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 "llvm/ADT/Optional.h" 20 #include "llvm/ADT/STLExtras.h" 21 #include "llvm/ADT/StringExtras.h" 22 #include "llvm/ADT/StringSet.h" 23 #include "llvm/ADT/Triple.h" 24 #include "llvm/CodeGen/FaultMaps.h" 25 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 26 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 27 #include "llvm/Demangle/Demangle.h" 28 #include "llvm/MC/MCAsmInfo.h" 29 #include "llvm/MC/MCContext.h" 30 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 31 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 32 #include "llvm/MC/MCInst.h" 33 #include "llvm/MC/MCInstPrinter.h" 34 #include "llvm/MC/MCInstrAnalysis.h" 35 #include "llvm/MC/MCInstrInfo.h" 36 #include "llvm/MC/MCObjectFileInfo.h" 37 #include "llvm/MC/MCRegisterInfo.h" 38 #include "llvm/MC/MCSubtargetInfo.h" 39 #include "llvm/Object/Archive.h" 40 #include "llvm/Object/COFF.h" 41 #include "llvm/Object/COFFImportFile.h" 42 #include "llvm/Object/ELFObjectFile.h" 43 #include "llvm/Object/MachO.h" 44 #include "llvm/Object/MachOUniversal.h" 45 #include "llvm/Object/ObjectFile.h" 46 #include "llvm/Object/Wasm.h" 47 #include "llvm/Support/Casting.h" 48 #include "llvm/Support/CommandLine.h" 49 #include "llvm/Support/Debug.h" 50 #include "llvm/Support/Errc.h" 51 #include "llvm/Support/FileSystem.h" 52 #include "llvm/Support/Format.h" 53 #include "llvm/Support/GraphWriter.h" 54 #include "llvm/Support/Host.h" 55 #include "llvm/Support/InitLLVM.h" 56 #include "llvm/Support/MemoryBuffer.h" 57 #include "llvm/Support/SourceMgr.h" 58 #include "llvm/Support/StringSaver.h" 59 #include "llvm/Support/TargetRegistry.h" 60 #include "llvm/Support/TargetSelect.h" 61 #include "llvm/Support/WithColor.h" 62 #include "llvm/Support/raw_ostream.h" 63 #include <algorithm> 64 #include <cctype> 65 #include <cstring> 66 #include <system_error> 67 #include <unordered_map> 68 #include <utility> 69 70 using namespace llvm; 71 using namespace object; 72 73 cl::opt<unsigned long long> AdjustVMA( 74 "adjust-vma", 75 cl::desc("Increase the displayed address by the specified offset"), 76 cl::value_desc("offset"), cl::init(0)); 77 78 cl::opt<bool> 79 llvm::AllHeaders("all-headers", 80 cl::desc("Display all available header information")); 81 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 82 cl::NotHidden, cl::Grouping, 83 cl::aliasopt(AllHeaders)); 84 85 static cl::list<std::string> 86 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 87 88 cl::opt<bool> 89 llvm::Disassemble("disassemble", 90 cl::desc("Display assembler mnemonics for the machine instructions")); 91 static cl::alias Disassembled("d", cl::desc("Alias for --disassemble"), 92 cl::NotHidden, cl::Grouping, 93 cl::aliasopt(Disassemble)); 94 95 cl::opt<bool> 96 llvm::DisassembleAll("disassemble-all", 97 cl::desc("Display assembler mnemonics for the machine instructions")); 98 static cl::alias DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 99 cl::NotHidden, cl::Grouping, 100 cl::aliasopt(DisassembleAll)); 101 102 cl::opt<bool> llvm::Demangle("demangle", cl::desc("Demangle symbols names"), 103 cl::init(false)); 104 105 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 106 cl::NotHidden, cl::Grouping, 107 cl::aliasopt(llvm::Demangle)); 108 109 static cl::list<std::string> 110 DisassembleFunctions("disassemble-functions", 111 cl::CommaSeparated, 112 cl::desc("List of functions to disassemble")); 113 static StringSet<> DisasmFuncsSet; 114 115 cl::opt<bool> 116 llvm::Relocations("reloc", 117 cl::desc("Display the relocation entries in the file")); 118 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 119 cl::NotHidden, cl::Grouping, 120 cl::aliasopt(llvm::Relocations)); 121 122 cl::opt<bool> 123 llvm::DynamicRelocations("dynamic-reloc", 124 cl::desc("Display the dynamic relocation entries in the file")); 125 static cl::alias DynamicRelocationsd("R", cl::desc("Alias for --dynamic-reloc"), 126 cl::NotHidden, cl::Grouping, 127 cl::aliasopt(DynamicRelocations)); 128 129 cl::opt<bool> 130 llvm::SectionContents("full-contents", 131 cl::desc("Display the content of each section")); 132 static cl::alias SectionContentsShort("s", 133 cl::desc("Alias for --full-contents"), 134 cl::NotHidden, cl::Grouping, 135 cl::aliasopt(SectionContents)); 136 137 cl::opt<bool> llvm::SymbolTable("syms", cl::desc("Display the symbol table")); 138 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 139 cl::NotHidden, cl::Grouping, 140 cl::aliasopt(llvm::SymbolTable)); 141 142 cl::opt<bool> 143 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 144 145 cl::opt<bool> 146 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 147 148 cl::opt<bool> 149 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 150 151 cl::opt<bool> 152 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 153 154 cl::opt<bool> 155 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 156 157 cl::opt<bool> 158 llvm::RawClangAST("raw-clang-ast", 159 cl::desc("Dump the raw binary contents of the clang AST section")); 160 161 static cl::opt<bool> 162 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 163 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 164 cl::Grouping, cl::aliasopt(MachOOpt)); 165 166 cl::opt<std::string> 167 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 168 "see -version for available targets")); 169 170 cl::opt<std::string> 171 llvm::MCPU("mcpu", 172 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 173 cl::value_desc("cpu-name"), 174 cl::init("")); 175 176 cl::opt<std::string> 177 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 178 "see -version for available targets")); 179 180 cl::opt<bool> 181 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 182 "headers for each section.")); 183 static cl::alias SectionHeadersShort("headers", 184 cl::desc("Alias for --section-headers"), 185 cl::NotHidden, 186 cl::aliasopt(SectionHeaders)); 187 static cl::alias SectionHeadersShorter("h", 188 cl::desc("Alias for --section-headers"), 189 cl::NotHidden, cl::Grouping, 190 cl::aliasopt(SectionHeaders)); 191 192 static cl::opt<bool> 193 ShowLMA("show-lma", 194 cl::desc("Display LMA column when dumping ELF section headers")); 195 196 cl::list<std::string> 197 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 198 "With -macho dump segment,section")); 199 cl::alias static FilterSectionsj("j", cl::desc("Alias for --section"), 200 cl::NotHidden, cl::Grouping, cl::Prefix, 201 cl::aliasopt(llvm::FilterSections)); 202 203 cl::list<std::string> 204 llvm::MAttrs("mattr", 205 cl::CommaSeparated, 206 cl::desc("Target specific attributes"), 207 cl::value_desc("a1,+a2,-a3,...")); 208 209 cl::opt<bool> 210 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 211 "instructions, do not print " 212 "the instruction bytes.")); 213 cl::opt<bool> 214 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 215 216 cl::opt<bool> 217 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 218 219 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 220 cl::NotHidden, cl::Grouping, 221 cl::aliasopt(UnwindInfo)); 222 223 cl::opt<bool> 224 llvm::PrivateHeaders("private-headers", 225 cl::desc("Display format specific file headers")); 226 227 cl::opt<bool> 228 llvm::FirstPrivateHeader("private-header", 229 cl::desc("Display only the first format specific file " 230 "header")); 231 232 static cl::alias PrivateHeadersShort("p", 233 cl::desc("Alias for --private-headers"), 234 cl::NotHidden, cl::Grouping, 235 cl::aliasopt(PrivateHeaders)); 236 237 cl::opt<bool> llvm::FileHeaders( 238 "file-headers", 239 cl::desc("Display the contents of the overall file header")); 240 241 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 242 cl::NotHidden, cl::Grouping, 243 cl::aliasopt(FileHeaders)); 244 245 cl::opt<bool> 246 llvm::ArchiveHeaders("archive-headers", 247 cl::desc("Display archive header information")); 248 249 cl::alias ArchiveHeadersShort("a", cl::desc("Alias for --archive-headers"), 250 cl::NotHidden, cl::Grouping, 251 cl::aliasopt(ArchiveHeaders)); 252 253 cl::opt<bool> 254 llvm::PrintImmHex("print-imm-hex", 255 cl::desc("Use hex format for immediate values")); 256 257 cl::opt<bool> PrintFaultMaps("fault-map-section", 258 cl::desc("Display contents of faultmap section")); 259 260 cl::opt<DIDumpType> llvm::DwarfDumpType( 261 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 262 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 263 264 cl::opt<bool> PrintSource( 265 "source", 266 cl::desc( 267 "Display source inlined with disassembly. Implies disassemble object")); 268 269 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), cl::NotHidden, 270 cl::Grouping, cl::aliasopt(PrintSource)); 271 272 cl::opt<bool> PrintLines("line-numbers", 273 cl::desc("Display source line numbers with " 274 "disassembly. Implies disassemble object")); 275 276 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 277 cl::NotHidden, cl::Grouping, 278 cl::aliasopt(PrintLines)); 279 280 cl::opt<unsigned long long> 281 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 282 cl::value_desc("address"), cl::init(0)); 283 cl::opt<unsigned long long> 284 StopAddress("stop-address", 285 cl::desc("Stop disassembly at address"), 286 cl::value_desc("address"), cl::init(UINT64_MAX)); 287 288 cl::opt<bool> DisassembleZeroes( 289 "disassemble-zeroes", 290 cl::desc("Do not skip blocks of zeroes when disassembling")); 291 cl::alias DisassembleZeroesShort("z", 292 cl::desc("Alias for --disassemble-zeroes"), 293 cl::NotHidden, cl::Grouping, 294 cl::aliasopt(DisassembleZeroes)); 295 296 static cl::list<std::string> 297 DisassemblerOptions("disassembler-options", 298 cl::desc("Pass target specific disassembler options"), 299 cl::value_desc("options"), cl::CommaSeparated); 300 static cl::alias 301 DisassemblerOptionsShort("M", cl::desc("Alias for --disassembler-options"), 302 cl::NotHidden, cl::Grouping, cl::Prefix, 303 cl::CommaSeparated, 304 cl::aliasopt(DisassemblerOptions)); 305 306 static StringRef ToolName; 307 308 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 309 310 SectionFilter llvm::ToolSectionFilter(llvm::object::ObjectFile const &O) { 311 return SectionFilter( 312 [](llvm::object::SectionRef const &S) { 313 if (FilterSections.empty()) 314 return true; 315 llvm::StringRef String; 316 std::error_code error = S.getName(String); 317 if (error) 318 return false; 319 return is_contained(FilterSections, String); 320 }, 321 O); 322 } 323 324 void llvm::error(std::error_code EC) { 325 if (!EC) 326 return; 327 WithColor::error(errs(), ToolName) 328 << "reading file: " << EC.message() << ".\n"; 329 errs().flush(); 330 exit(1); 331 } 332 333 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 334 WithColor::error(errs(), ToolName) << Message << ".\n"; 335 errs().flush(); 336 exit(1); 337 } 338 339 void llvm::warn(StringRef Message) { 340 WithColor::warning(errs(), ToolName) << Message << ".\n"; 341 errs().flush(); 342 } 343 344 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 345 Twine Message) { 346 WithColor::error(errs(), ToolName) 347 << "'" << File << "': " << Message << ".\n"; 348 exit(1); 349 } 350 351 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 352 std::error_code EC) { 353 assert(EC); 354 WithColor::error(errs(), ToolName) 355 << "'" << File << "': " << EC.message() << ".\n"; 356 exit(1); 357 } 358 359 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(Error E, StringRef File) { 360 assert(E); 361 std::string Buf; 362 raw_string_ostream OS(Buf); 363 logAllUnhandledErrors(std::move(E), OS); 364 OS.flush(); 365 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 366 exit(1); 367 } 368 369 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(Error E, StringRef ArchiveName, 370 StringRef FileName, 371 StringRef ArchitectureName) { 372 assert(E); 373 WithColor::error(errs(), ToolName); 374 if (ArchiveName != "") 375 errs() << ArchiveName << "(" << FileName << ")"; 376 else 377 errs() << "'" << FileName << "'"; 378 if (!ArchitectureName.empty()) 379 errs() << " (for architecture " << ArchitectureName << ")"; 380 std::string Buf; 381 raw_string_ostream OS(Buf); 382 logAllUnhandledErrors(std::move(E), OS); 383 OS.flush(); 384 errs() << ": " << Buf; 385 exit(1); 386 } 387 388 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(Error E, StringRef ArchiveName, 389 const object::Archive::Child &C, 390 StringRef ArchitectureName) { 391 Expected<StringRef> NameOrErr = C.getName(); 392 // TODO: if we have a error getting the name then it would be nice to print 393 // the index of which archive member this is and or its offset in the 394 // archive instead of "???" as the name. 395 if (!NameOrErr) { 396 consumeError(NameOrErr.takeError()); 397 llvm::report_error(std::move(E), ArchiveName, "???", ArchitectureName); 398 } else 399 llvm::report_error(std::move(E), ArchiveName, NameOrErr.get(), 400 ArchitectureName); 401 } 402 403 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 404 // Figure out the target triple. 405 llvm::Triple TheTriple("unknown-unknown-unknown"); 406 if (TripleName.empty()) { 407 if (Obj) 408 TheTriple = Obj->makeTriple(); 409 } else { 410 TheTriple.setTriple(Triple::normalize(TripleName)); 411 412 // Use the triple, but also try to combine with ARM build attributes. 413 if (Obj) { 414 auto Arch = Obj->getArch(); 415 if (Arch == Triple::arm || Arch == Triple::armeb) 416 Obj->setARMSubArch(TheTriple); 417 } 418 } 419 420 // Get the target specific parser. 421 std::string Error; 422 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 423 Error); 424 if (!TheTarget) { 425 if (Obj) 426 report_error(Obj->getFileName(), "can't find target: " + Error); 427 else 428 error("can't find target: " + Error); 429 } 430 431 // Update the triple name and return the found target. 432 TripleName = TheTriple.getTriple(); 433 return TheTarget; 434 } 435 436 bool llvm::isRelocAddressLess(RelocationRef A, RelocationRef B) { 437 return A.getOffset() < B.getOffset(); 438 } 439 440 static std::error_code getRelocationValueString(const RelocationRef &Rel, 441 SmallVectorImpl<char> &Result) { 442 const ObjectFile *Obj = Rel.getObject(); 443 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 444 return getELFRelocationValueString(ELF, Rel, Result); 445 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 446 return getCOFFRelocationValueString(COFF, Rel, Result); 447 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 448 return getWasmRelocationValueString(Wasm, Rel, Result); 449 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 450 return getMachORelocationValueString(MachO, Rel, Result); 451 llvm_unreachable("unknown object file format"); 452 } 453 454 /// Indicates whether this relocation should hidden when listing 455 /// relocations, usually because it is the trailing part of a multipart 456 /// relocation that will be printed as part of the leading relocation. 457 static bool getHidden(RelocationRef RelRef) { 458 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 459 if (!MachO) 460 return false; 461 462 unsigned Arch = MachO->getArch(); 463 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 464 uint64_t Type = MachO->getRelocationType(Rel); 465 466 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 467 // is always hidden. 468 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 469 return Type == MachO::GENERIC_RELOC_PAIR; 470 471 if (Arch == Triple::x86_64) { 472 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 473 // an X86_64_RELOC_SUBTRACTOR. 474 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 475 DataRefImpl RelPrev = Rel; 476 RelPrev.d.a--; 477 uint64_t PrevType = MachO->getRelocationType(RelPrev); 478 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 479 return true; 480 } 481 } 482 483 return false; 484 } 485 486 namespace { 487 class SourcePrinter { 488 protected: 489 DILineInfo OldLineInfo; 490 const ObjectFile *Obj = nullptr; 491 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 492 // File name to file contents of source 493 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 494 // Mark the line endings of the cached source 495 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 496 497 private: 498 bool cacheSource(const DILineInfo& LineInfoFile); 499 500 public: 501 SourcePrinter() = default; 502 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 503 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 504 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 505 DefaultArch); 506 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 507 } 508 virtual ~SourcePrinter() = default; 509 virtual void printSourceLine(raw_ostream &OS, 510 object::SectionedAddress Address, 511 StringRef Delimiter = "; "); 512 }; 513 514 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 515 std::unique_ptr<MemoryBuffer> Buffer; 516 if (LineInfo.Source) { 517 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 518 } else { 519 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 520 if (!BufferOrError) 521 return false; 522 Buffer = std::move(*BufferOrError); 523 } 524 // Chomp the file to get lines 525 const char *BufferStart = Buffer->getBufferStart(), 526 *BufferEnd = Buffer->getBufferEnd(); 527 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 528 const char *Start = BufferStart; 529 for (const char *I = BufferStart; I != BufferEnd; ++I) 530 if (*I == '\n') { 531 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 532 Start = I + 1; 533 } 534 if (Start < BufferEnd) 535 Lines.emplace_back(Start, BufferEnd - Start); 536 SourceCache[LineInfo.FileName] = std::move(Buffer); 537 return true; 538 } 539 540 void SourcePrinter::printSourceLine(raw_ostream &OS, 541 object::SectionedAddress Address, 542 StringRef Delimiter) { 543 if (!Symbolizer) 544 return; 545 DILineInfo LineInfo = DILineInfo(); 546 auto ExpectedLineInfo = 547 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 548 if (!ExpectedLineInfo) 549 consumeError(ExpectedLineInfo.takeError()); 550 else 551 LineInfo = *ExpectedLineInfo; 552 553 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 554 LineInfo.Line == 0) 555 return; 556 557 if (PrintLines) 558 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 559 if (PrintSource) { 560 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 561 if (!cacheSource(LineInfo)) 562 return; 563 auto LineBuffer = LineCache.find(LineInfo.FileName); 564 if (LineBuffer != LineCache.end()) { 565 if (LineInfo.Line > LineBuffer->second.size()) 566 return; 567 // Vector begins at 0, line numbers are non-zero 568 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1] << '\n'; 569 } 570 } 571 OldLineInfo = LineInfo; 572 } 573 574 static bool isArmElf(const ObjectFile *Obj) { 575 return (Obj->isELF() && 576 (Obj->getArch() == Triple::aarch64 || 577 Obj->getArch() == Triple::aarch64_be || 578 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 579 Obj->getArch() == Triple::thumb || 580 Obj->getArch() == Triple::thumbeb)); 581 } 582 583 static void printRelocation(const RelocationRef &Rel, uint64_t Address, 584 uint8_t AddrSize) { 585 StringRef Fmt = 586 AddrSize > 4 ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 587 SmallString<16> Name; 588 SmallString<32> Val; 589 Rel.getTypeName(Name); 590 error(getRelocationValueString(Rel, Val)); 591 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n"; 592 } 593 594 class PrettyPrinter { 595 public: 596 virtual ~PrettyPrinter() = default; 597 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 598 ArrayRef<uint8_t> Bytes, 599 object::SectionedAddress Address, raw_ostream &OS, 600 StringRef Annot, MCSubtargetInfo const &STI, 601 SourcePrinter *SP, 602 std::vector<RelocationRef> *Rels = nullptr) { 603 if (SP && (PrintSource || PrintLines)) 604 SP->printSourceLine(OS, Address); 605 if (!NoLeadingAddr) 606 OS << format("%8" PRIx64 ":", Address.Address); 607 if (!NoShowRawInsn) { 608 OS << "\t"; 609 dumpBytes(Bytes, OS); 610 } 611 if (MI) 612 IP.printInst(MI, OS, "", STI); 613 else 614 OS << " <unknown>"; 615 } 616 }; 617 PrettyPrinter PrettyPrinterInst; 618 class HexagonPrettyPrinter : public PrettyPrinter { 619 public: 620 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 621 raw_ostream &OS) { 622 uint32_t opcode = 623 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 624 if (!NoLeadingAddr) 625 OS << format("%8" PRIx64 ":", Address); 626 if (!NoShowRawInsn) { 627 OS << "\t"; 628 dumpBytes(Bytes.slice(0, 4), OS); 629 OS << format("%08" PRIx32, opcode); 630 } 631 } 632 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 633 object::SectionedAddress Address, raw_ostream &OS, 634 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 635 std::vector<RelocationRef> *Rels) override { 636 if (SP && (PrintSource || PrintLines)) 637 SP->printSourceLine(OS, Address, ""); 638 if (!MI) { 639 printLead(Bytes, Address.Address, OS); 640 OS << " <unknown>"; 641 return; 642 } 643 std::string Buffer; 644 { 645 raw_string_ostream TempStream(Buffer); 646 IP.printInst(MI, TempStream, "", STI); 647 } 648 StringRef Contents(Buffer); 649 // Split off bundle attributes 650 auto PacketBundle = Contents.rsplit('\n'); 651 // Split off first instruction from the rest 652 auto HeadTail = PacketBundle.first.split('\n'); 653 auto Preamble = " { "; 654 auto Separator = ""; 655 656 // Hexagon's packets require relocations to be inline rather than 657 // clustered at the end of the packet. 658 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 659 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 660 auto PrintReloc = [&]() -> void { 661 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 662 if (RelCur->getOffset() == Address.Address) { 663 printRelocation(*RelCur, Address.Address, 4); 664 return; 665 } 666 ++RelCur; 667 } 668 }; 669 670 while (!HeadTail.first.empty()) { 671 OS << Separator; 672 Separator = "\n"; 673 if (SP && (PrintSource || PrintLines)) 674 SP->printSourceLine(OS, Address, ""); 675 printLead(Bytes, Address.Address, OS); 676 OS << Preamble; 677 Preamble = " "; 678 StringRef Inst; 679 auto Duplex = HeadTail.first.split('\v'); 680 if (!Duplex.second.empty()) { 681 OS << Duplex.first; 682 OS << "; "; 683 Inst = Duplex.second; 684 } 685 else 686 Inst = HeadTail.first; 687 OS << Inst; 688 HeadTail = HeadTail.second.split('\n'); 689 if (HeadTail.first.empty()) 690 OS << " } " << PacketBundle.second; 691 PrintReloc(); 692 Bytes = Bytes.slice(4); 693 Address.Address += 4; 694 } 695 } 696 }; 697 HexagonPrettyPrinter HexagonPrettyPrinterInst; 698 699 class AMDGCNPrettyPrinter : public PrettyPrinter { 700 public: 701 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 702 object::SectionedAddress Address, raw_ostream &OS, 703 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 704 std::vector<RelocationRef> *Rels) override { 705 if (SP && (PrintSource || PrintLines)) 706 SP->printSourceLine(OS, Address); 707 708 typedef support::ulittle32_t U32; 709 710 if (MI) { 711 SmallString<40> InstStr; 712 raw_svector_ostream IS(InstStr); 713 714 IP.printInst(MI, IS, "", STI); 715 716 OS << left_justify(IS.str(), 60); 717 } else { 718 // an unrecognized encoding - this is probably data so represent it 719 // using the .long directive, or .byte directive if fewer than 4 bytes 720 // remaining 721 if (Bytes.size() >= 4) { 722 OS << format("\t.long 0x%08" PRIx32 " ", 723 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 724 OS.indent(42); 725 } else { 726 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 727 for (unsigned int i = 1; i < Bytes.size(); i++) 728 OS << format(", 0x%02" PRIx8, Bytes[i]); 729 OS.indent(55 - (6 * Bytes.size())); 730 } 731 } 732 733 OS << format("// %012" PRIX64 ": ", Address.Address); 734 if (Bytes.size() >=4) { 735 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 736 Bytes.size() / sizeof(U32))) 737 // D should be explicitly casted to uint32_t here as it is passed 738 // by format to snprintf as vararg. 739 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 740 } else { 741 for (unsigned int i = 0; i < Bytes.size(); i++) 742 OS << format("%02" PRIX8 " ", Bytes[i]); 743 } 744 745 if (!Annot.empty()) 746 OS << "// " << Annot; 747 } 748 }; 749 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 750 751 class BPFPrettyPrinter : public PrettyPrinter { 752 public: 753 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 754 object::SectionedAddress Address, raw_ostream &OS, 755 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 756 std::vector<RelocationRef> *Rels) override { 757 if (SP && (PrintSource || PrintLines)) 758 SP->printSourceLine(OS, Address); 759 if (!NoLeadingAddr) 760 OS << format("%8" PRId64 ":", Address.Address / 8); 761 if (!NoShowRawInsn) { 762 OS << "\t"; 763 dumpBytes(Bytes, OS); 764 } 765 if (MI) 766 IP.printInst(MI, OS, "", STI); 767 else 768 OS << " <unknown>"; 769 } 770 }; 771 BPFPrettyPrinter BPFPrettyPrinterInst; 772 773 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 774 switch(Triple.getArch()) { 775 default: 776 return PrettyPrinterInst; 777 case Triple::hexagon: 778 return HexagonPrettyPrinterInst; 779 case Triple::amdgcn: 780 return AMDGCNPrettyPrinterInst; 781 case Triple::bpfel: 782 case Triple::bpfeb: 783 return BPFPrettyPrinterInst; 784 } 785 } 786 } 787 788 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 789 assert(Obj->isELF()); 790 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 791 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 792 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 793 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 794 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 795 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 796 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 797 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 798 llvm_unreachable("Unsupported binary format"); 799 } 800 801 template <class ELFT> static void 802 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 803 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 804 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 805 uint8_t SymbolType = Symbol.getELFType(); 806 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 807 continue; 808 809 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 810 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 811 if (Name.empty()) 812 continue; 813 814 section_iterator SecI = 815 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 816 if (SecI == Obj->section_end()) 817 continue; 818 819 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 820 } 821 } 822 823 static void 824 addDynamicElfSymbols(const ObjectFile *Obj, 825 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 826 assert(Obj->isELF()); 827 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 828 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 829 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 830 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 831 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 832 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 833 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 834 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 835 else 836 llvm_unreachable("Unsupported binary format"); 837 } 838 839 static void addPltEntries(const ObjectFile *Obj, 840 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 841 StringSaver &Saver) { 842 Optional<SectionRef> Plt = None; 843 for (const SectionRef &Section : Obj->sections()) { 844 StringRef Name; 845 if (Section.getName(Name)) 846 continue; 847 if (Name == ".plt") 848 Plt = Section; 849 } 850 if (!Plt) 851 return; 852 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 853 for (auto PltEntry : ElfObj->getPltAddresses()) { 854 SymbolRef Symbol(PltEntry.first, ElfObj); 855 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 856 857 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 858 if (!Name.empty()) 859 AllSymbols[*Plt].emplace_back( 860 PltEntry.second, Saver.save((Name + "@plt").str()), SymbolType); 861 } 862 } 863 } 864 865 // Normally the disassembly output will skip blocks of zeroes. This function 866 // returns the number of zero bytes that can be skipped when dumping the 867 // disassembly of the instructions in Buf. 868 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 869 // Find the number of leading zeroes. 870 size_t N = 0; 871 while (N < Buf.size() && !Buf[N]) 872 ++N; 873 874 // We may want to skip blocks of zero bytes, but unless we see 875 // at least 8 of them in a row. 876 if (N < 8) 877 return 0; 878 879 // We skip zeroes in multiples of 4 because do not want to truncate an 880 // instruction if it starts with a zero byte. 881 return N & ~0x3; 882 } 883 884 // Returns a map from sections to their relocations. 885 static std::map<SectionRef, std::vector<RelocationRef>> 886 getRelocsMap(llvm::object::ObjectFile const &Obj) { 887 std::map<SectionRef, std::vector<RelocationRef>> Ret; 888 for (const SectionRef &Section : ToolSectionFilter(Obj)) { 889 section_iterator RelSec = Section.getRelocatedSection(); 890 if (RelSec == Obj.section_end()) 891 continue; 892 std::vector<RelocationRef> &V = Ret[*RelSec]; 893 for (const RelocationRef &R : Section.relocations()) 894 V.push_back(R); 895 // Sort relocations by address. 896 llvm::sort(V, isRelocAddressLess); 897 } 898 return Ret; 899 } 900 901 // Used for --adjust-vma to check if address should be adjusted by the 902 // specified value for a given section. 903 // For ELF we do not adjust non-allocatable sections like debug ones, 904 // because they are not loadable. 905 // TODO: implement for other file formats. 906 static bool shouldAdjustVA(const SectionRef &Section) { 907 const ObjectFile *Obj = Section.getObject(); 908 if (isa<object::ELFObjectFileBase>(Obj)) 909 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 910 return false; 911 } 912 913 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 914 MCContext &Ctx, MCDisassembler *DisAsm, 915 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 916 const MCSubtargetInfo *STI, PrettyPrinter &PIP, 917 SourcePrinter &SP, bool InlineRelocs) { 918 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 919 if (InlineRelocs) 920 RelocMap = getRelocsMap(*Obj); 921 922 // Create a mapping from virtual address to symbol name. This is used to 923 // pretty print the symbols while disassembling. 924 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 925 SectionSymbolsTy AbsoluteSymbols; 926 const StringRef FileName = Obj->getFileName(); 927 for (const SymbolRef &Symbol : Obj->symbols()) { 928 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName); 929 930 StringRef Name = unwrapOrError(Symbol.getName(), FileName); 931 if (Name.empty()) 932 continue; 933 934 uint8_t SymbolType = ELF::STT_NOTYPE; 935 if (Obj->isELF()) { 936 SymbolType = getElfSymbolType(Obj, Symbol); 937 if (SymbolType == ELF::STT_SECTION) 938 continue; 939 } 940 941 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 942 if (SecI != Obj->section_end()) 943 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 944 else 945 AbsoluteSymbols.emplace_back(Address, Name, SymbolType); 946 } 947 if (AllSymbols.empty() && Obj->isELF()) 948 addDynamicElfSymbols(Obj, AllSymbols); 949 950 BumpPtrAllocator A; 951 StringSaver Saver(A); 952 addPltEntries(Obj, AllSymbols, Saver); 953 954 // Create a mapping from virtual address to section. 955 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 956 for (SectionRef Sec : Obj->sections()) 957 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 958 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 959 960 // Linked executables (.exe and .dll files) typically don't include a real 961 // symbol table but they might contain an export table. 962 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 963 for (const auto &ExportEntry : COFFObj->export_directories()) { 964 StringRef Name; 965 error(ExportEntry.getSymbolName(Name)); 966 if (Name.empty()) 967 continue; 968 uint32_t RVA; 969 error(ExportEntry.getExportRVA(RVA)); 970 971 uint64_t VA = COFFObj->getImageBase() + RVA; 972 auto Sec = llvm::upper_bound( 973 SectionAddresses, VA, 974 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 975 return LHS < RHS.first; 976 }); 977 if (Sec != SectionAddresses.begin()) { 978 --Sec; 979 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 980 } else 981 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 982 } 983 } 984 985 // Sort all the symbols, this allows us to use a simple binary search to find 986 // a symbol near an address. 987 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 988 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 989 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 990 991 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 992 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 993 continue; 994 995 uint64_t SectionAddr = Section.getAddress(); 996 uint64_t SectSize = Section.getSize(); 997 if (!SectSize) 998 continue; 999 1000 // Get the list of all the symbols in this section. 1001 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1002 std::vector<uint64_t> DataMappingSymsAddr; 1003 std::vector<uint64_t> TextMappingSymsAddr; 1004 if (isArmElf(Obj)) { 1005 for (const auto &Symb : Symbols) { 1006 uint64_t Address = std::get<0>(Symb); 1007 StringRef Name = std::get<1>(Symb); 1008 if (Name.startswith("$d")) 1009 DataMappingSymsAddr.push_back(Address - SectionAddr); 1010 if (Name.startswith("$x")) 1011 TextMappingSymsAddr.push_back(Address - SectionAddr); 1012 if (Name.startswith("$a")) 1013 TextMappingSymsAddr.push_back(Address - SectionAddr); 1014 if (Name.startswith("$t")) 1015 TextMappingSymsAddr.push_back(Address - SectionAddr); 1016 } 1017 } 1018 1019 llvm::sort(DataMappingSymsAddr); 1020 llvm::sort(TextMappingSymsAddr); 1021 1022 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1023 // AMDGPU disassembler uses symbolizer for printing labels 1024 std::unique_ptr<MCRelocationInfo> RelInfo( 1025 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1026 if (RelInfo) { 1027 std::unique_ptr<MCSymbolizer> Symbolizer( 1028 TheTarget->createMCSymbolizer( 1029 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1030 DisAsm->setSymbolizer(std::move(Symbolizer)); 1031 } 1032 } 1033 1034 StringRef SegmentName = ""; 1035 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1036 DataRefImpl DR = Section.getRawDataRefImpl(); 1037 SegmentName = MachO->getSectionFinalSegmentName(DR); 1038 } 1039 StringRef SectionName; 1040 error(Section.getName(SectionName)); 1041 1042 // If the section has no symbol at the start, just insert a dummy one. 1043 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1044 Symbols.insert( 1045 Symbols.begin(), 1046 std::make_tuple(SectionAddr, SectionName, 1047 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1048 } 1049 1050 SmallString<40> Comments; 1051 raw_svector_ostream CommentStream(Comments); 1052 1053 StringRef BytesStr; 1054 error(Section.getContents(BytesStr)); 1055 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef(BytesStr); 1056 1057 uint64_t VMAAdjustment = 0; 1058 if (shouldAdjustVA(Section)) 1059 VMAAdjustment = AdjustVMA; 1060 1061 uint64_t Size; 1062 uint64_t Index; 1063 bool PrintedSection = false; 1064 std::vector<RelocationRef> Rels = RelocMap[Section]; 1065 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1066 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1067 // Disassemble symbol by symbol. 1068 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1069 uint64_t Start = std::get<0>(Symbols[SI]) - SectionAddr; 1070 // The end is either the section end or the beginning of the next 1071 // symbol. 1072 uint64_t End = (SI == SE - 1) 1073 ? SectSize 1074 : std::get<0>(Symbols[SI + 1]) - SectionAddr; 1075 // Don't try to disassemble beyond the end of section contents. 1076 if (End > SectSize) 1077 End = SectSize; 1078 // If this symbol has the same address as the next symbol, then skip it. 1079 if (Start >= End) 1080 continue; 1081 1082 // Check if we need to skip symbol 1083 // Skip if the symbol's data is not between StartAddress and StopAddress 1084 if (End + SectionAddr < StartAddress || 1085 Start + SectionAddr > StopAddress) { 1086 continue; 1087 } 1088 1089 /// Skip if user requested specific symbols and this is not in the list 1090 if (!DisasmFuncsSet.empty() && 1091 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1092 continue; 1093 1094 if (!PrintedSection) { 1095 PrintedSection = true; 1096 outs() << "Disassembly of section "; 1097 if (!SegmentName.empty()) 1098 outs() << SegmentName << ","; 1099 outs() << SectionName << ':'; 1100 } 1101 1102 // Stop disassembly at the stop address specified 1103 if (End + SectionAddr > StopAddress) 1104 End = StopAddress - SectionAddr; 1105 1106 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1107 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1108 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1109 Start += 256; 1110 } 1111 if (SI == SE - 1 || 1112 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1113 // cut trailing zeroes at the end of kernel 1114 // cut up to 256 bytes 1115 const uint64_t EndAlign = 256; 1116 const auto Limit = End - (std::min)(EndAlign, End - Start); 1117 while (End > Limit && 1118 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1119 End -= 4; 1120 } 1121 } 1122 1123 outs() << '\n'; 1124 if (!NoLeadingAddr) 1125 outs() << format("%016" PRIx64 " ", 1126 SectionAddr + Start + VMAAdjustment); 1127 1128 StringRef SymbolName = std::get<1>(Symbols[SI]); 1129 if (Demangle) 1130 outs() << demangle(SymbolName) << ":\n"; 1131 else 1132 outs() << SymbolName << ":\n"; 1133 1134 // Don't print raw contents of a virtual section. A virtual section 1135 // doesn't have any contents in the file. 1136 if (Section.isVirtual()) { 1137 outs() << "...\n"; 1138 continue; 1139 } 1140 1141 #ifndef NDEBUG 1142 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1143 #else 1144 raw_ostream &DebugOut = nulls(); 1145 #endif 1146 1147 // Some targets (like WebAssembly) have a special prelude at the start 1148 // of each symbol. 1149 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1150 SectionAddr + Start, DebugOut, CommentStream); 1151 Start += Size; 1152 1153 for (Index = Start; Index < End; Index += Size) { 1154 MCInst Inst; 1155 1156 if (Index + SectionAddr < StartAddress || 1157 Index + SectionAddr > StopAddress) { 1158 // skip byte by byte till StartAddress is reached 1159 Size = 1; 1160 continue; 1161 } 1162 // AArch64 ELF binaries can interleave data and text in the 1163 // same section. We rely on the markers introduced to 1164 // understand what we need to dump. If the data marker is within a 1165 // function, it is denoted as a word/short etc 1166 if (isArmElf(Obj) && std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1167 !DisassembleAll && 1168 std::binary_search(DataMappingSymsAddr.begin(), 1169 DataMappingSymsAddr.end(), Index)) { 1170 // Switch to data. 1171 support::endianness Endian = 1172 Obj->isLittleEndian() ? support::little : support::big; 1173 while (Index < End) { 1174 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1175 outs() << "\t"; 1176 if (Index + 4 <= End) { 1177 dumpBytes(Bytes.slice(Index, 4), outs()); 1178 outs() << "\t.word\t" 1179 << format_hex(support::endian::read32(Bytes.data() + Index, 1180 Endian), 1181 10); 1182 Index += 4; 1183 } else if (Index + 2 <= End) { 1184 dumpBytes(Bytes.slice(Index, 2), outs()); 1185 outs() << "\t\t.short\t" 1186 << format_hex(support::endian::read16(Bytes.data() + Index, 1187 Endian), 1188 6); 1189 Index += 2; 1190 } else { 1191 dumpBytes(Bytes.slice(Index, 1), outs()); 1192 outs() << "\t\t.byte\t" << format_hex(Bytes[0], 4); 1193 ++Index; 1194 } 1195 outs() << "\n"; 1196 if (std::binary_search(TextMappingSymsAddr.begin(), 1197 TextMappingSymsAddr.end(), Index)) 1198 break; 1199 } 1200 } 1201 1202 // If there is a data symbol inside an ELF text section and we are only 1203 // disassembling text (applicable all architectures), 1204 // we are in a situation where we must print the data and not 1205 // disassemble it. 1206 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1207 !DisassembleAll && Section.isText()) { 1208 // print out data up to 8 bytes at a time in hex and ascii 1209 uint8_t AsciiData[9] = {'\0'}; 1210 uint8_t Byte; 1211 int NumBytes = 0; 1212 1213 for (Index = Start; Index < End; Index += 1) { 1214 if (((SectionAddr + Index) < StartAddress) || 1215 ((SectionAddr + Index) > StopAddress)) 1216 continue; 1217 if (NumBytes == 0) { 1218 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1219 outs() << "\t"; 1220 } 1221 Byte = Bytes.slice(Index)[0]; 1222 outs() << format(" %02x", Byte); 1223 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1224 1225 uint8_t IndentOffset = 0; 1226 NumBytes++; 1227 if (Index == End - 1 || NumBytes > 8) { 1228 // Indent the space for less than 8 bytes data. 1229 // 2 spaces for byte and one for space between bytes 1230 IndentOffset = 3 * (8 - NumBytes); 1231 for (int Excess = NumBytes; Excess < 8; Excess++) 1232 AsciiData[Excess] = '\0'; 1233 NumBytes = 8; 1234 } 1235 if (NumBytes == 8) { 1236 AsciiData[8] = '\0'; 1237 outs() << std::string(IndentOffset, ' ') << " "; 1238 outs() << reinterpret_cast<char *>(AsciiData); 1239 outs() << '\n'; 1240 NumBytes = 0; 1241 } 1242 } 1243 } 1244 if (Index >= End) 1245 break; 1246 1247 // When -z or --disassemble-zeroes are given we always dissasemble them. 1248 // Otherwise we might want to skip zero bytes we see. 1249 if (!DisassembleZeroes) { 1250 uint64_t MaxOffset = End - Index; 1251 // For -reloc: print zero blocks patched by relocations, so that 1252 // relocations can be shown in the dump. 1253 if (RelCur != RelEnd) 1254 MaxOffset = RelCur->getOffset() - Index; 1255 1256 if (size_t N = 1257 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1258 outs() << "\t\t..." << '\n'; 1259 Index += N; 1260 if (Index >= End) 1261 break; 1262 } 1263 } 1264 1265 // Disassemble a real instruction or a data when disassemble all is 1266 // provided 1267 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1268 SectionAddr + Index, DebugOut, 1269 CommentStream); 1270 if (Size == 0) 1271 Size = 1; 1272 1273 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1274 Bytes.slice(Index, Size), 1275 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, 1276 outs(), "", *STI, &SP, &Rels); 1277 outs() << CommentStream.str(); 1278 Comments.clear(); 1279 1280 // Try to resolve the target of a call, tail call, etc. to a specific 1281 // symbol. 1282 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1283 MIA->isConditionalBranch(Inst))) { 1284 uint64_t Target; 1285 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1286 // In a relocatable object, the target's section must reside in 1287 // the same section as the call instruction or it is accessed 1288 // through a relocation. 1289 // 1290 // In a non-relocatable object, the target may be in any section. 1291 // 1292 // N.B. We don't walk the relocations in the relocatable case yet. 1293 auto *TargetSectionSymbols = &Symbols; 1294 if (!Obj->isRelocatableObject()) { 1295 auto SectionAddress = llvm::upper_bound( 1296 SectionAddresses, Target, 1297 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1298 return LHS < RHS.first; 1299 }); 1300 if (SectionAddress != SectionAddresses.begin()) { 1301 --SectionAddress; 1302 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1303 } else { 1304 TargetSectionSymbols = &AbsoluteSymbols; 1305 } 1306 } 1307 1308 // Find the first symbol in the section whose offset is less than 1309 // or equal to the target. If there isn't a section that contains 1310 // the target, find the nearest preceding absolute symbol. 1311 auto TargetSym = llvm::upper_bound( 1312 *TargetSectionSymbols, Target, 1313 [](uint64_t LHS, 1314 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1315 return LHS < std::get<0>(RHS); 1316 }); 1317 if (TargetSym == TargetSectionSymbols->begin()) { 1318 TargetSectionSymbols = &AbsoluteSymbols; 1319 TargetSym = llvm::upper_bound( 1320 AbsoluteSymbols, Target, 1321 [](uint64_t LHS, 1322 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1323 return LHS < std::get<0>(RHS); 1324 }); 1325 } 1326 if (TargetSym != TargetSectionSymbols->begin()) { 1327 --TargetSym; 1328 uint64_t TargetAddress = std::get<0>(*TargetSym); 1329 StringRef TargetName = std::get<1>(*TargetSym); 1330 outs() << " <" << TargetName; 1331 uint64_t Disp = Target - TargetAddress; 1332 if (Disp) 1333 outs() << "+0x" << Twine::utohexstr(Disp); 1334 outs() << '>'; 1335 } 1336 } 1337 } 1338 outs() << "\n"; 1339 1340 // Hexagon does this in pretty printer 1341 if (Obj->getArch() != Triple::hexagon) { 1342 // Print relocation for instruction. 1343 while (RelCur != RelEnd) { 1344 uint64_t Offset = RelCur->getOffset(); 1345 // If this relocation is hidden, skip it. 1346 if (getHidden(*RelCur) || ((SectionAddr + Offset) < StartAddress)) { 1347 ++RelCur; 1348 continue; 1349 } 1350 1351 // Stop when RelCur's offset is past the current instruction. 1352 if (Offset >= Index + Size) 1353 break; 1354 1355 // When --adjust-vma is used, update the address printed. 1356 if (RelCur->getSymbol() != Obj->symbol_end()) { 1357 Expected<section_iterator> SymSI = 1358 RelCur->getSymbol()->getSection(); 1359 if (SymSI && *SymSI != Obj->section_end() && 1360 (shouldAdjustVA(**SymSI))) 1361 Offset += AdjustVMA; 1362 } 1363 1364 printRelocation(*RelCur, SectionAddr + Offset, 1365 Obj->getBytesInAddress()); 1366 ++RelCur; 1367 } 1368 } 1369 } 1370 } 1371 } 1372 } 1373 1374 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1375 if (StartAddress > StopAddress) 1376 error("Start address should be less than stop address"); 1377 1378 const Target *TheTarget = getTarget(Obj); 1379 1380 // Package up features to be passed to target/subtarget 1381 SubtargetFeatures Features = Obj->getFeatures(); 1382 if (!MAttrs.empty()) 1383 for (unsigned I = 0; I != MAttrs.size(); ++I) 1384 Features.AddFeature(MAttrs[I]); 1385 1386 std::unique_ptr<const MCRegisterInfo> MRI( 1387 TheTarget->createMCRegInfo(TripleName)); 1388 if (!MRI) 1389 report_error(Obj->getFileName(), 1390 "no register info for target " + TripleName); 1391 1392 // Set up disassembler. 1393 std::unique_ptr<const MCAsmInfo> AsmInfo( 1394 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1395 if (!AsmInfo) 1396 report_error(Obj->getFileName(), 1397 "no assembly info for target " + TripleName); 1398 std::unique_ptr<const MCSubtargetInfo> STI( 1399 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1400 if (!STI) 1401 report_error(Obj->getFileName(), 1402 "no subtarget info for target " + TripleName); 1403 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1404 if (!MII) 1405 report_error(Obj->getFileName(), 1406 "no instruction info for target " + TripleName); 1407 MCObjectFileInfo MOFI; 1408 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1409 // FIXME: for now initialize MCObjectFileInfo with default values 1410 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1411 1412 std::unique_ptr<MCDisassembler> DisAsm( 1413 TheTarget->createMCDisassembler(*STI, Ctx)); 1414 if (!DisAsm) 1415 report_error(Obj->getFileName(), 1416 "no disassembler for target " + TripleName); 1417 1418 std::unique_ptr<const MCInstrAnalysis> MIA( 1419 TheTarget->createMCInstrAnalysis(MII.get())); 1420 1421 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1422 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1423 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1424 if (!IP) 1425 report_error(Obj->getFileName(), 1426 "no instruction printer for target " + TripleName); 1427 IP->setPrintImmHex(PrintImmHex); 1428 1429 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1430 SourcePrinter SP(Obj, TheTarget->getName()); 1431 1432 for (StringRef Opt : DisassemblerOptions) 1433 if (!IP->applyTargetSpecificCLOption(Opt)) 1434 error("Unrecognized disassembler option: " + Opt); 1435 1436 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), MIA.get(), IP.get(), 1437 STI.get(), PIP, SP, InlineRelocs); 1438 } 1439 1440 void llvm::printRelocations(const ObjectFile *Obj) { 1441 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1442 "%08" PRIx64; 1443 // Regular objdump doesn't print relocations in non-relocatable object 1444 // files. 1445 if (!Obj->isRelocatableObject()) 1446 return; 1447 1448 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1449 if (Section.relocation_begin() == Section.relocation_end()) 1450 continue; 1451 StringRef SecName; 1452 error(Section.getName(SecName)); 1453 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1454 for (const RelocationRef &Reloc : Section.relocations()) { 1455 uint64_t Address = Reloc.getOffset(); 1456 SmallString<32> RelocName; 1457 SmallString<32> ValueStr; 1458 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1459 continue; 1460 Reloc.getTypeName(RelocName); 1461 error(getRelocationValueString(Reloc, ValueStr)); 1462 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1463 << ValueStr << "\n"; 1464 } 1465 outs() << "\n"; 1466 } 1467 } 1468 1469 void llvm::printDynamicRelocations(const ObjectFile *Obj) { 1470 // For the moment, this option is for ELF only 1471 if (!Obj->isELF()) 1472 return; 1473 1474 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1475 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1476 error("not a dynamic object"); 1477 return; 1478 } 1479 1480 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1481 if (DynRelSec.empty()) 1482 return; 1483 1484 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1485 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1486 for (const SectionRef &Section : DynRelSec) { 1487 if (Section.relocation_begin() == Section.relocation_end()) 1488 continue; 1489 for (const RelocationRef &Reloc : Section.relocations()) { 1490 uint64_t Address = Reloc.getOffset(); 1491 SmallString<32> RelocName; 1492 SmallString<32> ValueStr; 1493 Reloc.getTypeName(RelocName); 1494 error(getRelocationValueString(Reloc, ValueStr)); 1495 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1496 << ValueStr << "\n"; 1497 } 1498 } 1499 } 1500 1501 // Returns true if we need to show LMA column when dumping section headers. We 1502 // show it only when the platform is ELF and either we have at least one section 1503 // whose VMA and LMA are different and/or when --show-lma flag is used. 1504 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1505 if (!Obj->isELF()) 1506 return false; 1507 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1508 if (S.getAddress() != getELFSectionLMA(S)) 1509 return true; 1510 return ShowLMA; 1511 } 1512 1513 void llvm::printSectionHeaders(const ObjectFile *Obj) { 1514 bool HasLMAColumn = shouldDisplayLMA(Obj); 1515 if (HasLMAColumn) 1516 outs() << "Sections:\n" 1517 "Idx Name Size VMA LMA " 1518 "Type\n"; 1519 else 1520 outs() << "Sections:\n" 1521 "Idx Name Size VMA Type\n"; 1522 1523 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1524 StringRef Name; 1525 error(Section.getName(Name)); 1526 uint64_t VMA = Section.getAddress(); 1527 if (shouldAdjustVA(Section)) 1528 VMA += AdjustVMA; 1529 1530 uint64_t Size = Section.getSize(); 1531 bool Text = Section.isText(); 1532 bool Data = Section.isData(); 1533 bool BSS = Section.isBSS(); 1534 std::string Type = (std::string(Text ? "TEXT " : "") + 1535 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1536 1537 if (HasLMAColumn) 1538 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %016" PRIx64 1539 " %s\n", 1540 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1541 VMA, getELFSectionLMA(Section), Type.c_str()); 1542 else 1543 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1544 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1545 VMA, Type.c_str()); 1546 } 1547 outs() << "\n"; 1548 } 1549 1550 void llvm::printSectionContents(const ObjectFile *Obj) { 1551 std::error_code EC; 1552 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1553 StringRef Name; 1554 StringRef Contents; 1555 error(Section.getName(Name)); 1556 uint64_t BaseAddr = Section.getAddress(); 1557 uint64_t Size = Section.getSize(); 1558 if (!Size) 1559 continue; 1560 1561 outs() << "Contents of section " << Name << ":\n"; 1562 if (Section.isBSS()) { 1563 outs() << format("<skipping contents of bss section at [%04" PRIx64 1564 ", %04" PRIx64 ")>\n", 1565 BaseAddr, BaseAddr + Size); 1566 continue; 1567 } 1568 1569 error(Section.getContents(Contents)); 1570 1571 // Dump out the content as hex and printable ascii characters. 1572 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1573 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1574 // Dump line of hex. 1575 for (std::size_t I = 0; I < 16; ++I) { 1576 if (I != 0 && I % 4 == 0) 1577 outs() << ' '; 1578 if (Addr + I < End) 1579 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1580 << hexdigit(Contents[Addr + I] & 0xF, true); 1581 else 1582 outs() << " "; 1583 } 1584 // Print ascii. 1585 outs() << " "; 1586 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1587 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1588 outs() << Contents[Addr + I]; 1589 else 1590 outs() << "."; 1591 } 1592 outs() << "\n"; 1593 } 1594 } 1595 } 1596 1597 void llvm::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1598 StringRef ArchitectureName) { 1599 outs() << "SYMBOL TABLE:\n"; 1600 1601 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1602 printCOFFSymbolTable(Coff); 1603 return; 1604 } 1605 1606 const StringRef FileName = O->getFileName(); 1607 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1608 // Skip printing the special zero symbol when dumping an ELF file. 1609 // This makes the output consistent with the GNU objdump. 1610 if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O)) 1611 continue; 1612 1613 const SymbolRef &Symbol = *I; 1614 uint64_t Address = unwrapOrError(Symbol.getAddress(), ArchiveName, FileName, 1615 ArchitectureName); 1616 if ((Address < StartAddress) || (Address > StopAddress)) 1617 continue; 1618 SymbolRef::Type Type = unwrapOrError(Symbol.getType(), ArchiveName, 1619 FileName, ArchitectureName); 1620 uint32_t Flags = Symbol.getFlags(); 1621 section_iterator Section = unwrapOrError(Symbol.getSection(), ArchiveName, 1622 FileName, ArchitectureName); 1623 StringRef Name; 1624 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) 1625 Section->getName(Name); 1626 else 1627 Name = unwrapOrError(Symbol.getName(), ArchiveName, FileName, 1628 ArchitectureName); 1629 1630 bool Global = Flags & SymbolRef::SF_Global; 1631 bool Weak = Flags & SymbolRef::SF_Weak; 1632 bool Absolute = Flags & SymbolRef::SF_Absolute; 1633 bool Common = Flags & SymbolRef::SF_Common; 1634 bool Hidden = Flags & SymbolRef::SF_Hidden; 1635 1636 char GlobLoc = ' '; 1637 if (Type != SymbolRef::ST_Unknown) 1638 GlobLoc = Global ? 'g' : 'l'; 1639 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1640 ? 'd' : ' '; 1641 char FileFunc = ' '; 1642 if (Type == SymbolRef::ST_File) 1643 FileFunc = 'f'; 1644 else if (Type == SymbolRef::ST_Function) 1645 FileFunc = 'F'; 1646 else if (Type == SymbolRef::ST_Data) 1647 FileFunc = 'O'; 1648 1649 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1650 "%08" PRIx64; 1651 1652 outs() << format(Fmt, Address) << " " 1653 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1654 << (Weak ? 'w' : ' ') // Weak? 1655 << ' ' // Constructor. Not supported yet. 1656 << ' ' // Warning. Not supported yet. 1657 << ' ' // Indirect reference to another symbol. 1658 << Debug // Debugging (d) or dynamic (D) symbol. 1659 << FileFunc // Name of function (F), file (f) or object (O). 1660 << ' '; 1661 if (Absolute) { 1662 outs() << "*ABS*"; 1663 } else if (Common) { 1664 outs() << "*COM*"; 1665 } else if (Section == O->section_end()) { 1666 outs() << "*UND*"; 1667 } else { 1668 if (const MachOObjectFile *MachO = 1669 dyn_cast<const MachOObjectFile>(O)) { 1670 DataRefImpl DR = Section->getRawDataRefImpl(); 1671 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1672 outs() << SegmentName << ","; 1673 } 1674 StringRef SectionName; 1675 error(Section->getName(SectionName)); 1676 outs() << SectionName; 1677 } 1678 1679 outs() << '\t'; 1680 if (Common || isa<ELFObjectFileBase>(O)) { 1681 uint64_t Val = 1682 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1683 outs() << format("\t %08" PRIx64 " ", Val); 1684 } 1685 1686 if (Hidden) 1687 outs() << ".hidden "; 1688 1689 if (Demangle) 1690 outs() << demangle(Name) << '\n'; 1691 else 1692 outs() << Name << '\n'; 1693 } 1694 } 1695 1696 static void printUnwindInfo(const ObjectFile *O) { 1697 outs() << "Unwind info:\n\n"; 1698 1699 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1700 printCOFFUnwindInfo(Coff); 1701 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1702 printMachOUnwindInfo(MachO); 1703 else 1704 // TODO: Extract DWARF dump tool to objdump. 1705 WithColor::error(errs(), ToolName) 1706 << "This operation is only currently supported " 1707 "for COFF and MachO object files.\n"; 1708 } 1709 1710 void llvm::printExportsTrie(const ObjectFile *o) { 1711 outs() << "Exports trie:\n"; 1712 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1713 printMachOExportsTrie(MachO); 1714 else 1715 WithColor::error(errs(), ToolName) 1716 << "This operation is only currently supported " 1717 "for Mach-O executable files.\n"; 1718 } 1719 1720 void llvm::printRebaseTable(ObjectFile *o) { 1721 outs() << "Rebase table:\n"; 1722 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1723 printMachORebaseTable(MachO); 1724 else 1725 WithColor::error(errs(), ToolName) 1726 << "This operation is only currently supported " 1727 "for Mach-O executable files.\n"; 1728 } 1729 1730 void llvm::printBindTable(ObjectFile *o) { 1731 outs() << "Bind table:\n"; 1732 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1733 printMachOBindTable(MachO); 1734 else 1735 WithColor::error(errs(), ToolName) 1736 << "This operation is only currently supported " 1737 "for Mach-O executable files.\n"; 1738 } 1739 1740 void llvm::printLazyBindTable(ObjectFile *o) { 1741 outs() << "Lazy bind table:\n"; 1742 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1743 printMachOLazyBindTable(MachO); 1744 else 1745 WithColor::error(errs(), ToolName) 1746 << "This operation is only currently supported " 1747 "for Mach-O executable files.\n"; 1748 } 1749 1750 void llvm::printWeakBindTable(ObjectFile *o) { 1751 outs() << "Weak bind table:\n"; 1752 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1753 printMachOWeakBindTable(MachO); 1754 else 1755 WithColor::error(errs(), ToolName) 1756 << "This operation is only currently supported " 1757 "for Mach-O executable files.\n"; 1758 } 1759 1760 /// Dump the raw contents of the __clangast section so the output can be piped 1761 /// into llvm-bcanalyzer. 1762 void llvm::printRawClangAST(const ObjectFile *Obj) { 1763 if (outs().is_displayed()) { 1764 WithColor::error(errs(), ToolName) 1765 << "The -raw-clang-ast option will dump the raw binary contents of " 1766 "the clang ast section.\n" 1767 "Please redirect the output to a file or another program such as " 1768 "llvm-bcanalyzer.\n"; 1769 return; 1770 } 1771 1772 StringRef ClangASTSectionName("__clangast"); 1773 if (isa<COFFObjectFile>(Obj)) { 1774 ClangASTSectionName = "clangast"; 1775 } 1776 1777 Optional<object::SectionRef> ClangASTSection; 1778 for (auto Sec : ToolSectionFilter(*Obj)) { 1779 StringRef Name; 1780 Sec.getName(Name); 1781 if (Name == ClangASTSectionName) { 1782 ClangASTSection = Sec; 1783 break; 1784 } 1785 } 1786 if (!ClangASTSection) 1787 return; 1788 1789 StringRef ClangASTContents; 1790 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1791 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1792 } 1793 1794 static void printFaultMaps(const ObjectFile *Obj) { 1795 StringRef FaultMapSectionName; 1796 1797 if (isa<ELFObjectFileBase>(Obj)) { 1798 FaultMapSectionName = ".llvm_faultmaps"; 1799 } else if (isa<MachOObjectFile>(Obj)) { 1800 FaultMapSectionName = "__llvm_faultmaps"; 1801 } else { 1802 WithColor::error(errs(), ToolName) 1803 << "This operation is only currently supported " 1804 "for ELF and Mach-O executable files.\n"; 1805 return; 1806 } 1807 1808 Optional<object::SectionRef> FaultMapSection; 1809 1810 for (auto Sec : ToolSectionFilter(*Obj)) { 1811 StringRef Name; 1812 Sec.getName(Name); 1813 if (Name == FaultMapSectionName) { 1814 FaultMapSection = Sec; 1815 break; 1816 } 1817 } 1818 1819 outs() << "FaultMap table:\n"; 1820 1821 if (!FaultMapSection.hasValue()) { 1822 outs() << "<not found>\n"; 1823 return; 1824 } 1825 1826 StringRef FaultMapContents; 1827 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1828 1829 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1830 FaultMapContents.bytes_end()); 1831 1832 outs() << FMP; 1833 } 1834 1835 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1836 if (O->isELF()) { 1837 printELFFileHeader(O); 1838 printELFDynamicSection(O); 1839 printELFSymbolVersionInfo(O); 1840 return; 1841 } 1842 if (O->isCOFF()) 1843 return printCOFFFileHeader(O); 1844 if (O->isWasm()) 1845 return printWasmFileHeader(O); 1846 if (O->isMachO()) { 1847 printMachOFileHeader(O); 1848 if (!OnlyFirst) 1849 printMachOLoadCommands(O); 1850 return; 1851 } 1852 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1853 } 1854 1855 static void printFileHeaders(const ObjectFile *O) { 1856 if (!O->isELF() && !O->isCOFF()) 1857 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1858 1859 Triple::ArchType AT = O->getArch(); 1860 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1861 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 1862 1863 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1864 outs() << "start address: " 1865 << "0x" << format(Fmt.data(), Address) << "\n\n"; 1866 } 1867 1868 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 1869 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 1870 if (!ModeOrErr) { 1871 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 1872 consumeError(ModeOrErr.takeError()); 1873 return; 1874 } 1875 sys::fs::perms Mode = ModeOrErr.get(); 1876 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1877 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1878 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1879 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1880 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1881 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1882 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1883 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1884 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1885 1886 outs() << " "; 1887 1888 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 1889 unwrapOrError(C.getGID(), Filename), 1890 unwrapOrError(C.getRawSize(), Filename)); 1891 1892 StringRef RawLastModified = C.getRawLastModified(); 1893 unsigned Seconds; 1894 if (RawLastModified.getAsInteger(10, Seconds)) 1895 outs() << "(date: \"" << RawLastModified 1896 << "\" contains non-decimal chars) "; 1897 else { 1898 // Since ctime(3) returns a 26 character string of the form: 1899 // "Sun Sep 16 01:03:52 1973\n\0" 1900 // just print 24 characters. 1901 time_t t = Seconds; 1902 outs() << format("%.24s ", ctime(&t)); 1903 } 1904 1905 StringRef Name = ""; 1906 Expected<StringRef> NameOrErr = C.getName(); 1907 if (!NameOrErr) { 1908 consumeError(NameOrErr.takeError()); 1909 Name = unwrapOrError(C.getRawName(), Filename); 1910 } else { 1911 Name = NameOrErr.get(); 1912 } 1913 outs() << Name << "\n"; 1914 } 1915 1916 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 1917 const Archive::Child *C = nullptr) { 1918 // Avoid other output when using a raw option. 1919 if (!RawClangAST) { 1920 outs() << '\n'; 1921 if (A) 1922 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 1923 else 1924 outs() << O->getFileName(); 1925 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 1926 } 1927 1928 StringRef ArchiveName = A ? A->getFileName() : ""; 1929 if (FileHeaders) 1930 printFileHeaders(O); 1931 if (ArchiveHeaders && !MachOOpt && C) 1932 printArchiveChild(ArchiveName, *C); 1933 if (Disassemble) 1934 disassembleObject(O, Relocations); 1935 if (Relocations && !Disassemble) 1936 printRelocations(O); 1937 if (DynamicRelocations) 1938 printDynamicRelocations(O); 1939 if (SectionHeaders) 1940 printSectionHeaders(O); 1941 if (SectionContents) 1942 printSectionContents(O); 1943 if (SymbolTable) 1944 printSymbolTable(O, ArchiveName); 1945 if (UnwindInfo) 1946 printUnwindInfo(O); 1947 if (PrivateHeaders || FirstPrivateHeader) 1948 printPrivateFileHeaders(O, FirstPrivateHeader); 1949 if (ExportsTrie) 1950 printExportsTrie(O); 1951 if (Rebase) 1952 printRebaseTable(O); 1953 if (Bind) 1954 printBindTable(O); 1955 if (LazyBind) 1956 printLazyBindTable(O); 1957 if (WeakBind) 1958 printWeakBindTable(O); 1959 if (RawClangAST) 1960 printRawClangAST(O); 1961 if (PrintFaultMaps) 1962 printFaultMaps(O); 1963 if (DwarfDumpType != DIDT_Null) { 1964 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 1965 // Dump the complete DWARF structure. 1966 DIDumpOptions DumpOpts; 1967 DumpOpts.DumpType = DwarfDumpType; 1968 DICtx->dump(outs(), DumpOpts); 1969 } 1970 } 1971 1972 static void dumpObject(const COFFImportFile *I, const Archive *A, 1973 const Archive::Child *C = nullptr) { 1974 StringRef ArchiveName = A ? A->getFileName() : ""; 1975 1976 // Avoid other output when using a raw option. 1977 if (!RawClangAST) 1978 outs() << '\n' 1979 << ArchiveName << "(" << I->getFileName() << ")" 1980 << ":\tfile format COFF-import-file" 1981 << "\n\n"; 1982 1983 if (ArchiveHeaders && !MachOOpt && C) 1984 printArchiveChild(ArchiveName, *C); 1985 if (SymbolTable) 1986 printCOFFSymbolTable(I); 1987 } 1988 1989 /// Dump each object file in \a a; 1990 static void dumpArchive(const Archive *A) { 1991 Error Err = Error::success(); 1992 for (auto &C : A->children(Err)) { 1993 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1994 if (!ChildOrErr) { 1995 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 1996 report_error(std::move(E), A->getFileName(), C); 1997 continue; 1998 } 1999 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2000 dumpObject(O, A, &C); 2001 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2002 dumpObject(I, A, &C); 2003 else 2004 report_error(A->getFileName(), object_error::invalid_file_type); 2005 } 2006 if (Err) 2007 report_error(std::move(Err), A->getFileName()); 2008 } 2009 2010 /// Open file and figure out how to dump it. 2011 static void dumpInput(StringRef file) { 2012 // If we are using the Mach-O specific object file parser, then let it parse 2013 // the file and process the command line options. So the -arch flags can 2014 // be used to select specific slices, etc. 2015 if (MachOOpt) { 2016 parseInputMachO(file); 2017 return; 2018 } 2019 2020 // Attempt to open the binary. 2021 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2022 Binary &Binary = *OBinary.getBinary(); 2023 2024 if (Archive *A = dyn_cast<Archive>(&Binary)) 2025 dumpArchive(A); 2026 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2027 dumpObject(O); 2028 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2029 parseInputMachO(UB); 2030 else 2031 report_error(file, object_error::invalid_file_type); 2032 } 2033 2034 int main(int argc, char **argv) { 2035 InitLLVM X(argc, argv); 2036 2037 // Initialize targets and assembly printers/parsers. 2038 llvm::InitializeAllTargetInfos(); 2039 llvm::InitializeAllTargetMCs(); 2040 llvm::InitializeAllDisassemblers(); 2041 2042 // Register the target printer for --version. 2043 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2044 2045 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2046 2047 ToolName = argv[0]; 2048 2049 // Defaults to a.out if no filenames specified. 2050 if (InputFilenames.empty()) 2051 InputFilenames.push_back("a.out"); 2052 2053 if (AllHeaders) 2054 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2055 SectionHeaders = SymbolTable = true; 2056 2057 if (DisassembleAll || PrintSource || PrintLines) 2058 Disassemble = true; 2059 2060 if (!Disassemble 2061 && !Relocations 2062 && !DynamicRelocations 2063 && !SectionHeaders 2064 && !SectionContents 2065 && !SymbolTable 2066 && !UnwindInfo 2067 && !PrivateHeaders 2068 && !FileHeaders 2069 && !FirstPrivateHeader 2070 && !ExportsTrie 2071 && !Rebase 2072 && !Bind 2073 && !LazyBind 2074 && !WeakBind 2075 && !RawClangAST 2076 && !(UniversalHeaders && MachOOpt) 2077 && !ArchiveHeaders 2078 && !(IndirectSymbols && MachOOpt) 2079 && !(DataInCode && MachOOpt) 2080 && !(LinkOptHints && MachOOpt) 2081 && !(InfoPlist && MachOOpt) 2082 && !(DylibsUsed && MachOOpt) 2083 && !(DylibId && MachOOpt) 2084 && !(ObjcMetaData && MachOOpt) 2085 && !(!FilterSections.empty() && MachOOpt) 2086 && !PrintFaultMaps 2087 && DwarfDumpType == DIDT_Null) { 2088 cl::PrintHelpMessage(); 2089 return 2; 2090 } 2091 2092 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2093 DisassembleFunctions.end()); 2094 2095 llvm::for_each(InputFilenames, dumpInput); 2096 2097 return EXIT_SUCCESS; 2098 } 2099