Clang] Fix expansion of response files in -Wp after integrated-cc1 change
[llvm-project.git] / llvm / tools / llvm-objdump / llvm-objdump.cpp
blob6bd37a1fb86c9bb7c105758751d0a796053476b4
1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This program is a utility that works like binutils "objdump", that is, it
10 // dumps out a plethora of information about an object file depending on the
11 // flags.
13 // The flags and output of this program should be near identical to those of
14 // binutils objdump.
16 //===----------------------------------------------------------------------===//
18 #include "llvm-objdump.h"
19 #include "llvm/ADT/Optional.h"
20 #include "llvm/ADT/STLExtras.h"
21 #include "llvm/ADT/SetOperations.h"
22 #include "llvm/ADT/StringExtras.h"
23 #include "llvm/ADT/StringSet.h"
24 #include "llvm/ADT/Triple.h"
25 #include "llvm/CodeGen/FaultMaps.h"
26 #include "llvm/DebugInfo/DWARF/DWARFContext.h"
27 #include "llvm/DebugInfo/Symbolize/Symbolize.h"
28 #include "llvm/Demangle/Demangle.h"
29 #include "llvm/MC/MCAsmInfo.h"
30 #include "llvm/MC/MCContext.h"
31 #include "llvm/MC/MCDisassembler/MCDisassembler.h"
32 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h"
33 #include "llvm/MC/MCInst.h"
34 #include "llvm/MC/MCInstPrinter.h"
35 #include "llvm/MC/MCInstrAnalysis.h"
36 #include "llvm/MC/MCInstrInfo.h"
37 #include "llvm/MC/MCObjectFileInfo.h"
38 #include "llvm/MC/MCRegisterInfo.h"
39 #include "llvm/MC/MCSubtargetInfo.h"
40 #include "llvm/MC/MCTargetOptions.h"
41 #include "llvm/Object/Archive.h"
42 #include "llvm/Object/COFF.h"
43 #include "llvm/Object/COFFImportFile.h"
44 #include "llvm/Object/ELFObjectFile.h"
45 #include "llvm/Object/MachO.h"
46 #include "llvm/Object/MachOUniversal.h"
47 #include "llvm/Object/ObjectFile.h"
48 #include "llvm/Object/Wasm.h"
49 #include "llvm/Support/Casting.h"
50 #include "llvm/Support/CommandLine.h"
51 #include "llvm/Support/Debug.h"
52 #include "llvm/Support/Errc.h"
53 #include "llvm/Support/FileSystem.h"
54 #include "llvm/Support/Format.h"
55 #include "llvm/Support/FormatVariadic.h"
56 #include "llvm/Support/GraphWriter.h"
57 #include "llvm/Support/Host.h"
58 #include "llvm/Support/InitLLVM.h"
59 #include "llvm/Support/MemoryBuffer.h"
60 #include "llvm/Support/SourceMgr.h"
61 #include "llvm/Support/StringSaver.h"
62 #include "llvm/Support/TargetRegistry.h"
63 #include "llvm/Support/TargetSelect.h"
64 #include "llvm/Support/WithColor.h"
65 #include "llvm/Support/raw_ostream.h"
66 #include <algorithm>
67 #include <cctype>
68 #include <cstring>
69 #include <system_error>
70 #include <unordered_map>
71 #include <utility>
73 using namespace llvm::object;
75 namespace llvm {
77 cl::OptionCategory ObjdumpCat("llvm-objdump Options");
79 // MachO specific
80 extern cl::OptionCategory MachOCat;
81 extern cl::opt<bool> Bind;
82 extern cl::opt<bool> DataInCode;
83 extern cl::opt<bool> DylibsUsed;
84 extern cl::opt<bool> DylibId;
85 extern cl::opt<bool> ExportsTrie;
86 extern cl::opt<bool> FirstPrivateHeader;
87 extern cl::opt<bool> IndirectSymbols;
88 extern cl::opt<bool> InfoPlist;
89 extern cl::opt<bool> LazyBind;
90 extern cl::opt<bool> LinkOptHints;
91 extern cl::opt<bool> ObjcMetaData;
92 extern cl::opt<bool> Rebase;
93 extern cl::opt<bool> UniversalHeaders;
94 extern cl::opt<bool> WeakBind;
96 static cl::opt<uint64_t> AdjustVMA(
97 "adjust-vma",
98 cl::desc("Increase the displayed address by the specified offset"),
99 cl::value_desc("offset"), cl::init(0), cl::cat(ObjdumpCat));
101 static cl::opt<bool>
102 AllHeaders("all-headers",
103 cl::desc("Display all available header information"),
104 cl::cat(ObjdumpCat));
105 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"),
106 cl::NotHidden, cl::Grouping,
107 cl::aliasopt(AllHeaders));
109 static cl::opt<std::string>
110 ArchName("arch-name",
111 cl::desc("Target arch to disassemble for, "
112 "see -version for available targets"),
113 cl::cat(ObjdumpCat));
115 cl::opt<bool> ArchiveHeaders("archive-headers",
116 cl::desc("Display archive header information"),
117 cl::cat(ObjdumpCat));
118 static cl::alias ArchiveHeadersShort("a",
119 cl::desc("Alias for --archive-headers"),
120 cl::NotHidden, cl::Grouping,
121 cl::aliasopt(ArchiveHeaders));
123 cl::opt<bool> Demangle("demangle", cl::desc("Demangle symbols names"),
124 cl::init(false), cl::cat(ObjdumpCat));
125 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"),
126 cl::NotHidden, cl::Grouping,
127 cl::aliasopt(Demangle));
129 cl::opt<bool> Disassemble(
130 "disassemble",
131 cl::desc("Display assembler mnemonics for the machine instructions"),
132 cl::cat(ObjdumpCat));
133 static cl::alias DisassembleShort("d", cl::desc("Alias for --disassemble"),
134 cl::NotHidden, cl::Grouping,
135 cl::aliasopt(Disassemble));
137 cl::opt<bool> DisassembleAll(
138 "disassemble-all",
139 cl::desc("Display assembler mnemonics for the machine instructions"),
140 cl::cat(ObjdumpCat));
141 static cl::alias DisassembleAllShort("D",
142 cl::desc("Alias for --disassemble-all"),
143 cl::NotHidden, cl::Grouping,
144 cl::aliasopt(DisassembleAll));
146 static cl::list<std::string>
147 DisassembleFunctions("disassemble-functions", cl::CommaSeparated,
148 cl::desc("List of functions to disassemble. "
149 "Accept demangled names when --demangle is "
150 "specified, otherwise accept mangled names"),
151 cl::cat(ObjdumpCat));
153 static cl::opt<bool> DisassembleZeroes(
154 "disassemble-zeroes",
155 cl::desc("Do not skip blocks of zeroes when disassembling"),
156 cl::cat(ObjdumpCat));
157 static cl::alias
158 DisassembleZeroesShort("z", cl::desc("Alias for --disassemble-zeroes"),
159 cl::NotHidden, cl::Grouping,
160 cl::aliasopt(DisassembleZeroes));
162 static cl::list<std::string>
163 DisassemblerOptions("disassembler-options",
164 cl::desc("Pass target specific disassembler options"),
165 cl::value_desc("options"), cl::CommaSeparated,
166 cl::cat(ObjdumpCat));
167 static cl::alias
168 DisassemblerOptionsShort("M", cl::desc("Alias for --disassembler-options"),
169 cl::NotHidden, cl::Grouping, cl::Prefix,
170 cl::CommaSeparated,
171 cl::aliasopt(DisassemblerOptions));
173 cl::opt<DIDumpType> DwarfDumpType(
174 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"),
175 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame")),
176 cl::cat(ObjdumpCat));
178 static cl::opt<bool> DynamicRelocations(
179 "dynamic-reloc",
180 cl::desc("Display the dynamic relocation entries in the file"),
181 cl::cat(ObjdumpCat));
182 static cl::alias DynamicRelocationShort("R",
183 cl::desc("Alias for --dynamic-reloc"),
184 cl::NotHidden, cl::Grouping,
185 cl::aliasopt(DynamicRelocations));
187 static cl::opt<bool>
188 FaultMapSection("fault-map-section",
189 cl::desc("Display contents of faultmap section"),
190 cl::cat(ObjdumpCat));
192 static cl::opt<bool>
193 FileHeaders("file-headers",
194 cl::desc("Display the contents of the overall file header"),
195 cl::cat(ObjdumpCat));
196 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"),
197 cl::NotHidden, cl::Grouping,
198 cl::aliasopt(FileHeaders));
200 cl::opt<bool> SectionContents("full-contents",
201 cl::desc("Display the content of each section"),
202 cl::cat(ObjdumpCat));
203 static cl::alias SectionContentsShort("s",
204 cl::desc("Alias for --full-contents"),
205 cl::NotHidden, cl::Grouping,
206 cl::aliasopt(SectionContents));
208 static cl::list<std::string> InputFilenames(cl::Positional,
209 cl::desc("<input object files>"),
210 cl::ZeroOrMore,
211 cl::cat(ObjdumpCat));
213 static cl::opt<bool>
214 PrintLines("line-numbers",
215 cl::desc("Display source line numbers with "
216 "disassembly. Implies disassemble object"),
217 cl::cat(ObjdumpCat));
218 static cl::alias PrintLinesShort("l", cl::desc("Alias for --line-numbers"),
219 cl::NotHidden, cl::Grouping,
220 cl::aliasopt(PrintLines));
222 static cl::opt<bool> MachOOpt("macho",
223 cl::desc("Use MachO specific object file parser"),
224 cl::cat(ObjdumpCat));
225 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden,
226 cl::Grouping, cl::aliasopt(MachOOpt));
228 cl::opt<std::string>
229 MCPU("mcpu",
230 cl::desc("Target a specific cpu type (-mcpu=help for details)"),
231 cl::value_desc("cpu-name"), cl::init(""), cl::cat(ObjdumpCat));
233 cl::list<std::string> MAttrs("mattr", cl::CommaSeparated,
234 cl::desc("Target specific attributes"),
235 cl::value_desc("a1,+a2,-a3,..."),
236 cl::cat(ObjdumpCat));
238 cl::opt<bool> NoShowRawInsn("no-show-raw-insn",
239 cl::desc("When disassembling "
240 "instructions, do not print "
241 "the instruction bytes."),
242 cl::cat(ObjdumpCat));
243 cl::opt<bool> NoLeadingAddr("no-leading-addr",
244 cl::desc("Print no leading address"),
245 cl::cat(ObjdumpCat));
247 static cl::opt<bool> RawClangAST(
248 "raw-clang-ast",
249 cl::desc("Dump the raw binary contents of the clang AST section"),
250 cl::cat(ObjdumpCat));
252 cl::opt<bool>
253 Relocations("reloc", cl::desc("Display the relocation entries in the file"),
254 cl::cat(ObjdumpCat));
255 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"),
256 cl::NotHidden, cl::Grouping,
257 cl::aliasopt(Relocations));
259 cl::opt<bool> PrintImmHex("print-imm-hex",
260 cl::desc("Use hex format for immediate values"),
261 cl::cat(ObjdumpCat));
263 cl::opt<bool> PrivateHeaders("private-headers",
264 cl::desc("Display format specific file headers"),
265 cl::cat(ObjdumpCat));
266 static cl::alias PrivateHeadersShort("p",
267 cl::desc("Alias for --private-headers"),
268 cl::NotHidden, cl::Grouping,
269 cl::aliasopt(PrivateHeaders));
271 cl::list<std::string>
272 FilterSections("section",
273 cl::desc("Operate on the specified sections only. "
274 "With -macho dump segment,section"),
275 cl::cat(ObjdumpCat));
276 static cl::alias FilterSectionsj("j", cl::desc("Alias for --section"),
277 cl::NotHidden, cl::Grouping, cl::Prefix,
278 cl::aliasopt(FilterSections));
280 cl::opt<bool> SectionHeaders("section-headers",
281 cl::desc("Display summaries of the "
282 "headers for each section."),
283 cl::cat(ObjdumpCat));
284 static cl::alias SectionHeadersShort("headers",
285 cl::desc("Alias for --section-headers"),
286 cl::NotHidden,
287 cl::aliasopt(SectionHeaders));
288 static cl::alias SectionHeadersShorter("h",
289 cl::desc("Alias for --section-headers"),
290 cl::NotHidden, cl::Grouping,
291 cl::aliasopt(SectionHeaders));
293 static cl::opt<bool>
294 ShowLMA("show-lma",
295 cl::desc("Display LMA column when dumping ELF section headers"),
296 cl::cat(ObjdumpCat));
298 static cl::opt<bool> PrintSource(
299 "source",
300 cl::desc(
301 "Display source inlined with disassembly. Implies disassemble object"),
302 cl::cat(ObjdumpCat));
303 static cl::alias PrintSourceShort("S", cl::desc("Alias for -source"),
304 cl::NotHidden, cl::Grouping,
305 cl::aliasopt(PrintSource));
307 static cl::opt<uint64_t>
308 StartAddress("start-address", cl::desc("Disassemble beginning at address"),
309 cl::value_desc("address"), cl::init(0), cl::cat(ObjdumpCat));
310 static cl::opt<uint64_t> StopAddress("stop-address",
311 cl::desc("Stop disassembly at address"),
312 cl::value_desc("address"),
313 cl::init(UINT64_MAX), cl::cat(ObjdumpCat));
315 cl::opt<bool> SymbolTable("syms", cl::desc("Display the symbol table"),
316 cl::cat(ObjdumpCat));
317 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"),
318 cl::NotHidden, cl::Grouping,
319 cl::aliasopt(SymbolTable));
321 cl::opt<std::string> TripleName("triple",
322 cl::desc("Target triple to disassemble for, "
323 "see -version for available targets"),
324 cl::cat(ObjdumpCat));
326 cl::opt<bool> UnwindInfo("unwind-info", cl::desc("Display unwind information"),
327 cl::cat(ObjdumpCat));
328 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"),
329 cl::NotHidden, cl::Grouping,
330 cl::aliasopt(UnwindInfo));
332 static cl::opt<bool>
333 Wide("wide", cl::desc("Ignored for compatibility with GNU objdump"),
334 cl::cat(ObjdumpCat));
335 static cl::alias WideShort("w", cl::Grouping, cl::aliasopt(Wide));
337 static cl::extrahelp
338 HelpResponse("\nPass @FILE as argument to read options from FILE.\n");
340 static StringSet<> DisasmFuncsSet;
341 static StringSet<> FoundSectionSet;
342 static StringRef ToolName;
344 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy;
346 namespace {
347 struct FilterResult {
348 // True if the section should not be skipped.
349 bool Keep;
351 // True if the index counter should be incremented, even if the section should
352 // be skipped. For example, sections may be skipped if they are not included
353 // in the --section flag, but we still want those to count toward the section
354 // count.
355 bool IncrementIndex;
357 } // namespace
359 static FilterResult checkSectionFilter(object::SectionRef S) {
360 if (FilterSections.empty())
361 return {/*Keep=*/true, /*IncrementIndex=*/true};
363 Expected<StringRef> SecNameOrErr = S.getName();
364 if (!SecNameOrErr) {
365 consumeError(SecNameOrErr.takeError());
366 return {/*Keep=*/false, /*IncrementIndex=*/false};
368 StringRef SecName = *SecNameOrErr;
370 // StringSet does not allow empty key so avoid adding sections with
371 // no name (such as the section with index 0) here.
372 if (!SecName.empty())
373 FoundSectionSet.insert(SecName);
375 // Only show the section if it's in the FilterSections list, but always
376 // increment so the indexing is stable.
377 return {/*Keep=*/is_contained(FilterSections, SecName),
378 /*IncrementIndex=*/true};
381 SectionFilter ToolSectionFilter(object::ObjectFile const &O, uint64_t *Idx) {
382 // Start at UINT64_MAX so that the first index returned after an increment is
383 // zero (after the unsigned wrap).
384 if (Idx)
385 *Idx = UINT64_MAX;
386 return SectionFilter(
387 [Idx](object::SectionRef S) {
388 FilterResult Result = checkSectionFilter(S);
389 if (Idx != nullptr && Result.IncrementIndex)
390 *Idx += 1;
391 return Result.Keep;
396 std::string getFileNameForError(const object::Archive::Child &C,
397 unsigned Index) {
398 Expected<StringRef> NameOrErr = C.getName();
399 if (NameOrErr)
400 return NameOrErr.get();
401 // If we have an error getting the name then we print the index of the archive
402 // member. Since we are already in an error state, we just ignore this error.
403 consumeError(NameOrErr.takeError());
404 return "<file index: " + std::to_string(Index) + ">";
407 void reportWarning(Twine Message, StringRef File) {
408 // Output order between errs() and outs() matters especially for archive
409 // files where the output is per member object.
410 outs().flush();
411 WithColor::warning(errs(), ToolName)
412 << "'" << File << "': " << Message << "\n";
413 errs().flush();
416 LLVM_ATTRIBUTE_NORETURN void reportError(StringRef File, Twine Message) {
417 WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n";
418 exit(1);
421 LLVM_ATTRIBUTE_NORETURN void reportError(Error E, StringRef FileName,
422 StringRef ArchiveName,
423 StringRef ArchitectureName) {
424 assert(E);
425 WithColor::error(errs(), ToolName);
426 if (ArchiveName != "")
427 errs() << ArchiveName << "(" << FileName << ")";
428 else
429 errs() << "'" << FileName << "'";
430 if (!ArchitectureName.empty())
431 errs() << " (for architecture " << ArchitectureName << ")";
432 std::string Buf;
433 raw_string_ostream OS(Buf);
434 logAllUnhandledErrors(std::move(E), OS);
435 OS.flush();
436 errs() << ": " << Buf;
437 exit(1);
440 static void reportCmdLineWarning(Twine Message) {
441 WithColor::warning(errs(), ToolName) << Message << "\n";
444 LLVM_ATTRIBUTE_NORETURN static void reportCmdLineError(Twine Message) {
445 WithColor::error(errs(), ToolName) << Message << "\n";
446 exit(1);
449 static void warnOnNoMatchForSections() {
450 SetVector<StringRef> MissingSections;
451 for (StringRef S : FilterSections) {
452 if (FoundSectionSet.count(S))
453 return;
454 // User may specify a unnamed section. Don't warn for it.
455 if (!S.empty())
456 MissingSections.insert(S);
459 // Warn only if no section in FilterSections is matched.
460 for (StringRef S : MissingSections)
461 reportCmdLineWarning("section '" + S +
462 "' mentioned in a -j/--section option, but not "
463 "found in any input file");
466 static const Target *getTarget(const ObjectFile *Obj) {
467 // Figure out the target triple.
468 Triple TheTriple("unknown-unknown-unknown");
469 if (TripleName.empty()) {
470 TheTriple = Obj->makeTriple();
471 } else {
472 TheTriple.setTriple(Triple::normalize(TripleName));
473 auto Arch = Obj->getArch();
474 if (Arch == Triple::arm || Arch == Triple::armeb)
475 Obj->setARMSubArch(TheTriple);
478 // Get the target specific parser.
479 std::string Error;
480 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple,
481 Error);
482 if (!TheTarget)
483 reportError(Obj->getFileName(), "can't find target: " + Error);
485 // Update the triple name and return the found target.
486 TripleName = TheTriple.getTriple();
487 return TheTarget;
490 bool isRelocAddressLess(RelocationRef A, RelocationRef B) {
491 return A.getOffset() < B.getOffset();
494 static Error getRelocationValueString(const RelocationRef &Rel,
495 SmallVectorImpl<char> &Result) {
496 const ObjectFile *Obj = Rel.getObject();
497 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj))
498 return getELFRelocationValueString(ELF, Rel, Result);
499 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj))
500 return getCOFFRelocationValueString(COFF, Rel, Result);
501 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj))
502 return getWasmRelocationValueString(Wasm, Rel, Result);
503 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj))
504 return getMachORelocationValueString(MachO, Rel, Result);
505 llvm_unreachable("unknown object file format");
508 /// Indicates whether this relocation should hidden when listing
509 /// relocations, usually because it is the trailing part of a multipart
510 /// relocation that will be printed as part of the leading relocation.
511 static bool getHidden(RelocationRef RelRef) {
512 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject());
513 if (!MachO)
514 return false;
516 unsigned Arch = MachO->getArch();
517 DataRefImpl Rel = RelRef.getRawDataRefImpl();
518 uint64_t Type = MachO->getRelocationType(Rel);
520 // On arches that use the generic relocations, GENERIC_RELOC_PAIR
521 // is always hidden.
522 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc)
523 return Type == MachO::GENERIC_RELOC_PAIR;
525 if (Arch == Triple::x86_64) {
526 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows
527 // an X86_64_RELOC_SUBTRACTOR.
528 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) {
529 DataRefImpl RelPrev = Rel;
530 RelPrev.d.a--;
531 uint64_t PrevType = MachO->getRelocationType(RelPrev);
532 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR)
533 return true;
537 return false;
540 namespace {
541 class SourcePrinter {
542 protected:
543 DILineInfo OldLineInfo;
544 const ObjectFile *Obj = nullptr;
545 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer;
546 // File name to file contents of source.
547 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache;
548 // Mark the line endings of the cached source.
549 std::unordered_map<std::string, std::vector<StringRef>> LineCache;
550 // Keep track of missing sources.
551 StringSet<> MissingSources;
552 // Only emit 'no debug info' warning once.
553 bool WarnedNoDebugInfo;
555 private:
556 bool cacheSource(const DILineInfo& LineInfoFile);
558 public:
559 SourcePrinter() = default;
560 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch)
561 : Obj(Obj), WarnedNoDebugInfo(false) {
562 symbolize::LLVMSymbolizer::Options SymbolizerOpts;
563 SymbolizerOpts.PrintFunctions = DILineInfoSpecifier::FunctionNameKind::None;
564 SymbolizerOpts.Demangle = false;
565 SymbolizerOpts.DefaultArch = DefaultArch;
566 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts));
568 virtual ~SourcePrinter() = default;
569 virtual void printSourceLine(raw_ostream &OS,
570 object::SectionedAddress Address,
571 StringRef ObjectFilename,
572 StringRef Delimiter = "; ");
575 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) {
576 std::unique_ptr<MemoryBuffer> Buffer;
577 if (LineInfo.Source) {
578 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source);
579 } else {
580 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName);
581 if (!BufferOrError) {
582 if (MissingSources.insert(LineInfo.FileName).second)
583 reportWarning("failed to find source " + LineInfo.FileName,
584 Obj->getFileName());
585 return false;
587 Buffer = std::move(*BufferOrError);
589 // Chomp the file to get lines
590 const char *BufferStart = Buffer->getBufferStart(),
591 *BufferEnd = Buffer->getBufferEnd();
592 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName];
593 const char *Start = BufferStart;
594 for (const char *I = BufferStart; I != BufferEnd; ++I)
595 if (*I == '\n') {
596 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r'));
597 Start = I + 1;
599 if (Start < BufferEnd)
600 Lines.emplace_back(Start, BufferEnd - Start);
601 SourceCache[LineInfo.FileName] = std::move(Buffer);
602 return true;
605 void SourcePrinter::printSourceLine(raw_ostream &OS,
606 object::SectionedAddress Address,
607 StringRef ObjectFilename,
608 StringRef Delimiter) {
609 if (!Symbolizer)
610 return;
612 DILineInfo LineInfo = DILineInfo();
613 auto ExpectedLineInfo = Symbolizer->symbolizeCode(*Obj, Address);
614 std::string ErrorMessage;
615 if (!ExpectedLineInfo)
616 ErrorMessage = toString(ExpectedLineInfo.takeError());
617 else
618 LineInfo = *ExpectedLineInfo;
620 if (LineInfo.FileName == DILineInfo::BadString) {
621 if (!WarnedNoDebugInfo) {
622 std::string Warning =
623 "failed to parse debug information for " + ObjectFilename.str();
624 if (!ErrorMessage.empty())
625 Warning += ": " + ErrorMessage;
626 reportWarning(Warning, ObjectFilename);
627 WarnedNoDebugInfo = true;
629 return;
632 if (LineInfo.Line == 0 || ((OldLineInfo.Line == LineInfo.Line) &&
633 (OldLineInfo.FileName == LineInfo.FileName)))
634 return;
636 if (PrintLines)
637 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n";
638 if (PrintSource) {
639 if (SourceCache.find(LineInfo.FileName) == SourceCache.end())
640 if (!cacheSource(LineInfo))
641 return;
642 auto LineBuffer = LineCache.find(LineInfo.FileName);
643 if (LineBuffer != LineCache.end()) {
644 if (LineInfo.Line > LineBuffer->second.size()) {
645 reportWarning(
646 formatv(
647 "debug info line number {0} exceeds the number of lines in {1}",
648 LineInfo.Line, LineInfo.FileName),
649 ObjectFilename);
650 return;
652 // Vector begins at 0, line numbers are non-zero
653 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1] << '\n';
656 OldLineInfo = LineInfo;
659 static bool isAArch64Elf(const ObjectFile *Obj) {
660 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj);
661 return Elf && Elf->getEMachine() == ELF::EM_AARCH64;
664 static bool isArmElf(const ObjectFile *Obj) {
665 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj);
666 return Elf && Elf->getEMachine() == ELF::EM_ARM;
669 static bool hasMappingSymbols(const ObjectFile *Obj) {
670 return isArmElf(Obj) || isAArch64Elf(Obj);
673 static void printRelocation(StringRef FileName, const RelocationRef &Rel,
674 uint64_t Address, bool Is64Bits) {
675 StringRef Fmt = Is64Bits ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": ";
676 SmallString<16> Name;
677 SmallString<32> Val;
678 Rel.getTypeName(Name);
679 if (Error E = getRelocationValueString(Rel, Val))
680 reportError(std::move(E), FileName);
681 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n";
684 class PrettyPrinter {
685 public:
686 virtual ~PrettyPrinter() = default;
687 virtual void printInst(MCInstPrinter &IP, const MCInst *MI,
688 ArrayRef<uint8_t> Bytes,
689 object::SectionedAddress Address, raw_ostream &OS,
690 StringRef Annot, MCSubtargetInfo const &STI,
691 SourcePrinter *SP, StringRef ObjectFilename,
692 std::vector<RelocationRef> *Rels = nullptr) {
693 if (SP && (PrintSource || PrintLines))
694 SP->printSourceLine(OS, Address, ObjectFilename);
696 size_t Start = OS.tell();
697 if (!NoLeadingAddr)
698 OS << format("%8" PRIx64 ":", Address.Address);
699 if (!NoShowRawInsn) {
700 OS << ' ';
701 dumpBytes(Bytes, OS);
704 // The output of printInst starts with a tab. Print some spaces so that
705 // the tab has 1 column and advances to the target tab stop.
706 unsigned TabStop = NoShowRawInsn ? 16 : 40;
707 unsigned Column = OS.tell() - Start;
708 OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8);
710 if (MI)
711 IP.printInst(MI, Address.Address, "", STI, OS);
712 else
713 OS << "\t<unknown>";
716 PrettyPrinter PrettyPrinterInst;
718 class HexagonPrettyPrinter : public PrettyPrinter {
719 public:
720 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address,
721 raw_ostream &OS) {
722 uint32_t opcode =
723 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0];
724 if (!NoLeadingAddr)
725 OS << format("%8" PRIx64 ":", Address);
726 if (!NoShowRawInsn) {
727 OS << "\t";
728 dumpBytes(Bytes.slice(0, 4), OS);
729 OS << format("\t%08" PRIx32, opcode);
732 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
733 object::SectionedAddress Address, raw_ostream &OS,
734 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
735 StringRef ObjectFilename,
736 std::vector<RelocationRef> *Rels) override {
737 if (SP && (PrintSource || PrintLines))
738 SP->printSourceLine(OS, Address, ObjectFilename, "");
739 if (!MI) {
740 printLead(Bytes, Address.Address, OS);
741 OS << " <unknown>";
742 return;
744 std::string Buffer;
746 raw_string_ostream TempStream(Buffer);
747 IP.printInst(MI, Address.Address, "", STI, TempStream);
749 StringRef Contents(Buffer);
750 // Split off bundle attributes
751 auto PacketBundle = Contents.rsplit('\n');
752 // Split off first instruction from the rest
753 auto HeadTail = PacketBundle.first.split('\n');
754 auto Preamble = " { ";
755 auto Separator = "";
757 // Hexagon's packets require relocations to be inline rather than
758 // clustered at the end of the packet.
759 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin();
760 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end();
761 auto PrintReloc = [&]() -> void {
762 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) {
763 if (RelCur->getOffset() == Address.Address) {
764 printRelocation(ObjectFilename, *RelCur, Address.Address, false);
765 return;
767 ++RelCur;
771 while (!HeadTail.first.empty()) {
772 OS << Separator;
773 Separator = "\n";
774 if (SP && (PrintSource || PrintLines))
775 SP->printSourceLine(OS, Address, ObjectFilename, "");
776 printLead(Bytes, Address.Address, OS);
777 OS << Preamble;
778 Preamble = " ";
779 StringRef Inst;
780 auto Duplex = HeadTail.first.split('\v');
781 if (!Duplex.second.empty()) {
782 OS << Duplex.first;
783 OS << "; ";
784 Inst = Duplex.second;
786 else
787 Inst = HeadTail.first;
788 OS << Inst;
789 HeadTail = HeadTail.second.split('\n');
790 if (HeadTail.first.empty())
791 OS << " } " << PacketBundle.second;
792 PrintReloc();
793 Bytes = Bytes.slice(4);
794 Address.Address += 4;
798 HexagonPrettyPrinter HexagonPrettyPrinterInst;
800 class AMDGCNPrettyPrinter : public PrettyPrinter {
801 public:
802 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
803 object::SectionedAddress Address, raw_ostream &OS,
804 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
805 StringRef ObjectFilename,
806 std::vector<RelocationRef> *Rels) override {
807 if (SP && (PrintSource || PrintLines))
808 SP->printSourceLine(OS, Address, ObjectFilename);
810 if (MI) {
811 SmallString<40> InstStr;
812 raw_svector_ostream IS(InstStr);
814 IP.printInst(MI, Address.Address, "", STI, IS);
816 OS << left_justify(IS.str(), 60);
817 } else {
818 // an unrecognized encoding - this is probably data so represent it
819 // using the .long directive, or .byte directive if fewer than 4 bytes
820 // remaining
821 if (Bytes.size() >= 4) {
822 OS << format("\t.long 0x%08" PRIx32 " ",
823 support::endian::read32<support::little>(Bytes.data()));
824 OS.indent(42);
825 } else {
826 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]);
827 for (unsigned int i = 1; i < Bytes.size(); i++)
828 OS << format(", 0x%02" PRIx8, Bytes[i]);
829 OS.indent(55 - (6 * Bytes.size()));
833 OS << format("// %012" PRIX64 ":", Address.Address);
834 if (Bytes.size() >= 4) {
835 // D should be casted to uint32_t here as it is passed by format to
836 // snprintf as vararg.
837 for (uint32_t D : makeArrayRef(
838 reinterpret_cast<const support::little32_t *>(Bytes.data()),
839 Bytes.size() / 4))
840 OS << format(" %08" PRIX32, D);
841 } else {
842 for (unsigned char B : Bytes)
843 OS << format(" %02" PRIX8, B);
846 if (!Annot.empty())
847 OS << " // " << Annot;
850 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst;
852 class BPFPrettyPrinter : public PrettyPrinter {
853 public:
854 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
855 object::SectionedAddress Address, raw_ostream &OS,
856 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
857 StringRef ObjectFilename,
858 std::vector<RelocationRef> *Rels) override {
859 if (SP && (PrintSource || PrintLines))
860 SP->printSourceLine(OS, Address, ObjectFilename);
861 if (!NoLeadingAddr)
862 OS << format("%8" PRId64 ":", Address.Address / 8);
863 if (!NoShowRawInsn) {
864 OS << "\t";
865 dumpBytes(Bytes, OS);
867 if (MI)
868 IP.printInst(MI, Address.Address, "", STI, OS);
869 else
870 OS << "\t<unknown>";
873 BPFPrettyPrinter BPFPrettyPrinterInst;
875 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) {
876 switch(Triple.getArch()) {
877 default:
878 return PrettyPrinterInst;
879 case Triple::hexagon:
880 return HexagonPrettyPrinterInst;
881 case Triple::amdgcn:
882 return AMDGCNPrettyPrinterInst;
883 case Triple::bpfel:
884 case Triple::bpfeb:
885 return BPFPrettyPrinterInst;
890 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) {
891 assert(Obj->isELF());
892 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj))
893 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType();
894 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj))
895 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType();
896 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj))
897 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType();
898 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj))
899 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType();
900 llvm_unreachable("Unsupported binary format");
903 template <class ELFT> static void
904 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj,
905 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
906 for (auto Symbol : Obj->getDynamicSymbolIterators()) {
907 uint8_t SymbolType = Symbol.getELFType();
908 if (SymbolType == ELF::STT_SECTION)
909 continue;
911 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName());
912 // ELFSymbolRef::getAddress() returns size instead of value for common
913 // symbols which is not desirable for disassembly output. Overriding.
914 if (SymbolType == ELF::STT_COMMON)
915 Address = Obj->getSymbol(Symbol.getRawDataRefImpl())->st_value;
917 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName());
918 if (Name.empty())
919 continue;
921 section_iterator SecI =
922 unwrapOrError(Symbol.getSection(), Obj->getFileName());
923 if (SecI == Obj->section_end())
924 continue;
926 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType);
930 static void
931 addDynamicElfSymbols(const ObjectFile *Obj,
932 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
933 assert(Obj->isELF());
934 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj))
935 addDynamicElfSymbols(Elf32LEObj, AllSymbols);
936 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj))
937 addDynamicElfSymbols(Elf64LEObj, AllSymbols);
938 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj))
939 addDynamicElfSymbols(Elf32BEObj, AllSymbols);
940 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj))
941 addDynamicElfSymbols(Elf64BEObj, AllSymbols);
942 else
943 llvm_unreachable("Unsupported binary format");
946 static void addPltEntries(const ObjectFile *Obj,
947 std::map<SectionRef, SectionSymbolsTy> &AllSymbols,
948 StringSaver &Saver) {
949 Optional<SectionRef> Plt = None;
950 for (const SectionRef &Section : Obj->sections()) {
951 Expected<StringRef> SecNameOrErr = Section.getName();
952 if (!SecNameOrErr) {
953 consumeError(SecNameOrErr.takeError());
954 continue;
956 if (*SecNameOrErr == ".plt")
957 Plt = Section;
959 if (!Plt)
960 return;
961 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) {
962 for (auto PltEntry : ElfObj->getPltAddresses()) {
963 SymbolRef Symbol(PltEntry.first, ElfObj);
964 uint8_t SymbolType = getElfSymbolType(Obj, Symbol);
966 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName());
967 if (!Name.empty())
968 AllSymbols[*Plt].emplace_back(
969 PltEntry.second, Saver.save((Name + "@plt").str()), SymbolType);
974 // Normally the disassembly output will skip blocks of zeroes. This function
975 // returns the number of zero bytes that can be skipped when dumping the
976 // disassembly of the instructions in Buf.
977 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) {
978 // Find the number of leading zeroes.
979 size_t N = 0;
980 while (N < Buf.size() && !Buf[N])
981 ++N;
983 // We may want to skip blocks of zero bytes, but unless we see
984 // at least 8 of them in a row.
985 if (N < 8)
986 return 0;
988 // We skip zeroes in multiples of 4 because do not want to truncate an
989 // instruction if it starts with a zero byte.
990 return N & ~0x3;
993 // Returns a map from sections to their relocations.
994 static std::map<SectionRef, std::vector<RelocationRef>>
995 getRelocsMap(object::ObjectFile const &Obj) {
996 std::map<SectionRef, std::vector<RelocationRef>> Ret;
997 uint64_t I = (uint64_t)-1;
998 for (SectionRef Sec : Obj.sections()) {
999 ++I;
1000 Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection();
1001 if (!RelocatedOrErr)
1002 reportError(Obj.getFileName(),
1003 "section (" + Twine(I) +
1004 "): failed to get a relocated section: " +
1005 toString(RelocatedOrErr.takeError()));
1007 section_iterator Relocated = *RelocatedOrErr;
1008 if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep)
1009 continue;
1010 std::vector<RelocationRef> &V = Ret[*Relocated];
1011 for (const RelocationRef &R : Sec.relocations())
1012 V.push_back(R);
1013 // Sort relocations by address.
1014 llvm::stable_sort(V, isRelocAddressLess);
1016 return Ret;
1019 // Used for --adjust-vma to check if address should be adjusted by the
1020 // specified value for a given section.
1021 // For ELF we do not adjust non-allocatable sections like debug ones,
1022 // because they are not loadable.
1023 // TODO: implement for other file formats.
1024 static bool shouldAdjustVA(const SectionRef &Section) {
1025 const ObjectFile *Obj = Section.getObject();
1026 if (isa<object::ELFObjectFileBase>(Obj))
1027 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC;
1028 return false;
1032 typedef std::pair<uint64_t, char> MappingSymbolPair;
1033 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols,
1034 uint64_t Address) {
1035 auto It =
1036 partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) {
1037 return Val.first <= Address;
1039 // Return zero for any address before the first mapping symbol; this means
1040 // we should use the default disassembly mode, depending on the target.
1041 if (It == MappingSymbols.begin())
1042 return '\x00';
1043 return (It - 1)->second;
1046 static uint64_t
1047 dumpARMELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End,
1048 const ObjectFile *Obj, ArrayRef<uint8_t> Bytes,
1049 ArrayRef<MappingSymbolPair> MappingSymbols) {
1050 support::endianness Endian =
1051 Obj->isLittleEndian() ? support::little : support::big;
1052 while (Index < End) {
1053 outs() << format("%8" PRIx64 ":", SectionAddr + Index);
1054 outs() << "\t";
1055 if (Index + 4 <= End) {
1056 dumpBytes(Bytes.slice(Index, 4), outs());
1057 outs() << "\t.word\t"
1058 << format_hex(
1059 support::endian::read32(Bytes.data() + Index, Endian), 10);
1060 Index += 4;
1061 } else if (Index + 2 <= End) {
1062 dumpBytes(Bytes.slice(Index, 2), outs());
1063 outs() << "\t\t.short\t"
1064 << format_hex(
1065 support::endian::read16(Bytes.data() + Index, Endian), 6);
1066 Index += 2;
1067 } else {
1068 dumpBytes(Bytes.slice(Index, 1), outs());
1069 outs() << "\t\t.byte\t" << format_hex(Bytes[0], 4);
1070 ++Index;
1072 outs() << "\n";
1073 if (getMappingSymbolKind(MappingSymbols, Index) != 'd')
1074 break;
1076 return Index;
1079 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End,
1080 ArrayRef<uint8_t> Bytes) {
1081 // print out data up to 8 bytes at a time in hex and ascii
1082 uint8_t AsciiData[9] = {'\0'};
1083 uint8_t Byte;
1084 int NumBytes = 0;
1086 for (; Index < End; ++Index) {
1087 if (NumBytes == 0)
1088 outs() << format("%8" PRIx64 ":", SectionAddr + Index);
1089 Byte = Bytes.slice(Index)[0];
1090 outs() << format(" %02x", Byte);
1091 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.';
1093 uint8_t IndentOffset = 0;
1094 NumBytes++;
1095 if (Index == End - 1 || NumBytes > 8) {
1096 // Indent the space for less than 8 bytes data.
1097 // 2 spaces for byte and one for space between bytes
1098 IndentOffset = 3 * (8 - NumBytes);
1099 for (int Excess = NumBytes; Excess < 8; Excess++)
1100 AsciiData[Excess] = '\0';
1101 NumBytes = 8;
1103 if (NumBytes == 8) {
1104 AsciiData[8] = '\0';
1105 outs() << std::string(IndentOffset, ' ') << " ";
1106 outs() << reinterpret_cast<char *>(AsciiData);
1107 outs() << '\n';
1108 NumBytes = 0;
1113 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj,
1114 MCContext &Ctx, MCDisassembler *PrimaryDisAsm,
1115 MCDisassembler *SecondaryDisAsm,
1116 const MCInstrAnalysis *MIA, MCInstPrinter *IP,
1117 const MCSubtargetInfo *PrimarySTI,
1118 const MCSubtargetInfo *SecondarySTI,
1119 PrettyPrinter &PIP,
1120 SourcePrinter &SP, bool InlineRelocs) {
1121 const MCSubtargetInfo *STI = PrimarySTI;
1122 MCDisassembler *DisAsm = PrimaryDisAsm;
1123 bool PrimaryIsThumb = false;
1124 if (isArmElf(Obj))
1125 PrimaryIsThumb = STI->checkFeatures("+thumb-mode");
1127 std::map<SectionRef, std::vector<RelocationRef>> RelocMap;
1128 if (InlineRelocs)
1129 RelocMap = getRelocsMap(*Obj);
1130 bool Is64Bits = Obj->getBytesInAddress() > 4;
1132 // Create a mapping from virtual address to symbol name. This is used to
1133 // pretty print the symbols while disassembling.
1134 std::map<SectionRef, SectionSymbolsTy> AllSymbols;
1135 SectionSymbolsTy AbsoluteSymbols;
1136 const StringRef FileName = Obj->getFileName();
1137 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj);
1138 for (const SymbolRef &Symbol : Obj->symbols()) {
1139 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName);
1141 StringRef Name = unwrapOrError(Symbol.getName(), FileName);
1142 if (Name.empty())
1143 continue;
1145 uint8_t SymbolType = ELF::STT_NOTYPE;
1146 if (Obj->isELF()) {
1147 SymbolType = getElfSymbolType(Obj, Symbol);
1148 if (SymbolType == ELF::STT_SECTION)
1149 continue;
1152 // Don't ask a Mach-O STAB symbol for its section unless you know that
1153 // STAB symbol's section field refers to a valid section index. Otherwise
1154 // the symbol may error trying to load a section that does not exist.
1155 if (MachO) {
1156 DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
1157 uint8_t NType = (MachO->is64Bit() ?
1158 MachO->getSymbol64TableEntry(SymDRI).n_type:
1159 MachO->getSymbolTableEntry(SymDRI).n_type);
1160 if (NType & MachO::N_STAB)
1161 continue;
1164 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName);
1165 if (SecI != Obj->section_end())
1166 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType);
1167 else
1168 AbsoluteSymbols.emplace_back(Address, Name, SymbolType);
1170 if (AllSymbols.empty() && Obj->isELF())
1171 addDynamicElfSymbols(Obj, AllSymbols);
1173 BumpPtrAllocator A;
1174 StringSaver Saver(A);
1175 addPltEntries(Obj, AllSymbols, Saver);
1177 // Create a mapping from virtual address to section.
1178 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses;
1179 for (SectionRef Sec : Obj->sections())
1180 SectionAddresses.emplace_back(Sec.getAddress(), Sec);
1181 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end());
1183 // Linked executables (.exe and .dll files) typically don't include a real
1184 // symbol table but they might contain an export table.
1185 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) {
1186 for (const auto &ExportEntry : COFFObj->export_directories()) {
1187 StringRef Name;
1188 if (std::error_code EC = ExportEntry.getSymbolName(Name))
1189 reportError(errorCodeToError(EC), Obj->getFileName());
1190 if (Name.empty())
1191 continue;
1193 uint32_t RVA;
1194 if (std::error_code EC = ExportEntry.getExportRVA(RVA))
1195 reportError(errorCodeToError(EC), Obj->getFileName());
1197 uint64_t VA = COFFObj->getImageBase() + RVA;
1198 auto Sec = partition_point(
1199 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) {
1200 return O.first <= VA;
1202 if (Sec != SectionAddresses.begin()) {
1203 --Sec;
1204 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE);
1205 } else
1206 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE);
1210 // Sort all the symbols, this allows us to use a simple binary search to find
1211 // a symbol near an address.
1212 StringSet<> FoundDisasmFuncsSet;
1213 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols)
1214 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end());
1215 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end());
1217 for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1218 if (FilterSections.empty() && !DisassembleAll &&
1219 (!Section.isText() || Section.isVirtual()))
1220 continue;
1222 uint64_t SectionAddr = Section.getAddress();
1223 uint64_t SectSize = Section.getSize();
1224 if (!SectSize)
1225 continue;
1227 // Get the list of all the symbols in this section.
1228 SectionSymbolsTy &Symbols = AllSymbols[Section];
1229 std::vector<MappingSymbolPair> MappingSymbols;
1230 if (hasMappingSymbols(Obj)) {
1231 for (const auto &Symb : Symbols) {
1232 uint64_t Address = std::get<0>(Symb);
1233 StringRef Name = std::get<1>(Symb);
1234 if (Name.startswith("$d"))
1235 MappingSymbols.emplace_back(Address - SectionAddr, 'd');
1236 if (Name.startswith("$x"))
1237 MappingSymbols.emplace_back(Address - SectionAddr, 'x');
1238 if (Name.startswith("$a"))
1239 MappingSymbols.emplace_back(Address - SectionAddr, 'a');
1240 if (Name.startswith("$t"))
1241 MappingSymbols.emplace_back(Address - SectionAddr, 't');
1245 llvm::sort(MappingSymbols);
1247 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) {
1248 // AMDGPU disassembler uses symbolizer for printing labels
1249 std::unique_ptr<MCRelocationInfo> RelInfo(
1250 TheTarget->createMCRelocationInfo(TripleName, Ctx));
1251 if (RelInfo) {
1252 std::unique_ptr<MCSymbolizer> Symbolizer(
1253 TheTarget->createMCSymbolizer(
1254 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo)));
1255 DisAsm->setSymbolizer(std::move(Symbolizer));
1259 StringRef SegmentName = "";
1260 if (MachO) {
1261 DataRefImpl DR = Section.getRawDataRefImpl();
1262 SegmentName = MachO->getSectionFinalSegmentName(DR);
1265 StringRef SectionName = unwrapOrError(Section.getName(), Obj->getFileName());
1266 // If the section has no symbol at the start, just insert a dummy one.
1267 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) {
1268 Symbols.insert(
1269 Symbols.begin(),
1270 std::make_tuple(SectionAddr, SectionName,
1271 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT));
1274 SmallString<40> Comments;
1275 raw_svector_ostream CommentStream(Comments);
1277 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef(
1278 unwrapOrError(Section.getContents(), Obj->getFileName()));
1280 uint64_t VMAAdjustment = 0;
1281 if (shouldAdjustVA(Section))
1282 VMAAdjustment = AdjustVMA;
1284 uint64_t Size;
1285 uint64_t Index;
1286 bool PrintedSection = false;
1287 std::vector<RelocationRef> Rels = RelocMap[Section];
1288 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin();
1289 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end();
1290 // Disassemble symbol by symbol.
1291 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) {
1292 std::string SymbolName = std::get<1>(Symbols[SI]).str();
1293 if (Demangle)
1294 SymbolName = demangle(SymbolName);
1296 // Skip if --disassemble-functions is not empty and the symbol is not in
1297 // the list.
1298 if (!DisasmFuncsSet.empty() && !DisasmFuncsSet.count(SymbolName))
1299 continue;
1301 uint64_t Start = std::get<0>(Symbols[SI]);
1302 if (Start < SectionAddr || StopAddress <= Start)
1303 continue;
1304 else
1305 FoundDisasmFuncsSet.insert(SymbolName);
1307 // The end is the section end, the beginning of the next symbol, or
1308 // --stop-address.
1309 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress);
1310 if (SI + 1 < SE)
1311 End = std::min(End, std::get<0>(Symbols[SI + 1]));
1312 if (Start >= End || End <= StartAddress)
1313 continue;
1314 Start -= SectionAddr;
1315 End -= SectionAddr;
1317 if (!PrintedSection) {
1318 PrintedSection = true;
1319 outs() << "\nDisassembly of section ";
1320 if (!SegmentName.empty())
1321 outs() << SegmentName << ",";
1322 outs() << SectionName << ":\n";
1325 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) {
1326 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) {
1327 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes)
1328 Start += 256;
1330 if (SI == SE - 1 ||
1331 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) {
1332 // cut trailing zeroes at the end of kernel
1333 // cut up to 256 bytes
1334 const uint64_t EndAlign = 256;
1335 const auto Limit = End - (std::min)(EndAlign, End - Start);
1336 while (End > Limit &&
1337 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0)
1338 End -= 4;
1342 outs() << '\n';
1343 if (!NoLeadingAddr)
1344 outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ",
1345 SectionAddr + Start + VMAAdjustment);
1347 outs() << SymbolName << ":\n";
1349 // Don't print raw contents of a virtual section. A virtual section
1350 // doesn't have any contents in the file.
1351 if (Section.isVirtual()) {
1352 outs() << "...\n";
1353 continue;
1356 // Some targets (like WebAssembly) have a special prelude at the start
1357 // of each symbol.
1358 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start),
1359 SectionAddr + Start, CommentStream);
1360 Start += Size;
1362 Index = Start;
1363 if (SectionAddr < StartAddress)
1364 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr);
1366 // If there is a data/common symbol inside an ELF text section and we are
1367 // only disassembling text (applicable all architectures), we are in a
1368 // situation where we must print the data and not disassemble it.
1369 if (Obj->isELF() && !DisassembleAll && Section.isText()) {
1370 uint8_t SymTy = std::get<2>(Symbols[SI]);
1371 if (SymTy == ELF::STT_OBJECT || SymTy == ELF::STT_COMMON) {
1372 dumpELFData(SectionAddr, Index, End, Bytes);
1373 Index = End;
1377 bool CheckARMELFData = hasMappingSymbols(Obj) &&
1378 std::get<2>(Symbols[SI]) != ELF::STT_OBJECT &&
1379 !DisassembleAll;
1380 while (Index < End) {
1381 // ARM and AArch64 ELF binaries can interleave data and text in the
1382 // same section. We rely on the markers introduced to understand what
1383 // we need to dump. If the data marker is within a function, it is
1384 // denoted as a word/short etc.
1385 if (CheckARMELFData &&
1386 getMappingSymbolKind(MappingSymbols, Index) == 'd') {
1387 Index = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes,
1388 MappingSymbols);
1389 continue;
1392 // When -z or --disassemble-zeroes are given we always dissasemble
1393 // them. Otherwise we might want to skip zero bytes we see.
1394 if (!DisassembleZeroes) {
1395 uint64_t MaxOffset = End - Index;
1396 // For -reloc: print zero blocks patched by relocations, so that
1397 // relocations can be shown in the dump.
1398 if (RelCur != RelEnd)
1399 MaxOffset = RelCur->getOffset() - Index;
1401 if (size_t N =
1402 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) {
1403 outs() << "\t\t..." << '\n';
1404 Index += N;
1405 continue;
1409 if (SecondarySTI) {
1410 if (getMappingSymbolKind(MappingSymbols, Index) == 'a') {
1411 STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI;
1412 DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm;
1413 } else if (getMappingSymbolKind(MappingSymbols, Index) == 't') {
1414 STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI;
1415 DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm;
1419 // Disassemble a real instruction or a data when disassemble all is
1420 // provided
1421 MCInst Inst;
1422 bool Disassembled = DisAsm->getInstruction(
1423 Inst, Size, Bytes.slice(Index), SectionAddr + Index, CommentStream);
1424 if (Size == 0)
1425 Size = 1;
1427 PIP.printInst(*IP, Disassembled ? &Inst : nullptr,
1428 Bytes.slice(Index, Size),
1429 {SectionAddr + Index + VMAAdjustment, Section.getIndex()},
1430 outs(), "", *STI, &SP, Obj->getFileName(), &Rels);
1431 outs() << CommentStream.str();
1432 Comments.clear();
1434 // Try to resolve the target of a call, tail call, etc. to a specific
1435 // symbol.
1436 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) ||
1437 MIA->isConditionalBranch(Inst))) {
1438 uint64_t Target;
1439 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) {
1440 // In a relocatable object, the target's section must reside in
1441 // the same section as the call instruction or it is accessed
1442 // through a relocation.
1444 // In a non-relocatable object, the target may be in any section.
1446 // N.B. We don't walk the relocations in the relocatable case yet.
1447 auto *TargetSectionSymbols = &Symbols;
1448 if (!Obj->isRelocatableObject()) {
1449 auto It = partition_point(
1450 SectionAddresses,
1451 [=](const std::pair<uint64_t, SectionRef> &O) {
1452 return O.first <= Target;
1454 if (It != SectionAddresses.begin()) {
1455 --It;
1456 TargetSectionSymbols = &AllSymbols[It->second];
1457 } else {
1458 TargetSectionSymbols = &AbsoluteSymbols;
1462 // Find the last symbol in the section whose offset is less than
1463 // or equal to the target. If there isn't a section that contains
1464 // the target, find the nearest preceding absolute symbol.
1465 auto TargetSym = partition_point(
1466 *TargetSectionSymbols,
1467 [=](const std::tuple<uint64_t, StringRef, uint8_t> &O) {
1468 return std::get<0>(O) <= Target;
1470 if (TargetSym == TargetSectionSymbols->begin()) {
1471 TargetSectionSymbols = &AbsoluteSymbols;
1472 TargetSym = partition_point(
1473 AbsoluteSymbols,
1474 [=](const std::tuple<uint64_t, StringRef, uint8_t> &O) {
1475 return std::get<0>(O) <= Target;
1478 if (TargetSym != TargetSectionSymbols->begin()) {
1479 --TargetSym;
1480 uint64_t TargetAddress = std::get<0>(*TargetSym);
1481 StringRef TargetName = std::get<1>(*TargetSym);
1482 outs() << " <" << TargetName;
1483 uint64_t Disp = Target - TargetAddress;
1484 if (Disp)
1485 outs() << "+0x" << Twine::utohexstr(Disp);
1486 outs() << '>';
1490 outs() << "\n";
1492 // Hexagon does this in pretty printer
1493 if (Obj->getArch() != Triple::hexagon) {
1494 // Print relocation for instruction.
1495 while (RelCur != RelEnd) {
1496 uint64_t Offset = RelCur->getOffset();
1497 // If this relocation is hidden, skip it.
1498 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) {
1499 ++RelCur;
1500 continue;
1503 // Stop when RelCur's offset is past the current instruction.
1504 if (Offset >= Index + Size)
1505 break;
1507 // When --adjust-vma is used, update the address printed.
1508 if (RelCur->getSymbol() != Obj->symbol_end()) {
1509 Expected<section_iterator> SymSI =
1510 RelCur->getSymbol()->getSection();
1511 if (SymSI && *SymSI != Obj->section_end() &&
1512 shouldAdjustVA(**SymSI))
1513 Offset += AdjustVMA;
1516 printRelocation(Obj->getFileName(), *RelCur, SectionAddr + Offset,
1517 Is64Bits);
1518 ++RelCur;
1522 Index += Size;
1526 StringSet<> MissingDisasmFuncsSet =
1527 set_difference(DisasmFuncsSet, FoundDisasmFuncsSet);
1528 for (StringRef MissingDisasmFunc : MissingDisasmFuncsSet.keys())
1529 reportWarning("failed to disassemble missing function " + MissingDisasmFunc,
1530 FileName);
1533 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) {
1534 const Target *TheTarget = getTarget(Obj);
1536 // Package up features to be passed to target/subtarget
1537 SubtargetFeatures Features = Obj->getFeatures();
1538 if (!MAttrs.empty())
1539 for (unsigned I = 0; I != MAttrs.size(); ++I)
1540 Features.AddFeature(MAttrs[I]);
1542 std::unique_ptr<const MCRegisterInfo> MRI(
1543 TheTarget->createMCRegInfo(TripleName));
1544 if (!MRI)
1545 reportError(Obj->getFileName(),
1546 "no register info for target " + TripleName);
1548 // Set up disassembler.
1549 MCTargetOptions MCOptions;
1550 std::unique_ptr<const MCAsmInfo> AsmInfo(
1551 TheTarget->createMCAsmInfo(*MRI, TripleName, MCOptions));
1552 if (!AsmInfo)
1553 reportError(Obj->getFileName(),
1554 "no assembly info for target " + TripleName);
1555 std::unique_ptr<const MCSubtargetInfo> STI(
1556 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString()));
1557 if (!STI)
1558 reportError(Obj->getFileName(),
1559 "no subtarget info for target " + TripleName);
1560 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo());
1561 if (!MII)
1562 reportError(Obj->getFileName(),
1563 "no instruction info for target " + TripleName);
1564 MCObjectFileInfo MOFI;
1565 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI);
1566 // FIXME: for now initialize MCObjectFileInfo with default values
1567 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx);
1569 std::unique_ptr<MCDisassembler> DisAsm(
1570 TheTarget->createMCDisassembler(*STI, Ctx));
1571 if (!DisAsm)
1572 reportError(Obj->getFileName(), "no disassembler for target " + TripleName);
1574 // If we have an ARM object file, we need a second disassembler, because
1575 // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode.
1576 // We use mapping symbols to switch between the two assemblers, where
1577 // appropriate.
1578 std::unique_ptr<MCDisassembler> SecondaryDisAsm;
1579 std::unique_ptr<const MCSubtargetInfo> SecondarySTI;
1580 if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) {
1581 if (STI->checkFeatures("+thumb-mode"))
1582 Features.AddFeature("-thumb-mode");
1583 else
1584 Features.AddFeature("+thumb-mode");
1585 SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU,
1586 Features.getString()));
1587 SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx));
1590 std::unique_ptr<const MCInstrAnalysis> MIA(
1591 TheTarget->createMCInstrAnalysis(MII.get()));
1593 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
1594 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
1595 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI));
1596 if (!IP)
1597 reportError(Obj->getFileName(),
1598 "no instruction printer for target " + TripleName);
1599 IP->setPrintImmHex(PrintImmHex);
1601 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName));
1602 SourcePrinter SP(Obj, TheTarget->getName());
1604 for (StringRef Opt : DisassemblerOptions)
1605 if (!IP->applyTargetSpecificCLOption(Opt))
1606 reportError(Obj->getFileName(),
1607 "Unrecognized disassembler option: " + Opt);
1609 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(),
1610 MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP,
1611 SP, InlineRelocs);
1614 void printRelocations(const ObjectFile *Obj) {
1615 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 :
1616 "%08" PRIx64;
1617 // Regular objdump doesn't print relocations in non-relocatable object
1618 // files.
1619 if (!Obj->isRelocatableObject())
1620 return;
1622 // Build a mapping from relocation target to a vector of relocation
1623 // sections. Usually, there is an only one relocation section for
1624 // each relocated section.
1625 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec;
1626 uint64_t Ndx;
1627 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Ndx)) {
1628 if (Section.relocation_begin() == Section.relocation_end())
1629 continue;
1630 Expected<section_iterator> SecOrErr = Section.getRelocatedSection();
1631 if (!SecOrErr)
1632 reportError(Obj->getFileName(),
1633 "section (" + Twine(Ndx) +
1634 "): unable to get a relocation target: " +
1635 toString(SecOrErr.takeError()));
1636 SecToRelSec[**SecOrErr].push_back(Section);
1639 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) {
1640 StringRef SecName = unwrapOrError(P.first.getName(), Obj->getFileName());
1641 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n";
1643 for (SectionRef Section : P.second) {
1644 for (const RelocationRef &Reloc : Section.relocations()) {
1645 uint64_t Address = Reloc.getOffset();
1646 SmallString<32> RelocName;
1647 SmallString<32> ValueStr;
1648 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc))
1649 continue;
1650 Reloc.getTypeName(RelocName);
1651 if (Error E = getRelocationValueString(Reloc, ValueStr))
1652 reportError(std::move(E), Obj->getFileName());
1654 outs() << format(Fmt.data(), Address) << " " << RelocName << " "
1655 << ValueStr << "\n";
1658 outs() << "\n";
1662 void printDynamicRelocations(const ObjectFile *Obj) {
1663 // For the moment, this option is for ELF only
1664 if (!Obj->isELF())
1665 return;
1667 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj);
1668 if (!Elf || Elf->getEType() != ELF::ET_DYN) {
1669 reportError(Obj->getFileName(), "not a dynamic object");
1670 return;
1673 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections();
1674 if (DynRelSec.empty())
1675 return;
1677 outs() << "DYNAMIC RELOCATION RECORDS\n";
1678 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
1679 for (const SectionRef &Section : DynRelSec)
1680 for (const RelocationRef &Reloc : Section.relocations()) {
1681 uint64_t Address = Reloc.getOffset();
1682 SmallString<32> RelocName;
1683 SmallString<32> ValueStr;
1684 Reloc.getTypeName(RelocName);
1685 if (Error E = getRelocationValueString(Reloc, ValueStr))
1686 reportError(std::move(E), Obj->getFileName());
1687 outs() << format(Fmt.data(), Address) << " " << RelocName << " "
1688 << ValueStr << "\n";
1692 // Returns true if we need to show LMA column when dumping section headers. We
1693 // show it only when the platform is ELF and either we have at least one section
1694 // whose VMA and LMA are different and/or when --show-lma flag is used.
1695 static bool shouldDisplayLMA(const ObjectFile *Obj) {
1696 if (!Obj->isELF())
1697 return false;
1698 for (const SectionRef &S : ToolSectionFilter(*Obj))
1699 if (S.getAddress() != getELFSectionLMA(S))
1700 return true;
1701 return ShowLMA;
1704 static size_t getMaxSectionNameWidth(const ObjectFile *Obj) {
1705 // Default column width for names is 13 even if no names are that long.
1706 size_t MaxWidth = 13;
1707 for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1708 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
1709 MaxWidth = std::max(MaxWidth, Name.size());
1711 return MaxWidth;
1714 void printSectionHeaders(const ObjectFile *Obj) {
1715 size_t NameWidth = getMaxSectionNameWidth(Obj);
1716 size_t AddressWidth = 2 * Obj->getBytesInAddress();
1717 bool HasLMAColumn = shouldDisplayLMA(Obj);
1718 if (HasLMAColumn)
1719 outs() << "Sections:\n"
1720 "Idx "
1721 << left_justify("Name", NameWidth) << " Size "
1722 << left_justify("VMA", AddressWidth) << " "
1723 << left_justify("LMA", AddressWidth) << " Type\n";
1724 else
1725 outs() << "Sections:\n"
1726 "Idx "
1727 << left_justify("Name", NameWidth) << " Size "
1728 << left_justify("VMA", AddressWidth) << " Type\n";
1730 uint64_t Idx;
1731 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Idx)) {
1732 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
1733 uint64_t VMA = Section.getAddress();
1734 if (shouldAdjustVA(Section))
1735 VMA += AdjustVMA;
1737 uint64_t Size = Section.getSize();
1739 std::string Type = Section.isText() ? "TEXT" : "";
1740 if (Section.isData())
1741 Type += Type.empty() ? "DATA" : " DATA";
1742 if (Section.isBSS())
1743 Type += Type.empty() ? "BSS" : " BSS";
1745 if (HasLMAColumn)
1746 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth,
1747 Name.str().c_str(), Size)
1748 << format_hex_no_prefix(VMA, AddressWidth) << " "
1749 << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth)
1750 << " " << Type << "\n";
1751 else
1752 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth,
1753 Name.str().c_str(), Size)
1754 << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n";
1756 outs() << "\n";
1759 void printSectionContents(const ObjectFile *Obj) {
1760 for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1761 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
1762 uint64_t BaseAddr = Section.getAddress();
1763 uint64_t Size = Section.getSize();
1764 if (!Size)
1765 continue;
1767 outs() << "Contents of section " << Name << ":\n";
1768 if (Section.isBSS()) {
1769 outs() << format("<skipping contents of bss section at [%04" PRIx64
1770 ", %04" PRIx64 ")>\n",
1771 BaseAddr, BaseAddr + Size);
1772 continue;
1775 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName());
1777 // Dump out the content as hex and printable ascii characters.
1778 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) {
1779 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr);
1780 // Dump line of hex.
1781 for (std::size_t I = 0; I < 16; ++I) {
1782 if (I != 0 && I % 4 == 0)
1783 outs() << ' ';
1784 if (Addr + I < End)
1785 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true)
1786 << hexdigit(Contents[Addr + I] & 0xF, true);
1787 else
1788 outs() << " ";
1790 // Print ascii.
1791 outs() << " ";
1792 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) {
1793 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF))
1794 outs() << Contents[Addr + I];
1795 else
1796 outs() << ".";
1798 outs() << "\n";
1803 void printSymbolTable(const ObjectFile *O, StringRef ArchiveName,
1804 StringRef ArchitectureName) {
1805 outs() << "SYMBOL TABLE:\n";
1807 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) {
1808 printCOFFSymbolTable(Coff);
1809 return;
1812 const StringRef FileName = O->getFileName();
1813 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(O);
1814 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) {
1815 const SymbolRef &Symbol = *I;
1816 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName, ArchiveName,
1817 ArchitectureName);
1818 if ((Address < StartAddress) || (Address > StopAddress))
1819 continue;
1820 SymbolRef::Type Type = unwrapOrError(Symbol.getType(), FileName,
1821 ArchiveName, ArchitectureName);
1822 uint32_t Flags = Symbol.getFlags();
1824 // Don't ask a Mach-O STAB symbol for its section unless you know that
1825 // STAB symbol's section field refers to a valid section index. Otherwise
1826 // the symbol may error trying to load a section that does not exist.
1827 bool isSTAB = false;
1828 if (MachO) {
1829 DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
1830 uint8_t NType = (MachO->is64Bit() ?
1831 MachO->getSymbol64TableEntry(SymDRI).n_type:
1832 MachO->getSymbolTableEntry(SymDRI).n_type);
1833 if (NType & MachO::N_STAB)
1834 isSTAB = true;
1836 section_iterator Section = isSTAB ? O->section_end() :
1837 unwrapOrError(Symbol.getSection(), FileName,
1838 ArchiveName, ArchitectureName);
1840 StringRef Name;
1841 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) {
1842 if (Expected<StringRef> NameOrErr = Section->getName())
1843 Name = *NameOrErr;
1844 else
1845 consumeError(NameOrErr.takeError());
1847 } else {
1848 Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName,
1849 ArchitectureName);
1852 bool Global = Flags & SymbolRef::SF_Global;
1853 bool Weak = Flags & SymbolRef::SF_Weak;
1854 bool Absolute = Flags & SymbolRef::SF_Absolute;
1855 bool Common = Flags & SymbolRef::SF_Common;
1856 bool Hidden = Flags & SymbolRef::SF_Hidden;
1858 char GlobLoc = ' ';
1859 if (Type != SymbolRef::ST_Unknown)
1860 GlobLoc = Global ? 'g' : 'l';
1861 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File)
1862 ? 'd' : ' ';
1863 char FileFunc = ' ';
1864 if (Type == SymbolRef::ST_File)
1865 FileFunc = 'f';
1866 else if (Type == SymbolRef::ST_Function)
1867 FileFunc = 'F';
1868 else if (Type == SymbolRef::ST_Data)
1869 FileFunc = 'O';
1871 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 :
1872 "%08" PRIx64;
1874 outs() << format(Fmt, Address) << " "
1875 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' '
1876 << (Weak ? 'w' : ' ') // Weak?
1877 << ' ' // Constructor. Not supported yet.
1878 << ' ' // Warning. Not supported yet.
1879 << ' ' // Indirect reference to another symbol.
1880 << Debug // Debugging (d) or dynamic (D) symbol.
1881 << FileFunc // Name of function (F), file (f) or object (O).
1882 << ' ';
1883 if (Absolute) {
1884 outs() << "*ABS*";
1885 } else if (Common) {
1886 outs() << "*COM*";
1887 } else if (Section == O->section_end()) {
1888 outs() << "*UND*";
1889 } else {
1890 if (const MachOObjectFile *MachO =
1891 dyn_cast<const MachOObjectFile>(O)) {
1892 DataRefImpl DR = Section->getRawDataRefImpl();
1893 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR);
1894 outs() << SegmentName << ",";
1896 StringRef SectionName =
1897 unwrapOrError(Section->getName(), O->getFileName());
1898 outs() << SectionName;
1901 if (Common || isa<ELFObjectFileBase>(O)) {
1902 uint64_t Val =
1903 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize();
1904 outs() << format("\t%08" PRIx64, Val);
1907 if (isa<ELFObjectFileBase>(O)) {
1908 uint8_t Other = ELFSymbolRef(Symbol).getOther();
1909 switch (Other) {
1910 case ELF::STV_DEFAULT:
1911 break;
1912 case ELF::STV_INTERNAL:
1913 outs() << " .internal";
1914 break;
1915 case ELF::STV_HIDDEN:
1916 outs() << " .hidden";
1917 break;
1918 case ELF::STV_PROTECTED:
1919 outs() << " .protected";
1920 break;
1921 default:
1922 outs() << format(" 0x%02x", Other);
1923 break;
1925 } else if (Hidden) {
1926 outs() << " .hidden";
1929 if (Demangle)
1930 outs() << ' ' << demangle(Name) << '\n';
1931 else
1932 outs() << ' ' << Name << '\n';
1936 static void printUnwindInfo(const ObjectFile *O) {
1937 outs() << "Unwind info:\n\n";
1939 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O))
1940 printCOFFUnwindInfo(Coff);
1941 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O))
1942 printMachOUnwindInfo(MachO);
1943 else
1944 // TODO: Extract DWARF dump tool to objdump.
1945 WithColor::error(errs(), ToolName)
1946 << "This operation is only currently supported "
1947 "for COFF and MachO object files.\n";
1950 /// Dump the raw contents of the __clangast section so the output can be piped
1951 /// into llvm-bcanalyzer.
1952 void printRawClangAST(const ObjectFile *Obj) {
1953 if (outs().is_displayed()) {
1954 WithColor::error(errs(), ToolName)
1955 << "The -raw-clang-ast option will dump the raw binary contents of "
1956 "the clang ast section.\n"
1957 "Please redirect the output to a file or another program such as "
1958 "llvm-bcanalyzer.\n";
1959 return;
1962 StringRef ClangASTSectionName("__clangast");
1963 if (isa<COFFObjectFile>(Obj)) {
1964 ClangASTSectionName = "clangast";
1967 Optional<object::SectionRef> ClangASTSection;
1968 for (auto Sec : ToolSectionFilter(*Obj)) {
1969 StringRef Name;
1970 if (Expected<StringRef> NameOrErr = Sec.getName())
1971 Name = *NameOrErr;
1972 else
1973 consumeError(NameOrErr.takeError());
1975 if (Name == ClangASTSectionName) {
1976 ClangASTSection = Sec;
1977 break;
1980 if (!ClangASTSection)
1981 return;
1983 StringRef ClangASTContents = unwrapOrError(
1984 ClangASTSection.getValue().getContents(), Obj->getFileName());
1985 outs().write(ClangASTContents.data(), ClangASTContents.size());
1988 static void printFaultMaps(const ObjectFile *Obj) {
1989 StringRef FaultMapSectionName;
1991 if (isa<ELFObjectFileBase>(Obj)) {
1992 FaultMapSectionName = ".llvm_faultmaps";
1993 } else if (isa<MachOObjectFile>(Obj)) {
1994 FaultMapSectionName = "__llvm_faultmaps";
1995 } else {
1996 WithColor::error(errs(), ToolName)
1997 << "This operation is only currently supported "
1998 "for ELF and Mach-O executable files.\n";
1999 return;
2002 Optional<object::SectionRef> FaultMapSection;
2004 for (auto Sec : ToolSectionFilter(*Obj)) {
2005 StringRef Name;
2006 if (Expected<StringRef> NameOrErr = Sec.getName())
2007 Name = *NameOrErr;
2008 else
2009 consumeError(NameOrErr.takeError());
2011 if (Name == FaultMapSectionName) {
2012 FaultMapSection = Sec;
2013 break;
2017 outs() << "FaultMap table:\n";
2019 if (!FaultMapSection.hasValue()) {
2020 outs() << "<not found>\n";
2021 return;
2024 StringRef FaultMapContents =
2025 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName());
2026 FaultMapParser FMP(FaultMapContents.bytes_begin(),
2027 FaultMapContents.bytes_end());
2029 outs() << FMP;
2032 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) {
2033 if (O->isELF()) {
2034 printELFFileHeader(O);
2035 printELFDynamicSection(O);
2036 printELFSymbolVersionInfo(O);
2037 return;
2039 if (O->isCOFF())
2040 return printCOFFFileHeader(O);
2041 if (O->isWasm())
2042 return printWasmFileHeader(O);
2043 if (O->isMachO()) {
2044 printMachOFileHeader(O);
2045 if (!OnlyFirst)
2046 printMachOLoadCommands(O);
2047 return;
2049 reportError(O->getFileName(), "Invalid/Unsupported object file format");
2052 static void printFileHeaders(const ObjectFile *O) {
2053 if (!O->isELF() && !O->isCOFF())
2054 reportError(O->getFileName(), "Invalid/Unsupported object file format");
2056 Triple::ArchType AT = O->getArch();
2057 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n";
2058 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName());
2060 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
2061 outs() << "start address: "
2062 << "0x" << format(Fmt.data(), Address) << "\n\n";
2065 static void printArchiveChild(StringRef Filename, const Archive::Child &C) {
2066 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode();
2067 if (!ModeOrErr) {
2068 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n";
2069 consumeError(ModeOrErr.takeError());
2070 return;
2072 sys::fs::perms Mode = ModeOrErr.get();
2073 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-");
2074 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-");
2075 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-");
2076 outs() << ((Mode & sys::fs::group_read) ? "r" : "-");
2077 outs() << ((Mode & sys::fs::group_write) ? "w" : "-");
2078 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-");
2079 outs() << ((Mode & sys::fs::others_read) ? "r" : "-");
2080 outs() << ((Mode & sys::fs::others_write) ? "w" : "-");
2081 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-");
2083 outs() << " ";
2085 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename),
2086 unwrapOrError(C.getGID(), Filename),
2087 unwrapOrError(C.getRawSize(), Filename));
2089 StringRef RawLastModified = C.getRawLastModified();
2090 unsigned Seconds;
2091 if (RawLastModified.getAsInteger(10, Seconds))
2092 outs() << "(date: \"" << RawLastModified
2093 << "\" contains non-decimal chars) ";
2094 else {
2095 // Since ctime(3) returns a 26 character string of the form:
2096 // "Sun Sep 16 01:03:52 1973\n\0"
2097 // just print 24 characters.
2098 time_t t = Seconds;
2099 outs() << format("%.24s ", ctime(&t));
2102 StringRef Name = "";
2103 Expected<StringRef> NameOrErr = C.getName();
2104 if (!NameOrErr) {
2105 consumeError(NameOrErr.takeError());
2106 Name = unwrapOrError(C.getRawName(), Filename);
2107 } else {
2108 Name = NameOrErr.get();
2110 outs() << Name << "\n";
2113 // For ELF only now.
2114 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) {
2115 if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) {
2116 if (Elf->getEType() != ELF::ET_REL)
2117 return true;
2119 return false;
2122 static void checkForInvalidStartStopAddress(ObjectFile *Obj,
2123 uint64_t Start, uint64_t Stop) {
2124 if (!shouldWarnForInvalidStartStopAddress(Obj))
2125 return;
2127 for (const SectionRef &Section : Obj->sections())
2128 if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) {
2129 uint64_t BaseAddr = Section.getAddress();
2130 uint64_t Size = Section.getSize();
2131 if ((Start < BaseAddr + Size) && Stop > BaseAddr)
2132 return;
2135 if (StartAddress.getNumOccurrences() == 0)
2136 reportWarning("no section has address less than 0x" +
2137 Twine::utohexstr(Stop) + " specified by --stop-address",
2138 Obj->getFileName());
2139 else if (StopAddress.getNumOccurrences() == 0)
2140 reportWarning("no section has address greater than or equal to 0x" +
2141 Twine::utohexstr(Start) + " specified by --start-address",
2142 Obj->getFileName());
2143 else
2144 reportWarning("no section overlaps the range [0x" +
2145 Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) +
2146 ") specified by --start-address/--stop-address",
2147 Obj->getFileName());
2150 static void dumpObject(ObjectFile *O, const Archive *A = nullptr,
2151 const Archive::Child *C = nullptr) {
2152 // Avoid other output when using a raw option.
2153 if (!RawClangAST) {
2154 outs() << '\n';
2155 if (A)
2156 outs() << A->getFileName() << "(" << O->getFileName() << ")";
2157 else
2158 outs() << O->getFileName();
2159 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n";
2162 if (StartAddress.getNumOccurrences() || StopAddress.getNumOccurrences())
2163 checkForInvalidStartStopAddress(O, StartAddress, StopAddress);
2165 // Note: the order here matches GNU objdump for compatability.
2166 StringRef ArchiveName = A ? A->getFileName() : "";
2167 if (ArchiveHeaders && !MachOOpt && C)
2168 printArchiveChild(ArchiveName, *C);
2169 if (FileHeaders)
2170 printFileHeaders(O);
2171 if (PrivateHeaders || FirstPrivateHeader)
2172 printPrivateFileHeaders(O, FirstPrivateHeader);
2173 if (SectionHeaders)
2174 printSectionHeaders(O);
2175 if (SymbolTable)
2176 printSymbolTable(O, ArchiveName);
2177 if (DwarfDumpType != DIDT_Null) {
2178 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O);
2179 // Dump the complete DWARF structure.
2180 DIDumpOptions DumpOpts;
2181 DumpOpts.DumpType = DwarfDumpType;
2182 DICtx->dump(outs(), DumpOpts);
2184 if (Relocations && !Disassemble)
2185 printRelocations(O);
2186 if (DynamicRelocations)
2187 printDynamicRelocations(O);
2188 if (SectionContents)
2189 printSectionContents(O);
2190 if (Disassemble)
2191 disassembleObject(O, Relocations);
2192 if (UnwindInfo)
2193 printUnwindInfo(O);
2195 // Mach-O specific options:
2196 if (ExportsTrie)
2197 printExportsTrie(O);
2198 if (Rebase)
2199 printRebaseTable(O);
2200 if (Bind)
2201 printBindTable(O);
2202 if (LazyBind)
2203 printLazyBindTable(O);
2204 if (WeakBind)
2205 printWeakBindTable(O);
2207 // Other special sections:
2208 if (RawClangAST)
2209 printRawClangAST(O);
2210 if (FaultMapSection)
2211 printFaultMaps(O);
2214 static void dumpObject(const COFFImportFile *I, const Archive *A,
2215 const Archive::Child *C = nullptr) {
2216 StringRef ArchiveName = A ? A->getFileName() : "";
2218 // Avoid other output when using a raw option.
2219 if (!RawClangAST)
2220 outs() << '\n'
2221 << ArchiveName << "(" << I->getFileName() << ")"
2222 << ":\tfile format COFF-import-file"
2223 << "\n\n";
2225 if (ArchiveHeaders && !MachOOpt && C)
2226 printArchiveChild(ArchiveName, *C);
2227 if (SymbolTable)
2228 printCOFFSymbolTable(I);
2231 /// Dump each object file in \a a;
2232 static void dumpArchive(const Archive *A) {
2233 Error Err = Error::success();
2234 unsigned I = -1;
2235 for (auto &C : A->children(Err)) {
2236 ++I;
2237 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary();
2238 if (!ChildOrErr) {
2239 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError()))
2240 reportError(std::move(E), getFileNameForError(C, I), A->getFileName());
2241 continue;
2243 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get()))
2244 dumpObject(O, A, &C);
2245 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get()))
2246 dumpObject(I, A, &C);
2247 else
2248 reportError(errorCodeToError(object_error::invalid_file_type),
2249 A->getFileName());
2251 if (Err)
2252 reportError(std::move(Err), A->getFileName());
2255 /// Open file and figure out how to dump it.
2256 static void dumpInput(StringRef file) {
2257 // If we are using the Mach-O specific object file parser, then let it parse
2258 // the file and process the command line options. So the -arch flags can
2259 // be used to select specific slices, etc.
2260 if (MachOOpt) {
2261 parseInputMachO(file);
2262 return;
2265 // Attempt to open the binary.
2266 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file);
2267 Binary &Binary = *OBinary.getBinary();
2269 if (Archive *A = dyn_cast<Archive>(&Binary))
2270 dumpArchive(A);
2271 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary))
2272 dumpObject(O);
2273 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary))
2274 parseInputMachO(UB);
2275 else
2276 reportError(errorCodeToError(object_error::invalid_file_type), file);
2278 } // namespace llvm
2280 int main(int argc, char **argv) {
2281 using namespace llvm;
2282 InitLLVM X(argc, argv);
2283 const cl::OptionCategory *OptionFilters[] = {&ObjdumpCat, &MachOCat};
2284 cl::HideUnrelatedOptions(OptionFilters);
2286 // Initialize targets and assembly printers/parsers.
2287 InitializeAllTargetInfos();
2288 InitializeAllTargetMCs();
2289 InitializeAllDisassemblers();
2291 // Register the target printer for --version.
2292 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion);
2294 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n");
2296 if (StartAddress >= StopAddress)
2297 reportCmdLineError("start address should be less than stop address");
2299 ToolName = argv[0];
2301 // Defaults to a.out if no filenames specified.
2302 if (InputFilenames.empty())
2303 InputFilenames.push_back("a.out");
2305 if (AllHeaders)
2306 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations =
2307 SectionHeaders = SymbolTable = true;
2309 if (DisassembleAll || PrintSource || PrintLines ||
2310 (!DisassembleFunctions.empty()))
2311 Disassemble = true;
2313 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null &&
2314 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST &&
2315 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable &&
2316 !UnwindInfo && !FaultMapSection &&
2317 !(MachOOpt &&
2318 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie ||
2319 FirstPrivateHeader || IndirectSymbols || InfoPlist || LazyBind ||
2320 LinkOptHints || ObjcMetaData || Rebase || UniversalHeaders ||
2321 WeakBind || !FilterSections.empty()))) {
2322 cl::PrintHelpMessage();
2323 return 2;
2326 DisasmFuncsSet.insert(DisassembleFunctions.begin(),
2327 DisassembleFunctions.end());
2329 llvm::for_each(InputFilenames, dumpInput);
2331 warnOnNoMatchForSections();
2333 return EXIT_SUCCESS;