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/FormatVariadic.h" 55 #include "llvm/Support/GraphWriter.h" 56 #include "llvm/Support/Host.h" 57 #include "llvm/Support/InitLLVM.h" 58 #include "llvm/Support/MemoryBuffer.h" 59 #include "llvm/Support/SourceMgr.h" 60 #include "llvm/Support/StringSaver.h" 61 #include "llvm/Support/TargetRegistry.h" 62 #include "llvm/Support/TargetSelect.h" 63 #include "llvm/Support/WithColor.h" 64 #include "llvm/Support/raw_ostream.h" 65 #include <algorithm> 66 #include <cctype> 67 #include <cstring> 68 #include <system_error> 69 #include <unordered_map> 70 #include <utility> 71 72 using namespace llvm::object; 73 74 namespace llvm { 75 76 cl::OptionCategory ObjdumpCat("llvm-objdump Options"); 77 78 // MachO specific 79 extern cl::OptionCategory MachOCat; 80 extern cl::opt<bool> Bind; 81 extern cl::opt<bool> DataInCode; 82 extern cl::opt<bool> DylibsUsed; 83 extern cl::opt<bool> DylibId; 84 extern cl::opt<bool> ExportsTrie; 85 extern cl::opt<bool> FirstPrivateHeader; 86 extern cl::opt<bool> IndirectSymbols; 87 extern cl::opt<bool> InfoPlist; 88 extern cl::opt<bool> LazyBind; 89 extern cl::opt<bool> LinkOptHints; 90 extern cl::opt<bool> ObjcMetaData; 91 extern cl::opt<bool> Rebase; 92 extern cl::opt<bool> UniversalHeaders; 93 extern cl::opt<bool> WeakBind; 94 95 static cl::opt<uint64_t> AdjustVMA( 96 "adjust-vma", 97 cl::desc("Increase the displayed address by the specified offset"), 98 cl::value_desc("offset"), cl::init(0), cl::cat(ObjdumpCat)); 99 100 static cl::opt<bool> 101 AllHeaders("all-headers", 102 cl::desc("Display all available header information"), 103 cl::cat(ObjdumpCat)); 104 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 105 cl::NotHidden, cl::Grouping, 106 cl::aliasopt(AllHeaders)); 107 108 static cl::opt<std::string> 109 ArchName("arch-name", 110 cl::desc("Target arch to disassemble for, " 111 "see -version for available targets"), 112 cl::cat(ObjdumpCat)); 113 114 cl::opt<bool> ArchiveHeaders("archive-headers", 115 cl::desc("Display archive header information"), 116 cl::cat(ObjdumpCat)); 117 static cl::alias ArchiveHeadersShort("a", 118 cl::desc("Alias for --archive-headers"), 119 cl::NotHidden, cl::Grouping, 120 cl::aliasopt(ArchiveHeaders)); 121 122 cl::opt<bool> Demangle("demangle", cl::desc("Demangle symbols names"), 123 cl::init(false), cl::cat(ObjdumpCat)); 124 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 125 cl::NotHidden, cl::Grouping, 126 cl::aliasopt(Demangle)); 127 128 cl::opt<bool> Disassemble( 129 "disassemble", 130 cl::desc("Display assembler mnemonics for the machine instructions"), 131 cl::cat(ObjdumpCat)); 132 static cl::alias DisassembleShort("d", cl::desc("Alias for --disassemble"), 133 cl::NotHidden, cl::Grouping, 134 cl::aliasopt(Disassemble)); 135 136 cl::opt<bool> DisassembleAll( 137 "disassemble-all", 138 cl::desc("Display assembler mnemonics for the machine instructions"), 139 cl::cat(ObjdumpCat)); 140 static cl::alias DisassembleAllShort("D", 141 cl::desc("Alias for --disassemble-all"), 142 cl::NotHidden, cl::Grouping, 143 cl::aliasopt(DisassembleAll)); 144 145 static cl::list<std::string> 146 DisassembleFunctions("disassemble-functions", cl::CommaSeparated, 147 cl::desc("List of functions to disassemble. " 148 "Accept demangled names when --demangle is " 149 "specified, otherwise accept mangled names"), 150 cl::cat(ObjdumpCat)); 151 152 static cl::opt<bool> DisassembleZeroes( 153 "disassemble-zeroes", 154 cl::desc("Do not skip blocks of zeroes when disassembling"), 155 cl::cat(ObjdumpCat)); 156 static cl::alias 157 DisassembleZeroesShort("z", cl::desc("Alias for --disassemble-zeroes"), 158 cl::NotHidden, cl::Grouping, 159 cl::aliasopt(DisassembleZeroes)); 160 161 static cl::list<std::string> 162 DisassemblerOptions("disassembler-options", 163 cl::desc("Pass target specific disassembler options"), 164 cl::value_desc("options"), cl::CommaSeparated, 165 cl::cat(ObjdumpCat)); 166 static cl::alias 167 DisassemblerOptionsShort("M", cl::desc("Alias for --disassembler-options"), 168 cl::NotHidden, cl::Grouping, cl::Prefix, 169 cl::CommaSeparated, 170 cl::aliasopt(DisassemblerOptions)); 171 172 cl::opt<DIDumpType> DwarfDumpType( 173 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 174 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame")), 175 cl::cat(ObjdumpCat)); 176 177 static cl::opt<bool> DynamicRelocations( 178 "dynamic-reloc", 179 cl::desc("Display the dynamic relocation entries in the file"), 180 cl::cat(ObjdumpCat)); 181 static cl::alias DynamicRelocationShort("R", 182 cl::desc("Alias for --dynamic-reloc"), 183 cl::NotHidden, cl::Grouping, 184 cl::aliasopt(DynamicRelocations)); 185 186 static cl::opt<bool> 187 FaultMapSection("fault-map-section", 188 cl::desc("Display contents of faultmap section"), 189 cl::cat(ObjdumpCat)); 190 191 static cl::opt<bool> 192 FileHeaders("file-headers", 193 cl::desc("Display the contents of the overall file header"), 194 cl::cat(ObjdumpCat)); 195 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 196 cl::NotHidden, cl::Grouping, 197 cl::aliasopt(FileHeaders)); 198 199 cl::opt<bool> SectionContents("full-contents", 200 cl::desc("Display the content of each section"), 201 cl::cat(ObjdumpCat)); 202 static cl::alias SectionContentsShort("s", 203 cl::desc("Alias for --full-contents"), 204 cl::NotHidden, cl::Grouping, 205 cl::aliasopt(SectionContents)); 206 207 static cl::list<std::string> InputFilenames(cl::Positional, 208 cl::desc("<input object files>"), 209 cl::ZeroOrMore, 210 cl::cat(ObjdumpCat)); 211 212 static cl::opt<bool> 213 PrintLines("line-numbers", 214 cl::desc("Display source line numbers with " 215 "disassembly. Implies disassemble object"), 216 cl::cat(ObjdumpCat)); 217 static cl::alias PrintLinesShort("l", cl::desc("Alias for --line-numbers"), 218 cl::NotHidden, cl::Grouping, 219 cl::aliasopt(PrintLines)); 220 221 static cl::opt<bool> MachOOpt("macho", 222 cl::desc("Use MachO specific object file parser"), 223 cl::cat(ObjdumpCat)); 224 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 225 cl::Grouping, cl::aliasopt(MachOOpt)); 226 227 cl::opt<std::string> 228 MCPU("mcpu", 229 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 230 cl::value_desc("cpu-name"), cl::init(""), cl::cat(ObjdumpCat)); 231 232 cl::list<std::string> MAttrs("mattr", cl::CommaSeparated, 233 cl::desc("Target specific attributes"), 234 cl::value_desc("a1,+a2,-a3,..."), 235 cl::cat(ObjdumpCat)); 236 237 cl::opt<bool> NoShowRawInsn("no-show-raw-insn", 238 cl::desc("When disassembling " 239 "instructions, do not print " 240 "the instruction bytes."), 241 cl::cat(ObjdumpCat)); 242 cl::opt<bool> NoLeadingAddr("no-leading-addr", 243 cl::desc("Print no leading address"), 244 cl::cat(ObjdumpCat)); 245 246 static cl::opt<bool> RawClangAST( 247 "raw-clang-ast", 248 cl::desc("Dump the raw binary contents of the clang AST section"), 249 cl::cat(ObjdumpCat)); 250 251 cl::opt<bool> 252 Relocations("reloc", cl::desc("Display the relocation entries in the file"), 253 cl::cat(ObjdumpCat)); 254 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 255 cl::NotHidden, cl::Grouping, 256 cl::aliasopt(Relocations)); 257 258 cl::opt<bool> PrintImmHex("print-imm-hex", 259 cl::desc("Use hex format for immediate values"), 260 cl::cat(ObjdumpCat)); 261 262 cl::opt<bool> PrivateHeaders("private-headers", 263 cl::desc("Display format specific file headers"), 264 cl::cat(ObjdumpCat)); 265 static cl::alias PrivateHeadersShort("p", 266 cl::desc("Alias for --private-headers"), 267 cl::NotHidden, cl::Grouping, 268 cl::aliasopt(PrivateHeaders)); 269 270 cl::list<std::string> 271 FilterSections("section", 272 cl::desc("Operate on the specified sections only. " 273 "With -macho dump segment,section"), 274 cl::cat(ObjdumpCat)); 275 static cl::alias FilterSectionsj("j", cl::desc("Alias for --section"), 276 cl::NotHidden, cl::Grouping, cl::Prefix, 277 cl::aliasopt(FilterSections)); 278 279 cl::opt<bool> SectionHeaders("section-headers", 280 cl::desc("Display summaries of the " 281 "headers for each section."), 282 cl::cat(ObjdumpCat)); 283 static cl::alias SectionHeadersShort("headers", 284 cl::desc("Alias for --section-headers"), 285 cl::NotHidden, 286 cl::aliasopt(SectionHeaders)); 287 static cl::alias SectionHeadersShorter("h", 288 cl::desc("Alias for --section-headers"), 289 cl::NotHidden, cl::Grouping, 290 cl::aliasopt(SectionHeaders)); 291 292 static cl::opt<bool> 293 ShowLMA("show-lma", 294 cl::desc("Display LMA column when dumping ELF section headers"), 295 cl::cat(ObjdumpCat)); 296 297 static cl::opt<bool> PrintSource( 298 "source", 299 cl::desc( 300 "Display source inlined with disassembly. Implies disassemble object"), 301 cl::cat(ObjdumpCat)); 302 static cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 303 cl::NotHidden, cl::Grouping, 304 cl::aliasopt(PrintSource)); 305 306 static cl::opt<uint64_t> 307 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 308 cl::value_desc("address"), cl::init(0), cl::cat(ObjdumpCat)); 309 static cl::opt<uint64_t> StopAddress("stop-address", 310 cl::desc("Stop disassembly at address"), 311 cl::value_desc("address"), 312 cl::init(UINT64_MAX), cl::cat(ObjdumpCat)); 313 314 cl::opt<bool> SymbolTable("syms", cl::desc("Display the symbol table"), 315 cl::cat(ObjdumpCat)); 316 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 317 cl::NotHidden, cl::Grouping, 318 cl::aliasopt(SymbolTable)); 319 320 cl::opt<std::string> TripleName("triple", 321 cl::desc("Target triple to disassemble for, " 322 "see -version for available targets"), 323 cl::cat(ObjdumpCat)); 324 325 cl::opt<bool> UnwindInfo("unwind-info", cl::desc("Display unwind information"), 326 cl::cat(ObjdumpCat)); 327 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 328 cl::NotHidden, cl::Grouping, 329 cl::aliasopt(UnwindInfo)); 330 331 static cl::opt<bool> 332 Wide("wide", cl::desc("Ignored for compatibility with GNU objdump"), 333 cl::cat(ObjdumpCat)); 334 static cl::alias WideShort("w", cl::Grouping, cl::aliasopt(Wide)); 335 336 static cl::extrahelp 337 HelpResponse("\nPass @FILE as argument to read options from FILE.\n"); 338 339 static StringSet<> DisasmFuncsSet; 340 static StringSet<> FoundSectionSet; 341 static StringRef ToolName; 342 343 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 344 345 static bool shouldKeep(object::SectionRef S) { 346 if (FilterSections.empty()) 347 return true; 348 349 Expected<StringRef> SecNameOrErr = S.getName(); 350 if (!SecNameOrErr) { 351 consumeError(SecNameOrErr.takeError()); 352 return false; 353 } 354 StringRef SecName = *SecNameOrErr; 355 356 // StringSet does not allow empty key so avoid adding sections with 357 // no name (such as the section with index 0) here. 358 if (!SecName.empty()) 359 FoundSectionSet.insert(SecName); 360 return is_contained(FilterSections, SecName); 361 } 362 363 SectionFilter ToolSectionFilter(object::ObjectFile const &O) { 364 return SectionFilter([](object::SectionRef S) { return shouldKeep(S); }, O); 365 } 366 367 void error(std::error_code EC) { 368 if (!EC) 369 return; 370 WithColor::error(errs(), ToolName) 371 << "reading file: " << EC.message() << ".\n"; 372 errs().flush(); 373 exit(1); 374 } 375 376 void error(Error E) { 377 if (!E) 378 return; 379 WithColor::error(errs(), ToolName) << toString(std::move(E)); 380 exit(1); 381 } 382 383 LLVM_ATTRIBUTE_NORETURN void error(Twine Message) { 384 WithColor::error(errs(), ToolName) << Message << ".\n"; 385 errs().flush(); 386 exit(1); 387 } 388 389 void warn(Twine Message) { 390 // Output order between errs() and outs() matters especially for archive 391 // files where the output is per member object. 392 outs().flush(); 393 WithColor::warning(errs(), ToolName) << Message << "\n"; 394 errs().flush(); 395 } 396 397 LLVM_ATTRIBUTE_NORETURN void report_error(StringRef File, Twine Message) { 398 WithColor::error(errs(), ToolName) 399 << "'" << File << "': " << Message << ".\n"; 400 exit(1); 401 } 402 403 LLVM_ATTRIBUTE_NORETURN void report_error(Error E, StringRef File) { 404 assert(E); 405 std::string Buf; 406 raw_string_ostream OS(Buf); 407 logAllUnhandledErrors(std::move(E), OS); 408 OS.flush(); 409 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 410 exit(1); 411 } 412 413 LLVM_ATTRIBUTE_NORETURN void report_error(Error E, StringRef ArchiveName, 414 StringRef FileName, 415 StringRef ArchitectureName) { 416 assert(E); 417 WithColor::error(errs(), ToolName); 418 if (ArchiveName != "") 419 errs() << ArchiveName << "(" << FileName << ")"; 420 else 421 errs() << "'" << FileName << "'"; 422 if (!ArchitectureName.empty()) 423 errs() << " (for architecture " << ArchitectureName << ")"; 424 std::string Buf; 425 raw_string_ostream OS(Buf); 426 logAllUnhandledErrors(std::move(E), OS); 427 OS.flush(); 428 errs() << ": " << Buf; 429 exit(1); 430 } 431 432 LLVM_ATTRIBUTE_NORETURN void report_error(Error E, StringRef ArchiveName, 433 const object::Archive::Child &C, 434 StringRef ArchitectureName) { 435 Expected<StringRef> NameOrErr = C.getName(); 436 // TODO: if we have a error getting the name then it would be nice to print 437 // the index of which archive member this is and or its offset in the 438 // archive instead of "???" as the name. 439 if (!NameOrErr) { 440 consumeError(NameOrErr.takeError()); 441 report_error(std::move(E), ArchiveName, "???", ArchitectureName); 442 } else 443 report_error(std::move(E), ArchiveName, NameOrErr.get(), ArchitectureName); 444 } 445 446 static void warnOnNoMatchForSections() { 447 SetVector<StringRef> MissingSections; 448 for (StringRef S : FilterSections) { 449 if (FoundSectionSet.count(S)) 450 return; 451 // User may specify a unnamed section. Don't warn for it. 452 if (!S.empty()) 453 MissingSections.insert(S); 454 } 455 456 // Warn only if no section in FilterSections is matched. 457 for (StringRef S : MissingSections) 458 warn("section '" + S + "' mentioned in a -j/--section option, but not " 459 "found in any input file"); 460 } 461 462 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 463 // Figure out the target triple. 464 Triple TheTriple("unknown-unknown-unknown"); 465 if (TripleName.empty()) { 466 if (Obj) 467 TheTriple = Obj->makeTriple(); 468 } else { 469 TheTriple.setTriple(Triple::normalize(TripleName)); 470 471 // Use the triple, but also try to combine with ARM build attributes. 472 if (Obj) { 473 auto Arch = Obj->getArch(); 474 if (Arch == Triple::arm || Arch == Triple::armeb) 475 Obj->setARMSubArch(TheTriple); 476 } 477 } 478 479 // Get the target specific parser. 480 std::string Error; 481 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 482 Error); 483 if (!TheTarget) { 484 if (Obj) 485 report_error(Obj->getFileName(), "can't find target: " + Error); 486 else 487 error("can't find target: " + Error); 488 } 489 490 // Update the triple name and return the found target. 491 TripleName = TheTriple.getTriple(); 492 return TheTarget; 493 } 494 495 bool isRelocAddressLess(RelocationRef A, RelocationRef B) { 496 return A.getOffset() < B.getOffset(); 497 } 498 499 static Error getRelocationValueString(const RelocationRef &Rel, 500 SmallVectorImpl<char> &Result) { 501 const ObjectFile *Obj = Rel.getObject(); 502 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 503 return getELFRelocationValueString(ELF, Rel, Result); 504 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 505 return getCOFFRelocationValueString(COFF, Rel, Result); 506 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 507 return getWasmRelocationValueString(Wasm, Rel, Result); 508 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 509 return getMachORelocationValueString(MachO, Rel, Result); 510 llvm_unreachable("unknown object file format"); 511 } 512 513 /// Indicates whether this relocation should hidden when listing 514 /// relocations, usually because it is the trailing part of a multipart 515 /// relocation that will be printed as part of the leading relocation. 516 static bool getHidden(RelocationRef RelRef) { 517 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 518 if (!MachO) 519 return false; 520 521 unsigned Arch = MachO->getArch(); 522 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 523 uint64_t Type = MachO->getRelocationType(Rel); 524 525 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 526 // is always hidden. 527 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 528 return Type == MachO::GENERIC_RELOC_PAIR; 529 530 if (Arch == Triple::x86_64) { 531 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 532 // an X86_64_RELOC_SUBTRACTOR. 533 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 534 DataRefImpl RelPrev = Rel; 535 RelPrev.d.a--; 536 uint64_t PrevType = MachO->getRelocationType(RelPrev); 537 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 538 return true; 539 } 540 } 541 542 return false; 543 } 544 545 namespace { 546 class SourcePrinter { 547 protected: 548 DILineInfo OldLineInfo; 549 const ObjectFile *Obj = nullptr; 550 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 551 // File name to file contents of source. 552 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 553 // Mark the line endings of the cached source. 554 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 555 // Keep track of missing sources. 556 StringSet<> MissingSources; 557 // Only emit 'no debug info' warning once. 558 bool WarnedNoDebugInfo; 559 560 private: 561 bool cacheSource(const DILineInfo& LineInfoFile); 562 563 public: 564 SourcePrinter() = default; 565 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) 566 : Obj(Obj), WarnedNoDebugInfo(false) { 567 symbolize::LLVMSymbolizer::Options SymbolizerOpts; 568 SymbolizerOpts.PrintFunctions = DILineInfoSpecifier::FunctionNameKind::None; 569 SymbolizerOpts.Demangle = false; 570 SymbolizerOpts.DefaultArch = DefaultArch; 571 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 572 } 573 virtual ~SourcePrinter() = default; 574 virtual void printSourceLine(raw_ostream &OS, 575 object::SectionedAddress Address, 576 StringRef ObjectFilename, 577 StringRef Delimiter = "; "); 578 }; 579 580 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 581 std::unique_ptr<MemoryBuffer> Buffer; 582 if (LineInfo.Source) { 583 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 584 } else { 585 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 586 if (!BufferOrError) { 587 if (MissingSources.insert(LineInfo.FileName).second) 588 warn("failed to find source " + LineInfo.FileName); 589 590 return false; 591 } 592 Buffer = std::move(*BufferOrError); 593 } 594 // Chomp the file to get lines 595 const char *BufferStart = Buffer->getBufferStart(), 596 *BufferEnd = Buffer->getBufferEnd(); 597 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 598 const char *Start = BufferStart; 599 for (const char *I = BufferStart; I != BufferEnd; ++I) 600 if (*I == '\n') { 601 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 602 Start = I + 1; 603 } 604 if (Start < BufferEnd) 605 Lines.emplace_back(Start, BufferEnd - Start); 606 SourceCache[LineInfo.FileName] = std::move(Buffer); 607 return true; 608 } 609 610 void SourcePrinter::printSourceLine(raw_ostream &OS, 611 object::SectionedAddress Address, 612 StringRef ObjectFilename, 613 StringRef Delimiter) { 614 if (!Symbolizer) 615 return; 616 617 DILineInfo LineInfo = DILineInfo(); 618 auto ExpectedLineInfo = Symbolizer->symbolizeCode(*Obj, Address); 619 std::string ErrorMessage; 620 if (!ExpectedLineInfo) 621 ErrorMessage = toString(ExpectedLineInfo.takeError()); 622 else 623 LineInfo = *ExpectedLineInfo; 624 625 if (LineInfo.FileName == DILineInfo::BadString) { 626 if (!WarnedNoDebugInfo) { 627 std::string Warning = 628 "failed to parse debug information for " + ObjectFilename.str(); 629 if (!ErrorMessage.empty()) 630 Warning += ": " + ErrorMessage; 631 warn(Warning); 632 WarnedNoDebugInfo = true; 633 } 634 return; 635 } 636 637 if (LineInfo.Line == 0 || ((OldLineInfo.Line == LineInfo.Line) && 638 (OldLineInfo.FileName == LineInfo.FileName))) 639 return; 640 641 if (PrintLines) 642 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 643 if (PrintSource) { 644 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 645 if (!cacheSource(LineInfo)) 646 return; 647 auto LineBuffer = LineCache.find(LineInfo.FileName); 648 if (LineBuffer != LineCache.end()) { 649 if (LineInfo.Line > LineBuffer->second.size()) { 650 warn(formatv( 651 "debug info line number {0} exceeds the number of lines in {1}", 652 LineInfo.Line, LineInfo.FileName)); 653 return; 654 } 655 // Vector begins at 0, line numbers are non-zero 656 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1] << '\n'; 657 } 658 } 659 OldLineInfo = LineInfo; 660 } 661 662 static bool isAArch64Elf(const ObjectFile *Obj) { 663 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 664 return Elf && Elf->getEMachine() == ELF::EM_AARCH64; 665 } 666 667 static bool isArmElf(const ObjectFile *Obj) { 668 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 669 return Elf && Elf->getEMachine() == ELF::EM_ARM; 670 } 671 672 static bool hasMappingSymbols(const ObjectFile *Obj) { 673 return isArmElf(Obj) || isAArch64Elf(Obj); 674 } 675 676 static void printRelocation(const RelocationRef &Rel, uint64_t Address, 677 bool Is64Bits) { 678 StringRef Fmt = Is64Bits ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 679 SmallString<16> Name; 680 SmallString<32> Val; 681 Rel.getTypeName(Name); 682 error(getRelocationValueString(Rel, Val)); 683 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n"; 684 } 685 686 class PrettyPrinter { 687 public: 688 virtual ~PrettyPrinter() = default; 689 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 690 ArrayRef<uint8_t> Bytes, 691 object::SectionedAddress Address, raw_ostream &OS, 692 StringRef Annot, MCSubtargetInfo const &STI, 693 SourcePrinter *SP, StringRef ObjectFilename, 694 std::vector<RelocationRef> *Rels = nullptr) { 695 if (SP && (PrintSource || PrintLines)) 696 SP->printSourceLine(OS, Address, ObjectFilename); 697 698 size_t Start = OS.tell(); 699 if (!NoLeadingAddr) 700 OS << format("%8" PRIx64 ":", Address.Address); 701 if (!NoShowRawInsn) { 702 OS << ' '; 703 dumpBytes(Bytes, OS); 704 } 705 706 // The output of printInst starts with a tab. Print some spaces so that 707 // the tab has 1 column and advances to the target tab stop. 708 unsigned TabStop = NoShowRawInsn ? 16 : 40; 709 unsigned Column = OS.tell() - Start; 710 OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8); 711 712 if (MI) 713 IP.printInst(MI, OS, "", STI); 714 else 715 OS << "\t<unknown>"; 716 } 717 }; 718 PrettyPrinter PrettyPrinterInst; 719 720 class HexagonPrettyPrinter : public PrettyPrinter { 721 public: 722 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 723 raw_ostream &OS) { 724 uint32_t opcode = 725 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 726 if (!NoLeadingAddr) 727 OS << format("%8" PRIx64 ":", Address); 728 if (!NoShowRawInsn) { 729 OS << "\t"; 730 dumpBytes(Bytes.slice(0, 4), OS); 731 OS << format("\t%08" PRIx32, opcode); 732 } 733 } 734 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 735 object::SectionedAddress Address, raw_ostream &OS, 736 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 737 StringRef ObjectFilename, 738 std::vector<RelocationRef> *Rels) override { 739 if (SP && (PrintSource || PrintLines)) 740 SP->printSourceLine(OS, Address, ObjectFilename, ""); 741 if (!MI) { 742 printLead(Bytes, Address.Address, OS); 743 OS << " <unknown>"; 744 return; 745 } 746 std::string Buffer; 747 { 748 raw_string_ostream TempStream(Buffer); 749 IP.printInst(MI, TempStream, "", STI); 750 } 751 StringRef Contents(Buffer); 752 // Split off bundle attributes 753 auto PacketBundle = Contents.rsplit('\n'); 754 // Split off first instruction from the rest 755 auto HeadTail = PacketBundle.first.split('\n'); 756 auto Preamble = " { "; 757 auto Separator = ""; 758 759 // Hexagon's packets require relocations to be inline rather than 760 // clustered at the end of the packet. 761 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 762 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 763 auto PrintReloc = [&]() -> void { 764 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 765 if (RelCur->getOffset() == Address.Address) { 766 printRelocation(*RelCur, Address.Address, false); 767 return; 768 } 769 ++RelCur; 770 } 771 }; 772 773 while (!HeadTail.first.empty()) { 774 OS << Separator; 775 Separator = "\n"; 776 if (SP && (PrintSource || PrintLines)) 777 SP->printSourceLine(OS, Address, ObjectFilename, ""); 778 printLead(Bytes, Address.Address, OS); 779 OS << Preamble; 780 Preamble = " "; 781 StringRef Inst; 782 auto Duplex = HeadTail.first.split('\v'); 783 if (!Duplex.second.empty()) { 784 OS << Duplex.first; 785 OS << "; "; 786 Inst = Duplex.second; 787 } 788 else 789 Inst = HeadTail.first; 790 OS << Inst; 791 HeadTail = HeadTail.second.split('\n'); 792 if (HeadTail.first.empty()) 793 OS << " } " << PacketBundle.second; 794 PrintReloc(); 795 Bytes = Bytes.slice(4); 796 Address.Address += 4; 797 } 798 } 799 }; 800 HexagonPrettyPrinter HexagonPrettyPrinterInst; 801 802 class AMDGCNPrettyPrinter : public PrettyPrinter { 803 public: 804 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 805 object::SectionedAddress Address, raw_ostream &OS, 806 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 807 StringRef ObjectFilename, 808 std::vector<RelocationRef> *Rels) override { 809 if (SP && (PrintSource || PrintLines)) 810 SP->printSourceLine(OS, Address, ObjectFilename); 811 812 if (MI) { 813 SmallString<40> InstStr; 814 raw_svector_ostream IS(InstStr); 815 816 IP.printInst(MI, IS, "", STI); 817 818 OS << left_justify(IS.str(), 60); 819 } else { 820 // an unrecognized encoding - this is probably data so represent it 821 // using the .long directive, or .byte directive if fewer than 4 bytes 822 // remaining 823 if (Bytes.size() >= 4) { 824 OS << format("\t.long 0x%08" PRIx32 " ", 825 support::endian::read32<support::little>(Bytes.data())); 826 OS.indent(42); 827 } else { 828 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 829 for (unsigned int i = 1; i < Bytes.size(); i++) 830 OS << format(", 0x%02" PRIx8, Bytes[i]); 831 OS.indent(55 - (6 * Bytes.size())); 832 } 833 } 834 835 OS << format("// %012" PRIX64 ":", Address.Address); 836 if (Bytes.size() >= 4) { 837 // D should be casted to uint32_t here as it is passed by format to 838 // snprintf as vararg. 839 for (uint32_t D : makeArrayRef( 840 reinterpret_cast<const support::little32_t *>(Bytes.data()), 841 Bytes.size() / 4)) 842 OS << format(" %08" PRIX32, D); 843 } else { 844 for (unsigned char B : Bytes) 845 OS << format(" %02" PRIX8, B); 846 } 847 848 if (!Annot.empty()) 849 OS << " // " << Annot; 850 } 851 }; 852 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 853 854 class BPFPrettyPrinter : public PrettyPrinter { 855 public: 856 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 857 object::SectionedAddress Address, raw_ostream &OS, 858 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 859 StringRef ObjectFilename, 860 std::vector<RelocationRef> *Rels) override { 861 if (SP && (PrintSource || PrintLines)) 862 SP->printSourceLine(OS, Address, ObjectFilename); 863 if (!NoLeadingAddr) 864 OS << format("%8" PRId64 ":", Address.Address / 8); 865 if (!NoShowRawInsn) { 866 OS << "\t"; 867 dumpBytes(Bytes, OS); 868 } 869 if (MI) 870 IP.printInst(MI, OS, "", STI); 871 else 872 OS << "\t<unknown>"; 873 } 874 }; 875 BPFPrettyPrinter BPFPrettyPrinterInst; 876 877 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 878 switch(Triple.getArch()) { 879 default: 880 return PrettyPrinterInst; 881 case Triple::hexagon: 882 return HexagonPrettyPrinterInst; 883 case Triple::amdgcn: 884 return AMDGCNPrettyPrinterInst; 885 case Triple::bpfel: 886 case Triple::bpfeb: 887 return BPFPrettyPrinterInst; 888 } 889 } 890 } 891 892 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 893 assert(Obj->isELF()); 894 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 895 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 896 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 897 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 898 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 899 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 900 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 901 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 902 llvm_unreachable("Unsupported binary format"); 903 } 904 905 template <class ELFT> static void 906 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 907 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 908 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 909 uint8_t SymbolType = Symbol.getELFType(); 910 if (SymbolType == ELF::STT_SECTION) 911 continue; 912 913 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 914 // ELFSymbolRef::getAddress() returns size instead of value for common 915 // symbols which is not desirable for disassembly output. Overriding. 916 if (SymbolType == ELF::STT_COMMON) 917 Address = Obj->getSymbol(Symbol.getRawDataRefImpl())->st_value; 918 919 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 920 if (Name.empty()) 921 continue; 922 923 section_iterator SecI = 924 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 925 if (SecI == Obj->section_end()) 926 continue; 927 928 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 929 } 930 } 931 932 static void 933 addDynamicElfSymbols(const ObjectFile *Obj, 934 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 935 assert(Obj->isELF()); 936 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 937 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 938 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 939 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 940 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 941 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 942 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 943 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 944 else 945 llvm_unreachable("Unsupported binary format"); 946 } 947 948 static void addPltEntries(const ObjectFile *Obj, 949 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 950 StringSaver &Saver) { 951 Optional<SectionRef> Plt = None; 952 for (const SectionRef &Section : Obj->sections()) { 953 Expected<StringRef> SecNameOrErr = Section.getName(); 954 if (!SecNameOrErr) { 955 consumeError(SecNameOrErr.takeError()); 956 continue; 957 } 958 if (*SecNameOrErr == ".plt") 959 Plt = Section; 960 } 961 if (!Plt) 962 return; 963 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 964 for (auto PltEntry : ElfObj->getPltAddresses()) { 965 SymbolRef Symbol(PltEntry.first, ElfObj); 966 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 967 968 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 969 if (!Name.empty()) 970 AllSymbols[*Plt].emplace_back( 971 PltEntry.second, Saver.save((Name + "@plt").str()), SymbolType); 972 } 973 } 974 } 975 976 // Normally the disassembly output will skip blocks of zeroes. This function 977 // returns the number of zero bytes that can be skipped when dumping the 978 // disassembly of the instructions in Buf. 979 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 980 // Find the number of leading zeroes. 981 size_t N = 0; 982 while (N < Buf.size() && !Buf[N]) 983 ++N; 984 985 // We may want to skip blocks of zero bytes, but unless we see 986 // at least 8 of them in a row. 987 if (N < 8) 988 return 0; 989 990 // We skip zeroes in multiples of 4 because do not want to truncate an 991 // instruction if it starts with a zero byte. 992 return N & ~0x3; 993 } 994 995 // Returns a map from sections to their relocations. 996 static std::map<SectionRef, std::vector<RelocationRef>> 997 getRelocsMap(object::ObjectFile const &Obj) { 998 std::map<SectionRef, std::vector<RelocationRef>> Ret; 999 for (SectionRef Sec : Obj.sections()) { 1000 section_iterator Relocated = Sec.getRelocatedSection(); 1001 if (Relocated == Obj.section_end() || !shouldKeep(*Relocated)) 1002 continue; 1003 std::vector<RelocationRef> &V = Ret[*Relocated]; 1004 for (const RelocationRef &R : Sec.relocations()) 1005 V.push_back(R); 1006 // Sort relocations by address. 1007 llvm::stable_sort(V, isRelocAddressLess); 1008 } 1009 return Ret; 1010 } 1011 1012 // Used for --adjust-vma to check if address should be adjusted by the 1013 // specified value for a given section. 1014 // For ELF we do not adjust non-allocatable sections like debug ones, 1015 // because they are not loadable. 1016 // TODO: implement for other file formats. 1017 static bool shouldAdjustVA(const SectionRef &Section) { 1018 const ObjectFile *Obj = Section.getObject(); 1019 if (isa<object::ELFObjectFileBase>(Obj)) 1020 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 1021 return false; 1022 } 1023 1024 1025 typedef std::pair<uint64_t, char> MappingSymbolPair; 1026 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols, 1027 uint64_t Address) { 1028 auto It = 1029 partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) { 1030 return Val.first <= Address; 1031 }); 1032 // Return zero for any address before the first mapping symbol; this means 1033 // we should use the default disassembly mode, depending on the target. 1034 if (It == MappingSymbols.begin()) 1035 return '\x00'; 1036 return (It - 1)->second; 1037 } 1038 1039 static uint64_t 1040 dumpARMELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 1041 const ObjectFile *Obj, ArrayRef<uint8_t> Bytes, 1042 ArrayRef<MappingSymbolPair> MappingSymbols) { 1043 support::endianness Endian = 1044 Obj->isLittleEndian() ? support::little : support::big; 1045 while (Index < End) { 1046 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1047 outs() << "\t"; 1048 if (Index + 4 <= End) { 1049 dumpBytes(Bytes.slice(Index, 4), outs()); 1050 outs() << "\t.word\t" 1051 << format_hex( 1052 support::endian::read32(Bytes.data() + Index, Endian), 10); 1053 Index += 4; 1054 } else if (Index + 2 <= End) { 1055 dumpBytes(Bytes.slice(Index, 2), outs()); 1056 outs() << "\t\t.short\t" 1057 << format_hex( 1058 support::endian::read16(Bytes.data() + Index, Endian), 6); 1059 Index += 2; 1060 } else { 1061 dumpBytes(Bytes.slice(Index, 1), outs()); 1062 outs() << "\t\t.byte\t" << format_hex(Bytes[0], 4); 1063 ++Index; 1064 } 1065 outs() << "\n"; 1066 if (getMappingSymbolKind(MappingSymbols, Index) != 'd') 1067 break; 1068 } 1069 return Index; 1070 } 1071 1072 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 1073 ArrayRef<uint8_t> Bytes) { 1074 // print out data up to 8 bytes at a time in hex and ascii 1075 uint8_t AsciiData[9] = {'\0'}; 1076 uint8_t Byte; 1077 int NumBytes = 0; 1078 1079 for (; Index < End; ++Index) { 1080 if (NumBytes == 0) 1081 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1082 Byte = Bytes.slice(Index)[0]; 1083 outs() << format(" %02x", Byte); 1084 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1085 1086 uint8_t IndentOffset = 0; 1087 NumBytes++; 1088 if (Index == End - 1 || NumBytes > 8) { 1089 // Indent the space for less than 8 bytes data. 1090 // 2 spaces for byte and one for space between bytes 1091 IndentOffset = 3 * (8 - NumBytes); 1092 for (int Excess = NumBytes; Excess < 8; Excess++) 1093 AsciiData[Excess] = '\0'; 1094 NumBytes = 8; 1095 } 1096 if (NumBytes == 8) { 1097 AsciiData[8] = '\0'; 1098 outs() << std::string(IndentOffset, ' ') << " "; 1099 outs() << reinterpret_cast<char *>(AsciiData); 1100 outs() << '\n'; 1101 NumBytes = 0; 1102 } 1103 } 1104 } 1105 1106 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 1107 MCContext &Ctx, MCDisassembler *PrimaryDisAsm, 1108 MCDisassembler *SecondaryDisAsm, 1109 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 1110 const MCSubtargetInfo *PrimarySTI, 1111 const MCSubtargetInfo *SecondarySTI, 1112 PrettyPrinter &PIP, 1113 SourcePrinter &SP, bool InlineRelocs) { 1114 const MCSubtargetInfo *STI = PrimarySTI; 1115 MCDisassembler *DisAsm = PrimaryDisAsm; 1116 bool PrimaryIsThumb = false; 1117 if (isArmElf(Obj)) 1118 PrimaryIsThumb = STI->checkFeatures("+thumb-mode"); 1119 1120 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 1121 if (InlineRelocs) 1122 RelocMap = getRelocsMap(*Obj); 1123 bool Is64Bits = Obj->getBytesInAddress() > 4; 1124 1125 // Create a mapping from virtual address to symbol name. This is used to 1126 // pretty print the symbols while disassembling. 1127 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1128 SectionSymbolsTy AbsoluteSymbols; 1129 const StringRef FileName = Obj->getFileName(); 1130 for (const SymbolRef &Symbol : Obj->symbols()) { 1131 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName); 1132 1133 StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1134 if (Name.empty()) 1135 continue; 1136 1137 uint8_t SymbolType = ELF::STT_NOTYPE; 1138 if (Obj->isELF()) { 1139 SymbolType = getElfSymbolType(Obj, Symbol); 1140 if (SymbolType == ELF::STT_SECTION) 1141 continue; 1142 } 1143 1144 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1145 if (SecI != Obj->section_end()) 1146 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 1147 else 1148 AbsoluteSymbols.emplace_back(Address, Name, SymbolType); 1149 } 1150 if (AllSymbols.empty() && Obj->isELF()) 1151 addDynamicElfSymbols(Obj, AllSymbols); 1152 1153 BumpPtrAllocator A; 1154 StringSaver Saver(A); 1155 addPltEntries(Obj, AllSymbols, Saver); 1156 1157 // Create a mapping from virtual address to section. 1158 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1159 for (SectionRef Sec : Obj->sections()) 1160 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1161 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1162 1163 // Linked executables (.exe and .dll files) typically don't include a real 1164 // symbol table but they might contain an export table. 1165 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1166 for (const auto &ExportEntry : COFFObj->export_directories()) { 1167 StringRef Name; 1168 error(ExportEntry.getSymbolName(Name)); 1169 if (Name.empty()) 1170 continue; 1171 uint32_t RVA; 1172 error(ExportEntry.getExportRVA(RVA)); 1173 1174 uint64_t VA = COFFObj->getImageBase() + RVA; 1175 auto Sec = partition_point( 1176 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) { 1177 return O.first <= VA; 1178 }); 1179 if (Sec != SectionAddresses.begin()) { 1180 --Sec; 1181 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1182 } else 1183 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1184 } 1185 } 1186 1187 // Sort all the symbols, this allows us to use a simple binary search to find 1188 // a symbol near an address. 1189 StringSet<> FoundDisasmFuncsSet; 1190 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1191 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1192 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1193 1194 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1195 if (FilterSections.empty() && !DisassembleAll && 1196 (!Section.isText() || Section.isVirtual())) 1197 continue; 1198 1199 uint64_t SectionAddr = Section.getAddress(); 1200 uint64_t SectSize = Section.getSize(); 1201 if (!SectSize) 1202 continue; 1203 1204 // Get the list of all the symbols in this section. 1205 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1206 std::vector<MappingSymbolPair> MappingSymbols; 1207 if (hasMappingSymbols(Obj)) { 1208 for (const auto &Symb : Symbols) { 1209 uint64_t Address = std::get<0>(Symb); 1210 StringRef Name = std::get<1>(Symb); 1211 if (Name.startswith("$d")) 1212 MappingSymbols.emplace_back(Address - SectionAddr, 'd'); 1213 if (Name.startswith("$x")) 1214 MappingSymbols.emplace_back(Address - SectionAddr, 'x'); 1215 if (Name.startswith("$a")) 1216 MappingSymbols.emplace_back(Address - SectionAddr, 'a'); 1217 if (Name.startswith("$t")) 1218 MappingSymbols.emplace_back(Address - SectionAddr, 't'); 1219 } 1220 } 1221 1222 llvm::sort(MappingSymbols); 1223 1224 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1225 // AMDGPU disassembler uses symbolizer for printing labels 1226 std::unique_ptr<MCRelocationInfo> RelInfo( 1227 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1228 if (RelInfo) { 1229 std::unique_ptr<MCSymbolizer> Symbolizer( 1230 TheTarget->createMCSymbolizer( 1231 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1232 DisAsm->setSymbolizer(std::move(Symbolizer)); 1233 } 1234 } 1235 1236 StringRef SegmentName = ""; 1237 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1238 DataRefImpl DR = Section.getRawDataRefImpl(); 1239 SegmentName = MachO->getSectionFinalSegmentName(DR); 1240 } 1241 1242 StringRef SectionName = unwrapOrError(Section.getName(), Obj->getFileName()); 1243 // If the section has no symbol at the start, just insert a dummy one. 1244 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1245 Symbols.insert( 1246 Symbols.begin(), 1247 std::make_tuple(SectionAddr, SectionName, 1248 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1249 } 1250 1251 SmallString<40> Comments; 1252 raw_svector_ostream CommentStream(Comments); 1253 1254 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef( 1255 unwrapOrError(Section.getContents(), Obj->getFileName())); 1256 1257 uint64_t VMAAdjustment = 0; 1258 if (shouldAdjustVA(Section)) 1259 VMAAdjustment = AdjustVMA; 1260 1261 uint64_t Size; 1262 uint64_t Index; 1263 bool PrintedSection = false; 1264 std::vector<RelocationRef> Rels = RelocMap[Section]; 1265 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1266 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1267 // Disassemble symbol by symbol. 1268 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1269 std::string SymbolName = std::get<1>(Symbols[SI]).str(); 1270 if (Demangle) 1271 SymbolName = demangle(SymbolName); 1272 1273 // Skip if --disassemble-functions is not empty and the symbol is not in 1274 // the list. 1275 if (!DisasmFuncsSet.empty() && !DisasmFuncsSet.count(SymbolName)) 1276 continue; 1277 1278 uint64_t Start = std::get<0>(Symbols[SI]); 1279 if (Start < SectionAddr || StopAddress <= Start) 1280 continue; 1281 else 1282 FoundDisasmFuncsSet.insert(SymbolName); 1283 1284 // The end is the section end, the beginning of the next symbol, or 1285 // --stop-address. 1286 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress); 1287 if (SI + 1 < SE) 1288 End = std::min(End, std::get<0>(Symbols[SI + 1])); 1289 if (Start >= End || End <= StartAddress) 1290 continue; 1291 Start -= SectionAddr; 1292 End -= SectionAddr; 1293 1294 if (!PrintedSection) { 1295 PrintedSection = true; 1296 outs() << "\nDisassembly of section "; 1297 if (!SegmentName.empty()) 1298 outs() << SegmentName << ","; 1299 outs() << SectionName << ":\n"; 1300 } 1301 1302 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1303 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1304 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1305 Start += 256; 1306 } 1307 if (SI == SE - 1 || 1308 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1309 // cut trailing zeroes at the end of kernel 1310 // cut up to 256 bytes 1311 const uint64_t EndAlign = 256; 1312 const auto Limit = End - (std::min)(EndAlign, End - Start); 1313 while (End > Limit && 1314 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1315 End -= 4; 1316 } 1317 } 1318 1319 outs() << '\n'; 1320 if (!NoLeadingAddr) 1321 outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ", 1322 SectionAddr + Start + VMAAdjustment); 1323 1324 outs() << SymbolName << ":\n"; 1325 1326 // Don't print raw contents of a virtual section. A virtual section 1327 // doesn't have any contents in the file. 1328 if (Section.isVirtual()) { 1329 outs() << "...\n"; 1330 continue; 1331 } 1332 1333 #ifndef NDEBUG 1334 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1335 #else 1336 raw_ostream &DebugOut = nulls(); 1337 #endif 1338 1339 // Some targets (like WebAssembly) have a special prelude at the start 1340 // of each symbol. 1341 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1342 SectionAddr + Start, DebugOut, CommentStream); 1343 Start += Size; 1344 1345 Index = Start; 1346 if (SectionAddr < StartAddress) 1347 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1348 1349 // If there is a data/common symbol inside an ELF text section and we are 1350 // only disassembling text (applicable all architectures), we are in a 1351 // situation where we must print the data and not disassemble it. 1352 if (Obj->isELF() && !DisassembleAll && Section.isText()) { 1353 uint8_t SymTy = std::get<2>(Symbols[SI]); 1354 if (SymTy == ELF::STT_OBJECT || SymTy == ELF::STT_COMMON) { 1355 dumpELFData(SectionAddr, Index, End, Bytes); 1356 Index = End; 1357 } 1358 } 1359 1360 bool CheckARMELFData = hasMappingSymbols(Obj) && 1361 std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1362 !DisassembleAll; 1363 while (Index < End) { 1364 // ARM and AArch64 ELF binaries can interleave data and text in the 1365 // same section. We rely on the markers introduced to understand what 1366 // we need to dump. If the data marker is within a function, it is 1367 // denoted as a word/short etc. 1368 if (CheckARMELFData && 1369 getMappingSymbolKind(MappingSymbols, Index) == 'd') { 1370 Index = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1371 MappingSymbols); 1372 continue; 1373 } 1374 1375 // When -z or --disassemble-zeroes are given we always dissasemble 1376 // them. Otherwise we might want to skip zero bytes we see. 1377 if (!DisassembleZeroes) { 1378 uint64_t MaxOffset = End - Index; 1379 // For -reloc: print zero blocks patched by relocations, so that 1380 // relocations can be shown in the dump. 1381 if (RelCur != RelEnd) 1382 MaxOffset = RelCur->getOffset() - Index; 1383 1384 if (size_t N = 1385 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1386 outs() << "\t\t..." << '\n'; 1387 Index += N; 1388 continue; 1389 } 1390 } 1391 1392 if (SecondarySTI) { 1393 if (getMappingSymbolKind(MappingSymbols, Index) == 'a') { 1394 STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI; 1395 DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm; 1396 } else if (getMappingSymbolKind(MappingSymbols, Index) == 't') { 1397 STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI; 1398 DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm; 1399 } 1400 } 1401 1402 // Disassemble a real instruction or a data when disassemble all is 1403 // provided 1404 MCInst Inst; 1405 bool Disassembled = DisAsm->getInstruction( 1406 Inst, Size, Bytes.slice(Index), SectionAddr + Index, DebugOut, 1407 CommentStream); 1408 if (Size == 0) 1409 Size = 1; 1410 1411 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1412 Bytes.slice(Index, Size), 1413 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, 1414 outs(), "", *STI, &SP, Obj->getFileName(), &Rels); 1415 outs() << CommentStream.str(); 1416 Comments.clear(); 1417 1418 // Try to resolve the target of a call, tail call, etc. to a specific 1419 // symbol. 1420 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1421 MIA->isConditionalBranch(Inst))) { 1422 uint64_t Target; 1423 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1424 // In a relocatable object, the target's section must reside in 1425 // the same section as the call instruction or it is accessed 1426 // through a relocation. 1427 // 1428 // In a non-relocatable object, the target may be in any section. 1429 // 1430 // N.B. We don't walk the relocations in the relocatable case yet. 1431 auto *TargetSectionSymbols = &Symbols; 1432 if (!Obj->isRelocatableObject()) { 1433 auto It = partition_point( 1434 SectionAddresses, 1435 [=](const std::pair<uint64_t, SectionRef> &O) { 1436 return O.first <= Target; 1437 }); 1438 if (It != SectionAddresses.begin()) { 1439 --It; 1440 TargetSectionSymbols = &AllSymbols[It->second]; 1441 } else { 1442 TargetSectionSymbols = &AbsoluteSymbols; 1443 } 1444 } 1445 1446 // Find the last symbol in the section whose offset is less than 1447 // or equal to the target. If there isn't a section that contains 1448 // the target, find the nearest preceding absolute symbol. 1449 auto TargetSym = partition_point( 1450 *TargetSectionSymbols, 1451 [=](const std::tuple<uint64_t, StringRef, uint8_t> &O) { 1452 return std::get<0>(O) <= Target; 1453 }); 1454 if (TargetSym == TargetSectionSymbols->begin()) { 1455 TargetSectionSymbols = &AbsoluteSymbols; 1456 TargetSym = partition_point( 1457 AbsoluteSymbols, 1458 [=](const std::tuple<uint64_t, StringRef, uint8_t> &O) { 1459 return std::get<0>(O) <= Target; 1460 }); 1461 } 1462 if (TargetSym != TargetSectionSymbols->begin()) { 1463 --TargetSym; 1464 uint64_t TargetAddress = std::get<0>(*TargetSym); 1465 StringRef TargetName = std::get<1>(*TargetSym); 1466 outs() << " <" << TargetName; 1467 uint64_t Disp = Target - TargetAddress; 1468 if (Disp) 1469 outs() << "+0x" << Twine::utohexstr(Disp); 1470 outs() << '>'; 1471 } 1472 } 1473 } 1474 outs() << "\n"; 1475 1476 // Hexagon does this in pretty printer 1477 if (Obj->getArch() != Triple::hexagon) { 1478 // Print relocation for instruction. 1479 while (RelCur != RelEnd) { 1480 uint64_t Offset = RelCur->getOffset(); 1481 // If this relocation is hidden, skip it. 1482 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 1483 ++RelCur; 1484 continue; 1485 } 1486 1487 // Stop when RelCur's offset is past the current instruction. 1488 if (Offset >= Index + Size) 1489 break; 1490 1491 // When --adjust-vma is used, update the address printed. 1492 if (RelCur->getSymbol() != Obj->symbol_end()) { 1493 Expected<section_iterator> SymSI = 1494 RelCur->getSymbol()->getSection(); 1495 if (SymSI && *SymSI != Obj->section_end() && 1496 shouldAdjustVA(**SymSI)) 1497 Offset += AdjustVMA; 1498 } 1499 1500 printRelocation(*RelCur, SectionAddr + Offset, Is64Bits); 1501 ++RelCur; 1502 } 1503 } 1504 1505 Index += Size; 1506 } 1507 } 1508 } 1509 StringSet<> MissingDisasmFuncsSet = 1510 set_difference(DisasmFuncsSet, FoundDisasmFuncsSet); 1511 for (StringRef MissingDisasmFunc : MissingDisasmFuncsSet.keys()) 1512 warn("failed to disassemble missing function " + MissingDisasmFunc); 1513 } 1514 1515 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1516 const Target *TheTarget = getTarget(Obj); 1517 1518 // Package up features to be passed to target/subtarget 1519 SubtargetFeatures Features = Obj->getFeatures(); 1520 if (!MAttrs.empty()) 1521 for (unsigned I = 0; I != MAttrs.size(); ++I) 1522 Features.AddFeature(MAttrs[I]); 1523 1524 std::unique_ptr<const MCRegisterInfo> MRI( 1525 TheTarget->createMCRegInfo(TripleName)); 1526 if (!MRI) 1527 report_error(Obj->getFileName(), 1528 "no register info for target " + TripleName); 1529 1530 // Set up disassembler. 1531 std::unique_ptr<const MCAsmInfo> AsmInfo( 1532 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1533 if (!AsmInfo) 1534 report_error(Obj->getFileName(), 1535 "no assembly info for target " + TripleName); 1536 std::unique_ptr<const MCSubtargetInfo> STI( 1537 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1538 if (!STI) 1539 report_error(Obj->getFileName(), 1540 "no subtarget info for target " + TripleName); 1541 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1542 if (!MII) 1543 report_error(Obj->getFileName(), 1544 "no instruction info for target " + TripleName); 1545 MCObjectFileInfo MOFI; 1546 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1547 // FIXME: for now initialize MCObjectFileInfo with default values 1548 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1549 1550 std::unique_ptr<MCDisassembler> DisAsm( 1551 TheTarget->createMCDisassembler(*STI, Ctx)); 1552 if (!DisAsm) 1553 report_error(Obj->getFileName(), 1554 "no disassembler for target " + TripleName); 1555 1556 // If we have an ARM object file, we need a second disassembler, because 1557 // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode. 1558 // We use mapping symbols to switch between the two assemblers, where 1559 // appropriate. 1560 std::unique_ptr<MCDisassembler> SecondaryDisAsm; 1561 std::unique_ptr<const MCSubtargetInfo> SecondarySTI; 1562 if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) { 1563 if (STI->checkFeatures("+thumb-mode")) 1564 Features.AddFeature("-thumb-mode"); 1565 else 1566 Features.AddFeature("+thumb-mode"); 1567 SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU, 1568 Features.getString())); 1569 SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx)); 1570 } 1571 1572 std::unique_ptr<const MCInstrAnalysis> MIA( 1573 TheTarget->createMCInstrAnalysis(MII.get())); 1574 1575 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1576 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1577 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1578 if (!IP) 1579 report_error(Obj->getFileName(), 1580 "no instruction printer for target " + TripleName); 1581 IP->setPrintImmHex(PrintImmHex); 1582 1583 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1584 SourcePrinter SP(Obj, TheTarget->getName()); 1585 1586 for (StringRef Opt : DisassemblerOptions) 1587 if (!IP->applyTargetSpecificCLOption(Opt)) 1588 error("Unrecognized disassembler option: " + Opt); 1589 1590 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(), 1591 MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP, 1592 SP, InlineRelocs); 1593 } 1594 1595 void printRelocations(const ObjectFile *Obj) { 1596 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1597 "%08" PRIx64; 1598 // Regular objdump doesn't print relocations in non-relocatable object 1599 // files. 1600 if (!Obj->isRelocatableObject()) 1601 return; 1602 1603 // Build a mapping from relocation target to a vector of relocation 1604 // sections. Usually, there is an only one relocation section for 1605 // each relocated section. 1606 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec; 1607 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1608 if (Section.relocation_begin() == Section.relocation_end()) 1609 continue; 1610 const SectionRef TargetSec = *Section.getRelocatedSection(); 1611 SecToRelSec[TargetSec].push_back(Section); 1612 } 1613 1614 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) { 1615 StringRef SecName = unwrapOrError(P.first.getName(), Obj->getFileName()); 1616 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1617 1618 for (SectionRef Section : P.second) { 1619 for (const RelocationRef &Reloc : Section.relocations()) { 1620 uint64_t Address = Reloc.getOffset(); 1621 SmallString<32> RelocName; 1622 SmallString<32> ValueStr; 1623 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1624 continue; 1625 Reloc.getTypeName(RelocName); 1626 error(getRelocationValueString(Reloc, ValueStr)); 1627 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1628 << ValueStr << "\n"; 1629 } 1630 } 1631 outs() << "\n"; 1632 } 1633 } 1634 1635 void printDynamicRelocations(const ObjectFile *Obj) { 1636 // For the moment, this option is for ELF only 1637 if (!Obj->isELF()) 1638 return; 1639 1640 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1641 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1642 error("not a dynamic object"); 1643 return; 1644 } 1645 1646 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1647 if (DynRelSec.empty()) 1648 return; 1649 1650 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1651 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1652 for (const SectionRef &Section : DynRelSec) 1653 for (const RelocationRef &Reloc : Section.relocations()) { 1654 uint64_t Address = Reloc.getOffset(); 1655 SmallString<32> RelocName; 1656 SmallString<32> ValueStr; 1657 Reloc.getTypeName(RelocName); 1658 error(getRelocationValueString(Reloc, ValueStr)); 1659 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1660 << ValueStr << "\n"; 1661 } 1662 } 1663 1664 // Returns true if we need to show LMA column when dumping section headers. We 1665 // show it only when the platform is ELF and either we have at least one section 1666 // whose VMA and LMA are different and/or when --show-lma flag is used. 1667 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1668 if (!Obj->isELF()) 1669 return false; 1670 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1671 if (S.getAddress() != getELFSectionLMA(S)) 1672 return true; 1673 return ShowLMA; 1674 } 1675 1676 void printSectionHeaders(const ObjectFile *Obj) { 1677 bool HasLMAColumn = shouldDisplayLMA(Obj); 1678 if (HasLMAColumn) 1679 outs() << "Sections:\n" 1680 "Idx Name Size VMA LMA " 1681 "Type\n"; 1682 else 1683 outs() << "Sections:\n" 1684 "Idx Name Size VMA Type\n"; 1685 1686 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1687 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 1688 uint64_t VMA = Section.getAddress(); 1689 if (shouldAdjustVA(Section)) 1690 VMA += AdjustVMA; 1691 1692 uint64_t Size = Section.getSize(); 1693 bool Text = Section.isText(); 1694 bool Data = Section.isData(); 1695 bool BSS = Section.isBSS(); 1696 std::string Type = (std::string(Text ? "TEXT " : "") + 1697 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1698 1699 if (HasLMAColumn) 1700 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %016" PRIx64 1701 " %s\n", 1702 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1703 VMA, getELFSectionLMA(Section), Type.c_str()); 1704 else 1705 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1706 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1707 VMA, Type.c_str()); 1708 } 1709 outs() << "\n"; 1710 } 1711 1712 void printSectionContents(const ObjectFile *Obj) { 1713 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1714 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 1715 uint64_t BaseAddr = Section.getAddress(); 1716 uint64_t Size = Section.getSize(); 1717 if (!Size) 1718 continue; 1719 1720 outs() << "Contents of section " << Name << ":\n"; 1721 if (Section.isBSS()) { 1722 outs() << format("<skipping contents of bss section at [%04" PRIx64 1723 ", %04" PRIx64 ")>\n", 1724 BaseAddr, BaseAddr + Size); 1725 continue; 1726 } 1727 1728 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName()); 1729 1730 // Dump out the content as hex and printable ascii characters. 1731 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1732 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1733 // Dump line of hex. 1734 for (std::size_t I = 0; I < 16; ++I) { 1735 if (I != 0 && I % 4 == 0) 1736 outs() << ' '; 1737 if (Addr + I < End) 1738 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1739 << hexdigit(Contents[Addr + I] & 0xF, true); 1740 else 1741 outs() << " "; 1742 } 1743 // Print ascii. 1744 outs() << " "; 1745 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1746 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1747 outs() << Contents[Addr + I]; 1748 else 1749 outs() << "."; 1750 } 1751 outs() << "\n"; 1752 } 1753 } 1754 } 1755 1756 void printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1757 StringRef ArchitectureName) { 1758 outs() << "SYMBOL TABLE:\n"; 1759 1760 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1761 printCOFFSymbolTable(Coff); 1762 return; 1763 } 1764 1765 const StringRef FileName = O->getFileName(); 1766 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1767 const SymbolRef &Symbol = *I; 1768 uint64_t Address = unwrapOrError(Symbol.getAddress(), ArchiveName, FileName, 1769 ArchitectureName); 1770 if ((Address < StartAddress) || (Address > StopAddress)) 1771 continue; 1772 SymbolRef::Type Type = unwrapOrError(Symbol.getType(), ArchiveName, 1773 FileName, ArchitectureName); 1774 uint32_t Flags = Symbol.getFlags(); 1775 section_iterator Section = unwrapOrError(Symbol.getSection(), ArchiveName, 1776 FileName, ArchitectureName); 1777 StringRef Name; 1778 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 1779 if (Expected<StringRef> NameOrErr = Section->getName()) 1780 Name = *NameOrErr; 1781 else 1782 consumeError(NameOrErr.takeError()); 1783 1784 } else { 1785 Name = unwrapOrError(Symbol.getName(), ArchiveName, FileName, 1786 ArchitectureName); 1787 } 1788 1789 bool Global = Flags & SymbolRef::SF_Global; 1790 bool Weak = Flags & SymbolRef::SF_Weak; 1791 bool Absolute = Flags & SymbolRef::SF_Absolute; 1792 bool Common = Flags & SymbolRef::SF_Common; 1793 bool Hidden = Flags & SymbolRef::SF_Hidden; 1794 1795 char GlobLoc = ' '; 1796 if (Type != SymbolRef::ST_Unknown) 1797 GlobLoc = Global ? 'g' : 'l'; 1798 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1799 ? 'd' : ' '; 1800 char FileFunc = ' '; 1801 if (Type == SymbolRef::ST_File) 1802 FileFunc = 'f'; 1803 else if (Type == SymbolRef::ST_Function) 1804 FileFunc = 'F'; 1805 else if (Type == SymbolRef::ST_Data) 1806 FileFunc = 'O'; 1807 1808 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1809 "%08" PRIx64; 1810 1811 outs() << format(Fmt, Address) << " " 1812 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1813 << (Weak ? 'w' : ' ') // Weak? 1814 << ' ' // Constructor. Not supported yet. 1815 << ' ' // Warning. Not supported yet. 1816 << ' ' // Indirect reference to another symbol. 1817 << Debug // Debugging (d) or dynamic (D) symbol. 1818 << FileFunc // Name of function (F), file (f) or object (O). 1819 << ' '; 1820 if (Absolute) { 1821 outs() << "*ABS*"; 1822 } else if (Common) { 1823 outs() << "*COM*"; 1824 } else if (Section == O->section_end()) { 1825 outs() << "*UND*"; 1826 } else { 1827 if (const MachOObjectFile *MachO = 1828 dyn_cast<const MachOObjectFile>(O)) { 1829 DataRefImpl DR = Section->getRawDataRefImpl(); 1830 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1831 outs() << SegmentName << ","; 1832 } 1833 StringRef SectionName = 1834 unwrapOrError(Section->getName(), O->getFileName()); 1835 outs() << SectionName; 1836 } 1837 1838 if (Common || isa<ELFObjectFileBase>(O)) { 1839 uint64_t Val = 1840 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1841 outs() << format("\t%08" PRIx64, Val); 1842 } 1843 1844 if (isa<ELFObjectFileBase>(O)) { 1845 uint8_t Other = ELFSymbolRef(Symbol).getOther(); 1846 switch (Other) { 1847 case ELF::STV_DEFAULT: 1848 break; 1849 case ELF::STV_INTERNAL: 1850 outs() << " .internal"; 1851 break; 1852 case ELF::STV_HIDDEN: 1853 outs() << " .hidden"; 1854 break; 1855 case ELF::STV_PROTECTED: 1856 outs() << " .protected"; 1857 break; 1858 default: 1859 outs() << format(" 0x%02x", Other); 1860 break; 1861 } 1862 } else if (Hidden) { 1863 outs() << " .hidden"; 1864 } 1865 1866 if (Demangle) 1867 outs() << ' ' << demangle(Name) << '\n'; 1868 else 1869 outs() << ' ' << Name << '\n'; 1870 } 1871 } 1872 1873 static void printUnwindInfo(const ObjectFile *O) { 1874 outs() << "Unwind info:\n\n"; 1875 1876 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1877 printCOFFUnwindInfo(Coff); 1878 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1879 printMachOUnwindInfo(MachO); 1880 else 1881 // TODO: Extract DWARF dump tool to objdump. 1882 WithColor::error(errs(), ToolName) 1883 << "This operation is only currently supported " 1884 "for COFF and MachO object files.\n"; 1885 } 1886 1887 /// Dump the raw contents of the __clangast section so the output can be piped 1888 /// into llvm-bcanalyzer. 1889 void printRawClangAST(const ObjectFile *Obj) { 1890 if (outs().is_displayed()) { 1891 WithColor::error(errs(), ToolName) 1892 << "The -raw-clang-ast option will dump the raw binary contents of " 1893 "the clang ast section.\n" 1894 "Please redirect the output to a file or another program such as " 1895 "llvm-bcanalyzer.\n"; 1896 return; 1897 } 1898 1899 StringRef ClangASTSectionName("__clangast"); 1900 if (isa<COFFObjectFile>(Obj)) { 1901 ClangASTSectionName = "clangast"; 1902 } 1903 1904 Optional<object::SectionRef> ClangASTSection; 1905 for (auto Sec : ToolSectionFilter(*Obj)) { 1906 StringRef Name; 1907 if (Expected<StringRef> NameOrErr = Sec.getName()) 1908 Name = *NameOrErr; 1909 else 1910 consumeError(NameOrErr.takeError()); 1911 1912 if (Name == ClangASTSectionName) { 1913 ClangASTSection = Sec; 1914 break; 1915 } 1916 } 1917 if (!ClangASTSection) 1918 return; 1919 1920 StringRef ClangASTContents = unwrapOrError( 1921 ClangASTSection.getValue().getContents(), Obj->getFileName()); 1922 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1923 } 1924 1925 static void printFaultMaps(const ObjectFile *Obj) { 1926 StringRef FaultMapSectionName; 1927 1928 if (isa<ELFObjectFileBase>(Obj)) { 1929 FaultMapSectionName = ".llvm_faultmaps"; 1930 } else if (isa<MachOObjectFile>(Obj)) { 1931 FaultMapSectionName = "__llvm_faultmaps"; 1932 } else { 1933 WithColor::error(errs(), ToolName) 1934 << "This operation is only currently supported " 1935 "for ELF and Mach-O executable files.\n"; 1936 return; 1937 } 1938 1939 Optional<object::SectionRef> FaultMapSection; 1940 1941 for (auto Sec : ToolSectionFilter(*Obj)) { 1942 StringRef Name; 1943 if (Expected<StringRef> NameOrErr = Sec.getName()) 1944 Name = *NameOrErr; 1945 else 1946 consumeError(NameOrErr.takeError()); 1947 1948 if (Name == FaultMapSectionName) { 1949 FaultMapSection = Sec; 1950 break; 1951 } 1952 } 1953 1954 outs() << "FaultMap table:\n"; 1955 1956 if (!FaultMapSection.hasValue()) { 1957 outs() << "<not found>\n"; 1958 return; 1959 } 1960 1961 StringRef FaultMapContents = 1962 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName()); 1963 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1964 FaultMapContents.bytes_end()); 1965 1966 outs() << FMP; 1967 } 1968 1969 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1970 if (O->isELF()) { 1971 printELFFileHeader(O); 1972 printELFDynamicSection(O); 1973 printELFSymbolVersionInfo(O); 1974 return; 1975 } 1976 if (O->isCOFF()) 1977 return printCOFFFileHeader(O); 1978 if (O->isWasm()) 1979 return printWasmFileHeader(O); 1980 if (O->isMachO()) { 1981 printMachOFileHeader(O); 1982 if (!OnlyFirst) 1983 printMachOLoadCommands(O); 1984 return; 1985 } 1986 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1987 } 1988 1989 static void printFileHeaders(const ObjectFile *O) { 1990 if (!O->isELF() && !O->isCOFF()) 1991 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1992 1993 Triple::ArchType AT = O->getArch(); 1994 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1995 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 1996 1997 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1998 outs() << "start address: " 1999 << "0x" << format(Fmt.data(), Address) << "\n\n"; 2000 } 2001 2002 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2003 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2004 if (!ModeOrErr) { 2005 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 2006 consumeError(ModeOrErr.takeError()); 2007 return; 2008 } 2009 sys::fs::perms Mode = ModeOrErr.get(); 2010 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2011 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2012 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2013 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2014 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2015 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2016 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2017 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2018 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2019 2020 outs() << " "; 2021 2022 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 2023 unwrapOrError(C.getGID(), Filename), 2024 unwrapOrError(C.getRawSize(), Filename)); 2025 2026 StringRef RawLastModified = C.getRawLastModified(); 2027 unsigned Seconds; 2028 if (RawLastModified.getAsInteger(10, Seconds)) 2029 outs() << "(date: \"" << RawLastModified 2030 << "\" contains non-decimal chars) "; 2031 else { 2032 // Since ctime(3) returns a 26 character string of the form: 2033 // "Sun Sep 16 01:03:52 1973\n\0" 2034 // just print 24 characters. 2035 time_t t = Seconds; 2036 outs() << format("%.24s ", ctime(&t)); 2037 } 2038 2039 StringRef Name = ""; 2040 Expected<StringRef> NameOrErr = C.getName(); 2041 if (!NameOrErr) { 2042 consumeError(NameOrErr.takeError()); 2043 Name = unwrapOrError(C.getRawName(), Filename); 2044 } else { 2045 Name = NameOrErr.get(); 2046 } 2047 outs() << Name << "\n"; 2048 } 2049 2050 // For ELF only now. 2051 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) { 2052 if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) { 2053 if (Elf->getEType() != ELF::ET_REL) 2054 return true; 2055 } 2056 return false; 2057 } 2058 2059 static void checkForInvalidStartStopAddress(ObjectFile *Obj, 2060 uint64_t Start, uint64_t Stop) { 2061 if (!shouldWarnForInvalidStartStopAddress(Obj)) 2062 return; 2063 2064 for (const SectionRef &Section : Obj->sections()) 2065 if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) { 2066 uint64_t BaseAddr = Section.getAddress(); 2067 uint64_t Size = Section.getSize(); 2068 if ((Start < BaseAddr + Size) && Stop > BaseAddr) 2069 return; 2070 } 2071 2072 if (StartAddress.getNumOccurrences() == 0) 2073 warn("no section has address less than 0x" + 2074 Twine::utohexstr(Stop) + " specified by --stop-address"); 2075 else if (StopAddress.getNumOccurrences() == 0) 2076 warn("no section has address greater than or equal to 0x" + 2077 Twine::utohexstr(Start) + " specified by --start-address"); 2078 else 2079 warn("no section overlaps the range [0x" + 2080 Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) + 2081 ") specified by --start-address/--stop-address"); 2082 } 2083 2084 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 2085 const Archive::Child *C = nullptr) { 2086 // Avoid other output when using a raw option. 2087 if (!RawClangAST) { 2088 outs() << '\n'; 2089 if (A) 2090 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 2091 else 2092 outs() << O->getFileName(); 2093 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 2094 } 2095 2096 if (StartAddress.getNumOccurrences() || StopAddress.getNumOccurrences()) 2097 checkForInvalidStartStopAddress(O, StartAddress, StopAddress); 2098 2099 StringRef ArchiveName = A ? A->getFileName() : ""; 2100 if (FileHeaders) 2101 printFileHeaders(O); 2102 if (ArchiveHeaders && !MachOOpt && C) 2103 printArchiveChild(ArchiveName, *C); 2104 if (Disassemble) 2105 disassembleObject(O, Relocations); 2106 if (Relocations && !Disassemble) 2107 printRelocations(O); 2108 if (DynamicRelocations) 2109 printDynamicRelocations(O); 2110 if (SectionHeaders) 2111 printSectionHeaders(O); 2112 if (SectionContents) 2113 printSectionContents(O); 2114 if (SymbolTable) 2115 printSymbolTable(O, ArchiveName); 2116 if (UnwindInfo) 2117 printUnwindInfo(O); 2118 if (PrivateHeaders || FirstPrivateHeader) 2119 printPrivateFileHeaders(O, FirstPrivateHeader); 2120 if (ExportsTrie) 2121 printExportsTrie(O); 2122 if (Rebase) 2123 printRebaseTable(O); 2124 if (Bind) 2125 printBindTable(O); 2126 if (LazyBind) 2127 printLazyBindTable(O); 2128 if (WeakBind) 2129 printWeakBindTable(O); 2130 if (RawClangAST) 2131 printRawClangAST(O); 2132 if (FaultMapSection) 2133 printFaultMaps(O); 2134 if (DwarfDumpType != DIDT_Null) { 2135 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2136 // Dump the complete DWARF structure. 2137 DIDumpOptions DumpOpts; 2138 DumpOpts.DumpType = DwarfDumpType; 2139 DICtx->dump(outs(), DumpOpts); 2140 } 2141 } 2142 2143 static void dumpObject(const COFFImportFile *I, const Archive *A, 2144 const Archive::Child *C = nullptr) { 2145 StringRef ArchiveName = A ? A->getFileName() : ""; 2146 2147 // Avoid other output when using a raw option. 2148 if (!RawClangAST) 2149 outs() << '\n' 2150 << ArchiveName << "(" << I->getFileName() << ")" 2151 << ":\tfile format COFF-import-file" 2152 << "\n\n"; 2153 2154 if (ArchiveHeaders && !MachOOpt && C) 2155 printArchiveChild(ArchiveName, *C); 2156 if (SymbolTable) 2157 printCOFFSymbolTable(I); 2158 } 2159 2160 /// Dump each object file in \a a; 2161 static void dumpArchive(const Archive *A) { 2162 Error Err = Error::success(); 2163 for (auto &C : A->children(Err)) { 2164 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2165 if (!ChildOrErr) { 2166 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2167 report_error(std::move(E), A->getFileName(), C); 2168 continue; 2169 } 2170 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2171 dumpObject(O, A, &C); 2172 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2173 dumpObject(I, A, &C); 2174 else 2175 report_error(errorCodeToError(object_error::invalid_file_type), 2176 A->getFileName()); 2177 } 2178 if (Err) 2179 report_error(std::move(Err), A->getFileName()); 2180 } 2181 2182 /// Open file and figure out how to dump it. 2183 static void dumpInput(StringRef file) { 2184 // If we are using the Mach-O specific object file parser, then let it parse 2185 // the file and process the command line options. So the -arch flags can 2186 // be used to select specific slices, etc. 2187 if (MachOOpt) { 2188 parseInputMachO(file); 2189 return; 2190 } 2191 2192 // Attempt to open the binary. 2193 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2194 Binary &Binary = *OBinary.getBinary(); 2195 2196 if (Archive *A = dyn_cast<Archive>(&Binary)) 2197 dumpArchive(A); 2198 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2199 dumpObject(O); 2200 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2201 parseInputMachO(UB); 2202 else 2203 report_error(errorCodeToError(object_error::invalid_file_type), file); 2204 } 2205 } // namespace llvm 2206 2207 int main(int argc, char **argv) { 2208 using namespace llvm; 2209 InitLLVM X(argc, argv); 2210 const cl::OptionCategory *OptionFilters[] = {&ObjdumpCat, &MachOCat}; 2211 cl::HideUnrelatedOptions(OptionFilters); 2212 2213 // Initialize targets and assembly printers/parsers. 2214 InitializeAllTargetInfos(); 2215 InitializeAllTargetMCs(); 2216 InitializeAllDisassemblers(); 2217 2218 // Register the target printer for --version. 2219 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2220 2221 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2222 2223 if (StartAddress >= StopAddress) 2224 error("start address should be less than stop address"); 2225 2226 ToolName = argv[0]; 2227 2228 // Defaults to a.out if no filenames specified. 2229 if (InputFilenames.empty()) 2230 InputFilenames.push_back("a.out"); 2231 2232 if (AllHeaders) 2233 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2234 SectionHeaders = SymbolTable = true; 2235 2236 if (DisassembleAll || PrintSource || PrintLines || 2237 (!DisassembleFunctions.empty())) 2238 Disassemble = true; 2239 2240 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null && 2241 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST && 2242 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable && 2243 !UnwindInfo && !FaultMapSection && 2244 !(MachOOpt && 2245 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie || 2246 FirstPrivateHeader || IndirectSymbols || InfoPlist || LazyBind || 2247 LinkOptHints || ObjcMetaData || Rebase || UniversalHeaders || 2248 WeakBind || !FilterSections.empty()))) { 2249 cl::PrintHelpMessage(); 2250 return 2; 2251 } 2252 2253 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2254 DisassembleFunctions.end()); 2255 2256 llvm::for_each(InputFilenames, dumpInput); 2257 2258 warnOnNoMatchForSections(); 2259 2260 return EXIT_SUCCESS; 2261 } 2262