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