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, 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 StringRef ToolName; 297 298 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 299 300 SectionFilter llvm::ToolSectionFilter(llvm::object::ObjectFile const &O) { 301 return SectionFilter( 302 [](llvm::object::SectionRef const &S) { 303 if (FilterSections.empty()) 304 return true; 305 llvm::StringRef String; 306 std::error_code error = S.getName(String); 307 if (error) 308 return false; 309 return is_contained(FilterSections, String); 310 }, 311 O); 312 } 313 314 void llvm::error(std::error_code EC) { 315 if (!EC) 316 return; 317 WithColor::error(errs(), ToolName) 318 << "reading file: " << EC.message() << ".\n"; 319 errs().flush(); 320 exit(1); 321 } 322 323 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 324 WithColor::error(errs(), ToolName) << Message << ".\n"; 325 errs().flush(); 326 exit(1); 327 } 328 329 void llvm::warn(StringRef Message) { 330 WithColor::warning(errs(), ToolName) << Message << ".\n"; 331 errs().flush(); 332 } 333 334 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 335 Twine Message) { 336 WithColor::error(errs(), ToolName) 337 << "'" << File << "': " << Message << ".\n"; 338 exit(1); 339 } 340 341 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 342 std::error_code EC) { 343 assert(EC); 344 WithColor::error(errs(), ToolName) 345 << "'" << File << "': " << EC.message() << ".\n"; 346 exit(1); 347 } 348 349 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 350 llvm::Error E) { 351 assert(E); 352 std::string Buf; 353 raw_string_ostream OS(Buf); 354 logAllUnhandledErrors(std::move(E), OS); 355 OS.flush(); 356 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 357 exit(1); 358 } 359 360 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 361 StringRef FileName, 362 llvm::Error E, 363 StringRef ArchitectureName) { 364 assert(E); 365 WithColor::error(errs(), ToolName); 366 if (ArchiveName != "") 367 errs() << ArchiveName << "(" << FileName << ")"; 368 else 369 errs() << "'" << FileName << "'"; 370 if (!ArchitectureName.empty()) 371 errs() << " (for architecture " << ArchitectureName << ")"; 372 std::string Buf; 373 raw_string_ostream OS(Buf); 374 logAllUnhandledErrors(std::move(E), OS); 375 OS.flush(); 376 errs() << ": " << Buf; 377 exit(1); 378 } 379 380 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 381 const object::Archive::Child &C, 382 llvm::Error E, 383 StringRef ArchitectureName) { 384 Expected<StringRef> NameOrErr = C.getName(); 385 // TODO: if we have a error getting the name then it would be nice to print 386 // the index of which archive member this is and or its offset in the 387 // archive instead of "???" as the name. 388 if (!NameOrErr) { 389 consumeError(NameOrErr.takeError()); 390 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 391 } else 392 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 393 ArchitectureName); 394 } 395 396 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 397 // Figure out the target triple. 398 llvm::Triple TheTriple("unknown-unknown-unknown"); 399 if (TripleName.empty()) { 400 if (Obj) 401 TheTriple = Obj->makeTriple(); 402 } else { 403 TheTriple.setTriple(Triple::normalize(TripleName)); 404 405 // Use the triple, but also try to combine with ARM build attributes. 406 if (Obj) { 407 auto Arch = Obj->getArch(); 408 if (Arch == Triple::arm || Arch == Triple::armeb) 409 Obj->setARMSubArch(TheTriple); 410 } 411 } 412 413 // Get the target specific parser. 414 std::string Error; 415 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 416 Error); 417 if (!TheTarget) { 418 if (Obj) 419 report_error(Obj->getFileName(), "can't find target: " + Error); 420 else 421 error("can't find target: " + Error); 422 } 423 424 // Update the triple name and return the found target. 425 TripleName = TheTriple.getTriple(); 426 return TheTarget; 427 } 428 429 bool llvm::isRelocAddressLess(RelocationRef A, RelocationRef B) { 430 return A.getOffset() < B.getOffset(); 431 } 432 433 static std::error_code getRelocationValueString(const RelocationRef &Rel, 434 SmallVectorImpl<char> &Result) { 435 const ObjectFile *Obj = Rel.getObject(); 436 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 437 return getELFRelocationValueString(ELF, Rel, Result); 438 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 439 return getCOFFRelocationValueString(COFF, Rel, Result); 440 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 441 return getWasmRelocationValueString(Wasm, Rel, Result); 442 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 443 return getMachORelocationValueString(MachO, Rel, Result); 444 llvm_unreachable("unknown object file format"); 445 } 446 447 /// Indicates whether this relocation should hidden when listing 448 /// relocations, usually because it is the trailing part of a multipart 449 /// relocation that will be printed as part of the leading relocation. 450 static bool getHidden(RelocationRef RelRef) { 451 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 452 if (!MachO) 453 return false; 454 455 unsigned Arch = MachO->getArch(); 456 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 457 uint64_t Type = MachO->getRelocationType(Rel); 458 459 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 460 // is always hidden. 461 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 462 return Type == MachO::GENERIC_RELOC_PAIR; 463 464 if (Arch == Triple::x86_64) { 465 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 466 // an X86_64_RELOC_SUBTRACTOR. 467 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 468 DataRefImpl RelPrev = Rel; 469 RelPrev.d.a--; 470 uint64_t PrevType = MachO->getRelocationType(RelPrev); 471 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 472 return true; 473 } 474 } 475 476 return false; 477 } 478 479 namespace { 480 class SourcePrinter { 481 protected: 482 DILineInfo OldLineInfo; 483 const ObjectFile *Obj = nullptr; 484 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 485 // File name to file contents of source 486 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 487 // Mark the line endings of the cached source 488 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 489 490 private: 491 bool cacheSource(const DILineInfo& LineInfoFile); 492 493 public: 494 SourcePrinter() = default; 495 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 496 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 497 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 498 DefaultArch); 499 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 500 } 501 virtual ~SourcePrinter() = default; 502 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 503 StringRef Delimiter = "; "); 504 }; 505 506 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 507 std::unique_ptr<MemoryBuffer> Buffer; 508 if (LineInfo.Source) { 509 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 510 } else { 511 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 512 if (!BufferOrError) 513 return false; 514 Buffer = std::move(*BufferOrError); 515 } 516 // Chomp the file to get lines 517 size_t BufferSize = Buffer->getBufferSize(); 518 const char *BufferStart = Buffer->getBufferStart(); 519 for (const char *Start = BufferStart, *End = BufferStart; 520 End < BufferStart + BufferSize; End++) 521 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 522 (*End == '\r' && *(End + 1) == '\n')) { 523 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 524 if (*End == '\r') 525 End++; 526 Start = End + 1; 527 } 528 SourceCache[LineInfo.FileName] = std::move(Buffer); 529 return true; 530 } 531 532 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 533 StringRef Delimiter) { 534 if (!Symbolizer) 535 return; 536 DILineInfo LineInfo = DILineInfo(); 537 auto ExpectecLineInfo = 538 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 539 if (!ExpectecLineInfo) 540 consumeError(ExpectecLineInfo.takeError()); 541 else 542 LineInfo = *ExpectecLineInfo; 543 544 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 545 LineInfo.Line == 0) 546 return; 547 548 if (PrintLines) 549 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 550 if (PrintSource) { 551 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 552 if (!cacheSource(LineInfo)) 553 return; 554 auto FileBuffer = SourceCache.find(LineInfo.FileName); 555 if (FileBuffer != SourceCache.end()) { 556 auto LineBuffer = LineCache.find(LineInfo.FileName); 557 if (LineBuffer != LineCache.end()) { 558 if (LineInfo.Line > LineBuffer->second.size()) 559 return; 560 // Vector begins at 0, line numbers are non-zero 561 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 562 << "\n"; 563 } 564 } 565 } 566 OldLineInfo = LineInfo; 567 } 568 569 static bool isArmElf(const ObjectFile *Obj) { 570 return (Obj->isELF() && 571 (Obj->getArch() == Triple::aarch64 || 572 Obj->getArch() == Triple::aarch64_be || 573 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 574 Obj->getArch() == Triple::thumb || 575 Obj->getArch() == Triple::thumbeb)); 576 } 577 578 static void printRelocation(const RelocationRef &Rel, uint64_t Address, 579 uint8_t AddrSize) { 580 StringRef Fmt = 581 AddrSize > 4 ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 582 SmallString<16> Name; 583 SmallString<32> Val; 584 Rel.getTypeName(Name); 585 error(getRelocationValueString(Rel, Val)); 586 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n"; 587 } 588 589 class PrettyPrinter { 590 public: 591 virtual ~PrettyPrinter() = default; 592 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 593 ArrayRef<uint8_t> Bytes, uint64_t Address, 594 raw_ostream &OS, StringRef Annot, 595 MCSubtargetInfo const &STI, SourcePrinter *SP, 596 std::vector<RelocationRef> *Rels = nullptr) { 597 if (SP && (PrintSource || PrintLines)) 598 SP->printSourceLine(OS, Address); 599 if (!NoLeadingAddr) 600 OS << format("%8" PRIx64 ":", Address); 601 if (!NoShowRawInsn) { 602 OS << "\t"; 603 dumpBytes(Bytes, OS); 604 } 605 if (MI) 606 IP.printInst(MI, OS, "", STI); 607 else 608 OS << " <unknown>"; 609 } 610 }; 611 PrettyPrinter PrettyPrinterInst; 612 class HexagonPrettyPrinter : public PrettyPrinter { 613 public: 614 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 615 raw_ostream &OS) { 616 uint32_t opcode = 617 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 618 if (!NoLeadingAddr) 619 OS << format("%8" PRIx64 ":", Address); 620 if (!NoShowRawInsn) { 621 OS << "\t"; 622 dumpBytes(Bytes.slice(0, 4), OS); 623 OS << format("%08" PRIx32, opcode); 624 } 625 } 626 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 627 uint64_t Address, raw_ostream &OS, StringRef Annot, 628 MCSubtargetInfo const &STI, SourcePrinter *SP, 629 std::vector<RelocationRef> *Rels) override { 630 if (SP && (PrintSource || PrintLines)) 631 SP->printSourceLine(OS, Address, ""); 632 if (!MI) { 633 printLead(Bytes, Address, OS); 634 OS << " <unknown>"; 635 return; 636 } 637 std::string Buffer; 638 { 639 raw_string_ostream TempStream(Buffer); 640 IP.printInst(MI, TempStream, "", STI); 641 } 642 StringRef Contents(Buffer); 643 // Split off bundle attributes 644 auto PacketBundle = Contents.rsplit('\n'); 645 // Split off first instruction from the rest 646 auto HeadTail = PacketBundle.first.split('\n'); 647 auto Preamble = " { "; 648 auto Separator = ""; 649 650 // Hexagon's packets require relocations to be inline rather than 651 // clustered at the end of the packet. 652 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 653 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 654 auto PrintReloc = [&]() -> void { 655 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address)) { 656 if (RelCur->getOffset() == Address) { 657 printRelocation(*RelCur, Address, 4); 658 return; 659 } 660 ++RelCur; 661 } 662 }; 663 664 while (!HeadTail.first.empty()) { 665 OS << Separator; 666 Separator = "\n"; 667 if (SP && (PrintSource || PrintLines)) 668 SP->printSourceLine(OS, Address, ""); 669 printLead(Bytes, Address, OS); 670 OS << Preamble; 671 Preamble = " "; 672 StringRef Inst; 673 auto Duplex = HeadTail.first.split('\v'); 674 if (!Duplex.second.empty()) { 675 OS << Duplex.first; 676 OS << "; "; 677 Inst = Duplex.second; 678 } 679 else 680 Inst = HeadTail.first; 681 OS << Inst; 682 HeadTail = HeadTail.second.split('\n'); 683 if (HeadTail.first.empty()) 684 OS << " } " << PacketBundle.second; 685 PrintReloc(); 686 Bytes = Bytes.slice(4); 687 Address += 4; 688 } 689 } 690 }; 691 HexagonPrettyPrinter HexagonPrettyPrinterInst; 692 693 class AMDGCNPrettyPrinter : public PrettyPrinter { 694 public: 695 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 696 uint64_t Address, raw_ostream &OS, StringRef Annot, 697 MCSubtargetInfo const &STI, SourcePrinter *SP, 698 std::vector<RelocationRef> *Rels) override { 699 if (SP && (PrintSource || PrintLines)) 700 SP->printSourceLine(OS, Address); 701 702 typedef support::ulittle32_t U32; 703 704 if (MI) { 705 SmallString<40> InstStr; 706 raw_svector_ostream IS(InstStr); 707 708 IP.printInst(MI, IS, "", STI); 709 710 OS << left_justify(IS.str(), 60); 711 } else { 712 // an unrecognized encoding - this is probably data so represent it 713 // using the .long directive, or .byte directive if fewer than 4 bytes 714 // remaining 715 if (Bytes.size() >= 4) { 716 OS << format("\t.long 0x%08" PRIx32 " ", 717 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 718 OS.indent(42); 719 } else { 720 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 721 for (unsigned int i = 1; i < Bytes.size(); i++) 722 OS << format(", 0x%02" PRIx8, Bytes[i]); 723 OS.indent(55 - (6 * Bytes.size())); 724 } 725 } 726 727 OS << format("// %012" PRIX64 ": ", Address); 728 if (Bytes.size() >=4) { 729 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 730 Bytes.size() / sizeof(U32))) 731 // D should be explicitly casted to uint32_t here as it is passed 732 // by format to snprintf as vararg. 733 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 734 } else { 735 for (unsigned int i = 0; i < Bytes.size(); i++) 736 OS << format("%02" PRIX8 " ", Bytes[i]); 737 } 738 739 if (!Annot.empty()) 740 OS << "// " << Annot; 741 } 742 }; 743 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 744 745 class BPFPrettyPrinter : public PrettyPrinter { 746 public: 747 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 748 uint64_t Address, raw_ostream &OS, StringRef Annot, 749 MCSubtargetInfo const &STI, SourcePrinter *SP, 750 std::vector<RelocationRef> *Rels) override { 751 if (SP && (PrintSource || PrintLines)) 752 SP->printSourceLine(OS, Address); 753 if (!NoLeadingAddr) 754 OS << format("%8" PRId64 ":", Address / 8); 755 if (!NoShowRawInsn) { 756 OS << "\t"; 757 dumpBytes(Bytes, OS); 758 } 759 if (MI) 760 IP.printInst(MI, OS, "", STI); 761 else 762 OS << " <unknown>"; 763 } 764 }; 765 BPFPrettyPrinter BPFPrettyPrinterInst; 766 767 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 768 switch(Triple.getArch()) { 769 default: 770 return PrettyPrinterInst; 771 case Triple::hexagon: 772 return HexagonPrettyPrinterInst; 773 case Triple::amdgcn: 774 return AMDGCNPrettyPrinterInst; 775 case Triple::bpfel: 776 case Triple::bpfeb: 777 return BPFPrettyPrinterInst; 778 } 779 } 780 } 781 782 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 783 assert(Obj->isELF()); 784 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 785 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 786 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 787 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 788 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 789 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 790 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 791 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 792 llvm_unreachable("Unsupported binary format"); 793 } 794 795 template <class ELFT> static void 796 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 797 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 798 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 799 uint8_t SymbolType = Symbol.getELFType(); 800 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 801 continue; 802 803 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 804 if (!AddressOrErr) 805 report_error(Obj->getFileName(), AddressOrErr.takeError()); 806 807 Expected<StringRef> Name = Symbol.getName(); 808 if (!Name) 809 report_error(Obj->getFileName(), Name.takeError()); 810 if (Name->empty()) 811 continue; 812 813 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 814 if (!SectionOrErr) 815 report_error(Obj->getFileName(), SectionOrErr.takeError()); 816 section_iterator SecI = *SectionOrErr; 817 if (SecI == Obj->section_end()) 818 continue; 819 820 AllSymbols[*SecI].emplace_back(*AddressOrErr, *Name, SymbolType); 821 } 822 } 823 824 static void 825 addDynamicElfSymbols(const ObjectFile *Obj, 826 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 827 assert(Obj->isELF()); 828 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 829 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 830 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 831 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 832 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 833 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 834 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 835 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 836 else 837 llvm_unreachable("Unsupported binary format"); 838 } 839 840 static void addPltEntries(const ObjectFile *Obj, 841 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 842 StringSaver &Saver) { 843 Optional<SectionRef> Plt = None; 844 for (const SectionRef &Section : Obj->sections()) { 845 StringRef Name; 846 if (Section.getName(Name)) 847 continue; 848 if (Name == ".plt") 849 Plt = Section; 850 } 851 if (!Plt) 852 return; 853 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 854 for (auto PltEntry : ElfObj->getPltAddresses()) { 855 SymbolRef Symbol(PltEntry.first, ElfObj); 856 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 857 858 Expected<StringRef> NameOrErr = Symbol.getName(); 859 if (!NameOrErr) 860 report_error(Obj->getFileName(), NameOrErr.takeError()); 861 if (NameOrErr->empty()) 862 continue; 863 StringRef Name = Saver.save((*NameOrErr + "@plt").str()); 864 865 AllSymbols[*Plt].emplace_back(PltEntry.second, Name, SymbolType); 866 } 867 } 868 } 869 870 // Normally the disassembly output will skip blocks of zeroes. This function 871 // returns the number of zero bytes that can be skipped when dumping the 872 // disassembly of the instructions in Buf. 873 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 874 // When -z or --disassemble-zeroes are given we always dissasemble them. 875 if (DisassembleZeroes) 876 return 0; 877 878 // Find the number of leading zeroes. 879 size_t N = 0; 880 while (N < Buf.size() && !Buf[N]) 881 ++N; 882 883 // We may want to skip blocks of zero bytes, but unless we see 884 // at least 8 of them in a row. 885 if (N < 8) 886 return 0; 887 888 // We skip zeroes in multiples of 4 because do not want to truncate an 889 // instruction if it starts with a zero byte. 890 return N & ~0x3; 891 } 892 893 // Returns a map from sections to their relocations. 894 static std::map<SectionRef, std::vector<RelocationRef>> 895 getRelocsMap(llvm::object::ObjectFile const &Obj) { 896 std::map<SectionRef, std::vector<RelocationRef>> Ret; 897 for (const SectionRef &Section : ToolSectionFilter(Obj)) { 898 section_iterator RelSec = Section.getRelocatedSection(); 899 if (RelSec == Obj.section_end()) 900 continue; 901 std::vector<RelocationRef> &V = Ret[*RelSec]; 902 for (const RelocationRef &R : Section.relocations()) 903 V.push_back(R); 904 // Sort relocations by address. 905 llvm::sort(V, isRelocAddressLess); 906 } 907 return Ret; 908 } 909 910 // Used for --adjust-vma to check if address should be adjusted by the 911 // specified value for a given section. 912 // For ELF we do not adjust non-allocatable sections like debug ones, 913 // because they are not loadable. 914 // TODO: implement for other file formats. 915 static bool shouldAdjustVA(const SectionRef &Section) { 916 const ObjectFile *Obj = Section.getObject(); 917 if (isa<object::ELFObjectFileBase>(Obj)) 918 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 919 return false; 920 } 921 922 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 923 MCContext &Ctx, MCDisassembler *DisAsm, 924 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 925 const MCSubtargetInfo *STI, PrettyPrinter &PIP, 926 SourcePrinter &SP, bool InlineRelocs) { 927 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 928 if (InlineRelocs) 929 RelocMap = getRelocsMap(*Obj); 930 931 // Create a mapping from virtual address to symbol name. This is used to 932 // pretty print the symbols while disassembling. 933 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 934 SectionSymbolsTy AbsoluteSymbols; 935 for (const SymbolRef &Symbol : Obj->symbols()) { 936 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 937 if (!AddressOrErr) 938 report_error(Obj->getFileName(), AddressOrErr.takeError()); 939 uint64_t Address = *AddressOrErr; 940 941 Expected<StringRef> Name = Symbol.getName(); 942 if (!Name) 943 report_error(Obj->getFileName(), Name.takeError()); 944 if (Name->empty()) 945 continue; 946 947 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 948 if (!SectionOrErr) 949 report_error(Obj->getFileName(), SectionOrErr.takeError()); 950 951 uint8_t SymbolType = ELF::STT_NOTYPE; 952 if (Obj->isELF()) { 953 SymbolType = getElfSymbolType(Obj, Symbol); 954 if (SymbolType == ELF::STT_SECTION) 955 continue; 956 } 957 958 section_iterator SecI = *SectionOrErr; 959 if (SecI != Obj->section_end()) 960 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 961 else 962 AbsoluteSymbols.emplace_back(Address, *Name, SymbolType); 963 964 965 } 966 if (AllSymbols.empty() && Obj->isELF()) 967 addDynamicElfSymbols(Obj, AllSymbols); 968 969 BumpPtrAllocator A; 970 StringSaver Saver(A); 971 addPltEntries(Obj, AllSymbols, Saver); 972 973 // Create a mapping from virtual address to section. 974 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 975 for (SectionRef Sec : Obj->sections()) 976 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 977 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 978 979 // Linked executables (.exe and .dll files) typically don't include a real 980 // symbol table but they might contain an export table. 981 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 982 for (const auto &ExportEntry : COFFObj->export_directories()) { 983 StringRef Name; 984 error(ExportEntry.getSymbolName(Name)); 985 if (Name.empty()) 986 continue; 987 uint32_t RVA; 988 error(ExportEntry.getExportRVA(RVA)); 989 990 uint64_t VA = COFFObj->getImageBase() + RVA; 991 auto Sec = std::upper_bound( 992 SectionAddresses.begin(), SectionAddresses.end(), VA, 993 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 994 return LHS < RHS.first; 995 }); 996 if (Sec != SectionAddresses.begin()) 997 --Sec; 998 else 999 Sec = SectionAddresses.end(); 1000 1001 if (Sec != SectionAddresses.end()) 1002 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1003 else 1004 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1005 } 1006 } 1007 1008 // Sort all the symbols, this allows us to use a simple binary search to find 1009 // a symbol near an address. 1010 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1011 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1012 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1013 1014 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1015 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1016 continue; 1017 1018 uint64_t SectionAddr = Section.getAddress(); 1019 uint64_t SectSize = Section.getSize(); 1020 if (!SectSize) 1021 continue; 1022 1023 // Get the list of all the symbols in this section. 1024 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1025 std::vector<uint64_t> DataMappingSymsAddr; 1026 std::vector<uint64_t> TextMappingSymsAddr; 1027 if (isArmElf(Obj)) { 1028 for (const auto &Symb : Symbols) { 1029 uint64_t Address = std::get<0>(Symb); 1030 StringRef Name = std::get<1>(Symb); 1031 if (Name.startswith("$d")) 1032 DataMappingSymsAddr.push_back(Address - SectionAddr); 1033 if (Name.startswith("$x")) 1034 TextMappingSymsAddr.push_back(Address - SectionAddr); 1035 if (Name.startswith("$a")) 1036 TextMappingSymsAddr.push_back(Address - SectionAddr); 1037 if (Name.startswith("$t")) 1038 TextMappingSymsAddr.push_back(Address - SectionAddr); 1039 } 1040 } 1041 1042 llvm::sort(DataMappingSymsAddr); 1043 llvm::sort(TextMappingSymsAddr); 1044 1045 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1046 // AMDGPU disassembler uses symbolizer for printing labels 1047 std::unique_ptr<MCRelocationInfo> RelInfo( 1048 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1049 if (RelInfo) { 1050 std::unique_ptr<MCSymbolizer> Symbolizer( 1051 TheTarget->createMCSymbolizer( 1052 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1053 DisAsm->setSymbolizer(std::move(Symbolizer)); 1054 } 1055 } 1056 1057 StringRef SegmentName = ""; 1058 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1059 DataRefImpl DR = Section.getRawDataRefImpl(); 1060 SegmentName = MachO->getSectionFinalSegmentName(DR); 1061 } 1062 StringRef SectionName; 1063 error(Section.getName(SectionName)); 1064 1065 // If the section has no symbol at the start, just insert a dummy one. 1066 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1067 Symbols.insert( 1068 Symbols.begin(), 1069 std::make_tuple(SectionAddr, SectionName, 1070 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1071 } 1072 1073 SmallString<40> Comments; 1074 raw_svector_ostream CommentStream(Comments); 1075 1076 StringRef BytesStr; 1077 error(Section.getContents(BytesStr)); 1078 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1079 BytesStr.size()); 1080 1081 uint64_t VMAAdjustment = 0; 1082 if (shouldAdjustVA(Section)) 1083 VMAAdjustment = AdjustVMA; 1084 1085 uint64_t Size; 1086 uint64_t Index; 1087 bool PrintedSection = false; 1088 std::vector<RelocationRef> Rels = RelocMap[Section]; 1089 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1090 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1091 // Disassemble symbol by symbol. 1092 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1093 uint64_t Start = std::get<0>(Symbols[SI]) - SectionAddr; 1094 // The end is either the section end or the beginning of the next 1095 // symbol. 1096 uint64_t End = (SI == SE - 1) 1097 ? SectSize 1098 : std::get<0>(Symbols[SI + 1]) - SectionAddr; 1099 // Don't try to disassemble beyond the end of section contents. 1100 if (End > SectSize) 1101 End = SectSize; 1102 // If this symbol has the same address as the next symbol, then skip it. 1103 if (Start >= End) 1104 continue; 1105 1106 // Check if we need to skip symbol 1107 // Skip if the symbol's data is not between StartAddress and StopAddress 1108 if (End + SectionAddr < StartAddress || 1109 Start + SectionAddr > StopAddress) { 1110 continue; 1111 } 1112 1113 /// Skip if user requested specific symbols and this is not in the list 1114 if (!DisasmFuncsSet.empty() && 1115 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1116 continue; 1117 1118 if (!PrintedSection) { 1119 PrintedSection = true; 1120 outs() << "Disassembly of section "; 1121 if (!SegmentName.empty()) 1122 outs() << SegmentName << ","; 1123 outs() << SectionName << ':'; 1124 } 1125 1126 // Stop disassembly at the stop address specified 1127 if (End + SectionAddr > StopAddress) 1128 End = StopAddress - SectionAddr; 1129 1130 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1131 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1132 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1133 Start += 256; 1134 } 1135 if (SI == SE - 1 || 1136 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1137 // cut trailing zeroes at the end of kernel 1138 // cut up to 256 bytes 1139 const uint64_t EndAlign = 256; 1140 const auto Limit = End - (std::min)(EndAlign, End - Start); 1141 while (End > Limit && 1142 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1143 End -= 4; 1144 } 1145 } 1146 1147 outs() << '\n'; 1148 if (!NoLeadingAddr) 1149 outs() << format("%016" PRIx64 " ", 1150 SectionAddr + Start + VMAAdjustment); 1151 1152 StringRef SymbolName = std::get<1>(Symbols[SI]); 1153 if (Demangle) 1154 outs() << demangle(SymbolName) << ":\n"; 1155 else 1156 outs() << SymbolName << ":\n"; 1157 1158 // Don't print raw contents of a virtual section. A virtual section 1159 // doesn't have any contents in the file. 1160 if (Section.isVirtual()) { 1161 outs() << "...\n"; 1162 continue; 1163 } 1164 1165 #ifndef NDEBUG 1166 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1167 #else 1168 raw_ostream &DebugOut = nulls(); 1169 #endif 1170 1171 // Some targets (like WebAssembly) have a special prelude at the start 1172 // of each symbol. 1173 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1174 SectionAddr + Start, DebugOut, CommentStream); 1175 Start += Size; 1176 1177 for (Index = Start; Index < End; Index += Size) { 1178 MCInst Inst; 1179 1180 if (Index + SectionAddr < StartAddress || 1181 Index + SectionAddr > StopAddress) { 1182 // skip byte by byte till StartAddress is reached 1183 Size = 1; 1184 continue; 1185 } 1186 // AArch64 ELF binaries can interleave data and text in the 1187 // same section. We rely on the markers introduced to 1188 // understand what we need to dump. If the data marker is within a 1189 // function, it is denoted as a word/short etc 1190 if (isArmElf(Obj) && std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1191 !DisassembleAll) { 1192 uint64_t Stride = 0; 1193 1194 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1195 DataMappingSymsAddr.end(), Index); 1196 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1197 // Switch to data. 1198 while (Index < End) { 1199 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1200 outs() << "\t"; 1201 if (Index + 4 <= End) { 1202 Stride = 4; 1203 dumpBytes(Bytes.slice(Index, 4), outs()); 1204 outs() << "\t.word\t"; 1205 uint32_t Data = 0; 1206 if (Obj->isLittleEndian()) { 1207 const auto Word = 1208 reinterpret_cast<const support::ulittle32_t *>( 1209 Bytes.data() + Index); 1210 Data = *Word; 1211 } else { 1212 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1213 Bytes.data() + Index); 1214 Data = *Word; 1215 } 1216 outs() << "0x" << format("%08" PRIx32, Data); 1217 } else if (Index + 2 <= End) { 1218 Stride = 2; 1219 dumpBytes(Bytes.slice(Index, 2), outs()); 1220 outs() << "\t\t.short\t"; 1221 uint16_t Data = 0; 1222 if (Obj->isLittleEndian()) { 1223 const auto Short = 1224 reinterpret_cast<const support::ulittle16_t *>( 1225 Bytes.data() + Index); 1226 Data = *Short; 1227 } else { 1228 const auto Short = 1229 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1230 Index); 1231 Data = *Short; 1232 } 1233 outs() << "0x" << format("%04" PRIx16, Data); 1234 } else { 1235 Stride = 1; 1236 dumpBytes(Bytes.slice(Index, 1), outs()); 1237 outs() << "\t\t.byte\t"; 1238 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1239 } 1240 Index += Stride; 1241 outs() << "\n"; 1242 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1243 TextMappingSymsAddr.end(), Index); 1244 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1245 break; 1246 } 1247 } 1248 } 1249 1250 // If there is a data symbol inside an ELF text section and we are only 1251 // disassembling text (applicable all architectures), 1252 // we are in a situation where we must print the data and not 1253 // disassemble it. 1254 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1255 !DisassembleAll && Section.isText()) { 1256 // print out data up to 8 bytes at a time in hex and ascii 1257 uint8_t AsciiData[9] = {'\0'}; 1258 uint8_t Byte; 1259 int NumBytes = 0; 1260 1261 for (Index = Start; Index < End; Index += 1) { 1262 if (((SectionAddr + Index) < StartAddress) || 1263 ((SectionAddr + Index) > StopAddress)) 1264 continue; 1265 if (NumBytes == 0) { 1266 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1267 outs() << "\t"; 1268 } 1269 Byte = Bytes.slice(Index)[0]; 1270 outs() << format(" %02x", Byte); 1271 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1272 1273 uint8_t IndentOffset = 0; 1274 NumBytes++; 1275 if (Index == End - 1 || NumBytes > 8) { 1276 // Indent the space for less than 8 bytes data. 1277 // 2 spaces for byte and one for space between bytes 1278 IndentOffset = 3 * (8 - NumBytes); 1279 for (int Excess = NumBytes; Excess < 8; Excess++) 1280 AsciiData[Excess] = '\0'; 1281 NumBytes = 8; 1282 } 1283 if (NumBytes == 8) { 1284 AsciiData[8] = '\0'; 1285 outs() << std::string(IndentOffset, ' ') << " "; 1286 outs() << reinterpret_cast<char *>(AsciiData); 1287 outs() << '\n'; 1288 NumBytes = 0; 1289 } 1290 } 1291 } 1292 if (Index >= End) 1293 break; 1294 1295 if (size_t N = 1296 countSkippableZeroBytes(Bytes.slice(Index, End - Index))) { 1297 outs() << "\t\t..." << '\n'; 1298 Index += N; 1299 if (Index >= End) 1300 break; 1301 } 1302 1303 // Disassemble a real instruction or a data when disassemble all is 1304 // provided 1305 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1306 SectionAddr + Index, DebugOut, 1307 CommentStream); 1308 if (Size == 0) 1309 Size = 1; 1310 1311 PIP.printInst( 1312 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 1313 SectionAddr + Index + VMAAdjustment, outs(), "", *STI, &SP, &Rels); 1314 outs() << CommentStream.str(); 1315 Comments.clear(); 1316 1317 // Try to resolve the target of a call, tail call, etc. to a specific 1318 // symbol. 1319 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1320 MIA->isConditionalBranch(Inst))) { 1321 uint64_t Target; 1322 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1323 // In a relocatable object, the target's section must reside in 1324 // the same section as the call instruction or it is accessed 1325 // through a relocation. 1326 // 1327 // In a non-relocatable object, the target may be in any section. 1328 // 1329 // N.B. We don't walk the relocations in the relocatable case yet. 1330 auto *TargetSectionSymbols = &Symbols; 1331 if (!Obj->isRelocatableObject()) { 1332 auto SectionAddress = std::upper_bound( 1333 SectionAddresses.begin(), SectionAddresses.end(), Target, 1334 [](uint64_t LHS, 1335 const std::pair<uint64_t, SectionRef> &RHS) { 1336 return LHS < RHS.first; 1337 }); 1338 if (SectionAddress != SectionAddresses.begin()) { 1339 --SectionAddress; 1340 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1341 } else { 1342 TargetSectionSymbols = &AbsoluteSymbols; 1343 } 1344 } 1345 1346 // Find the first symbol in the section whose offset is less than 1347 // or equal to the target. If there isn't a section that contains 1348 // the target, find the nearest preceding absolute symbol. 1349 auto TargetSym = std::upper_bound( 1350 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1351 Target, [](uint64_t LHS, 1352 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1353 return LHS < std::get<0>(RHS); 1354 }); 1355 if (TargetSym == TargetSectionSymbols->begin()) { 1356 TargetSectionSymbols = &AbsoluteSymbols; 1357 TargetSym = std::upper_bound( 1358 AbsoluteSymbols.begin(), AbsoluteSymbols.end(), 1359 Target, [](uint64_t LHS, 1360 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1361 return LHS < std::get<0>(RHS); 1362 }); 1363 } 1364 if (TargetSym != TargetSectionSymbols->begin()) { 1365 --TargetSym; 1366 uint64_t TargetAddress = std::get<0>(*TargetSym); 1367 StringRef TargetName = std::get<1>(*TargetSym); 1368 outs() << " <" << TargetName; 1369 uint64_t Disp = Target - TargetAddress; 1370 if (Disp) 1371 outs() << "+0x" << Twine::utohexstr(Disp); 1372 outs() << '>'; 1373 } 1374 } 1375 } 1376 outs() << "\n"; 1377 1378 // Hexagon does this in pretty printer 1379 if (Obj->getArch() != Triple::hexagon) { 1380 // Print relocation for instruction. 1381 while (RelCur != RelEnd) { 1382 uint64_t Offset = RelCur->getOffset(); 1383 // If this relocation is hidden, skip it. 1384 if (getHidden(*RelCur) || ((SectionAddr + Offset) < StartAddress)) { 1385 ++RelCur; 1386 continue; 1387 } 1388 1389 // Stop when RelCur's offset is past the current instruction. 1390 if (Offset >= Index + Size) 1391 break; 1392 1393 // When --adjust-vma is used, update the address printed. 1394 if (RelCur->getSymbol() != Obj->symbol_end()) { 1395 Expected<section_iterator> SymSI = 1396 RelCur->getSymbol()->getSection(); 1397 if (SymSI && *SymSI != Obj->section_end() && 1398 (shouldAdjustVA(**SymSI))) 1399 Offset += AdjustVMA; 1400 } 1401 1402 printRelocation(*RelCur, SectionAddr + Offset, 1403 Obj->getBytesInAddress()); 1404 ++RelCur; 1405 } 1406 } 1407 } 1408 } 1409 } 1410 } 1411 1412 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1413 if (StartAddress > StopAddress) 1414 error("Start address should be less than stop address"); 1415 1416 const Target *TheTarget = getTarget(Obj); 1417 1418 // Package up features to be passed to target/subtarget 1419 SubtargetFeatures Features = Obj->getFeatures(); 1420 if (!MAttrs.empty()) 1421 for (unsigned I = 0; I != MAttrs.size(); ++I) 1422 Features.AddFeature(MAttrs[I]); 1423 1424 std::unique_ptr<const MCRegisterInfo> MRI( 1425 TheTarget->createMCRegInfo(TripleName)); 1426 if (!MRI) 1427 report_error(Obj->getFileName(), 1428 "no register info for target " + TripleName); 1429 1430 // Set up disassembler. 1431 std::unique_ptr<const MCAsmInfo> AsmInfo( 1432 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1433 if (!AsmInfo) 1434 report_error(Obj->getFileName(), 1435 "no assembly info for target " + TripleName); 1436 std::unique_ptr<const MCSubtargetInfo> STI( 1437 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1438 if (!STI) 1439 report_error(Obj->getFileName(), 1440 "no subtarget info for target " + TripleName); 1441 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1442 if (!MII) 1443 report_error(Obj->getFileName(), 1444 "no instruction info for target " + TripleName); 1445 MCObjectFileInfo MOFI; 1446 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1447 // FIXME: for now initialize MCObjectFileInfo with default values 1448 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1449 1450 std::unique_ptr<MCDisassembler> DisAsm( 1451 TheTarget->createMCDisassembler(*STI, Ctx)); 1452 if (!DisAsm) 1453 report_error(Obj->getFileName(), 1454 "no disassembler for target " + TripleName); 1455 1456 std::unique_ptr<const MCInstrAnalysis> MIA( 1457 TheTarget->createMCInstrAnalysis(MII.get())); 1458 1459 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1460 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1461 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1462 if (!IP) 1463 report_error(Obj->getFileName(), 1464 "no instruction printer for target " + TripleName); 1465 IP->setPrintImmHex(PrintImmHex); 1466 1467 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1468 SourcePrinter SP(Obj, TheTarget->getName()); 1469 1470 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), MIA.get(), IP.get(), 1471 STI.get(), PIP, SP, InlineRelocs); 1472 } 1473 1474 void llvm::printRelocations(const ObjectFile *Obj) { 1475 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1476 "%08" PRIx64; 1477 // Regular objdump doesn't print relocations in non-relocatable object 1478 // files. 1479 if (!Obj->isRelocatableObject()) 1480 return; 1481 1482 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1483 if (Section.relocation_begin() == Section.relocation_end()) 1484 continue; 1485 StringRef SecName; 1486 error(Section.getName(SecName)); 1487 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1488 for (const RelocationRef &Reloc : Section.relocations()) { 1489 uint64_t Address = Reloc.getOffset(); 1490 SmallString<32> RelocName; 1491 SmallString<32> ValueStr; 1492 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1493 continue; 1494 Reloc.getTypeName(RelocName); 1495 error(getRelocationValueString(Reloc, ValueStr)); 1496 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1497 << ValueStr << "\n"; 1498 } 1499 outs() << "\n"; 1500 } 1501 } 1502 1503 void llvm::printDynamicRelocations(const ObjectFile *Obj) { 1504 // For the moment, this option is for ELF only 1505 if (!Obj->isELF()) 1506 return; 1507 1508 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1509 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1510 error("not a dynamic object"); 1511 return; 1512 } 1513 1514 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1515 if (DynRelSec.empty()) 1516 return; 1517 1518 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1519 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1520 for (const SectionRef &Section : DynRelSec) { 1521 if (Section.relocation_begin() == Section.relocation_end()) 1522 continue; 1523 for (const RelocationRef &Reloc : Section.relocations()) { 1524 uint64_t Address = Reloc.getOffset(); 1525 SmallString<32> RelocName; 1526 SmallString<32> ValueStr; 1527 Reloc.getTypeName(RelocName); 1528 error(getRelocationValueString(Reloc, ValueStr)); 1529 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1530 << ValueStr << "\n"; 1531 } 1532 } 1533 } 1534 1535 // Returns true if we need to show LMA column when dumping section headers. We 1536 // show it only when the platform is ELF and either we have at least one section 1537 // whose VMA and LMA are different and/or when --show-lma flag is used. 1538 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1539 if (!Obj->isELF()) 1540 return false; 1541 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1542 if (S.getAddress() != getELFSectionLMA(S)) 1543 return true; 1544 return ShowLMA; 1545 } 1546 1547 void llvm::printSectionHeaders(const ObjectFile *Obj) { 1548 bool HasLMAColumn = shouldDisplayLMA(Obj); 1549 if (HasLMAColumn) 1550 outs() << "Sections:\n" 1551 "Idx Name Size VMA LMA " 1552 "Type\n"; 1553 else 1554 outs() << "Sections:\n" 1555 "Idx Name Size VMA Type\n"; 1556 1557 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1558 StringRef Name; 1559 error(Section.getName(Name)); 1560 uint64_t VMA = Section.getAddress(); 1561 if (shouldAdjustVA(Section)) 1562 VMA += AdjustVMA; 1563 1564 uint64_t Size = Section.getSize(); 1565 bool Text = Section.isText(); 1566 bool Data = Section.isData(); 1567 bool BSS = Section.isBSS(); 1568 std::string Type = (std::string(Text ? "TEXT " : "") + 1569 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1570 1571 if (HasLMAColumn) 1572 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %016" PRIx64 1573 " %s\n", 1574 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1575 VMA, getELFSectionLMA(Section), Type.c_str()); 1576 else 1577 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1578 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1579 VMA, Type.c_str()); 1580 } 1581 outs() << "\n"; 1582 } 1583 1584 void llvm::printSectionContents(const ObjectFile *Obj) { 1585 std::error_code EC; 1586 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1587 StringRef Name; 1588 StringRef Contents; 1589 error(Section.getName(Name)); 1590 uint64_t BaseAddr = Section.getAddress(); 1591 uint64_t Size = Section.getSize(); 1592 if (!Size) 1593 continue; 1594 1595 outs() << "Contents of section " << Name << ":\n"; 1596 if (Section.isBSS()) { 1597 outs() << format("<skipping contents of bss section at [%04" PRIx64 1598 ", %04" PRIx64 ")>\n", 1599 BaseAddr, BaseAddr + Size); 1600 continue; 1601 } 1602 1603 error(Section.getContents(Contents)); 1604 1605 // Dump out the content as hex and printable ascii characters. 1606 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1607 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1608 // Dump line of hex. 1609 for (std::size_t I = 0; I < 16; ++I) { 1610 if (I != 0 && I % 4 == 0) 1611 outs() << ' '; 1612 if (Addr + I < End) 1613 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1614 << hexdigit(Contents[Addr + I] & 0xF, true); 1615 else 1616 outs() << " "; 1617 } 1618 // Print ascii. 1619 outs() << " "; 1620 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1621 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1622 outs() << Contents[Addr + I]; 1623 else 1624 outs() << "."; 1625 } 1626 outs() << "\n"; 1627 } 1628 } 1629 } 1630 1631 void llvm::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1632 StringRef ArchitectureName) { 1633 outs() << "SYMBOL TABLE:\n"; 1634 1635 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1636 printCOFFSymbolTable(Coff); 1637 return; 1638 } 1639 1640 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1641 // Skip printing the special zero symbol when dumping an ELF file. 1642 // This makes the output consistent with the GNU objdump. 1643 if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O)) 1644 continue; 1645 1646 const SymbolRef &Symbol = *I; 1647 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1648 if (!AddressOrError) 1649 report_error(ArchiveName, O->getFileName(), AddressOrError.takeError(), 1650 ArchitectureName); 1651 uint64_t Address = *AddressOrError; 1652 if ((Address < StartAddress) || (Address > StopAddress)) 1653 continue; 1654 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1655 if (!TypeOrError) 1656 report_error(ArchiveName, O->getFileName(), TypeOrError.takeError(), 1657 ArchitectureName); 1658 SymbolRef::Type Type = *TypeOrError; 1659 uint32_t Flags = Symbol.getFlags(); 1660 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1661 if (!SectionOrErr) 1662 report_error(ArchiveName, O->getFileName(), SectionOrErr.takeError(), 1663 ArchitectureName); 1664 section_iterator Section = *SectionOrErr; 1665 StringRef Name; 1666 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 1667 Section->getName(Name); 1668 } else { 1669 Expected<StringRef> NameOrErr = Symbol.getName(); 1670 if (!NameOrErr) 1671 report_error(ArchiveName, O->getFileName(), NameOrErr.takeError(), 1672 ArchitectureName); 1673 Name = *NameOrErr; 1674 } 1675 1676 bool Global = Flags & SymbolRef::SF_Global; 1677 bool Weak = Flags & SymbolRef::SF_Weak; 1678 bool Absolute = Flags & SymbolRef::SF_Absolute; 1679 bool Common = Flags & SymbolRef::SF_Common; 1680 bool Hidden = Flags & SymbolRef::SF_Hidden; 1681 1682 char GlobLoc = ' '; 1683 if (Type != SymbolRef::ST_Unknown) 1684 GlobLoc = Global ? 'g' : 'l'; 1685 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1686 ? 'd' : ' '; 1687 char FileFunc = ' '; 1688 if (Type == SymbolRef::ST_File) 1689 FileFunc = 'f'; 1690 else if (Type == SymbolRef::ST_Function) 1691 FileFunc = 'F'; 1692 else if (Type == SymbolRef::ST_Data) 1693 FileFunc = 'O'; 1694 1695 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1696 "%08" PRIx64; 1697 1698 outs() << format(Fmt, Address) << " " 1699 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1700 << (Weak ? 'w' : ' ') // Weak? 1701 << ' ' // Constructor. Not supported yet. 1702 << ' ' // Warning. Not supported yet. 1703 << ' ' // Indirect reference to another symbol. 1704 << Debug // Debugging (d) or dynamic (D) symbol. 1705 << FileFunc // Name of function (F), file (f) or object (O). 1706 << ' '; 1707 if (Absolute) { 1708 outs() << "*ABS*"; 1709 } else if (Common) { 1710 outs() << "*COM*"; 1711 } else if (Section == O->section_end()) { 1712 outs() << "*UND*"; 1713 } else { 1714 if (const MachOObjectFile *MachO = 1715 dyn_cast<const MachOObjectFile>(O)) { 1716 DataRefImpl DR = Section->getRawDataRefImpl(); 1717 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1718 outs() << SegmentName << ","; 1719 } 1720 StringRef SectionName; 1721 error(Section->getName(SectionName)); 1722 outs() << SectionName; 1723 } 1724 1725 outs() << '\t'; 1726 if (Common || isa<ELFObjectFileBase>(O)) { 1727 uint64_t Val = 1728 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1729 outs() << format("\t %08" PRIx64 " ", Val); 1730 } 1731 1732 if (Hidden) 1733 outs() << ".hidden "; 1734 1735 if (Demangle) 1736 outs() << demangle(Name) << '\n'; 1737 else 1738 outs() << Name << '\n'; 1739 } 1740 } 1741 1742 static void printUnwindInfo(const ObjectFile *O) { 1743 outs() << "Unwind info:\n\n"; 1744 1745 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1746 printCOFFUnwindInfo(Coff); 1747 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1748 printMachOUnwindInfo(MachO); 1749 else 1750 // TODO: Extract DWARF dump tool to objdump. 1751 WithColor::error(errs(), ToolName) 1752 << "This operation is only currently supported " 1753 "for COFF and MachO object files.\n"; 1754 } 1755 1756 void llvm::printExportsTrie(const ObjectFile *o) { 1757 outs() << "Exports trie:\n"; 1758 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1759 printMachOExportsTrie(MachO); 1760 else 1761 WithColor::error(errs(), ToolName) 1762 << "This operation is only currently supported " 1763 "for Mach-O executable files.\n"; 1764 } 1765 1766 void llvm::printRebaseTable(ObjectFile *o) { 1767 outs() << "Rebase table:\n"; 1768 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1769 printMachORebaseTable(MachO); 1770 else 1771 WithColor::error(errs(), ToolName) 1772 << "This operation is only currently supported " 1773 "for Mach-O executable files.\n"; 1774 } 1775 1776 void llvm::printBindTable(ObjectFile *o) { 1777 outs() << "Bind table:\n"; 1778 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1779 printMachOBindTable(MachO); 1780 else 1781 WithColor::error(errs(), ToolName) 1782 << "This operation is only currently supported " 1783 "for Mach-O executable files.\n"; 1784 } 1785 1786 void llvm::printLazyBindTable(ObjectFile *o) { 1787 outs() << "Lazy bind table:\n"; 1788 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1789 printMachOLazyBindTable(MachO); 1790 else 1791 WithColor::error(errs(), ToolName) 1792 << "This operation is only currently supported " 1793 "for Mach-O executable files.\n"; 1794 } 1795 1796 void llvm::printWeakBindTable(ObjectFile *o) { 1797 outs() << "Weak bind table:\n"; 1798 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1799 printMachOWeakBindTable(MachO); 1800 else 1801 WithColor::error(errs(), ToolName) 1802 << "This operation is only currently supported " 1803 "for Mach-O executable files.\n"; 1804 } 1805 1806 /// Dump the raw contents of the __clangast section so the output can be piped 1807 /// into llvm-bcanalyzer. 1808 void llvm::printRawClangAST(const ObjectFile *Obj) { 1809 if (outs().is_displayed()) { 1810 WithColor::error(errs(), ToolName) 1811 << "The -raw-clang-ast option will dump the raw binary contents of " 1812 "the clang ast section.\n" 1813 "Please redirect the output to a file or another program such as " 1814 "llvm-bcanalyzer.\n"; 1815 return; 1816 } 1817 1818 StringRef ClangASTSectionName("__clangast"); 1819 if (isa<COFFObjectFile>(Obj)) { 1820 ClangASTSectionName = "clangast"; 1821 } 1822 1823 Optional<object::SectionRef> ClangASTSection; 1824 for (auto Sec : ToolSectionFilter(*Obj)) { 1825 StringRef Name; 1826 Sec.getName(Name); 1827 if (Name == ClangASTSectionName) { 1828 ClangASTSection = Sec; 1829 break; 1830 } 1831 } 1832 if (!ClangASTSection) 1833 return; 1834 1835 StringRef ClangASTContents; 1836 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1837 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1838 } 1839 1840 static void printFaultMaps(const ObjectFile *Obj) { 1841 StringRef FaultMapSectionName; 1842 1843 if (isa<ELFObjectFileBase>(Obj)) { 1844 FaultMapSectionName = ".llvm_faultmaps"; 1845 } else if (isa<MachOObjectFile>(Obj)) { 1846 FaultMapSectionName = "__llvm_faultmaps"; 1847 } else { 1848 WithColor::error(errs(), ToolName) 1849 << "This operation is only currently supported " 1850 "for ELF and Mach-O executable files.\n"; 1851 return; 1852 } 1853 1854 Optional<object::SectionRef> FaultMapSection; 1855 1856 for (auto Sec : ToolSectionFilter(*Obj)) { 1857 StringRef Name; 1858 Sec.getName(Name); 1859 if (Name == FaultMapSectionName) { 1860 FaultMapSection = Sec; 1861 break; 1862 } 1863 } 1864 1865 outs() << "FaultMap table:\n"; 1866 1867 if (!FaultMapSection.hasValue()) { 1868 outs() << "<not found>\n"; 1869 return; 1870 } 1871 1872 StringRef FaultMapContents; 1873 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1874 1875 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1876 FaultMapContents.bytes_end()); 1877 1878 outs() << FMP; 1879 } 1880 1881 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1882 if (O->isELF()) { 1883 printELFFileHeader(O); 1884 return printELFDynamicSection(O); 1885 } 1886 if (O->isCOFF()) 1887 return printCOFFFileHeader(O); 1888 if (O->isWasm()) 1889 return printWasmFileHeader(O); 1890 if (O->isMachO()) { 1891 printMachOFileHeader(O); 1892 if (!OnlyFirst) 1893 printMachOLoadCommands(O); 1894 return; 1895 } 1896 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1897 } 1898 1899 static void printFileHeaders(const ObjectFile *O) { 1900 if (!O->isELF() && !O->isCOFF()) 1901 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1902 1903 Triple::ArchType AT = O->getArch(); 1904 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1905 Expected<uint64_t> StartAddrOrErr = O->getStartAddress(); 1906 if (!StartAddrOrErr) 1907 report_error(O->getFileName(), StartAddrOrErr.takeError()); 1908 1909 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1910 uint64_t Address = StartAddrOrErr.get(); 1911 outs() << "start address: " 1912 << "0x" << format(Fmt.data(), Address) << "\n\n"; 1913 } 1914 1915 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 1916 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 1917 if (!ModeOrErr) { 1918 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 1919 consumeError(ModeOrErr.takeError()); 1920 return; 1921 } 1922 sys::fs::perms Mode = ModeOrErr.get(); 1923 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1924 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1925 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1926 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1927 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1928 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1929 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1930 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1931 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1932 1933 outs() << " "; 1934 1935 Expected<unsigned> UIDOrErr = C.getUID(); 1936 if (!UIDOrErr) 1937 report_error(Filename, UIDOrErr.takeError()); 1938 unsigned UID = UIDOrErr.get(); 1939 outs() << format("%d/", UID); 1940 1941 Expected<unsigned> GIDOrErr = C.getGID(); 1942 if (!GIDOrErr) 1943 report_error(Filename, GIDOrErr.takeError()); 1944 unsigned GID = GIDOrErr.get(); 1945 outs() << format("%-d ", GID); 1946 1947 Expected<uint64_t> Size = C.getRawSize(); 1948 if (!Size) 1949 report_error(Filename, Size.takeError()); 1950 outs() << format("%6" PRId64, Size.get()) << " "; 1951 1952 StringRef RawLastModified = C.getRawLastModified(); 1953 unsigned Seconds; 1954 if (RawLastModified.getAsInteger(10, Seconds)) 1955 outs() << "(date: \"" << RawLastModified 1956 << "\" contains non-decimal chars) "; 1957 else { 1958 // Since ctime(3) returns a 26 character string of the form: 1959 // "Sun Sep 16 01:03:52 1973\n\0" 1960 // just print 24 characters. 1961 time_t t = Seconds; 1962 outs() << format("%.24s ", ctime(&t)); 1963 } 1964 1965 StringRef Name = ""; 1966 Expected<StringRef> NameOrErr = C.getName(); 1967 if (!NameOrErr) { 1968 consumeError(NameOrErr.takeError()); 1969 Expected<StringRef> RawNameOrErr = C.getRawName(); 1970 if (!RawNameOrErr) 1971 report_error(Filename, NameOrErr.takeError()); 1972 Name = RawNameOrErr.get(); 1973 } else { 1974 Name = NameOrErr.get(); 1975 } 1976 outs() << Name << "\n"; 1977 } 1978 1979 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 1980 const Archive::Child *C = nullptr) { 1981 // Avoid other output when using a raw option. 1982 if (!RawClangAST) { 1983 outs() << '\n'; 1984 if (A) 1985 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 1986 else 1987 outs() << O->getFileName(); 1988 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 1989 } 1990 1991 StringRef ArchiveName = A ? A->getFileName() : ""; 1992 if (FileHeaders) 1993 printFileHeaders(O); 1994 if (ArchiveHeaders && !MachOOpt && C) 1995 printArchiveChild(ArchiveName, *C); 1996 if (Disassemble) 1997 disassembleObject(O, Relocations); 1998 if (Relocations && !Disassemble) 1999 printRelocations(O); 2000 if (DynamicRelocations) 2001 printDynamicRelocations(O); 2002 if (SectionHeaders) 2003 printSectionHeaders(O); 2004 if (SectionContents) 2005 printSectionContents(O); 2006 if (SymbolTable) 2007 printSymbolTable(O, ArchiveName); 2008 if (UnwindInfo) 2009 printUnwindInfo(O); 2010 if (PrivateHeaders || FirstPrivateHeader) 2011 printPrivateFileHeaders(O, FirstPrivateHeader); 2012 if (ExportsTrie) 2013 printExportsTrie(O); 2014 if (Rebase) 2015 printRebaseTable(O); 2016 if (Bind) 2017 printBindTable(O); 2018 if (LazyBind) 2019 printLazyBindTable(O); 2020 if (WeakBind) 2021 printWeakBindTable(O); 2022 if (RawClangAST) 2023 printRawClangAST(O); 2024 if (PrintFaultMaps) 2025 printFaultMaps(O); 2026 if (DwarfDumpType != DIDT_Null) { 2027 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2028 // Dump the complete DWARF structure. 2029 DIDumpOptions DumpOpts; 2030 DumpOpts.DumpType = DwarfDumpType; 2031 DICtx->dump(outs(), DumpOpts); 2032 } 2033 } 2034 2035 static void dumpObject(const COFFImportFile *I, const Archive *A, 2036 const Archive::Child *C = nullptr) { 2037 StringRef ArchiveName = A ? A->getFileName() : ""; 2038 2039 // Avoid other output when using a raw option. 2040 if (!RawClangAST) 2041 outs() << '\n' 2042 << ArchiveName << "(" << I->getFileName() << ")" 2043 << ":\tfile format COFF-import-file" 2044 << "\n\n"; 2045 2046 if (ArchiveHeaders && !MachOOpt && C) 2047 printArchiveChild(ArchiveName, *C); 2048 if (SymbolTable) 2049 printCOFFSymbolTable(I); 2050 } 2051 2052 /// Dump each object file in \a a; 2053 static void dumpArchive(const Archive *A) { 2054 Error Err = Error::success(); 2055 for (auto &C : A->children(Err)) { 2056 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2057 if (!ChildOrErr) { 2058 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2059 report_error(A->getFileName(), C, std::move(E)); 2060 continue; 2061 } 2062 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2063 dumpObject(O, A, &C); 2064 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2065 dumpObject(I, A, &C); 2066 else 2067 report_error(A->getFileName(), object_error::invalid_file_type); 2068 } 2069 if (Err) 2070 report_error(A->getFileName(), std::move(Err)); 2071 } 2072 2073 /// Open file and figure out how to dump it. 2074 static void dumpInput(StringRef file) { 2075 // If we are using the Mach-O specific object file parser, then let it parse 2076 // the file and process the command line options. So the -arch flags can 2077 // be used to select specific slices, etc. 2078 if (MachOOpt) { 2079 parseInputMachO(file); 2080 return; 2081 } 2082 2083 // Attempt to open the binary. 2084 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2085 if (!BinaryOrErr) 2086 report_error(file, BinaryOrErr.takeError()); 2087 Binary &Binary = *BinaryOrErr.get().getBinary(); 2088 2089 if (Archive *A = dyn_cast<Archive>(&Binary)) 2090 dumpArchive(A); 2091 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2092 dumpObject(O); 2093 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2094 parseInputMachO(UB); 2095 else 2096 report_error(file, object_error::invalid_file_type); 2097 } 2098 2099 int main(int argc, char **argv) { 2100 InitLLVM X(argc, argv); 2101 2102 // Initialize targets and assembly printers/parsers. 2103 llvm::InitializeAllTargetInfos(); 2104 llvm::InitializeAllTargetMCs(); 2105 llvm::InitializeAllDisassemblers(); 2106 2107 // Register the target printer for --version. 2108 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2109 2110 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2111 2112 ToolName = argv[0]; 2113 2114 // Defaults to a.out if no filenames specified. 2115 if (InputFilenames.empty()) 2116 InputFilenames.push_back("a.out"); 2117 2118 if (AllHeaders) 2119 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2120 SectionHeaders = SymbolTable = true; 2121 2122 if (DisassembleAll || PrintSource || PrintLines) 2123 Disassemble = true; 2124 2125 if (!Disassemble 2126 && !Relocations 2127 && !DynamicRelocations 2128 && !SectionHeaders 2129 && !SectionContents 2130 && !SymbolTable 2131 && !UnwindInfo 2132 && !PrivateHeaders 2133 && !FileHeaders 2134 && !FirstPrivateHeader 2135 && !ExportsTrie 2136 && !Rebase 2137 && !Bind 2138 && !LazyBind 2139 && !WeakBind 2140 && !RawClangAST 2141 && !(UniversalHeaders && MachOOpt) 2142 && !ArchiveHeaders 2143 && !(IndirectSymbols && MachOOpt) 2144 && !(DataInCode && MachOOpt) 2145 && !(LinkOptHints && MachOOpt) 2146 && !(InfoPlist && MachOOpt) 2147 && !(DylibsUsed && MachOOpt) 2148 && !(DylibId && MachOOpt) 2149 && !(ObjcMetaData && MachOOpt) 2150 && !(!FilterSections.empty() && MachOOpt) 2151 && !PrintFaultMaps 2152 && DwarfDumpType == DIDT_Null) { 2153 cl::PrintHelpMessage(); 2154 return 2; 2155 } 2156 2157 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2158 DisassembleFunctions.end()); 2159 2160 llvm::for_each(InputFilenames, dumpInput); 2161 2162 return EXIT_SUCCESS; 2163 } 2164