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 size_t BufferSize = Buffer->getBufferSize(); 526 const char *BufferStart = Buffer->getBufferStart(); 527 for (const char *Start = BufferStart, *End = BufferStart; 528 End < BufferStart + BufferSize; End++) 529 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 530 (*End == '\r' && *(End + 1) == '\n')) { 531 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 532 if (*End == '\r') 533 End++; 534 Start = End + 1; 535 } 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 FileBuffer = SourceCache.find(LineInfo.FileName); 564 if (FileBuffer != SourceCache.end()) { 565 auto LineBuffer = LineCache.find(LineInfo.FileName); 566 if (LineBuffer != LineCache.end()) { 567 if (LineInfo.Line > LineBuffer->second.size()) 568 return; 569 // Vector begins at 0, line numbers are non-zero 570 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 571 << "\n"; 572 } 573 } 574 } 575 OldLineInfo = LineInfo; 576 } 577 578 static bool isArmElf(const ObjectFile *Obj) { 579 return (Obj->isELF() && 580 (Obj->getArch() == Triple::aarch64 || 581 Obj->getArch() == Triple::aarch64_be || 582 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 583 Obj->getArch() == Triple::thumb || 584 Obj->getArch() == Triple::thumbeb)); 585 } 586 587 static void printRelocation(const RelocationRef &Rel, uint64_t Address, 588 uint8_t AddrSize) { 589 StringRef Fmt = 590 AddrSize > 4 ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 591 SmallString<16> Name; 592 SmallString<32> Val; 593 Rel.getTypeName(Name); 594 error(getRelocationValueString(Rel, Val)); 595 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n"; 596 } 597 598 class PrettyPrinter { 599 public: 600 virtual ~PrettyPrinter() = default; 601 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 602 ArrayRef<uint8_t> Bytes, 603 object::SectionedAddress Address, raw_ostream &OS, 604 StringRef Annot, MCSubtargetInfo const &STI, 605 SourcePrinter *SP, 606 std::vector<RelocationRef> *Rels = nullptr) { 607 if (SP && (PrintSource || PrintLines)) 608 SP->printSourceLine(OS, Address); 609 if (!NoLeadingAddr) 610 OS << format("%8" PRIx64 ":", Address.Address); 611 if (!NoShowRawInsn) { 612 OS << "\t"; 613 dumpBytes(Bytes, OS); 614 } 615 if (MI) 616 IP.printInst(MI, OS, "", STI); 617 else 618 OS << " <unknown>"; 619 } 620 }; 621 PrettyPrinter PrettyPrinterInst; 622 class HexagonPrettyPrinter : public PrettyPrinter { 623 public: 624 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 625 raw_ostream &OS) { 626 uint32_t opcode = 627 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 628 if (!NoLeadingAddr) 629 OS << format("%8" PRIx64 ":", Address); 630 if (!NoShowRawInsn) { 631 OS << "\t"; 632 dumpBytes(Bytes.slice(0, 4), OS); 633 OS << format("%08" PRIx32, opcode); 634 } 635 } 636 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 637 object::SectionedAddress Address, raw_ostream &OS, 638 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 639 std::vector<RelocationRef> *Rels) override { 640 if (SP && (PrintSource || PrintLines)) 641 SP->printSourceLine(OS, Address, ""); 642 if (!MI) { 643 printLead(Bytes, Address.Address, OS); 644 OS << " <unknown>"; 645 return; 646 } 647 std::string Buffer; 648 { 649 raw_string_ostream TempStream(Buffer); 650 IP.printInst(MI, TempStream, "", STI); 651 } 652 StringRef Contents(Buffer); 653 // Split off bundle attributes 654 auto PacketBundle = Contents.rsplit('\n'); 655 // Split off first instruction from the rest 656 auto HeadTail = PacketBundle.first.split('\n'); 657 auto Preamble = " { "; 658 auto Separator = ""; 659 660 // Hexagon's packets require relocations to be inline rather than 661 // clustered at the end of the packet. 662 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 663 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 664 auto PrintReloc = [&]() -> void { 665 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 666 if (RelCur->getOffset() == Address.Address) { 667 printRelocation(*RelCur, Address.Address, 4); 668 return; 669 } 670 ++RelCur; 671 } 672 }; 673 674 while (!HeadTail.first.empty()) { 675 OS << Separator; 676 Separator = "\n"; 677 if (SP && (PrintSource || PrintLines)) 678 SP->printSourceLine(OS, Address, ""); 679 printLead(Bytes, Address.Address, OS); 680 OS << Preamble; 681 Preamble = " "; 682 StringRef Inst; 683 auto Duplex = HeadTail.first.split('\v'); 684 if (!Duplex.second.empty()) { 685 OS << Duplex.first; 686 OS << "; "; 687 Inst = Duplex.second; 688 } 689 else 690 Inst = HeadTail.first; 691 OS << Inst; 692 HeadTail = HeadTail.second.split('\n'); 693 if (HeadTail.first.empty()) 694 OS << " } " << PacketBundle.second; 695 PrintReloc(); 696 Bytes = Bytes.slice(4); 697 Address.Address += 4; 698 } 699 } 700 }; 701 HexagonPrettyPrinter HexagonPrettyPrinterInst; 702 703 class AMDGCNPrettyPrinter : public PrettyPrinter { 704 public: 705 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 706 object::SectionedAddress Address, raw_ostream &OS, 707 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 708 std::vector<RelocationRef> *Rels) override { 709 if (SP && (PrintSource || PrintLines)) 710 SP->printSourceLine(OS, Address); 711 712 typedef support::ulittle32_t U32; 713 714 if (MI) { 715 SmallString<40> InstStr; 716 raw_svector_ostream IS(InstStr); 717 718 IP.printInst(MI, IS, "", STI); 719 720 OS << left_justify(IS.str(), 60); 721 } else { 722 // an unrecognized encoding - this is probably data so represent it 723 // using the .long directive, or .byte directive if fewer than 4 bytes 724 // remaining 725 if (Bytes.size() >= 4) { 726 OS << format("\t.long 0x%08" PRIx32 " ", 727 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 728 OS.indent(42); 729 } else { 730 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 731 for (unsigned int i = 1; i < Bytes.size(); i++) 732 OS << format(", 0x%02" PRIx8, Bytes[i]); 733 OS.indent(55 - (6 * Bytes.size())); 734 } 735 } 736 737 OS << format("// %012" PRIX64 ": ", Address.Address); 738 if (Bytes.size() >=4) { 739 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 740 Bytes.size() / sizeof(U32))) 741 // D should be explicitly casted to uint32_t here as it is passed 742 // by format to snprintf as vararg. 743 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 744 } else { 745 for (unsigned int i = 0; i < Bytes.size(); i++) 746 OS << format("%02" PRIX8 " ", Bytes[i]); 747 } 748 749 if (!Annot.empty()) 750 OS << "// " << Annot; 751 } 752 }; 753 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 754 755 class BPFPrettyPrinter : public PrettyPrinter { 756 public: 757 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 758 object::SectionedAddress Address, raw_ostream &OS, 759 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 760 std::vector<RelocationRef> *Rels) override { 761 if (SP && (PrintSource || PrintLines)) 762 SP->printSourceLine(OS, Address); 763 if (!NoLeadingAddr) 764 OS << format("%8" PRId64 ":", Address.Address / 8); 765 if (!NoShowRawInsn) { 766 OS << "\t"; 767 dumpBytes(Bytes, OS); 768 } 769 if (MI) 770 IP.printInst(MI, OS, "", STI); 771 else 772 OS << " <unknown>"; 773 } 774 }; 775 BPFPrettyPrinter BPFPrettyPrinterInst; 776 777 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 778 switch(Triple.getArch()) { 779 default: 780 return PrettyPrinterInst; 781 case Triple::hexagon: 782 return HexagonPrettyPrinterInst; 783 case Triple::amdgcn: 784 return AMDGCNPrettyPrinterInst; 785 case Triple::bpfel: 786 case Triple::bpfeb: 787 return BPFPrettyPrinterInst; 788 } 789 } 790 } 791 792 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 793 assert(Obj->isELF()); 794 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 795 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 796 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 797 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 798 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 799 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 800 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 801 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 802 llvm_unreachable("Unsupported binary format"); 803 } 804 805 template <class ELFT> static void 806 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 807 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 808 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 809 uint8_t SymbolType = Symbol.getELFType(); 810 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 811 continue; 812 813 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 814 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 815 if (Name.empty()) 816 continue; 817 818 section_iterator SecI = 819 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 820 if (SecI == Obj->section_end()) 821 continue; 822 823 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 824 } 825 } 826 827 static void 828 addDynamicElfSymbols(const ObjectFile *Obj, 829 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 830 assert(Obj->isELF()); 831 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 832 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 833 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 834 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 835 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 836 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 837 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 838 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 839 else 840 llvm_unreachable("Unsupported binary format"); 841 } 842 843 static void addPltEntries(const ObjectFile *Obj, 844 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 845 StringSaver &Saver) { 846 Optional<SectionRef> Plt = None; 847 for (const SectionRef &Section : Obj->sections()) { 848 StringRef Name; 849 if (Section.getName(Name)) 850 continue; 851 if (Name == ".plt") 852 Plt = Section; 853 } 854 if (!Plt) 855 return; 856 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 857 for (auto PltEntry : ElfObj->getPltAddresses()) { 858 SymbolRef Symbol(PltEntry.first, ElfObj); 859 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 860 861 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 862 if (!Name.empty()) 863 AllSymbols[*Plt].emplace_back( 864 PltEntry.second, Saver.save((Name + "@plt").str()), SymbolType); 865 } 866 } 867 } 868 869 // Normally the disassembly output will skip blocks of zeroes. This function 870 // returns the number of zero bytes that can be skipped when dumping the 871 // disassembly of the instructions in Buf. 872 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 873 // Find the number of leading zeroes. 874 size_t N = 0; 875 while (N < Buf.size() && !Buf[N]) 876 ++N; 877 878 // We may want to skip blocks of zero bytes, but unless we see 879 // at least 8 of them in a row. 880 if (N < 8) 881 return 0; 882 883 // We skip zeroes in multiples of 4 because do not want to truncate an 884 // instruction if it starts with a zero byte. 885 return N & ~0x3; 886 } 887 888 // Returns a map from sections to their relocations. 889 static std::map<SectionRef, std::vector<RelocationRef>> 890 getRelocsMap(llvm::object::ObjectFile const &Obj) { 891 std::map<SectionRef, std::vector<RelocationRef>> Ret; 892 for (const SectionRef &Section : ToolSectionFilter(Obj)) { 893 section_iterator RelSec = Section.getRelocatedSection(); 894 if (RelSec == Obj.section_end()) 895 continue; 896 std::vector<RelocationRef> &V = Ret[*RelSec]; 897 for (const RelocationRef &R : Section.relocations()) 898 V.push_back(R); 899 // Sort relocations by address. 900 llvm::sort(V, isRelocAddressLess); 901 } 902 return Ret; 903 } 904 905 // Used for --adjust-vma to check if address should be adjusted by the 906 // specified value for a given section. 907 // For ELF we do not adjust non-allocatable sections like debug ones, 908 // because they are not loadable. 909 // TODO: implement for other file formats. 910 static bool shouldAdjustVA(const SectionRef &Section) { 911 const ObjectFile *Obj = Section.getObject(); 912 if (isa<object::ELFObjectFileBase>(Obj)) 913 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 914 return false; 915 } 916 917 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 918 MCContext &Ctx, MCDisassembler *DisAsm, 919 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 920 const MCSubtargetInfo *STI, PrettyPrinter &PIP, 921 SourcePrinter &SP, bool InlineRelocs) { 922 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 923 if (InlineRelocs) 924 RelocMap = getRelocsMap(*Obj); 925 926 // Create a mapping from virtual address to symbol name. This is used to 927 // pretty print the symbols while disassembling. 928 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 929 SectionSymbolsTy AbsoluteSymbols; 930 const StringRef FileName = Obj->getFileName(); 931 for (const SymbolRef &Symbol : Obj->symbols()) { 932 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName); 933 934 StringRef Name = unwrapOrError(Symbol.getName(), FileName); 935 if (Name.empty()) 936 continue; 937 938 uint8_t SymbolType = ELF::STT_NOTYPE; 939 if (Obj->isELF()) { 940 SymbolType = getElfSymbolType(Obj, Symbol); 941 if (SymbolType == ELF::STT_SECTION) 942 continue; 943 } 944 945 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 946 if (SecI != Obj->section_end()) 947 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 948 else 949 AbsoluteSymbols.emplace_back(Address, Name, SymbolType); 950 } 951 if (AllSymbols.empty() && Obj->isELF()) 952 addDynamicElfSymbols(Obj, AllSymbols); 953 954 BumpPtrAllocator A; 955 StringSaver Saver(A); 956 addPltEntries(Obj, AllSymbols, Saver); 957 958 // Create a mapping from virtual address to section. 959 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 960 for (SectionRef Sec : Obj->sections()) 961 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 962 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 963 964 // Linked executables (.exe and .dll files) typically don't include a real 965 // symbol table but they might contain an export table. 966 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 967 for (const auto &ExportEntry : COFFObj->export_directories()) { 968 StringRef Name; 969 error(ExportEntry.getSymbolName(Name)); 970 if (Name.empty()) 971 continue; 972 uint32_t RVA; 973 error(ExportEntry.getExportRVA(RVA)); 974 975 uint64_t VA = COFFObj->getImageBase() + RVA; 976 auto Sec = llvm::upper_bound( 977 SectionAddresses, VA, 978 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 979 return LHS < RHS.first; 980 }); 981 if (Sec != SectionAddresses.begin()) { 982 --Sec; 983 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 984 } else 985 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 986 } 987 } 988 989 // Sort all the symbols, this allows us to use a simple binary search to find 990 // a symbol near an address. 991 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 992 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 993 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 994 995 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 996 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 997 continue; 998 999 uint64_t SectionAddr = Section.getAddress(); 1000 uint64_t SectSize = Section.getSize(); 1001 if (!SectSize) 1002 continue; 1003 1004 // Get the list of all the symbols in this section. 1005 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1006 std::vector<uint64_t> DataMappingSymsAddr; 1007 std::vector<uint64_t> TextMappingSymsAddr; 1008 if (isArmElf(Obj)) { 1009 for (const auto &Symb : Symbols) { 1010 uint64_t Address = std::get<0>(Symb); 1011 StringRef Name = std::get<1>(Symb); 1012 if (Name.startswith("$d")) 1013 DataMappingSymsAddr.push_back(Address - SectionAddr); 1014 if (Name.startswith("$x")) 1015 TextMappingSymsAddr.push_back(Address - SectionAddr); 1016 if (Name.startswith("$a")) 1017 TextMappingSymsAddr.push_back(Address - SectionAddr); 1018 if (Name.startswith("$t")) 1019 TextMappingSymsAddr.push_back(Address - SectionAddr); 1020 } 1021 } 1022 1023 llvm::sort(DataMappingSymsAddr); 1024 llvm::sort(TextMappingSymsAddr); 1025 1026 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1027 // AMDGPU disassembler uses symbolizer for printing labels 1028 std::unique_ptr<MCRelocationInfo> RelInfo( 1029 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1030 if (RelInfo) { 1031 std::unique_ptr<MCSymbolizer> Symbolizer( 1032 TheTarget->createMCSymbolizer( 1033 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1034 DisAsm->setSymbolizer(std::move(Symbolizer)); 1035 } 1036 } 1037 1038 StringRef SegmentName = ""; 1039 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1040 DataRefImpl DR = Section.getRawDataRefImpl(); 1041 SegmentName = MachO->getSectionFinalSegmentName(DR); 1042 } 1043 StringRef SectionName; 1044 error(Section.getName(SectionName)); 1045 1046 // If the section has no symbol at the start, just insert a dummy one. 1047 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1048 Symbols.insert( 1049 Symbols.begin(), 1050 std::make_tuple(SectionAddr, SectionName, 1051 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1052 } 1053 1054 SmallString<40> Comments; 1055 raw_svector_ostream CommentStream(Comments); 1056 1057 StringRef BytesStr; 1058 error(Section.getContents(BytesStr)); 1059 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef(BytesStr); 1060 1061 uint64_t VMAAdjustment = 0; 1062 if (shouldAdjustVA(Section)) 1063 VMAAdjustment = AdjustVMA; 1064 1065 uint64_t Size; 1066 uint64_t Index; 1067 bool PrintedSection = false; 1068 std::vector<RelocationRef> Rels = RelocMap[Section]; 1069 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1070 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1071 // Disassemble symbol by symbol. 1072 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1073 uint64_t Start = std::get<0>(Symbols[SI]) - SectionAddr; 1074 // The end is either the section end or the beginning of the next 1075 // symbol. 1076 uint64_t End = (SI == SE - 1) 1077 ? SectSize 1078 : std::get<0>(Symbols[SI + 1]) - SectionAddr; 1079 // Don't try to disassemble beyond the end of section contents. 1080 if (End > SectSize) 1081 End = SectSize; 1082 // If this symbol has the same address as the next symbol, then skip it. 1083 if (Start >= End) 1084 continue; 1085 1086 // Check if we need to skip symbol 1087 // Skip if the symbol's data is not between StartAddress and StopAddress 1088 if (End + SectionAddr < StartAddress || 1089 Start + SectionAddr > StopAddress) { 1090 continue; 1091 } 1092 1093 /// Skip if user requested specific symbols and this is not in the list 1094 if (!DisasmFuncsSet.empty() && 1095 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1096 continue; 1097 1098 if (!PrintedSection) { 1099 PrintedSection = true; 1100 outs() << "Disassembly of section "; 1101 if (!SegmentName.empty()) 1102 outs() << SegmentName << ","; 1103 outs() << SectionName << ':'; 1104 } 1105 1106 // Stop disassembly at the stop address specified 1107 if (End + SectionAddr > StopAddress) 1108 End = StopAddress - SectionAddr; 1109 1110 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1111 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1112 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1113 Start += 256; 1114 } 1115 if (SI == SE - 1 || 1116 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1117 // cut trailing zeroes at the end of kernel 1118 // cut up to 256 bytes 1119 const uint64_t EndAlign = 256; 1120 const auto Limit = End - (std::min)(EndAlign, End - Start); 1121 while (End > Limit && 1122 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1123 End -= 4; 1124 } 1125 } 1126 1127 outs() << '\n'; 1128 if (!NoLeadingAddr) 1129 outs() << format("%016" PRIx64 " ", 1130 SectionAddr + Start + VMAAdjustment); 1131 1132 StringRef SymbolName = std::get<1>(Symbols[SI]); 1133 if (Demangle) 1134 outs() << demangle(SymbolName) << ":\n"; 1135 else 1136 outs() << SymbolName << ":\n"; 1137 1138 // Don't print raw contents of a virtual section. A virtual section 1139 // doesn't have any contents in the file. 1140 if (Section.isVirtual()) { 1141 outs() << "...\n"; 1142 continue; 1143 } 1144 1145 #ifndef NDEBUG 1146 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1147 #else 1148 raw_ostream &DebugOut = nulls(); 1149 #endif 1150 1151 // Some targets (like WebAssembly) have a special prelude at the start 1152 // of each symbol. 1153 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1154 SectionAddr + Start, DebugOut, CommentStream); 1155 Start += Size; 1156 1157 for (Index = Start; Index < End; Index += Size) { 1158 MCInst Inst; 1159 1160 if (Index + SectionAddr < StartAddress || 1161 Index + SectionAddr > StopAddress) { 1162 // skip byte by byte till StartAddress is reached 1163 Size = 1; 1164 continue; 1165 } 1166 // AArch64 ELF binaries can interleave data and text in the 1167 // same section. We rely on the markers introduced to 1168 // understand what we need to dump. If the data marker is within a 1169 // function, it is denoted as a word/short etc 1170 if (isArmElf(Obj) && std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1171 !DisassembleAll && 1172 std::binary_search(DataMappingSymsAddr.begin(), 1173 DataMappingSymsAddr.end(), Index)) { 1174 // Switch to data. 1175 support::endianness Endian = 1176 Obj->isLittleEndian() ? support::little : support::big; 1177 while (Index < End) { 1178 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1179 outs() << "\t"; 1180 if (Index + 4 <= End) { 1181 dumpBytes(Bytes.slice(Index, 4), outs()); 1182 outs() << "\t.word\t" 1183 << format_hex(support::endian::read32(Bytes.data() + Index, 1184 Endian), 1185 10); 1186 Index += 4; 1187 } else if (Index + 2 <= End) { 1188 dumpBytes(Bytes.slice(Index, 2), outs()); 1189 outs() << "\t\t.short\t" 1190 << format_hex(support::endian::read16(Bytes.data() + Index, 1191 Endian), 1192 6); 1193 Index += 2; 1194 } else { 1195 dumpBytes(Bytes.slice(Index, 1), outs()); 1196 outs() << "\t\t.byte\t" << format_hex(Bytes[0], 4); 1197 ++Index; 1198 } 1199 outs() << "\n"; 1200 if (std::binary_search(TextMappingSymsAddr.begin(), 1201 TextMappingSymsAddr.end(), Index)) 1202 break; 1203 } 1204 } 1205 1206 // If there is a data symbol inside an ELF text section and we are only 1207 // disassembling text (applicable all architectures), 1208 // we are in a situation where we must print the data and not 1209 // disassemble it. 1210 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1211 !DisassembleAll && Section.isText()) { 1212 // print out data up to 8 bytes at a time in hex and ascii 1213 uint8_t AsciiData[9] = {'\0'}; 1214 uint8_t Byte; 1215 int NumBytes = 0; 1216 1217 for (Index = Start; Index < End; Index += 1) { 1218 if (((SectionAddr + Index) < StartAddress) || 1219 ((SectionAddr + Index) > StopAddress)) 1220 continue; 1221 if (NumBytes == 0) { 1222 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1223 outs() << "\t"; 1224 } 1225 Byte = Bytes.slice(Index)[0]; 1226 outs() << format(" %02x", Byte); 1227 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1228 1229 uint8_t IndentOffset = 0; 1230 NumBytes++; 1231 if (Index == End - 1 || NumBytes > 8) { 1232 // Indent the space for less than 8 bytes data. 1233 // 2 spaces for byte and one for space between bytes 1234 IndentOffset = 3 * (8 - NumBytes); 1235 for (int Excess = NumBytes; Excess < 8; Excess++) 1236 AsciiData[Excess] = '\0'; 1237 NumBytes = 8; 1238 } 1239 if (NumBytes == 8) { 1240 AsciiData[8] = '\0'; 1241 outs() << std::string(IndentOffset, ' ') << " "; 1242 outs() << reinterpret_cast<char *>(AsciiData); 1243 outs() << '\n'; 1244 NumBytes = 0; 1245 } 1246 } 1247 } 1248 if (Index >= End) 1249 break; 1250 1251 // When -z or --disassemble-zeroes are given we always dissasemble them. 1252 // Otherwise we might want to skip zero bytes we see. 1253 if (!DisassembleZeroes) { 1254 uint64_t MaxOffset = End - Index; 1255 // For -reloc: print zero blocks patched by relocations, so that 1256 // relocations can be shown in the dump. 1257 if (RelCur != RelEnd) 1258 MaxOffset = RelCur->getOffset() - Index; 1259 1260 if (size_t N = 1261 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1262 outs() << "\t\t..." << '\n'; 1263 Index += N; 1264 if (Index >= End) 1265 break; 1266 } 1267 } 1268 1269 // Disassemble a real instruction or a data when disassemble all is 1270 // provided 1271 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1272 SectionAddr + Index, DebugOut, 1273 CommentStream); 1274 if (Size == 0) 1275 Size = 1; 1276 1277 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1278 Bytes.slice(Index, Size), 1279 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, 1280 outs(), "", *STI, &SP, &Rels); 1281 outs() << CommentStream.str(); 1282 Comments.clear(); 1283 1284 // Try to resolve the target of a call, tail call, etc. to a specific 1285 // symbol. 1286 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1287 MIA->isConditionalBranch(Inst))) { 1288 uint64_t Target; 1289 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1290 // In a relocatable object, the target's section must reside in 1291 // the same section as the call instruction or it is accessed 1292 // through a relocation. 1293 // 1294 // In a non-relocatable object, the target may be in any section. 1295 // 1296 // N.B. We don't walk the relocations in the relocatable case yet. 1297 auto *TargetSectionSymbols = &Symbols; 1298 if (!Obj->isRelocatableObject()) { 1299 auto SectionAddress = llvm::upper_bound( 1300 SectionAddresses, Target, 1301 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1302 return LHS < RHS.first; 1303 }); 1304 if (SectionAddress != SectionAddresses.begin()) { 1305 --SectionAddress; 1306 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1307 } else { 1308 TargetSectionSymbols = &AbsoluteSymbols; 1309 } 1310 } 1311 1312 // Find the first symbol in the section whose offset is less than 1313 // or equal to the target. If there isn't a section that contains 1314 // the target, find the nearest preceding absolute symbol. 1315 auto TargetSym = llvm::upper_bound( 1316 *TargetSectionSymbols, Target, 1317 [](uint64_t LHS, 1318 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1319 return LHS < std::get<0>(RHS); 1320 }); 1321 if (TargetSym == TargetSectionSymbols->begin()) { 1322 TargetSectionSymbols = &AbsoluteSymbols; 1323 TargetSym = llvm::upper_bound( 1324 AbsoluteSymbols, Target, 1325 [](uint64_t LHS, 1326 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1327 return LHS < std::get<0>(RHS); 1328 }); 1329 } 1330 if (TargetSym != TargetSectionSymbols->begin()) { 1331 --TargetSym; 1332 uint64_t TargetAddress = std::get<0>(*TargetSym); 1333 StringRef TargetName = std::get<1>(*TargetSym); 1334 outs() << " <" << TargetName; 1335 uint64_t Disp = Target - TargetAddress; 1336 if (Disp) 1337 outs() << "+0x" << Twine::utohexstr(Disp); 1338 outs() << '>'; 1339 } 1340 } 1341 } 1342 outs() << "\n"; 1343 1344 // Hexagon does this in pretty printer 1345 if (Obj->getArch() != Triple::hexagon) { 1346 // Print relocation for instruction. 1347 while (RelCur != RelEnd) { 1348 uint64_t Offset = RelCur->getOffset(); 1349 // If this relocation is hidden, skip it. 1350 if (getHidden(*RelCur) || ((SectionAddr + Offset) < StartAddress)) { 1351 ++RelCur; 1352 continue; 1353 } 1354 1355 // Stop when RelCur's offset is past the current instruction. 1356 if (Offset >= Index + Size) 1357 break; 1358 1359 // When --adjust-vma is used, update the address printed. 1360 if (RelCur->getSymbol() != Obj->symbol_end()) { 1361 Expected<section_iterator> SymSI = 1362 RelCur->getSymbol()->getSection(); 1363 if (SymSI && *SymSI != Obj->section_end() && 1364 (shouldAdjustVA(**SymSI))) 1365 Offset += AdjustVMA; 1366 } 1367 1368 printRelocation(*RelCur, SectionAddr + Offset, 1369 Obj->getBytesInAddress()); 1370 ++RelCur; 1371 } 1372 } 1373 } 1374 } 1375 } 1376 } 1377 1378 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1379 if (StartAddress > StopAddress) 1380 error("Start address should be less than stop address"); 1381 1382 const Target *TheTarget = getTarget(Obj); 1383 1384 // Package up features to be passed to target/subtarget 1385 SubtargetFeatures Features = Obj->getFeatures(); 1386 if (!MAttrs.empty()) 1387 for (unsigned I = 0; I != MAttrs.size(); ++I) 1388 Features.AddFeature(MAttrs[I]); 1389 1390 std::unique_ptr<const MCRegisterInfo> MRI( 1391 TheTarget->createMCRegInfo(TripleName)); 1392 if (!MRI) 1393 report_error(Obj->getFileName(), 1394 "no register info for target " + TripleName); 1395 1396 // Set up disassembler. 1397 std::unique_ptr<const MCAsmInfo> AsmInfo( 1398 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1399 if (!AsmInfo) 1400 report_error(Obj->getFileName(), 1401 "no assembly info for target " + TripleName); 1402 std::unique_ptr<const MCSubtargetInfo> STI( 1403 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1404 if (!STI) 1405 report_error(Obj->getFileName(), 1406 "no subtarget info for target " + TripleName); 1407 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1408 if (!MII) 1409 report_error(Obj->getFileName(), 1410 "no instruction info for target " + TripleName); 1411 MCObjectFileInfo MOFI; 1412 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1413 // FIXME: for now initialize MCObjectFileInfo with default values 1414 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1415 1416 std::unique_ptr<MCDisassembler> DisAsm( 1417 TheTarget->createMCDisassembler(*STI, Ctx)); 1418 if (!DisAsm) 1419 report_error(Obj->getFileName(), 1420 "no disassembler for target " + TripleName); 1421 1422 std::unique_ptr<const MCInstrAnalysis> MIA( 1423 TheTarget->createMCInstrAnalysis(MII.get())); 1424 1425 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1426 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1427 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1428 if (!IP) 1429 report_error(Obj->getFileName(), 1430 "no instruction printer for target " + TripleName); 1431 IP->setPrintImmHex(PrintImmHex); 1432 1433 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1434 SourcePrinter SP(Obj, TheTarget->getName()); 1435 1436 for (StringRef Opt : DisassemblerOptions) 1437 if (!IP->applyTargetSpecificCLOption(Opt)) 1438 error("Unrecognized disassembler option: " + Opt); 1439 1440 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), MIA.get(), IP.get(), 1441 STI.get(), PIP, SP, InlineRelocs); 1442 } 1443 1444 void llvm::printRelocations(const ObjectFile *Obj) { 1445 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1446 "%08" PRIx64; 1447 // Regular objdump doesn't print relocations in non-relocatable object 1448 // files. 1449 if (!Obj->isRelocatableObject()) 1450 return; 1451 1452 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1453 if (Section.relocation_begin() == Section.relocation_end()) 1454 continue; 1455 StringRef SecName; 1456 error(Section.getName(SecName)); 1457 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1458 for (const RelocationRef &Reloc : Section.relocations()) { 1459 uint64_t Address = Reloc.getOffset(); 1460 SmallString<32> RelocName; 1461 SmallString<32> ValueStr; 1462 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1463 continue; 1464 Reloc.getTypeName(RelocName); 1465 error(getRelocationValueString(Reloc, ValueStr)); 1466 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1467 << ValueStr << "\n"; 1468 } 1469 outs() << "\n"; 1470 } 1471 } 1472 1473 void llvm::printDynamicRelocations(const ObjectFile *Obj) { 1474 // For the moment, this option is for ELF only 1475 if (!Obj->isELF()) 1476 return; 1477 1478 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1479 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1480 error("not a dynamic object"); 1481 return; 1482 } 1483 1484 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1485 if (DynRelSec.empty()) 1486 return; 1487 1488 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1489 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1490 for (const SectionRef &Section : DynRelSec) { 1491 if (Section.relocation_begin() == Section.relocation_end()) 1492 continue; 1493 for (const RelocationRef &Reloc : Section.relocations()) { 1494 uint64_t Address = Reloc.getOffset(); 1495 SmallString<32> RelocName; 1496 SmallString<32> ValueStr; 1497 Reloc.getTypeName(RelocName); 1498 error(getRelocationValueString(Reloc, ValueStr)); 1499 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1500 << ValueStr << "\n"; 1501 } 1502 } 1503 } 1504 1505 // Returns true if we need to show LMA column when dumping section headers. We 1506 // show it only when the platform is ELF and either we have at least one section 1507 // whose VMA and LMA are different and/or when --show-lma flag is used. 1508 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1509 if (!Obj->isELF()) 1510 return false; 1511 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1512 if (S.getAddress() != getELFSectionLMA(S)) 1513 return true; 1514 return ShowLMA; 1515 } 1516 1517 void llvm::printSectionHeaders(const ObjectFile *Obj) { 1518 bool HasLMAColumn = shouldDisplayLMA(Obj); 1519 if (HasLMAColumn) 1520 outs() << "Sections:\n" 1521 "Idx Name Size VMA LMA " 1522 "Type\n"; 1523 else 1524 outs() << "Sections:\n" 1525 "Idx Name Size VMA Type\n"; 1526 1527 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1528 StringRef Name; 1529 error(Section.getName(Name)); 1530 uint64_t VMA = Section.getAddress(); 1531 if (shouldAdjustVA(Section)) 1532 VMA += AdjustVMA; 1533 1534 uint64_t Size = Section.getSize(); 1535 bool Text = Section.isText(); 1536 bool Data = Section.isData(); 1537 bool BSS = Section.isBSS(); 1538 std::string Type = (std::string(Text ? "TEXT " : "") + 1539 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1540 1541 if (HasLMAColumn) 1542 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %016" PRIx64 1543 " %s\n", 1544 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1545 VMA, getELFSectionLMA(Section), Type.c_str()); 1546 else 1547 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1548 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1549 VMA, Type.c_str()); 1550 } 1551 outs() << "\n"; 1552 } 1553 1554 void llvm::printSectionContents(const ObjectFile *Obj) { 1555 std::error_code EC; 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(A->getFileName(), object_error::invalid_file_type); 2009 } 2010 if (Err) 2011 report_error(std::move(Err), A->getFileName()); 2012 } 2013 2014 /// Open file and figure out how to dump it. 2015 static void dumpInput(StringRef file) { 2016 // If we are using the Mach-O specific object file parser, then let it parse 2017 // the file and process the command line options. So the -arch flags can 2018 // be used to select specific slices, etc. 2019 if (MachOOpt) { 2020 parseInputMachO(file); 2021 return; 2022 } 2023 2024 // Attempt to open the binary. 2025 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2026 Binary &Binary = *OBinary.getBinary(); 2027 2028 if (Archive *A = dyn_cast<Archive>(&Binary)) 2029 dumpArchive(A); 2030 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2031 dumpObject(O); 2032 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2033 parseInputMachO(UB); 2034 else 2035 report_error(file, object_error::invalid_file_type); 2036 } 2037 2038 int main(int argc, char **argv) { 2039 InitLLVM X(argc, argv); 2040 2041 // Initialize targets and assembly printers/parsers. 2042 llvm::InitializeAllTargetInfos(); 2043 llvm::InitializeAllTargetMCs(); 2044 llvm::InitializeAllDisassemblers(); 2045 2046 // Register the target printer for --version. 2047 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2048 2049 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2050 2051 ToolName = argv[0]; 2052 2053 // Defaults to a.out if no filenames specified. 2054 if (InputFilenames.empty()) 2055 InputFilenames.push_back("a.out"); 2056 2057 if (AllHeaders) 2058 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2059 SectionHeaders = SymbolTable = true; 2060 2061 if (DisassembleAll || PrintSource || PrintLines) 2062 Disassemble = true; 2063 2064 if (!Disassemble 2065 && !Relocations 2066 && !DynamicRelocations 2067 && !SectionHeaders 2068 && !SectionContents 2069 && !SymbolTable 2070 && !UnwindInfo 2071 && !PrivateHeaders 2072 && !FileHeaders 2073 && !FirstPrivateHeader 2074 && !ExportsTrie 2075 && !Rebase 2076 && !Bind 2077 && !LazyBind 2078 && !WeakBind 2079 && !RawClangAST 2080 && !(UniversalHeaders && MachOOpt) 2081 && !ArchiveHeaders 2082 && !(IndirectSymbols && MachOOpt) 2083 && !(DataInCode && MachOOpt) 2084 && !(LinkOptHints && MachOOpt) 2085 && !(InfoPlist && MachOOpt) 2086 && !(DylibsUsed && MachOOpt) 2087 && !(DylibId && MachOOpt) 2088 && !(ObjcMetaData && MachOOpt) 2089 && !(!FilterSections.empty() && MachOOpt) 2090 && !PrintFaultMaps 2091 && DwarfDumpType == DIDT_Null) { 2092 cl::PrintHelpMessage(); 2093 return 2; 2094 } 2095 2096 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2097 DisassembleFunctions.end()); 2098 2099 llvm::for_each(InputFilenames, dumpInput); 2100 2101 return EXIT_SUCCESS; 2102 } 2103