);
69 llvm::Disassemble("disassemble",
70 cl::desc("Display assembler mnemonics for the machine instructions"));
72 Disassembled("d", cl::desc("Alias for --disassemble"),
73 cl::aliasopt(Disassemble));
76 llvm::Relocations("r", cl::desc("Display the relocation entries in the file"));
79 llvm::SectionContents("s", cl::desc("Display the content of each section"));
82 llvm::SymbolTable("t", cl::desc("Display the symbol table"));
85 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols"));
88 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info"));
91 llvm::Bind("bind", cl::desc("Display mach-o binding info"));
94 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info"));
97 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info"));
100 MachOOpt("macho", cl::desc("Use MachO specific object file parser"));
102 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt));
105 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, "
106 "see -version for available targets"));
110 cl::desc("Target a specific cpu type (-mcpu=help for details)"),
111 cl::value_desc("cpu-name"),
115 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, "
116 "see -version for available targets"));
119 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the "
120 "headers for each section."));
122 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"),
123 cl::aliasopt(SectionHeaders));
125 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"),
126 cl::aliasopt(SectionHeaders));
128 cl::list<std::string>
129 llvm::MAttrs("mattr",
131 cl::desc("Target specific attributes"),
132 cl::value_desc("a1,+a2,-a3,..."));
135 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling "
136 "instructions, do not print "
137 "the instruction bytes."));
140 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information"));
143 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"),
144 cl::aliasopt(UnwindInfo));
147 llvm::PrivateHeaders("private-headers",
148 cl::desc("Display format specific file headers"));
151 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"),
152 cl::aliasopt(PrivateHeaders));
155 llvm::PrintImmHex("print-imm-hex",
156 cl::desc("Use hex format for immediate values"));
158 cl::opt<bool> PrintFaultMaps("fault-map-section",
159 cl::desc("Display contents of faultmap section"));
161 static StringRef ToolName;
162 static int ReturnValue = EXIT_SUCCESS;
164 bool llvm::error(std::error_code EC) {
168 outs() << ToolName << ": error reading file: " << EC.message() << ".\n";
170 ReturnValue = EXIT_FAILURE;
174 static void report_error(StringRef File, std::error_code EC) {
176 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n";
177 ReturnValue = EXIT_FAILURE;
180 static const Target *getTarget(const ObjectFile *Obj = nullptr) {
181 // Figure out the target triple.
182 llvm::Triple TheTriple("unknown-unknown-unknown");
183 if (TripleName.empty()) {
185 TheTriple.setArch(Triple::ArchType(Obj->getArch()));
186 // TheTriple defaults to ELF, and COFF doesn't have an environment:
187 // the best we can do here is indicate that it is mach-o.
189 TheTriple.setObjectFormat(Triple::MachO);
192 const auto COFFObj = dyn_cast<COFFObjectFile>(Obj);
193 if (COFFObj->getArch() == Triple::thumb)
194 TheTriple.setTriple("thumbv7-windows");
198 TheTriple.setTriple(Triple::normalize(TripleName));
200 // Get the target specific parser.
202 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple,
205 errs() << ToolName << ": " << Error;
209 // Update the triple name and return the found target.
210 TripleName = TheTriple.getTriple();
214 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) {
215 return a.getOffset() < b.getOffset();
219 class PrettyPrinter {
221 virtual ~PrettyPrinter(){}
222 virtual void printInst(MCInstPrinter &IP, const MCInst *MI,
223 ArrayRef<uint8_t> Bytes, uint64_t Address,
224 raw_ostream &OS, StringRef Annot,
225 MCSubtargetInfo const &STI) {
226 outs() << format("%8" PRIx64 ":", Address);
227 if (!NoShowRawInsn) {
229 dumpBytes(Bytes, outs());
231 IP.printInst(MI, outs(), "", STI);
234 PrettyPrinter PrettyPrinterInst;
235 class HexagonPrettyPrinter : public PrettyPrinter {
237 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address,
240 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0];
241 OS << format("%8" PRIx64 ":", Address);
242 if (!NoShowRawInsn) {
244 dumpBytes(Bytes.slice(0, 4), OS);
245 OS << format("%08" PRIx32, opcode);
248 void printInst(MCInstPrinter &IP, const MCInst *MI,
249 ArrayRef<uint8_t> Bytes, uint64_t Address,
250 raw_ostream &OS, StringRef Annot,
251 MCSubtargetInfo const &STI) override {
254 raw_string_ostream TempStream(Buffer);
255 IP.printInst(MI, TempStream, "", STI);
257 StringRef Contents(Buffer);
258 // Split off bundle attributes
259 auto PacketBundle = Contents.rsplit('\n');
260 // Split off first instruction from the rest
261 auto HeadTail = PacketBundle.first.split('\n');
262 auto Preamble = " { ";
264 while(!HeadTail.first.empty()) {
267 printLead(Bytes, Address, OS);
271 auto Duplex = HeadTail.first.split('\v');
272 if(!Duplex.second.empty()){
275 Inst = Duplex.second;
278 Inst = HeadTail.first;
280 Bytes = Bytes.slice(4);
282 HeadTail = HeadTail.second.split('\n');
284 OS << " } " << PacketBundle.second;
287 HexagonPrettyPrinter HexagonPrettyPrinterInst;
288 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) {
289 switch(Triple.getArch()) {
291 return PrettyPrinterInst;
292 case Triple::hexagon:
293 return HexagonPrettyPrinterInst;
298 template <class ELFT>
299 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj,
301 SmallVectorImpl<char> &Result) {
302 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym;
303 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr;
304 typedef typename ELFObjectFile<ELFT>::Elf_Rel Elf_Rel;
305 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela;
307 const ELFFile<ELFT> &EF = *Obj->getELFFile();
309 ErrorOr<const Elf_Shdr *> SecOrErr = EF.getSection(Rel.d.a);
310 if (std::error_code EC = SecOrErr.getError())
312 const Elf_Shdr *Sec = *SecOrErr;
313 ErrorOr<const Elf_Shdr *> SymTabOrErr = EF.getSection(Sec->sh_link);
314 if (std::error_code EC = SymTabOrErr.getError())
316 const Elf_Shdr *SymTab = *SymTabOrErr;
317 assert(SymTab->sh_type == ELF::SHT_SYMTAB ||
318 SymTab->sh_type == ELF::SHT_DYNSYM);
319 ErrorOr<const Elf_Shdr *> StrTabSec = EF.getSection(SymTab->sh_link);
320 if (std::error_code EC = StrTabSec.getError())
322 ErrorOr<StringRef> StrTabOrErr = EF.getStringTable(*StrTabSec);
323 if (std::error_code EC = StrTabOrErr.getError())
325 StringRef StrTab = *StrTabOrErr;
329 uint16_t symbol_index = 0;
330 switch (Sec->sh_type) {
332 return object_error::parse_failed;
334 const Elf_Rel *ERel = Obj->getRel(Rel);
335 type = ERel->getType(EF.isMips64EL());
336 symbol_index = ERel->getSymbol(EF.isMips64EL());
337 // TODO: Read implicit addend from section data.
340 case ELF::SHT_RELA: {
341 const Elf_Rela *ERela = Obj->getRela(Rel);
342 type = ERela->getType(EF.isMips64EL());
343 symbol_index = ERela->getSymbol(EF.isMips64EL());
344 addend = ERela->r_addend;
348 const Elf_Sym *symb =
349 EF.template getEntry<Elf_Sym>(Sec->sh_link, symbol_index);
351 ErrorOr<const Elf_Shdr *> SymSec = EF.getSection(symb);
352 if (std::error_code EC = SymSec.getError())
354 if (symb->getType() == ELF::STT_SECTION) {
355 ErrorOr<StringRef> SecName = EF.getSectionName(*SymSec);
356 if (std::error_code EC = SecName.getError())
360 ErrorOr<StringRef> SymName = symb->getName(StrTab);
362 return SymName.getError();
365 switch (EF.getHeader()->e_machine) {
368 case ELF::R_X86_64_PC8:
369 case ELF::R_X86_64_PC16:
370 case ELF::R_X86_64_PC32: {
372 raw_string_ostream fmt(fmtbuf);
373 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P";
375 Result.append(fmtbuf.begin(), fmtbuf.end());
377 case ELF::R_X86_64_8:
378 case ELF::R_X86_64_16:
379 case ELF::R_X86_64_32:
380 case ELF::R_X86_64_32S:
381 case ELF::R_X86_64_64: {
383 raw_string_ostream fmt(fmtbuf);
384 fmt << Target << (addend < 0 ? "" : "+") << addend;
386 Result.append(fmtbuf.begin(), fmtbuf.end());
392 case ELF::EM_AARCH64: {
394 raw_string_ostream fmt(fmtbuf);
397 fmt << (addend < 0 ? "" : "+") << addend;
399 Result.append(fmtbuf.begin(), fmtbuf.end());
404 case ELF::EM_HEXAGON:
412 Result.append(res.begin(), res.end());
413 return std::error_code();
416 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj,
417 const RelocationRef &RelRef,
418 SmallVectorImpl<char> &Result) {
419 DataRefImpl Rel = RelRef.getRawDataRefImpl();
420 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj))
421 return getRelocationValueString(ELF32LE, Rel, Result);
422 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj))
423 return getRelocationValueString(ELF64LE, Rel, Result);
424 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj))
425 return getRelocationValueString(ELF32BE, Rel, Result);
426 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj);
427 return getRelocationValueString(ELF64BE, Rel, Result);
430 static std::error_code getRelocationValueString(const COFFObjectFile *Obj,
431 const RelocationRef &Rel,
432 SmallVectorImpl<char> &Result) {
433 symbol_iterator SymI = Rel.getSymbol();
434 ErrorOr<StringRef> SymNameOrErr = SymI->getName();
435 if (std::error_code EC = SymNameOrErr.getError())
437 StringRef SymName = *SymNameOrErr;
438 Result.append(SymName.begin(), SymName.end());
439 return std::error_code();
442 static void printRelocationTargetName(const MachOObjectFile *O,
443 const MachO::any_relocation_info &RE,
444 raw_string_ostream &fmt) {
445 bool IsScattered = O->isRelocationScattered(RE);
447 // Target of a scattered relocation is an address. In the interest of
448 // generating pretty output, scan through the symbol table looking for a
449 // symbol that aligns with that address. If we find one, print it.
450 // Otherwise, we just print the hex address of the target.
452 uint32_t Val = O->getPlainRelocationSymbolNum(RE);
454 for (const SymbolRef &Symbol : O->symbols()) {
456 ErrorOr<uint64_t> Addr = Symbol.getAddress();
457 if ((ec = Addr.getError()))
458 report_fatal_error(ec.message());
461 ErrorOr<StringRef> Name = Symbol.getName();
462 if (std::error_code EC = Name.getError())
463 report_fatal_error(EC.message());
468 // If we couldn't find a symbol that this relocation refers to, try
469 // to find a section beginning instead.
470 for (const SectionRef &Section : O->sections()) {
474 uint64_t Addr = Section.getAddress();
477 if ((ec = Section.getName(Name)))
478 report_fatal_error(ec.message());
483 fmt << format("0x%x", Val);
488 bool isExtern = O->getPlainRelocationExternal(RE);
489 uint64_t Val = O->getPlainRelocationSymbolNum(RE);
492 symbol_iterator SI = O->symbol_begin();
494 ErrorOr<StringRef> SOrErr = SI->getName();
495 if (!error(SOrErr.getError()))
498 section_iterator SI = O->section_begin();
499 // Adjust for the fact that sections are 1-indexed.
500 advance(SI, Val - 1);
507 static std::error_code getRelocationValueString(const MachOObjectFile *Obj,
508 const RelocationRef &RelRef,
509 SmallVectorImpl<char> &Result) {
510 DataRefImpl Rel = RelRef.getRawDataRefImpl();
511 MachO::any_relocation_info RE = Obj->getRelocation(Rel);
513 unsigned Arch = Obj->getArch();
516 raw_string_ostream fmt(fmtbuf);
517 unsigned Type = Obj->getAnyRelocationType(RE);
518 bool IsPCRel = Obj->getAnyRelocationPCRel(RE);
520 // Determine any addends that should be displayed with the relocation.
521 // These require decoding the relocation type, which is triple-specific.
523 // X86_64 has entirely custom relocation types.
524 if (Arch == Triple::x86_64) {
525 bool isPCRel = Obj->getAnyRelocationPCRel(RE);
528 case MachO::X86_64_RELOC_GOT_LOAD:
529 case MachO::X86_64_RELOC_GOT: {
530 printRelocationTargetName(Obj, RE, fmt);
536 case MachO::X86_64_RELOC_SUBTRACTOR: {
537 DataRefImpl RelNext = Rel;
538 Obj->moveRelocationNext(RelNext);
539 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext);
541 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type
542 // X86_64_RELOC_UNSIGNED.
543 // NOTE: Scattered relocations don't exist on x86_64.
544 unsigned RType = Obj->getAnyRelocationType(RENext);
545 if (RType != MachO::X86_64_RELOC_UNSIGNED)
546 report_fatal_error("Expected X86_64_RELOC_UNSIGNED after "
547 "X86_64_RELOC_SUBTRACTOR.");
549 // The X86_64_RELOC_UNSIGNED contains the minuend symbol;
550 // X86_64_RELOC_SUBTRACTOR contains the subtrahend.
551 printRelocationTargetName(Obj, RENext, fmt);
553 printRelocationTargetName(Obj, RE, fmt);
556 case MachO::X86_64_RELOC_TLV:
557 printRelocationTargetName(Obj, RE, fmt);
562 case MachO::X86_64_RELOC_SIGNED_1:
563 printRelocationTargetName(Obj, RE, fmt);
566 case MachO::X86_64_RELOC_SIGNED_2:
567 printRelocationTargetName(Obj, RE, fmt);
570 case MachO::X86_64_RELOC_SIGNED_4:
571 printRelocationTargetName(Obj, RE, fmt);
575 printRelocationTargetName(Obj, RE, fmt);
578 // X86 and ARM share some relocation types in common.
579 } else if (Arch == Triple::x86 || Arch == Triple::arm ||
580 Arch == Triple::ppc) {
581 // Generic relocation types...
583 case MachO::GENERIC_RELOC_PAIR: // prints no info
584 return std::error_code();
585 case MachO::GENERIC_RELOC_SECTDIFF: {
586 DataRefImpl RelNext = Rel;
587 Obj->moveRelocationNext(RelNext);
588 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext);
590 // X86 sect diff's must be followed by a relocation of type
591 // GENERIC_RELOC_PAIR.
592 unsigned RType = Obj->getAnyRelocationType(RENext);
594 if (RType != MachO::GENERIC_RELOC_PAIR)
595 report_fatal_error("Expected GENERIC_RELOC_PAIR after "
596 "GENERIC_RELOC_SECTDIFF.");
598 printRelocationTargetName(Obj, RE, fmt);
600 printRelocationTargetName(Obj, RENext, fmt);
605 if (Arch == Triple::x86 || Arch == Triple::ppc) {
607 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: {
608 DataRefImpl RelNext = Rel;
609 Obj->moveRelocationNext(RelNext);
610 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext);
612 // X86 sect diff's must be followed by a relocation of type
613 // GENERIC_RELOC_PAIR.
614 unsigned RType = Obj->getAnyRelocationType(RENext);
615 if (RType != MachO::GENERIC_RELOC_PAIR)
616 report_fatal_error("Expected GENERIC_RELOC_PAIR after "
617 "GENERIC_RELOC_LOCAL_SECTDIFF.");
619 printRelocationTargetName(Obj, RE, fmt);
621 printRelocationTargetName(Obj, RENext, fmt);
624 case MachO::GENERIC_RELOC_TLV: {
625 printRelocationTargetName(Obj, RE, fmt);
632 printRelocationTargetName(Obj, RE, fmt);
634 } else { // ARM-specific relocations
636 case MachO::ARM_RELOC_HALF:
637 case MachO::ARM_RELOC_HALF_SECTDIFF: {
638 // Half relocations steal a bit from the length field to encode
639 // whether this is an upper16 or a lower16 relocation.
640 bool isUpper = Obj->getAnyRelocationLength(RE) >> 1;
646 printRelocationTargetName(Obj, RE, fmt);
648 DataRefImpl RelNext = Rel;
649 Obj->moveRelocationNext(RelNext);
650 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext);
652 // ARM half relocs must be followed by a relocation of type
654 unsigned RType = Obj->getAnyRelocationType(RENext);
655 if (RType != MachO::ARM_RELOC_PAIR)
656 report_fatal_error("Expected ARM_RELOC_PAIR after "
659 // NOTE: The half of the target virtual address is stashed in the
660 // address field of the secondary relocation, but we can't reverse
661 // engineer the constant offset from it without decoding the movw/movt
662 // instruction to find the other half in its immediate field.
664 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the
665 // symbol/section pointer of the follow-on relocation.
666 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) {
668 printRelocationTargetName(Obj, RENext, fmt);
674 default: { printRelocationTargetName(Obj, RE, fmt); }
678 printRelocationTargetName(Obj, RE, fmt);
681 Result.append(fmtbuf.begin(), fmtbuf.end());
682 return std::error_code();
685 static std::error_code getRelocationValueString(const RelocationRef &Rel,
686 SmallVectorImpl<char> &Result) {
687 const ObjectFile *Obj = Rel.getObject();
688 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj))
689 return getRelocationValueString(ELF, Rel, Result);
690 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj))
691 return getRelocationValueString(COFF, Rel, Result);
692 auto *MachO = cast<MachOObjectFile>(Obj);
693 return getRelocationValueString(MachO, Rel, Result);
696 /// @brief Indicates whether this relocation should hidden when listing
697 /// relocations, usually because it is the trailing part of a multipart
698 /// relocation that will be printed as part of the leading relocation.
699 static bool getHidden(RelocationRef RelRef) {
700 const ObjectFile *Obj = RelRef.getObject();
701 auto *MachO = dyn_cast<MachOObjectFile>(Obj);
705 unsigned Arch = MachO->getArch();
706 DataRefImpl Rel = RelRef.getRawDataRefImpl();
707 uint64_t Type = MachO->getRelocationType(Rel);
709 // On arches that use the generic relocations, GENERIC_RELOC_PAIR
711 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) {
712 if (Type == MachO::GENERIC_RELOC_PAIR)
714 } else if (Arch == Triple::x86_64) {
715 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows
716 // an X86_64_RELOC_SUBTRACTOR.
717 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) {
718 DataRefImpl RelPrev = Rel;
720 uint64_t PrevType = MachO->getRelocationType(RelPrev);
721 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR)
729 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) {
730 const Target *TheTarget = getTarget(Obj);
731 // getTarget() will have already issued a diagnostic if necessary, so
732 // just bail here if it failed.
736 // Package up features to be passed to target/subtarget
737 std::string FeaturesStr;
739 SubtargetFeatures Features;
740 for (unsigned i = 0; i != MAttrs.size(); ++i)
741 Features.AddFeature(MAttrs[i]);
742 FeaturesStr = Features.getString();
745 std::unique_ptr<const MCRegisterInfo> MRI(
746 TheTarget->createMCRegInfo(TripleName));
748 errs() << "error: no register info for target " << TripleName << "\n";
752 // Set up disassembler.
753 std::unique_ptr<const MCAsmInfo> AsmInfo(
754 TheTarget->createMCAsmInfo(*MRI, TripleName));
756 errs() << "error: no assembly info for target " << TripleName << "\n";
760 std::unique_ptr<const MCSubtargetInfo> STI(
761 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr));
763 errs() << "error: no subtarget info for target " << TripleName << "\n";
767 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo());
769 errs() << "error: no instruction info for target " << TripleName << "\n";
773 std::unique_ptr<const MCObjectFileInfo> MOFI(new MCObjectFileInfo);
774 MCContext Ctx(AsmInfo.get(), MRI.get(), MOFI.get());
776 std::unique_ptr<MCDisassembler> DisAsm(
777 TheTarget->createMCDisassembler(*STI, Ctx));
780 errs() << "error: no disassembler for target " << TripleName << "\n";
784 std::unique_ptr<const MCInstrAnalysis> MIA(
785 TheTarget->createMCInstrAnalysis(MII.get()));
787 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
788 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
789 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI));
791 errs() << "error: no instruction printer for target " << TripleName
795 IP->setPrintImmHex(PrintImmHex);
796 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName));
798 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " :
799 "\t\t\t%08" PRIx64 ": ";
801 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections
802 // in RelocSecs contain the relocations for section S.
804 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap;
805 for (const SectionRef &Section : Obj->sections()) {
806 section_iterator Sec2 = Section.getRelocatedSection();
807 if (Sec2 != Obj->section_end())
808 SectionRelocMap[*Sec2].push_back(Section);
811 // Create a mapping from virtual address to symbol name. This is used to
812 // pretty print the target of a call.
813 std::vector<std::pair<uint64_t, StringRef>> AllSymbols;
815 for (const SymbolRef &Symbol : Obj->symbols()) {
816 ErrorOr<uint64_t> AddressOrErr = Symbol.getAddress();
817 if (error(AddressOrErr.getError()))
819 uint64_t Address = *AddressOrErr;
821 ErrorOr<StringRef> Name = Symbol.getName();
822 if (error(Name.getError()))
826 AllSymbols.push_back(std::make_pair(Address, *Name));
829 array_pod_sort(AllSymbols.begin(), AllSymbols.end());
832 for (const SectionRef &Section : Obj->sections()) {
833 if (!Section.isText() || Section.isVirtual())
836 uint64_t SectionAddr = Section.getAddress();
837 uint64_t SectSize = Section.getSize();
841 // Make a list of all the symbols in this section.
842 std::vector<std::pair<uint64_t, StringRef>> Symbols;
843 for (const SymbolRef &Symbol : Obj->symbols()) {
844 if (Section.containsSymbol(Symbol)) {
845 ErrorOr<uint64_t> AddressOrErr = Symbol.getAddress();
846 if (error(AddressOrErr.getError()))
848 uint64_t Address = *AddressOrErr;
849 Address -= SectionAddr;
850 if (Address >= SectSize)
853 ErrorOr<StringRef> Name = Symbol.getName();
854 if (error(Name.getError()))
856 Symbols.push_back(std::make_pair(Address, *Name));
860 // Sort the symbols by address, just in case they didn't come in that way.
861 array_pod_sort(Symbols.begin(), Symbols.end());
863 // Make a list of all the relocations for this section.
864 std::vector<RelocationRef> Rels;
866 for (const SectionRef &RelocSec : SectionRelocMap[Section]) {
867 for (const RelocationRef &Reloc : RelocSec.relocations()) {
868 Rels.push_back(Reloc);
873 // Sort relocations by address.
874 std::sort(Rels.begin(), Rels.end(), RelocAddressLess);
876 StringRef SegmentName = "";
877 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) {
878 DataRefImpl DR = Section.getRawDataRefImpl();
879 SegmentName = MachO->getSectionFinalSegmentName(DR);
882 if (error(Section.getName(name)))
884 outs() << "Disassembly of section ";
885 if (!SegmentName.empty())
886 outs() << SegmentName << ",";
887 outs() << name << ':';
889 // If the section has no symbol at the start, just insert a dummy one.
890 if (Symbols.empty() || Symbols[0].first != 0)
891 Symbols.insert(Symbols.begin(), std::make_pair(0, name));
893 SmallString<40> Comments;
894 raw_svector_ostream CommentStream(Comments);
897 if (error(Section.getContents(BytesStr)))
899 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()),
905 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin();
906 std::vector<RelocationRef>::const_iterator rel_end = Rels.end();
907 // Disassemble symbol by symbol.
908 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) {
910 uint64_t Start = Symbols[si].first;
911 // The end is either the section end or the beginning of the next symbol.
912 uint64_t End = (si == se - 1) ? SectSize : Symbols[si + 1].first;
913 // If this symbol has the same address as the next symbol, then skip it.
917 outs() << '\n' << Symbols[si].second << ":\n";
920 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
922 raw_ostream &DebugOut = nulls();
925 for (Index = Start; Index < End; Index += Size) {
928 if (DisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
929 SectionAddr + Index, DebugOut,
931 PIP.printInst(*IP, &Inst,
932 Bytes.slice(Index, Size),
933 SectionAddr + Index, outs(), "", *STI);
934 outs() << CommentStream.str();
936 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst))) {
938 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) {
939 const auto &TargetSym =
940 std::lower_bound(AllSymbols.begin(), AllSymbols.end(),
941 std::make_pair(Target, StringRef()));
942 if (TargetSym != AllSymbols.end()) {
943 outs() << " <" << TargetSym->second;
944 uint64_t Disp = TargetSym->first - Target;
946 outs() << '-' << Disp;
953 errs() << ToolName << ": warning: invalid instruction encoding\n";
955 Size = 1; // skip illegible bytes
958 // Print relocation for instruction.
959 while (rel_cur != rel_end) {
960 bool hidden = getHidden(*rel_cur);
961 uint64_t addr = rel_cur->getOffset();
962 SmallString<16> name;
965 // If this relocation is hidden, skip it.
966 if (hidden) goto skip_print_rel;
968 // Stop when rel_cur's address is past the current instruction.
969 if (addr >= Index + Size) break;
970 rel_cur->getTypeName(name);
971 if (error(getRelocationValueString(*rel_cur, val)))
973 outs() << format(Fmt.data(), SectionAddr + addr) << name
974 << "\t" << val << "\n";
984 void llvm::PrintRelocations(const ObjectFile *Obj) {
985 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 :
987 // Regular objdump doesn't print relocations in non-relocatable object
989 if (!Obj->isRelocatableObject())
992 for (const SectionRef &Section : Obj->sections()) {
993 if (Section.relocation_begin() == Section.relocation_end())
996 if (error(Section.getName(secname)))
998 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n";
999 for (const RelocationRef &Reloc : Section.relocations()) {
1000 bool hidden = getHidden(Reloc);
1001 uint64_t address = Reloc.getOffset();
1002 SmallString<32> relocname;
1003 SmallString<32> valuestr;
1006 Reloc.getTypeName(relocname);
1007 if (error(getRelocationValueString(Reloc, valuestr)))
1009 outs() << format(Fmt.data(), address) << " " << relocname << " "
1010 << valuestr << "\n";
1016 void llvm::PrintSectionHeaders(const ObjectFile *Obj) {
1017 outs() << "Sections:\n"
1018 "Idx Name Size Address Type\n";
1020 for (const SectionRef &Section : Obj->sections()) {
1022 if (error(Section.getName(Name)))
1024 uint64_t Address = Section.getAddress();
1025 uint64_t Size = Section.getSize();
1026 bool Text = Section.isText();
1027 bool Data = Section.isData();
1028 bool BSS = Section.isBSS();
1029 std::string Type = (std::string(Text ? "TEXT " : "") +
1030 (Data ? "DATA " : "") + (BSS ? "BSS" : ""));
1031 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i,
1032 Name.str().c_str(), Size, Address, Type.c_str());
1037 void llvm::PrintSectionContents(const ObjectFile *Obj) {
1039 for (const SectionRef &Section : Obj->sections()) {
1042 if (error(Section.getName(Name)))
1044 uint64_t BaseAddr = Section.getAddress();
1045 uint64_t Size = Section.getSize();
1049 outs() << "Contents of section " << Name << ":\n";
1050 if (Section.isBSS()) {
1051 outs() << format("<skipping contents of bss section at [%04" PRIx64
1052 ", %04" PRIx64 ")>\n",
1053 BaseAddr, BaseAddr + Size);
1057 if (error(Section.getContents(Contents)))
1060 // Dump out the content as hex and printable ascii characters.
1061 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) {
1062 outs() << format(" %04" PRIx64 " ", BaseAddr + addr);
1063 // Dump line of hex.
1064 for (std::size_t i = 0; i < 16; ++i) {
1065 if (i != 0 && i % 4 == 0)
1068 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true)
1069 << hexdigit(Contents[addr + i] & 0xF, true);
1075 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) {
1076 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF))
1077 outs() << Contents[addr + i];
1086 static void PrintCOFFSymbolTable(const COFFObjectFile *coff) {
1087 for (unsigned SI = 0, SE = coff->getNumberOfSymbols(); SI != SE; ++SI) {
1088 ErrorOr<COFFSymbolRef> Symbol = coff->getSymbol(SI);
1090 if (error(Symbol.getError()))
1093 if (error(coff->getSymbolName(*Symbol, Name)))
1096 outs() << "[" << format("%2d", SI) << "]"
1097 << "(sec " << format("%2d", int(Symbol->getSectionNumber())) << ")"
1098 << "(fl 0x00)" // Flag bits, which COFF doesn't have.
1099 << "(ty " << format("%3x", unsigned(Symbol->getType())) << ")"
1100 << "(scl " << format("%3x", unsigned(Symbol->getStorageClass())) << ") "
1101 << "(nx " << unsigned(Symbol->getNumberOfAuxSymbols()) << ") "
1102 << "0x" << format("%08x", unsigned(Symbol->getValue())) << " "
1105 for (unsigned AI = 0, AE = Symbol->getNumberOfAuxSymbols(); AI < AE; ++AI, ++SI) {
1106 if (Symbol->isSectionDefinition()) {
1107 const coff_aux_section_definition *asd;
1108 if (error(coff->getAuxSymbol<coff_aux_section_definition>(SI + 1, asd)))
1111 int32_t AuxNumber = asd->getNumber(Symbol->isBigObj());
1114 << format("scnlen 0x%x nreloc %d nlnno %d checksum 0x%x "
1115 , unsigned(asd->Length)
1116 , unsigned(asd->NumberOfRelocations)
1117 , unsigned(asd->NumberOfLinenumbers)
1118 , unsigned(asd->CheckSum))
1119 << format("assoc %d comdat %d\n"
1120 , unsigned(AuxNumber)
1121 , unsigned(asd->Selection));
1122 } else if (Symbol->isFileRecord()) {
1123 const char *FileName;
1124 if (error(coff->getAuxSymbol<char>(SI + 1, FileName)))
1127 StringRef Name(FileName, Symbol->getNumberOfAuxSymbols() *
1128 coff->getSymbolTableEntrySize());
1129 outs() << "AUX " << Name.rtrim(StringRef("\0", 1)) << '\n';
1131 SI = SI + Symbol->getNumberOfAuxSymbols();
1134 outs() << "AUX Unknown\n";
1140 void llvm::PrintSymbolTable(const ObjectFile *o) {
1141 outs() << "SYMBOL TABLE:\n";
1143 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) {
1144 PrintCOFFSymbolTable(coff);
1147 for (const SymbolRef &Symbol : o->symbols()) {
1148 ErrorOr<uint64_t> AddressOrError = Symbol.getAddress();
1149 if (error(AddressOrError.getError()))
1151 uint64_t Address = *AddressOrError;
1152 SymbolRef::Type Type = Symbol.getType();
1153 uint32_t Flags = Symbol.getFlags();
1154 section_iterator Section = o->section_end();
1155 if (error(Symbol.getSection(Section)))
1158 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) {
1159 Section->getName(Name);
1161 ErrorOr<StringRef> NameOrErr = Symbol.getName();
1162 if (error(NameOrErr.getError()))
1167 bool Global = Flags & SymbolRef::SF_Global;
1168 bool Weak = Flags & SymbolRef::SF_Weak;
1169 bool Absolute = Flags & SymbolRef::SF_Absolute;
1170 bool Common = Flags & SymbolRef::SF_Common;
1171 bool Hidden = Flags & SymbolRef::SF_Hidden;
1174 if (Type != SymbolRef::ST_Unknown)
1175 GlobLoc = Global ? 'g' : 'l';
1176 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File)
1178 char FileFunc = ' ';
1179 if (Type == SymbolRef::ST_File)
1181 else if (Type == SymbolRef::ST_Function)
1184 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 :
1187 outs() << format(Fmt, Address) << " "
1188 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' '
1189 << (Weak ? 'w' : ' ') // Weak?
1190 << ' ' // Constructor. Not supported yet.
1191 << ' ' // Warning. Not supported yet.
1192 << ' ' // Indirect reference to another symbol.
1193 << Debug // Debugging (d) or dynamic (D) symbol.
1194 << FileFunc // Name of function (F), file (f) or object (O).
1198 } else if (Common) {
1200 } else if (Section == o->section_end()) {
1203 if (const MachOObjectFile *MachO =
1204 dyn_cast<const MachOObjectFile>(o)) {
1205 DataRefImpl DR = Section->getRawDataRefImpl();
1206 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR);
1207 outs() << SegmentName << ",";
1209 StringRef SectionName;
1210 if (error(Section->getName(SectionName)))
1212 outs() << SectionName;
1216 if (Common || isa<ELFObjectFileBase>(o)) {
1218 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize();
1219 outs() << format("\t %08" PRIx64 " ", Val);
1223 outs() << ".hidden ";
1230 static void PrintUnwindInfo(const ObjectFile *o) {
1231 outs() << "Unwind info:\n\n";
1233 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) {
1234 printCOFFUnwindInfo(coff);
1235 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o))
1236 printMachOUnwindInfo(MachO);
1238 // TODO: Extract DWARF dump tool to objdump.
1239 errs() << "This operation is only currently supported "
1240 "for COFF and MachO object files.\n";
1245 void llvm::printExportsTrie(const ObjectFile *o) {
1246 outs() << "Exports trie:\n";
1247 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o))
1248 printMachOExportsTrie(MachO);
1250 errs() << "This operation is only currently supported "
1251 "for Mach-O executable files.\n";
1256 void llvm::printRebaseTable(const ObjectFile *o) {
1257 outs() << "Rebase table:\n";
1258 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o))
1259 printMachORebaseTable(MachO);
1261 errs() << "This operation is only currently supported "
1262 "for Mach-O executable files.\n";
1267 void llvm::printBindTable(const ObjectFile *o) {
1268 outs() << "Bind table:\n";
1269 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o))
1270 printMachOBindTable(MachO);
1272 errs() << "This operation is only currently supported "
1273 "for Mach-O executable files.\n";
1278 void llvm::printLazyBindTable(const ObjectFile *o) {
1279 outs() << "Lazy bind table:\n";
1280 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o))
1281 printMachOLazyBindTable(MachO);
1283 errs() << "This operation is only currently supported "
1284 "for Mach-O executable files.\n";
1289 void llvm::printWeakBindTable(const ObjectFile *o) {
1290 outs() << "Weak bind table:\n";
1291 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o))
1292 printMachOWeakBindTable(MachO);
1294 errs() << "This operation is only currently supported "
1295 "for Mach-O executable files.\n";
1300 static void printFaultMaps(const ObjectFile *Obj) {
1301 const char *FaultMapSectionName = nullptr;
1303 if (isa<ELFObjectFileBase>(Obj)) {
1304 FaultMapSectionName = ".llvm_faultmaps";
1305 } else if (isa<MachOObjectFile>(Obj)) {
1306 FaultMapSectionName = "__llvm_faultmaps";
1308 errs() << "This operation is only currently supported "
1309 "for ELF and Mach-O executable files.\n";
1313 Optional<object::SectionRef> FaultMapSection;
1315 for (auto Sec : Obj->sections()) {
1318 if (Name == FaultMapSectionName) {
1319 FaultMapSection = Sec;
1324 outs() << "FaultMap table:\n";
1326 if (!FaultMapSection.hasValue()) {
1327 outs() << "<not found>\n";
1331 StringRef FaultMapContents;
1332 if (error(FaultMapSection.getValue().getContents(FaultMapContents))) {
1333 errs() << "Could not read the " << FaultMapContents << " section!\n";
1337 FaultMapParser FMP(FaultMapContents.bytes_begin(),
1338 FaultMapContents.bytes_end());
1343 static void printPrivateFileHeader(const ObjectFile *o) {
1345 printELFFileHeader(o);
1346 } else if (o->isCOFF()) {
1347 printCOFFFileHeader(o);
1348 } else if (o->isMachO()) {
1349 printMachOFileHeader(o);
1353 static void DumpObject(const ObjectFile *o) {
1355 outs() << o->getFileName()
1356 << ":\tfile format " << o->getFileFormatName() << "\n\n";
1359 DisassembleObject(o, Relocations);
1360 if (Relocations && !Disassemble)
1361 PrintRelocations(o);
1363 PrintSectionHeaders(o);
1364 if (SectionContents)
1365 PrintSectionContents(o);
1367 PrintSymbolTable(o);
1371 printPrivateFileHeader(o);
1373 printExportsTrie(o);
1375 printRebaseTable(o);
1379 printLazyBindTable(o);
1381 printWeakBindTable(o);
1386 /// @brief Dump each object file in \a a;
1387 static void DumpArchive(const Archive *a) {
1388 for (Archive::child_iterator i = a->child_begin(), e = a->child_end(); i != e;
1390 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = i->getAsBinary();
1391 if (std::error_code EC = ChildOrErr.getError()) {
1392 // Ignore non-object files.
1393 if (EC != object_error::invalid_file_type)
1394 report_error(a->getFileName(), EC);
1397 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get()))
1400 report_error(a->getFileName(), object_error::invalid_file_type);
1404 /// @brief Open file and figure out how to dump it.
1405 static void DumpInput(StringRef file) {
1406 // If file isn't stdin, check that it exists.
1407 if (file != "-" && !sys::fs::exists(file)) {
1408 report_error(file, errc::no_such_file_or_directory);
1412 // If we are using the Mach-O specific object file parser, then let it parse
1413 // the file and process the command line options. So the -arch flags can
1414 // be used to select specific slices, etc.
1416 ParseInputMachO(file);
1420 // Attempt to open the binary.
1421 ErrorOr<OwningBinary<Binary>> BinaryOrErr = createBinary(file);
1422 if (std::error_code EC = BinaryOrErr.getError()) {
1423 report_error(file, EC);
1426 Binary &Binary = *BinaryOrErr.get().getBinary();
1428 if (Archive *a = dyn_cast<Archive>(&Binary))
1430 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary))
1433 report_error(file, object_error::invalid_file_type);
1436 int main(int argc, char **argv) {
1437 // Print a stack trace if we signal out.
1438 sys::PrintStackTraceOnErrorSignal();
1439 PrettyStackTraceProgram X(argc, argv);
1440 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit.
1442 // Initialize targets and assembly printers/parsers.
1443 llvm::InitializeAllTargetInfos();
1444 llvm::InitializeAllTargetMCs();
1445 llvm::InitializeAllAsmParsers();
1446 llvm::InitializeAllDisassemblers();
1448 // Register the target printer for --version.
1449 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion);
1451 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n");
1452 TripleName = Triple::normalize(TripleName);
1456 // Defaults to a.out if no filenames specified.
1457 if (InputFilenames.size() == 0)
1458 InputFilenames.push_back("a.out");
1472 && !(UniversalHeaders && MachOOpt)
1473 && !(ArchiveHeaders && MachOOpt)
1474 && !(IndirectSymbols && MachOOpt)
1475 && !(DataInCode && MachOOpt)
1476 && !(LinkOptHints && MachOOpt)
1477 && !(InfoPlist && MachOOpt)
1478 && !(DylibsUsed && MachOOpt)
1479 && !(DylibId && MachOOpt)
1480 && !(ObjcMetaData && MachOOpt)
1481 && !(DumpSections.size() != 0 && MachOOpt)
1482 && !PrintFaultMaps) {
1483 cl::PrintHelpMessage();
1487 std::for_each(InputFilenames.begin(), InputFilenames.end(),