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 void llvm::error(Error E) { 334 if (!E) 335 return; 336 WithColor::error(errs(), ToolName) << toString(std::move(E)); 337 exit(1); 338 } 339 340 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 341 WithColor::error(errs(), ToolName) << Message << ".\n"; 342 errs().flush(); 343 exit(1); 344 } 345 346 void llvm::warn(StringRef Message) { 347 WithColor::warning(errs(), ToolName) << Message << ".\n"; 348 errs().flush(); 349 } 350 351 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 352 Twine Message) { 353 WithColor::error(errs(), ToolName) 354 << "'" << File << "': " << Message << ".\n"; 355 exit(1); 356 } 357 358 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(Error E, StringRef File) { 359 assert(E); 360 std::string Buf; 361 raw_string_ostream OS(Buf); 362 logAllUnhandledErrors(std::move(E), OS); 363 OS.flush(); 364 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 365 exit(1); 366 } 367 368 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(Error E, StringRef ArchiveName, 369 StringRef FileName, 370 StringRef ArchitectureName) { 371 assert(E); 372 WithColor::error(errs(), ToolName); 373 if (ArchiveName != "") 374 errs() << ArchiveName << "(" << FileName << ")"; 375 else 376 errs() << "'" << FileName << "'"; 377 if (!ArchitectureName.empty()) 378 errs() << " (for architecture " << ArchitectureName << ")"; 379 std::string Buf; 380 raw_string_ostream OS(Buf); 381 logAllUnhandledErrors(std::move(E), OS); 382 OS.flush(); 383 errs() << ": " << Buf; 384 exit(1); 385 } 386 387 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(Error E, StringRef ArchiveName, 388 const object::Archive::Child &C, 389 StringRef ArchitectureName) { 390 Expected<StringRef> NameOrErr = C.getName(); 391 // TODO: if we have a error getting the name then it would be nice to print 392 // the index of which archive member this is and or its offset in the 393 // archive instead of "???" as the name. 394 if (!NameOrErr) { 395 consumeError(NameOrErr.takeError()); 396 llvm::report_error(std::move(E), ArchiveName, "???", ArchitectureName); 397 } else 398 llvm::report_error(std::move(E), ArchiveName, NameOrErr.get(), 399 ArchitectureName); 400 } 401 402 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 403 // Figure out the target triple. 404 llvm::Triple TheTriple("unknown-unknown-unknown"); 405 if (TripleName.empty()) { 406 if (Obj) 407 TheTriple = Obj->makeTriple(); 408 } else { 409 TheTriple.setTriple(Triple::normalize(TripleName)); 410 411 // Use the triple, but also try to combine with ARM build attributes. 412 if (Obj) { 413 auto Arch = Obj->getArch(); 414 if (Arch == Triple::arm || Arch == Triple::armeb) 415 Obj->setARMSubArch(TheTriple); 416 } 417 } 418 419 // Get the target specific parser. 420 std::string Error; 421 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 422 Error); 423 if (!TheTarget) { 424 if (Obj) 425 report_error(Obj->getFileName(), "can't find target: " + Error); 426 else 427 error("can't find target: " + Error); 428 } 429 430 // Update the triple name and return the found target. 431 TripleName = TheTriple.getTriple(); 432 return TheTarget; 433 } 434 435 bool llvm::isRelocAddressLess(RelocationRef A, RelocationRef B) { 436 return A.getOffset() < B.getOffset(); 437 } 438 439 static Error getRelocationValueString(const RelocationRef &Rel, 440 SmallVectorImpl<char> &Result) { 441 const ObjectFile *Obj = Rel.getObject(); 442 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 443 return getELFRelocationValueString(ELF, Rel, Result); 444 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 445 return getCOFFRelocationValueString(COFF, Rel, Result); 446 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 447 return getWasmRelocationValueString(Wasm, Rel, Result); 448 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 449 return getMachORelocationValueString(MachO, Rel, Result); 450 llvm_unreachable("unknown object file format"); 451 } 452 453 /// Indicates whether this relocation should hidden when listing 454 /// relocations, usually because it is the trailing part of a multipart 455 /// relocation that will be printed as part of the leading relocation. 456 static bool getHidden(RelocationRef RelRef) { 457 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 458 if (!MachO) 459 return false; 460 461 unsigned Arch = MachO->getArch(); 462 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 463 uint64_t Type = MachO->getRelocationType(Rel); 464 465 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 466 // is always hidden. 467 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 468 return Type == MachO::GENERIC_RELOC_PAIR; 469 470 if (Arch == Triple::x86_64) { 471 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 472 // an X86_64_RELOC_SUBTRACTOR. 473 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 474 DataRefImpl RelPrev = Rel; 475 RelPrev.d.a--; 476 uint64_t PrevType = MachO->getRelocationType(RelPrev); 477 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 478 return true; 479 } 480 } 481 482 return false; 483 } 484 485 namespace { 486 class SourcePrinter { 487 protected: 488 DILineInfo OldLineInfo; 489 const ObjectFile *Obj = nullptr; 490 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 491 // File name to file contents of source 492 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 493 // Mark the line endings of the cached source 494 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 495 496 private: 497 bool cacheSource(const DILineInfo& LineInfoFile); 498 499 public: 500 SourcePrinter() = default; 501 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 502 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 503 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 504 DefaultArch); 505 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 506 } 507 virtual ~SourcePrinter() = default; 508 virtual void printSourceLine(raw_ostream &OS, 509 object::SectionedAddress Address, 510 StringRef Delimiter = "; "); 511 }; 512 513 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 514 std::unique_ptr<MemoryBuffer> Buffer; 515 if (LineInfo.Source) { 516 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 517 } else { 518 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 519 if (!BufferOrError) 520 return false; 521 Buffer = std::move(*BufferOrError); 522 } 523 // Chomp the file to get lines 524 const char *BufferStart = Buffer->getBufferStart(), 525 *BufferEnd = Buffer->getBufferEnd(); 526 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 527 const char *Start = BufferStart; 528 for (const char *I = BufferStart; I != BufferEnd; ++I) 529 if (*I == '\n') { 530 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 531 Start = I + 1; 532 } 533 if (Start < BufferEnd) 534 Lines.emplace_back(Start, BufferEnd - Start); 535 SourceCache[LineInfo.FileName] = std::move(Buffer); 536 return true; 537 } 538 539 void SourcePrinter::printSourceLine(raw_ostream &OS, 540 object::SectionedAddress Address, 541 StringRef Delimiter) { 542 if (!Symbolizer) 543 return; 544 DILineInfo LineInfo = DILineInfo(); 545 auto ExpectedLineInfo = 546 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 547 if (!ExpectedLineInfo) 548 consumeError(ExpectedLineInfo.takeError()); 549 else 550 LineInfo = *ExpectedLineInfo; 551 552 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 553 LineInfo.Line == 0) 554 return; 555 556 if (PrintLines) 557 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 558 if (PrintSource) { 559 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 560 if (!cacheSource(LineInfo)) 561 return; 562 auto LineBuffer = LineCache.find(LineInfo.FileName); 563 if (LineBuffer != LineCache.end()) { 564 if (LineInfo.Line > LineBuffer->second.size()) 565 return; 566 // Vector begins at 0, line numbers are non-zero 567 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1] << '\n'; 568 } 569 } 570 OldLineInfo = LineInfo; 571 } 572 573 static bool isArmElf(const ObjectFile *Obj) { 574 return (Obj->isELF() && 575 (Obj->getArch() == Triple::aarch64 || 576 Obj->getArch() == Triple::aarch64_be || 577 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 578 Obj->getArch() == Triple::thumb || 579 Obj->getArch() == Triple::thumbeb)); 580 } 581 582 static void printRelocation(const RelocationRef &Rel, uint64_t Address, 583 uint8_t AddrSize) { 584 StringRef Fmt = 585 AddrSize > 4 ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 586 SmallString<16> Name; 587 SmallString<32> Val; 588 Rel.getTypeName(Name); 589 error(getRelocationValueString(Rel, Val)); 590 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n"; 591 } 592 593 class PrettyPrinter { 594 public: 595 virtual ~PrettyPrinter() = default; 596 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 597 ArrayRef<uint8_t> Bytes, 598 object::SectionedAddress Address, raw_ostream &OS, 599 StringRef Annot, MCSubtargetInfo const &STI, 600 SourcePrinter *SP, 601 std::vector<RelocationRef> *Rels = nullptr) { 602 if (SP && (PrintSource || PrintLines)) 603 SP->printSourceLine(OS, Address); 604 if (!NoLeadingAddr) 605 OS << format("%8" PRIx64 ":", Address.Address); 606 if (!NoShowRawInsn) { 607 OS << "\t"; 608 dumpBytes(Bytes, OS); 609 } 610 if (MI) 611 IP.printInst(MI, OS, "", STI); 612 else 613 OS << " <unknown>"; 614 } 615 }; 616 PrettyPrinter PrettyPrinterInst; 617 class HexagonPrettyPrinter : public PrettyPrinter { 618 public: 619 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 620 raw_ostream &OS) { 621 uint32_t opcode = 622 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 623 if (!NoLeadingAddr) 624 OS << format("%8" PRIx64 ":", Address); 625 if (!NoShowRawInsn) { 626 OS << "\t"; 627 dumpBytes(Bytes.slice(0, 4), OS); 628 OS << format("%08" PRIx32, opcode); 629 } 630 } 631 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 632 object::SectionedAddress Address, raw_ostream &OS, 633 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 634 std::vector<RelocationRef> *Rels) override { 635 if (SP && (PrintSource || PrintLines)) 636 SP->printSourceLine(OS, Address, ""); 637 if (!MI) { 638 printLead(Bytes, Address.Address, OS); 639 OS << " <unknown>"; 640 return; 641 } 642 std::string Buffer; 643 { 644 raw_string_ostream TempStream(Buffer); 645 IP.printInst(MI, TempStream, "", STI); 646 } 647 StringRef Contents(Buffer); 648 // Split off bundle attributes 649 auto PacketBundle = Contents.rsplit('\n'); 650 // Split off first instruction from the rest 651 auto HeadTail = PacketBundle.first.split('\n'); 652 auto Preamble = " { "; 653 auto Separator = ""; 654 655 // Hexagon's packets require relocations to be inline rather than 656 // clustered at the end of the packet. 657 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 658 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 659 auto PrintReloc = [&]() -> void { 660 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 661 if (RelCur->getOffset() == Address.Address) { 662 printRelocation(*RelCur, Address.Address, 4); 663 return; 664 } 665 ++RelCur; 666 } 667 }; 668 669 while (!HeadTail.first.empty()) { 670 OS << Separator; 671 Separator = "\n"; 672 if (SP && (PrintSource || PrintLines)) 673 SP->printSourceLine(OS, Address, ""); 674 printLead(Bytes, Address.Address, OS); 675 OS << Preamble; 676 Preamble = " "; 677 StringRef Inst; 678 auto Duplex = HeadTail.first.split('\v'); 679 if (!Duplex.second.empty()) { 680 OS << Duplex.first; 681 OS << "; "; 682 Inst = Duplex.second; 683 } 684 else 685 Inst = HeadTail.first; 686 OS << Inst; 687 HeadTail = HeadTail.second.split('\n'); 688 if (HeadTail.first.empty()) 689 OS << " } " << PacketBundle.second; 690 PrintReloc(); 691 Bytes = Bytes.slice(4); 692 Address.Address += 4; 693 } 694 } 695 }; 696 HexagonPrettyPrinter HexagonPrettyPrinterInst; 697 698 class AMDGCNPrettyPrinter : public PrettyPrinter { 699 public: 700 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 701 object::SectionedAddress Address, raw_ostream &OS, 702 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 703 std::vector<RelocationRef> *Rels) override { 704 if (SP && (PrintSource || PrintLines)) 705 SP->printSourceLine(OS, Address); 706 707 typedef support::ulittle32_t U32; 708 709 if (MI) { 710 SmallString<40> InstStr; 711 raw_svector_ostream IS(InstStr); 712 713 IP.printInst(MI, IS, "", STI); 714 715 OS << left_justify(IS.str(), 60); 716 } else { 717 // an unrecognized encoding - this is probably data so represent it 718 // using the .long directive, or .byte directive if fewer than 4 bytes 719 // remaining 720 if (Bytes.size() >= 4) { 721 OS << format("\t.long 0x%08" PRIx32 " ", 722 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 723 OS.indent(42); 724 } else { 725 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 726 for (unsigned int i = 1; i < Bytes.size(); i++) 727 OS << format(", 0x%02" PRIx8, Bytes[i]); 728 OS.indent(55 - (6 * Bytes.size())); 729 } 730 } 731 732 OS << format("// %012" PRIX64 ": ", Address.Address); 733 if (Bytes.size() >=4) { 734 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 735 Bytes.size() / sizeof(U32))) 736 // D should be explicitly casted to uint32_t here as it is passed 737 // by format to snprintf as vararg. 738 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 739 } else { 740 for (unsigned int i = 0; i < Bytes.size(); i++) 741 OS << format("%02" PRIX8 " ", Bytes[i]); 742 } 743 744 if (!Annot.empty()) 745 OS << "// " << Annot; 746 } 747 }; 748 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 749 750 class BPFPrettyPrinter : public PrettyPrinter { 751 public: 752 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 753 object::SectionedAddress Address, raw_ostream &OS, 754 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 755 std::vector<RelocationRef> *Rels) override { 756 if (SP && (PrintSource || PrintLines)) 757 SP->printSourceLine(OS, Address); 758 if (!NoLeadingAddr) 759 OS << format("%8" PRId64 ":", Address.Address / 8); 760 if (!NoShowRawInsn) { 761 OS << "\t"; 762 dumpBytes(Bytes, OS); 763 } 764 if (MI) 765 IP.printInst(MI, OS, "", STI); 766 else 767 OS << " <unknown>"; 768 } 769 }; 770 BPFPrettyPrinter BPFPrettyPrinterInst; 771 772 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 773 switch(Triple.getArch()) { 774 default: 775 return PrettyPrinterInst; 776 case Triple::hexagon: 777 return HexagonPrettyPrinterInst; 778 case Triple::amdgcn: 779 return AMDGCNPrettyPrinterInst; 780 case Triple::bpfel: 781 case Triple::bpfeb: 782 return BPFPrettyPrinterInst; 783 } 784 } 785 } 786 787 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 788 assert(Obj->isELF()); 789 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 790 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 791 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 792 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 793 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 794 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 795 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 796 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 797 llvm_unreachable("Unsupported binary format"); 798 } 799 800 template <class ELFT> static void 801 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 802 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 803 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 804 uint8_t SymbolType = Symbol.getELFType(); 805 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 806 continue; 807 808 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 809 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 810 if (Name.empty()) 811 continue; 812 813 section_iterator SecI = 814 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 815 if (SecI == Obj->section_end()) 816 continue; 817 818 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 819 } 820 } 821 822 static void 823 addDynamicElfSymbols(const ObjectFile *Obj, 824 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 825 assert(Obj->isELF()); 826 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 827 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 828 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 829 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 830 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 831 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 832 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 833 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 834 else 835 llvm_unreachable("Unsupported binary format"); 836 } 837 838 static void addPltEntries(const ObjectFile *Obj, 839 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 840 StringSaver &Saver) { 841 Optional<SectionRef> Plt = None; 842 for (const SectionRef &Section : Obj->sections()) { 843 StringRef Name; 844 if (Section.getName(Name)) 845 continue; 846 if (Name == ".plt") 847 Plt = Section; 848 } 849 if (!Plt) 850 return; 851 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 852 for (auto PltEntry : ElfObj->getPltAddresses()) { 853 SymbolRef Symbol(PltEntry.first, ElfObj); 854 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 855 856 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 857 if (!Name.empty()) 858 AllSymbols[*Plt].emplace_back( 859 PltEntry.second, Saver.save((Name + "@plt").str()), SymbolType); 860 } 861 } 862 } 863 864 // Normally the disassembly output will skip blocks of zeroes. This function 865 // returns the number of zero bytes that can be skipped when dumping the 866 // disassembly of the instructions in Buf. 867 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 868 // Find the number of leading zeroes. 869 size_t N = 0; 870 while (N < Buf.size() && !Buf[N]) 871 ++N; 872 873 // We may want to skip blocks of zero bytes, but unless we see 874 // at least 8 of them in a row. 875 if (N < 8) 876 return 0; 877 878 // We skip zeroes in multiples of 4 because do not want to truncate an 879 // instruction if it starts with a zero byte. 880 return N & ~0x3; 881 } 882 883 // Returns a map from sections to their relocations. 884 static std::map<SectionRef, std::vector<RelocationRef>> 885 getRelocsMap(llvm::object::ObjectFile const &Obj) { 886 std::map<SectionRef, std::vector<RelocationRef>> Ret; 887 for (const SectionRef &Section : ToolSectionFilter(Obj)) { 888 section_iterator RelSec = Section.getRelocatedSection(); 889 if (RelSec == Obj.section_end()) 890 continue; 891 std::vector<RelocationRef> &V = Ret[*RelSec]; 892 for (const RelocationRef &R : Section.relocations()) 893 V.push_back(R); 894 // Sort relocations by address. 895 llvm::sort(V, isRelocAddressLess); 896 } 897 return Ret; 898 } 899 900 // Used for --adjust-vma to check if address should be adjusted by the 901 // specified value for a given section. 902 // For ELF we do not adjust non-allocatable sections like debug ones, 903 // because they are not loadable. 904 // TODO: implement for other file formats. 905 static bool shouldAdjustVA(const SectionRef &Section) { 906 const ObjectFile *Obj = Section.getObject(); 907 if (isa<object::ELFObjectFileBase>(Obj)) 908 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 909 return false; 910 } 911 912 static uint64_t 913 dumpARMELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 914 const ObjectFile *Obj, ArrayRef<uint8_t> Bytes, 915 const std::vector<uint64_t> &TextMappingSymsAddr) { 916 support::endianness Endian = 917 Obj->isLittleEndian() ? support::little : support::big; 918 while (Index < End) { 919 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 920 outs() << "\t"; 921 if (Index + 4 <= End) { 922 dumpBytes(Bytes.slice(Index, 4), outs()); 923 outs() << "\t.word\t" 924 << format_hex( 925 support::endian::read32(Bytes.data() + Index, Endian), 10); 926 Index += 4; 927 } else if (Index + 2 <= End) { 928 dumpBytes(Bytes.slice(Index, 2), outs()); 929 outs() << "\t\t.short\t" 930 << format_hex( 931 support::endian::read16(Bytes.data() + Index, Endian), 6); 932 Index += 2; 933 } else { 934 dumpBytes(Bytes.slice(Index, 1), outs()); 935 outs() << "\t\t.byte\t" << format_hex(Bytes[0], 4); 936 ++Index; 937 } 938 outs() << "\n"; 939 if (std::binary_search(TextMappingSymsAddr.begin(), 940 TextMappingSymsAddr.end(), Index)) 941 break; 942 } 943 return Index; 944 } 945 946 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 947 ArrayRef<uint8_t> Bytes) { 948 // print out data up to 8 bytes at a time in hex and ascii 949 uint8_t AsciiData[9] = {'\0'}; 950 uint8_t Byte; 951 int NumBytes = 0; 952 953 for (; Index < End; ++Index) { 954 if (NumBytes == 0) { 955 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 956 outs() << "\t"; 957 } 958 Byte = Bytes.slice(Index)[0]; 959 outs() << format(" %02x", Byte); 960 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 961 962 uint8_t IndentOffset = 0; 963 NumBytes++; 964 if (Index == End - 1 || NumBytes > 8) { 965 // Indent the space for less than 8 bytes data. 966 // 2 spaces for byte and one for space between bytes 967 IndentOffset = 3 * (8 - NumBytes); 968 for (int Excess = NumBytes; Excess < 8; Excess++) 969 AsciiData[Excess] = '\0'; 970 NumBytes = 8; 971 } 972 if (NumBytes == 8) { 973 AsciiData[8] = '\0'; 974 outs() << std::string(IndentOffset, ' ') << " "; 975 outs() << reinterpret_cast<char *>(AsciiData); 976 outs() << '\n'; 977 NumBytes = 0; 978 } 979 } 980 } 981 982 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 983 MCContext &Ctx, MCDisassembler *DisAsm, 984 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 985 const MCSubtargetInfo *STI, PrettyPrinter &PIP, 986 SourcePrinter &SP, bool InlineRelocs) { 987 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 988 if (InlineRelocs) 989 RelocMap = getRelocsMap(*Obj); 990 991 // Create a mapping from virtual address to symbol name. This is used to 992 // pretty print the symbols while disassembling. 993 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 994 SectionSymbolsTy AbsoluteSymbols; 995 const StringRef FileName = Obj->getFileName(); 996 for (const SymbolRef &Symbol : Obj->symbols()) { 997 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName); 998 999 StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1000 if (Name.empty()) 1001 continue; 1002 1003 uint8_t SymbolType = ELF::STT_NOTYPE; 1004 if (Obj->isELF()) { 1005 SymbolType = getElfSymbolType(Obj, Symbol); 1006 if (SymbolType == ELF::STT_SECTION) 1007 continue; 1008 } 1009 1010 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1011 if (SecI != Obj->section_end()) 1012 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 1013 else 1014 AbsoluteSymbols.emplace_back(Address, Name, SymbolType); 1015 } 1016 if (AllSymbols.empty() && Obj->isELF()) 1017 addDynamicElfSymbols(Obj, AllSymbols); 1018 1019 BumpPtrAllocator A; 1020 StringSaver Saver(A); 1021 addPltEntries(Obj, AllSymbols, Saver); 1022 1023 // Create a mapping from virtual address to section. 1024 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1025 for (SectionRef Sec : Obj->sections()) 1026 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1027 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1028 1029 // Linked executables (.exe and .dll files) typically don't include a real 1030 // symbol table but they might contain an export table. 1031 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1032 for (const auto &ExportEntry : COFFObj->export_directories()) { 1033 StringRef Name; 1034 error(ExportEntry.getSymbolName(Name)); 1035 if (Name.empty()) 1036 continue; 1037 uint32_t RVA; 1038 error(ExportEntry.getExportRVA(RVA)); 1039 1040 uint64_t VA = COFFObj->getImageBase() + RVA; 1041 auto Sec = llvm::upper_bound( 1042 SectionAddresses, VA, 1043 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1044 return LHS < RHS.first; 1045 }); 1046 if (Sec != SectionAddresses.begin()) { 1047 --Sec; 1048 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1049 } else 1050 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1051 } 1052 } 1053 1054 // Sort all the symbols, this allows us to use a simple binary search to find 1055 // a symbol near an address. 1056 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1057 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1058 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1059 1060 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1061 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1062 continue; 1063 1064 uint64_t SectionAddr = Section.getAddress(); 1065 uint64_t SectSize = Section.getSize(); 1066 if (!SectSize) 1067 continue; 1068 1069 // Get the list of all the symbols in this section. 1070 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1071 std::vector<uint64_t> DataMappingSymsAddr; 1072 std::vector<uint64_t> TextMappingSymsAddr; 1073 if (isArmElf(Obj)) { 1074 for (const auto &Symb : Symbols) { 1075 uint64_t Address = std::get<0>(Symb); 1076 StringRef Name = std::get<1>(Symb); 1077 if (Name.startswith("$d")) 1078 DataMappingSymsAddr.push_back(Address - SectionAddr); 1079 if (Name.startswith("$x")) 1080 TextMappingSymsAddr.push_back(Address - SectionAddr); 1081 if (Name.startswith("$a")) 1082 TextMappingSymsAddr.push_back(Address - SectionAddr); 1083 if (Name.startswith("$t")) 1084 TextMappingSymsAddr.push_back(Address - SectionAddr); 1085 } 1086 } 1087 1088 llvm::sort(DataMappingSymsAddr); 1089 llvm::sort(TextMappingSymsAddr); 1090 1091 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1092 // AMDGPU disassembler uses symbolizer for printing labels 1093 std::unique_ptr<MCRelocationInfo> RelInfo( 1094 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1095 if (RelInfo) { 1096 std::unique_ptr<MCSymbolizer> Symbolizer( 1097 TheTarget->createMCSymbolizer( 1098 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1099 DisAsm->setSymbolizer(std::move(Symbolizer)); 1100 } 1101 } 1102 1103 StringRef SegmentName = ""; 1104 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1105 DataRefImpl DR = Section.getRawDataRefImpl(); 1106 SegmentName = MachO->getSectionFinalSegmentName(DR); 1107 } 1108 StringRef SectionName; 1109 error(Section.getName(SectionName)); 1110 1111 // If the section has no symbol at the start, just insert a dummy one. 1112 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1113 Symbols.insert( 1114 Symbols.begin(), 1115 std::make_tuple(SectionAddr, SectionName, 1116 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1117 } 1118 1119 SmallString<40> Comments; 1120 raw_svector_ostream CommentStream(Comments); 1121 1122 StringRef BytesStr; 1123 error(Section.getContents(BytesStr)); 1124 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef(BytesStr); 1125 1126 uint64_t VMAAdjustment = 0; 1127 if (shouldAdjustVA(Section)) 1128 VMAAdjustment = AdjustVMA; 1129 1130 uint64_t Size; 1131 uint64_t Index; 1132 bool PrintedSection = false; 1133 std::vector<RelocationRef> Rels = RelocMap[Section]; 1134 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1135 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1136 // Disassemble symbol by symbol. 1137 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1138 uint64_t Start = std::get<0>(Symbols[SI]) - SectionAddr; 1139 // The end is either the section end or the beginning of the next 1140 // symbol. 1141 uint64_t End = (SI == SE - 1) 1142 ? SectSize 1143 : std::get<0>(Symbols[SI + 1]) - SectionAddr; 1144 // Don't try to disassemble beyond the end of section contents. 1145 if (End > SectSize) 1146 End = SectSize; 1147 // If this symbol has the same address as the next symbol, then skip it. 1148 if (Start >= End) 1149 continue; 1150 1151 // Check if we need to skip symbol 1152 // Skip if the symbol's data is not between StartAddress and StopAddress 1153 if (End + SectionAddr <= StartAddress || 1154 Start + SectionAddr >= StopAddress) 1155 continue; 1156 1157 // Stop disassembly at the stop address specified 1158 if (End + SectionAddr > StopAddress) 1159 End = StopAddress - SectionAddr; 1160 1161 /// Skip if user requested specific symbols and this is not in the list 1162 if (!DisasmFuncsSet.empty() && 1163 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1164 continue; 1165 1166 if (!PrintedSection) { 1167 PrintedSection = true; 1168 outs() << "Disassembly of section "; 1169 if (!SegmentName.empty()) 1170 outs() << SegmentName << ","; 1171 outs() << SectionName << ':'; 1172 } 1173 1174 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1175 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1176 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1177 Start += 256; 1178 } 1179 if (SI == SE - 1 || 1180 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1181 // cut trailing zeroes at the end of kernel 1182 // cut up to 256 bytes 1183 const uint64_t EndAlign = 256; 1184 const auto Limit = End - (std::min)(EndAlign, End - Start); 1185 while (End > Limit && 1186 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1187 End -= 4; 1188 } 1189 } 1190 1191 outs() << '\n'; 1192 if (!NoLeadingAddr) 1193 outs() << format("%016" PRIx64 " ", 1194 SectionAddr + Start + VMAAdjustment); 1195 1196 StringRef SymbolName = std::get<1>(Symbols[SI]); 1197 if (Demangle) 1198 outs() << demangle(SymbolName) << ":\n"; 1199 else 1200 outs() << SymbolName << ":\n"; 1201 1202 // Don't print raw contents of a virtual section. A virtual section 1203 // doesn't have any contents in the file. 1204 if (Section.isVirtual()) { 1205 outs() << "...\n"; 1206 continue; 1207 } 1208 1209 #ifndef NDEBUG 1210 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1211 #else 1212 raw_ostream &DebugOut = nulls(); 1213 #endif 1214 1215 // Some targets (like WebAssembly) have a special prelude at the start 1216 // of each symbol. 1217 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1218 SectionAddr + Start, DebugOut, CommentStream); 1219 Start += Size; 1220 1221 Index = Start; 1222 if (SectionAddr < StartAddress) 1223 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1224 1225 // If there is a data symbol inside an ELF text section and we are 1226 // only disassembling text (applicable all architectures), we are in a 1227 // situation where we must print the data and not disassemble it. 1228 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1229 !DisassembleAll && Section.isText()) { 1230 dumpELFData(SectionAddr, Index, End, Bytes); 1231 Index = End; 1232 } 1233 1234 bool CheckARMELFData = isArmElf(Obj) && 1235 std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1236 !DisassembleAll; 1237 while (Index < End) { 1238 // AArch64 ELF binaries can interleave data and text in the same 1239 // section. We rely on the markers introduced to understand what we 1240 // need to dump. If the data marker is within a function, it is 1241 // denoted as a word/short etc. 1242 if (CheckARMELFData && 1243 std::binary_search(DataMappingSymsAddr.begin(), 1244 DataMappingSymsAddr.end(), Index)) { 1245 Index = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1246 TextMappingSymsAddr); 1247 continue; 1248 } 1249 1250 // When -z or --disassemble-zeroes are given we always dissasemble 1251 // them. Otherwise we might want to skip zero bytes we see. 1252 if (!DisassembleZeroes) { 1253 uint64_t MaxOffset = End - Index; 1254 // For -reloc: print zero blocks patched by relocations, so that 1255 // relocations can be shown in the dump. 1256 if (RelCur != RelEnd) 1257 MaxOffset = RelCur->getOffset() - Index; 1258 1259 if (size_t N = 1260 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1261 outs() << "\t\t..." << '\n'; 1262 Index += N; 1263 continue; 1264 } 1265 } 1266 1267 // Disassemble a real instruction or a data when disassemble all is 1268 // provided 1269 MCInst Inst; 1270 bool Disassembled = DisAsm->getInstruction( 1271 Inst, Size, Bytes.slice(Index), SectionAddr + Index, DebugOut, 1272 CommentStream); 1273 if (Size == 0) 1274 Size = 1; 1275 1276 PIP.printInst( 1277 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 1278 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, outs(), 1279 "", *STI, &SP, &Rels); 1280 outs() << CommentStream.str(); 1281 Comments.clear(); 1282 1283 // Try to resolve the target of a call, tail call, etc. to a specific 1284 // symbol. 1285 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1286 MIA->isConditionalBranch(Inst))) { 1287 uint64_t Target; 1288 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1289 // In a relocatable object, the target's section must reside in 1290 // the same section as the call instruction or it is accessed 1291 // through a relocation. 1292 // 1293 // In a non-relocatable object, the target may be in any section. 1294 // 1295 // N.B. We don't walk the relocations in the relocatable case yet. 1296 auto *TargetSectionSymbols = &Symbols; 1297 if (!Obj->isRelocatableObject()) { 1298 auto SectionAddress = llvm::upper_bound( 1299 SectionAddresses, Target, 1300 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1301 return LHS < RHS.first; 1302 }); 1303 if (SectionAddress != SectionAddresses.begin()) { 1304 --SectionAddress; 1305 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1306 } else { 1307 TargetSectionSymbols = &AbsoluteSymbols; 1308 } 1309 } 1310 1311 // Find the first symbol in the section whose offset is less than 1312 // or equal to the target. If there isn't a section that contains 1313 // the target, find the nearest preceding absolute symbol. 1314 auto TargetSym = llvm::upper_bound( 1315 *TargetSectionSymbols, Target, 1316 [](uint64_t LHS, 1317 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1318 return LHS < std::get<0>(RHS); 1319 }); 1320 if (TargetSym == TargetSectionSymbols->begin()) { 1321 TargetSectionSymbols = &AbsoluteSymbols; 1322 TargetSym = llvm::upper_bound( 1323 AbsoluteSymbols, Target, 1324 [](uint64_t LHS, 1325 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1326 return LHS < std::get<0>(RHS); 1327 }); 1328 } 1329 if (TargetSym != TargetSectionSymbols->begin()) { 1330 --TargetSym; 1331 uint64_t TargetAddress = std::get<0>(*TargetSym); 1332 StringRef TargetName = std::get<1>(*TargetSym); 1333 outs() << " <" << TargetName; 1334 uint64_t Disp = Target - TargetAddress; 1335 if (Disp) 1336 outs() << "+0x" << Twine::utohexstr(Disp); 1337 outs() << '>'; 1338 } 1339 } 1340 } 1341 outs() << "\n"; 1342 1343 // Hexagon does this in pretty printer 1344 if (Obj->getArch() != Triple::hexagon) { 1345 // Print relocation for instruction. 1346 while (RelCur != RelEnd) { 1347 uint64_t Offset = RelCur->getOffset(); 1348 // If this relocation is hidden, skip it. 1349 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 1350 ++RelCur; 1351 continue; 1352 } 1353 1354 // Stop when RelCur's offset is past the current instruction. 1355 if (Offset >= Index + Size) 1356 break; 1357 1358 // When --adjust-vma is used, update the address printed. 1359 if (RelCur->getSymbol() != Obj->symbol_end()) { 1360 Expected<section_iterator> SymSI = 1361 RelCur->getSymbol()->getSection(); 1362 if (SymSI && *SymSI != Obj->section_end() && 1363 shouldAdjustVA(**SymSI)) 1364 Offset += AdjustVMA; 1365 } 1366 1367 printRelocation(*RelCur, SectionAddr + Offset, 1368 Obj->getBytesInAddress()); 1369 ++RelCur; 1370 } 1371 } 1372 1373 Index += Size; 1374 } 1375 } 1376 } 1377 } 1378 1379 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1380 if (StartAddress > StopAddress) 1381 error("Start address should be less than stop address"); 1382 1383 const Target *TheTarget = getTarget(Obj); 1384 1385 // Package up features to be passed to target/subtarget 1386 SubtargetFeatures Features = Obj->getFeatures(); 1387 if (!MAttrs.empty()) 1388 for (unsigned I = 0; I != MAttrs.size(); ++I) 1389 Features.AddFeature(MAttrs[I]); 1390 1391 std::unique_ptr<const MCRegisterInfo> MRI( 1392 TheTarget->createMCRegInfo(TripleName)); 1393 if (!MRI) 1394 report_error(Obj->getFileName(), 1395 "no register info for target " + TripleName); 1396 1397 // Set up disassembler. 1398 std::unique_ptr<const MCAsmInfo> AsmInfo( 1399 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1400 if (!AsmInfo) 1401 report_error(Obj->getFileName(), 1402 "no assembly info for target " + TripleName); 1403 std::unique_ptr<const MCSubtargetInfo> STI( 1404 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1405 if (!STI) 1406 report_error(Obj->getFileName(), 1407 "no subtarget info for target " + TripleName); 1408 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1409 if (!MII) 1410 report_error(Obj->getFileName(), 1411 "no instruction info for target " + TripleName); 1412 MCObjectFileInfo MOFI; 1413 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1414 // FIXME: for now initialize MCObjectFileInfo with default values 1415 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1416 1417 std::unique_ptr<MCDisassembler> DisAsm( 1418 TheTarget->createMCDisassembler(*STI, Ctx)); 1419 if (!DisAsm) 1420 report_error(Obj->getFileName(), 1421 "no disassembler for target " + TripleName); 1422 1423 std::unique_ptr<const MCInstrAnalysis> MIA( 1424 TheTarget->createMCInstrAnalysis(MII.get())); 1425 1426 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1427 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1428 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1429 if (!IP) 1430 report_error(Obj->getFileName(), 1431 "no instruction printer for target " + TripleName); 1432 IP->setPrintImmHex(PrintImmHex); 1433 1434 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1435 SourcePrinter SP(Obj, TheTarget->getName()); 1436 1437 for (StringRef Opt : DisassemblerOptions) 1438 if (!IP->applyTargetSpecificCLOption(Opt)) 1439 error("Unrecognized disassembler option: " + Opt); 1440 1441 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), MIA.get(), IP.get(), 1442 STI.get(), PIP, SP, InlineRelocs); 1443 } 1444 1445 void llvm::printRelocations(const ObjectFile *Obj) { 1446 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1447 "%08" PRIx64; 1448 // Regular objdump doesn't print relocations in non-relocatable object 1449 // files. 1450 if (!Obj->isRelocatableObject()) 1451 return; 1452 1453 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1454 if (Section.relocation_begin() == Section.relocation_end()) 1455 continue; 1456 StringRef SecName; 1457 error(Section.getName(SecName)); 1458 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1459 for (const RelocationRef &Reloc : Section.relocations()) { 1460 uint64_t Address = Reloc.getOffset(); 1461 SmallString<32> RelocName; 1462 SmallString<32> ValueStr; 1463 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1464 continue; 1465 Reloc.getTypeName(RelocName); 1466 error(getRelocationValueString(Reloc, ValueStr)); 1467 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1468 << ValueStr << "\n"; 1469 } 1470 outs() << "\n"; 1471 } 1472 } 1473 1474 void llvm::printDynamicRelocations(const ObjectFile *Obj) { 1475 // For the moment, this option is for ELF only 1476 if (!Obj->isELF()) 1477 return; 1478 1479 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1480 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1481 error("not a dynamic object"); 1482 return; 1483 } 1484 1485 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1486 if (DynRelSec.empty()) 1487 return; 1488 1489 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1490 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1491 for (const SectionRef &Section : DynRelSec) { 1492 if (Section.relocation_begin() == Section.relocation_end()) 1493 continue; 1494 for (const RelocationRef &Reloc : Section.relocations()) { 1495 uint64_t Address = Reloc.getOffset(); 1496 SmallString<32> RelocName; 1497 SmallString<32> ValueStr; 1498 Reloc.getTypeName(RelocName); 1499 error(getRelocationValueString(Reloc, ValueStr)); 1500 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1501 << ValueStr << "\n"; 1502 } 1503 } 1504 } 1505 1506 // Returns true if we need to show LMA column when dumping section headers. We 1507 // show it only when the platform is ELF and either we have at least one section 1508 // whose VMA and LMA are different and/or when --show-lma flag is used. 1509 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1510 if (!Obj->isELF()) 1511 return false; 1512 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1513 if (S.getAddress() != getELFSectionLMA(S)) 1514 return true; 1515 return ShowLMA; 1516 } 1517 1518 void llvm::printSectionHeaders(const ObjectFile *Obj) { 1519 bool HasLMAColumn = shouldDisplayLMA(Obj); 1520 if (HasLMAColumn) 1521 outs() << "Sections:\n" 1522 "Idx Name Size VMA LMA " 1523 "Type\n"; 1524 else 1525 outs() << "Sections:\n" 1526 "Idx Name Size VMA Type\n"; 1527 1528 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1529 StringRef Name; 1530 error(Section.getName(Name)); 1531 uint64_t VMA = Section.getAddress(); 1532 if (shouldAdjustVA(Section)) 1533 VMA += AdjustVMA; 1534 1535 uint64_t Size = Section.getSize(); 1536 bool Text = Section.isText(); 1537 bool Data = Section.isData(); 1538 bool BSS = Section.isBSS(); 1539 std::string Type = (std::string(Text ? "TEXT " : "") + 1540 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1541 1542 if (HasLMAColumn) 1543 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %016" PRIx64 1544 " %s\n", 1545 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1546 VMA, getELFSectionLMA(Section), Type.c_str()); 1547 else 1548 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1549 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1550 VMA, Type.c_str()); 1551 } 1552 outs() << "\n"; 1553 } 1554 1555 void llvm::printSectionContents(const ObjectFile *Obj) { 1556 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1557 StringRef Name; 1558 StringRef Contents; 1559 error(Section.getName(Name)); 1560 uint64_t BaseAddr = Section.getAddress(); 1561 uint64_t Size = Section.getSize(); 1562 if (!Size) 1563 continue; 1564 1565 outs() << "Contents of section " << Name << ":\n"; 1566 if (Section.isBSS()) { 1567 outs() << format("<skipping contents of bss section at [%04" PRIx64 1568 ", %04" PRIx64 ")>\n", 1569 BaseAddr, BaseAddr + Size); 1570 continue; 1571 } 1572 1573 error(Section.getContents(Contents)); 1574 1575 // Dump out the content as hex and printable ascii characters. 1576 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1577 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1578 // Dump line of hex. 1579 for (std::size_t I = 0; I < 16; ++I) { 1580 if (I != 0 && I % 4 == 0) 1581 outs() << ' '; 1582 if (Addr + I < End) 1583 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1584 << hexdigit(Contents[Addr + I] & 0xF, true); 1585 else 1586 outs() << " "; 1587 } 1588 // Print ascii. 1589 outs() << " "; 1590 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1591 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1592 outs() << Contents[Addr + I]; 1593 else 1594 outs() << "."; 1595 } 1596 outs() << "\n"; 1597 } 1598 } 1599 } 1600 1601 void llvm::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1602 StringRef ArchitectureName) { 1603 outs() << "SYMBOL TABLE:\n"; 1604 1605 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1606 printCOFFSymbolTable(Coff); 1607 return; 1608 } 1609 1610 const StringRef FileName = O->getFileName(); 1611 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1612 // Skip printing the special zero symbol when dumping an ELF file. 1613 // This makes the output consistent with the GNU objdump. 1614 if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O)) 1615 continue; 1616 1617 const SymbolRef &Symbol = *I; 1618 uint64_t Address = unwrapOrError(Symbol.getAddress(), ArchiveName, FileName, 1619 ArchitectureName); 1620 if ((Address < StartAddress) || (Address > StopAddress)) 1621 continue; 1622 SymbolRef::Type Type = unwrapOrError(Symbol.getType(), ArchiveName, 1623 FileName, ArchitectureName); 1624 uint32_t Flags = Symbol.getFlags(); 1625 section_iterator Section = unwrapOrError(Symbol.getSection(), ArchiveName, 1626 FileName, ArchitectureName); 1627 StringRef Name; 1628 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) 1629 Section->getName(Name); 1630 else 1631 Name = unwrapOrError(Symbol.getName(), ArchiveName, FileName, 1632 ArchitectureName); 1633 1634 bool Global = Flags & SymbolRef::SF_Global; 1635 bool Weak = Flags & SymbolRef::SF_Weak; 1636 bool Absolute = Flags & SymbolRef::SF_Absolute; 1637 bool Common = Flags & SymbolRef::SF_Common; 1638 bool Hidden = Flags & SymbolRef::SF_Hidden; 1639 1640 char GlobLoc = ' '; 1641 if (Type != SymbolRef::ST_Unknown) 1642 GlobLoc = Global ? 'g' : 'l'; 1643 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1644 ? 'd' : ' '; 1645 char FileFunc = ' '; 1646 if (Type == SymbolRef::ST_File) 1647 FileFunc = 'f'; 1648 else if (Type == SymbolRef::ST_Function) 1649 FileFunc = 'F'; 1650 else if (Type == SymbolRef::ST_Data) 1651 FileFunc = 'O'; 1652 1653 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1654 "%08" PRIx64; 1655 1656 outs() << format(Fmt, Address) << " " 1657 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1658 << (Weak ? 'w' : ' ') // Weak? 1659 << ' ' // Constructor. Not supported yet. 1660 << ' ' // Warning. Not supported yet. 1661 << ' ' // Indirect reference to another symbol. 1662 << Debug // Debugging (d) or dynamic (D) symbol. 1663 << FileFunc // Name of function (F), file (f) or object (O). 1664 << ' '; 1665 if (Absolute) { 1666 outs() << "*ABS*"; 1667 } else if (Common) { 1668 outs() << "*COM*"; 1669 } else if (Section == O->section_end()) { 1670 outs() << "*UND*"; 1671 } else { 1672 if (const MachOObjectFile *MachO = 1673 dyn_cast<const MachOObjectFile>(O)) { 1674 DataRefImpl DR = Section->getRawDataRefImpl(); 1675 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1676 outs() << SegmentName << ","; 1677 } 1678 StringRef SectionName; 1679 error(Section->getName(SectionName)); 1680 outs() << SectionName; 1681 } 1682 1683 outs() << '\t'; 1684 if (Common || isa<ELFObjectFileBase>(O)) { 1685 uint64_t Val = 1686 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1687 outs() << format("\t %08" PRIx64 " ", Val); 1688 } 1689 1690 if (Hidden) 1691 outs() << ".hidden "; 1692 1693 if (Demangle) 1694 outs() << demangle(Name) << '\n'; 1695 else 1696 outs() << Name << '\n'; 1697 } 1698 } 1699 1700 static void printUnwindInfo(const ObjectFile *O) { 1701 outs() << "Unwind info:\n\n"; 1702 1703 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1704 printCOFFUnwindInfo(Coff); 1705 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1706 printMachOUnwindInfo(MachO); 1707 else 1708 // TODO: Extract DWARF dump tool to objdump. 1709 WithColor::error(errs(), ToolName) 1710 << "This operation is only currently supported " 1711 "for COFF and MachO object files.\n"; 1712 } 1713 1714 void llvm::printExportsTrie(const ObjectFile *o) { 1715 outs() << "Exports trie:\n"; 1716 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1717 printMachOExportsTrie(MachO); 1718 else 1719 WithColor::error(errs(), ToolName) 1720 << "This operation is only currently supported " 1721 "for Mach-O executable files.\n"; 1722 } 1723 1724 void llvm::printRebaseTable(ObjectFile *o) { 1725 outs() << "Rebase table:\n"; 1726 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1727 printMachORebaseTable(MachO); 1728 else 1729 WithColor::error(errs(), ToolName) 1730 << "This operation is only currently supported " 1731 "for Mach-O executable files.\n"; 1732 } 1733 1734 void llvm::printBindTable(ObjectFile *o) { 1735 outs() << "Bind table:\n"; 1736 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1737 printMachOBindTable(MachO); 1738 else 1739 WithColor::error(errs(), ToolName) 1740 << "This operation is only currently supported " 1741 "for Mach-O executable files.\n"; 1742 } 1743 1744 void llvm::printLazyBindTable(ObjectFile *o) { 1745 outs() << "Lazy bind table:\n"; 1746 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1747 printMachOLazyBindTable(MachO); 1748 else 1749 WithColor::error(errs(), ToolName) 1750 << "This operation is only currently supported " 1751 "for Mach-O executable files.\n"; 1752 } 1753 1754 void llvm::printWeakBindTable(ObjectFile *o) { 1755 outs() << "Weak bind table:\n"; 1756 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1757 printMachOWeakBindTable(MachO); 1758 else 1759 WithColor::error(errs(), ToolName) 1760 << "This operation is only currently supported " 1761 "for Mach-O executable files.\n"; 1762 } 1763 1764 /// Dump the raw contents of the __clangast section so the output can be piped 1765 /// into llvm-bcanalyzer. 1766 void llvm::printRawClangAST(const ObjectFile *Obj) { 1767 if (outs().is_displayed()) { 1768 WithColor::error(errs(), ToolName) 1769 << "The -raw-clang-ast option will dump the raw binary contents of " 1770 "the clang ast section.\n" 1771 "Please redirect the output to a file or another program such as " 1772 "llvm-bcanalyzer.\n"; 1773 return; 1774 } 1775 1776 StringRef ClangASTSectionName("__clangast"); 1777 if (isa<COFFObjectFile>(Obj)) { 1778 ClangASTSectionName = "clangast"; 1779 } 1780 1781 Optional<object::SectionRef> ClangASTSection; 1782 for (auto Sec : ToolSectionFilter(*Obj)) { 1783 StringRef Name; 1784 Sec.getName(Name); 1785 if (Name == ClangASTSectionName) { 1786 ClangASTSection = Sec; 1787 break; 1788 } 1789 } 1790 if (!ClangASTSection) 1791 return; 1792 1793 StringRef ClangASTContents; 1794 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1795 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1796 } 1797 1798 static void printFaultMaps(const ObjectFile *Obj) { 1799 StringRef FaultMapSectionName; 1800 1801 if (isa<ELFObjectFileBase>(Obj)) { 1802 FaultMapSectionName = ".llvm_faultmaps"; 1803 } else if (isa<MachOObjectFile>(Obj)) { 1804 FaultMapSectionName = "__llvm_faultmaps"; 1805 } else { 1806 WithColor::error(errs(), ToolName) 1807 << "This operation is only currently supported " 1808 "for ELF and Mach-O executable files.\n"; 1809 return; 1810 } 1811 1812 Optional<object::SectionRef> FaultMapSection; 1813 1814 for (auto Sec : ToolSectionFilter(*Obj)) { 1815 StringRef Name; 1816 Sec.getName(Name); 1817 if (Name == FaultMapSectionName) { 1818 FaultMapSection = Sec; 1819 break; 1820 } 1821 } 1822 1823 outs() << "FaultMap table:\n"; 1824 1825 if (!FaultMapSection.hasValue()) { 1826 outs() << "<not found>\n"; 1827 return; 1828 } 1829 1830 StringRef FaultMapContents; 1831 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1832 1833 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1834 FaultMapContents.bytes_end()); 1835 1836 outs() << FMP; 1837 } 1838 1839 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1840 if (O->isELF()) { 1841 printELFFileHeader(O); 1842 printELFDynamicSection(O); 1843 printELFSymbolVersionInfo(O); 1844 return; 1845 } 1846 if (O->isCOFF()) 1847 return printCOFFFileHeader(O); 1848 if (O->isWasm()) 1849 return printWasmFileHeader(O); 1850 if (O->isMachO()) { 1851 printMachOFileHeader(O); 1852 if (!OnlyFirst) 1853 printMachOLoadCommands(O); 1854 return; 1855 } 1856 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1857 } 1858 1859 static void printFileHeaders(const ObjectFile *O) { 1860 if (!O->isELF() && !O->isCOFF()) 1861 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1862 1863 Triple::ArchType AT = O->getArch(); 1864 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1865 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 1866 1867 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1868 outs() << "start address: " 1869 << "0x" << format(Fmt.data(), Address) << "\n\n"; 1870 } 1871 1872 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 1873 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 1874 if (!ModeOrErr) { 1875 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 1876 consumeError(ModeOrErr.takeError()); 1877 return; 1878 } 1879 sys::fs::perms Mode = ModeOrErr.get(); 1880 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1881 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1882 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1883 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1884 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1885 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1886 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1887 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1888 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1889 1890 outs() << " "; 1891 1892 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 1893 unwrapOrError(C.getGID(), Filename), 1894 unwrapOrError(C.getRawSize(), Filename)); 1895 1896 StringRef RawLastModified = C.getRawLastModified(); 1897 unsigned Seconds; 1898 if (RawLastModified.getAsInteger(10, Seconds)) 1899 outs() << "(date: \"" << RawLastModified 1900 << "\" contains non-decimal chars) "; 1901 else { 1902 // Since ctime(3) returns a 26 character string of the form: 1903 // "Sun Sep 16 01:03:52 1973\n\0" 1904 // just print 24 characters. 1905 time_t t = Seconds; 1906 outs() << format("%.24s ", ctime(&t)); 1907 } 1908 1909 StringRef Name = ""; 1910 Expected<StringRef> NameOrErr = C.getName(); 1911 if (!NameOrErr) { 1912 consumeError(NameOrErr.takeError()); 1913 Name = unwrapOrError(C.getRawName(), Filename); 1914 } else { 1915 Name = NameOrErr.get(); 1916 } 1917 outs() << Name << "\n"; 1918 } 1919 1920 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 1921 const Archive::Child *C = nullptr) { 1922 // Avoid other output when using a raw option. 1923 if (!RawClangAST) { 1924 outs() << '\n'; 1925 if (A) 1926 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 1927 else 1928 outs() << O->getFileName(); 1929 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 1930 } 1931 1932 StringRef ArchiveName = A ? A->getFileName() : ""; 1933 if (FileHeaders) 1934 printFileHeaders(O); 1935 if (ArchiveHeaders && !MachOOpt && C) 1936 printArchiveChild(ArchiveName, *C); 1937 if (Disassemble) 1938 disassembleObject(O, Relocations); 1939 if (Relocations && !Disassemble) 1940 printRelocations(O); 1941 if (DynamicRelocations) 1942 printDynamicRelocations(O); 1943 if (SectionHeaders) 1944 printSectionHeaders(O); 1945 if (SectionContents) 1946 printSectionContents(O); 1947 if (SymbolTable) 1948 printSymbolTable(O, ArchiveName); 1949 if (UnwindInfo) 1950 printUnwindInfo(O); 1951 if (PrivateHeaders || FirstPrivateHeader) 1952 printPrivateFileHeaders(O, FirstPrivateHeader); 1953 if (ExportsTrie) 1954 printExportsTrie(O); 1955 if (Rebase) 1956 printRebaseTable(O); 1957 if (Bind) 1958 printBindTable(O); 1959 if (LazyBind) 1960 printLazyBindTable(O); 1961 if (WeakBind) 1962 printWeakBindTable(O); 1963 if (RawClangAST) 1964 printRawClangAST(O); 1965 if (PrintFaultMaps) 1966 printFaultMaps(O); 1967 if (DwarfDumpType != DIDT_Null) { 1968 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 1969 // Dump the complete DWARF structure. 1970 DIDumpOptions DumpOpts; 1971 DumpOpts.DumpType = DwarfDumpType; 1972 DICtx->dump(outs(), DumpOpts); 1973 } 1974 } 1975 1976 static void dumpObject(const COFFImportFile *I, const Archive *A, 1977 const Archive::Child *C = nullptr) { 1978 StringRef ArchiveName = A ? A->getFileName() : ""; 1979 1980 // Avoid other output when using a raw option. 1981 if (!RawClangAST) 1982 outs() << '\n' 1983 << ArchiveName << "(" << I->getFileName() << ")" 1984 << ":\tfile format COFF-import-file" 1985 << "\n\n"; 1986 1987 if (ArchiveHeaders && !MachOOpt && C) 1988 printArchiveChild(ArchiveName, *C); 1989 if (SymbolTable) 1990 printCOFFSymbolTable(I); 1991 } 1992 1993 /// Dump each object file in \a a; 1994 static void dumpArchive(const Archive *A) { 1995 Error Err = Error::success(); 1996 for (auto &C : A->children(Err)) { 1997 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1998 if (!ChildOrErr) { 1999 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2000 report_error(std::move(E), A->getFileName(), C); 2001 continue; 2002 } 2003 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2004 dumpObject(O, A, &C); 2005 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2006 dumpObject(I, A, &C); 2007 else 2008 report_error(errorCodeToError(object_error::invalid_file_type), 2009 A->getFileName()); 2010 } 2011 if (Err) 2012 report_error(std::move(Err), A->getFileName()); 2013 } 2014 2015 /// Open file and figure out how to dump it. 2016 static void dumpInput(StringRef file) { 2017 // If we are using the Mach-O specific object file parser, then let it parse 2018 // the file and process the command line options. So the -arch flags can 2019 // be used to select specific slices, etc. 2020 if (MachOOpt) { 2021 parseInputMachO(file); 2022 return; 2023 } 2024 2025 // Attempt to open the binary. 2026 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2027 Binary &Binary = *OBinary.getBinary(); 2028 2029 if (Archive *A = dyn_cast<Archive>(&Binary)) 2030 dumpArchive(A); 2031 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2032 dumpObject(O); 2033 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2034 parseInputMachO(UB); 2035 else 2036 report_error(errorCodeToError(object_error::invalid_file_type), file); 2037 } 2038 2039 int main(int argc, char **argv) { 2040 InitLLVM X(argc, argv); 2041 2042 // Initialize targets and assembly printers/parsers. 2043 llvm::InitializeAllTargetInfos(); 2044 llvm::InitializeAllTargetMCs(); 2045 llvm::InitializeAllDisassemblers(); 2046 2047 // Register the target printer for --version. 2048 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2049 2050 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2051 2052 ToolName = argv[0]; 2053 2054 // Defaults to a.out if no filenames specified. 2055 if (InputFilenames.empty()) 2056 InputFilenames.push_back("a.out"); 2057 2058 if (AllHeaders) 2059 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2060 SectionHeaders = SymbolTable = true; 2061 2062 if (DisassembleAll || PrintSource || PrintLines) 2063 Disassemble = true; 2064 2065 if (!Disassemble 2066 && !Relocations 2067 && !DynamicRelocations 2068 && !SectionHeaders 2069 && !SectionContents 2070 && !SymbolTable 2071 && !UnwindInfo 2072 && !PrivateHeaders 2073 && !FileHeaders 2074 && !FirstPrivateHeader 2075 && !ExportsTrie 2076 && !Rebase 2077 && !Bind 2078 && !LazyBind 2079 && !WeakBind 2080 && !RawClangAST 2081 && !(UniversalHeaders && MachOOpt) 2082 && !ArchiveHeaders 2083 && !(IndirectSymbols && MachOOpt) 2084 && !(DataInCode && MachOOpt) 2085 && !(LinkOptHints && MachOOpt) 2086 && !(InfoPlist && MachOOpt) 2087 && !(DylibsUsed && MachOOpt) 2088 && !(DylibId && MachOOpt) 2089 && !(ObjcMetaData && MachOOpt) 2090 && !(!FilterSections.empty() && MachOOpt) 2091 && !PrintFaultMaps 2092 && DwarfDumpType == DIDT_Null) { 2093 cl::PrintHelpMessage(); 2094 return 2; 2095 } 2096 2097 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2098 DisassembleFunctions.end()); 2099 2100 llvm::for_each(InputFilenames, dumpInput); 2101 2102 return EXIT_SUCCESS; 2103 } 2104