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. 12 // 13 // The flags and output of this program should be near identical to those of 14 // binutils objdump. 15 // 16 //===----------------------------------------------------------------------===// 17 18 #include "llvm-objdump.h" 19 #include "COFFDump.h" 20 #include "ELFDump.h" 21 #include "MachODump.h" 22 #include "ObjdumpOptID.h" 23 #include "OffloadDump.h" 24 #include "SourcePrinter.h" 25 #include "WasmDump.h" 26 #include "XCOFFDump.h" 27 #include "llvm/ADT/IndexedMap.h" 28 #include "llvm/ADT/STLExtras.h" 29 #include "llvm/ADT/SetOperations.h" 30 #include "llvm/ADT/SmallSet.h" 31 #include "llvm/ADT/StringExtras.h" 32 #include "llvm/ADT/StringSet.h" 33 #include "llvm/ADT/Twine.h" 34 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 35 #include "llvm/DebugInfo/Symbolize/SymbolizableModule.h" 36 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 37 #include "llvm/Debuginfod/BuildIDFetcher.h" 38 #include "llvm/Debuginfod/Debuginfod.h" 39 #include "llvm/Debuginfod/HTTPClient.h" 40 #include "llvm/Demangle/Demangle.h" 41 #include "llvm/MC/MCAsmInfo.h" 42 #include "llvm/MC/MCContext.h" 43 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 44 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 45 #include "llvm/MC/MCInst.h" 46 #include "llvm/MC/MCInstPrinter.h" 47 #include "llvm/MC/MCInstrAnalysis.h" 48 #include "llvm/MC/MCInstrInfo.h" 49 #include "llvm/MC/MCObjectFileInfo.h" 50 #include "llvm/MC/MCRegisterInfo.h" 51 #include "llvm/MC/MCTargetOptions.h" 52 #include "llvm/MC/TargetRegistry.h" 53 #include "llvm/Object/Archive.h" 54 #include "llvm/Object/BuildID.h" 55 #include "llvm/Object/COFF.h" 56 #include "llvm/Object/COFFImportFile.h" 57 #include "llvm/Object/ELFObjectFile.h" 58 #include "llvm/Object/ELFTypes.h" 59 #include "llvm/Object/FaultMapParser.h" 60 #include "llvm/Object/MachO.h" 61 #include "llvm/Object/MachOUniversal.h" 62 #include "llvm/Object/ObjectFile.h" 63 #include "llvm/Object/OffloadBinary.h" 64 #include "llvm/Object/Wasm.h" 65 #include "llvm/Option/Arg.h" 66 #include "llvm/Option/ArgList.h" 67 #include "llvm/Option/Option.h" 68 #include "llvm/Support/Casting.h" 69 #include "llvm/Support/Debug.h" 70 #include "llvm/Support/Errc.h" 71 #include "llvm/Support/FileSystem.h" 72 #include "llvm/Support/Format.h" 73 #include "llvm/Support/FormatVariadic.h" 74 #include "llvm/Support/GraphWriter.h" 75 #include "llvm/Support/InitLLVM.h" 76 #include "llvm/Support/MemoryBuffer.h" 77 #include "llvm/Support/SourceMgr.h" 78 #include "llvm/Support/StringSaver.h" 79 #include "llvm/Support/TargetSelect.h" 80 #include "llvm/Support/WithColor.h" 81 #include "llvm/Support/raw_ostream.h" 82 #include "llvm/TargetParser/Host.h" 83 #include "llvm/TargetParser/Triple.h" 84 #include <algorithm> 85 #include <cctype> 86 #include <cstring> 87 #include <optional> 88 #include <system_error> 89 #include <unordered_map> 90 #include <utility> 91 92 using namespace llvm; 93 using namespace llvm::object; 94 using namespace llvm::objdump; 95 using namespace llvm::opt; 96 97 namespace { 98 99 class CommonOptTable : public opt::GenericOptTable { 100 public: 101 CommonOptTable(ArrayRef<Info> OptionInfos, const char *Usage, 102 const char *Description) 103 : opt::GenericOptTable(OptionInfos), Usage(Usage), 104 Description(Description) { 105 setGroupedShortOptions(true); 106 } 107 108 void printHelp(StringRef Argv0, bool ShowHidden = false) const { 109 Argv0 = sys::path::filename(Argv0); 110 opt::GenericOptTable::printHelp(outs(), (Argv0 + Usage).str().c_str(), 111 Description, ShowHidden, ShowHidden); 112 // TODO Replace this with OptTable API once it adds extrahelp support. 113 outs() << "\nPass @FILE as argument to read options from FILE.\n"; 114 } 115 116 private: 117 const char *Usage; 118 const char *Description; 119 }; 120 121 // ObjdumpOptID is in ObjdumpOptID.h 122 namespace objdump_opt { 123 #define PREFIX(NAME, VALUE) \ 124 static constexpr StringLiteral NAME##_init[] = VALUE; \ 125 static constexpr ArrayRef<StringLiteral> NAME(NAME##_init, \ 126 std::size(NAME##_init) - 1); 127 #include "ObjdumpOpts.inc" 128 #undef PREFIX 129 130 static constexpr opt::OptTable::Info ObjdumpInfoTable[] = { 131 #define OPTION(PREFIX, NAME, ID, KIND, GROUP, ALIAS, ALIASARGS, FLAGS, PARAM, \ 132 HELPTEXT, METAVAR, VALUES) \ 133 {PREFIX, NAME, HELPTEXT, \ 134 METAVAR, OBJDUMP_##ID, opt::Option::KIND##Class, \ 135 PARAM, FLAGS, OBJDUMP_##GROUP, \ 136 OBJDUMP_##ALIAS, ALIASARGS, VALUES}, 137 #include "ObjdumpOpts.inc" 138 #undef OPTION 139 }; 140 } // namespace objdump_opt 141 142 class ObjdumpOptTable : public CommonOptTable { 143 public: 144 ObjdumpOptTable() 145 : CommonOptTable(objdump_opt::ObjdumpInfoTable, 146 " [options] <input object files>", 147 "llvm object file dumper") {} 148 }; 149 150 enum OtoolOptID { 151 OTOOL_INVALID = 0, // This is not an option ID. 152 #define OPTION(PREFIX, NAME, ID, KIND, GROUP, ALIAS, ALIASARGS, FLAGS, PARAM, \ 153 HELPTEXT, METAVAR, VALUES) \ 154 OTOOL_##ID, 155 #include "OtoolOpts.inc" 156 #undef OPTION 157 }; 158 159 namespace otool { 160 #define PREFIX(NAME, VALUE) \ 161 static constexpr StringLiteral NAME##_init[] = VALUE; \ 162 static constexpr ArrayRef<StringLiteral> NAME(NAME##_init, \ 163 std::size(NAME##_init) - 1); 164 #include "OtoolOpts.inc" 165 #undef PREFIX 166 167 static constexpr opt::OptTable::Info OtoolInfoTable[] = { 168 #define OPTION(PREFIX, NAME, ID, KIND, GROUP, ALIAS, ALIASARGS, FLAGS, PARAM, \ 169 HELPTEXT, METAVAR, VALUES) \ 170 {PREFIX, NAME, HELPTEXT, \ 171 METAVAR, OTOOL_##ID, opt::Option::KIND##Class, \ 172 PARAM, FLAGS, OTOOL_##GROUP, \ 173 OTOOL_##ALIAS, ALIASARGS, VALUES}, 174 #include "OtoolOpts.inc" 175 #undef OPTION 176 }; 177 } // namespace otool 178 179 class OtoolOptTable : public CommonOptTable { 180 public: 181 OtoolOptTable() 182 : CommonOptTable(otool::OtoolInfoTable, " [option...] [file...]", 183 "Mach-O object file displaying tool") {} 184 }; 185 186 } // namespace 187 188 #define DEBUG_TYPE "objdump" 189 190 static uint64_t AdjustVMA; 191 static bool AllHeaders; 192 static std::string ArchName; 193 bool objdump::ArchiveHeaders; 194 bool objdump::Demangle; 195 bool objdump::Disassemble; 196 bool objdump::DisassembleAll; 197 bool objdump::SymbolDescription; 198 bool objdump::TracebackTable; 199 static std::vector<std::string> DisassembleSymbols; 200 static bool DisassembleZeroes; 201 static std::vector<std::string> DisassemblerOptions; 202 DIDumpType objdump::DwarfDumpType; 203 static bool DynamicRelocations; 204 static bool FaultMapSection; 205 static bool FileHeaders; 206 bool objdump::SectionContents; 207 static std::vector<std::string> InputFilenames; 208 bool objdump::PrintLines; 209 static bool MachOOpt; 210 std::string objdump::MCPU; 211 std::vector<std::string> objdump::MAttrs; 212 bool objdump::ShowRawInsn; 213 bool objdump::LeadingAddr; 214 static bool Offloading; 215 static bool RawClangAST; 216 bool objdump::Relocations; 217 bool objdump::PrintImmHex; 218 bool objdump::PrivateHeaders; 219 std::vector<std::string> objdump::FilterSections; 220 bool objdump::SectionHeaders; 221 static bool ShowAllSymbols; 222 static bool ShowLMA; 223 bool objdump::PrintSource; 224 225 static uint64_t StartAddress; 226 static bool HasStartAddressFlag; 227 static uint64_t StopAddress = UINT64_MAX; 228 static bool HasStopAddressFlag; 229 230 bool objdump::SymbolTable; 231 static bool SymbolizeOperands; 232 static bool DynamicSymbolTable; 233 std::string objdump::TripleName; 234 bool objdump::UnwindInfo; 235 static bool Wide; 236 std::string objdump::Prefix; 237 uint32_t objdump::PrefixStrip; 238 239 DebugVarsFormat objdump::DbgVariables = DVDisabled; 240 241 int objdump::DbgIndent = 52; 242 243 static StringSet<> DisasmSymbolSet; 244 StringSet<> objdump::FoundSectionSet; 245 static StringRef ToolName; 246 247 std::unique_ptr<BuildIDFetcher> BIDFetcher; 248 ExitOnError ExitOnErr; 249 250 void Dumper::reportUniqueWarning(Error Err) { 251 reportUniqueWarning(toString(std::move(Err))); 252 } 253 254 void Dumper::reportUniqueWarning(const Twine &Msg) { 255 if (Warnings.insert(StringRef(Msg.str())).second) 256 reportWarning(Msg, O.getFileName()); 257 } 258 259 static Expected<std::unique_ptr<Dumper>> createDumper(const ObjectFile &Obj) { 260 if (const auto *O = dyn_cast<COFFObjectFile>(&Obj)) 261 return createCOFFDumper(*O); 262 if (const auto *O = dyn_cast<ELFObjectFileBase>(&Obj)) 263 return createELFDumper(*O); 264 if (const auto *O = dyn_cast<MachOObjectFile>(&Obj)) 265 return createMachODumper(*O); 266 if (const auto *O = dyn_cast<WasmObjectFile>(&Obj)) 267 return createWasmDumper(*O); 268 if (const auto *O = dyn_cast<XCOFFObjectFile>(&Obj)) 269 return createXCOFFDumper(*O); 270 271 return createStringError(errc::invalid_argument, 272 "unsupported object file format"); 273 } 274 275 namespace { 276 struct FilterResult { 277 // True if the section should not be skipped. 278 bool Keep; 279 280 // True if the index counter should be incremented, even if the section should 281 // be skipped. For example, sections may be skipped if they are not included 282 // in the --section flag, but we still want those to count toward the section 283 // count. 284 bool IncrementIndex; 285 }; 286 } // namespace 287 288 static FilterResult checkSectionFilter(object::SectionRef S) { 289 if (FilterSections.empty()) 290 return {/*Keep=*/true, /*IncrementIndex=*/true}; 291 292 Expected<StringRef> SecNameOrErr = S.getName(); 293 if (!SecNameOrErr) { 294 consumeError(SecNameOrErr.takeError()); 295 return {/*Keep=*/false, /*IncrementIndex=*/false}; 296 } 297 StringRef SecName = *SecNameOrErr; 298 299 // StringSet does not allow empty key so avoid adding sections with 300 // no name (such as the section with index 0) here. 301 if (!SecName.empty()) 302 FoundSectionSet.insert(SecName); 303 304 // Only show the section if it's in the FilterSections list, but always 305 // increment so the indexing is stable. 306 return {/*Keep=*/is_contained(FilterSections, SecName), 307 /*IncrementIndex=*/true}; 308 } 309 310 SectionFilter objdump::ToolSectionFilter(object::ObjectFile const &O, 311 uint64_t *Idx) { 312 // Start at UINT64_MAX so that the first index returned after an increment is 313 // zero (after the unsigned wrap). 314 if (Idx) 315 *Idx = UINT64_MAX; 316 return SectionFilter( 317 [Idx](object::SectionRef S) { 318 FilterResult Result = checkSectionFilter(S); 319 if (Idx != nullptr && Result.IncrementIndex) 320 *Idx += 1; 321 return Result.Keep; 322 }, 323 O); 324 } 325 326 std::string objdump::getFileNameForError(const object::Archive::Child &C, 327 unsigned Index) { 328 Expected<StringRef> NameOrErr = C.getName(); 329 if (NameOrErr) 330 return std::string(NameOrErr.get()); 331 // If we have an error getting the name then we print the index of the archive 332 // member. Since we are already in an error state, we just ignore this error. 333 consumeError(NameOrErr.takeError()); 334 return "<file index: " + std::to_string(Index) + ">"; 335 } 336 337 void objdump::reportWarning(const Twine &Message, StringRef File) { 338 // Output order between errs() and outs() matters especially for archive 339 // files where the output is per member object. 340 outs().flush(); 341 WithColor::warning(errs(), ToolName) 342 << "'" << File << "': " << Message << "\n"; 343 } 344 345 [[noreturn]] void objdump::reportError(StringRef File, const Twine &Message) { 346 outs().flush(); 347 WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n"; 348 exit(1); 349 } 350 351 [[noreturn]] void objdump::reportError(Error E, StringRef FileName, 352 StringRef ArchiveName, 353 StringRef ArchitectureName) { 354 assert(E); 355 outs().flush(); 356 WithColor::error(errs(), ToolName); 357 if (ArchiveName != "") 358 errs() << ArchiveName << "(" << FileName << ")"; 359 else 360 errs() << "'" << FileName << "'"; 361 if (!ArchitectureName.empty()) 362 errs() << " (for architecture " << ArchitectureName << ")"; 363 errs() << ": "; 364 logAllUnhandledErrors(std::move(E), errs()); 365 exit(1); 366 } 367 368 static void reportCmdLineWarning(const Twine &Message) { 369 WithColor::warning(errs(), ToolName) << Message << "\n"; 370 } 371 372 [[noreturn]] static void reportCmdLineError(const Twine &Message) { 373 WithColor::error(errs(), ToolName) << Message << "\n"; 374 exit(1); 375 } 376 377 static void warnOnNoMatchForSections() { 378 SetVector<StringRef> MissingSections; 379 for (StringRef S : FilterSections) { 380 if (FoundSectionSet.count(S)) 381 return; 382 // User may specify a unnamed section. Don't warn for it. 383 if (!S.empty()) 384 MissingSections.insert(S); 385 } 386 387 // Warn only if no section in FilterSections is matched. 388 for (StringRef S : MissingSections) 389 reportCmdLineWarning("section '" + S + 390 "' mentioned in a -j/--section option, but not " 391 "found in any input file"); 392 } 393 394 static const Target *getTarget(const ObjectFile *Obj) { 395 // Figure out the target triple. 396 Triple TheTriple("unknown-unknown-unknown"); 397 if (TripleName.empty()) { 398 TheTriple = Obj->makeTriple(); 399 } else { 400 TheTriple.setTriple(Triple::normalize(TripleName)); 401 auto Arch = Obj->getArch(); 402 if (Arch == Triple::arm || Arch == Triple::armeb) 403 Obj->setARMSubArch(TheTriple); 404 } 405 406 // Get the target specific parser. 407 std::string Error; 408 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 409 Error); 410 if (!TheTarget) 411 reportError(Obj->getFileName(), "can't find target: " + Error); 412 413 // Update the triple name and return the found target. 414 TripleName = TheTriple.getTriple(); 415 return TheTarget; 416 } 417 418 bool objdump::isRelocAddressLess(RelocationRef A, RelocationRef B) { 419 return A.getOffset() < B.getOffset(); 420 } 421 422 static Error getRelocationValueString(const RelocationRef &Rel, 423 SmallVectorImpl<char> &Result) { 424 const ObjectFile *Obj = Rel.getObject(); 425 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 426 return getELFRelocationValueString(ELF, Rel, Result); 427 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 428 return getCOFFRelocationValueString(COFF, Rel, Result); 429 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 430 return getWasmRelocationValueString(Wasm, Rel, Result); 431 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 432 return getMachORelocationValueString(MachO, Rel, Result); 433 if (auto *XCOFF = dyn_cast<XCOFFObjectFile>(Obj)) 434 return getXCOFFRelocationValueString(*XCOFF, Rel, Result); 435 llvm_unreachable("unknown object file format"); 436 } 437 438 /// Indicates whether this relocation should hidden when listing 439 /// relocations, usually because it is the trailing part of a multipart 440 /// relocation that will be printed as part of the leading relocation. 441 static bool getHidden(RelocationRef RelRef) { 442 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 443 if (!MachO) 444 return false; 445 446 unsigned Arch = MachO->getArch(); 447 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 448 uint64_t Type = MachO->getRelocationType(Rel); 449 450 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 451 // is always hidden. 452 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 453 return Type == MachO::GENERIC_RELOC_PAIR; 454 455 if (Arch == Triple::x86_64) { 456 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 457 // an X86_64_RELOC_SUBTRACTOR. 458 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 459 DataRefImpl RelPrev = Rel; 460 RelPrev.d.a--; 461 uint64_t PrevType = MachO->getRelocationType(RelPrev); 462 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 463 return true; 464 } 465 } 466 467 return false; 468 } 469 470 /// Get the column at which we want to start printing the instruction 471 /// disassembly, taking into account anything which appears to the left of it. 472 unsigned objdump::getInstStartColumn(const MCSubtargetInfo &STI) { 473 return !ShowRawInsn ? 16 : STI.getTargetTriple().isX86() ? 40 : 24; 474 } 475 476 static void AlignToInstStartColumn(size_t Start, const MCSubtargetInfo &STI, 477 raw_ostream &OS) { 478 // The output of printInst starts with a tab. Print some spaces so that 479 // the tab has 1 column and advances to the target tab stop. 480 unsigned TabStop = getInstStartColumn(STI); 481 unsigned Column = OS.tell() - Start; 482 OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8); 483 } 484 485 void objdump::printRawData(ArrayRef<uint8_t> Bytes, uint64_t Address, 486 formatted_raw_ostream &OS, 487 MCSubtargetInfo const &STI) { 488 size_t Start = OS.tell(); 489 if (LeadingAddr) 490 OS << format("%8" PRIx64 ":", Address); 491 if (ShowRawInsn) { 492 OS << ' '; 493 dumpBytes(Bytes, OS); 494 } 495 AlignToInstStartColumn(Start, STI, OS); 496 } 497 498 namespace { 499 500 static bool isAArch64Elf(const ObjectFile &Obj) { 501 const auto *Elf = dyn_cast<ELFObjectFileBase>(&Obj); 502 return Elf && Elf->getEMachine() == ELF::EM_AARCH64; 503 } 504 505 static bool isArmElf(const ObjectFile &Obj) { 506 const auto *Elf = dyn_cast<ELFObjectFileBase>(&Obj); 507 return Elf && Elf->getEMachine() == ELF::EM_ARM; 508 } 509 510 static bool isCSKYElf(const ObjectFile &Obj) { 511 const auto *Elf = dyn_cast<ELFObjectFileBase>(&Obj); 512 return Elf && Elf->getEMachine() == ELF::EM_CSKY; 513 } 514 515 static bool hasMappingSymbols(const ObjectFile &Obj) { 516 return isArmElf(Obj) || isAArch64Elf(Obj) || isCSKYElf(Obj) ; 517 } 518 519 static void printRelocation(formatted_raw_ostream &OS, StringRef FileName, 520 const RelocationRef &Rel, uint64_t Address, 521 bool Is64Bits) { 522 StringRef Fmt = Is64Bits ? "%016" PRIx64 ": " : "%08" PRIx64 ": "; 523 SmallString<16> Name; 524 SmallString<32> Val; 525 Rel.getTypeName(Name); 526 if (Error E = getRelocationValueString(Rel, Val)) 527 reportError(std::move(E), FileName); 528 OS << (Is64Bits || !LeadingAddr ? "\t\t" : "\t\t\t"); 529 if (LeadingAddr) 530 OS << format(Fmt.data(), Address); 531 OS << Name << "\t" << Val; 532 } 533 534 class PrettyPrinter { 535 public: 536 virtual ~PrettyPrinter() = default; 537 virtual void 538 printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 539 object::SectionedAddress Address, formatted_raw_ostream &OS, 540 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 541 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 542 LiveVariablePrinter &LVP) { 543 if (SP && (PrintSource || PrintLines)) 544 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 545 LVP.printBetweenInsts(OS, false); 546 547 printRawData(Bytes, Address.Address, OS, STI); 548 549 if (MI) { 550 // See MCInstPrinter::printInst. On targets where a PC relative immediate 551 // is relative to the next instruction and the length of a MCInst is 552 // difficult to measure (x86), this is the address of the next 553 // instruction. 554 uint64_t Addr = 555 Address.Address + (STI.getTargetTriple().isX86() ? Bytes.size() : 0); 556 IP.printInst(MI, Addr, "", STI, OS); 557 } else 558 OS << "\t<unknown>"; 559 } 560 }; 561 PrettyPrinter PrettyPrinterInst; 562 563 class HexagonPrettyPrinter : public PrettyPrinter { 564 public: 565 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 566 formatted_raw_ostream &OS) { 567 uint32_t opcode = 568 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 569 if (LeadingAddr) 570 OS << format("%8" PRIx64 ":", Address); 571 if (ShowRawInsn) { 572 OS << "\t"; 573 dumpBytes(Bytes.slice(0, 4), OS); 574 OS << format("\t%08" PRIx32, opcode); 575 } 576 } 577 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 578 object::SectionedAddress Address, formatted_raw_ostream &OS, 579 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 580 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 581 LiveVariablePrinter &LVP) override { 582 if (SP && (PrintSource || PrintLines)) 583 SP->printSourceLine(OS, Address, ObjectFilename, LVP, ""); 584 if (!MI) { 585 printLead(Bytes, Address.Address, OS); 586 OS << " <unknown>"; 587 return; 588 } 589 std::string Buffer; 590 { 591 raw_string_ostream TempStream(Buffer); 592 IP.printInst(MI, Address.Address, "", STI, TempStream); 593 } 594 StringRef Contents(Buffer); 595 // Split off bundle attributes 596 auto PacketBundle = Contents.rsplit('\n'); 597 // Split off first instruction from the rest 598 auto HeadTail = PacketBundle.first.split('\n'); 599 auto Preamble = " { "; 600 auto Separator = ""; 601 602 // Hexagon's packets require relocations to be inline rather than 603 // clustered at the end of the packet. 604 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 605 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 606 auto PrintReloc = [&]() -> void { 607 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 608 if (RelCur->getOffset() == Address.Address) { 609 printRelocation(OS, ObjectFilename, *RelCur, Address.Address, false); 610 return; 611 } 612 ++RelCur; 613 } 614 }; 615 616 while (!HeadTail.first.empty()) { 617 OS << Separator; 618 Separator = "\n"; 619 if (SP && (PrintSource || PrintLines)) 620 SP->printSourceLine(OS, Address, ObjectFilename, LVP, ""); 621 printLead(Bytes, Address.Address, OS); 622 OS << Preamble; 623 Preamble = " "; 624 StringRef Inst; 625 auto Duplex = HeadTail.first.split('\v'); 626 if (!Duplex.second.empty()) { 627 OS << Duplex.first; 628 OS << "; "; 629 Inst = Duplex.second; 630 } 631 else 632 Inst = HeadTail.first; 633 OS << Inst; 634 HeadTail = HeadTail.second.split('\n'); 635 if (HeadTail.first.empty()) 636 OS << " } " << PacketBundle.second; 637 PrintReloc(); 638 Bytes = Bytes.slice(4); 639 Address.Address += 4; 640 } 641 } 642 }; 643 HexagonPrettyPrinter HexagonPrettyPrinterInst; 644 645 class AMDGCNPrettyPrinter : public PrettyPrinter { 646 public: 647 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 648 object::SectionedAddress Address, formatted_raw_ostream &OS, 649 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 650 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 651 LiveVariablePrinter &LVP) override { 652 if (SP && (PrintSource || PrintLines)) 653 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 654 655 if (MI) { 656 SmallString<40> InstStr; 657 raw_svector_ostream IS(InstStr); 658 659 IP.printInst(MI, Address.Address, "", STI, IS); 660 661 OS << left_justify(IS.str(), 60); 662 } else { 663 // an unrecognized encoding - this is probably data so represent it 664 // using the .long directive, or .byte directive if fewer than 4 bytes 665 // remaining 666 if (Bytes.size() >= 4) { 667 OS << format("\t.long 0x%08" PRIx32 " ", 668 support::endian::read32<support::little>(Bytes.data())); 669 OS.indent(42); 670 } else { 671 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 672 for (unsigned int i = 1; i < Bytes.size(); i++) 673 OS << format(", 0x%02" PRIx8, Bytes[i]); 674 OS.indent(55 - (6 * Bytes.size())); 675 } 676 } 677 678 OS << format("// %012" PRIX64 ":", Address.Address); 679 if (Bytes.size() >= 4) { 680 // D should be casted to uint32_t here as it is passed by format to 681 // snprintf as vararg. 682 for (uint32_t D : 683 ArrayRef(reinterpret_cast<const support::little32_t *>(Bytes.data()), 684 Bytes.size() / 4)) 685 OS << format(" %08" PRIX32, D); 686 } else { 687 for (unsigned char B : Bytes) 688 OS << format(" %02" PRIX8, B); 689 } 690 691 if (!Annot.empty()) 692 OS << " // " << Annot; 693 } 694 }; 695 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 696 697 class BPFPrettyPrinter : public PrettyPrinter { 698 public: 699 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 700 object::SectionedAddress Address, formatted_raw_ostream &OS, 701 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 702 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 703 LiveVariablePrinter &LVP) override { 704 if (SP && (PrintSource || PrintLines)) 705 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 706 if (LeadingAddr) 707 OS << format("%8" PRId64 ":", Address.Address / 8); 708 if (ShowRawInsn) { 709 OS << "\t"; 710 dumpBytes(Bytes, OS); 711 } 712 if (MI) 713 IP.printInst(MI, Address.Address, "", STI, OS); 714 else 715 OS << "\t<unknown>"; 716 } 717 }; 718 BPFPrettyPrinter BPFPrettyPrinterInst; 719 720 class ARMPrettyPrinter : public PrettyPrinter { 721 public: 722 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 723 object::SectionedAddress Address, formatted_raw_ostream &OS, 724 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 725 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 726 LiveVariablePrinter &LVP) override { 727 if (SP && (PrintSource || PrintLines)) 728 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 729 LVP.printBetweenInsts(OS, false); 730 731 size_t Start = OS.tell(); 732 if (LeadingAddr) 733 OS << format("%8" PRIx64 ":", Address.Address); 734 if (ShowRawInsn) { 735 size_t Pos = 0, End = Bytes.size(); 736 if (STI.checkFeatures("+thumb-mode")) { 737 for (; Pos + 2 <= End; Pos += 2) 738 OS << ' ' 739 << format_hex_no_prefix( 740 llvm::support::endian::read<uint16_t>( 741 Bytes.data() + Pos, InstructionEndianness), 742 4); 743 } else { 744 for (; Pos + 4 <= End; Pos += 4) 745 OS << ' ' 746 << format_hex_no_prefix( 747 llvm::support::endian::read<uint32_t>( 748 Bytes.data() + Pos, InstructionEndianness), 749 8); 750 } 751 if (Pos < End) { 752 OS << ' '; 753 dumpBytes(Bytes.slice(Pos), OS); 754 } 755 } 756 757 AlignToInstStartColumn(Start, STI, OS); 758 759 if (MI) { 760 IP.printInst(MI, Address.Address, "", STI, OS); 761 } else 762 OS << "\t<unknown>"; 763 } 764 765 void setInstructionEndianness(llvm::support::endianness Endianness) { 766 InstructionEndianness = Endianness; 767 } 768 769 private: 770 llvm::support::endianness InstructionEndianness = llvm::support::little; 771 }; 772 ARMPrettyPrinter ARMPrettyPrinterInst; 773 774 class AArch64PrettyPrinter : public PrettyPrinter { 775 public: 776 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 777 object::SectionedAddress Address, formatted_raw_ostream &OS, 778 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 779 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 780 LiveVariablePrinter &LVP) override { 781 if (SP && (PrintSource || PrintLines)) 782 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 783 LVP.printBetweenInsts(OS, false); 784 785 size_t Start = OS.tell(); 786 if (LeadingAddr) 787 OS << format("%8" PRIx64 ":", Address.Address); 788 if (ShowRawInsn) { 789 size_t Pos = 0, End = Bytes.size(); 790 for (; Pos + 4 <= End; Pos += 4) 791 OS << ' ' 792 << format_hex_no_prefix( 793 llvm::support::endian::read<uint32_t>(Bytes.data() + Pos, 794 llvm::support::little), 795 8); 796 if (Pos < End) { 797 OS << ' '; 798 dumpBytes(Bytes.slice(Pos), OS); 799 } 800 } 801 802 AlignToInstStartColumn(Start, STI, OS); 803 804 if (MI) { 805 IP.printInst(MI, Address.Address, "", STI, OS); 806 } else 807 OS << "\t<unknown>"; 808 } 809 }; 810 AArch64PrettyPrinter AArch64PrettyPrinterInst; 811 812 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 813 switch(Triple.getArch()) { 814 default: 815 return PrettyPrinterInst; 816 case Triple::hexagon: 817 return HexagonPrettyPrinterInst; 818 case Triple::amdgcn: 819 return AMDGCNPrettyPrinterInst; 820 case Triple::bpfel: 821 case Triple::bpfeb: 822 return BPFPrettyPrinterInst; 823 case Triple::arm: 824 case Triple::armeb: 825 case Triple::thumb: 826 case Triple::thumbeb: 827 return ARMPrettyPrinterInst; 828 case Triple::aarch64: 829 case Triple::aarch64_be: 830 case Triple::aarch64_32: 831 return AArch64PrettyPrinterInst; 832 } 833 } 834 835 class DisassemblerTarget { 836 public: 837 const Target *TheTarget; 838 std::unique_ptr<const MCSubtargetInfo> SubtargetInfo; 839 std::shared_ptr<MCContext> Context; 840 std::unique_ptr<MCDisassembler> DisAsm; 841 std::shared_ptr<const MCInstrAnalysis> InstrAnalysis; 842 std::shared_ptr<MCInstPrinter> InstPrinter; 843 PrettyPrinter *Printer; 844 845 DisassemblerTarget(const Target *TheTarget, ObjectFile &Obj, 846 StringRef TripleName, StringRef MCPU, 847 SubtargetFeatures &Features); 848 DisassemblerTarget(DisassemblerTarget &Other, SubtargetFeatures &Features); 849 850 private: 851 MCTargetOptions Options; 852 std::shared_ptr<const MCRegisterInfo> RegisterInfo; 853 std::shared_ptr<const MCAsmInfo> AsmInfo; 854 std::shared_ptr<const MCInstrInfo> InstrInfo; 855 std::shared_ptr<MCObjectFileInfo> ObjectFileInfo; 856 }; 857 858 DisassemblerTarget::DisassemblerTarget(const Target *TheTarget, ObjectFile &Obj, 859 StringRef TripleName, StringRef MCPU, 860 SubtargetFeatures &Features) 861 : TheTarget(TheTarget), 862 Printer(&selectPrettyPrinter(Triple(TripleName))), 863 RegisterInfo(TheTarget->createMCRegInfo(TripleName)) { 864 if (!RegisterInfo) 865 reportError(Obj.getFileName(), "no register info for target " + TripleName); 866 867 // Set up disassembler. 868 AsmInfo.reset(TheTarget->createMCAsmInfo(*RegisterInfo, TripleName, Options)); 869 if (!AsmInfo) 870 reportError(Obj.getFileName(), "no assembly info for target " + TripleName); 871 872 SubtargetInfo.reset( 873 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 874 if (!SubtargetInfo) 875 reportError(Obj.getFileName(), 876 "no subtarget info for target " + TripleName); 877 InstrInfo.reset(TheTarget->createMCInstrInfo()); 878 if (!InstrInfo) 879 reportError(Obj.getFileName(), 880 "no instruction info for target " + TripleName); 881 Context = 882 std::make_shared<MCContext>(Triple(TripleName), AsmInfo.get(), 883 RegisterInfo.get(), SubtargetInfo.get()); 884 885 // FIXME: for now initialize MCObjectFileInfo with default values 886 ObjectFileInfo.reset( 887 TheTarget->createMCObjectFileInfo(*Context, /*PIC=*/false)); 888 Context->setObjectFileInfo(ObjectFileInfo.get()); 889 890 DisAsm.reset(TheTarget->createMCDisassembler(*SubtargetInfo, *Context)); 891 if (!DisAsm) 892 reportError(Obj.getFileName(), "no disassembler for target " + TripleName); 893 894 InstrAnalysis.reset(TheTarget->createMCInstrAnalysis(InstrInfo.get())); 895 896 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 897 InstPrinter.reset(TheTarget->createMCInstPrinter(Triple(TripleName), 898 AsmPrinterVariant, *AsmInfo, 899 *InstrInfo, *RegisterInfo)); 900 if (!InstPrinter) 901 reportError(Obj.getFileName(), 902 "no instruction printer for target " + TripleName); 903 InstPrinter->setPrintImmHex(PrintImmHex); 904 InstPrinter->setPrintBranchImmAsAddress(true); 905 InstPrinter->setSymbolizeOperands(SymbolizeOperands); 906 InstPrinter->setMCInstrAnalysis(InstrAnalysis.get()); 907 } 908 909 DisassemblerTarget::DisassemblerTarget(DisassemblerTarget &Other, 910 SubtargetFeatures &Features) 911 : TheTarget(Other.TheTarget), 912 SubtargetInfo(TheTarget->createMCSubtargetInfo(TripleName, MCPU, 913 Features.getString())), 914 Context(Other.Context), 915 DisAsm(TheTarget->createMCDisassembler(*SubtargetInfo, *Context)), 916 InstrAnalysis(Other.InstrAnalysis), InstPrinter(Other.InstPrinter), 917 Printer(Other.Printer), RegisterInfo(Other.RegisterInfo), 918 AsmInfo(Other.AsmInfo), InstrInfo(Other.InstrInfo), 919 ObjectFileInfo(Other.ObjectFileInfo) {} 920 } // namespace 921 922 static uint8_t getElfSymbolType(const ObjectFile &Obj, const SymbolRef &Sym) { 923 assert(Obj.isELF()); 924 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(&Obj)) 925 return unwrapOrError(Elf32LEObj->getSymbol(Sym.getRawDataRefImpl()), 926 Obj.getFileName()) 927 ->getType(); 928 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(&Obj)) 929 return unwrapOrError(Elf64LEObj->getSymbol(Sym.getRawDataRefImpl()), 930 Obj.getFileName()) 931 ->getType(); 932 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(&Obj)) 933 return unwrapOrError(Elf32BEObj->getSymbol(Sym.getRawDataRefImpl()), 934 Obj.getFileName()) 935 ->getType(); 936 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(&Obj)) 937 return unwrapOrError(Elf64BEObj->getSymbol(Sym.getRawDataRefImpl()), 938 Obj.getFileName()) 939 ->getType(); 940 llvm_unreachable("Unsupported binary format"); 941 } 942 943 template <class ELFT> 944 static void 945 addDynamicElfSymbols(const ELFObjectFile<ELFT> &Obj, 946 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 947 for (auto Symbol : Obj.getDynamicSymbolIterators()) { 948 uint8_t SymbolType = Symbol.getELFType(); 949 if (SymbolType == ELF::STT_SECTION) 950 continue; 951 952 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj.getFileName()); 953 // ELFSymbolRef::getAddress() returns size instead of value for common 954 // symbols which is not desirable for disassembly output. Overriding. 955 if (SymbolType == ELF::STT_COMMON) 956 Address = unwrapOrError(Obj.getSymbol(Symbol.getRawDataRefImpl()), 957 Obj.getFileName()) 958 ->st_value; 959 960 StringRef Name = unwrapOrError(Symbol.getName(), Obj.getFileName()); 961 if (Name.empty()) 962 continue; 963 964 section_iterator SecI = 965 unwrapOrError(Symbol.getSection(), Obj.getFileName()); 966 if (SecI == Obj.section_end()) 967 continue; 968 969 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 970 } 971 } 972 973 static void 974 addDynamicElfSymbols(const ELFObjectFileBase &Obj, 975 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 976 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(&Obj)) 977 addDynamicElfSymbols(*Elf32LEObj, AllSymbols); 978 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(&Obj)) 979 addDynamicElfSymbols(*Elf64LEObj, AllSymbols); 980 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(&Obj)) 981 addDynamicElfSymbols(*Elf32BEObj, AllSymbols); 982 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(&Obj)) 983 addDynamicElfSymbols(*Elf64BEObj, AllSymbols); 984 else 985 llvm_unreachable("Unsupported binary format"); 986 } 987 988 static std::optional<SectionRef> getWasmCodeSection(const WasmObjectFile &Obj) { 989 for (auto SecI : Obj.sections()) { 990 const WasmSection &Section = Obj.getWasmSection(SecI); 991 if (Section.Type == wasm::WASM_SEC_CODE) 992 return SecI; 993 } 994 return std::nullopt; 995 } 996 997 static void 998 addMissingWasmCodeSymbols(const WasmObjectFile &Obj, 999 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1000 std::optional<SectionRef> Section = getWasmCodeSection(Obj); 1001 if (!Section) 1002 return; 1003 SectionSymbolsTy &Symbols = AllSymbols[*Section]; 1004 1005 std::set<uint64_t> SymbolAddresses; 1006 for (const auto &Sym : Symbols) 1007 SymbolAddresses.insert(Sym.Addr); 1008 1009 for (const wasm::WasmFunction &Function : Obj.functions()) { 1010 uint64_t Address = Function.CodeSectionOffset; 1011 // Only add fallback symbols for functions not already present in the symbol 1012 // table. 1013 if (SymbolAddresses.count(Address)) 1014 continue; 1015 // This function has no symbol, so it should have no SymbolName. 1016 assert(Function.SymbolName.empty()); 1017 // We use DebugName for the name, though it may be empty if there is no 1018 // "name" custom section, or that section is missing a name for this 1019 // function. 1020 StringRef Name = Function.DebugName; 1021 Symbols.emplace_back(Address, Name, ELF::STT_NOTYPE); 1022 } 1023 } 1024 1025 static void addPltEntries(const ObjectFile &Obj, 1026 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 1027 StringSaver &Saver) { 1028 auto *ElfObj = dyn_cast<ELFObjectFileBase>(&Obj); 1029 if (!ElfObj) 1030 return; 1031 DenseMap<StringRef, SectionRef> Sections; 1032 for (SectionRef Section : Obj.sections()) { 1033 Expected<StringRef> SecNameOrErr = Section.getName(); 1034 if (!SecNameOrErr) { 1035 consumeError(SecNameOrErr.takeError()); 1036 continue; 1037 } 1038 Sections[*SecNameOrErr] = Section; 1039 } 1040 for (auto Plt : ElfObj->getPltEntries()) { 1041 if (Plt.Symbol) { 1042 SymbolRef Symbol(*Plt.Symbol, ElfObj); 1043 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1044 if (Expected<StringRef> NameOrErr = Symbol.getName()) { 1045 if (!NameOrErr->empty()) 1046 AllSymbols[Sections[Plt.Section]].emplace_back( 1047 Plt.Address, Saver.save((*NameOrErr + "@plt").str()), SymbolType); 1048 continue; 1049 } else { 1050 // The warning has been reported in disassembleObject(). 1051 consumeError(NameOrErr.takeError()); 1052 } 1053 } 1054 reportWarning("PLT entry at 0x" + Twine::utohexstr(Plt.Address) + 1055 " references an invalid symbol", 1056 Obj.getFileName()); 1057 } 1058 } 1059 1060 // Normally the disassembly output will skip blocks of zeroes. This function 1061 // returns the number of zero bytes that can be skipped when dumping the 1062 // disassembly of the instructions in Buf. 1063 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 1064 // Find the number of leading zeroes. 1065 size_t N = 0; 1066 while (N < Buf.size() && !Buf[N]) 1067 ++N; 1068 1069 // We may want to skip blocks of zero bytes, but unless we see 1070 // at least 8 of them in a row. 1071 if (N < 8) 1072 return 0; 1073 1074 // We skip zeroes in multiples of 4 because do not want to truncate an 1075 // instruction if it starts with a zero byte. 1076 return N & ~0x3; 1077 } 1078 1079 // Returns a map from sections to their relocations. 1080 static std::map<SectionRef, std::vector<RelocationRef>> 1081 getRelocsMap(object::ObjectFile const &Obj) { 1082 std::map<SectionRef, std::vector<RelocationRef>> Ret; 1083 uint64_t I = (uint64_t)-1; 1084 for (SectionRef Sec : Obj.sections()) { 1085 ++I; 1086 Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection(); 1087 if (!RelocatedOrErr) 1088 reportError(Obj.getFileName(), 1089 "section (" + Twine(I) + 1090 "): failed to get a relocated section: " + 1091 toString(RelocatedOrErr.takeError())); 1092 1093 section_iterator Relocated = *RelocatedOrErr; 1094 if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep) 1095 continue; 1096 std::vector<RelocationRef> &V = Ret[*Relocated]; 1097 append_range(V, Sec.relocations()); 1098 // Sort relocations by address. 1099 llvm::stable_sort(V, isRelocAddressLess); 1100 } 1101 return Ret; 1102 } 1103 1104 // Used for --adjust-vma to check if address should be adjusted by the 1105 // specified value for a given section. 1106 // For ELF we do not adjust non-allocatable sections like debug ones, 1107 // because they are not loadable. 1108 // TODO: implement for other file formats. 1109 static bool shouldAdjustVA(const SectionRef &Section) { 1110 const ObjectFile *Obj = Section.getObject(); 1111 if (Obj->isELF()) 1112 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 1113 return false; 1114 } 1115 1116 1117 typedef std::pair<uint64_t, char> MappingSymbolPair; 1118 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols, 1119 uint64_t Address) { 1120 auto It = 1121 partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) { 1122 return Val.first <= Address; 1123 }); 1124 // Return zero for any address before the first mapping symbol; this means 1125 // we should use the default disassembly mode, depending on the target. 1126 if (It == MappingSymbols.begin()) 1127 return '\x00'; 1128 return (It - 1)->second; 1129 } 1130 1131 static uint64_t dumpARMELFData(uint64_t SectionAddr, uint64_t Index, 1132 uint64_t End, const ObjectFile &Obj, 1133 ArrayRef<uint8_t> Bytes, 1134 ArrayRef<MappingSymbolPair> MappingSymbols, 1135 const MCSubtargetInfo &STI, raw_ostream &OS) { 1136 support::endianness Endian = 1137 Obj.isLittleEndian() ? support::little : support::big; 1138 size_t Start = OS.tell(); 1139 OS << format("%8" PRIx64 ": ", SectionAddr + Index); 1140 if (Index + 4 <= End) { 1141 dumpBytes(Bytes.slice(Index, 4), OS); 1142 AlignToInstStartColumn(Start, STI, OS); 1143 OS << "\t.word\t" 1144 << format_hex(support::endian::read32(Bytes.data() + Index, Endian), 1145 10); 1146 return 4; 1147 } 1148 if (Index + 2 <= End) { 1149 dumpBytes(Bytes.slice(Index, 2), OS); 1150 AlignToInstStartColumn(Start, STI, OS); 1151 OS << "\t.short\t" 1152 << format_hex(support::endian::read16(Bytes.data() + Index, Endian), 6); 1153 return 2; 1154 } 1155 dumpBytes(Bytes.slice(Index, 1), OS); 1156 AlignToInstStartColumn(Start, STI, OS); 1157 OS << "\t.byte\t" << format_hex(Bytes[Index], 4); 1158 return 1; 1159 } 1160 1161 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 1162 ArrayRef<uint8_t> Bytes) { 1163 // print out data up to 8 bytes at a time in hex and ascii 1164 uint8_t AsciiData[9] = {'\0'}; 1165 uint8_t Byte; 1166 int NumBytes = 0; 1167 1168 for (; Index < End; ++Index) { 1169 if (NumBytes == 0) 1170 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1171 Byte = Bytes.slice(Index)[0]; 1172 outs() << format(" %02x", Byte); 1173 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1174 1175 uint8_t IndentOffset = 0; 1176 NumBytes++; 1177 if (Index == End - 1 || NumBytes > 8) { 1178 // Indent the space for less than 8 bytes data. 1179 // 2 spaces for byte and one for space between bytes 1180 IndentOffset = 3 * (8 - NumBytes); 1181 for (int Excess = NumBytes; Excess < 8; Excess++) 1182 AsciiData[Excess] = '\0'; 1183 NumBytes = 8; 1184 } 1185 if (NumBytes == 8) { 1186 AsciiData[8] = '\0'; 1187 outs() << std::string(IndentOffset, ' ') << " "; 1188 outs() << reinterpret_cast<char *>(AsciiData); 1189 outs() << '\n'; 1190 NumBytes = 0; 1191 } 1192 } 1193 } 1194 1195 SymbolInfoTy objdump::createSymbolInfo(const ObjectFile &Obj, 1196 const SymbolRef &Symbol, 1197 bool IsMappingSymbol) { 1198 const StringRef FileName = Obj.getFileName(); 1199 const uint64_t Addr = unwrapOrError(Symbol.getAddress(), FileName); 1200 const StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1201 1202 if (Obj.isXCOFF() && (SymbolDescription || TracebackTable)) { 1203 const auto &XCOFFObj = cast<XCOFFObjectFile>(Obj); 1204 DataRefImpl SymbolDRI = Symbol.getRawDataRefImpl(); 1205 1206 const uint32_t SymbolIndex = XCOFFObj.getSymbolIndex(SymbolDRI.p); 1207 std::optional<XCOFF::StorageMappingClass> Smc = 1208 getXCOFFSymbolCsectSMC(XCOFFObj, Symbol); 1209 return SymbolInfoTy(Smc, Addr, Name, SymbolIndex, 1210 isLabel(XCOFFObj, Symbol)); 1211 } else if (Obj.isXCOFF()) { 1212 const SymbolRef::Type SymType = unwrapOrError(Symbol.getType(), FileName); 1213 return SymbolInfoTy(Addr, Name, SymType, /*IsMappingSymbol=*/false, 1214 /*IsXCOFF=*/true); 1215 } else { 1216 uint8_t Type = 1217 Obj.isELF() ? getElfSymbolType(Obj, Symbol) : (uint8_t)ELF::STT_NOTYPE; 1218 return SymbolInfoTy(Addr, Name, Type, IsMappingSymbol); 1219 } 1220 } 1221 1222 static SymbolInfoTy createDummySymbolInfo(const ObjectFile &Obj, 1223 const uint64_t Addr, StringRef &Name, 1224 uint8_t Type) { 1225 if (Obj.isXCOFF() && (SymbolDescription || TracebackTable)) 1226 return SymbolInfoTy(std::nullopt, Addr, Name, std::nullopt, false); 1227 else 1228 return SymbolInfoTy(Addr, Name, Type); 1229 } 1230 1231 static void 1232 collectBBAddrMapLabels(const std::unordered_map<uint64_t, BBAddrMap> &AddrToBBAddrMap, 1233 uint64_t SectionAddr, uint64_t Start, uint64_t End, 1234 std::unordered_map<uint64_t, std::vector<std::string>> &Labels) { 1235 if (AddrToBBAddrMap.empty()) 1236 return; 1237 Labels.clear(); 1238 uint64_t StartAddress = SectionAddr + Start; 1239 uint64_t EndAddress = SectionAddr + End; 1240 auto Iter = AddrToBBAddrMap.find(StartAddress); 1241 if (Iter == AddrToBBAddrMap.end()) 1242 return; 1243 for (const BBAddrMap::BBEntry &BBEntry : Iter->second.BBEntries) { 1244 uint64_t BBAddress = BBEntry.Offset + Iter->second.Addr; 1245 if (BBAddress >= EndAddress) 1246 continue; 1247 Labels[BBAddress].push_back(("BB" + Twine(BBEntry.ID)).str()); 1248 } 1249 } 1250 1251 static void collectLocalBranchTargets( 1252 ArrayRef<uint8_t> Bytes, const MCInstrAnalysis *MIA, MCDisassembler *DisAsm, 1253 MCInstPrinter *IP, const MCSubtargetInfo *STI, uint64_t SectionAddr, 1254 uint64_t Start, uint64_t End, std::unordered_map<uint64_t, std::string> &Labels) { 1255 // So far only supports PowerPC and X86. 1256 if (!STI->getTargetTriple().isPPC() && !STI->getTargetTriple().isX86()) 1257 return; 1258 1259 Labels.clear(); 1260 unsigned LabelCount = 0; 1261 Start += SectionAddr; 1262 End += SectionAddr; 1263 uint64_t Index = Start; 1264 while (Index < End) { 1265 // Disassemble a real instruction and record function-local branch labels. 1266 MCInst Inst; 1267 uint64_t Size; 1268 ArrayRef<uint8_t> ThisBytes = Bytes.slice(Index - SectionAddr); 1269 bool Disassembled = 1270 DisAsm->getInstruction(Inst, Size, ThisBytes, Index, nulls()); 1271 if (Size == 0) 1272 Size = std::min<uint64_t>(ThisBytes.size(), 1273 DisAsm->suggestBytesToSkip(ThisBytes, Index)); 1274 1275 if (Disassembled && MIA) { 1276 uint64_t Target; 1277 bool TargetKnown = MIA->evaluateBranch(Inst, Index, Size, Target); 1278 // On PowerPC, if the address of a branch is the same as the target, it 1279 // means that it's a function call. Do not mark the label for this case. 1280 if (TargetKnown && (Target >= Start && Target < End) && 1281 !Labels.count(Target) && 1282 !(STI->getTargetTriple().isPPC() && Target == Index)) 1283 Labels[Target] = ("L" + Twine(LabelCount++)).str(); 1284 } 1285 Index += Size; 1286 } 1287 } 1288 1289 // Create an MCSymbolizer for the target and add it to the MCDisassembler. 1290 // This is currently only used on AMDGPU, and assumes the format of the 1291 // void * argument passed to AMDGPU's createMCSymbolizer. 1292 static void addSymbolizer( 1293 MCContext &Ctx, const Target *Target, StringRef TripleName, 1294 MCDisassembler *DisAsm, uint64_t SectionAddr, ArrayRef<uint8_t> Bytes, 1295 SectionSymbolsTy &Symbols, 1296 std::vector<std::unique_ptr<std::string>> &SynthesizedLabelNames) { 1297 1298 std::unique_ptr<MCRelocationInfo> RelInfo( 1299 Target->createMCRelocationInfo(TripleName, Ctx)); 1300 if (!RelInfo) 1301 return; 1302 std::unique_ptr<MCSymbolizer> Symbolizer(Target->createMCSymbolizer( 1303 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1304 MCSymbolizer *SymbolizerPtr = &*Symbolizer; 1305 DisAsm->setSymbolizer(std::move(Symbolizer)); 1306 1307 if (!SymbolizeOperands) 1308 return; 1309 1310 // Synthesize labels referenced by branch instructions by 1311 // disassembling, discarding the output, and collecting the referenced 1312 // addresses from the symbolizer. 1313 for (size_t Index = 0; Index != Bytes.size();) { 1314 MCInst Inst; 1315 uint64_t Size; 1316 ArrayRef<uint8_t> ThisBytes = Bytes.slice(Index); 1317 const uint64_t ThisAddr = SectionAddr + Index; 1318 DisAsm->getInstruction(Inst, Size, ThisBytes, ThisAddr, nulls()); 1319 if (Size == 0) 1320 Size = std::min<uint64_t>(ThisBytes.size(), 1321 DisAsm->suggestBytesToSkip(ThisBytes, Index)); 1322 Index += Size; 1323 } 1324 ArrayRef<uint64_t> LabelAddrsRef = SymbolizerPtr->getReferencedAddresses(); 1325 // Copy and sort to remove duplicates. 1326 std::vector<uint64_t> LabelAddrs; 1327 LabelAddrs.insert(LabelAddrs.end(), LabelAddrsRef.begin(), 1328 LabelAddrsRef.end()); 1329 llvm::sort(LabelAddrs); 1330 LabelAddrs.resize(std::unique(LabelAddrs.begin(), LabelAddrs.end()) - 1331 LabelAddrs.begin()); 1332 // Add the labels. 1333 for (unsigned LabelNum = 0; LabelNum != LabelAddrs.size(); ++LabelNum) { 1334 auto Name = std::make_unique<std::string>(); 1335 *Name = (Twine("L") + Twine(LabelNum)).str(); 1336 SynthesizedLabelNames.push_back(std::move(Name)); 1337 Symbols.push_back(SymbolInfoTy( 1338 LabelAddrs[LabelNum], *SynthesizedLabelNames.back(), ELF::STT_NOTYPE)); 1339 } 1340 llvm::stable_sort(Symbols); 1341 // Recreate the symbolizer with the new symbols list. 1342 RelInfo.reset(Target->createMCRelocationInfo(TripleName, Ctx)); 1343 Symbolizer.reset(Target->createMCSymbolizer( 1344 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1345 DisAsm->setSymbolizer(std::move(Symbolizer)); 1346 } 1347 1348 static StringRef getSegmentName(const MachOObjectFile *MachO, 1349 const SectionRef &Section) { 1350 if (MachO) { 1351 DataRefImpl DR = Section.getRawDataRefImpl(); 1352 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1353 return SegmentName; 1354 } 1355 return ""; 1356 } 1357 1358 static void emitPostInstructionInfo(formatted_raw_ostream &FOS, 1359 const MCAsmInfo &MAI, 1360 const MCSubtargetInfo &STI, 1361 StringRef Comments, 1362 LiveVariablePrinter &LVP) { 1363 do { 1364 if (!Comments.empty()) { 1365 // Emit a line of comments. 1366 StringRef Comment; 1367 std::tie(Comment, Comments) = Comments.split('\n'); 1368 // MAI.getCommentColumn() assumes that instructions are printed at the 1369 // position of 8, while getInstStartColumn() returns the actual position. 1370 unsigned CommentColumn = 1371 MAI.getCommentColumn() - 8 + getInstStartColumn(STI); 1372 FOS.PadToColumn(CommentColumn); 1373 FOS << MAI.getCommentString() << ' ' << Comment; 1374 } 1375 LVP.printAfterInst(FOS); 1376 FOS << '\n'; 1377 } while (!Comments.empty()); 1378 FOS.flush(); 1379 } 1380 1381 static void createFakeELFSections(ObjectFile &Obj) { 1382 assert(Obj.isELF()); 1383 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(&Obj)) 1384 Elf32LEObj->createFakeSections(); 1385 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(&Obj)) 1386 Elf64LEObj->createFakeSections(); 1387 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(&Obj)) 1388 Elf32BEObj->createFakeSections(); 1389 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(&Obj)) 1390 Elf64BEObj->createFakeSections(); 1391 else 1392 llvm_unreachable("Unsupported binary format"); 1393 } 1394 1395 // Tries to fetch a more complete version of the given object file using its 1396 // Build ID. Returns std::nullopt if nothing was found. 1397 static std::optional<OwningBinary<Binary>> 1398 fetchBinaryByBuildID(const ObjectFile &Obj) { 1399 object::BuildIDRef BuildID = getBuildID(&Obj); 1400 if (BuildID.empty()) 1401 return std::nullopt; 1402 std::optional<std::string> Path = BIDFetcher->fetch(BuildID); 1403 if (!Path) 1404 return std::nullopt; 1405 Expected<OwningBinary<Binary>> DebugBinary = createBinary(*Path); 1406 if (!DebugBinary) { 1407 reportWarning(toString(DebugBinary.takeError()), *Path); 1408 return std::nullopt; 1409 } 1410 return std::move(*DebugBinary); 1411 } 1412 1413 static void 1414 disassembleObject(ObjectFile &Obj, const ObjectFile &DbgObj, 1415 DisassemblerTarget &PrimaryTarget, 1416 std::optional<DisassemblerTarget> &SecondaryTarget, 1417 SourcePrinter &SP, bool InlineRelocs) { 1418 DisassemblerTarget *DT = &PrimaryTarget; 1419 bool PrimaryIsThumb = false; 1420 if (isArmElf(Obj)) 1421 PrimaryIsThumb = PrimaryTarget.SubtargetInfo->checkFeatures("+thumb-mode"); 1422 1423 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 1424 if (InlineRelocs) 1425 RelocMap = getRelocsMap(Obj); 1426 bool Is64Bits = Obj.getBytesInAddress() > 4; 1427 1428 // Create a mapping from virtual address to symbol name. This is used to 1429 // pretty print the symbols while disassembling. 1430 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1431 std::map<SectionRef, SmallVector<MappingSymbolPair, 0>> AllMappingSymbols; 1432 SectionSymbolsTy AbsoluteSymbols; 1433 const StringRef FileName = Obj.getFileName(); 1434 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(&Obj); 1435 for (const SymbolRef &Symbol : Obj.symbols()) { 1436 Expected<StringRef> NameOrErr = Symbol.getName(); 1437 if (!NameOrErr) { 1438 reportWarning(toString(NameOrErr.takeError()), FileName); 1439 continue; 1440 } 1441 if (NameOrErr->empty() && !(Obj.isXCOFF() && SymbolDescription)) 1442 continue; 1443 1444 if (Obj.isELF() && 1445 (cantFail(Symbol.getFlags()) & SymbolRef::SF_FormatSpecific)) { 1446 // Symbol is intended not to be displayed by default (STT_FILE, 1447 // STT_SECTION, or a mapping symbol). Ignore STT_SECTION symbols. We will 1448 // synthesize a section symbol if no symbol is defined at offset 0. 1449 // 1450 // For a mapping symbol, store it within both AllSymbols and 1451 // AllMappingSymbols. If --show-all-symbols is unspecified, its label will 1452 // not be printed in disassembly listing. 1453 if (getElfSymbolType(Obj, Symbol) != ELF::STT_SECTION && 1454 hasMappingSymbols(Obj)) { 1455 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1456 if (SecI != Obj.section_end()) { 1457 uint64_t SectionAddr = SecI->getAddress(); 1458 uint64_t Address = cantFail(Symbol.getAddress()); 1459 StringRef Name = *NameOrErr; 1460 if (Name.consume_front("$") && Name.size() && 1461 strchr("adtx", Name[0])) { 1462 AllMappingSymbols[*SecI].emplace_back(Address - SectionAddr, 1463 Name[0]); 1464 AllSymbols[*SecI].push_back( 1465 createSymbolInfo(Obj, Symbol, /*MappingSymbol=*/true)); 1466 } 1467 } 1468 } 1469 continue; 1470 } 1471 1472 if (MachO) { 1473 // __mh_(execute|dylib|dylinker|bundle|preload|object)_header are special 1474 // symbols that support MachO header introspection. They do not bind to 1475 // code locations and are irrelevant for disassembly. 1476 if (NameOrErr->startswith("__mh_") && NameOrErr->endswith("_header")) 1477 continue; 1478 // Don't ask a Mach-O STAB symbol for its section unless you know that 1479 // STAB symbol's section field refers to a valid section index. Otherwise 1480 // the symbol may error trying to load a section that does not exist. 1481 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1482 uint8_t NType = (MachO->is64Bit() ? 1483 MachO->getSymbol64TableEntry(SymDRI).n_type: 1484 MachO->getSymbolTableEntry(SymDRI).n_type); 1485 if (NType & MachO::N_STAB) 1486 continue; 1487 } 1488 1489 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1490 if (SecI != Obj.section_end()) 1491 AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol)); 1492 else 1493 AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol)); 1494 } 1495 1496 if (AllSymbols.empty() && Obj.isELF()) 1497 addDynamicElfSymbols(cast<ELFObjectFileBase>(Obj), AllSymbols); 1498 1499 if (Obj.isWasm()) 1500 addMissingWasmCodeSymbols(cast<WasmObjectFile>(Obj), AllSymbols); 1501 1502 if (Obj.isELF() && Obj.sections().empty()) 1503 createFakeELFSections(Obj); 1504 1505 BumpPtrAllocator A; 1506 StringSaver Saver(A); 1507 addPltEntries(Obj, AllSymbols, Saver); 1508 1509 // Create a mapping from virtual address to section. An empty section can 1510 // cause more than one section at the same address. Sort such sections to be 1511 // before same-addressed non-empty sections so that symbol lookups prefer the 1512 // non-empty section. 1513 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1514 for (SectionRef Sec : Obj.sections()) 1515 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1516 llvm::stable_sort(SectionAddresses, [](const auto &LHS, const auto &RHS) { 1517 if (LHS.first != RHS.first) 1518 return LHS.first < RHS.first; 1519 return LHS.second.getSize() < RHS.second.getSize(); 1520 }); 1521 1522 // Linked executables (.exe and .dll files) typically don't include a real 1523 // symbol table but they might contain an export table. 1524 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(&Obj)) { 1525 for (const auto &ExportEntry : COFFObj->export_directories()) { 1526 StringRef Name; 1527 if (Error E = ExportEntry.getSymbolName(Name)) 1528 reportError(std::move(E), Obj.getFileName()); 1529 if (Name.empty()) 1530 continue; 1531 1532 uint32_t RVA; 1533 if (Error E = ExportEntry.getExportRVA(RVA)) 1534 reportError(std::move(E), Obj.getFileName()); 1535 1536 uint64_t VA = COFFObj->getImageBase() + RVA; 1537 auto Sec = partition_point( 1538 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) { 1539 return O.first <= VA; 1540 }); 1541 if (Sec != SectionAddresses.begin()) { 1542 --Sec; 1543 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1544 } else 1545 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1546 } 1547 } 1548 1549 // Sort all the symbols, this allows us to use a simple binary search to find 1550 // Multiple symbols can have the same address. Use a stable sort to stabilize 1551 // the output. 1552 StringSet<> FoundDisasmSymbolSet; 1553 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1554 llvm::stable_sort(SecSyms.second); 1555 llvm::stable_sort(AbsoluteSymbols); 1556 1557 std::unique_ptr<DWARFContext> DICtx; 1558 LiveVariablePrinter LVP(*DT->Context->getRegisterInfo(), *DT->SubtargetInfo); 1559 1560 if (DbgVariables != DVDisabled) { 1561 DICtx = DWARFContext::create(DbgObj); 1562 for (const std::unique_ptr<DWARFUnit> &CU : DICtx->compile_units()) 1563 LVP.addCompileUnit(CU->getUnitDIE(false)); 1564 } 1565 1566 LLVM_DEBUG(LVP.dump()); 1567 1568 std::unordered_map<uint64_t, BBAddrMap> AddrToBBAddrMap; 1569 auto ReadBBAddrMap = [&](std::optional<unsigned> SectionIndex = 1570 std::nullopt) { 1571 AddrToBBAddrMap.clear(); 1572 if (const auto *Elf = dyn_cast<ELFObjectFileBase>(&Obj)) { 1573 auto BBAddrMapsOrErr = Elf->readBBAddrMap(SectionIndex); 1574 if (!BBAddrMapsOrErr) { 1575 reportWarning(toString(BBAddrMapsOrErr.takeError()), Obj.getFileName()); 1576 return; 1577 } 1578 for (auto &FunctionBBAddrMap : *BBAddrMapsOrErr) 1579 AddrToBBAddrMap.emplace(FunctionBBAddrMap.Addr, 1580 std::move(FunctionBBAddrMap)); 1581 } 1582 }; 1583 1584 // For non-relocatable objects, Read all LLVM_BB_ADDR_MAP sections into a 1585 // single mapping, since they don't have any conflicts. 1586 if (SymbolizeOperands && !Obj.isRelocatableObject()) 1587 ReadBBAddrMap(); 1588 1589 for (const SectionRef &Section : ToolSectionFilter(Obj)) { 1590 if (FilterSections.empty() && !DisassembleAll && 1591 (!Section.isText() || Section.isVirtual())) 1592 continue; 1593 1594 uint64_t SectionAddr = Section.getAddress(); 1595 uint64_t SectSize = Section.getSize(); 1596 if (!SectSize) 1597 continue; 1598 1599 // For relocatable object files, read the LLVM_BB_ADDR_MAP section 1600 // corresponding to this section, if present. 1601 if (SymbolizeOperands && Obj.isRelocatableObject()) 1602 ReadBBAddrMap(Section.getIndex()); 1603 1604 // Get the list of all the symbols in this section. 1605 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1606 auto &MappingSymbols = AllMappingSymbols[Section]; 1607 llvm::sort(MappingSymbols); 1608 1609 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef( 1610 unwrapOrError(Section.getContents(), Obj.getFileName())); 1611 1612 std::vector<std::unique_ptr<std::string>> SynthesizedLabelNames; 1613 if (Obj.isELF() && Obj.getArch() == Triple::amdgcn) { 1614 // AMDGPU disassembler uses symbolizer for printing labels 1615 addSymbolizer(*DT->Context, DT->TheTarget, TripleName, DT->DisAsm.get(), 1616 SectionAddr, Bytes, Symbols, SynthesizedLabelNames); 1617 } 1618 1619 StringRef SegmentName = getSegmentName(MachO, Section); 1620 StringRef SectionName = unwrapOrError(Section.getName(), Obj.getFileName()); 1621 // If the section has no symbol at the start, just insert a dummy one. 1622 // Without --show-all-symbols, also insert one if all symbols at the start 1623 // are mapping symbols. 1624 bool CreateDummy = Symbols.empty(); 1625 if (!CreateDummy) { 1626 CreateDummy = true; 1627 for (auto &Sym : Symbols) { 1628 if (Sym.Addr != SectionAddr) 1629 break; 1630 if (!Sym.IsMappingSymbol || ShowAllSymbols) 1631 CreateDummy = false; 1632 } 1633 } 1634 if (CreateDummy) { 1635 SymbolInfoTy Sym = createDummySymbolInfo( 1636 Obj, SectionAddr, SectionName, 1637 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT); 1638 if (Obj.isXCOFF()) 1639 Symbols.insert(Symbols.begin(), Sym); 1640 else 1641 Symbols.insert(llvm::lower_bound(Symbols, Sym), Sym); 1642 } 1643 1644 SmallString<40> Comments; 1645 raw_svector_ostream CommentStream(Comments); 1646 1647 uint64_t VMAAdjustment = 0; 1648 if (shouldAdjustVA(Section)) 1649 VMAAdjustment = AdjustVMA; 1650 1651 // In executable and shared objects, r_offset holds a virtual address. 1652 // Subtract SectionAddr from the r_offset field of a relocation to get 1653 // the section offset. 1654 uint64_t RelAdjustment = Obj.isRelocatableObject() ? 0 : SectionAddr; 1655 uint64_t Size; 1656 uint64_t Index; 1657 bool PrintedSection = false; 1658 std::vector<RelocationRef> Rels = RelocMap[Section]; 1659 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1660 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1661 1662 // Loop over each chunk of code between two points where at least 1663 // one symbol is defined. 1664 for (size_t SI = 0, SE = Symbols.size(); SI != SE;) { 1665 // Advance SI past all the symbols starting at the same address, 1666 // and make an ArrayRef of them. 1667 unsigned FirstSI = SI; 1668 uint64_t Start = Symbols[SI].Addr; 1669 ArrayRef<SymbolInfoTy> SymbolsHere; 1670 while (SI != SE && Symbols[SI].Addr == Start) 1671 ++SI; 1672 SymbolsHere = ArrayRef<SymbolInfoTy>(&Symbols[FirstSI], SI - FirstSI); 1673 1674 // Get the demangled names of all those symbols. We end up with a vector 1675 // of StringRef that holds the names we're going to use, and a vector of 1676 // std::string that stores the new strings returned by demangle(), if 1677 // any. If we don't call demangle() then that vector can stay empty. 1678 std::vector<StringRef> SymNamesHere; 1679 std::vector<std::string> DemangledSymNamesHere; 1680 if (Demangle) { 1681 // Fetch the demangled names and store them locally. 1682 for (const SymbolInfoTy &Symbol : SymbolsHere) 1683 DemangledSymNamesHere.push_back(demangle(Symbol.Name)); 1684 // Now we've finished modifying that vector, it's safe to make 1685 // a vector of StringRefs pointing into it. 1686 SymNamesHere.insert(SymNamesHere.begin(), DemangledSymNamesHere.begin(), 1687 DemangledSymNamesHere.end()); 1688 } else { 1689 for (const SymbolInfoTy &Symbol : SymbolsHere) 1690 SymNamesHere.push_back(Symbol.Name); 1691 } 1692 1693 // Distinguish ELF data from code symbols, which will be used later on to 1694 // decide whether to 'disassemble' this chunk as a data declaration via 1695 // dumpELFData(), or whether to treat it as code. 1696 // 1697 // If data _and_ code symbols are defined at the same address, the code 1698 // takes priority, on the grounds that disassembling code is our main 1699 // purpose here, and it would be a worse failure to _not_ interpret 1700 // something that _was_ meaningful as code than vice versa. 1701 // 1702 // Any ELF symbol type that is not clearly data will be regarded as code. 1703 // In particular, one of the uses of STT_NOTYPE is for branch targets 1704 // inside functions, for which STT_FUNC would be inaccurate. 1705 // 1706 // So here, we spot whether there's any non-data symbol present at all, 1707 // and only set the DisassembleAsData flag if there isn't. Also, we use 1708 // this distinction to inform the decision of which symbol to print at 1709 // the head of the section, so that if we're printing code, we print a 1710 // code-related symbol name to go with it. 1711 bool DisassembleAsData = false; 1712 size_t DisplaySymIndex = SymbolsHere.size() - 1; 1713 if (Obj.isELF() && !DisassembleAll && Section.isText()) { 1714 DisassembleAsData = true; // unless we find a code symbol below 1715 1716 for (size_t i = 0; i < SymbolsHere.size(); ++i) { 1717 uint8_t SymTy = SymbolsHere[i].Type; 1718 if (SymTy != ELF::STT_OBJECT && SymTy != ELF::STT_COMMON) { 1719 DisassembleAsData = false; 1720 DisplaySymIndex = i; 1721 } 1722 } 1723 } 1724 1725 // Decide which symbol(s) from this collection we're going to print. 1726 std::vector<bool> SymsToPrint(SymbolsHere.size(), false); 1727 // If the user has given the --disassemble-symbols option, then we must 1728 // display every symbol in that set, and no others. 1729 if (!DisasmSymbolSet.empty()) { 1730 bool FoundAny = false; 1731 for (size_t i = 0; i < SymbolsHere.size(); ++i) { 1732 if (DisasmSymbolSet.count(SymNamesHere[i])) { 1733 SymsToPrint[i] = true; 1734 FoundAny = true; 1735 } 1736 } 1737 1738 // And if none of the symbols here is one that the user asked for, skip 1739 // disassembling this entire chunk of code. 1740 if (!FoundAny) 1741 continue; 1742 } else if (!SymbolsHere[DisplaySymIndex].IsMappingSymbol) { 1743 // Otherwise, print whichever symbol at this location is last in the 1744 // Symbols array, because that array is pre-sorted in a way intended to 1745 // correlate with priority of which symbol to display. 1746 SymsToPrint[DisplaySymIndex] = true; 1747 } 1748 1749 // Now that we know we're disassembling this section, override the choice 1750 // of which symbols to display by printing _all_ of them at this address 1751 // if the user asked for all symbols. 1752 // 1753 // That way, '--show-all-symbols --disassemble-symbol=foo' will print 1754 // only the chunk of code headed by 'foo', but also show any other 1755 // symbols defined at that address, such as aliases for 'foo', or the ARM 1756 // mapping symbol preceding its code. 1757 if (ShowAllSymbols) { 1758 for (size_t i = 0; i < SymbolsHere.size(); ++i) 1759 SymsToPrint[i] = true; 1760 } 1761 1762 if (Start < SectionAddr || StopAddress <= Start) 1763 continue; 1764 1765 for (size_t i = 0; i < SymbolsHere.size(); ++i) 1766 FoundDisasmSymbolSet.insert(SymNamesHere[i]); 1767 1768 // The end is the section end, the beginning of the next symbol, or 1769 // --stop-address. 1770 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress); 1771 if (SI < SE) 1772 End = std::min(End, Symbols[SI].Addr); 1773 if (Start >= End || End <= StartAddress) 1774 continue; 1775 Start -= SectionAddr; 1776 End -= SectionAddr; 1777 1778 if (!PrintedSection) { 1779 PrintedSection = true; 1780 outs() << "\nDisassembly of section "; 1781 if (!SegmentName.empty()) 1782 outs() << SegmentName << ","; 1783 outs() << SectionName << ":\n"; 1784 } 1785 1786 bool PrintedLabel = false; 1787 for (size_t i = 0; i < SymbolsHere.size(); ++i) { 1788 if (!SymsToPrint[i]) 1789 continue; 1790 1791 const SymbolInfoTy &Symbol = SymbolsHere[i]; 1792 const StringRef SymbolName = SymNamesHere[i]; 1793 1794 if (!PrintedLabel) { 1795 outs() << '\n'; 1796 PrintedLabel = true; 1797 } 1798 if (LeadingAddr) 1799 outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ", 1800 SectionAddr + Start + VMAAdjustment); 1801 if (Obj.isXCOFF() && SymbolDescription) { 1802 outs() << getXCOFFSymbolDescription(Symbol, SymbolName) << ":\n"; 1803 } else 1804 outs() << '<' << SymbolName << ">:\n"; 1805 } 1806 1807 // Don't print raw contents of a virtual section. A virtual section 1808 // doesn't have any contents in the file. 1809 if (Section.isVirtual()) { 1810 outs() << "...\n"; 1811 continue; 1812 } 1813 1814 // See if any of the symbols defined at this location triggers target- 1815 // specific disassembly behavior, e.g. of special descriptors or function 1816 // prelude information. 1817 // 1818 // We stop this loop at the first symbol that triggers some kind of 1819 // interesting behavior (if any), on the assumption that if two symbols 1820 // defined at the same address trigger two conflicting symbol handlers, 1821 // the object file is probably confused anyway, and it would make even 1822 // less sense to present the output of _both_ handlers, because that 1823 // would describe the same data twice. 1824 for (size_t SHI = 0; SHI < SymbolsHere.size(); ++SHI) { 1825 SymbolInfoTy Symbol = SymbolsHere[SHI]; 1826 1827 auto Status = DT->DisAsm->onSymbolStart( 1828 Symbol, Size, Bytes.slice(Start, End - Start), SectionAddr + Start, 1829 CommentStream); 1830 1831 if (!Status) { 1832 // If onSymbolStart returns std::nullopt, that means it didn't trigger 1833 // any interesting handling for this symbol. Try the other symbols 1834 // defined at this address. 1835 continue; 1836 } 1837 1838 if (*Status == MCDisassembler::Fail) { 1839 // If onSymbolStart returns Fail, that means it identified some kind 1840 // of special data at this address, but wasn't able to disassemble it 1841 // meaningfully. So we fall back to disassembling the failed region 1842 // as bytes, assuming that the target detected the failure before 1843 // printing anything. 1844 // 1845 // Return values Success or SoftFail (i.e no 'real' failure) are 1846 // expected to mean that the target has emitted its own output. 1847 // 1848 // Either way, 'Size' will have been set to the amount of data 1849 // covered by whatever prologue the target identified. So we advance 1850 // our own position to beyond that. Sometimes that will be the entire 1851 // distance to the next symbol, and sometimes it will be just a 1852 // prologue and we should start disassembling instructions from where 1853 // it left off. 1854 outs() << DT->Context->getAsmInfo()->getCommentString() 1855 << " error in decoding " << SymNamesHere[SHI] 1856 << " : decoding failed region as bytes.\n"; 1857 for (uint64_t I = 0; I < Size; ++I) { 1858 outs() << "\t.byte\t " << format_hex(Bytes[I], 1, /*Upper=*/true) 1859 << "\n"; 1860 } 1861 } 1862 Start += Size; 1863 break; 1864 } 1865 1866 Index = Start; 1867 if (SectionAddr < StartAddress) 1868 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1869 1870 if (DisassembleAsData) { 1871 dumpELFData(SectionAddr, Index, End, Bytes); 1872 Index = End; 1873 continue; 1874 } 1875 1876 bool DumpARMELFData = false; 1877 bool DumpTracebackTableForXCOFFFunction = 1878 Obj.isXCOFF() && Section.isText() && TracebackTable && 1879 Symbols[SI - 1].XCOFFSymInfo.StorageMappingClass && 1880 (*Symbols[SI - 1].XCOFFSymInfo.StorageMappingClass == XCOFF::XMC_PR); 1881 1882 formatted_raw_ostream FOS(outs()); 1883 1884 std::unordered_map<uint64_t, std::string> AllLabels; 1885 std::unordered_map<uint64_t, std::vector<std::string>> BBAddrMapLabels; 1886 if (SymbolizeOperands) { 1887 collectLocalBranchTargets(Bytes, DT->InstrAnalysis.get(), 1888 DT->DisAsm.get(), DT->InstPrinter.get(), 1889 PrimaryTarget.SubtargetInfo.get(), 1890 SectionAddr, Index, End, AllLabels); 1891 collectBBAddrMapLabels(AddrToBBAddrMap, SectionAddr, Index, End, 1892 BBAddrMapLabels); 1893 } 1894 1895 while (Index < End) { 1896 // ARM and AArch64 ELF binaries can interleave data and text in the 1897 // same section. We rely on the markers introduced to understand what 1898 // we need to dump. If the data marker is within a function, it is 1899 // denoted as a word/short etc. 1900 if (!MappingSymbols.empty()) { 1901 char Kind = getMappingSymbolKind(MappingSymbols, Index); 1902 DumpARMELFData = Kind == 'd'; 1903 if (SecondaryTarget) { 1904 if (Kind == 'a') { 1905 DT = PrimaryIsThumb ? &*SecondaryTarget : &PrimaryTarget; 1906 } else if (Kind == 't') { 1907 DT = PrimaryIsThumb ? &PrimaryTarget : &*SecondaryTarget; 1908 } 1909 } 1910 } 1911 1912 if (DumpARMELFData) { 1913 Size = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1914 MappingSymbols, *DT->SubtargetInfo, FOS); 1915 } else { 1916 // When -z or --disassemble-zeroes are given we always dissasemble 1917 // them. Otherwise we might want to skip zero bytes we see. 1918 if (!DisassembleZeroes) { 1919 uint64_t MaxOffset = End - Index; 1920 // For --reloc: print zero blocks patched by relocations, so that 1921 // relocations can be shown in the dump. 1922 if (RelCur != RelEnd) 1923 MaxOffset = std::min(RelCur->getOffset() - RelAdjustment - Index, 1924 MaxOffset); 1925 1926 if (size_t N = 1927 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1928 FOS << "\t\t..." << '\n'; 1929 Index += N; 1930 continue; 1931 } 1932 } 1933 1934 if (DumpTracebackTableForXCOFFFunction && 1935 doesXCOFFTracebackTableBegin(Bytes.slice(Index, 4))) { 1936 dumpTracebackTable(Bytes.slice(Index), 1937 SectionAddr + Index + VMAAdjustment, FOS, 1938 SectionAddr + End + VMAAdjustment, 1939 *DT->SubtargetInfo, cast<XCOFFObjectFile>(&Obj)); 1940 Index = End; 1941 continue; 1942 } 1943 1944 // Print local label if there's any. 1945 auto Iter1 = BBAddrMapLabels.find(SectionAddr + Index); 1946 if (Iter1 != BBAddrMapLabels.end()) { 1947 for (StringRef Label : Iter1->second) 1948 FOS << "<" << Label << ">:\n"; 1949 } else { 1950 auto Iter2 = AllLabels.find(SectionAddr + Index); 1951 if (Iter2 != AllLabels.end()) 1952 FOS << "<" << Iter2->second << ">:\n"; 1953 } 1954 1955 // Disassemble a real instruction or a data when disassemble all is 1956 // provided 1957 MCInst Inst; 1958 ArrayRef<uint8_t> ThisBytes = Bytes.slice(Index); 1959 uint64_t ThisAddr = SectionAddr + Index; 1960 bool Disassembled = DT->DisAsm->getInstruction( 1961 Inst, Size, ThisBytes, ThisAddr, CommentStream); 1962 if (Size == 0) 1963 Size = std::min<uint64_t>( 1964 ThisBytes.size(), 1965 DT->DisAsm->suggestBytesToSkip(ThisBytes, ThisAddr)); 1966 1967 LVP.update({Index, Section.getIndex()}, 1968 {Index + Size, Section.getIndex()}, Index + Size != End); 1969 1970 DT->InstPrinter->setCommentStream(CommentStream); 1971 1972 DT->Printer->printInst( 1973 *DT->InstPrinter, Disassembled ? &Inst : nullptr, 1974 Bytes.slice(Index, Size), 1975 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, FOS, 1976 "", *DT->SubtargetInfo, &SP, Obj.getFileName(), &Rels, LVP); 1977 1978 DT->InstPrinter->setCommentStream(llvm::nulls()); 1979 1980 // If disassembly has failed, avoid analysing invalid/incomplete 1981 // instruction information. Otherwise, try to resolve the target 1982 // address (jump target or memory operand address) and print it on the 1983 // right of the instruction. 1984 if (Disassembled && DT->InstrAnalysis) { 1985 // Branch targets are printed just after the instructions. 1986 llvm::raw_ostream *TargetOS = &FOS; 1987 uint64_t Target; 1988 bool PrintTarget = DT->InstrAnalysis->evaluateBranch( 1989 Inst, SectionAddr + Index, Size, Target); 1990 if (!PrintTarget) 1991 if (std::optional<uint64_t> MaybeTarget = 1992 DT->InstrAnalysis->evaluateMemoryOperandAddress( 1993 Inst, DT->SubtargetInfo.get(), SectionAddr + Index, 1994 Size)) { 1995 Target = *MaybeTarget; 1996 PrintTarget = true; 1997 // Do not print real address when symbolizing. 1998 if (!SymbolizeOperands) { 1999 // Memory operand addresses are printed as comments. 2000 TargetOS = &CommentStream; 2001 *TargetOS << "0x" << Twine::utohexstr(Target); 2002 } 2003 } 2004 if (PrintTarget) { 2005 // In a relocatable object, the target's section must reside in 2006 // the same section as the call instruction or it is accessed 2007 // through a relocation. 2008 // 2009 // In a non-relocatable object, the target may be in any section. 2010 // In that case, locate the section(s) containing the target 2011 // address and find the symbol in one of those, if possible. 2012 // 2013 // N.B. We don't walk the relocations in the relocatable case yet. 2014 std::vector<const SectionSymbolsTy *> TargetSectionSymbols; 2015 if (!Obj.isRelocatableObject()) { 2016 auto It = llvm::partition_point( 2017 SectionAddresses, 2018 [=](const std::pair<uint64_t, SectionRef> &O) { 2019 return O.first <= Target; 2020 }); 2021 uint64_t TargetSecAddr = 0; 2022 while (It != SectionAddresses.begin()) { 2023 --It; 2024 if (TargetSecAddr == 0) 2025 TargetSecAddr = It->first; 2026 if (It->first != TargetSecAddr) 2027 break; 2028 TargetSectionSymbols.push_back(&AllSymbols[It->second]); 2029 } 2030 } else { 2031 TargetSectionSymbols.push_back(&Symbols); 2032 } 2033 TargetSectionSymbols.push_back(&AbsoluteSymbols); 2034 2035 // Find the last symbol in the first candidate section whose 2036 // offset is less than or equal to the target. If there are no 2037 // such symbols, try in the next section and so on, before finally 2038 // using the nearest preceding absolute symbol (if any), if there 2039 // are no other valid symbols. 2040 const SymbolInfoTy *TargetSym = nullptr; 2041 for (const SectionSymbolsTy *TargetSymbols : 2042 TargetSectionSymbols) { 2043 auto It = llvm::partition_point( 2044 *TargetSymbols, 2045 [=](const SymbolInfoTy &O) { return O.Addr <= Target; }); 2046 while (It != TargetSymbols->begin()) { 2047 --It; 2048 // Skip mapping symbols to avoid possible ambiguity as they 2049 // do not allow uniquely identifying the target address. 2050 if (!It->IsMappingSymbol) { 2051 TargetSym = &*It; 2052 break; 2053 } 2054 } 2055 if (TargetSym) 2056 break; 2057 } 2058 2059 // Print the labels corresponding to the target if there's any. 2060 bool BBAddrMapLabelAvailable = BBAddrMapLabels.count(Target); 2061 bool LabelAvailable = AllLabels.count(Target); 2062 if (TargetSym != nullptr) { 2063 uint64_t TargetAddress = TargetSym->Addr; 2064 uint64_t Disp = Target - TargetAddress; 2065 std::string TargetName = Demangle ? demangle(TargetSym->Name) 2066 : TargetSym->Name.str(); 2067 2068 *TargetOS << " <"; 2069 if (!Disp) { 2070 // Always Print the binary symbol precisely corresponding to 2071 // the target address. 2072 *TargetOS << TargetName; 2073 } else if (BBAddrMapLabelAvailable) { 2074 *TargetOS << BBAddrMapLabels[Target].front(); 2075 } else if (LabelAvailable) { 2076 *TargetOS << AllLabels[Target]; 2077 } else { 2078 // Always Print the binary symbol plus an offset if there's no 2079 // local label corresponding to the target address. 2080 *TargetOS << TargetName << "+0x" << Twine::utohexstr(Disp); 2081 } 2082 *TargetOS << ">"; 2083 } else if (BBAddrMapLabelAvailable) { 2084 *TargetOS << " <" << BBAddrMapLabels[Target].front() << ">"; 2085 } else if (LabelAvailable) { 2086 *TargetOS << " <" << AllLabels[Target] << ">"; 2087 } 2088 // By convention, each record in the comment stream should be 2089 // terminated. 2090 if (TargetOS == &CommentStream) 2091 *TargetOS << "\n"; 2092 } 2093 } 2094 } 2095 2096 assert(DT->Context->getAsmInfo()); 2097 emitPostInstructionInfo(FOS, *DT->Context->getAsmInfo(), 2098 *DT->SubtargetInfo, CommentStream.str(), LVP); 2099 Comments.clear(); 2100 2101 // Hexagon does this in pretty printer 2102 if (Obj.getArch() != Triple::hexagon) { 2103 // Print relocation for instruction and data. 2104 while (RelCur != RelEnd) { 2105 uint64_t Offset = RelCur->getOffset() - RelAdjustment; 2106 // If this relocation is hidden, skip it. 2107 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 2108 ++RelCur; 2109 continue; 2110 } 2111 2112 // Stop when RelCur's offset is past the disassembled 2113 // instruction/data. Note that it's possible the disassembled data 2114 // is not the complete data: we might see the relocation printed in 2115 // the middle of the data, but this matches the binutils objdump 2116 // output. 2117 if (Offset >= Index + Size) 2118 break; 2119 2120 // When --adjust-vma is used, update the address printed. 2121 if (RelCur->getSymbol() != Obj.symbol_end()) { 2122 Expected<section_iterator> SymSI = 2123 RelCur->getSymbol()->getSection(); 2124 if (SymSI && *SymSI != Obj.section_end() && 2125 shouldAdjustVA(**SymSI)) 2126 Offset += AdjustVMA; 2127 } 2128 2129 printRelocation(FOS, Obj.getFileName(), *RelCur, 2130 SectionAddr + Offset, Is64Bits); 2131 LVP.printAfterOtherLine(FOS, true); 2132 ++RelCur; 2133 } 2134 } 2135 2136 Index += Size; 2137 } 2138 } 2139 } 2140 StringSet<> MissingDisasmSymbolSet = 2141 set_difference(DisasmSymbolSet, FoundDisasmSymbolSet); 2142 for (StringRef Sym : MissingDisasmSymbolSet.keys()) 2143 reportWarning("failed to disassemble missing symbol " + Sym, FileName); 2144 } 2145 2146 static void disassembleObject(ObjectFile *Obj, bool InlineRelocs) { 2147 // If information useful for showing the disassembly is missing, try to find a 2148 // more complete binary and disassemble that instead. 2149 OwningBinary<Binary> FetchedBinary; 2150 if (Obj->symbols().empty()) { 2151 if (std::optional<OwningBinary<Binary>> FetchedBinaryOpt = 2152 fetchBinaryByBuildID(*Obj)) { 2153 if (auto *O = dyn_cast<ObjectFile>(FetchedBinaryOpt->getBinary())) { 2154 if (!O->symbols().empty() || 2155 (!O->sections().empty() && Obj->sections().empty())) { 2156 FetchedBinary = std::move(*FetchedBinaryOpt); 2157 Obj = O; 2158 } 2159 } 2160 } 2161 } 2162 2163 const Target *TheTarget = getTarget(Obj); 2164 2165 // Package up features to be passed to target/subtarget 2166 Expected<SubtargetFeatures> FeaturesValue = Obj->getFeatures(); 2167 if (!FeaturesValue) 2168 reportError(FeaturesValue.takeError(), Obj->getFileName()); 2169 SubtargetFeatures Features = *FeaturesValue; 2170 if (!MAttrs.empty()) { 2171 for (unsigned I = 0; I != MAttrs.size(); ++I) 2172 Features.AddFeature(MAttrs[I]); 2173 } else if (MCPU.empty() && Obj->getArch() == llvm::Triple::aarch64) { 2174 Features.AddFeature("+all"); 2175 } 2176 2177 if (MCPU.empty()) 2178 MCPU = Obj->tryGetCPUName().value_or("").str(); 2179 2180 if (isArmElf(*Obj)) { 2181 // When disassembling big-endian Arm ELF, the instruction endianness is 2182 // determined in a complex way. In relocatable objects, AAELF32 mandates 2183 // that instruction endianness matches the ELF file endianness; in 2184 // executable images, that's true unless the file header has the EF_ARM_BE8 2185 // flag, in which case instructions are little-endian regardless of data 2186 // endianness. 2187 // 2188 // We must set the big-endian-instructions SubtargetFeature to make the 2189 // disassembler read the instructions the right way round, and also tell 2190 // our own prettyprinter to retrieve the encodings the same way to print in 2191 // hex. 2192 const auto *Elf32BE = dyn_cast<ELF32BEObjectFile>(Obj); 2193 2194 if (Elf32BE && (Elf32BE->isRelocatableObject() || 2195 !(Elf32BE->getPlatformFlags() & ELF::EF_ARM_BE8))) { 2196 Features.AddFeature("+big-endian-instructions"); 2197 ARMPrettyPrinterInst.setInstructionEndianness(llvm::support::big); 2198 } else { 2199 ARMPrettyPrinterInst.setInstructionEndianness(llvm::support::little); 2200 } 2201 } 2202 2203 DisassemblerTarget PrimaryTarget(TheTarget, *Obj, TripleName, MCPU, Features); 2204 2205 // If we have an ARM object file, we need a second disassembler, because 2206 // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode. 2207 // We use mapping symbols to switch between the two assemblers, where 2208 // appropriate. 2209 std::optional<DisassemblerTarget> SecondaryTarget; 2210 2211 if (isArmElf(*Obj)) { 2212 if (!PrimaryTarget.SubtargetInfo->checkFeatures("+mclass")) { 2213 if (PrimaryTarget.SubtargetInfo->checkFeatures("+thumb-mode")) 2214 Features.AddFeature("-thumb-mode"); 2215 else 2216 Features.AddFeature("+thumb-mode"); 2217 SecondaryTarget.emplace(PrimaryTarget, Features); 2218 } 2219 } 2220 2221 const ObjectFile *DbgObj = Obj; 2222 if (!FetchedBinary.getBinary() && !Obj->hasDebugInfo()) { 2223 if (std::optional<OwningBinary<Binary>> DebugBinaryOpt = 2224 fetchBinaryByBuildID(*Obj)) { 2225 if (auto *FetchedObj = 2226 dyn_cast<const ObjectFile>(DebugBinaryOpt->getBinary())) { 2227 if (FetchedObj->hasDebugInfo()) { 2228 FetchedBinary = std::move(*DebugBinaryOpt); 2229 DbgObj = FetchedObj; 2230 } 2231 } 2232 } 2233 } 2234 2235 std::unique_ptr<object::Binary> DSYMBinary; 2236 std::unique_ptr<MemoryBuffer> DSYMBuf; 2237 if (!DbgObj->hasDebugInfo()) { 2238 if (const MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&*Obj)) { 2239 DbgObj = objdump::getMachODSymObject(MachOOF, Obj->getFileName(), 2240 DSYMBinary, DSYMBuf); 2241 if (!DbgObj) 2242 return; 2243 } 2244 } 2245 2246 SourcePrinter SP(DbgObj, TheTarget->getName()); 2247 2248 for (StringRef Opt : DisassemblerOptions) 2249 if (!PrimaryTarget.InstPrinter->applyTargetSpecificCLOption(Opt)) 2250 reportError(Obj->getFileName(), 2251 "Unrecognized disassembler option: " + Opt); 2252 2253 disassembleObject(*Obj, *DbgObj, PrimaryTarget, SecondaryTarget, SP, 2254 InlineRelocs); 2255 } 2256 2257 void Dumper::printRelocations() { 2258 StringRef Fmt = O.getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2259 2260 // Build a mapping from relocation target to a vector of relocation 2261 // sections. Usually, there is an only one relocation section for 2262 // each relocated section. 2263 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec; 2264 uint64_t Ndx; 2265 for (const SectionRef &Section : ToolSectionFilter(O, &Ndx)) { 2266 if (O.isELF() && (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC)) 2267 continue; 2268 if (Section.relocation_begin() == Section.relocation_end()) 2269 continue; 2270 Expected<section_iterator> SecOrErr = Section.getRelocatedSection(); 2271 if (!SecOrErr) 2272 reportError(O.getFileName(), 2273 "section (" + Twine(Ndx) + 2274 "): unable to get a relocation target: " + 2275 toString(SecOrErr.takeError())); 2276 SecToRelSec[**SecOrErr].push_back(Section); 2277 } 2278 2279 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) { 2280 StringRef SecName = unwrapOrError(P.first.getName(), O.getFileName()); 2281 outs() << "\nRELOCATION RECORDS FOR [" << SecName << "]:\n"; 2282 uint32_t OffsetPadding = (O.getBytesInAddress() > 4 ? 16 : 8); 2283 uint32_t TypePadding = 24; 2284 outs() << left_justify("OFFSET", OffsetPadding) << " " 2285 << left_justify("TYPE", TypePadding) << " " 2286 << "VALUE\n"; 2287 2288 for (SectionRef Section : P.second) { 2289 for (const RelocationRef &Reloc : Section.relocations()) { 2290 uint64_t Address = Reloc.getOffset(); 2291 SmallString<32> RelocName; 2292 SmallString<32> ValueStr; 2293 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 2294 continue; 2295 Reloc.getTypeName(RelocName); 2296 if (Error E = getRelocationValueString(Reloc, ValueStr)) 2297 reportUniqueWarning(std::move(E)); 2298 2299 outs() << format(Fmt.data(), Address) << " " 2300 << left_justify(RelocName, TypePadding) << " " << ValueStr 2301 << "\n"; 2302 } 2303 } 2304 } 2305 } 2306 2307 // Returns true if we need to show LMA column when dumping section headers. We 2308 // show it only when the platform is ELF and either we have at least one section 2309 // whose VMA and LMA are different and/or when --show-lma flag is used. 2310 static bool shouldDisplayLMA(const ObjectFile &Obj) { 2311 if (!Obj.isELF()) 2312 return false; 2313 for (const SectionRef &S : ToolSectionFilter(Obj)) 2314 if (S.getAddress() != getELFSectionLMA(S)) 2315 return true; 2316 return ShowLMA; 2317 } 2318 2319 static size_t getMaxSectionNameWidth(const ObjectFile &Obj) { 2320 // Default column width for names is 13 even if no names are that long. 2321 size_t MaxWidth = 13; 2322 for (const SectionRef &Section : ToolSectionFilter(Obj)) { 2323 StringRef Name = unwrapOrError(Section.getName(), Obj.getFileName()); 2324 MaxWidth = std::max(MaxWidth, Name.size()); 2325 } 2326 return MaxWidth; 2327 } 2328 2329 void objdump::printSectionHeaders(ObjectFile &Obj) { 2330 if (Obj.isELF() && Obj.sections().empty()) 2331 createFakeELFSections(Obj); 2332 2333 size_t NameWidth = getMaxSectionNameWidth(Obj); 2334 size_t AddressWidth = 2 * Obj.getBytesInAddress(); 2335 bool HasLMAColumn = shouldDisplayLMA(Obj); 2336 outs() << "\nSections:\n"; 2337 if (HasLMAColumn) 2338 outs() << "Idx " << left_justify("Name", NameWidth) << " Size " 2339 << left_justify("VMA", AddressWidth) << " " 2340 << left_justify("LMA", AddressWidth) << " Type\n"; 2341 else 2342 outs() << "Idx " << left_justify("Name", NameWidth) << " Size " 2343 << left_justify("VMA", AddressWidth) << " Type\n"; 2344 2345 uint64_t Idx; 2346 for (const SectionRef &Section : ToolSectionFilter(Obj, &Idx)) { 2347 StringRef Name = unwrapOrError(Section.getName(), Obj.getFileName()); 2348 uint64_t VMA = Section.getAddress(); 2349 if (shouldAdjustVA(Section)) 2350 VMA += AdjustVMA; 2351 2352 uint64_t Size = Section.getSize(); 2353 2354 std::string Type = Section.isText() ? "TEXT" : ""; 2355 if (Section.isData()) 2356 Type += Type.empty() ? "DATA" : ", DATA"; 2357 if (Section.isBSS()) 2358 Type += Type.empty() ? "BSS" : ", BSS"; 2359 if (Section.isDebugSection()) 2360 Type += Type.empty() ? "DEBUG" : ", DEBUG"; 2361 2362 if (HasLMAColumn) 2363 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 2364 Name.str().c_str(), Size) 2365 << format_hex_no_prefix(VMA, AddressWidth) << " " 2366 << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth) 2367 << " " << Type << "\n"; 2368 else 2369 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 2370 Name.str().c_str(), Size) 2371 << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n"; 2372 } 2373 } 2374 2375 void objdump::printSectionContents(const ObjectFile *Obj) { 2376 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj); 2377 2378 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 2379 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2380 uint64_t BaseAddr = Section.getAddress(); 2381 uint64_t Size = Section.getSize(); 2382 if (!Size) 2383 continue; 2384 2385 outs() << "Contents of section "; 2386 StringRef SegmentName = getSegmentName(MachO, Section); 2387 if (!SegmentName.empty()) 2388 outs() << SegmentName << ","; 2389 outs() << Name << ":\n"; 2390 if (Section.isBSS()) { 2391 outs() << format("<skipping contents of bss section at [%04" PRIx64 2392 ", %04" PRIx64 ")>\n", 2393 BaseAddr, BaseAddr + Size); 2394 continue; 2395 } 2396 2397 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName()); 2398 2399 // Dump out the content as hex and printable ascii characters. 2400 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 2401 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 2402 // Dump line of hex. 2403 for (std::size_t I = 0; I < 16; ++I) { 2404 if (I != 0 && I % 4 == 0) 2405 outs() << ' '; 2406 if (Addr + I < End) 2407 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 2408 << hexdigit(Contents[Addr + I] & 0xF, true); 2409 else 2410 outs() << " "; 2411 } 2412 // Print ascii. 2413 outs() << " "; 2414 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 2415 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 2416 outs() << Contents[Addr + I]; 2417 else 2418 outs() << "."; 2419 } 2420 outs() << "\n"; 2421 } 2422 } 2423 } 2424 2425 void Dumper::printSymbolTable(StringRef ArchiveName, StringRef ArchitectureName, 2426 bool DumpDynamic) { 2427 if (O.isCOFF() && !DumpDynamic) { 2428 outs() << "\nSYMBOL TABLE:\n"; 2429 printCOFFSymbolTable(cast<const COFFObjectFile>(O)); 2430 return; 2431 } 2432 2433 const StringRef FileName = O.getFileName(); 2434 2435 if (!DumpDynamic) { 2436 outs() << "\nSYMBOL TABLE:\n"; 2437 for (auto I = O.symbol_begin(); I != O.symbol_end(); ++I) 2438 printSymbol(*I, {}, FileName, ArchiveName, ArchitectureName, DumpDynamic); 2439 return; 2440 } 2441 2442 outs() << "\nDYNAMIC SYMBOL TABLE:\n"; 2443 if (!O.isELF()) { 2444 reportWarning( 2445 "this operation is not currently supported for this file format", 2446 FileName); 2447 return; 2448 } 2449 2450 const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(&O); 2451 auto Symbols = ELF->getDynamicSymbolIterators(); 2452 Expected<std::vector<VersionEntry>> SymbolVersionsOrErr = 2453 ELF->readDynsymVersions(); 2454 if (!SymbolVersionsOrErr) { 2455 reportWarning(toString(SymbolVersionsOrErr.takeError()), FileName); 2456 SymbolVersionsOrErr = std::vector<VersionEntry>(); 2457 (void)!SymbolVersionsOrErr; 2458 } 2459 for (auto &Sym : Symbols) 2460 printSymbol(Sym, *SymbolVersionsOrErr, FileName, ArchiveName, 2461 ArchitectureName, DumpDynamic); 2462 } 2463 2464 void Dumper::printSymbol(const SymbolRef &Symbol, 2465 ArrayRef<VersionEntry> SymbolVersions, 2466 StringRef FileName, StringRef ArchiveName, 2467 StringRef ArchitectureName, bool DumpDynamic) { 2468 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(&O); 2469 Expected<uint64_t> AddrOrErr = Symbol.getAddress(); 2470 if (!AddrOrErr) { 2471 reportUniqueWarning(AddrOrErr.takeError()); 2472 return; 2473 } 2474 uint64_t Address = *AddrOrErr; 2475 if ((Address < StartAddress) || (Address > StopAddress)) 2476 return; 2477 SymbolRef::Type Type = 2478 unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName); 2479 uint32_t Flags = 2480 unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName); 2481 2482 // Don't ask a Mach-O STAB symbol for its section unless you know that 2483 // STAB symbol's section field refers to a valid section index. Otherwise 2484 // the symbol may error trying to load a section that does not exist. 2485 bool IsSTAB = false; 2486 if (MachO) { 2487 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 2488 uint8_t NType = 2489 (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type 2490 : MachO->getSymbolTableEntry(SymDRI).n_type); 2491 if (NType & MachO::N_STAB) 2492 IsSTAB = true; 2493 } 2494 section_iterator Section = IsSTAB 2495 ? O.section_end() 2496 : unwrapOrError(Symbol.getSection(), FileName, 2497 ArchiveName, ArchitectureName); 2498 2499 StringRef Name; 2500 if (Type == SymbolRef::ST_Debug && Section != O.section_end()) { 2501 if (Expected<StringRef> NameOrErr = Section->getName()) 2502 Name = *NameOrErr; 2503 else 2504 consumeError(NameOrErr.takeError()); 2505 2506 } else { 2507 Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName, 2508 ArchitectureName); 2509 } 2510 2511 bool Global = Flags & SymbolRef::SF_Global; 2512 bool Weak = Flags & SymbolRef::SF_Weak; 2513 bool Absolute = Flags & SymbolRef::SF_Absolute; 2514 bool Common = Flags & SymbolRef::SF_Common; 2515 bool Hidden = Flags & SymbolRef::SF_Hidden; 2516 2517 char GlobLoc = ' '; 2518 if ((Section != O.section_end() || Absolute) && !Weak) 2519 GlobLoc = Global ? 'g' : 'l'; 2520 char IFunc = ' '; 2521 if (O.isELF()) { 2522 if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC) 2523 IFunc = 'i'; 2524 if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE) 2525 GlobLoc = 'u'; 2526 } 2527 2528 char Debug = ' '; 2529 if (DumpDynamic) 2530 Debug = 'D'; 2531 else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2532 Debug = 'd'; 2533 2534 char FileFunc = ' '; 2535 if (Type == SymbolRef::ST_File) 2536 FileFunc = 'f'; 2537 else if (Type == SymbolRef::ST_Function) 2538 FileFunc = 'F'; 2539 else if (Type == SymbolRef::ST_Data) 2540 FileFunc = 'O'; 2541 2542 const char *Fmt = O.getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2543 2544 outs() << format(Fmt, Address) << " " 2545 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2546 << (Weak ? 'w' : ' ') // Weak? 2547 << ' ' // Constructor. Not supported yet. 2548 << ' ' // Warning. Not supported yet. 2549 << IFunc // Indirect reference to another symbol. 2550 << Debug // Debugging (d) or dynamic (D) symbol. 2551 << FileFunc // Name of function (F), file (f) or object (O). 2552 << ' '; 2553 if (Absolute) { 2554 outs() << "*ABS*"; 2555 } else if (Common) { 2556 outs() << "*COM*"; 2557 } else if (Section == O.section_end()) { 2558 if (O.isXCOFF()) { 2559 XCOFFSymbolRef XCOFFSym = cast<const XCOFFObjectFile>(O).toSymbolRef( 2560 Symbol.getRawDataRefImpl()); 2561 if (XCOFF::N_DEBUG == XCOFFSym.getSectionNumber()) 2562 outs() << "*DEBUG*"; 2563 else 2564 outs() << "*UND*"; 2565 } else 2566 outs() << "*UND*"; 2567 } else { 2568 StringRef SegmentName = getSegmentName(MachO, *Section); 2569 if (!SegmentName.empty()) 2570 outs() << SegmentName << ","; 2571 StringRef SectionName = unwrapOrError(Section->getName(), FileName); 2572 outs() << SectionName; 2573 if (O.isXCOFF()) { 2574 std::optional<SymbolRef> SymRef = 2575 getXCOFFSymbolContainingSymbolRef(cast<XCOFFObjectFile>(O), Symbol); 2576 if (SymRef) { 2577 2578 Expected<StringRef> NameOrErr = SymRef->getName(); 2579 2580 if (NameOrErr) { 2581 outs() << " (csect:"; 2582 std::string SymName = 2583 Demangle ? demangle(*NameOrErr) : NameOrErr->str(); 2584 2585 if (SymbolDescription) 2586 SymName = getXCOFFSymbolDescription(createSymbolInfo(O, *SymRef), 2587 SymName); 2588 2589 outs() << ' ' << SymName; 2590 outs() << ") "; 2591 } else 2592 reportWarning(toString(NameOrErr.takeError()), FileName); 2593 } 2594 } 2595 } 2596 2597 if (Common) 2598 outs() << '\t' << format(Fmt, static_cast<uint64_t>(Symbol.getAlignment())); 2599 else if (O.isXCOFF()) 2600 outs() << '\t' 2601 << format(Fmt, cast<XCOFFObjectFile>(O).getSymbolSize( 2602 Symbol.getRawDataRefImpl())); 2603 else if (O.isELF()) 2604 outs() << '\t' << format(Fmt, ELFSymbolRef(Symbol).getSize()); 2605 2606 if (O.isELF()) { 2607 if (!SymbolVersions.empty()) { 2608 const VersionEntry &Ver = 2609 SymbolVersions[Symbol.getRawDataRefImpl().d.b - 1]; 2610 std::string Str; 2611 if (!Ver.Name.empty()) 2612 Str = Ver.IsVerDef ? ' ' + Ver.Name : '(' + Ver.Name + ')'; 2613 outs() << ' ' << left_justify(Str, 12); 2614 } 2615 2616 uint8_t Other = ELFSymbolRef(Symbol).getOther(); 2617 switch (Other) { 2618 case ELF::STV_DEFAULT: 2619 break; 2620 case ELF::STV_INTERNAL: 2621 outs() << " .internal"; 2622 break; 2623 case ELF::STV_HIDDEN: 2624 outs() << " .hidden"; 2625 break; 2626 case ELF::STV_PROTECTED: 2627 outs() << " .protected"; 2628 break; 2629 default: 2630 outs() << format(" 0x%02x", Other); 2631 break; 2632 } 2633 } else if (Hidden) { 2634 outs() << " .hidden"; 2635 } 2636 2637 std::string SymName = Demangle ? demangle(Name) : Name.str(); 2638 if (O.isXCOFF() && SymbolDescription) 2639 SymName = getXCOFFSymbolDescription(createSymbolInfo(O, Symbol), SymName); 2640 2641 outs() << ' ' << SymName << '\n'; 2642 } 2643 2644 static void printUnwindInfo(const ObjectFile *O) { 2645 outs() << "Unwind info:\n\n"; 2646 2647 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 2648 printCOFFUnwindInfo(Coff); 2649 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 2650 printMachOUnwindInfo(MachO); 2651 else 2652 // TODO: Extract DWARF dump tool to objdump. 2653 WithColor::error(errs(), ToolName) 2654 << "This operation is only currently supported " 2655 "for COFF and MachO object files.\n"; 2656 } 2657 2658 /// Dump the raw contents of the __clangast section so the output can be piped 2659 /// into llvm-bcanalyzer. 2660 static void printRawClangAST(const ObjectFile *Obj) { 2661 if (outs().is_displayed()) { 2662 WithColor::error(errs(), ToolName) 2663 << "The -raw-clang-ast option will dump the raw binary contents of " 2664 "the clang ast section.\n" 2665 "Please redirect the output to a file or another program such as " 2666 "llvm-bcanalyzer.\n"; 2667 return; 2668 } 2669 2670 StringRef ClangASTSectionName("__clangast"); 2671 if (Obj->isCOFF()) { 2672 ClangASTSectionName = "clangast"; 2673 } 2674 2675 std::optional<object::SectionRef> ClangASTSection; 2676 for (auto Sec : ToolSectionFilter(*Obj)) { 2677 StringRef Name; 2678 if (Expected<StringRef> NameOrErr = Sec.getName()) 2679 Name = *NameOrErr; 2680 else 2681 consumeError(NameOrErr.takeError()); 2682 2683 if (Name == ClangASTSectionName) { 2684 ClangASTSection = Sec; 2685 break; 2686 } 2687 } 2688 if (!ClangASTSection) 2689 return; 2690 2691 StringRef ClangASTContents = 2692 unwrapOrError(ClangASTSection->getContents(), Obj->getFileName()); 2693 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2694 } 2695 2696 static void printFaultMaps(const ObjectFile *Obj) { 2697 StringRef FaultMapSectionName; 2698 2699 if (Obj->isELF()) { 2700 FaultMapSectionName = ".llvm_faultmaps"; 2701 } else if (Obj->isMachO()) { 2702 FaultMapSectionName = "__llvm_faultmaps"; 2703 } else { 2704 WithColor::error(errs(), ToolName) 2705 << "This operation is only currently supported " 2706 "for ELF and Mach-O executable files.\n"; 2707 return; 2708 } 2709 2710 std::optional<object::SectionRef> FaultMapSection; 2711 2712 for (auto Sec : ToolSectionFilter(*Obj)) { 2713 StringRef Name; 2714 if (Expected<StringRef> NameOrErr = Sec.getName()) 2715 Name = *NameOrErr; 2716 else 2717 consumeError(NameOrErr.takeError()); 2718 2719 if (Name == FaultMapSectionName) { 2720 FaultMapSection = Sec; 2721 break; 2722 } 2723 } 2724 2725 outs() << "FaultMap table:\n"; 2726 2727 if (!FaultMapSection) { 2728 outs() << "<not found>\n"; 2729 return; 2730 } 2731 2732 StringRef FaultMapContents = 2733 unwrapOrError(FaultMapSection->getContents(), Obj->getFileName()); 2734 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2735 FaultMapContents.bytes_end()); 2736 2737 outs() << FMP; 2738 } 2739 2740 void Dumper::printPrivateHeaders() { 2741 reportError(O.getFileName(), "Invalid/Unsupported object file format"); 2742 } 2743 2744 static void printFileHeaders(const ObjectFile *O) { 2745 if (!O->isELF() && !O->isCOFF()) 2746 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2747 2748 Triple::ArchType AT = O->getArch(); 2749 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2750 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 2751 2752 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2753 outs() << "start address: " 2754 << "0x" << format(Fmt.data(), Address) << "\n"; 2755 } 2756 2757 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2758 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2759 if (!ModeOrErr) { 2760 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 2761 consumeError(ModeOrErr.takeError()); 2762 return; 2763 } 2764 sys::fs::perms Mode = ModeOrErr.get(); 2765 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2766 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2767 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2768 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2769 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2770 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2771 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2772 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2773 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2774 2775 outs() << " "; 2776 2777 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 2778 unwrapOrError(C.getGID(), Filename), 2779 unwrapOrError(C.getRawSize(), Filename)); 2780 2781 StringRef RawLastModified = C.getRawLastModified(); 2782 unsigned Seconds; 2783 if (RawLastModified.getAsInteger(10, Seconds)) 2784 outs() << "(date: \"" << RawLastModified 2785 << "\" contains non-decimal chars) "; 2786 else { 2787 // Since ctime(3) returns a 26 character string of the form: 2788 // "Sun Sep 16 01:03:52 1973\n\0" 2789 // just print 24 characters. 2790 time_t t = Seconds; 2791 outs() << format("%.24s ", ctime(&t)); 2792 } 2793 2794 StringRef Name = ""; 2795 Expected<StringRef> NameOrErr = C.getName(); 2796 if (!NameOrErr) { 2797 consumeError(NameOrErr.takeError()); 2798 Name = unwrapOrError(C.getRawName(), Filename); 2799 } else { 2800 Name = NameOrErr.get(); 2801 } 2802 outs() << Name << "\n"; 2803 } 2804 2805 // For ELF only now. 2806 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) { 2807 if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) { 2808 if (Elf->getEType() != ELF::ET_REL) 2809 return true; 2810 } 2811 return false; 2812 } 2813 2814 static void checkForInvalidStartStopAddress(ObjectFile *Obj, 2815 uint64_t Start, uint64_t Stop) { 2816 if (!shouldWarnForInvalidStartStopAddress(Obj)) 2817 return; 2818 2819 for (const SectionRef &Section : Obj->sections()) 2820 if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) { 2821 uint64_t BaseAddr = Section.getAddress(); 2822 uint64_t Size = Section.getSize(); 2823 if ((Start < BaseAddr + Size) && Stop > BaseAddr) 2824 return; 2825 } 2826 2827 if (!HasStartAddressFlag) 2828 reportWarning("no section has address less than 0x" + 2829 Twine::utohexstr(Stop) + " specified by --stop-address", 2830 Obj->getFileName()); 2831 else if (!HasStopAddressFlag) 2832 reportWarning("no section has address greater than or equal to 0x" + 2833 Twine::utohexstr(Start) + " specified by --start-address", 2834 Obj->getFileName()); 2835 else 2836 reportWarning("no section overlaps the range [0x" + 2837 Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) + 2838 ") specified by --start-address/--stop-address", 2839 Obj->getFileName()); 2840 } 2841 2842 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 2843 const Archive::Child *C = nullptr) { 2844 Expected<std::unique_ptr<Dumper>> DumperOrErr = createDumper(*O); 2845 if (!DumperOrErr) { 2846 reportError(DumperOrErr.takeError(), O->getFileName(), 2847 A ? A->getFileName() : ""); 2848 return; 2849 } 2850 Dumper &D = **DumperOrErr; 2851 2852 // Avoid other output when using a raw option. 2853 if (!RawClangAST) { 2854 outs() << '\n'; 2855 if (A) 2856 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 2857 else 2858 outs() << O->getFileName(); 2859 outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n"; 2860 } 2861 2862 if (HasStartAddressFlag || HasStopAddressFlag) 2863 checkForInvalidStartStopAddress(O, StartAddress, StopAddress); 2864 2865 // TODO: Change print* free functions to Dumper member functions to utilitize 2866 // stateful functions like reportUniqueWarning. 2867 2868 // Note: the order here matches GNU objdump for compatability. 2869 StringRef ArchiveName = A ? A->getFileName() : ""; 2870 if (ArchiveHeaders && !MachOOpt && C) 2871 printArchiveChild(ArchiveName, *C); 2872 if (FileHeaders) 2873 printFileHeaders(O); 2874 if (PrivateHeaders || FirstPrivateHeader) 2875 D.printPrivateHeaders(); 2876 if (SectionHeaders) 2877 printSectionHeaders(*O); 2878 if (SymbolTable) 2879 D.printSymbolTable(ArchiveName); 2880 if (DynamicSymbolTable) 2881 D.printSymbolTable(ArchiveName, /*ArchitectureName=*/"", 2882 /*DumpDynamic=*/true); 2883 if (DwarfDumpType != DIDT_Null) { 2884 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2885 // Dump the complete DWARF structure. 2886 DIDumpOptions DumpOpts; 2887 DumpOpts.DumpType = DwarfDumpType; 2888 DICtx->dump(outs(), DumpOpts); 2889 } 2890 if (Relocations && !Disassemble) 2891 D.printRelocations(); 2892 if (DynamicRelocations) 2893 D.printDynamicRelocations(); 2894 if (SectionContents) 2895 printSectionContents(O); 2896 if (Disassemble) 2897 disassembleObject(O, Relocations); 2898 if (UnwindInfo) 2899 printUnwindInfo(O); 2900 2901 // Mach-O specific options: 2902 if (ExportsTrie) 2903 printExportsTrie(O); 2904 if (Rebase) 2905 printRebaseTable(O); 2906 if (Bind) 2907 printBindTable(O); 2908 if (LazyBind) 2909 printLazyBindTable(O); 2910 if (WeakBind) 2911 printWeakBindTable(O); 2912 2913 // Other special sections: 2914 if (RawClangAST) 2915 printRawClangAST(O); 2916 if (FaultMapSection) 2917 printFaultMaps(O); 2918 if (Offloading) 2919 dumpOffloadBinary(*O); 2920 } 2921 2922 static void dumpObject(const COFFImportFile *I, const Archive *A, 2923 const Archive::Child *C = nullptr) { 2924 StringRef ArchiveName = A ? A->getFileName() : ""; 2925 2926 // Avoid other output when using a raw option. 2927 if (!RawClangAST) 2928 outs() << '\n' 2929 << ArchiveName << "(" << I->getFileName() << ")" 2930 << ":\tfile format COFF-import-file" 2931 << "\n\n"; 2932 2933 if (ArchiveHeaders && !MachOOpt && C) 2934 printArchiveChild(ArchiveName, *C); 2935 if (SymbolTable) 2936 printCOFFSymbolTable(*I); 2937 } 2938 2939 /// Dump each object file in \a a; 2940 static void dumpArchive(const Archive *A) { 2941 Error Err = Error::success(); 2942 unsigned I = -1; 2943 for (auto &C : A->children(Err)) { 2944 ++I; 2945 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2946 if (!ChildOrErr) { 2947 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2948 reportError(std::move(E), getFileNameForError(C, I), A->getFileName()); 2949 continue; 2950 } 2951 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2952 dumpObject(O, A, &C); 2953 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2954 dumpObject(I, A, &C); 2955 else 2956 reportError(errorCodeToError(object_error::invalid_file_type), 2957 A->getFileName()); 2958 } 2959 if (Err) 2960 reportError(std::move(Err), A->getFileName()); 2961 } 2962 2963 /// Open file and figure out how to dump it. 2964 static void dumpInput(StringRef file) { 2965 // If we are using the Mach-O specific object file parser, then let it parse 2966 // the file and process the command line options. So the -arch flags can 2967 // be used to select specific slices, etc. 2968 if (MachOOpt) { 2969 parseInputMachO(file); 2970 return; 2971 } 2972 2973 // Attempt to open the binary. 2974 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2975 Binary &Binary = *OBinary.getBinary(); 2976 2977 if (Archive *A = dyn_cast<Archive>(&Binary)) 2978 dumpArchive(A); 2979 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2980 dumpObject(O); 2981 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2982 parseInputMachO(UB); 2983 else if (OffloadBinary *OB = dyn_cast<OffloadBinary>(&Binary)) 2984 dumpOffloadSections(*OB); 2985 else 2986 reportError(errorCodeToError(object_error::invalid_file_type), file); 2987 } 2988 2989 template <typename T> 2990 static void parseIntArg(const llvm::opt::InputArgList &InputArgs, int ID, 2991 T &Value) { 2992 if (const opt::Arg *A = InputArgs.getLastArg(ID)) { 2993 StringRef V(A->getValue()); 2994 if (!llvm::to_integer(V, Value, 0)) { 2995 reportCmdLineError(A->getSpelling() + 2996 ": expected a non-negative integer, but got '" + V + 2997 "'"); 2998 } 2999 } 3000 } 3001 3002 static object::BuildID parseBuildIDArg(const opt::Arg *A) { 3003 StringRef V(A->getValue()); 3004 object::BuildID BID = parseBuildID(V); 3005 if (BID.empty()) 3006 reportCmdLineError(A->getSpelling() + ": expected a build ID, but got '" + 3007 V + "'"); 3008 return BID; 3009 } 3010 3011 void objdump::invalidArgValue(const opt::Arg *A) { 3012 reportCmdLineError("'" + StringRef(A->getValue()) + 3013 "' is not a valid value for '" + A->getSpelling() + "'"); 3014 } 3015 3016 static std::vector<std::string> 3017 commaSeparatedValues(const llvm::opt::InputArgList &InputArgs, int ID) { 3018 std::vector<std::string> Values; 3019 for (StringRef Value : InputArgs.getAllArgValues(ID)) { 3020 llvm::SmallVector<StringRef, 2> SplitValues; 3021 llvm::SplitString(Value, SplitValues, ","); 3022 for (StringRef SplitValue : SplitValues) 3023 Values.push_back(SplitValue.str()); 3024 } 3025 return Values; 3026 } 3027 3028 static void parseOtoolOptions(const llvm::opt::InputArgList &InputArgs) { 3029 MachOOpt = true; 3030 FullLeadingAddr = true; 3031 PrintImmHex = true; 3032 3033 ArchName = InputArgs.getLastArgValue(OTOOL_arch).str(); 3034 LinkOptHints = InputArgs.hasArg(OTOOL_C); 3035 if (InputArgs.hasArg(OTOOL_d)) 3036 FilterSections.push_back("__DATA,__data"); 3037 DylibId = InputArgs.hasArg(OTOOL_D); 3038 UniversalHeaders = InputArgs.hasArg(OTOOL_f); 3039 DataInCode = InputArgs.hasArg(OTOOL_G); 3040 FirstPrivateHeader = InputArgs.hasArg(OTOOL_h); 3041 IndirectSymbols = InputArgs.hasArg(OTOOL_I); 3042 ShowRawInsn = InputArgs.hasArg(OTOOL_j); 3043 PrivateHeaders = InputArgs.hasArg(OTOOL_l); 3044 DylibsUsed = InputArgs.hasArg(OTOOL_L); 3045 MCPU = InputArgs.getLastArgValue(OTOOL_mcpu_EQ).str(); 3046 ObjcMetaData = InputArgs.hasArg(OTOOL_o); 3047 DisSymName = InputArgs.getLastArgValue(OTOOL_p).str(); 3048 InfoPlist = InputArgs.hasArg(OTOOL_P); 3049 Relocations = InputArgs.hasArg(OTOOL_r); 3050 if (const Arg *A = InputArgs.getLastArg(OTOOL_s)) { 3051 auto Filter = (A->getValue(0) + StringRef(",") + A->getValue(1)).str(); 3052 FilterSections.push_back(Filter); 3053 } 3054 if (InputArgs.hasArg(OTOOL_t)) 3055 FilterSections.push_back("__TEXT,__text"); 3056 Verbose = InputArgs.hasArg(OTOOL_v) || InputArgs.hasArg(OTOOL_V) || 3057 InputArgs.hasArg(OTOOL_o); 3058 SymbolicOperands = InputArgs.hasArg(OTOOL_V); 3059 if (InputArgs.hasArg(OTOOL_x)) 3060 FilterSections.push_back(",__text"); 3061 LeadingAddr = LeadingHeaders = !InputArgs.hasArg(OTOOL_X); 3062 3063 ChainedFixups = InputArgs.hasArg(OTOOL_chained_fixups); 3064 DyldInfo = InputArgs.hasArg(OTOOL_dyld_info); 3065 3066 InputFilenames = InputArgs.getAllArgValues(OTOOL_INPUT); 3067 if (InputFilenames.empty()) 3068 reportCmdLineError("no input file"); 3069 3070 for (const Arg *A : InputArgs) { 3071 const Option &O = A->getOption(); 3072 if (O.getGroup().isValid() && O.getGroup().getID() == OTOOL_grp_obsolete) { 3073 reportCmdLineWarning(O.getPrefixedName() + 3074 " is obsolete and not implemented"); 3075 } 3076 } 3077 } 3078 3079 static void parseObjdumpOptions(const llvm::opt::InputArgList &InputArgs) { 3080 parseIntArg(InputArgs, OBJDUMP_adjust_vma_EQ, AdjustVMA); 3081 AllHeaders = InputArgs.hasArg(OBJDUMP_all_headers); 3082 ArchName = InputArgs.getLastArgValue(OBJDUMP_arch_name_EQ).str(); 3083 ArchiveHeaders = InputArgs.hasArg(OBJDUMP_archive_headers); 3084 Demangle = InputArgs.hasArg(OBJDUMP_demangle); 3085 Disassemble = InputArgs.hasArg(OBJDUMP_disassemble); 3086 DisassembleAll = InputArgs.hasArg(OBJDUMP_disassemble_all); 3087 SymbolDescription = InputArgs.hasArg(OBJDUMP_symbol_description); 3088 TracebackTable = InputArgs.hasArg(OBJDUMP_traceback_table); 3089 DisassembleSymbols = 3090 commaSeparatedValues(InputArgs, OBJDUMP_disassemble_symbols_EQ); 3091 DisassembleZeroes = InputArgs.hasArg(OBJDUMP_disassemble_zeroes); 3092 if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_dwarf_EQ)) { 3093 DwarfDumpType = StringSwitch<DIDumpType>(A->getValue()) 3094 .Case("frames", DIDT_DebugFrame) 3095 .Default(DIDT_Null); 3096 if (DwarfDumpType == DIDT_Null) 3097 invalidArgValue(A); 3098 } 3099 DynamicRelocations = InputArgs.hasArg(OBJDUMP_dynamic_reloc); 3100 FaultMapSection = InputArgs.hasArg(OBJDUMP_fault_map_section); 3101 Offloading = InputArgs.hasArg(OBJDUMP_offloading); 3102 FileHeaders = InputArgs.hasArg(OBJDUMP_file_headers); 3103 SectionContents = InputArgs.hasArg(OBJDUMP_full_contents); 3104 PrintLines = InputArgs.hasArg(OBJDUMP_line_numbers); 3105 InputFilenames = InputArgs.getAllArgValues(OBJDUMP_INPUT); 3106 MachOOpt = InputArgs.hasArg(OBJDUMP_macho); 3107 MCPU = InputArgs.getLastArgValue(OBJDUMP_mcpu_EQ).str(); 3108 MAttrs = commaSeparatedValues(InputArgs, OBJDUMP_mattr_EQ); 3109 ShowRawInsn = !InputArgs.hasArg(OBJDUMP_no_show_raw_insn); 3110 LeadingAddr = !InputArgs.hasArg(OBJDUMP_no_leading_addr); 3111 RawClangAST = InputArgs.hasArg(OBJDUMP_raw_clang_ast); 3112 Relocations = InputArgs.hasArg(OBJDUMP_reloc); 3113 PrintImmHex = 3114 InputArgs.hasFlag(OBJDUMP_print_imm_hex, OBJDUMP_no_print_imm_hex, true); 3115 PrivateHeaders = InputArgs.hasArg(OBJDUMP_private_headers); 3116 FilterSections = InputArgs.getAllArgValues(OBJDUMP_section_EQ); 3117 SectionHeaders = InputArgs.hasArg(OBJDUMP_section_headers); 3118 ShowAllSymbols = InputArgs.hasArg(OBJDUMP_show_all_symbols); 3119 ShowLMA = InputArgs.hasArg(OBJDUMP_show_lma); 3120 PrintSource = InputArgs.hasArg(OBJDUMP_source); 3121 parseIntArg(InputArgs, OBJDUMP_start_address_EQ, StartAddress); 3122 HasStartAddressFlag = InputArgs.hasArg(OBJDUMP_start_address_EQ); 3123 parseIntArg(InputArgs, OBJDUMP_stop_address_EQ, StopAddress); 3124 HasStopAddressFlag = InputArgs.hasArg(OBJDUMP_stop_address_EQ); 3125 SymbolTable = InputArgs.hasArg(OBJDUMP_syms); 3126 SymbolizeOperands = InputArgs.hasArg(OBJDUMP_symbolize_operands); 3127 DynamicSymbolTable = InputArgs.hasArg(OBJDUMP_dynamic_syms); 3128 TripleName = InputArgs.getLastArgValue(OBJDUMP_triple_EQ).str(); 3129 UnwindInfo = InputArgs.hasArg(OBJDUMP_unwind_info); 3130 Wide = InputArgs.hasArg(OBJDUMP_wide); 3131 Prefix = InputArgs.getLastArgValue(OBJDUMP_prefix).str(); 3132 parseIntArg(InputArgs, OBJDUMP_prefix_strip, PrefixStrip); 3133 if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_debug_vars_EQ)) { 3134 DbgVariables = StringSwitch<DebugVarsFormat>(A->getValue()) 3135 .Case("ascii", DVASCII) 3136 .Case("unicode", DVUnicode) 3137 .Default(DVInvalid); 3138 if (DbgVariables == DVInvalid) 3139 invalidArgValue(A); 3140 } 3141 parseIntArg(InputArgs, OBJDUMP_debug_vars_indent_EQ, DbgIndent); 3142 3143 parseMachOOptions(InputArgs); 3144 3145 // Parse -M (--disassembler-options) and deprecated 3146 // --x86-asm-syntax={att,intel}. 3147 // 3148 // Note, for x86, the asm dialect (AssemblerDialect) is initialized when the 3149 // MCAsmInfo is constructed. MCInstPrinter::applyTargetSpecificCLOption is 3150 // called too late. For now we have to use the internal cl::opt option. 3151 const char *AsmSyntax = nullptr; 3152 for (const auto *A : InputArgs.filtered(OBJDUMP_disassembler_options_EQ, 3153 OBJDUMP_x86_asm_syntax_att, 3154 OBJDUMP_x86_asm_syntax_intel)) { 3155 switch (A->getOption().getID()) { 3156 case OBJDUMP_x86_asm_syntax_att: 3157 AsmSyntax = "--x86-asm-syntax=att"; 3158 continue; 3159 case OBJDUMP_x86_asm_syntax_intel: 3160 AsmSyntax = "--x86-asm-syntax=intel"; 3161 continue; 3162 } 3163 3164 SmallVector<StringRef, 2> Values; 3165 llvm::SplitString(A->getValue(), Values, ","); 3166 for (StringRef V : Values) { 3167 if (V == "att") 3168 AsmSyntax = "--x86-asm-syntax=att"; 3169 else if (V == "intel") 3170 AsmSyntax = "--x86-asm-syntax=intel"; 3171 else 3172 DisassemblerOptions.push_back(V.str()); 3173 } 3174 } 3175 if (AsmSyntax) { 3176 const char *Argv[] = {"llvm-objdump", AsmSyntax}; 3177 llvm::cl::ParseCommandLineOptions(2, Argv); 3178 } 3179 3180 // Look up any provided build IDs, then append them to the input filenames. 3181 for (const opt::Arg *A : InputArgs.filtered(OBJDUMP_build_id)) { 3182 object::BuildID BuildID = parseBuildIDArg(A); 3183 std::optional<std::string> Path = BIDFetcher->fetch(BuildID); 3184 if (!Path) { 3185 reportCmdLineError(A->getSpelling() + ": could not find build ID '" + 3186 A->getValue() + "'"); 3187 } 3188 InputFilenames.push_back(std::move(*Path)); 3189 } 3190 3191 // objdump defaults to a.out if no filenames specified. 3192 if (InputFilenames.empty()) 3193 InputFilenames.push_back("a.out"); 3194 } 3195 3196 int main(int argc, char **argv) { 3197 using namespace llvm; 3198 InitLLVM X(argc, argv); 3199 3200 ToolName = argv[0]; 3201 std::unique_ptr<CommonOptTable> T; 3202 OptSpecifier Unknown, HelpFlag, HelpHiddenFlag, VersionFlag; 3203 3204 StringRef Stem = sys::path::stem(ToolName); 3205 auto Is = [=](StringRef Tool) { 3206 // We need to recognize the following filenames: 3207 // 3208 // llvm-objdump -> objdump 3209 // llvm-otool-10.exe -> otool 3210 // powerpc64-unknown-freebsd13-objdump -> objdump 3211 auto I = Stem.rfind_insensitive(Tool); 3212 return I != StringRef::npos && 3213 (I + Tool.size() == Stem.size() || !isAlnum(Stem[I + Tool.size()])); 3214 }; 3215 if (Is("otool")) { 3216 T = std::make_unique<OtoolOptTable>(); 3217 Unknown = OTOOL_UNKNOWN; 3218 HelpFlag = OTOOL_help; 3219 HelpHiddenFlag = OTOOL_help_hidden; 3220 VersionFlag = OTOOL_version; 3221 } else { 3222 T = std::make_unique<ObjdumpOptTable>(); 3223 Unknown = OBJDUMP_UNKNOWN; 3224 HelpFlag = OBJDUMP_help; 3225 HelpHiddenFlag = OBJDUMP_help_hidden; 3226 VersionFlag = OBJDUMP_version; 3227 } 3228 3229 BumpPtrAllocator A; 3230 StringSaver Saver(A); 3231 opt::InputArgList InputArgs = 3232 T->parseArgs(argc, argv, Unknown, Saver, 3233 [&](StringRef Msg) { reportCmdLineError(Msg); }); 3234 3235 if (InputArgs.size() == 0 || InputArgs.hasArg(HelpFlag)) { 3236 T->printHelp(ToolName); 3237 return 0; 3238 } 3239 if (InputArgs.hasArg(HelpHiddenFlag)) { 3240 T->printHelp(ToolName, /*ShowHidden=*/true); 3241 return 0; 3242 } 3243 3244 // Initialize targets and assembly printers/parsers. 3245 InitializeAllTargetInfos(); 3246 InitializeAllTargetMCs(); 3247 InitializeAllDisassemblers(); 3248 3249 if (InputArgs.hasArg(VersionFlag)) { 3250 cl::PrintVersionMessage(); 3251 if (!Is("otool")) { 3252 outs() << '\n'; 3253 TargetRegistry::printRegisteredTargetsForVersion(outs()); 3254 } 3255 return 0; 3256 } 3257 3258 // Initialize debuginfod. 3259 const bool ShouldUseDebuginfodByDefault = 3260 InputArgs.hasArg(OBJDUMP_build_id) || canUseDebuginfod(); 3261 std::vector<std::string> DebugFileDirectories = 3262 InputArgs.getAllArgValues(OBJDUMP_debug_file_directory); 3263 if (InputArgs.hasFlag(OBJDUMP_debuginfod, OBJDUMP_no_debuginfod, 3264 ShouldUseDebuginfodByDefault)) { 3265 HTTPClient::initialize(); 3266 BIDFetcher = 3267 std::make_unique<DebuginfodFetcher>(std::move(DebugFileDirectories)); 3268 } else { 3269 BIDFetcher = 3270 std::make_unique<BuildIDFetcher>(std::move(DebugFileDirectories)); 3271 } 3272 3273 if (Is("otool")) 3274 parseOtoolOptions(InputArgs); 3275 else 3276 parseObjdumpOptions(InputArgs); 3277 3278 if (StartAddress >= StopAddress) 3279 reportCmdLineError("start address should be less than stop address"); 3280 3281 // Removes trailing separators from prefix. 3282 while (!Prefix.empty() && sys::path::is_separator(Prefix.back())) 3283 Prefix.pop_back(); 3284 3285 if (AllHeaders) 3286 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 3287 SectionHeaders = SymbolTable = true; 3288 3289 if (DisassembleAll || PrintSource || PrintLines || TracebackTable || 3290 !DisassembleSymbols.empty()) 3291 Disassemble = true; 3292 3293 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null && 3294 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST && 3295 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable && 3296 !DynamicSymbolTable && !UnwindInfo && !FaultMapSection && !Offloading && 3297 !(MachOOpt && 3298 (Bind || DataInCode || ChainedFixups || DyldInfo || DylibId || 3299 DylibsUsed || ExportsTrie || FirstPrivateHeader || 3300 FunctionStartsType != FunctionStartsMode::None || IndirectSymbols || 3301 InfoPlist || LazyBind || LinkOptHints || ObjcMetaData || Rebase || 3302 Rpaths || UniversalHeaders || WeakBind || !FilterSections.empty()))) { 3303 T->printHelp(ToolName); 3304 return 2; 3305 } 3306 3307 DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end()); 3308 3309 llvm::for_each(InputFilenames, dumpInput); 3310 3311 warnOnNoMatchForSections(); 3312 3313 return EXIT_SUCCESS; 3314 } 3315