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 cl::extrahelp 334 HelpResponse("\nPass @FILE as argument to read options from FILE.\n"); 335 336 static StringSet<> DisasmFuncsSet; 337 static StringRef ToolName; 338 339 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 340 341 static bool shouldKeep(object::SectionRef S) { 342 if (FilterSections.empty()) 343 return true; 344 StringRef String; 345 std::error_code error = S.getName(String); 346 if (error) 347 return false; 348 return is_contained(FilterSections, String); 349 } 350 351 SectionFilter ToolSectionFilter(object::ObjectFile const &O) { 352 return SectionFilter([](object::SectionRef S) { return shouldKeep(S); }, O); 353 } 354 355 void error(std::error_code EC) { 356 if (!EC) 357 return; 358 WithColor::error(errs(), ToolName) 359 << "reading file: " << EC.message() << ".\n"; 360 errs().flush(); 361 exit(1); 362 } 363 364 void error(Error E) { 365 if (!E) 366 return; 367 WithColor::error(errs(), ToolName) << toString(std::move(E)); 368 exit(1); 369 } 370 371 LLVM_ATTRIBUTE_NORETURN void error(Twine Message) { 372 WithColor::error(errs(), ToolName) << Message << ".\n"; 373 errs().flush(); 374 exit(1); 375 } 376 377 void warn(StringRef Message) { 378 WithColor::warning(errs(), ToolName) << Message << ".\n"; 379 errs().flush(); 380 } 381 382 void warn(Twine Message) { 383 WithColor::warning(errs(), ToolName) << Message << "\n"; 384 } 385 386 LLVM_ATTRIBUTE_NORETURN void report_error(StringRef File, Twine Message) { 387 WithColor::error(errs(), ToolName) 388 << "'" << File << "': " << Message << ".\n"; 389 exit(1); 390 } 391 392 LLVM_ATTRIBUTE_NORETURN void report_error(Error E, StringRef File) { 393 assert(E); 394 std::string Buf; 395 raw_string_ostream OS(Buf); 396 logAllUnhandledErrors(std::move(E), OS); 397 OS.flush(); 398 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 399 exit(1); 400 } 401 402 LLVM_ATTRIBUTE_NORETURN void report_error(Error E, StringRef ArchiveName, 403 StringRef FileName, 404 StringRef ArchitectureName) { 405 assert(E); 406 WithColor::error(errs(), ToolName); 407 if (ArchiveName != "") 408 errs() << ArchiveName << "(" << FileName << ")"; 409 else 410 errs() << "'" << FileName << "'"; 411 if (!ArchitectureName.empty()) 412 errs() << " (for architecture " << ArchitectureName << ")"; 413 std::string Buf; 414 raw_string_ostream OS(Buf); 415 logAllUnhandledErrors(std::move(E), OS); 416 OS.flush(); 417 errs() << ": " << Buf; 418 exit(1); 419 } 420 421 LLVM_ATTRIBUTE_NORETURN void report_error(Error E, StringRef ArchiveName, 422 const object::Archive::Child &C, 423 StringRef ArchitectureName) { 424 Expected<StringRef> NameOrErr = C.getName(); 425 // TODO: if we have a error getting the name then it would be nice to print 426 // the index of which archive member this is and or its offset in the 427 // archive instead of "???" as the name. 428 if (!NameOrErr) { 429 consumeError(NameOrErr.takeError()); 430 report_error(std::move(E), ArchiveName, "???", ArchitectureName); 431 } else 432 report_error(std::move(E), ArchiveName, NameOrErr.get(), ArchitectureName); 433 } 434 435 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 436 // Figure out the target triple. 437 Triple TheTriple("unknown-unknown-unknown"); 438 if (TripleName.empty()) { 439 if (Obj) 440 TheTriple = Obj->makeTriple(); 441 } else { 442 TheTriple.setTriple(Triple::normalize(TripleName)); 443 444 // Use the triple, but also try to combine with ARM build attributes. 445 if (Obj) { 446 auto Arch = Obj->getArch(); 447 if (Arch == Triple::arm || Arch == Triple::armeb) 448 Obj->setARMSubArch(TheTriple); 449 } 450 } 451 452 // Get the target specific parser. 453 std::string Error; 454 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 455 Error); 456 if (!TheTarget) { 457 if (Obj) 458 report_error(Obj->getFileName(), "can't find target: " + Error); 459 else 460 error("can't find target: " + Error); 461 } 462 463 // Update the triple name and return the found target. 464 TripleName = TheTriple.getTriple(); 465 return TheTarget; 466 } 467 468 bool isRelocAddressLess(RelocationRef A, RelocationRef B) { 469 return A.getOffset() < B.getOffset(); 470 } 471 472 static Error getRelocationValueString(const RelocationRef &Rel, 473 SmallVectorImpl<char> &Result) { 474 const ObjectFile *Obj = Rel.getObject(); 475 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 476 return getELFRelocationValueString(ELF, Rel, Result); 477 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 478 return getCOFFRelocationValueString(COFF, Rel, Result); 479 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 480 return getWasmRelocationValueString(Wasm, Rel, Result); 481 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 482 return getMachORelocationValueString(MachO, Rel, Result); 483 llvm_unreachable("unknown object file format"); 484 } 485 486 /// Indicates whether this relocation should hidden when listing 487 /// relocations, usually because it is the trailing part of a multipart 488 /// relocation that will be printed as part of the leading relocation. 489 static bool getHidden(RelocationRef RelRef) { 490 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 491 if (!MachO) 492 return false; 493 494 unsigned Arch = MachO->getArch(); 495 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 496 uint64_t Type = MachO->getRelocationType(Rel); 497 498 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 499 // is always hidden. 500 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 501 return Type == MachO::GENERIC_RELOC_PAIR; 502 503 if (Arch == Triple::x86_64) { 504 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 505 // an X86_64_RELOC_SUBTRACTOR. 506 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 507 DataRefImpl RelPrev = Rel; 508 RelPrev.d.a--; 509 uint64_t PrevType = MachO->getRelocationType(RelPrev); 510 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 511 return true; 512 } 513 } 514 515 return false; 516 } 517 518 namespace { 519 class SourcePrinter { 520 protected: 521 DILineInfo OldLineInfo; 522 const ObjectFile *Obj = nullptr; 523 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 524 // File name to file contents of source 525 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 526 // Mark the line endings of the cached source 527 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 528 529 private: 530 bool cacheSource(const DILineInfo& LineInfoFile); 531 532 public: 533 SourcePrinter() = default; 534 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 535 symbolize::LLVMSymbolizer::Options SymbolizerOpts; 536 SymbolizerOpts.PrintFunctions = DILineInfoSpecifier::FunctionNameKind::None; 537 SymbolizerOpts.Demangle = false; 538 SymbolizerOpts.DefaultArch = DefaultArch; 539 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 540 } 541 virtual ~SourcePrinter() = default; 542 virtual void printSourceLine(raw_ostream &OS, 543 object::SectionedAddress Address, 544 StringRef Delimiter = "; "); 545 }; 546 547 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 548 std::unique_ptr<MemoryBuffer> Buffer; 549 if (LineInfo.Source) { 550 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 551 } else { 552 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 553 if (!BufferOrError) 554 return false; 555 Buffer = std::move(*BufferOrError); 556 } 557 // Chomp the file to get lines 558 const char *BufferStart = Buffer->getBufferStart(), 559 *BufferEnd = Buffer->getBufferEnd(); 560 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 561 const char *Start = BufferStart; 562 for (const char *I = BufferStart; I != BufferEnd; ++I) 563 if (*I == '\n') { 564 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 565 Start = I + 1; 566 } 567 if (Start < BufferEnd) 568 Lines.emplace_back(Start, BufferEnd - Start); 569 SourceCache[LineInfo.FileName] = std::move(Buffer); 570 return true; 571 } 572 573 void SourcePrinter::printSourceLine(raw_ostream &OS, 574 object::SectionedAddress Address, 575 StringRef Delimiter) { 576 if (!Symbolizer) 577 return; 578 579 DILineInfo LineInfo = DILineInfo(); 580 auto ExpectedLineInfo = 581 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 582 if (!ExpectedLineInfo) 583 consumeError(ExpectedLineInfo.takeError()); 584 else 585 LineInfo = *ExpectedLineInfo; 586 587 if ((LineInfo.FileName == "<invalid>") || LineInfo.Line == 0 || 588 ((OldLineInfo.Line == LineInfo.Line) && 589 (OldLineInfo.FileName == LineInfo.FileName))) 590 return; 591 592 if (PrintLines) 593 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 594 if (PrintSource) { 595 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 596 if (!cacheSource(LineInfo)) 597 return; 598 auto LineBuffer = LineCache.find(LineInfo.FileName); 599 if (LineBuffer != LineCache.end()) { 600 if (LineInfo.Line > LineBuffer->second.size()) 601 return; 602 // Vector begins at 0, line numbers are non-zero 603 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1] << '\n'; 604 } 605 } 606 OldLineInfo = LineInfo; 607 } 608 609 static bool isAArch64Elf(const ObjectFile *Obj) { 610 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 611 return Elf && Elf->getEMachine() == ELF::EM_AARCH64; 612 } 613 614 static bool isArmElf(const ObjectFile *Obj) { 615 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 616 return Elf && Elf->getEMachine() == ELF::EM_ARM; 617 } 618 619 static bool hasMappingSymbols(const ObjectFile *Obj) { 620 return isArmElf(Obj) || isAArch64Elf(Obj); 621 } 622 623 static void printRelocation(const RelocationRef &Rel, uint64_t Address, 624 bool Is64Bits) { 625 StringRef Fmt = Is64Bits ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 626 SmallString<16> Name; 627 SmallString<32> Val; 628 Rel.getTypeName(Name); 629 error(getRelocationValueString(Rel, Val)); 630 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n"; 631 } 632 633 class PrettyPrinter { 634 public: 635 virtual ~PrettyPrinter() = default; 636 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 637 ArrayRef<uint8_t> Bytes, 638 object::SectionedAddress Address, raw_ostream &OS, 639 StringRef Annot, MCSubtargetInfo const &STI, 640 SourcePrinter *SP, 641 std::vector<RelocationRef> *Rels = nullptr) { 642 if (SP && (PrintSource || PrintLines)) 643 SP->printSourceLine(OS, Address); 644 645 { 646 formatted_raw_ostream FOS(OS); 647 if (!NoLeadingAddr) 648 FOS << format("%8" PRIx64 ":", Address.Address); 649 if (!NoShowRawInsn) { 650 FOS << ' '; 651 dumpBytes(Bytes, FOS); 652 } 653 FOS.flush(); 654 // The output of printInst starts with a tab. Print some spaces so that 655 // the tab has 1 column and advances to the target tab stop. 656 unsigned TabStop = NoShowRawInsn ? 16 : 40; 657 unsigned Column = FOS.getColumn(); 658 FOS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8); 659 660 // The dtor calls flush() to ensure the indent comes before printInst(). 661 } 662 663 if (MI) 664 IP.printInst(MI, OS, "", STI); 665 else 666 OS << "\t<unknown>"; 667 } 668 }; 669 PrettyPrinter PrettyPrinterInst; 670 671 class HexagonPrettyPrinter : public PrettyPrinter { 672 public: 673 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 674 raw_ostream &OS) { 675 uint32_t opcode = 676 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 677 if (!NoLeadingAddr) 678 OS << format("%8" PRIx64 ":", Address); 679 if (!NoShowRawInsn) { 680 OS << "\t"; 681 dumpBytes(Bytes.slice(0, 4), OS); 682 OS << format("\t%08" PRIx32, opcode); 683 } 684 } 685 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 686 object::SectionedAddress Address, raw_ostream &OS, 687 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 688 std::vector<RelocationRef> *Rels) override { 689 if (SP && (PrintSource || PrintLines)) 690 SP->printSourceLine(OS, Address, ""); 691 if (!MI) { 692 printLead(Bytes, Address.Address, OS); 693 OS << " <unknown>"; 694 return; 695 } 696 std::string Buffer; 697 { 698 raw_string_ostream TempStream(Buffer); 699 IP.printInst(MI, TempStream, "", STI); 700 } 701 StringRef Contents(Buffer); 702 // Split off bundle attributes 703 auto PacketBundle = Contents.rsplit('\n'); 704 // Split off first instruction from the rest 705 auto HeadTail = PacketBundle.first.split('\n'); 706 auto Preamble = " { "; 707 auto Separator = ""; 708 709 // Hexagon's packets require relocations to be inline rather than 710 // clustered at the end of the packet. 711 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 712 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 713 auto PrintReloc = [&]() -> void { 714 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 715 if (RelCur->getOffset() == Address.Address) { 716 printRelocation(*RelCur, Address.Address, false); 717 return; 718 } 719 ++RelCur; 720 } 721 }; 722 723 while (!HeadTail.first.empty()) { 724 OS << Separator; 725 Separator = "\n"; 726 if (SP && (PrintSource || PrintLines)) 727 SP->printSourceLine(OS, Address, ""); 728 printLead(Bytes, Address.Address, OS); 729 OS << Preamble; 730 Preamble = " "; 731 StringRef Inst; 732 auto Duplex = HeadTail.first.split('\v'); 733 if (!Duplex.second.empty()) { 734 OS << Duplex.first; 735 OS << "; "; 736 Inst = Duplex.second; 737 } 738 else 739 Inst = HeadTail.first; 740 OS << Inst; 741 HeadTail = HeadTail.second.split('\n'); 742 if (HeadTail.first.empty()) 743 OS << " } " << PacketBundle.second; 744 PrintReloc(); 745 Bytes = Bytes.slice(4); 746 Address.Address += 4; 747 } 748 } 749 }; 750 HexagonPrettyPrinter HexagonPrettyPrinterInst; 751 752 class AMDGCNPrettyPrinter : public PrettyPrinter { 753 public: 754 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 755 object::SectionedAddress Address, raw_ostream &OS, 756 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 757 std::vector<RelocationRef> *Rels) override { 758 if (SP && (PrintSource || PrintLines)) 759 SP->printSourceLine(OS, Address); 760 761 if (MI) { 762 SmallString<40> InstStr; 763 raw_svector_ostream IS(InstStr); 764 765 IP.printInst(MI, IS, "", STI); 766 767 OS << left_justify(IS.str(), 60); 768 } else { 769 // an unrecognized encoding - this is probably data so represent it 770 // using the .long directive, or .byte directive if fewer than 4 bytes 771 // remaining 772 if (Bytes.size() >= 4) { 773 OS << format("\t.long 0x%08" PRIx32 " ", 774 support::endian::read32<support::little>(Bytes.data())); 775 OS.indent(42); 776 } else { 777 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 778 for (unsigned int i = 1; i < Bytes.size(); i++) 779 OS << format(", 0x%02" PRIx8, Bytes[i]); 780 OS.indent(55 - (6 * Bytes.size())); 781 } 782 } 783 784 OS << format("// %012" PRIX64 ":", Address.Address); 785 if (Bytes.size() >= 4) { 786 // D should be casted to uint32_t here as it is passed by format to 787 // snprintf as vararg. 788 for (uint32_t D : makeArrayRef( 789 reinterpret_cast<const support::little32_t *>(Bytes.data()), 790 Bytes.size() / 4)) 791 OS << format(" %08" PRIX32, D); 792 } else { 793 for (unsigned char B : Bytes) 794 OS << format(" %02" PRIX8, B); 795 } 796 797 if (!Annot.empty()) 798 OS << " // " << Annot; 799 } 800 }; 801 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 802 803 class BPFPrettyPrinter : public PrettyPrinter { 804 public: 805 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 806 object::SectionedAddress Address, raw_ostream &OS, 807 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 808 std::vector<RelocationRef> *Rels) override { 809 if (SP && (PrintSource || PrintLines)) 810 SP->printSourceLine(OS, Address); 811 if (!NoLeadingAddr) 812 OS << format("%8" PRId64 ":", Address.Address / 8); 813 if (!NoShowRawInsn) { 814 OS << "\t"; 815 dumpBytes(Bytes, OS); 816 } 817 if (MI) 818 IP.printInst(MI, OS, "", STI); 819 else 820 OS << "\t<unknown>"; 821 } 822 }; 823 BPFPrettyPrinter BPFPrettyPrinterInst; 824 825 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 826 switch(Triple.getArch()) { 827 default: 828 return PrettyPrinterInst; 829 case Triple::hexagon: 830 return HexagonPrettyPrinterInst; 831 case Triple::amdgcn: 832 return AMDGCNPrettyPrinterInst; 833 case Triple::bpfel: 834 case Triple::bpfeb: 835 return BPFPrettyPrinterInst; 836 } 837 } 838 } 839 840 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 841 assert(Obj->isELF()); 842 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 843 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 844 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 845 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 846 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 847 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 848 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 849 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 850 llvm_unreachable("Unsupported binary format"); 851 } 852 853 template <class ELFT> static void 854 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 855 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 856 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 857 uint8_t SymbolType = Symbol.getELFType(); 858 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 859 continue; 860 861 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 862 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 863 if (Name.empty()) 864 continue; 865 866 section_iterator SecI = 867 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 868 if (SecI == Obj->section_end()) 869 continue; 870 871 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 872 } 873 } 874 875 static void 876 addDynamicElfSymbols(const ObjectFile *Obj, 877 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 878 assert(Obj->isELF()); 879 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 880 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 881 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 882 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 883 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 884 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 885 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 886 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 887 else 888 llvm_unreachable("Unsupported binary format"); 889 } 890 891 static void addPltEntries(const ObjectFile *Obj, 892 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 893 StringSaver &Saver) { 894 Optional<SectionRef> Plt = None; 895 for (const SectionRef &Section : Obj->sections()) { 896 StringRef Name; 897 if (Section.getName(Name)) 898 continue; 899 if (Name == ".plt") 900 Plt = Section; 901 } 902 if (!Plt) 903 return; 904 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 905 for (auto PltEntry : ElfObj->getPltAddresses()) { 906 SymbolRef Symbol(PltEntry.first, ElfObj); 907 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 908 909 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 910 if (!Name.empty()) 911 AllSymbols[*Plt].emplace_back( 912 PltEntry.second, Saver.save((Name + "@plt").str()), SymbolType); 913 } 914 } 915 } 916 917 // Normally the disassembly output will skip blocks of zeroes. This function 918 // returns the number of zero bytes that can be skipped when dumping the 919 // disassembly of the instructions in Buf. 920 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 921 // Find the number of leading zeroes. 922 size_t N = 0; 923 while (N < Buf.size() && !Buf[N]) 924 ++N; 925 926 // We may want to skip blocks of zero bytes, but unless we see 927 // at least 8 of them in a row. 928 if (N < 8) 929 return 0; 930 931 // We skip zeroes in multiples of 4 because do not want to truncate an 932 // instruction if it starts with a zero byte. 933 return N & ~0x3; 934 } 935 936 // Returns a map from sections to their relocations. 937 static std::map<SectionRef, std::vector<RelocationRef>> 938 getRelocsMap(object::ObjectFile const &Obj) { 939 std::map<SectionRef, std::vector<RelocationRef>> Ret; 940 for (SectionRef Sec : Obj.sections()) { 941 section_iterator Relocated = Sec.getRelocatedSection(); 942 if (Relocated == Obj.section_end() || !shouldKeep(*Relocated)) 943 continue; 944 std::vector<RelocationRef> &V = Ret[*Relocated]; 945 for (const RelocationRef &R : Sec.relocations()) 946 V.push_back(R); 947 // Sort relocations by address. 948 llvm::stable_sort(V, isRelocAddressLess); 949 } 950 return Ret; 951 } 952 953 // Used for --adjust-vma to check if address should be adjusted by the 954 // specified value for a given section. 955 // For ELF we do not adjust non-allocatable sections like debug ones, 956 // because they are not loadable. 957 // TODO: implement for other file formats. 958 static bool shouldAdjustVA(const SectionRef &Section) { 959 const ObjectFile *Obj = Section.getObject(); 960 if (isa<object::ELFObjectFileBase>(Obj)) 961 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 962 return false; 963 } 964 965 966 typedef std::pair<uint64_t, char> MappingSymbolPair; 967 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols, 968 uint64_t Address) { 969 auto Sym = bsearch(MappingSymbols, [Address](const MappingSymbolPair &Val) { 970 return Val.first > Address; 971 }); 972 // Return zero for any address before the first mapping symbol; this means 973 // we should use the default disassembly mode, depending on the target. 974 if (Sym == MappingSymbols.begin()) 975 return '\x00'; 976 return (Sym - 1)->second; 977 } 978 979 static uint64_t 980 dumpARMELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 981 const ObjectFile *Obj, ArrayRef<uint8_t> Bytes, 982 ArrayRef<MappingSymbolPair> MappingSymbols) { 983 support::endianness Endian = 984 Obj->isLittleEndian() ? support::little : support::big; 985 while (Index < End) { 986 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 987 outs() << "\t"; 988 if (Index + 4 <= End) { 989 dumpBytes(Bytes.slice(Index, 4), outs()); 990 outs() << "\t.word\t" 991 << format_hex( 992 support::endian::read32(Bytes.data() + Index, Endian), 10); 993 Index += 4; 994 } else if (Index + 2 <= End) { 995 dumpBytes(Bytes.slice(Index, 2), outs()); 996 outs() << "\t\t.short\t" 997 << format_hex( 998 support::endian::read16(Bytes.data() + Index, Endian), 6); 999 Index += 2; 1000 } else { 1001 dumpBytes(Bytes.slice(Index, 1), outs()); 1002 outs() << "\t\t.byte\t" << format_hex(Bytes[0], 4); 1003 ++Index; 1004 } 1005 outs() << "\n"; 1006 if (getMappingSymbolKind(MappingSymbols, Index) != 'd') 1007 break; 1008 } 1009 return Index; 1010 } 1011 1012 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 1013 ArrayRef<uint8_t> Bytes) { 1014 // print out data up to 8 bytes at a time in hex and ascii 1015 uint8_t AsciiData[9] = {'\0'}; 1016 uint8_t Byte; 1017 int NumBytes = 0; 1018 1019 for (; Index < End; ++Index) { 1020 if (NumBytes == 0) 1021 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1022 Byte = Bytes.slice(Index)[0]; 1023 outs() << format(" %02x", Byte); 1024 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1025 1026 uint8_t IndentOffset = 0; 1027 NumBytes++; 1028 if (Index == End - 1 || NumBytes > 8) { 1029 // Indent the space for less than 8 bytes data. 1030 // 2 spaces for byte and one for space between bytes 1031 IndentOffset = 3 * (8 - NumBytes); 1032 for (int Excess = NumBytes; Excess < 8; Excess++) 1033 AsciiData[Excess] = '\0'; 1034 NumBytes = 8; 1035 } 1036 if (NumBytes == 8) { 1037 AsciiData[8] = '\0'; 1038 outs() << std::string(IndentOffset, ' ') << " "; 1039 outs() << reinterpret_cast<char *>(AsciiData); 1040 outs() << '\n'; 1041 NumBytes = 0; 1042 } 1043 } 1044 } 1045 1046 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 1047 MCContext &Ctx, MCDisassembler *PrimaryDisAsm, 1048 MCDisassembler *SecondaryDisAsm, 1049 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 1050 const MCSubtargetInfo *PrimarySTI, 1051 const MCSubtargetInfo *SecondarySTI, 1052 PrettyPrinter &PIP, 1053 SourcePrinter &SP, bool InlineRelocs) { 1054 const MCSubtargetInfo *STI = PrimarySTI; 1055 MCDisassembler *DisAsm = PrimaryDisAsm; 1056 bool PrimaryIsThumb = false; 1057 if (isArmElf(Obj)) 1058 PrimaryIsThumb = STI->checkFeatures("+thumb-mode"); 1059 1060 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 1061 if (InlineRelocs) 1062 RelocMap = getRelocsMap(*Obj); 1063 bool Is64Bits = Obj->getBytesInAddress() > 4; 1064 1065 // Create a mapping from virtual address to symbol name. This is used to 1066 // pretty print the symbols while disassembling. 1067 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1068 SectionSymbolsTy AbsoluteSymbols; 1069 const StringRef FileName = Obj->getFileName(); 1070 for (const SymbolRef &Symbol : Obj->symbols()) { 1071 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName); 1072 1073 StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1074 if (Name.empty()) 1075 continue; 1076 1077 uint8_t SymbolType = ELF::STT_NOTYPE; 1078 if (Obj->isELF()) { 1079 SymbolType = getElfSymbolType(Obj, Symbol); 1080 if (SymbolType == ELF::STT_SECTION) 1081 continue; 1082 } 1083 1084 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1085 if (SecI != Obj->section_end()) 1086 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 1087 else 1088 AbsoluteSymbols.emplace_back(Address, Name, SymbolType); 1089 } 1090 if (AllSymbols.empty() && Obj->isELF()) 1091 addDynamicElfSymbols(Obj, AllSymbols); 1092 1093 BumpPtrAllocator A; 1094 StringSaver Saver(A); 1095 addPltEntries(Obj, AllSymbols, Saver); 1096 1097 // Create a mapping from virtual address to section. 1098 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1099 for (SectionRef Sec : Obj->sections()) 1100 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1101 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1102 1103 // Linked executables (.exe and .dll files) typically don't include a real 1104 // symbol table but they might contain an export table. 1105 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1106 for (const auto &ExportEntry : COFFObj->export_directories()) { 1107 StringRef Name; 1108 error(ExportEntry.getSymbolName(Name)); 1109 if (Name.empty()) 1110 continue; 1111 uint32_t RVA; 1112 error(ExportEntry.getExportRVA(RVA)); 1113 1114 uint64_t VA = COFFObj->getImageBase() + RVA; 1115 auto Sec = llvm::bsearch( 1116 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &RHS) { 1117 return VA < RHS.first; 1118 }); 1119 if (Sec != SectionAddresses.begin()) { 1120 --Sec; 1121 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1122 } else 1123 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1124 } 1125 } 1126 1127 // Sort all the symbols, this allows us to use a simple binary search to find 1128 // a symbol near an address. 1129 StringSet<> FoundDisasmFuncsSet; 1130 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1131 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1132 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1133 1134 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1135 if (FilterSections.empty() && !DisassembleAll && 1136 (!Section.isText() || Section.isVirtual())) 1137 continue; 1138 1139 uint64_t SectionAddr = Section.getAddress(); 1140 uint64_t SectSize = Section.getSize(); 1141 if (!SectSize) 1142 continue; 1143 1144 // Get the list of all the symbols in this section. 1145 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1146 std::vector<MappingSymbolPair> MappingSymbols; 1147 if (hasMappingSymbols(Obj)) { 1148 for (const auto &Symb : Symbols) { 1149 uint64_t Address = std::get<0>(Symb); 1150 StringRef Name = std::get<1>(Symb); 1151 if (Name.startswith("$d")) 1152 MappingSymbols.emplace_back(Address - SectionAddr, 'd'); 1153 if (Name.startswith("$x")) 1154 MappingSymbols.emplace_back(Address - SectionAddr, 'x'); 1155 if (Name.startswith("$a")) 1156 MappingSymbols.emplace_back(Address - SectionAddr, 'a'); 1157 if (Name.startswith("$t")) 1158 MappingSymbols.emplace_back(Address - SectionAddr, 't'); 1159 } 1160 } 1161 1162 llvm::sort(MappingSymbols); 1163 1164 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1165 // AMDGPU disassembler uses symbolizer for printing labels 1166 std::unique_ptr<MCRelocationInfo> RelInfo( 1167 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1168 if (RelInfo) { 1169 std::unique_ptr<MCSymbolizer> Symbolizer( 1170 TheTarget->createMCSymbolizer( 1171 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1172 DisAsm->setSymbolizer(std::move(Symbolizer)); 1173 } 1174 } 1175 1176 StringRef SegmentName = ""; 1177 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1178 DataRefImpl DR = Section.getRawDataRefImpl(); 1179 SegmentName = MachO->getSectionFinalSegmentName(DR); 1180 } 1181 StringRef SectionName; 1182 error(Section.getName(SectionName)); 1183 1184 // If the section has no symbol at the start, just insert a dummy one. 1185 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1186 Symbols.insert( 1187 Symbols.begin(), 1188 std::make_tuple(SectionAddr, SectionName, 1189 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1190 } 1191 1192 SmallString<40> Comments; 1193 raw_svector_ostream CommentStream(Comments); 1194 1195 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef( 1196 unwrapOrError(Section.getContents(), Obj->getFileName())); 1197 1198 uint64_t VMAAdjustment = 0; 1199 if (shouldAdjustVA(Section)) 1200 VMAAdjustment = AdjustVMA; 1201 1202 uint64_t Size; 1203 uint64_t Index; 1204 bool PrintedSection = false; 1205 std::vector<RelocationRef> Rels = RelocMap[Section]; 1206 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1207 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1208 // Disassemble symbol by symbol. 1209 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1210 // Skip if --disassemble-functions is not empty and the symbol is not in 1211 // the list. 1212 if (!DisasmFuncsSet.empty() && 1213 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1214 continue; 1215 1216 uint64_t Start = std::get<0>(Symbols[SI]); 1217 if (Start < SectionAddr || StopAddress <= Start) 1218 continue; 1219 else 1220 FoundDisasmFuncsSet.insert(std::get<1>(Symbols[SI])); 1221 1222 // The end is the section end, the beginning of the next symbol, or 1223 // --stop-address. 1224 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress); 1225 if (SI + 1 < SE) 1226 End = std::min(End, std::get<0>(Symbols[SI + 1])); 1227 if (Start >= End || End <= StartAddress) 1228 continue; 1229 Start -= SectionAddr; 1230 End -= SectionAddr; 1231 1232 if (!PrintedSection) { 1233 PrintedSection = true; 1234 outs() << "\nDisassembly of section "; 1235 if (!SegmentName.empty()) 1236 outs() << SegmentName << ","; 1237 outs() << SectionName << ":\n"; 1238 } 1239 1240 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1241 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1242 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1243 Start += 256; 1244 } 1245 if (SI == SE - 1 || 1246 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1247 // cut trailing zeroes at the end of kernel 1248 // cut up to 256 bytes 1249 const uint64_t EndAlign = 256; 1250 const auto Limit = End - (std::min)(EndAlign, End - Start); 1251 while (End > Limit && 1252 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1253 End -= 4; 1254 } 1255 } 1256 1257 outs() << '\n'; 1258 if (!NoLeadingAddr) 1259 outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ", 1260 SectionAddr + Start + VMAAdjustment); 1261 1262 StringRef SymbolName = std::get<1>(Symbols[SI]); 1263 if (Demangle) 1264 outs() << demangle(SymbolName) << ":\n"; 1265 else 1266 outs() << SymbolName << ":\n"; 1267 1268 // Don't print raw contents of a virtual section. A virtual section 1269 // doesn't have any contents in the file. 1270 if (Section.isVirtual()) { 1271 outs() << "...\n"; 1272 continue; 1273 } 1274 1275 #ifndef NDEBUG 1276 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1277 #else 1278 raw_ostream &DebugOut = nulls(); 1279 #endif 1280 1281 // Some targets (like WebAssembly) have a special prelude at the start 1282 // of each symbol. 1283 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1284 SectionAddr + Start, DebugOut, CommentStream); 1285 Start += Size; 1286 1287 Index = Start; 1288 if (SectionAddr < StartAddress) 1289 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1290 1291 // If there is a data symbol inside an ELF text section and we are 1292 // only disassembling text (applicable all architectures), we are in a 1293 // situation where we must print the data and not disassemble it. 1294 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1295 !DisassembleAll && Section.isText()) { 1296 dumpELFData(SectionAddr, Index, End, Bytes); 1297 Index = End; 1298 } 1299 1300 bool CheckARMELFData = hasMappingSymbols(Obj) && 1301 std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1302 !DisassembleAll; 1303 while (Index < End) { 1304 // ARM and AArch64 ELF binaries can interleave data and text in the 1305 // same section. We rely on the markers introduced to understand what 1306 // we need to dump. If the data marker is within a function, it is 1307 // denoted as a word/short etc. 1308 if (CheckARMELFData && 1309 getMappingSymbolKind(MappingSymbols, Index) == 'd') { 1310 Index = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1311 MappingSymbols); 1312 continue; 1313 } 1314 1315 // When -z or --disassemble-zeroes are given we always dissasemble 1316 // them. Otherwise we might want to skip zero bytes we see. 1317 if (!DisassembleZeroes) { 1318 uint64_t MaxOffset = End - Index; 1319 // For -reloc: print zero blocks patched by relocations, so that 1320 // relocations can be shown in the dump. 1321 if (RelCur != RelEnd) 1322 MaxOffset = RelCur->getOffset() - Index; 1323 1324 if (size_t N = 1325 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1326 outs() << "\t\t..." << '\n'; 1327 Index += N; 1328 continue; 1329 } 1330 } 1331 1332 if (SecondarySTI) { 1333 if (getMappingSymbolKind(MappingSymbols, Index) == 'a') { 1334 STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI; 1335 DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm; 1336 } else if (getMappingSymbolKind(MappingSymbols, Index) == 't') { 1337 STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI; 1338 DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm; 1339 } 1340 } 1341 1342 // Disassemble a real instruction or a data when disassemble all is 1343 // provided 1344 MCInst Inst; 1345 bool Disassembled = DisAsm->getInstruction( 1346 Inst, Size, Bytes.slice(Index), SectionAddr + Index, DebugOut, 1347 CommentStream); 1348 if (Size == 0) 1349 Size = 1; 1350 1351 PIP.printInst( 1352 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 1353 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, outs(), 1354 "", *STI, &SP, &Rels); 1355 outs() << CommentStream.str(); 1356 Comments.clear(); 1357 1358 // Try to resolve the target of a call, tail call, etc. to a specific 1359 // symbol. 1360 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1361 MIA->isConditionalBranch(Inst))) { 1362 uint64_t Target; 1363 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1364 // In a relocatable object, the target's section must reside in 1365 // the same section as the call instruction or it is accessed 1366 // through a relocation. 1367 // 1368 // In a non-relocatable object, the target may be in any section. 1369 // 1370 // N.B. We don't walk the relocations in the relocatable case yet. 1371 auto *TargetSectionSymbols = &Symbols; 1372 if (!Obj->isRelocatableObject()) { 1373 auto It = llvm::bsearch( 1374 SectionAddresses, 1375 [=](const std::pair<uint64_t, SectionRef> &RHS) { 1376 return Target < RHS.first; 1377 }); 1378 if (It != SectionAddresses.begin()) { 1379 --It; 1380 TargetSectionSymbols = &AllSymbols[It->second]; 1381 } else { 1382 TargetSectionSymbols = &AbsoluteSymbols; 1383 } 1384 } 1385 1386 // Find the last symbol in the section whose offset is less than 1387 // or equal to the target. If there isn't a section that contains 1388 // the target, find the nearest preceding absolute symbol. 1389 auto TargetSym = llvm::bsearch( 1390 *TargetSectionSymbols, 1391 [=](const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1392 return Target < std::get<0>(RHS); 1393 }); 1394 if (TargetSym == TargetSectionSymbols->begin()) { 1395 TargetSectionSymbols = &AbsoluteSymbols; 1396 TargetSym = llvm::bsearch( 1397 AbsoluteSymbols, 1398 [=](const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1399 return Target < std::get<0>(RHS); 1400 }); 1401 } 1402 if (TargetSym != TargetSectionSymbols->begin()) { 1403 --TargetSym; 1404 uint64_t TargetAddress = std::get<0>(*TargetSym); 1405 StringRef TargetName = std::get<1>(*TargetSym); 1406 outs() << " <" << TargetName; 1407 uint64_t Disp = Target - TargetAddress; 1408 if (Disp) 1409 outs() << "+0x" << Twine::utohexstr(Disp); 1410 outs() << '>'; 1411 } 1412 } 1413 } 1414 outs() << "\n"; 1415 1416 // Hexagon does this in pretty printer 1417 if (Obj->getArch() != Triple::hexagon) { 1418 // Print relocation for instruction. 1419 while (RelCur != RelEnd) { 1420 uint64_t Offset = RelCur->getOffset(); 1421 // If this relocation is hidden, skip it. 1422 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 1423 ++RelCur; 1424 continue; 1425 } 1426 1427 // Stop when RelCur's offset is past the current instruction. 1428 if (Offset >= Index + Size) 1429 break; 1430 1431 // When --adjust-vma is used, update the address printed. 1432 if (RelCur->getSymbol() != Obj->symbol_end()) { 1433 Expected<section_iterator> SymSI = 1434 RelCur->getSymbol()->getSection(); 1435 if (SymSI && *SymSI != Obj->section_end() && 1436 shouldAdjustVA(**SymSI)) 1437 Offset += AdjustVMA; 1438 } 1439 1440 printRelocation(*RelCur, SectionAddr + Offset, Is64Bits); 1441 ++RelCur; 1442 } 1443 } 1444 1445 Index += Size; 1446 } 1447 } 1448 } 1449 StringSet<> MissingDisasmFuncsSet = 1450 set_difference(DisasmFuncsSet, FoundDisasmFuncsSet); 1451 for (StringRef MissingDisasmFunc : MissingDisasmFuncsSet.keys()) 1452 warn("failed to disassemble missing function " + MissingDisasmFunc); 1453 } 1454 1455 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1456 if (StartAddress >= StopAddress) 1457 error("start address should be less than stop address"); 1458 1459 const Target *TheTarget = getTarget(Obj); 1460 1461 // Package up features to be passed to target/subtarget 1462 SubtargetFeatures Features = Obj->getFeatures(); 1463 if (!MAttrs.empty()) 1464 for (unsigned I = 0; I != MAttrs.size(); ++I) 1465 Features.AddFeature(MAttrs[I]); 1466 1467 std::unique_ptr<const MCRegisterInfo> MRI( 1468 TheTarget->createMCRegInfo(TripleName)); 1469 if (!MRI) 1470 report_error(Obj->getFileName(), 1471 "no register info for target " + TripleName); 1472 1473 // Set up disassembler. 1474 std::unique_ptr<const MCAsmInfo> AsmInfo( 1475 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1476 if (!AsmInfo) 1477 report_error(Obj->getFileName(), 1478 "no assembly info for target " + TripleName); 1479 std::unique_ptr<const MCSubtargetInfo> STI( 1480 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1481 if (!STI) 1482 report_error(Obj->getFileName(), 1483 "no subtarget info for target " + TripleName); 1484 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1485 if (!MII) 1486 report_error(Obj->getFileName(), 1487 "no instruction info for target " + TripleName); 1488 MCObjectFileInfo MOFI; 1489 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1490 // FIXME: for now initialize MCObjectFileInfo with default values 1491 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1492 1493 std::unique_ptr<MCDisassembler> DisAsm( 1494 TheTarget->createMCDisassembler(*STI, Ctx)); 1495 if (!DisAsm) 1496 report_error(Obj->getFileName(), 1497 "no disassembler for target " + TripleName); 1498 1499 // If we have an ARM object file, we need a second disassembler, because 1500 // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode. 1501 // We use mapping symbols to switch between the two assemblers, where 1502 // appropriate. 1503 std::unique_ptr<MCDisassembler> SecondaryDisAsm; 1504 std::unique_ptr<const MCSubtargetInfo> SecondarySTI; 1505 if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) { 1506 if (STI->checkFeatures("+thumb-mode")) 1507 Features.AddFeature("-thumb-mode"); 1508 else 1509 Features.AddFeature("+thumb-mode"); 1510 SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU, 1511 Features.getString())); 1512 SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx)); 1513 } 1514 1515 std::unique_ptr<const MCInstrAnalysis> MIA( 1516 TheTarget->createMCInstrAnalysis(MII.get())); 1517 1518 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1519 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1520 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1521 if (!IP) 1522 report_error(Obj->getFileName(), 1523 "no instruction printer for target " + TripleName); 1524 IP->setPrintImmHex(PrintImmHex); 1525 1526 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1527 SourcePrinter SP(Obj, TheTarget->getName()); 1528 1529 for (StringRef Opt : DisassemblerOptions) 1530 if (!IP->applyTargetSpecificCLOption(Opt)) 1531 error("Unrecognized disassembler option: " + Opt); 1532 1533 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(), 1534 MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP, 1535 SP, InlineRelocs); 1536 } 1537 1538 void printRelocations(const ObjectFile *Obj) { 1539 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1540 "%08" PRIx64; 1541 // Regular objdump doesn't print relocations in non-relocatable object 1542 // files. 1543 if (!Obj->isRelocatableObject()) 1544 return; 1545 1546 // Build a mapping from relocation target to a vector of relocation 1547 // sections. Usually, there is an only one relocation section for 1548 // each relocated section. 1549 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec; 1550 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1551 if (Section.relocation_begin() == Section.relocation_end()) 1552 continue; 1553 const SectionRef TargetSec = *Section.getRelocatedSection(); 1554 SecToRelSec[TargetSec].push_back(Section); 1555 } 1556 1557 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) { 1558 StringRef SecName; 1559 error(P.first.getName(SecName)); 1560 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1561 1562 for (SectionRef Section : P.second) { 1563 for (const RelocationRef &Reloc : Section.relocations()) { 1564 uint64_t Address = Reloc.getOffset(); 1565 SmallString<32> RelocName; 1566 SmallString<32> ValueStr; 1567 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1568 continue; 1569 Reloc.getTypeName(RelocName); 1570 error(getRelocationValueString(Reloc, ValueStr)); 1571 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1572 << ValueStr << "\n"; 1573 } 1574 } 1575 outs() << "\n"; 1576 } 1577 } 1578 1579 void printDynamicRelocations(const ObjectFile *Obj) { 1580 // For the moment, this option is for ELF only 1581 if (!Obj->isELF()) 1582 return; 1583 1584 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1585 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1586 error("not a dynamic object"); 1587 return; 1588 } 1589 1590 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1591 if (DynRelSec.empty()) 1592 return; 1593 1594 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1595 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1596 for (const SectionRef &Section : DynRelSec) 1597 for (const RelocationRef &Reloc : Section.relocations()) { 1598 uint64_t Address = Reloc.getOffset(); 1599 SmallString<32> RelocName; 1600 SmallString<32> ValueStr; 1601 Reloc.getTypeName(RelocName); 1602 error(getRelocationValueString(Reloc, ValueStr)); 1603 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1604 << ValueStr << "\n"; 1605 } 1606 } 1607 1608 // Returns true if we need to show LMA column when dumping section headers. We 1609 // show it only when the platform is ELF and either we have at least one section 1610 // whose VMA and LMA are different and/or when --show-lma flag is used. 1611 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1612 if (!Obj->isELF()) 1613 return false; 1614 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1615 if (S.getAddress() != getELFSectionLMA(S)) 1616 return true; 1617 return ShowLMA; 1618 } 1619 1620 void printSectionHeaders(const ObjectFile *Obj) { 1621 bool HasLMAColumn = shouldDisplayLMA(Obj); 1622 if (HasLMAColumn) 1623 outs() << "Sections:\n" 1624 "Idx Name Size VMA LMA " 1625 "Type\n"; 1626 else 1627 outs() << "Sections:\n" 1628 "Idx Name Size VMA Type\n"; 1629 1630 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1631 StringRef Name; 1632 error(Section.getName(Name)); 1633 uint64_t VMA = Section.getAddress(); 1634 if (shouldAdjustVA(Section)) 1635 VMA += AdjustVMA; 1636 1637 uint64_t Size = Section.getSize(); 1638 bool Text = Section.isText(); 1639 bool Data = Section.isData(); 1640 bool BSS = Section.isBSS(); 1641 std::string Type = (std::string(Text ? "TEXT " : "") + 1642 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1643 1644 if (HasLMAColumn) 1645 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %016" PRIx64 1646 " %s\n", 1647 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1648 VMA, getELFSectionLMA(Section), Type.c_str()); 1649 else 1650 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1651 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1652 VMA, Type.c_str()); 1653 } 1654 outs() << "\n"; 1655 } 1656 1657 void printSectionContents(const ObjectFile *Obj) { 1658 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1659 StringRef Name; 1660 error(Section.getName(Name)); 1661 uint64_t BaseAddr = Section.getAddress(); 1662 uint64_t Size = Section.getSize(); 1663 if (!Size) 1664 continue; 1665 1666 outs() << "Contents of section " << Name << ":\n"; 1667 if (Section.isBSS()) { 1668 outs() << format("<skipping contents of bss section at [%04" PRIx64 1669 ", %04" PRIx64 ")>\n", 1670 BaseAddr, BaseAddr + Size); 1671 continue; 1672 } 1673 1674 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName()); 1675 1676 // Dump out the content as hex and printable ascii characters. 1677 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1678 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1679 // Dump line of hex. 1680 for (std::size_t I = 0; I < 16; ++I) { 1681 if (I != 0 && I % 4 == 0) 1682 outs() << ' '; 1683 if (Addr + I < End) 1684 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1685 << hexdigit(Contents[Addr + I] & 0xF, true); 1686 else 1687 outs() << " "; 1688 } 1689 // Print ascii. 1690 outs() << " "; 1691 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1692 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1693 outs() << Contents[Addr + I]; 1694 else 1695 outs() << "."; 1696 } 1697 outs() << "\n"; 1698 } 1699 } 1700 } 1701 1702 void printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1703 StringRef ArchitectureName) { 1704 outs() << "SYMBOL TABLE:\n"; 1705 1706 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1707 printCOFFSymbolTable(Coff); 1708 return; 1709 } 1710 1711 const StringRef FileName = O->getFileName(); 1712 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1713 const SymbolRef &Symbol = *I; 1714 uint64_t Address = unwrapOrError(Symbol.getAddress(), ArchiveName, FileName, 1715 ArchitectureName); 1716 if ((Address < StartAddress) || (Address > StopAddress)) 1717 continue; 1718 SymbolRef::Type Type = unwrapOrError(Symbol.getType(), ArchiveName, 1719 FileName, ArchitectureName); 1720 uint32_t Flags = Symbol.getFlags(); 1721 section_iterator Section = unwrapOrError(Symbol.getSection(), ArchiveName, 1722 FileName, ArchitectureName); 1723 StringRef Name; 1724 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) 1725 Section->getName(Name); 1726 else 1727 Name = unwrapOrError(Symbol.getName(), ArchiveName, FileName, 1728 ArchitectureName); 1729 1730 bool Global = Flags & SymbolRef::SF_Global; 1731 bool Weak = Flags & SymbolRef::SF_Weak; 1732 bool Absolute = Flags & SymbolRef::SF_Absolute; 1733 bool Common = Flags & SymbolRef::SF_Common; 1734 bool Hidden = Flags & SymbolRef::SF_Hidden; 1735 1736 char GlobLoc = ' '; 1737 if (Type != SymbolRef::ST_Unknown) 1738 GlobLoc = Global ? 'g' : 'l'; 1739 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1740 ? 'd' : ' '; 1741 char FileFunc = ' '; 1742 if (Type == SymbolRef::ST_File) 1743 FileFunc = 'f'; 1744 else if (Type == SymbolRef::ST_Function) 1745 FileFunc = 'F'; 1746 else if (Type == SymbolRef::ST_Data) 1747 FileFunc = 'O'; 1748 1749 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1750 "%08" PRIx64; 1751 1752 outs() << format(Fmt, Address) << " " 1753 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1754 << (Weak ? 'w' : ' ') // Weak? 1755 << ' ' // Constructor. Not supported yet. 1756 << ' ' // Warning. Not supported yet. 1757 << ' ' // Indirect reference to another symbol. 1758 << Debug // Debugging (d) or dynamic (D) symbol. 1759 << FileFunc // Name of function (F), file (f) or object (O). 1760 << ' '; 1761 if (Absolute) { 1762 outs() << "*ABS*"; 1763 } else if (Common) { 1764 outs() << "*COM*"; 1765 } else if (Section == O->section_end()) { 1766 outs() << "*UND*"; 1767 } else { 1768 if (const MachOObjectFile *MachO = 1769 dyn_cast<const MachOObjectFile>(O)) { 1770 DataRefImpl DR = Section->getRawDataRefImpl(); 1771 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1772 outs() << SegmentName << ","; 1773 } 1774 StringRef SectionName; 1775 error(Section->getName(SectionName)); 1776 outs() << SectionName; 1777 } 1778 1779 if (Common || isa<ELFObjectFileBase>(O)) { 1780 uint64_t Val = 1781 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1782 outs() << format("\t%08" PRIx64, Val); 1783 } 1784 1785 if (isa<ELFObjectFileBase>(O)) { 1786 uint8_t Other = ELFSymbolRef(Symbol).getOther(); 1787 switch (Other) { 1788 case ELF::STV_DEFAULT: 1789 break; 1790 case ELF::STV_INTERNAL: 1791 outs() << " .internal"; 1792 break; 1793 case ELF::STV_HIDDEN: 1794 outs() << " .hidden"; 1795 break; 1796 case ELF::STV_PROTECTED: 1797 outs() << " .protected"; 1798 break; 1799 default: 1800 outs() << format(" 0x%02x", Other); 1801 break; 1802 } 1803 } else if (Hidden) { 1804 outs() << " .hidden"; 1805 } 1806 1807 if (Demangle) 1808 outs() << ' ' << demangle(Name) << '\n'; 1809 else 1810 outs() << ' ' << Name << '\n'; 1811 } 1812 } 1813 1814 static void printUnwindInfo(const ObjectFile *O) { 1815 outs() << "Unwind info:\n\n"; 1816 1817 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1818 printCOFFUnwindInfo(Coff); 1819 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1820 printMachOUnwindInfo(MachO); 1821 else 1822 // TODO: Extract DWARF dump tool to objdump. 1823 WithColor::error(errs(), ToolName) 1824 << "This operation is only currently supported " 1825 "for COFF and MachO object files.\n"; 1826 } 1827 1828 /// Dump the raw contents of the __clangast section so the output can be piped 1829 /// into llvm-bcanalyzer. 1830 void printRawClangAST(const ObjectFile *Obj) { 1831 if (outs().is_displayed()) { 1832 WithColor::error(errs(), ToolName) 1833 << "The -raw-clang-ast option will dump the raw binary contents of " 1834 "the clang ast section.\n" 1835 "Please redirect the output to a file or another program such as " 1836 "llvm-bcanalyzer.\n"; 1837 return; 1838 } 1839 1840 StringRef ClangASTSectionName("__clangast"); 1841 if (isa<COFFObjectFile>(Obj)) { 1842 ClangASTSectionName = "clangast"; 1843 } 1844 1845 Optional<object::SectionRef> ClangASTSection; 1846 for (auto Sec : ToolSectionFilter(*Obj)) { 1847 StringRef Name; 1848 Sec.getName(Name); 1849 if (Name == ClangASTSectionName) { 1850 ClangASTSection = Sec; 1851 break; 1852 } 1853 } 1854 if (!ClangASTSection) 1855 return; 1856 1857 StringRef ClangASTContents = unwrapOrError( 1858 ClangASTSection.getValue().getContents(), Obj->getFileName()); 1859 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1860 } 1861 1862 static void printFaultMaps(const ObjectFile *Obj) { 1863 StringRef FaultMapSectionName; 1864 1865 if (isa<ELFObjectFileBase>(Obj)) { 1866 FaultMapSectionName = ".llvm_faultmaps"; 1867 } else if (isa<MachOObjectFile>(Obj)) { 1868 FaultMapSectionName = "__llvm_faultmaps"; 1869 } else { 1870 WithColor::error(errs(), ToolName) 1871 << "This operation is only currently supported " 1872 "for ELF and Mach-O executable files.\n"; 1873 return; 1874 } 1875 1876 Optional<object::SectionRef> FaultMapSection; 1877 1878 for (auto Sec : ToolSectionFilter(*Obj)) { 1879 StringRef Name; 1880 Sec.getName(Name); 1881 if (Name == FaultMapSectionName) { 1882 FaultMapSection = Sec; 1883 break; 1884 } 1885 } 1886 1887 outs() << "FaultMap table:\n"; 1888 1889 if (!FaultMapSection.hasValue()) { 1890 outs() << "<not found>\n"; 1891 return; 1892 } 1893 1894 StringRef FaultMapContents = 1895 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName()); 1896 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1897 FaultMapContents.bytes_end()); 1898 1899 outs() << FMP; 1900 } 1901 1902 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1903 if (O->isELF()) { 1904 printELFFileHeader(O); 1905 printELFDynamicSection(O); 1906 printELFSymbolVersionInfo(O); 1907 return; 1908 } 1909 if (O->isCOFF()) 1910 return printCOFFFileHeader(O); 1911 if (O->isWasm()) 1912 return printWasmFileHeader(O); 1913 if (O->isMachO()) { 1914 printMachOFileHeader(O); 1915 if (!OnlyFirst) 1916 printMachOLoadCommands(O); 1917 return; 1918 } 1919 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1920 } 1921 1922 static void printFileHeaders(const ObjectFile *O) { 1923 if (!O->isELF() && !O->isCOFF()) 1924 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1925 1926 Triple::ArchType AT = O->getArch(); 1927 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1928 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 1929 1930 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1931 outs() << "start address: " 1932 << "0x" << format(Fmt.data(), Address) << "\n\n"; 1933 } 1934 1935 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 1936 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 1937 if (!ModeOrErr) { 1938 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 1939 consumeError(ModeOrErr.takeError()); 1940 return; 1941 } 1942 sys::fs::perms Mode = ModeOrErr.get(); 1943 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1944 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1945 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1946 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1947 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1948 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1949 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1950 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1951 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1952 1953 outs() << " "; 1954 1955 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 1956 unwrapOrError(C.getGID(), Filename), 1957 unwrapOrError(C.getRawSize(), Filename)); 1958 1959 StringRef RawLastModified = C.getRawLastModified(); 1960 unsigned Seconds; 1961 if (RawLastModified.getAsInteger(10, Seconds)) 1962 outs() << "(date: \"" << RawLastModified 1963 << "\" contains non-decimal chars) "; 1964 else { 1965 // Since ctime(3) returns a 26 character string of the form: 1966 // "Sun Sep 16 01:03:52 1973\n\0" 1967 // just print 24 characters. 1968 time_t t = Seconds; 1969 outs() << format("%.24s ", ctime(&t)); 1970 } 1971 1972 StringRef Name = ""; 1973 Expected<StringRef> NameOrErr = C.getName(); 1974 if (!NameOrErr) { 1975 consumeError(NameOrErr.takeError()); 1976 Name = unwrapOrError(C.getRawName(), Filename); 1977 } else { 1978 Name = NameOrErr.get(); 1979 } 1980 outs() << Name << "\n"; 1981 } 1982 1983 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 1984 const Archive::Child *C = nullptr) { 1985 // Avoid other output when using a raw option. 1986 if (!RawClangAST) { 1987 outs() << '\n'; 1988 if (A) 1989 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 1990 else 1991 outs() << O->getFileName(); 1992 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 1993 } 1994 1995 StringRef ArchiveName = A ? A->getFileName() : ""; 1996 if (FileHeaders) 1997 printFileHeaders(O); 1998 if (ArchiveHeaders && !MachOOpt && C) 1999 printArchiveChild(ArchiveName, *C); 2000 if (Disassemble) 2001 disassembleObject(O, Relocations); 2002 if (Relocations && !Disassemble) 2003 printRelocations(O); 2004 if (DynamicRelocations) 2005 printDynamicRelocations(O); 2006 if (SectionHeaders) 2007 printSectionHeaders(O); 2008 if (SectionContents) 2009 printSectionContents(O); 2010 if (SymbolTable) 2011 printSymbolTable(O, ArchiveName); 2012 if (UnwindInfo) 2013 printUnwindInfo(O); 2014 if (PrivateHeaders || FirstPrivateHeader) 2015 printPrivateFileHeaders(O, FirstPrivateHeader); 2016 if (ExportsTrie) 2017 printExportsTrie(O); 2018 if (Rebase) 2019 printRebaseTable(O); 2020 if (Bind) 2021 printBindTable(O); 2022 if (LazyBind) 2023 printLazyBindTable(O); 2024 if (WeakBind) 2025 printWeakBindTable(O); 2026 if (RawClangAST) 2027 printRawClangAST(O); 2028 if (FaultMapSection) 2029 printFaultMaps(O); 2030 if (DwarfDumpType != DIDT_Null) { 2031 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2032 // Dump the complete DWARF structure. 2033 DIDumpOptions DumpOpts; 2034 DumpOpts.DumpType = DwarfDumpType; 2035 DICtx->dump(outs(), DumpOpts); 2036 } 2037 } 2038 2039 static void dumpObject(const COFFImportFile *I, const Archive *A, 2040 const Archive::Child *C = nullptr) { 2041 StringRef ArchiveName = A ? A->getFileName() : ""; 2042 2043 // Avoid other output when using a raw option. 2044 if (!RawClangAST) 2045 outs() << '\n' 2046 << ArchiveName << "(" << I->getFileName() << ")" 2047 << ":\tfile format COFF-import-file" 2048 << "\n\n"; 2049 2050 if (ArchiveHeaders && !MachOOpt && C) 2051 printArchiveChild(ArchiveName, *C); 2052 if (SymbolTable) 2053 printCOFFSymbolTable(I); 2054 } 2055 2056 /// Dump each object file in \a a; 2057 static void dumpArchive(const Archive *A) { 2058 Error Err = Error::success(); 2059 for (auto &C : A->children(Err)) { 2060 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2061 if (!ChildOrErr) { 2062 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2063 report_error(std::move(E), A->getFileName(), C); 2064 continue; 2065 } 2066 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2067 dumpObject(O, A, &C); 2068 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2069 dumpObject(I, A, &C); 2070 else 2071 report_error(errorCodeToError(object_error::invalid_file_type), 2072 A->getFileName()); 2073 } 2074 if (Err) 2075 report_error(std::move(Err), A->getFileName()); 2076 } 2077 2078 /// Open file and figure out how to dump it. 2079 static void dumpInput(StringRef file) { 2080 // If we are using the Mach-O specific object file parser, then let it parse 2081 // the file and process the command line options. So the -arch flags can 2082 // be used to select specific slices, etc. 2083 if (MachOOpt) { 2084 parseInputMachO(file); 2085 return; 2086 } 2087 2088 // Attempt to open the binary. 2089 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2090 Binary &Binary = *OBinary.getBinary(); 2091 2092 if (Archive *A = dyn_cast<Archive>(&Binary)) 2093 dumpArchive(A); 2094 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2095 dumpObject(O); 2096 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2097 parseInputMachO(UB); 2098 else 2099 report_error(errorCodeToError(object_error::invalid_file_type), file); 2100 } 2101 } // namespace llvm 2102 2103 int main(int argc, char **argv) { 2104 using namespace llvm; 2105 InitLLVM X(argc, argv); 2106 const cl::OptionCategory *OptionFilters[] = {&ObjdumpCat, &MachOCat}; 2107 cl::HideUnrelatedOptions(OptionFilters); 2108 2109 // Initialize targets and assembly printers/parsers. 2110 InitializeAllTargetInfos(); 2111 InitializeAllTargetMCs(); 2112 InitializeAllDisassemblers(); 2113 2114 // Register the target printer for --version. 2115 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2116 2117 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2118 2119 ToolName = argv[0]; 2120 2121 // Defaults to a.out if no filenames specified. 2122 if (InputFilenames.empty()) 2123 InputFilenames.push_back("a.out"); 2124 2125 if (AllHeaders) 2126 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2127 SectionHeaders = SymbolTable = true; 2128 2129 if (DisassembleAll || PrintSource || PrintLines || 2130 (!DisassembleFunctions.empty())) 2131 Disassemble = true; 2132 2133 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null && 2134 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST && 2135 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable && 2136 !UnwindInfo && !FaultMapSection && 2137 !(MachOOpt && 2138 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie || 2139 FirstPrivateHeader || IndirectSymbols || InfoPlist || LazyBind || 2140 LinkOptHints || ObjcMetaData || Rebase || UniversalHeaders || 2141 WeakBind || !FilterSections.empty()))) { 2142 cl::PrintHelpMessage(); 2143 return 2; 2144 } 2145 2146 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2147 DisassembleFunctions.end()); 2148 2149 llvm::for_each(InputFilenames, dumpInput); 2150 2151 return EXIT_SUCCESS; 2152 } 2153