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/SetOperations.h" 22 #include "llvm/ADT/StringExtras.h" 23 #include "llvm/ADT/StringSet.h" 24 #include "llvm/ADT/Triple.h" 25 #include "llvm/CodeGen/FaultMaps.h" 26 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 27 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 28 #include "llvm/Demangle/Demangle.h" 29 #include "llvm/MC/MCAsmInfo.h" 30 #include "llvm/MC/MCContext.h" 31 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 32 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 33 #include "llvm/MC/MCInst.h" 34 #include "llvm/MC/MCInstPrinter.h" 35 #include "llvm/MC/MCInstrAnalysis.h" 36 #include "llvm/MC/MCInstrInfo.h" 37 #include "llvm/MC/MCObjectFileInfo.h" 38 #include "llvm/MC/MCRegisterInfo.h" 39 #include "llvm/MC/MCSubtargetInfo.h" 40 #include "llvm/Object/Archive.h" 41 #include "llvm/Object/COFF.h" 42 #include "llvm/Object/COFFImportFile.h" 43 #include "llvm/Object/ELFObjectFile.h" 44 #include "llvm/Object/MachO.h" 45 #include "llvm/Object/MachOUniversal.h" 46 #include "llvm/Object/ObjectFile.h" 47 #include "llvm/Object/Wasm.h" 48 #include "llvm/Support/Casting.h" 49 #include "llvm/Support/CommandLine.h" 50 #include "llvm/Support/Debug.h" 51 #include "llvm/Support/Errc.h" 52 #include "llvm/Support/FileSystem.h" 53 #include "llvm/Support/Format.h" 54 #include "llvm/Support/GraphWriter.h" 55 #include "llvm/Support/Host.h" 56 #include "llvm/Support/InitLLVM.h" 57 #include "llvm/Support/MemoryBuffer.h" 58 #include "llvm/Support/SourceMgr.h" 59 #include "llvm/Support/StringSaver.h" 60 #include "llvm/Support/TargetRegistry.h" 61 #include "llvm/Support/TargetSelect.h" 62 #include "llvm/Support/WithColor.h" 63 #include "llvm/Support/raw_ostream.h" 64 #include <algorithm> 65 #include <cctype> 66 #include <cstring> 67 #include <system_error> 68 #include <unordered_map> 69 #include <utility> 70 71 using namespace llvm::object; 72 73 namespace llvm { 74 75 cl::OptionCategory ObjdumpCat("llvm-objdump Options"); 76 77 // MachO specific 78 extern cl::OptionCategory MachOCat; 79 extern cl::opt<bool> Bind; 80 extern cl::opt<bool> DataInCode; 81 extern cl::opt<bool> DylibsUsed; 82 extern cl::opt<bool> DylibId; 83 extern cl::opt<bool> ExportsTrie; 84 extern cl::opt<bool> FirstPrivateHeader; 85 extern cl::opt<bool> IndirectSymbols; 86 extern cl::opt<bool> InfoPlist; 87 extern cl::opt<bool> LazyBind; 88 extern cl::opt<bool> LinkOptHints; 89 extern cl::opt<bool> ObjcMetaData; 90 extern cl::opt<bool> Rebase; 91 extern cl::opt<bool> UniversalHeaders; 92 extern cl::opt<bool> WeakBind; 93 94 static cl::opt<uint64_t> AdjustVMA( 95 "adjust-vma", 96 cl::desc("Increase the displayed address by the specified offset"), 97 cl::value_desc("offset"), cl::init(0), cl::cat(ObjdumpCat)); 98 99 static cl::opt<bool> 100 AllHeaders("all-headers", 101 cl::desc("Display all available header information"), 102 cl::cat(ObjdumpCat)); 103 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 104 cl::NotHidden, cl::Grouping, 105 cl::aliasopt(AllHeaders)); 106 107 static cl::opt<std::string> 108 ArchName("arch-name", 109 cl::desc("Target arch to disassemble for, " 110 "see -version for available targets"), 111 cl::cat(ObjdumpCat)); 112 113 cl::opt<bool> ArchiveHeaders("archive-headers", 114 cl::desc("Display archive header information"), 115 cl::cat(ObjdumpCat)); 116 static cl::alias ArchiveHeadersShort("a", 117 cl::desc("Alias for --archive-headers"), 118 cl::NotHidden, cl::Grouping, 119 cl::aliasopt(ArchiveHeaders)); 120 121 cl::opt<bool> Demangle("demangle", cl::desc("Demangle symbols names"), 122 cl::init(false), cl::cat(ObjdumpCat)); 123 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 124 cl::NotHidden, cl::Grouping, 125 cl::aliasopt(Demangle)); 126 127 cl::opt<bool> Disassemble( 128 "disassemble", 129 cl::desc("Display assembler mnemonics for the machine instructions"), 130 cl::cat(ObjdumpCat)); 131 static cl::alias DisassembleShort("d", cl::desc("Alias for --disassemble"), 132 cl::NotHidden, cl::Grouping, 133 cl::aliasopt(Disassemble)); 134 135 cl::opt<bool> DisassembleAll( 136 "disassemble-all", 137 cl::desc("Display assembler mnemonics for the machine instructions"), 138 cl::cat(ObjdumpCat)); 139 static cl::alias DisassembleAllShort("D", 140 cl::desc("Alias for --disassemble-all"), 141 cl::NotHidden, cl::Grouping, 142 cl::aliasopt(DisassembleAll)); 143 144 static cl::list<std::string> 145 DisassembleFunctions("disassemble-functions", cl::CommaSeparated, 146 cl::desc("List of functions to disassemble"), 147 cl::cat(ObjdumpCat)); 148 149 static cl::opt<bool> DisassembleZeroes( 150 "disassemble-zeroes", 151 cl::desc("Do not skip blocks of zeroes when disassembling"), 152 cl::cat(ObjdumpCat)); 153 static cl::alias 154 DisassembleZeroesShort("z", cl::desc("Alias for --disassemble-zeroes"), 155 cl::NotHidden, cl::Grouping, 156 cl::aliasopt(DisassembleZeroes)); 157 158 static cl::list<std::string> 159 DisassemblerOptions("disassembler-options", 160 cl::desc("Pass target specific disassembler options"), 161 cl::value_desc("options"), cl::CommaSeparated, 162 cl::cat(ObjdumpCat)); 163 static cl::alias 164 DisassemblerOptionsShort("M", cl::desc("Alias for --disassembler-options"), 165 cl::NotHidden, cl::Grouping, cl::Prefix, 166 cl::CommaSeparated, 167 cl::aliasopt(DisassemblerOptions)); 168 169 cl::opt<DIDumpType> DwarfDumpType( 170 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 171 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame")), 172 cl::cat(ObjdumpCat)); 173 174 static cl::opt<bool> DynamicRelocations( 175 "dynamic-reloc", 176 cl::desc("Display the dynamic relocation entries in the file"), 177 cl::cat(ObjdumpCat)); 178 static cl::alias DynamicRelocationShort("R", 179 cl::desc("Alias for --dynamic-reloc"), 180 cl::NotHidden, cl::Grouping, 181 cl::aliasopt(DynamicRelocations)); 182 183 static cl::opt<bool> 184 FaultMapSection("fault-map-section", 185 cl::desc("Display contents of faultmap section"), 186 cl::cat(ObjdumpCat)); 187 188 static cl::opt<bool> 189 FileHeaders("file-headers", 190 cl::desc("Display the contents of the overall file header"), 191 cl::cat(ObjdumpCat)); 192 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 193 cl::NotHidden, cl::Grouping, 194 cl::aliasopt(FileHeaders)); 195 196 cl::opt<bool> SectionContents("full-contents", 197 cl::desc("Display the content of each section"), 198 cl::cat(ObjdumpCat)); 199 static cl::alias SectionContentsShort("s", 200 cl::desc("Alias for --full-contents"), 201 cl::NotHidden, cl::Grouping, 202 cl::aliasopt(SectionContents)); 203 204 static cl::list<std::string> InputFilenames(cl::Positional, 205 cl::desc("<input object files>"), 206 cl::ZeroOrMore, 207 cl::cat(ObjdumpCat)); 208 209 static cl::opt<bool> 210 PrintLines("line-numbers", 211 cl::desc("Display source line numbers with " 212 "disassembly. Implies disassemble object"), 213 cl::cat(ObjdumpCat)); 214 static cl::alias PrintLinesShort("l", cl::desc("Alias for --line-numbers"), 215 cl::NotHidden, cl::Grouping, 216 cl::aliasopt(PrintLines)); 217 218 static cl::opt<bool> MachOOpt("macho", 219 cl::desc("Use MachO specific object file parser"), 220 cl::cat(ObjdumpCat)); 221 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 222 cl::Grouping, cl::aliasopt(MachOOpt)); 223 224 cl::opt<std::string> 225 MCPU("mcpu", 226 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 227 cl::value_desc("cpu-name"), cl::init(""), cl::cat(ObjdumpCat)); 228 229 cl::list<std::string> MAttrs("mattr", cl::CommaSeparated, 230 cl::desc("Target specific attributes"), 231 cl::value_desc("a1,+a2,-a3,..."), 232 cl::cat(ObjdumpCat)); 233 234 cl::opt<bool> NoShowRawInsn("no-show-raw-insn", 235 cl::desc("When disassembling " 236 "instructions, do not print " 237 "the instruction bytes."), 238 cl::cat(ObjdumpCat)); 239 cl::opt<bool> NoLeadingAddr("no-leading-addr", 240 cl::desc("Print no leading address"), 241 cl::cat(ObjdumpCat)); 242 243 static cl::opt<bool> RawClangAST( 244 "raw-clang-ast", 245 cl::desc("Dump the raw binary contents of the clang AST section"), 246 cl::cat(ObjdumpCat)); 247 248 cl::opt<bool> 249 Relocations("reloc", cl::desc("Display the relocation entries in the file"), 250 cl::cat(ObjdumpCat)); 251 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 252 cl::NotHidden, cl::Grouping, 253 cl::aliasopt(Relocations)); 254 255 cl::opt<bool> PrintImmHex("print-imm-hex", 256 cl::desc("Use hex format for immediate values"), 257 cl::cat(ObjdumpCat)); 258 259 cl::opt<bool> PrivateHeaders("private-headers", 260 cl::desc("Display format specific file headers"), 261 cl::cat(ObjdumpCat)); 262 static cl::alias PrivateHeadersShort("p", 263 cl::desc("Alias for --private-headers"), 264 cl::NotHidden, cl::Grouping, 265 cl::aliasopt(PrivateHeaders)); 266 267 cl::list<std::string> 268 FilterSections("section", 269 cl::desc("Operate on the specified sections only. " 270 "With -macho dump segment,section"), 271 cl::cat(ObjdumpCat)); 272 static cl::alias FilterSectionsj("j", cl::desc("Alias for --section"), 273 cl::NotHidden, cl::Grouping, cl::Prefix, 274 cl::aliasopt(FilterSections)); 275 276 cl::opt<bool> SectionHeaders("section-headers", 277 cl::desc("Display summaries of the " 278 "headers for each section."), 279 cl::cat(ObjdumpCat)); 280 static cl::alias SectionHeadersShort("headers", 281 cl::desc("Alias for --section-headers"), 282 cl::NotHidden, 283 cl::aliasopt(SectionHeaders)); 284 static cl::alias SectionHeadersShorter("h", 285 cl::desc("Alias for --section-headers"), 286 cl::NotHidden, cl::Grouping, 287 cl::aliasopt(SectionHeaders)); 288 289 static cl::opt<bool> 290 ShowLMA("show-lma", 291 cl::desc("Display LMA column when dumping ELF section headers"), 292 cl::cat(ObjdumpCat)); 293 294 static cl::opt<bool> PrintSource( 295 "source", 296 cl::desc( 297 "Display source inlined with disassembly. Implies disassemble object"), 298 cl::cat(ObjdumpCat)); 299 static cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 300 cl::NotHidden, cl::Grouping, 301 cl::aliasopt(PrintSource)); 302 303 static cl::opt<uint64_t> 304 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 305 cl::value_desc("address"), cl::init(0), cl::cat(ObjdumpCat)); 306 static cl::opt<uint64_t> StopAddress("stop-address", 307 cl::desc("Stop disassembly at address"), 308 cl::value_desc("address"), 309 cl::init(UINT64_MAX), cl::cat(ObjdumpCat)); 310 311 cl::opt<bool> SymbolTable("syms", cl::desc("Display the symbol table"), 312 cl::cat(ObjdumpCat)); 313 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 314 cl::NotHidden, cl::Grouping, 315 cl::aliasopt(SymbolTable)); 316 317 cl::opt<std::string> TripleName("triple", 318 cl::desc("Target triple to disassemble for, " 319 "see -version for available targets"), 320 cl::cat(ObjdumpCat)); 321 322 cl::opt<bool> UnwindInfo("unwind-info", cl::desc("Display unwind information"), 323 cl::cat(ObjdumpCat)); 324 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 325 cl::NotHidden, cl::Grouping, 326 cl::aliasopt(UnwindInfo)); 327 328 static cl::opt<bool> 329 Wide("wide", cl::desc("Ignored for compatibility with GNU objdump"), 330 cl::cat(ObjdumpCat)); 331 static cl::alias WideShort("w", cl::Grouping, cl::aliasopt(Wide)); 332 333 static StringSet<> DisasmFuncsSet; 334 static StringRef ToolName; 335 336 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 337 338 static bool shouldKeep(object::SectionRef S) { 339 if (FilterSections.empty()) 340 return true; 341 StringRef String; 342 std::error_code error = S.getName(String); 343 if (error) 344 return false; 345 return is_contained(FilterSections, String); 346 } 347 348 SectionFilter ToolSectionFilter(object::ObjectFile const &O) { 349 return SectionFilter([](object::SectionRef S) { return shouldKeep(S); }, O); 350 } 351 352 void error(std::error_code EC) { 353 if (!EC) 354 return; 355 WithColor::error(errs(), ToolName) 356 << "reading file: " << EC.message() << ".\n"; 357 errs().flush(); 358 exit(1); 359 } 360 361 void error(Error E) { 362 if (!E) 363 return; 364 WithColor::error(errs(), ToolName) << toString(std::move(E)); 365 exit(1); 366 } 367 368 LLVM_ATTRIBUTE_NORETURN void error(Twine Message) { 369 WithColor::error(errs(), ToolName) << Message << ".\n"; 370 errs().flush(); 371 exit(1); 372 } 373 374 void warn(StringRef Message) { 375 WithColor::warning(errs(), ToolName) << Message << ".\n"; 376 errs().flush(); 377 } 378 379 void warn(Twine Message) { 380 WithColor::warning(errs(), ToolName) << Message << "\n"; 381 } 382 383 LLVM_ATTRIBUTE_NORETURN void report_error(StringRef File, Twine Message) { 384 WithColor::error(errs(), ToolName) 385 << "'" << File << "': " << Message << ".\n"; 386 exit(1); 387 } 388 389 LLVM_ATTRIBUTE_NORETURN void report_error(Error E, StringRef File) { 390 assert(E); 391 std::string Buf; 392 raw_string_ostream OS(Buf); 393 logAllUnhandledErrors(std::move(E), OS); 394 OS.flush(); 395 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 396 exit(1); 397 } 398 399 LLVM_ATTRIBUTE_NORETURN void report_error(Error E, StringRef ArchiveName, 400 StringRef FileName, 401 StringRef ArchitectureName) { 402 assert(E); 403 WithColor::error(errs(), ToolName); 404 if (ArchiveName != "") 405 errs() << ArchiveName << "(" << FileName << ")"; 406 else 407 errs() << "'" << FileName << "'"; 408 if (!ArchitectureName.empty()) 409 errs() << " (for architecture " << ArchitectureName << ")"; 410 std::string Buf; 411 raw_string_ostream OS(Buf); 412 logAllUnhandledErrors(std::move(E), OS); 413 OS.flush(); 414 errs() << ": " << Buf; 415 exit(1); 416 } 417 418 LLVM_ATTRIBUTE_NORETURN void report_error(Error E, StringRef ArchiveName, 419 const object::Archive::Child &C, 420 StringRef ArchitectureName) { 421 Expected<StringRef> NameOrErr = C.getName(); 422 // TODO: if we have a error getting the name then it would be nice to print 423 // the index of which archive member this is and or its offset in the 424 // archive instead of "???" as the name. 425 if (!NameOrErr) { 426 consumeError(NameOrErr.takeError()); 427 report_error(std::move(E), ArchiveName, "???", ArchitectureName); 428 } else 429 report_error(std::move(E), ArchiveName, NameOrErr.get(), ArchitectureName); 430 } 431 432 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 433 // Figure out the target triple. 434 Triple TheTriple("unknown-unknown-unknown"); 435 if (TripleName.empty()) { 436 if (Obj) 437 TheTriple = Obj->makeTriple(); 438 } else { 439 TheTriple.setTriple(Triple::normalize(TripleName)); 440 441 // Use the triple, but also try to combine with ARM build attributes. 442 if (Obj) { 443 auto Arch = Obj->getArch(); 444 if (Arch == Triple::arm || Arch == Triple::armeb) 445 Obj->setARMSubArch(TheTriple); 446 } 447 } 448 449 // Get the target specific parser. 450 std::string Error; 451 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 452 Error); 453 if (!TheTarget) { 454 if (Obj) 455 report_error(Obj->getFileName(), "can't find target: " + Error); 456 else 457 error("can't find target: " + Error); 458 } 459 460 // Update the triple name and return the found target. 461 TripleName = TheTriple.getTriple(); 462 return TheTarget; 463 } 464 465 bool isRelocAddressLess(RelocationRef A, RelocationRef B) { 466 return A.getOffset() < B.getOffset(); 467 } 468 469 static Error getRelocationValueString(const RelocationRef &Rel, 470 SmallVectorImpl<char> &Result) { 471 const ObjectFile *Obj = Rel.getObject(); 472 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 473 return getELFRelocationValueString(ELF, Rel, Result); 474 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 475 return getCOFFRelocationValueString(COFF, Rel, Result); 476 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 477 return getWasmRelocationValueString(Wasm, Rel, Result); 478 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 479 return getMachORelocationValueString(MachO, Rel, Result); 480 llvm_unreachable("unknown object file format"); 481 } 482 483 /// Indicates whether this relocation should hidden when listing 484 /// relocations, usually because it is the trailing part of a multipart 485 /// relocation that will be printed as part of the leading relocation. 486 static bool getHidden(RelocationRef RelRef) { 487 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 488 if (!MachO) 489 return false; 490 491 unsigned Arch = MachO->getArch(); 492 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 493 uint64_t Type = MachO->getRelocationType(Rel); 494 495 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 496 // is always hidden. 497 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 498 return Type == MachO::GENERIC_RELOC_PAIR; 499 500 if (Arch == Triple::x86_64) { 501 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 502 // an X86_64_RELOC_SUBTRACTOR. 503 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 504 DataRefImpl RelPrev = Rel; 505 RelPrev.d.a--; 506 uint64_t PrevType = MachO->getRelocationType(RelPrev); 507 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 508 return true; 509 } 510 } 511 512 return false; 513 } 514 515 namespace { 516 class SourcePrinter { 517 protected: 518 DILineInfo OldLineInfo; 519 const ObjectFile *Obj = nullptr; 520 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 521 // File name to file contents of source 522 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 523 // Mark the line endings of the cached source 524 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 525 526 private: 527 bool cacheSource(const DILineInfo& LineInfoFile); 528 529 public: 530 SourcePrinter() = default; 531 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 532 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 533 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 534 DefaultArch); 535 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 536 } 537 virtual ~SourcePrinter() = default; 538 virtual void printSourceLine(raw_ostream &OS, 539 object::SectionedAddress Address, 540 StringRef Delimiter = "; "); 541 }; 542 543 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 544 std::unique_ptr<MemoryBuffer> Buffer; 545 if (LineInfo.Source) { 546 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 547 } else { 548 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 549 if (!BufferOrError) 550 return false; 551 Buffer = std::move(*BufferOrError); 552 } 553 // Chomp the file to get lines 554 const char *BufferStart = Buffer->getBufferStart(), 555 *BufferEnd = Buffer->getBufferEnd(); 556 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 557 const char *Start = BufferStart; 558 for (const char *I = BufferStart; I != BufferEnd; ++I) 559 if (*I == '\n') { 560 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 561 Start = I + 1; 562 } 563 if (Start < BufferEnd) 564 Lines.emplace_back(Start, BufferEnd - Start); 565 SourceCache[LineInfo.FileName] = std::move(Buffer); 566 return true; 567 } 568 569 void SourcePrinter::printSourceLine(raw_ostream &OS, 570 object::SectionedAddress Address, 571 StringRef Delimiter) { 572 if (!Symbolizer) 573 return; 574 DILineInfo LineInfo = DILineInfo(); 575 auto ExpectedLineInfo = 576 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 577 if (!ExpectedLineInfo) 578 consumeError(ExpectedLineInfo.takeError()); 579 else 580 LineInfo = *ExpectedLineInfo; 581 582 if ((LineInfo.FileName == "<invalid>") || LineInfo.Line == 0 || 583 ((OldLineInfo.Line == LineInfo.Line) && 584 (OldLineInfo.FileName == LineInfo.FileName))) 585 return; 586 587 if (PrintLines) 588 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 589 if (PrintSource) { 590 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 591 if (!cacheSource(LineInfo)) 592 return; 593 auto LineBuffer = LineCache.find(LineInfo.FileName); 594 if (LineBuffer != LineCache.end()) { 595 if (LineInfo.Line > LineBuffer->second.size()) 596 return; 597 // Vector begins at 0, line numbers are non-zero 598 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1] << '\n'; 599 } 600 } 601 OldLineInfo = LineInfo; 602 } 603 604 static bool isArmElf(const ObjectFile *Obj) { 605 return (Obj->isELF() && 606 (Obj->getArch() == Triple::aarch64 || 607 Obj->getArch() == Triple::aarch64_be || 608 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 609 Obj->getArch() == Triple::thumb || 610 Obj->getArch() == Triple::thumbeb)); 611 } 612 613 static void printRelocation(const RelocationRef &Rel, uint64_t Address, 614 uint8_t AddrSize) { 615 StringRef Fmt = 616 AddrSize > 4 ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 617 SmallString<16> Name; 618 SmallString<32> Val; 619 Rel.getTypeName(Name); 620 error(getRelocationValueString(Rel, Val)); 621 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n"; 622 } 623 624 class PrettyPrinter { 625 public: 626 virtual ~PrettyPrinter() = default; 627 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 628 ArrayRef<uint8_t> Bytes, 629 object::SectionedAddress Address, raw_ostream &OS, 630 StringRef Annot, MCSubtargetInfo const &STI, 631 SourcePrinter *SP, 632 std::vector<RelocationRef> *Rels = nullptr) { 633 if (SP && (PrintSource || PrintLines)) 634 SP->printSourceLine(OS, Address); 635 636 { 637 formatted_raw_ostream FOS(OS); 638 if (!NoLeadingAddr) 639 FOS << format("%8" PRIx64 ":", Address.Address); 640 if (!NoShowRawInsn) { 641 FOS << ' '; 642 dumpBytes(Bytes, FOS); 643 } 644 FOS.flush(); 645 // The output of printInst starts with a tab. Print some spaces so that 646 // the tab has 1 column and advances to the target tab stop. 647 unsigned TabStop = NoShowRawInsn ? 16 : 40; 648 unsigned Column = FOS.getColumn(); 649 FOS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8); 650 651 // The dtor calls flush() to ensure the indent comes before printInst(). 652 } 653 654 if (MI) 655 IP.printInst(MI, OS, "", STI); 656 else 657 OS << "\t<unknown>"; 658 } 659 }; 660 PrettyPrinter PrettyPrinterInst; 661 class HexagonPrettyPrinter : public PrettyPrinter { 662 public: 663 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 664 raw_ostream &OS) { 665 uint32_t opcode = 666 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 667 if (!NoLeadingAddr) 668 OS << format("%8" PRIx64 ":", Address); 669 if (!NoShowRawInsn) { 670 OS << "\t"; 671 dumpBytes(Bytes.slice(0, 4), OS); 672 OS << format("\t%08" PRIx32, opcode); 673 } 674 } 675 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 676 object::SectionedAddress Address, raw_ostream &OS, 677 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 678 std::vector<RelocationRef> *Rels) override { 679 if (SP && (PrintSource || PrintLines)) 680 SP->printSourceLine(OS, Address, ""); 681 if (!MI) { 682 printLead(Bytes, Address.Address, OS); 683 OS << " <unknown>"; 684 return; 685 } 686 std::string Buffer; 687 { 688 raw_string_ostream TempStream(Buffer); 689 IP.printInst(MI, TempStream, "", STI); 690 } 691 StringRef Contents(Buffer); 692 // Split off bundle attributes 693 auto PacketBundle = Contents.rsplit('\n'); 694 // Split off first instruction from the rest 695 auto HeadTail = PacketBundle.first.split('\n'); 696 auto Preamble = " { "; 697 auto Separator = ""; 698 699 // Hexagon's packets require relocations to be inline rather than 700 // clustered at the end of the packet. 701 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 702 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 703 auto PrintReloc = [&]() -> void { 704 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 705 if (RelCur->getOffset() == Address.Address) { 706 printRelocation(*RelCur, Address.Address, 4); 707 return; 708 } 709 ++RelCur; 710 } 711 }; 712 713 while (!HeadTail.first.empty()) { 714 OS << Separator; 715 Separator = "\n"; 716 if (SP && (PrintSource || PrintLines)) 717 SP->printSourceLine(OS, Address, ""); 718 printLead(Bytes, Address.Address, OS); 719 OS << Preamble; 720 Preamble = " "; 721 StringRef Inst; 722 auto Duplex = HeadTail.first.split('\v'); 723 if (!Duplex.second.empty()) { 724 OS << Duplex.first; 725 OS << "; "; 726 Inst = Duplex.second; 727 } 728 else 729 Inst = HeadTail.first; 730 OS << Inst; 731 HeadTail = HeadTail.second.split('\n'); 732 if (HeadTail.first.empty()) 733 OS << " } " << PacketBundle.second; 734 PrintReloc(); 735 Bytes = Bytes.slice(4); 736 Address.Address += 4; 737 } 738 } 739 }; 740 HexagonPrettyPrinter HexagonPrettyPrinterInst; 741 742 class AMDGCNPrettyPrinter : public PrettyPrinter { 743 public: 744 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 745 object::SectionedAddress Address, raw_ostream &OS, 746 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 747 std::vector<RelocationRef> *Rels) override { 748 if (SP && (PrintSource || PrintLines)) 749 SP->printSourceLine(OS, Address); 750 751 typedef support::ulittle32_t U32; 752 753 if (MI) { 754 SmallString<40> InstStr; 755 raw_svector_ostream IS(InstStr); 756 757 IP.printInst(MI, IS, "", STI); 758 759 OS << left_justify(IS.str(), 60); 760 } else { 761 // an unrecognized encoding - this is probably data so represent it 762 // using the .long directive, or .byte directive if fewer than 4 bytes 763 // remaining 764 if (Bytes.size() >= 4) { 765 OS << format("\t.long 0x%08" PRIx32 " ", 766 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 767 OS.indent(42); 768 } else { 769 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 770 for (unsigned int i = 1; i < Bytes.size(); i++) 771 OS << format(", 0x%02" PRIx8, Bytes[i]); 772 OS.indent(55 - (6 * Bytes.size())); 773 } 774 } 775 776 OS << format("// %012" PRIX64 ": ", Address.Address); 777 if (Bytes.size() >=4) { 778 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 779 Bytes.size() / sizeof(U32))) 780 // D should be explicitly casted to uint32_t here as it is passed 781 // by format to snprintf as vararg. 782 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 783 } else { 784 for (unsigned int i = 0; i < Bytes.size(); i++) 785 OS << format("%02" PRIX8 " ", Bytes[i]); 786 } 787 788 if (!Annot.empty()) 789 OS << "// " << Annot; 790 } 791 }; 792 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 793 794 class BPFPrettyPrinter : public PrettyPrinter { 795 public: 796 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 797 object::SectionedAddress Address, raw_ostream &OS, 798 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 799 std::vector<RelocationRef> *Rels) override { 800 if (SP && (PrintSource || PrintLines)) 801 SP->printSourceLine(OS, Address); 802 if (!NoLeadingAddr) 803 OS << format("%8" PRId64 ":", Address.Address / 8); 804 if (!NoShowRawInsn) { 805 OS << "\t"; 806 dumpBytes(Bytes, OS); 807 } 808 if (MI) 809 IP.printInst(MI, OS, "", STI); 810 else 811 OS << "\t<unknown>"; 812 } 813 }; 814 BPFPrettyPrinter BPFPrettyPrinterInst; 815 816 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 817 switch(Triple.getArch()) { 818 default: 819 return PrettyPrinterInst; 820 case Triple::hexagon: 821 return HexagonPrettyPrinterInst; 822 case Triple::amdgcn: 823 return AMDGCNPrettyPrinterInst; 824 case Triple::bpfel: 825 case Triple::bpfeb: 826 return BPFPrettyPrinterInst; 827 } 828 } 829 } 830 831 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 832 assert(Obj->isELF()); 833 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 834 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 835 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 836 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 837 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 838 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 839 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 840 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 841 llvm_unreachable("Unsupported binary format"); 842 } 843 844 template <class ELFT> static void 845 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 846 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 847 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 848 uint8_t SymbolType = Symbol.getELFType(); 849 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 850 continue; 851 852 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 853 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 854 if (Name.empty()) 855 continue; 856 857 section_iterator SecI = 858 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 859 if (SecI == Obj->section_end()) 860 continue; 861 862 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 863 } 864 } 865 866 static void 867 addDynamicElfSymbols(const ObjectFile *Obj, 868 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 869 assert(Obj->isELF()); 870 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 871 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 872 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 873 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 874 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 875 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 876 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 877 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 878 else 879 llvm_unreachable("Unsupported binary format"); 880 } 881 882 static void addPltEntries(const ObjectFile *Obj, 883 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 884 StringSaver &Saver) { 885 Optional<SectionRef> Plt = None; 886 for (const SectionRef &Section : Obj->sections()) { 887 StringRef Name; 888 if (Section.getName(Name)) 889 continue; 890 if (Name == ".plt") 891 Plt = Section; 892 } 893 if (!Plt) 894 return; 895 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 896 for (auto PltEntry : ElfObj->getPltAddresses()) { 897 SymbolRef Symbol(PltEntry.first, ElfObj); 898 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 899 900 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 901 if (!Name.empty()) 902 AllSymbols[*Plt].emplace_back( 903 PltEntry.second, Saver.save((Name + "@plt").str()), SymbolType); 904 } 905 } 906 } 907 908 // Normally the disassembly output will skip blocks of zeroes. This function 909 // returns the number of zero bytes that can be skipped when dumping the 910 // disassembly of the instructions in Buf. 911 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 912 // Find the number of leading zeroes. 913 size_t N = 0; 914 while (N < Buf.size() && !Buf[N]) 915 ++N; 916 917 // We may want to skip blocks of zero bytes, but unless we see 918 // at least 8 of them in a row. 919 if (N < 8) 920 return 0; 921 922 // We skip zeroes in multiples of 4 because do not want to truncate an 923 // instruction if it starts with a zero byte. 924 return N & ~0x3; 925 } 926 927 // Returns a map from sections to their relocations. 928 static std::map<SectionRef, std::vector<RelocationRef>> 929 getRelocsMap(object::ObjectFile const &Obj) { 930 std::map<SectionRef, std::vector<RelocationRef>> Ret; 931 for (SectionRef Sec : Obj.sections()) { 932 section_iterator Relocated = Sec.getRelocatedSection(); 933 if (Relocated == Obj.section_end() || !shouldKeep(*Relocated)) 934 continue; 935 std::vector<RelocationRef> &V = Ret[*Relocated]; 936 for (const RelocationRef &R : Sec.relocations()) 937 V.push_back(R); 938 // Sort relocations by address. 939 llvm::stable_sort(V, isRelocAddressLess); 940 } 941 return Ret; 942 } 943 944 // Used for --adjust-vma to check if address should be adjusted by the 945 // specified value for a given section. 946 // For ELF we do not adjust non-allocatable sections like debug ones, 947 // because they are not loadable. 948 // TODO: implement for other file formats. 949 static bool shouldAdjustVA(const SectionRef &Section) { 950 const ObjectFile *Obj = Section.getObject(); 951 if (isa<object::ELFObjectFileBase>(Obj)) 952 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 953 return false; 954 } 955 956 static uint64_t 957 dumpARMELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 958 const ObjectFile *Obj, ArrayRef<uint8_t> Bytes, 959 const std::vector<uint64_t> &TextMappingSymsAddr) { 960 support::endianness Endian = 961 Obj->isLittleEndian() ? support::little : support::big; 962 while (Index < End) { 963 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 964 outs() << "\t"; 965 if (Index + 4 <= End) { 966 dumpBytes(Bytes.slice(Index, 4), outs()); 967 outs() << "\t.word\t" 968 << format_hex( 969 support::endian::read32(Bytes.data() + Index, Endian), 10); 970 Index += 4; 971 } else if (Index + 2 <= End) { 972 dumpBytes(Bytes.slice(Index, 2), outs()); 973 outs() << "\t\t.short\t" 974 << format_hex( 975 support::endian::read16(Bytes.data() + Index, Endian), 6); 976 Index += 2; 977 } else { 978 dumpBytes(Bytes.slice(Index, 1), outs()); 979 outs() << "\t\t.byte\t" << format_hex(Bytes[0], 4); 980 ++Index; 981 } 982 outs() << "\n"; 983 if (std::binary_search(TextMappingSymsAddr.begin(), 984 TextMappingSymsAddr.end(), Index)) 985 break; 986 } 987 return Index; 988 } 989 990 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 991 ArrayRef<uint8_t> Bytes) { 992 // print out data up to 8 bytes at a time in hex and ascii 993 uint8_t AsciiData[9] = {'\0'}; 994 uint8_t Byte; 995 int NumBytes = 0; 996 997 for (; Index < End; ++Index) { 998 if (NumBytes == 0) { 999 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1000 outs() << "\t"; 1001 } 1002 Byte = Bytes.slice(Index)[0]; 1003 outs() << format(" %02x", Byte); 1004 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1005 1006 uint8_t IndentOffset = 0; 1007 NumBytes++; 1008 if (Index == End - 1 || NumBytes > 8) { 1009 // Indent the space for less than 8 bytes data. 1010 // 2 spaces for byte and one for space between bytes 1011 IndentOffset = 3 * (8 - NumBytes); 1012 for (int Excess = NumBytes; Excess < 8; Excess++) 1013 AsciiData[Excess] = '\0'; 1014 NumBytes = 8; 1015 } 1016 if (NumBytes == 8) { 1017 AsciiData[8] = '\0'; 1018 outs() << std::string(IndentOffset, ' ') << " "; 1019 outs() << reinterpret_cast<char *>(AsciiData); 1020 outs() << '\n'; 1021 NumBytes = 0; 1022 } 1023 } 1024 } 1025 1026 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 1027 MCContext &Ctx, MCDisassembler *DisAsm, 1028 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 1029 const MCSubtargetInfo *STI, PrettyPrinter &PIP, 1030 SourcePrinter &SP, bool InlineRelocs) { 1031 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 1032 if (InlineRelocs) 1033 RelocMap = getRelocsMap(*Obj); 1034 1035 // Create a mapping from virtual address to symbol name. This is used to 1036 // pretty print the symbols while disassembling. 1037 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1038 SectionSymbolsTy AbsoluteSymbols; 1039 const StringRef FileName = Obj->getFileName(); 1040 for (const SymbolRef &Symbol : Obj->symbols()) { 1041 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName); 1042 1043 StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1044 if (Name.empty()) 1045 continue; 1046 1047 uint8_t SymbolType = ELF::STT_NOTYPE; 1048 if (Obj->isELF()) { 1049 SymbolType = getElfSymbolType(Obj, Symbol); 1050 if (SymbolType == ELF::STT_SECTION) 1051 continue; 1052 } 1053 1054 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1055 if (SecI != Obj->section_end()) 1056 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 1057 else 1058 AbsoluteSymbols.emplace_back(Address, Name, SymbolType); 1059 } 1060 if (AllSymbols.empty() && Obj->isELF()) 1061 addDynamicElfSymbols(Obj, AllSymbols); 1062 1063 BumpPtrAllocator A; 1064 StringSaver Saver(A); 1065 addPltEntries(Obj, AllSymbols, Saver); 1066 1067 // Create a mapping from virtual address to section. 1068 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1069 for (SectionRef Sec : Obj->sections()) 1070 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1071 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1072 1073 // Linked executables (.exe and .dll files) typically don't include a real 1074 // symbol table but they might contain an export table. 1075 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1076 for (const auto &ExportEntry : COFFObj->export_directories()) { 1077 StringRef Name; 1078 error(ExportEntry.getSymbolName(Name)); 1079 if (Name.empty()) 1080 continue; 1081 uint32_t RVA; 1082 error(ExportEntry.getExportRVA(RVA)); 1083 1084 uint64_t VA = COFFObj->getImageBase() + RVA; 1085 auto Sec = llvm::bsearch( 1086 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &RHS) { 1087 return VA < RHS.first; 1088 }); 1089 if (Sec != SectionAddresses.begin()) { 1090 --Sec; 1091 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1092 } else 1093 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1094 } 1095 } 1096 1097 // Sort all the symbols, this allows us to use a simple binary search to find 1098 // a symbol near an address. 1099 StringSet<> FoundDisasmFuncsSet; 1100 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1101 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1102 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1103 1104 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1105 if (FilterSections.empty() && !DisassembleAll && 1106 (!Section.isText() || Section.isVirtual())) 1107 continue; 1108 1109 uint64_t SectionAddr = Section.getAddress(); 1110 uint64_t SectSize = Section.getSize(); 1111 if (!SectSize) 1112 continue; 1113 1114 // Get the list of all the symbols in this section. 1115 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1116 std::vector<uint64_t> DataMappingSymsAddr; 1117 std::vector<uint64_t> TextMappingSymsAddr; 1118 if (isArmElf(Obj)) { 1119 for (const auto &Symb : Symbols) { 1120 uint64_t Address = std::get<0>(Symb); 1121 StringRef Name = std::get<1>(Symb); 1122 if (Name.startswith("$d")) 1123 DataMappingSymsAddr.push_back(Address - SectionAddr); 1124 if (Name.startswith("$x")) 1125 TextMappingSymsAddr.push_back(Address - SectionAddr); 1126 if (Name.startswith("$a")) 1127 TextMappingSymsAddr.push_back(Address - SectionAddr); 1128 if (Name.startswith("$t")) 1129 TextMappingSymsAddr.push_back(Address - SectionAddr); 1130 } 1131 } 1132 1133 llvm::sort(DataMappingSymsAddr); 1134 llvm::sort(TextMappingSymsAddr); 1135 1136 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1137 // AMDGPU disassembler uses symbolizer for printing labels 1138 std::unique_ptr<MCRelocationInfo> RelInfo( 1139 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1140 if (RelInfo) { 1141 std::unique_ptr<MCSymbolizer> Symbolizer( 1142 TheTarget->createMCSymbolizer( 1143 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1144 DisAsm->setSymbolizer(std::move(Symbolizer)); 1145 } 1146 } 1147 1148 StringRef SegmentName = ""; 1149 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1150 DataRefImpl DR = Section.getRawDataRefImpl(); 1151 SegmentName = MachO->getSectionFinalSegmentName(DR); 1152 } 1153 StringRef SectionName; 1154 error(Section.getName(SectionName)); 1155 1156 // If the section has no symbol at the start, just insert a dummy one. 1157 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1158 Symbols.insert( 1159 Symbols.begin(), 1160 std::make_tuple(SectionAddr, SectionName, 1161 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1162 } 1163 1164 SmallString<40> Comments; 1165 raw_svector_ostream CommentStream(Comments); 1166 1167 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef( 1168 unwrapOrError(Section.getContents(), Obj->getFileName())); 1169 1170 uint64_t VMAAdjustment = 0; 1171 if (shouldAdjustVA(Section)) 1172 VMAAdjustment = AdjustVMA; 1173 1174 uint64_t Size; 1175 uint64_t Index; 1176 bool PrintedSection = false; 1177 std::vector<RelocationRef> Rels = RelocMap[Section]; 1178 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1179 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1180 // Disassemble symbol by symbol. 1181 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1182 // Skip if --disassemble-functions is not empty and the symbol is not in 1183 // the list. 1184 if (!DisasmFuncsSet.empty() && 1185 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1186 continue; 1187 1188 uint64_t Start = std::get<0>(Symbols[SI]); 1189 if (Start < SectionAddr || StopAddress <= Start) 1190 continue; 1191 else 1192 FoundDisasmFuncsSet.insert(std::get<1>(Symbols[SI])); 1193 1194 // The end is the section end, the beginning of the next symbol, or 1195 // --stop-address. 1196 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress); 1197 if (SI + 1 < SE) 1198 End = std::min(End, std::get<0>(Symbols[SI + 1])); 1199 if (Start >= End || End <= StartAddress) 1200 continue; 1201 Start -= SectionAddr; 1202 End -= SectionAddr; 1203 1204 if (!PrintedSection) { 1205 PrintedSection = true; 1206 outs() << "\nDisassembly of section "; 1207 if (!SegmentName.empty()) 1208 outs() << SegmentName << ","; 1209 outs() << SectionName << ":\n"; 1210 } 1211 1212 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1213 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1214 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1215 Start += 256; 1216 } 1217 if (SI == SE - 1 || 1218 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1219 // cut trailing zeroes at the end of kernel 1220 // cut up to 256 bytes 1221 const uint64_t EndAlign = 256; 1222 const auto Limit = End - (std::min)(EndAlign, End - Start); 1223 while (End > Limit && 1224 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1225 End -= 4; 1226 } 1227 } 1228 1229 outs() << '\n'; 1230 if (!NoLeadingAddr) 1231 outs() << format("%016" PRIx64 " ", 1232 SectionAddr + Start + VMAAdjustment); 1233 1234 StringRef SymbolName = std::get<1>(Symbols[SI]); 1235 if (Demangle) 1236 outs() << demangle(SymbolName) << ":\n"; 1237 else 1238 outs() << SymbolName << ":\n"; 1239 1240 // Don't print raw contents of a virtual section. A virtual section 1241 // doesn't have any contents in the file. 1242 if (Section.isVirtual()) { 1243 outs() << "...\n"; 1244 continue; 1245 } 1246 1247 #ifndef NDEBUG 1248 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1249 #else 1250 raw_ostream &DebugOut = nulls(); 1251 #endif 1252 1253 // Some targets (like WebAssembly) have a special prelude at the start 1254 // of each symbol. 1255 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1256 SectionAddr + Start, DebugOut, CommentStream); 1257 Start += Size; 1258 1259 Index = Start; 1260 if (SectionAddr < StartAddress) 1261 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1262 1263 // If there is a data symbol inside an ELF text section and we are 1264 // only disassembling text (applicable all architectures), we are in a 1265 // situation where we must print the data and not disassemble it. 1266 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1267 !DisassembleAll && Section.isText()) { 1268 dumpELFData(SectionAddr, Index, End, Bytes); 1269 Index = End; 1270 } 1271 1272 bool CheckARMELFData = isArmElf(Obj) && 1273 std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1274 !DisassembleAll; 1275 while (Index < End) { 1276 // AArch64 ELF binaries can interleave data and text in the same 1277 // section. We rely on the markers introduced to understand what we 1278 // need to dump. If the data marker is within a function, it is 1279 // denoted as a word/short etc. 1280 if (CheckARMELFData && 1281 std::binary_search(DataMappingSymsAddr.begin(), 1282 DataMappingSymsAddr.end(), Index)) { 1283 Index = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1284 TextMappingSymsAddr); 1285 continue; 1286 } 1287 1288 // When -z or --disassemble-zeroes are given we always dissasemble 1289 // them. Otherwise we might want to skip zero bytes we see. 1290 if (!DisassembleZeroes) { 1291 uint64_t MaxOffset = End - Index; 1292 // For -reloc: print zero blocks patched by relocations, so that 1293 // relocations can be shown in the dump. 1294 if (RelCur != RelEnd) 1295 MaxOffset = RelCur->getOffset() - Index; 1296 1297 if (size_t N = 1298 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1299 outs() << "\t\t..." << '\n'; 1300 Index += N; 1301 continue; 1302 } 1303 } 1304 1305 // Disassemble a real instruction or a data when disassemble all is 1306 // provided 1307 MCInst Inst; 1308 bool Disassembled = DisAsm->getInstruction( 1309 Inst, Size, Bytes.slice(Index), SectionAddr + Index, DebugOut, 1310 CommentStream); 1311 if (Size == 0) 1312 Size = 1; 1313 1314 PIP.printInst( 1315 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 1316 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, outs(), 1317 "", *STI, &SP, &Rels); 1318 outs() << CommentStream.str(); 1319 Comments.clear(); 1320 1321 // Try to resolve the target of a call, tail call, etc. to a specific 1322 // symbol. 1323 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1324 MIA->isConditionalBranch(Inst))) { 1325 uint64_t Target; 1326 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1327 // In a relocatable object, the target's section must reside in 1328 // the same section as the call instruction or it is accessed 1329 // through a relocation. 1330 // 1331 // In a non-relocatable object, the target may be in any section. 1332 // 1333 // N.B. We don't walk the relocations in the relocatable case yet. 1334 auto *TargetSectionSymbols = &Symbols; 1335 if (!Obj->isRelocatableObject()) { 1336 auto It = llvm::bsearch( 1337 SectionAddresses, 1338 [=](const std::pair<uint64_t, SectionRef> &RHS) { 1339 return Target < RHS.first; 1340 }); 1341 if (It != SectionAddresses.begin()) { 1342 --It; 1343 TargetSectionSymbols = &AllSymbols[It->second]; 1344 } else { 1345 TargetSectionSymbols = &AbsoluteSymbols; 1346 } 1347 } 1348 1349 // Find the last symbol in the section whose offset is less than 1350 // or equal to the target. If there isn't a section that contains 1351 // the target, find the nearest preceding absolute symbol. 1352 auto TargetSym = llvm::bsearch( 1353 *TargetSectionSymbols, 1354 [=](const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1355 return Target < std::get<0>(RHS); 1356 }); 1357 if (TargetSym == TargetSectionSymbols->begin()) { 1358 TargetSectionSymbols = &AbsoluteSymbols; 1359 TargetSym = llvm::bsearch( 1360 AbsoluteSymbols, 1361 [=](const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1362 return Target < std::get<0>(RHS); 1363 }); 1364 } 1365 if (TargetSym != TargetSectionSymbols->begin()) { 1366 --TargetSym; 1367 uint64_t TargetAddress = std::get<0>(*TargetSym); 1368 StringRef TargetName = std::get<1>(*TargetSym); 1369 outs() << " <" << TargetName; 1370 uint64_t Disp = Target - TargetAddress; 1371 if (Disp) 1372 outs() << "+0x" << Twine::utohexstr(Disp); 1373 outs() << '>'; 1374 } 1375 } 1376 } 1377 outs() << "\n"; 1378 1379 // Hexagon does this in pretty printer 1380 if (Obj->getArch() != Triple::hexagon) { 1381 // Print relocation for instruction. 1382 while (RelCur != RelEnd) { 1383 uint64_t Offset = RelCur->getOffset(); 1384 // If this relocation is hidden, skip it. 1385 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 1386 ++RelCur; 1387 continue; 1388 } 1389 1390 // Stop when RelCur's offset is past the current instruction. 1391 if (Offset >= Index + Size) 1392 break; 1393 1394 // When --adjust-vma is used, update the address printed. 1395 if (RelCur->getSymbol() != Obj->symbol_end()) { 1396 Expected<section_iterator> SymSI = 1397 RelCur->getSymbol()->getSection(); 1398 if (SymSI && *SymSI != Obj->section_end() && 1399 shouldAdjustVA(**SymSI)) 1400 Offset += AdjustVMA; 1401 } 1402 1403 printRelocation(*RelCur, SectionAddr + Offset, 1404 Obj->getBytesInAddress()); 1405 ++RelCur; 1406 } 1407 } 1408 1409 Index += Size; 1410 } 1411 } 1412 } 1413 StringSet<> MissingDisasmFuncsSet = 1414 set_difference(DisasmFuncsSet, FoundDisasmFuncsSet); 1415 for (StringRef MissingDisasmFunc : MissingDisasmFuncsSet.keys()) 1416 warn("failed to disassemble missing function " + MissingDisasmFunc); 1417 } 1418 1419 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1420 if (StartAddress > StopAddress) 1421 error("Start address should be less than stop address"); 1422 1423 const Target *TheTarget = getTarget(Obj); 1424 1425 // Package up features to be passed to target/subtarget 1426 SubtargetFeatures Features = Obj->getFeatures(); 1427 if (!MAttrs.empty()) 1428 for (unsigned I = 0; I != MAttrs.size(); ++I) 1429 Features.AddFeature(MAttrs[I]); 1430 1431 std::unique_ptr<const MCRegisterInfo> MRI( 1432 TheTarget->createMCRegInfo(TripleName)); 1433 if (!MRI) 1434 report_error(Obj->getFileName(), 1435 "no register info for target " + TripleName); 1436 1437 // Set up disassembler. 1438 std::unique_ptr<const MCAsmInfo> AsmInfo( 1439 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1440 if (!AsmInfo) 1441 report_error(Obj->getFileName(), 1442 "no assembly info for target " + TripleName); 1443 std::unique_ptr<const MCSubtargetInfo> STI( 1444 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1445 if (!STI) 1446 report_error(Obj->getFileName(), 1447 "no subtarget info for target " + TripleName); 1448 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1449 if (!MII) 1450 report_error(Obj->getFileName(), 1451 "no instruction info for target " + TripleName); 1452 MCObjectFileInfo MOFI; 1453 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1454 // FIXME: for now initialize MCObjectFileInfo with default values 1455 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1456 1457 std::unique_ptr<MCDisassembler> DisAsm( 1458 TheTarget->createMCDisassembler(*STI, Ctx)); 1459 if (!DisAsm) 1460 report_error(Obj->getFileName(), 1461 "no disassembler for target " + TripleName); 1462 1463 std::unique_ptr<const MCInstrAnalysis> MIA( 1464 TheTarget->createMCInstrAnalysis(MII.get())); 1465 1466 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1467 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1468 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1469 if (!IP) 1470 report_error(Obj->getFileName(), 1471 "no instruction printer for target " + TripleName); 1472 IP->setPrintImmHex(PrintImmHex); 1473 1474 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1475 SourcePrinter SP(Obj, TheTarget->getName()); 1476 1477 for (StringRef Opt : DisassemblerOptions) 1478 if (!IP->applyTargetSpecificCLOption(Opt)) 1479 error("Unrecognized disassembler option: " + Opt); 1480 1481 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), MIA.get(), IP.get(), 1482 STI.get(), PIP, SP, InlineRelocs); 1483 } 1484 1485 void printRelocations(const ObjectFile *Obj) { 1486 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1487 "%08" PRIx64; 1488 // Regular objdump doesn't print relocations in non-relocatable object 1489 // files. 1490 if (!Obj->isRelocatableObject()) 1491 return; 1492 1493 // Build a mapping from relocation target to a vector of relocation 1494 // sections. Usually, there is an only one relocation section for 1495 // each relocated section. 1496 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec; 1497 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1498 if (Section.relocation_begin() == Section.relocation_end()) 1499 continue; 1500 const SectionRef TargetSec = *Section.getRelocatedSection(); 1501 SecToRelSec[TargetSec].push_back(Section); 1502 } 1503 1504 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) { 1505 StringRef SecName; 1506 error(P.first.getName(SecName)); 1507 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1508 1509 for (SectionRef Section : P.second) { 1510 for (const RelocationRef &Reloc : Section.relocations()) { 1511 uint64_t Address = Reloc.getOffset(); 1512 SmallString<32> RelocName; 1513 SmallString<32> ValueStr; 1514 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1515 continue; 1516 Reloc.getTypeName(RelocName); 1517 error(getRelocationValueString(Reloc, ValueStr)); 1518 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1519 << ValueStr << "\n"; 1520 } 1521 } 1522 outs() << "\n"; 1523 } 1524 } 1525 1526 void printDynamicRelocations(const ObjectFile *Obj) { 1527 // For the moment, this option is for ELF only 1528 if (!Obj->isELF()) 1529 return; 1530 1531 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1532 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1533 error("not a dynamic object"); 1534 return; 1535 } 1536 1537 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1538 if (DynRelSec.empty()) 1539 return; 1540 1541 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1542 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1543 for (const SectionRef &Section : DynRelSec) 1544 for (const RelocationRef &Reloc : Section.relocations()) { 1545 uint64_t Address = Reloc.getOffset(); 1546 SmallString<32> RelocName; 1547 SmallString<32> ValueStr; 1548 Reloc.getTypeName(RelocName); 1549 error(getRelocationValueString(Reloc, ValueStr)); 1550 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1551 << ValueStr << "\n"; 1552 } 1553 } 1554 1555 // Returns true if we need to show LMA column when dumping section headers. We 1556 // show it only when the platform is ELF and either we have at least one section 1557 // whose VMA and LMA are different and/or when --show-lma flag is used. 1558 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1559 if (!Obj->isELF()) 1560 return false; 1561 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1562 if (S.getAddress() != getELFSectionLMA(S)) 1563 return true; 1564 return ShowLMA; 1565 } 1566 1567 void printSectionHeaders(const ObjectFile *Obj) { 1568 bool HasLMAColumn = shouldDisplayLMA(Obj); 1569 if (HasLMAColumn) 1570 outs() << "Sections:\n" 1571 "Idx Name Size VMA LMA " 1572 "Type\n"; 1573 else 1574 outs() << "Sections:\n" 1575 "Idx Name Size VMA Type\n"; 1576 1577 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1578 StringRef Name; 1579 error(Section.getName(Name)); 1580 uint64_t VMA = Section.getAddress(); 1581 if (shouldAdjustVA(Section)) 1582 VMA += AdjustVMA; 1583 1584 uint64_t Size = Section.getSize(); 1585 bool Text = Section.isText(); 1586 bool Data = Section.isData(); 1587 bool BSS = Section.isBSS(); 1588 std::string Type = (std::string(Text ? "TEXT " : "") + 1589 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1590 1591 if (HasLMAColumn) 1592 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %016" PRIx64 1593 " %s\n", 1594 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1595 VMA, getELFSectionLMA(Section), Type.c_str()); 1596 else 1597 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1598 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1599 VMA, Type.c_str()); 1600 } 1601 outs() << "\n"; 1602 } 1603 1604 void printSectionContents(const ObjectFile *Obj) { 1605 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1606 StringRef Name; 1607 error(Section.getName(Name)); 1608 uint64_t BaseAddr = Section.getAddress(); 1609 uint64_t Size = Section.getSize(); 1610 if (!Size) 1611 continue; 1612 1613 outs() << "Contents of section " << Name << ":\n"; 1614 if (Section.isBSS()) { 1615 outs() << format("<skipping contents of bss section at [%04" PRIx64 1616 ", %04" PRIx64 ")>\n", 1617 BaseAddr, BaseAddr + Size); 1618 continue; 1619 } 1620 1621 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName()); 1622 1623 // Dump out the content as hex and printable ascii characters. 1624 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1625 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1626 // Dump line of hex. 1627 for (std::size_t I = 0; I < 16; ++I) { 1628 if (I != 0 && I % 4 == 0) 1629 outs() << ' '; 1630 if (Addr + I < End) 1631 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1632 << hexdigit(Contents[Addr + I] & 0xF, true); 1633 else 1634 outs() << " "; 1635 } 1636 // Print ascii. 1637 outs() << " "; 1638 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1639 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1640 outs() << Contents[Addr + I]; 1641 else 1642 outs() << "."; 1643 } 1644 outs() << "\n"; 1645 } 1646 } 1647 } 1648 1649 void printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1650 StringRef ArchitectureName) { 1651 outs() << "SYMBOL TABLE:\n"; 1652 1653 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1654 printCOFFSymbolTable(Coff); 1655 return; 1656 } 1657 1658 const StringRef FileName = O->getFileName(); 1659 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1660 const SymbolRef &Symbol = *I; 1661 uint64_t Address = unwrapOrError(Symbol.getAddress(), ArchiveName, FileName, 1662 ArchitectureName); 1663 if ((Address < StartAddress) || (Address > StopAddress)) 1664 continue; 1665 SymbolRef::Type Type = unwrapOrError(Symbol.getType(), ArchiveName, 1666 FileName, ArchitectureName); 1667 uint32_t Flags = Symbol.getFlags(); 1668 section_iterator Section = unwrapOrError(Symbol.getSection(), ArchiveName, 1669 FileName, ArchitectureName); 1670 StringRef Name; 1671 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) 1672 Section->getName(Name); 1673 else 1674 Name = unwrapOrError(Symbol.getName(), ArchiveName, FileName, 1675 ArchitectureName); 1676 1677 bool Global = Flags & SymbolRef::SF_Global; 1678 bool Weak = Flags & SymbolRef::SF_Weak; 1679 bool Absolute = Flags & SymbolRef::SF_Absolute; 1680 bool Common = Flags & SymbolRef::SF_Common; 1681 bool Hidden = Flags & SymbolRef::SF_Hidden; 1682 1683 char GlobLoc = ' '; 1684 if (Type != SymbolRef::ST_Unknown) 1685 GlobLoc = Global ? 'g' : 'l'; 1686 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1687 ? 'd' : ' '; 1688 char FileFunc = ' '; 1689 if (Type == SymbolRef::ST_File) 1690 FileFunc = 'f'; 1691 else if (Type == SymbolRef::ST_Function) 1692 FileFunc = 'F'; 1693 else if (Type == SymbolRef::ST_Data) 1694 FileFunc = 'O'; 1695 1696 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1697 "%08" PRIx64; 1698 1699 outs() << format(Fmt, Address) << " " 1700 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1701 << (Weak ? 'w' : ' ') // Weak? 1702 << ' ' // Constructor. Not supported yet. 1703 << ' ' // Warning. Not supported yet. 1704 << ' ' // Indirect reference to another symbol. 1705 << Debug // Debugging (d) or dynamic (D) symbol. 1706 << FileFunc // Name of function (F), file (f) or object (O). 1707 << ' '; 1708 if (Absolute) { 1709 outs() << "*ABS*"; 1710 } else if (Common) { 1711 outs() << "*COM*"; 1712 } else if (Section == O->section_end()) { 1713 outs() << "*UND*"; 1714 } else { 1715 if (const MachOObjectFile *MachO = 1716 dyn_cast<const MachOObjectFile>(O)) { 1717 DataRefImpl DR = Section->getRawDataRefImpl(); 1718 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1719 outs() << SegmentName << ","; 1720 } 1721 StringRef SectionName; 1722 error(Section->getName(SectionName)); 1723 outs() << SectionName; 1724 } 1725 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 (isa<ELFObjectFileBase>(O)) { 1733 uint8_t Other = ELFSymbolRef(Symbol).getOther(); 1734 switch (Other) { 1735 case ELF::STV_DEFAULT: 1736 break; 1737 case ELF::STV_INTERNAL: 1738 outs() << " .internal"; 1739 break; 1740 case ELF::STV_HIDDEN: 1741 outs() << " .hidden"; 1742 break; 1743 case ELF::STV_PROTECTED: 1744 outs() << " .protected"; 1745 break; 1746 default: 1747 outs() << format(" 0x%02x", Other); 1748 break; 1749 } 1750 } else if (Hidden) { 1751 outs() << " .hidden"; 1752 } 1753 1754 if (Demangle) 1755 outs() << ' ' << demangle(Name) << '\n'; 1756 else 1757 outs() << ' ' << Name << '\n'; 1758 } 1759 } 1760 1761 static void printUnwindInfo(const ObjectFile *O) { 1762 outs() << "Unwind info:\n\n"; 1763 1764 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1765 printCOFFUnwindInfo(Coff); 1766 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1767 printMachOUnwindInfo(MachO); 1768 else 1769 // TODO: Extract DWARF dump tool to objdump. 1770 WithColor::error(errs(), ToolName) 1771 << "This operation is only currently supported " 1772 "for COFF and MachO object files.\n"; 1773 } 1774 1775 /// Dump the raw contents of the __clangast section so the output can be piped 1776 /// into llvm-bcanalyzer. 1777 void printRawClangAST(const ObjectFile *Obj) { 1778 if (outs().is_displayed()) { 1779 WithColor::error(errs(), ToolName) 1780 << "The -raw-clang-ast option will dump the raw binary contents of " 1781 "the clang ast section.\n" 1782 "Please redirect the output to a file or another program such as " 1783 "llvm-bcanalyzer.\n"; 1784 return; 1785 } 1786 1787 StringRef ClangASTSectionName("__clangast"); 1788 if (isa<COFFObjectFile>(Obj)) { 1789 ClangASTSectionName = "clangast"; 1790 } 1791 1792 Optional<object::SectionRef> ClangASTSection; 1793 for (auto Sec : ToolSectionFilter(*Obj)) { 1794 StringRef Name; 1795 Sec.getName(Name); 1796 if (Name == ClangASTSectionName) { 1797 ClangASTSection = Sec; 1798 break; 1799 } 1800 } 1801 if (!ClangASTSection) 1802 return; 1803 1804 StringRef ClangASTContents = unwrapOrError( 1805 ClangASTSection.getValue().getContents(), Obj->getFileName()); 1806 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1807 } 1808 1809 static void printFaultMaps(const ObjectFile *Obj) { 1810 StringRef FaultMapSectionName; 1811 1812 if (isa<ELFObjectFileBase>(Obj)) { 1813 FaultMapSectionName = ".llvm_faultmaps"; 1814 } else if (isa<MachOObjectFile>(Obj)) { 1815 FaultMapSectionName = "__llvm_faultmaps"; 1816 } else { 1817 WithColor::error(errs(), ToolName) 1818 << "This operation is only currently supported " 1819 "for ELF and Mach-O executable files.\n"; 1820 return; 1821 } 1822 1823 Optional<object::SectionRef> FaultMapSection; 1824 1825 for (auto Sec : ToolSectionFilter(*Obj)) { 1826 StringRef Name; 1827 Sec.getName(Name); 1828 if (Name == FaultMapSectionName) { 1829 FaultMapSection = Sec; 1830 break; 1831 } 1832 } 1833 1834 outs() << "FaultMap table:\n"; 1835 1836 if (!FaultMapSection.hasValue()) { 1837 outs() << "<not found>\n"; 1838 return; 1839 } 1840 1841 StringRef FaultMapContents = 1842 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName()); 1843 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1844 FaultMapContents.bytes_end()); 1845 1846 outs() << FMP; 1847 } 1848 1849 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1850 if (O->isELF()) { 1851 printELFFileHeader(O); 1852 printELFDynamicSection(O); 1853 printELFSymbolVersionInfo(O); 1854 return; 1855 } 1856 if (O->isCOFF()) 1857 return printCOFFFileHeader(O); 1858 if (O->isWasm()) 1859 return printWasmFileHeader(O); 1860 if (O->isMachO()) { 1861 printMachOFileHeader(O); 1862 if (!OnlyFirst) 1863 printMachOLoadCommands(O); 1864 return; 1865 } 1866 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1867 } 1868 1869 static void printFileHeaders(const ObjectFile *O) { 1870 if (!O->isELF() && !O->isCOFF()) 1871 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1872 1873 Triple::ArchType AT = O->getArch(); 1874 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1875 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 1876 1877 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1878 outs() << "start address: " 1879 << "0x" << format(Fmt.data(), Address) << "\n\n"; 1880 } 1881 1882 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 1883 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 1884 if (!ModeOrErr) { 1885 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 1886 consumeError(ModeOrErr.takeError()); 1887 return; 1888 } 1889 sys::fs::perms Mode = ModeOrErr.get(); 1890 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1891 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1892 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1893 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1894 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1895 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1896 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1897 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1898 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1899 1900 outs() << " "; 1901 1902 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 1903 unwrapOrError(C.getGID(), Filename), 1904 unwrapOrError(C.getRawSize(), Filename)); 1905 1906 StringRef RawLastModified = C.getRawLastModified(); 1907 unsigned Seconds; 1908 if (RawLastModified.getAsInteger(10, Seconds)) 1909 outs() << "(date: \"" << RawLastModified 1910 << "\" contains non-decimal chars) "; 1911 else { 1912 // Since ctime(3) returns a 26 character string of the form: 1913 // "Sun Sep 16 01:03:52 1973\n\0" 1914 // just print 24 characters. 1915 time_t t = Seconds; 1916 outs() << format("%.24s ", ctime(&t)); 1917 } 1918 1919 StringRef Name = ""; 1920 Expected<StringRef> NameOrErr = C.getName(); 1921 if (!NameOrErr) { 1922 consumeError(NameOrErr.takeError()); 1923 Name = unwrapOrError(C.getRawName(), Filename); 1924 } else { 1925 Name = NameOrErr.get(); 1926 } 1927 outs() << Name << "\n"; 1928 } 1929 1930 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 1931 const Archive::Child *C = nullptr) { 1932 // Avoid other output when using a raw option. 1933 if (!RawClangAST) { 1934 outs() << '\n'; 1935 if (A) 1936 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 1937 else 1938 outs() << O->getFileName(); 1939 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 1940 } 1941 1942 StringRef ArchiveName = A ? A->getFileName() : ""; 1943 if (FileHeaders) 1944 printFileHeaders(O); 1945 if (ArchiveHeaders && !MachOOpt && C) 1946 printArchiveChild(ArchiveName, *C); 1947 if (Disassemble) 1948 disassembleObject(O, Relocations); 1949 if (Relocations && !Disassemble) 1950 printRelocations(O); 1951 if (DynamicRelocations) 1952 printDynamicRelocations(O); 1953 if (SectionHeaders) 1954 printSectionHeaders(O); 1955 if (SectionContents) 1956 printSectionContents(O); 1957 if (SymbolTable) 1958 printSymbolTable(O, ArchiveName); 1959 if (UnwindInfo) 1960 printUnwindInfo(O); 1961 if (PrivateHeaders || FirstPrivateHeader) 1962 printPrivateFileHeaders(O, FirstPrivateHeader); 1963 if (ExportsTrie) 1964 printExportsTrie(O); 1965 if (Rebase) 1966 printRebaseTable(O); 1967 if (Bind) 1968 printBindTable(O); 1969 if (LazyBind) 1970 printLazyBindTable(O); 1971 if (WeakBind) 1972 printWeakBindTable(O); 1973 if (RawClangAST) 1974 printRawClangAST(O); 1975 if (FaultMapSection) 1976 printFaultMaps(O); 1977 if (DwarfDumpType != DIDT_Null) { 1978 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 1979 // Dump the complete DWARF structure. 1980 DIDumpOptions DumpOpts; 1981 DumpOpts.DumpType = DwarfDumpType; 1982 DICtx->dump(outs(), DumpOpts); 1983 } 1984 } 1985 1986 static void dumpObject(const COFFImportFile *I, const Archive *A, 1987 const Archive::Child *C = nullptr) { 1988 StringRef ArchiveName = A ? A->getFileName() : ""; 1989 1990 // Avoid other output when using a raw option. 1991 if (!RawClangAST) 1992 outs() << '\n' 1993 << ArchiveName << "(" << I->getFileName() << ")" 1994 << ":\tfile format COFF-import-file" 1995 << "\n\n"; 1996 1997 if (ArchiveHeaders && !MachOOpt && C) 1998 printArchiveChild(ArchiveName, *C); 1999 if (SymbolTable) 2000 printCOFFSymbolTable(I); 2001 } 2002 2003 /// Dump each object file in \a a; 2004 static void dumpArchive(const Archive *A) { 2005 Error Err = Error::success(); 2006 for (auto &C : A->children(Err)) { 2007 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2008 if (!ChildOrErr) { 2009 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2010 report_error(std::move(E), A->getFileName(), C); 2011 continue; 2012 } 2013 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2014 dumpObject(O, A, &C); 2015 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2016 dumpObject(I, A, &C); 2017 else 2018 report_error(errorCodeToError(object_error::invalid_file_type), 2019 A->getFileName()); 2020 } 2021 if (Err) 2022 report_error(std::move(Err), A->getFileName()); 2023 } 2024 2025 /// Open file and figure out how to dump it. 2026 static void dumpInput(StringRef file) { 2027 // If we are using the Mach-O specific object file parser, then let it parse 2028 // the file and process the command line options. So the -arch flags can 2029 // be used to select specific slices, etc. 2030 if (MachOOpt) { 2031 parseInputMachO(file); 2032 return; 2033 } 2034 2035 // Attempt to open the binary. 2036 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2037 Binary &Binary = *OBinary.getBinary(); 2038 2039 if (Archive *A = dyn_cast<Archive>(&Binary)) 2040 dumpArchive(A); 2041 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2042 dumpObject(O); 2043 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2044 parseInputMachO(UB); 2045 else 2046 report_error(errorCodeToError(object_error::invalid_file_type), file); 2047 } 2048 } // namespace llvm 2049 2050 int main(int argc, char **argv) { 2051 using namespace llvm; 2052 InitLLVM X(argc, argv); 2053 const cl::OptionCategory *OptionFilters[] = {&ObjdumpCat, &MachOCat}; 2054 cl::HideUnrelatedOptions(OptionFilters); 2055 2056 // Initialize targets and assembly printers/parsers. 2057 InitializeAllTargetInfos(); 2058 InitializeAllTargetMCs(); 2059 InitializeAllDisassemblers(); 2060 2061 // Register the target printer for --version. 2062 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2063 2064 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2065 2066 ToolName = argv[0]; 2067 2068 // Defaults to a.out if no filenames specified. 2069 if (InputFilenames.empty()) 2070 InputFilenames.push_back("a.out"); 2071 2072 if (AllHeaders) 2073 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2074 SectionHeaders = SymbolTable = true; 2075 2076 if (DisassembleAll || PrintSource || PrintLines || 2077 (!DisassembleFunctions.empty())) 2078 Disassemble = true; 2079 2080 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null && 2081 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST && 2082 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable && 2083 !UnwindInfo && !FaultMapSection && 2084 !(MachOOpt && 2085 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie || 2086 FirstPrivateHeader || IndirectSymbols || InfoPlist || LazyBind || 2087 LinkOptHints || ObjcMetaData || Rebase || UniversalHeaders || 2088 WeakBind || !FilterSections.empty()))) { 2089 cl::PrintHelpMessage(); 2090 return 2; 2091 } 2092 2093 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2094 DisassembleFunctions.end()); 2095 2096 llvm::for_each(InputFilenames, dumpInput); 2097 2098 return EXIT_SUCCESS; 2099 } 2100