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 "COFFDump.h" 20 #include "ELFDump.h" 21 #include "MachODump.h" 22 #include "WasmDump.h" 23 #include "XCOFFDump.h" 24 #include "llvm/ADT/Optional.h" 25 #include "llvm/ADT/STLExtras.h" 26 #include "llvm/ADT/SetOperations.h" 27 #include "llvm/ADT/StringExtras.h" 28 #include "llvm/ADT/StringSet.h" 29 #include "llvm/ADT/Triple.h" 30 #include "llvm/CodeGen/FaultMaps.h" 31 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 32 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 33 #include "llvm/Demangle/Demangle.h" 34 #include "llvm/MC/MCAsmInfo.h" 35 #include "llvm/MC/MCContext.h" 36 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 37 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 38 #include "llvm/MC/MCInst.h" 39 #include "llvm/MC/MCInstPrinter.h" 40 #include "llvm/MC/MCInstrAnalysis.h" 41 #include "llvm/MC/MCInstrInfo.h" 42 #include "llvm/MC/MCObjectFileInfo.h" 43 #include "llvm/MC/MCRegisterInfo.h" 44 #include "llvm/MC/MCSubtargetInfo.h" 45 #include "llvm/MC/MCTargetOptions.h" 46 #include "llvm/Object/Archive.h" 47 #include "llvm/Object/COFF.h" 48 #include "llvm/Object/COFFImportFile.h" 49 #include "llvm/Object/ELFObjectFile.h" 50 #include "llvm/Object/MachO.h" 51 #include "llvm/Object/MachOUniversal.h" 52 #include "llvm/Object/ObjectFile.h" 53 #include "llvm/Object/Wasm.h" 54 #include "llvm/Support/Casting.h" 55 #include "llvm/Support/CommandLine.h" 56 #include "llvm/Support/Debug.h" 57 #include "llvm/Support/Errc.h" 58 #include "llvm/Support/FileSystem.h" 59 #include "llvm/Support/Format.h" 60 #include "llvm/Support/FormatVariadic.h" 61 #include "llvm/Support/GraphWriter.h" 62 #include "llvm/Support/Host.h" 63 #include "llvm/Support/InitLLVM.h" 64 #include "llvm/Support/MemoryBuffer.h" 65 #include "llvm/Support/SourceMgr.h" 66 #include "llvm/Support/StringSaver.h" 67 #include "llvm/Support/TargetRegistry.h" 68 #include "llvm/Support/TargetSelect.h" 69 #include "llvm/Support/WithColor.h" 70 #include "llvm/Support/raw_ostream.h" 71 #include <algorithm> 72 #include <cctype> 73 #include <cstring> 74 #include <system_error> 75 #include <unordered_map> 76 #include <utility> 77 78 using namespace llvm; 79 using namespace llvm::object; 80 using namespace llvm::objdump; 81 82 static cl::OptionCategory ObjdumpCat("llvm-objdump Options"); 83 84 static cl::opt<uint64_t> AdjustVMA( 85 "adjust-vma", 86 cl::desc("Increase the displayed address by the specified offset"), 87 cl::value_desc("offset"), cl::init(0), cl::cat(ObjdumpCat)); 88 89 static cl::opt<bool> 90 AllHeaders("all-headers", 91 cl::desc("Display all available header information"), 92 cl::cat(ObjdumpCat)); 93 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 94 cl::NotHidden, cl::Grouping, 95 cl::aliasopt(AllHeaders)); 96 97 static cl::opt<std::string> 98 ArchName("arch-name", 99 cl::desc("Target arch to disassemble for, " 100 "see -version for available targets"), 101 cl::cat(ObjdumpCat)); 102 103 cl::opt<bool> 104 objdump::ArchiveHeaders("archive-headers", 105 cl::desc("Display archive header information"), 106 cl::cat(ObjdumpCat)); 107 static cl::alias ArchiveHeadersShort("a", 108 cl::desc("Alias for --archive-headers"), 109 cl::NotHidden, cl::Grouping, 110 cl::aliasopt(ArchiveHeaders)); 111 112 cl::opt<bool> objdump::Demangle("demangle", cl::desc("Demangle symbols names"), 113 cl::init(false), cl::cat(ObjdumpCat)); 114 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 115 cl::NotHidden, cl::Grouping, 116 cl::aliasopt(Demangle)); 117 118 cl::opt<bool> objdump::Disassemble( 119 "disassemble", 120 cl::desc("Display assembler mnemonics for the machine instructions"), 121 cl::cat(ObjdumpCat)); 122 static cl::alias DisassembleShort("d", cl::desc("Alias for --disassemble"), 123 cl::NotHidden, cl::Grouping, 124 cl::aliasopt(Disassemble)); 125 126 cl::opt<bool> objdump::DisassembleAll( 127 "disassemble-all", 128 cl::desc("Display assembler mnemonics for the machine instructions"), 129 cl::cat(ObjdumpCat)); 130 static cl::alias DisassembleAllShort("D", 131 cl::desc("Alias for --disassemble-all"), 132 cl::NotHidden, cl::Grouping, 133 cl::aliasopt(DisassembleAll)); 134 135 cl::opt<bool> objdump::SymbolDescription( 136 "symbol-description", 137 cl::desc("Add symbol description for disassembly. This " 138 "option is for XCOFF files only"), 139 cl::init(false), cl::cat(ObjdumpCat)); 140 141 static cl::list<std::string> 142 DisassembleSymbols("disassemble-symbols", cl::CommaSeparated, 143 cl::desc("List of symbols to disassemble. " 144 "Accept demangled names when --demangle is " 145 "specified, otherwise accept mangled names"), 146 cl::cat(ObjdumpCat)); 147 148 static cl::opt<bool> DisassembleZeroes( 149 "disassemble-zeroes", 150 cl::desc("Do not skip blocks of zeroes when disassembling"), 151 cl::cat(ObjdumpCat)); 152 static cl::alias 153 DisassembleZeroesShort("z", cl::desc("Alias for --disassemble-zeroes"), 154 cl::NotHidden, cl::Grouping, 155 cl::aliasopt(DisassembleZeroes)); 156 157 static cl::list<std::string> 158 DisassemblerOptions("disassembler-options", 159 cl::desc("Pass target specific disassembler options"), 160 cl::value_desc("options"), cl::CommaSeparated, 161 cl::cat(ObjdumpCat)); 162 static cl::alias 163 DisassemblerOptionsShort("M", cl::desc("Alias for --disassembler-options"), 164 cl::NotHidden, cl::Grouping, cl::Prefix, 165 cl::CommaSeparated, 166 cl::aliasopt(DisassemblerOptions)); 167 168 cl::opt<DIDumpType> objdump::DwarfDumpType( 169 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 170 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame")), 171 cl::cat(ObjdumpCat)); 172 173 static cl::opt<bool> DynamicRelocations( 174 "dynamic-reloc", 175 cl::desc("Display the dynamic relocation entries in the file"), 176 cl::cat(ObjdumpCat)); 177 static cl::alias DynamicRelocationShort("R", 178 cl::desc("Alias for --dynamic-reloc"), 179 cl::NotHidden, cl::Grouping, 180 cl::aliasopt(DynamicRelocations)); 181 182 static cl::opt<bool> 183 FaultMapSection("fault-map-section", 184 cl::desc("Display contents of faultmap section"), 185 cl::cat(ObjdumpCat)); 186 187 static cl::opt<bool> 188 FileHeaders("file-headers", 189 cl::desc("Display the contents of the overall file header"), 190 cl::cat(ObjdumpCat)); 191 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 192 cl::NotHidden, cl::Grouping, 193 cl::aliasopt(FileHeaders)); 194 195 cl::opt<bool> 196 objdump::SectionContents("full-contents", 197 cl::desc("Display the content of each section"), 198 cl::cat(ObjdumpCat)); 199 static cl::alias SectionContentsShort("s", 200 cl::desc("Alias for --full-contents"), 201 cl::NotHidden, cl::Grouping, 202 cl::aliasopt(SectionContents)); 203 204 static cl::list<std::string> InputFilenames(cl::Positional, 205 cl::desc("<input object files>"), 206 cl::ZeroOrMore, 207 cl::cat(ObjdumpCat)); 208 209 static cl::opt<bool> 210 PrintLines("line-numbers", 211 cl::desc("Display source line numbers with " 212 "disassembly. Implies disassemble object"), 213 cl::cat(ObjdumpCat)); 214 static cl::alias PrintLinesShort("l", cl::desc("Alias for --line-numbers"), 215 cl::NotHidden, cl::Grouping, 216 cl::aliasopt(PrintLines)); 217 218 static cl::opt<bool> MachOOpt("macho", 219 cl::desc("Use MachO specific object file parser"), 220 cl::cat(ObjdumpCat)); 221 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 222 cl::Grouping, cl::aliasopt(MachOOpt)); 223 224 cl::opt<std::string> objdump::MCPU( 225 "mcpu", cl::desc("Target a specific cpu type (-mcpu=help for details)"), 226 cl::value_desc("cpu-name"), cl::init(""), cl::cat(ObjdumpCat)); 227 228 cl::list<std::string> objdump::MAttrs("mattr", cl::CommaSeparated, 229 cl::desc("Target specific attributes"), 230 cl::value_desc("a1,+a2,-a3,..."), 231 cl::cat(ObjdumpCat)); 232 233 cl::opt<bool> objdump::NoShowRawInsn( 234 "no-show-raw-insn", 235 cl::desc( 236 "When disassembling instructions, do not print the instruction bytes."), 237 cl::cat(ObjdumpCat)); 238 239 cl::opt<bool> objdump::NoLeadingAddr("no-leading-addr", 240 cl::desc("Print no leading address"), 241 cl::cat(ObjdumpCat)); 242 243 static cl::opt<bool> RawClangAST( 244 "raw-clang-ast", 245 cl::desc("Dump the raw binary contents of the clang AST section"), 246 cl::cat(ObjdumpCat)); 247 248 cl::opt<bool> 249 objdump::Relocations("reloc", 250 cl::desc("Display the relocation entries in the file"), 251 cl::cat(ObjdumpCat)); 252 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 253 cl::NotHidden, cl::Grouping, 254 cl::aliasopt(Relocations)); 255 256 cl::opt<bool> 257 objdump::PrintImmHex("print-imm-hex", 258 cl::desc("Use hex format for immediate values"), 259 cl::cat(ObjdumpCat)); 260 261 cl::opt<bool> 262 objdump::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 objdump::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> objdump::SectionHeaders( 280 "section-headers", 281 cl::desc("Display summaries of the 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> objdump::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 static cl::opt<bool> DynamicSymbolTable( 321 "dynamic-syms", 322 cl::desc("Display the contents of the dynamic symbol table"), 323 cl::cat(ObjdumpCat)); 324 static cl::alias DynamicSymbolTableShort("T", 325 cl::desc("Alias for --dynamic-syms"), 326 cl::NotHidden, cl::Grouping, 327 cl::aliasopt(DynamicSymbolTable)); 328 329 cl::opt<std::string> objdump::TripleName( 330 "triple", 331 cl::desc( 332 "Target triple to disassemble for, see -version for available targets"), 333 cl::cat(ObjdumpCat)); 334 335 cl::opt<bool> objdump::UnwindInfo("unwind-info", 336 cl::desc("Display unwind information"), 337 cl::cat(ObjdumpCat)); 338 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 339 cl::NotHidden, cl::Grouping, 340 cl::aliasopt(UnwindInfo)); 341 342 static cl::opt<bool> 343 Wide("wide", cl::desc("Ignored for compatibility with GNU objdump"), 344 cl::cat(ObjdumpCat)); 345 static cl::alias WideShort("w", cl::Grouping, cl::aliasopt(Wide)); 346 347 static cl::extrahelp 348 HelpResponse("\nPass @FILE as argument to read options from FILE.\n"); 349 350 static StringSet<> DisasmSymbolSet; 351 StringSet<> objdump::FoundSectionSet; 352 static StringRef ToolName; 353 354 namespace { 355 struct FilterResult { 356 // True if the section should not be skipped. 357 bool Keep; 358 359 // True if the index counter should be incremented, even if the section should 360 // be skipped. For example, sections may be skipped if they are not included 361 // in the --section flag, but we still want those to count toward the section 362 // count. 363 bool IncrementIndex; 364 }; 365 } // namespace 366 367 static FilterResult checkSectionFilter(object::SectionRef S) { 368 if (FilterSections.empty()) 369 return {/*Keep=*/true, /*IncrementIndex=*/true}; 370 371 Expected<StringRef> SecNameOrErr = S.getName(); 372 if (!SecNameOrErr) { 373 consumeError(SecNameOrErr.takeError()); 374 return {/*Keep=*/false, /*IncrementIndex=*/false}; 375 } 376 StringRef SecName = *SecNameOrErr; 377 378 // StringSet does not allow empty key so avoid adding sections with 379 // no name (such as the section with index 0) here. 380 if (!SecName.empty()) 381 FoundSectionSet.insert(SecName); 382 383 // Only show the section if it's in the FilterSections list, but always 384 // increment so the indexing is stable. 385 return {/*Keep=*/is_contained(FilterSections, SecName), 386 /*IncrementIndex=*/true}; 387 } 388 389 namespace llvm { 390 391 SectionFilter ToolSectionFilter(object::ObjectFile const &O, uint64_t *Idx) { 392 // Start at UINT64_MAX so that the first index returned after an increment is 393 // zero (after the unsigned wrap). 394 if (Idx) 395 *Idx = UINT64_MAX; 396 return SectionFilter( 397 [Idx](object::SectionRef S) { 398 FilterResult Result = checkSectionFilter(S); 399 if (Idx != nullptr && Result.IncrementIndex) 400 *Idx += 1; 401 return Result.Keep; 402 }, 403 O); 404 } 405 406 std::string getFileNameForError(const object::Archive::Child &C, 407 unsigned Index) { 408 Expected<StringRef> NameOrErr = C.getName(); 409 if (NameOrErr) 410 return std::string(NameOrErr.get()); 411 // If we have an error getting the name then we print the index of the archive 412 // member. Since we are already in an error state, we just ignore this error. 413 consumeError(NameOrErr.takeError()); 414 return "<file index: " + std::to_string(Index) + ">"; 415 } 416 417 void reportWarning(Twine Message, StringRef File) { 418 // Output order between errs() and outs() matters especially for archive 419 // files where the output is per member object. 420 outs().flush(); 421 WithColor::warning(errs(), ToolName) 422 << "'" << File << "': " << Message << "\n"; 423 errs().flush(); 424 } 425 426 LLVM_ATTRIBUTE_NORETURN void reportError(StringRef File, Twine Message) { 427 WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n"; 428 exit(1); 429 } 430 431 LLVM_ATTRIBUTE_NORETURN void reportError(Error E, StringRef FileName, 432 StringRef ArchiveName, 433 StringRef ArchitectureName) { 434 assert(E); 435 WithColor::error(errs(), ToolName); 436 if (ArchiveName != "") 437 errs() << ArchiveName << "(" << FileName << ")"; 438 else 439 errs() << "'" << FileName << "'"; 440 if (!ArchitectureName.empty()) 441 errs() << " (for architecture " << ArchitectureName << ")"; 442 std::string Buf; 443 raw_string_ostream OS(Buf); 444 logAllUnhandledErrors(std::move(E), OS); 445 OS.flush(); 446 errs() << ": " << Buf; 447 exit(1); 448 } 449 450 static void reportCmdLineWarning(Twine Message) { 451 WithColor::warning(errs(), ToolName) << Message << "\n"; 452 } 453 454 LLVM_ATTRIBUTE_NORETURN static void reportCmdLineError(Twine Message) { 455 WithColor::error(errs(), ToolName) << Message << "\n"; 456 exit(1); 457 } 458 459 static void warnOnNoMatchForSections() { 460 SetVector<StringRef> MissingSections; 461 for (StringRef S : FilterSections) { 462 if (FoundSectionSet.count(S)) 463 return; 464 // User may specify a unnamed section. Don't warn for it. 465 if (!S.empty()) 466 MissingSections.insert(S); 467 } 468 469 // Warn only if no section in FilterSections is matched. 470 for (StringRef S : MissingSections) 471 reportCmdLineWarning("section '" + S + 472 "' mentioned in a -j/--section option, but not " 473 "found in any input file"); 474 } 475 476 static const Target *getTarget(const ObjectFile *Obj) { 477 // Figure out the target triple. 478 Triple TheTriple("unknown-unknown-unknown"); 479 if (TripleName.empty()) { 480 TheTriple = Obj->makeTriple(); 481 } else { 482 TheTriple.setTriple(Triple::normalize(TripleName)); 483 auto Arch = Obj->getArch(); 484 if (Arch == Triple::arm || Arch == Triple::armeb) 485 Obj->setARMSubArch(TheTriple); 486 } 487 488 // Get the target specific parser. 489 std::string Error; 490 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 491 Error); 492 if (!TheTarget) 493 reportError(Obj->getFileName(), "can't find target: " + Error); 494 495 // Update the triple name and return the found target. 496 TripleName = TheTriple.getTriple(); 497 return TheTarget; 498 } 499 500 bool isRelocAddressLess(RelocationRef A, RelocationRef B) { 501 return A.getOffset() < B.getOffset(); 502 } 503 504 static Error getRelocationValueString(const RelocationRef &Rel, 505 SmallVectorImpl<char> &Result) { 506 const ObjectFile *Obj = Rel.getObject(); 507 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 508 return getELFRelocationValueString(ELF, Rel, Result); 509 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 510 return getCOFFRelocationValueString(COFF, Rel, Result); 511 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 512 return getWasmRelocationValueString(Wasm, Rel, Result); 513 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 514 return getMachORelocationValueString(MachO, Rel, Result); 515 if (auto *XCOFF = dyn_cast<XCOFFObjectFile>(Obj)) 516 return getXCOFFRelocationValueString(XCOFF, Rel, Result); 517 llvm_unreachable("unknown object file format"); 518 } 519 520 /// Indicates whether this relocation should hidden when listing 521 /// relocations, usually because it is the trailing part of a multipart 522 /// relocation that will be printed as part of the leading relocation. 523 static bool getHidden(RelocationRef RelRef) { 524 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 525 if (!MachO) 526 return false; 527 528 unsigned Arch = MachO->getArch(); 529 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 530 uint64_t Type = MachO->getRelocationType(Rel); 531 532 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 533 // is always hidden. 534 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 535 return Type == MachO::GENERIC_RELOC_PAIR; 536 537 if (Arch == Triple::x86_64) { 538 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 539 // an X86_64_RELOC_SUBTRACTOR. 540 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 541 DataRefImpl RelPrev = Rel; 542 RelPrev.d.a--; 543 uint64_t PrevType = MachO->getRelocationType(RelPrev); 544 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 545 return true; 546 } 547 } 548 549 return false; 550 } 551 552 namespace { 553 class SourcePrinter { 554 protected: 555 DILineInfo OldLineInfo; 556 const ObjectFile *Obj = nullptr; 557 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 558 // File name to file contents of source. 559 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 560 // Mark the line endings of the cached source. 561 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 562 // Keep track of missing sources. 563 StringSet<> MissingSources; 564 // Only emit 'no debug info' warning once. 565 bool WarnedNoDebugInfo; 566 567 private: 568 bool cacheSource(const DILineInfo& LineInfoFile); 569 570 void printLines(raw_ostream &OS, const DILineInfo &LineInfo, 571 StringRef Delimiter); 572 573 void printSources(raw_ostream &OS, const DILineInfo &LineInfo, 574 StringRef ObjectFilename, StringRef Delimiter); 575 576 public: 577 SourcePrinter() = default; 578 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) 579 : Obj(Obj), WarnedNoDebugInfo(false) { 580 symbolize::LLVMSymbolizer::Options SymbolizerOpts; 581 SymbolizerOpts.PrintFunctions = 582 DILineInfoSpecifier::FunctionNameKind::LinkageName; 583 SymbolizerOpts.Demangle = Demangle; 584 SymbolizerOpts.DefaultArch = std::string(DefaultArch); 585 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 586 } 587 virtual ~SourcePrinter() = default; 588 virtual void printSourceLine(raw_ostream &OS, 589 object::SectionedAddress Address, 590 StringRef ObjectFilename, 591 StringRef Delimiter = "; "); 592 }; 593 594 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 595 std::unique_ptr<MemoryBuffer> Buffer; 596 if (LineInfo.Source) { 597 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 598 } else { 599 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 600 if (!BufferOrError) { 601 if (MissingSources.insert(LineInfo.FileName).second) 602 reportWarning("failed to find source " + LineInfo.FileName, 603 Obj->getFileName()); 604 return false; 605 } 606 Buffer = std::move(*BufferOrError); 607 } 608 // Chomp the file to get lines 609 const char *BufferStart = Buffer->getBufferStart(), 610 *BufferEnd = Buffer->getBufferEnd(); 611 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 612 const char *Start = BufferStart; 613 for (const char *I = BufferStart; I != BufferEnd; ++I) 614 if (*I == '\n') { 615 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 616 Start = I + 1; 617 } 618 if (Start < BufferEnd) 619 Lines.emplace_back(Start, BufferEnd - Start); 620 SourceCache[LineInfo.FileName] = std::move(Buffer); 621 return true; 622 } 623 624 void SourcePrinter::printSourceLine(raw_ostream &OS, 625 object::SectionedAddress Address, 626 StringRef ObjectFilename, 627 StringRef Delimiter) { 628 if (!Symbolizer) 629 return; 630 631 DILineInfo LineInfo = DILineInfo(); 632 auto ExpectedLineInfo = Symbolizer->symbolizeCode(*Obj, Address); 633 std::string ErrorMessage; 634 if (!ExpectedLineInfo) 635 ErrorMessage = toString(ExpectedLineInfo.takeError()); 636 else 637 LineInfo = *ExpectedLineInfo; 638 639 if (LineInfo.FileName == DILineInfo::BadString) { 640 if (!WarnedNoDebugInfo) { 641 std::string Warning = 642 "failed to parse debug information for " + ObjectFilename.str(); 643 if (!ErrorMessage.empty()) 644 Warning += ": " + ErrorMessage; 645 reportWarning(Warning, ObjectFilename); 646 WarnedNoDebugInfo = true; 647 } 648 } 649 650 if (PrintLines) 651 printLines(OS, LineInfo, Delimiter); 652 if (PrintSource) 653 printSources(OS, LineInfo, ObjectFilename, Delimiter); 654 OldLineInfo = LineInfo; 655 } 656 657 void SourcePrinter::printLines(raw_ostream &OS, const DILineInfo &LineInfo, 658 StringRef Delimiter) { 659 bool PrintFunctionName = LineInfo.FunctionName != DILineInfo::BadString && 660 LineInfo.FunctionName != OldLineInfo.FunctionName; 661 if (PrintFunctionName) { 662 OS << Delimiter << LineInfo.FunctionName; 663 // If demangling is successful, FunctionName will end with "()". Print it 664 // only if demangling did not run or was unsuccessful. 665 if (!StringRef(LineInfo.FunctionName).endswith("()")) 666 OS << "()"; 667 OS << ":\n"; 668 } 669 if (LineInfo.FileName != DILineInfo::BadString && LineInfo.Line != 0 && 670 (OldLineInfo.Line != LineInfo.Line || 671 OldLineInfo.FileName != LineInfo.FileName || PrintFunctionName)) 672 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 673 } 674 675 void SourcePrinter::printSources(raw_ostream &OS, const DILineInfo &LineInfo, 676 StringRef ObjectFilename, 677 StringRef Delimiter) { 678 if (LineInfo.FileName == DILineInfo::BadString || LineInfo.Line == 0 || 679 (OldLineInfo.Line == LineInfo.Line && 680 OldLineInfo.FileName == LineInfo.FileName)) 681 return; 682 683 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 684 if (!cacheSource(LineInfo)) 685 return; 686 auto LineBuffer = LineCache.find(LineInfo.FileName); 687 if (LineBuffer != LineCache.end()) { 688 if (LineInfo.Line > LineBuffer->second.size()) { 689 reportWarning( 690 formatv( 691 "debug info line number {0} exceeds the number of lines in {1}", 692 LineInfo.Line, LineInfo.FileName), 693 ObjectFilename); 694 return; 695 } 696 // Vector begins at 0, line numbers are non-zero 697 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1] << '\n'; 698 } 699 } 700 701 static bool isAArch64Elf(const ObjectFile *Obj) { 702 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 703 return Elf && Elf->getEMachine() == ELF::EM_AARCH64; 704 } 705 706 static bool isArmElf(const ObjectFile *Obj) { 707 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 708 return Elf && Elf->getEMachine() == ELF::EM_ARM; 709 } 710 711 static bool hasMappingSymbols(const ObjectFile *Obj) { 712 return isArmElf(Obj) || isAArch64Elf(Obj); 713 } 714 715 static void printRelocation(StringRef FileName, const RelocationRef &Rel, 716 uint64_t Address, bool Is64Bits) { 717 StringRef Fmt = Is64Bits ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 718 SmallString<16> Name; 719 SmallString<32> Val; 720 Rel.getTypeName(Name); 721 if (Error E = getRelocationValueString(Rel, Val)) 722 reportError(std::move(E), FileName); 723 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n"; 724 } 725 726 class PrettyPrinter { 727 public: 728 virtual ~PrettyPrinter() = default; 729 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 730 ArrayRef<uint8_t> Bytes, 731 object::SectionedAddress Address, raw_ostream &OS, 732 StringRef Annot, MCSubtargetInfo const &STI, 733 SourcePrinter *SP, StringRef ObjectFilename, 734 std::vector<RelocationRef> *Rels = nullptr) { 735 if (SP && (PrintSource || PrintLines)) 736 SP->printSourceLine(OS, Address, ObjectFilename); 737 738 size_t Start = OS.tell(); 739 if (!NoLeadingAddr) 740 OS << format("%8" PRIx64 ":", Address.Address); 741 if (!NoShowRawInsn) { 742 OS << ' '; 743 dumpBytes(Bytes, OS); 744 } 745 746 // The output of printInst starts with a tab. Print some spaces so that 747 // the tab has 1 column and advances to the target tab stop. 748 unsigned TabStop = NoShowRawInsn ? 16 : 40; 749 unsigned Column = OS.tell() - Start; 750 OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8); 751 752 if (MI) { 753 // See MCInstPrinter::printInst. On targets where a PC relative immediate 754 // is relative to the next instruction and the length of a MCInst is 755 // difficult to measure (x86), this is the address of the next 756 // instruction. 757 uint64_t Addr = 758 Address.Address + (STI.getTargetTriple().isX86() ? Bytes.size() : 0); 759 IP.printInst(MI, Addr, "", STI, OS); 760 } else 761 OS << "\t<unknown>"; 762 } 763 }; 764 PrettyPrinter PrettyPrinterInst; 765 766 class HexagonPrettyPrinter : public PrettyPrinter { 767 public: 768 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 769 raw_ostream &OS) { 770 uint32_t opcode = 771 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 772 if (!NoLeadingAddr) 773 OS << format("%8" PRIx64 ":", Address); 774 if (!NoShowRawInsn) { 775 OS << "\t"; 776 dumpBytes(Bytes.slice(0, 4), OS); 777 OS << format("\t%08" PRIx32, opcode); 778 } 779 } 780 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 781 object::SectionedAddress Address, raw_ostream &OS, 782 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 783 StringRef ObjectFilename, 784 std::vector<RelocationRef> *Rels) override { 785 if (SP && (PrintSource || PrintLines)) 786 SP->printSourceLine(OS, Address, ObjectFilename, ""); 787 if (!MI) { 788 printLead(Bytes, Address.Address, OS); 789 OS << " <unknown>"; 790 return; 791 } 792 std::string Buffer; 793 { 794 raw_string_ostream TempStream(Buffer); 795 IP.printInst(MI, Address.Address, "", STI, TempStream); 796 } 797 StringRef Contents(Buffer); 798 // Split off bundle attributes 799 auto PacketBundle = Contents.rsplit('\n'); 800 // Split off first instruction from the rest 801 auto HeadTail = PacketBundle.first.split('\n'); 802 auto Preamble = " { "; 803 auto Separator = ""; 804 805 // Hexagon's packets require relocations to be inline rather than 806 // clustered at the end of the packet. 807 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 808 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 809 auto PrintReloc = [&]() -> void { 810 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 811 if (RelCur->getOffset() == Address.Address) { 812 printRelocation(ObjectFilename, *RelCur, Address.Address, false); 813 return; 814 } 815 ++RelCur; 816 } 817 }; 818 819 while (!HeadTail.first.empty()) { 820 OS << Separator; 821 Separator = "\n"; 822 if (SP && (PrintSource || PrintLines)) 823 SP->printSourceLine(OS, Address, ObjectFilename, ""); 824 printLead(Bytes, Address.Address, OS); 825 OS << Preamble; 826 Preamble = " "; 827 StringRef Inst; 828 auto Duplex = HeadTail.first.split('\v'); 829 if (!Duplex.second.empty()) { 830 OS << Duplex.first; 831 OS << "; "; 832 Inst = Duplex.second; 833 } 834 else 835 Inst = HeadTail.first; 836 OS << Inst; 837 HeadTail = HeadTail.second.split('\n'); 838 if (HeadTail.first.empty()) 839 OS << " } " << PacketBundle.second; 840 PrintReloc(); 841 Bytes = Bytes.slice(4); 842 Address.Address += 4; 843 } 844 } 845 }; 846 HexagonPrettyPrinter HexagonPrettyPrinterInst; 847 848 class AMDGCNPrettyPrinter : public PrettyPrinter { 849 public: 850 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 851 object::SectionedAddress Address, raw_ostream &OS, 852 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 853 StringRef ObjectFilename, 854 std::vector<RelocationRef> *Rels) override { 855 if (SP && (PrintSource || PrintLines)) 856 SP->printSourceLine(OS, Address, ObjectFilename); 857 858 if (MI) { 859 SmallString<40> InstStr; 860 raw_svector_ostream IS(InstStr); 861 862 IP.printInst(MI, Address.Address, "", STI, IS); 863 864 OS << left_justify(IS.str(), 60); 865 } else { 866 // an unrecognized encoding - this is probably data so represent it 867 // using the .long directive, or .byte directive if fewer than 4 bytes 868 // remaining 869 if (Bytes.size() >= 4) { 870 OS << format("\t.long 0x%08" PRIx32 " ", 871 support::endian::read32<support::little>(Bytes.data())); 872 OS.indent(42); 873 } else { 874 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 875 for (unsigned int i = 1; i < Bytes.size(); i++) 876 OS << format(", 0x%02" PRIx8, Bytes[i]); 877 OS.indent(55 - (6 * Bytes.size())); 878 } 879 } 880 881 OS << format("// %012" PRIX64 ":", Address.Address); 882 if (Bytes.size() >= 4) { 883 // D should be casted to uint32_t here as it is passed by format to 884 // snprintf as vararg. 885 for (uint32_t D : makeArrayRef( 886 reinterpret_cast<const support::little32_t *>(Bytes.data()), 887 Bytes.size() / 4)) 888 OS << format(" %08" PRIX32, D); 889 } else { 890 for (unsigned char B : Bytes) 891 OS << format(" %02" PRIX8, B); 892 } 893 894 if (!Annot.empty()) 895 OS << " // " << Annot; 896 } 897 }; 898 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 899 900 class BPFPrettyPrinter : public PrettyPrinter { 901 public: 902 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 903 object::SectionedAddress Address, raw_ostream &OS, 904 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 905 StringRef ObjectFilename, 906 std::vector<RelocationRef> *Rels) override { 907 if (SP && (PrintSource || PrintLines)) 908 SP->printSourceLine(OS, Address, ObjectFilename); 909 if (!NoLeadingAddr) 910 OS << format("%8" PRId64 ":", Address.Address / 8); 911 if (!NoShowRawInsn) { 912 OS << "\t"; 913 dumpBytes(Bytes, OS); 914 } 915 if (MI) 916 IP.printInst(MI, Address.Address, "", STI, OS); 917 else 918 OS << "\t<unknown>"; 919 } 920 }; 921 BPFPrettyPrinter BPFPrettyPrinterInst; 922 923 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 924 switch(Triple.getArch()) { 925 default: 926 return PrettyPrinterInst; 927 case Triple::hexagon: 928 return HexagonPrettyPrinterInst; 929 case Triple::amdgcn: 930 return AMDGCNPrettyPrinterInst; 931 case Triple::bpfel: 932 case Triple::bpfeb: 933 return BPFPrettyPrinterInst; 934 } 935 } 936 } 937 938 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 939 assert(Obj->isELF()); 940 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 941 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 942 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 943 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 944 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 945 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 946 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 947 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 948 llvm_unreachable("Unsupported binary format"); 949 } 950 951 template <class ELFT> static void 952 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 953 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 954 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 955 uint8_t SymbolType = Symbol.getELFType(); 956 if (SymbolType == ELF::STT_SECTION) 957 continue; 958 959 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 960 // ELFSymbolRef::getAddress() returns size instead of value for common 961 // symbols which is not desirable for disassembly output. Overriding. 962 if (SymbolType == ELF::STT_COMMON) 963 Address = Obj->getSymbol(Symbol.getRawDataRefImpl())->st_value; 964 965 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 966 if (Name.empty()) 967 continue; 968 969 section_iterator SecI = 970 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 971 if (SecI == Obj->section_end()) 972 continue; 973 974 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 975 } 976 } 977 978 static void 979 addDynamicElfSymbols(const ObjectFile *Obj, 980 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 981 assert(Obj->isELF()); 982 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 983 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 984 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 985 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 986 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 987 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 988 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 989 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 990 else 991 llvm_unreachable("Unsupported binary format"); 992 } 993 994 static void addPltEntries(const ObjectFile *Obj, 995 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 996 StringSaver &Saver) { 997 Optional<SectionRef> Plt = None; 998 for (const SectionRef &Section : Obj->sections()) { 999 Expected<StringRef> SecNameOrErr = Section.getName(); 1000 if (!SecNameOrErr) { 1001 consumeError(SecNameOrErr.takeError()); 1002 continue; 1003 } 1004 if (*SecNameOrErr == ".plt") 1005 Plt = Section; 1006 } 1007 if (!Plt) 1008 return; 1009 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1010 for (auto PltEntry : ElfObj->getPltAddresses()) { 1011 SymbolRef Symbol(PltEntry.first, ElfObj); 1012 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1013 1014 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 1015 if (!Name.empty()) 1016 AllSymbols[*Plt].emplace_back( 1017 PltEntry.second, Saver.save((Name + "@plt").str()), SymbolType); 1018 } 1019 } 1020 } 1021 1022 // Normally the disassembly output will skip blocks of zeroes. This function 1023 // returns the number of zero bytes that can be skipped when dumping the 1024 // disassembly of the instructions in Buf. 1025 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 1026 // Find the number of leading zeroes. 1027 size_t N = 0; 1028 while (N < Buf.size() && !Buf[N]) 1029 ++N; 1030 1031 // We may want to skip blocks of zero bytes, but unless we see 1032 // at least 8 of them in a row. 1033 if (N < 8) 1034 return 0; 1035 1036 // We skip zeroes in multiples of 4 because do not want to truncate an 1037 // instruction if it starts with a zero byte. 1038 return N & ~0x3; 1039 } 1040 1041 // Returns a map from sections to their relocations. 1042 static std::map<SectionRef, std::vector<RelocationRef>> 1043 getRelocsMap(object::ObjectFile const &Obj) { 1044 std::map<SectionRef, std::vector<RelocationRef>> Ret; 1045 uint64_t I = (uint64_t)-1; 1046 for (SectionRef Sec : Obj.sections()) { 1047 ++I; 1048 Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection(); 1049 if (!RelocatedOrErr) 1050 reportError(Obj.getFileName(), 1051 "section (" + Twine(I) + 1052 "): failed to get a relocated section: " + 1053 toString(RelocatedOrErr.takeError())); 1054 1055 section_iterator Relocated = *RelocatedOrErr; 1056 if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep) 1057 continue; 1058 std::vector<RelocationRef> &V = Ret[*Relocated]; 1059 for (const RelocationRef &R : Sec.relocations()) 1060 V.push_back(R); 1061 // Sort relocations by address. 1062 llvm::stable_sort(V, isRelocAddressLess); 1063 } 1064 return Ret; 1065 } 1066 1067 // Used for --adjust-vma to check if address should be adjusted by the 1068 // specified value for a given section. 1069 // For ELF we do not adjust non-allocatable sections like debug ones, 1070 // because they are not loadable. 1071 // TODO: implement for other file formats. 1072 static bool shouldAdjustVA(const SectionRef &Section) { 1073 const ObjectFile *Obj = Section.getObject(); 1074 if (Obj->isELF()) 1075 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 1076 return false; 1077 } 1078 1079 1080 typedef std::pair<uint64_t, char> MappingSymbolPair; 1081 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols, 1082 uint64_t Address) { 1083 auto It = 1084 partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) { 1085 return Val.first <= Address; 1086 }); 1087 // Return zero for any address before the first mapping symbol; this means 1088 // we should use the default disassembly mode, depending on the target. 1089 if (It == MappingSymbols.begin()) 1090 return '\x00'; 1091 return (It - 1)->second; 1092 } 1093 1094 static uint64_t dumpARMELFData(uint64_t SectionAddr, uint64_t Index, 1095 uint64_t End, const ObjectFile *Obj, 1096 ArrayRef<uint8_t> Bytes, 1097 ArrayRef<MappingSymbolPair> MappingSymbols) { 1098 support::endianness Endian = 1099 Obj->isLittleEndian() ? support::little : support::big; 1100 outs() << format("%8" PRIx64 ":\t", SectionAddr + Index); 1101 if (Index + 4 <= End) { 1102 dumpBytes(Bytes.slice(Index, 4), outs()); 1103 outs() << "\t.word\t" 1104 << format_hex(support::endian::read32(Bytes.data() + Index, Endian), 1105 10); 1106 return 4; 1107 } 1108 if (Index + 2 <= End) { 1109 dumpBytes(Bytes.slice(Index, 2), outs()); 1110 outs() << "\t\t.short\t" 1111 << format_hex(support::endian::read16(Bytes.data() + Index, Endian), 1112 6); 1113 return 2; 1114 } 1115 dumpBytes(Bytes.slice(Index, 1), outs()); 1116 outs() << "\t\t.byte\t" << format_hex(Bytes[0], 4); 1117 return 1; 1118 } 1119 1120 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 1121 ArrayRef<uint8_t> Bytes) { 1122 // print out data up to 8 bytes at a time in hex and ascii 1123 uint8_t AsciiData[9] = {'\0'}; 1124 uint8_t Byte; 1125 int NumBytes = 0; 1126 1127 for (; Index < End; ++Index) { 1128 if (NumBytes == 0) 1129 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1130 Byte = Bytes.slice(Index)[0]; 1131 outs() << format(" %02x", Byte); 1132 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1133 1134 uint8_t IndentOffset = 0; 1135 NumBytes++; 1136 if (Index == End - 1 || NumBytes > 8) { 1137 // Indent the space for less than 8 bytes data. 1138 // 2 spaces for byte and one for space between bytes 1139 IndentOffset = 3 * (8 - NumBytes); 1140 for (int Excess = NumBytes; Excess < 8; Excess++) 1141 AsciiData[Excess] = '\0'; 1142 NumBytes = 8; 1143 } 1144 if (NumBytes == 8) { 1145 AsciiData[8] = '\0'; 1146 outs() << std::string(IndentOffset, ' ') << " "; 1147 outs() << reinterpret_cast<char *>(AsciiData); 1148 outs() << '\n'; 1149 NumBytes = 0; 1150 } 1151 } 1152 } 1153 1154 SymbolInfoTy createSymbolInfo(const ObjectFile *Obj, const SymbolRef &Symbol) { 1155 const StringRef FileName = Obj->getFileName(); 1156 const uint64_t Addr = unwrapOrError(Symbol.getAddress(), FileName); 1157 const StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1158 1159 if (Obj->isXCOFF() && SymbolDescription) { 1160 const auto *XCOFFObj = cast<XCOFFObjectFile>(Obj); 1161 DataRefImpl SymbolDRI = Symbol.getRawDataRefImpl(); 1162 1163 const uint32_t SymbolIndex = XCOFFObj->getSymbolIndex(SymbolDRI.p); 1164 Optional<XCOFF::StorageMappingClass> Smc = 1165 getXCOFFSymbolCsectSMC(XCOFFObj, Symbol); 1166 return SymbolInfoTy(Addr, Name, Smc, SymbolIndex, 1167 isLabel(XCOFFObj, Symbol)); 1168 } else 1169 return SymbolInfoTy(Addr, Name, 1170 Obj->isELF() ? getElfSymbolType(Obj, Symbol) 1171 : (uint8_t)ELF::STT_NOTYPE); 1172 } 1173 1174 SymbolInfoTy createDummySymbolInfo(const ObjectFile *Obj, const uint64_t Addr, 1175 StringRef &Name, uint8_t Type) { 1176 if (Obj->isXCOFF() && SymbolDescription) 1177 return SymbolInfoTy(Addr, Name, None, None, false); 1178 else 1179 return SymbolInfoTy(Addr, Name, Type); 1180 } 1181 1182 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 1183 MCContext &Ctx, MCDisassembler *PrimaryDisAsm, 1184 MCDisassembler *SecondaryDisAsm, 1185 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 1186 const MCSubtargetInfo *PrimarySTI, 1187 const MCSubtargetInfo *SecondarySTI, 1188 PrettyPrinter &PIP, 1189 SourcePrinter &SP, bool InlineRelocs) { 1190 const MCSubtargetInfo *STI = PrimarySTI; 1191 MCDisassembler *DisAsm = PrimaryDisAsm; 1192 bool PrimaryIsThumb = false; 1193 if (isArmElf(Obj)) 1194 PrimaryIsThumb = STI->checkFeatures("+thumb-mode"); 1195 1196 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 1197 if (InlineRelocs) 1198 RelocMap = getRelocsMap(*Obj); 1199 bool Is64Bits = Obj->getBytesInAddress() > 4; 1200 1201 // Create a mapping from virtual address to symbol name. This is used to 1202 // pretty print the symbols while disassembling. 1203 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1204 SectionSymbolsTy AbsoluteSymbols; 1205 const StringRef FileName = Obj->getFileName(); 1206 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj); 1207 for (const SymbolRef &Symbol : Obj->symbols()) { 1208 StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1209 if (Name.empty() && !(Obj->isXCOFF() && SymbolDescription)) 1210 continue; 1211 1212 if (Obj->isELF() && getElfSymbolType(Obj, Symbol) == ELF::STT_SECTION) 1213 continue; 1214 1215 // Don't ask a Mach-O STAB symbol for its section unless you know that 1216 // STAB symbol's section field refers to a valid section index. Otherwise 1217 // the symbol may error trying to load a section that does not exist. 1218 if (MachO) { 1219 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1220 uint8_t NType = (MachO->is64Bit() ? 1221 MachO->getSymbol64TableEntry(SymDRI).n_type: 1222 MachO->getSymbolTableEntry(SymDRI).n_type); 1223 if (NType & MachO::N_STAB) 1224 continue; 1225 } 1226 1227 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1228 if (SecI != Obj->section_end()) 1229 AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol)); 1230 else 1231 AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol)); 1232 } 1233 1234 if (AllSymbols.empty() && Obj->isELF()) 1235 addDynamicElfSymbols(Obj, AllSymbols); 1236 1237 BumpPtrAllocator A; 1238 StringSaver Saver(A); 1239 addPltEntries(Obj, AllSymbols, Saver); 1240 1241 // Create a mapping from virtual address to section. An empty section can 1242 // cause more than one section at the same address. Sort such sections to be 1243 // before same-addressed non-empty sections so that symbol lookups prefer the 1244 // non-empty section. 1245 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1246 for (SectionRef Sec : Obj->sections()) 1247 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1248 llvm::stable_sort(SectionAddresses, [](const auto &LHS, const auto &RHS) { 1249 if (LHS.first != RHS.first) 1250 return LHS.first < RHS.first; 1251 return LHS.second.getSize() < RHS.second.getSize(); 1252 }); 1253 1254 // Linked executables (.exe and .dll files) typically don't include a real 1255 // symbol table but they might contain an export table. 1256 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1257 for (const auto &ExportEntry : COFFObj->export_directories()) { 1258 StringRef Name; 1259 if (std::error_code EC = ExportEntry.getSymbolName(Name)) 1260 reportError(errorCodeToError(EC), Obj->getFileName()); 1261 if (Name.empty()) 1262 continue; 1263 1264 uint32_t RVA; 1265 if (std::error_code EC = ExportEntry.getExportRVA(RVA)) 1266 reportError(errorCodeToError(EC), Obj->getFileName()); 1267 1268 uint64_t VA = COFFObj->getImageBase() + RVA; 1269 auto Sec = partition_point( 1270 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) { 1271 return O.first <= VA; 1272 }); 1273 if (Sec != SectionAddresses.begin()) { 1274 --Sec; 1275 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1276 } else 1277 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1278 } 1279 } 1280 1281 // Sort all the symbols, this allows us to use a simple binary search to find 1282 // Multiple symbols can have the same address. Use a stable sort to stabilize 1283 // the output. 1284 StringSet<> FoundDisasmSymbolSet; 1285 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1286 stable_sort(SecSyms.second); 1287 stable_sort(AbsoluteSymbols); 1288 1289 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1290 if (FilterSections.empty() && !DisassembleAll && 1291 (!Section.isText() || Section.isVirtual())) 1292 continue; 1293 1294 uint64_t SectionAddr = Section.getAddress(); 1295 uint64_t SectSize = Section.getSize(); 1296 if (!SectSize) 1297 continue; 1298 1299 // Get the list of all the symbols in this section. 1300 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1301 std::vector<MappingSymbolPair> MappingSymbols; 1302 if (hasMappingSymbols(Obj)) { 1303 for (const auto &Symb : Symbols) { 1304 uint64_t Address = Symb.Addr; 1305 StringRef Name = Symb.Name; 1306 if (Name.startswith("$d")) 1307 MappingSymbols.emplace_back(Address - SectionAddr, 'd'); 1308 if (Name.startswith("$x")) 1309 MappingSymbols.emplace_back(Address - SectionAddr, 'x'); 1310 if (Name.startswith("$a")) 1311 MappingSymbols.emplace_back(Address - SectionAddr, 'a'); 1312 if (Name.startswith("$t")) 1313 MappingSymbols.emplace_back(Address - SectionAddr, 't'); 1314 } 1315 } 1316 1317 llvm::sort(MappingSymbols); 1318 1319 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1320 // AMDGPU disassembler uses symbolizer for printing labels 1321 std::unique_ptr<MCRelocationInfo> RelInfo( 1322 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1323 if (RelInfo) { 1324 std::unique_ptr<MCSymbolizer> Symbolizer( 1325 TheTarget->createMCSymbolizer( 1326 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1327 DisAsm->setSymbolizer(std::move(Symbolizer)); 1328 } 1329 } 1330 1331 StringRef SegmentName = ""; 1332 if (MachO) { 1333 DataRefImpl DR = Section.getRawDataRefImpl(); 1334 SegmentName = MachO->getSectionFinalSegmentName(DR); 1335 } 1336 1337 StringRef SectionName = unwrapOrError(Section.getName(), Obj->getFileName()); 1338 // If the section has no symbol at the start, just insert a dummy one. 1339 if (Symbols.empty() || Symbols[0].Addr != 0) { 1340 Symbols.insert(Symbols.begin(), 1341 createDummySymbolInfo(Obj, SectionAddr, SectionName, 1342 Section.isText() ? ELF::STT_FUNC 1343 : ELF::STT_OBJECT)); 1344 } 1345 1346 SmallString<40> Comments; 1347 raw_svector_ostream CommentStream(Comments); 1348 1349 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef( 1350 unwrapOrError(Section.getContents(), Obj->getFileName())); 1351 1352 uint64_t VMAAdjustment = 0; 1353 if (shouldAdjustVA(Section)) 1354 VMAAdjustment = AdjustVMA; 1355 1356 uint64_t Size; 1357 uint64_t Index; 1358 bool PrintedSection = false; 1359 std::vector<RelocationRef> Rels = RelocMap[Section]; 1360 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1361 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1362 // Disassemble symbol by symbol. 1363 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1364 std::string SymbolName = Symbols[SI].Name.str(); 1365 if (Demangle) 1366 SymbolName = demangle(SymbolName); 1367 1368 // Skip if --disassemble-symbols is not empty and the symbol is not in 1369 // the list. 1370 if (!DisasmSymbolSet.empty() && !DisasmSymbolSet.count(SymbolName)) 1371 continue; 1372 1373 uint64_t Start = Symbols[SI].Addr; 1374 if (Start < SectionAddr || StopAddress <= Start) 1375 continue; 1376 else 1377 FoundDisasmSymbolSet.insert(SymbolName); 1378 1379 // The end is the section end, the beginning of the next symbol, or 1380 // --stop-address. 1381 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress); 1382 if (SI + 1 < SE) 1383 End = std::min(End, Symbols[SI + 1].Addr); 1384 if (Start >= End || End <= StartAddress) 1385 continue; 1386 Start -= SectionAddr; 1387 End -= SectionAddr; 1388 1389 if (!PrintedSection) { 1390 PrintedSection = true; 1391 outs() << "\nDisassembly of section "; 1392 if (!SegmentName.empty()) 1393 outs() << SegmentName << ","; 1394 outs() << SectionName << ":\n"; 1395 } 1396 1397 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1398 if (Symbols[SI].Type == ELF::STT_AMDGPU_HSA_KERNEL) { 1399 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1400 Start += 256; 1401 } 1402 if (SI == SE - 1 || 1403 Symbols[SI + 1].Type == ELF::STT_AMDGPU_HSA_KERNEL) { 1404 // cut trailing zeroes at the end of kernel 1405 // cut up to 256 bytes 1406 const uint64_t EndAlign = 256; 1407 const auto Limit = End - (std::min)(EndAlign, End - Start); 1408 while (End > Limit && 1409 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1410 End -= 4; 1411 } 1412 } 1413 1414 outs() << '\n'; 1415 if (!NoLeadingAddr) 1416 outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ", 1417 SectionAddr + Start + VMAAdjustment); 1418 if (Obj->isXCOFF() && SymbolDescription) { 1419 outs() << getXCOFFSymbolDescription(Symbols[SI], SymbolName) << ":\n"; 1420 } else 1421 outs() << '<' << SymbolName << ">:\n"; 1422 1423 // Don't print raw contents of a virtual section. A virtual section 1424 // doesn't have any contents in the file. 1425 if (Section.isVirtual()) { 1426 outs() << "...\n"; 1427 continue; 1428 } 1429 1430 // Some targets (like WebAssembly) have a special prelude at the start 1431 // of each symbol. 1432 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1433 SectionAddr + Start, CommentStream); 1434 Start += Size; 1435 1436 Index = Start; 1437 if (SectionAddr < StartAddress) 1438 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1439 1440 // If there is a data/common symbol inside an ELF text section and we are 1441 // only disassembling text (applicable all architectures), we are in a 1442 // situation where we must print the data and not disassemble it. 1443 if (Obj->isELF() && !DisassembleAll && Section.isText()) { 1444 uint8_t SymTy = Symbols[SI].Type; 1445 if (SymTy == ELF::STT_OBJECT || SymTy == ELF::STT_COMMON) { 1446 dumpELFData(SectionAddr, Index, End, Bytes); 1447 Index = End; 1448 } 1449 } 1450 1451 bool CheckARMELFData = hasMappingSymbols(Obj) && 1452 Symbols[SI].Type != ELF::STT_OBJECT && 1453 !DisassembleAll; 1454 bool DumpARMELFData = false; 1455 while (Index < End) { 1456 // ARM and AArch64 ELF binaries can interleave data and text in the 1457 // same section. We rely on the markers introduced to understand what 1458 // we need to dump. If the data marker is within a function, it is 1459 // denoted as a word/short etc. 1460 if (CheckARMELFData) { 1461 char Kind = getMappingSymbolKind(MappingSymbols, Index); 1462 DumpARMELFData = Kind == 'd'; 1463 if (SecondarySTI) { 1464 if (Kind == 'a') { 1465 STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI; 1466 DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm; 1467 } else if (Kind == 't') { 1468 STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI; 1469 DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm; 1470 } 1471 } 1472 } 1473 1474 if (DumpARMELFData) { 1475 Size = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1476 MappingSymbols); 1477 } else { 1478 // When -z or --disassemble-zeroes are given we always dissasemble 1479 // them. Otherwise we might want to skip zero bytes we see. 1480 if (!DisassembleZeroes) { 1481 uint64_t MaxOffset = End - Index; 1482 // For --reloc: print zero blocks patched by relocations, so that 1483 // relocations can be shown in the dump. 1484 if (RelCur != RelEnd) 1485 MaxOffset = RelCur->getOffset() - Index; 1486 1487 if (size_t N = 1488 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1489 outs() << "\t\t..." << '\n'; 1490 Index += N; 1491 continue; 1492 } 1493 } 1494 1495 // Disassemble a real instruction or a data when disassemble all is 1496 // provided 1497 MCInst Inst; 1498 bool Disassembled = 1499 DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1500 SectionAddr + Index, CommentStream); 1501 if (Size == 0) 1502 Size = 1; 1503 1504 PIP.printInst( 1505 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 1506 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, outs(), 1507 "", *STI, &SP, Obj->getFileName(), &Rels); 1508 outs() << CommentStream.str(); 1509 Comments.clear(); 1510 1511 // If disassembly has failed, avoid analysing invalid/incomplete 1512 // instruction information. Otherwise, try to resolve the target 1513 // address (jump target or memory operand address) and print it on the 1514 // right of the instruction. 1515 if (Disassembled && MIA) { 1516 uint64_t Target; 1517 bool PrintTarget = 1518 MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target); 1519 if (!PrintTarget) 1520 if (Optional<uint64_t> MaybeTarget = 1521 MIA->evaluateMemoryOperandAddress( 1522 Inst, SectionAddr + Index, Size)) { 1523 Target = *MaybeTarget; 1524 PrintTarget = true; 1525 outs() << " # " << Twine::utohexstr(Target); 1526 } 1527 if (PrintTarget) { 1528 // In a relocatable object, the target's section must reside in 1529 // the same section as the call instruction or it is accessed 1530 // through a relocation. 1531 // 1532 // In a non-relocatable object, the target may be in any section. 1533 // In that case, locate the section(s) containing the target 1534 // address and find the symbol in one of those, if possible. 1535 // 1536 // N.B. We don't walk the relocations in the relocatable case yet. 1537 std::vector<const SectionSymbolsTy *> TargetSectionSymbols; 1538 if (!Obj->isRelocatableObject()) { 1539 auto It = llvm::partition_point( 1540 SectionAddresses, 1541 [=](const std::pair<uint64_t, SectionRef> &O) { 1542 return O.first <= Target; 1543 }); 1544 uint64_t TargetSecAddr = 0; 1545 while (It != SectionAddresses.begin()) { 1546 --It; 1547 if (TargetSecAddr == 0) 1548 TargetSecAddr = It->first; 1549 if (It->first != TargetSecAddr) 1550 break; 1551 TargetSectionSymbols.push_back(&AllSymbols[It->second]); 1552 } 1553 } else { 1554 TargetSectionSymbols.push_back(&Symbols); 1555 } 1556 TargetSectionSymbols.push_back(&AbsoluteSymbols); 1557 1558 // Find the last symbol in the first candidate section whose 1559 // offset is less than or equal to the target. If there are no 1560 // such symbols, try in the next section and so on, before finally 1561 // using the nearest preceding absolute symbol (if any), if there 1562 // are no other valid symbols. 1563 const SymbolInfoTy *TargetSym = nullptr; 1564 for (const SectionSymbolsTy *TargetSymbols : 1565 TargetSectionSymbols) { 1566 auto It = llvm::partition_point( 1567 *TargetSymbols, 1568 [=](const SymbolInfoTy &O) { return O.Addr <= Target; }); 1569 if (It != TargetSymbols->begin()) { 1570 TargetSym = &*(It - 1); 1571 break; 1572 } 1573 } 1574 1575 if (TargetSym != nullptr) { 1576 uint64_t TargetAddress = TargetSym->Addr; 1577 std::string TargetName = TargetSym->Name.str(); 1578 if (Demangle) 1579 TargetName = demangle(TargetName); 1580 1581 outs() << " <" << TargetName; 1582 uint64_t Disp = Target - TargetAddress; 1583 if (Disp) 1584 outs() << "+0x" << Twine::utohexstr(Disp); 1585 outs() << '>'; 1586 } 1587 } 1588 } 1589 } 1590 outs() << "\n"; 1591 1592 // Hexagon does this in pretty printer 1593 if (Obj->getArch() != Triple::hexagon) { 1594 // Print relocation for instruction and data. 1595 while (RelCur != RelEnd) { 1596 uint64_t Offset = RelCur->getOffset(); 1597 // If this relocation is hidden, skip it. 1598 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 1599 ++RelCur; 1600 continue; 1601 } 1602 1603 // Stop when RelCur's offset is past the disassembled 1604 // instruction/data. Note that it's possible the disassembled data 1605 // is not the complete data: we might see the relocation printed in 1606 // the middle of the data, but this matches the binutils objdump 1607 // output. 1608 if (Offset >= Index + Size) 1609 break; 1610 1611 // When --adjust-vma is used, update the address printed. 1612 if (RelCur->getSymbol() != Obj->symbol_end()) { 1613 Expected<section_iterator> SymSI = 1614 RelCur->getSymbol()->getSection(); 1615 if (SymSI && *SymSI != Obj->section_end() && 1616 shouldAdjustVA(**SymSI)) 1617 Offset += AdjustVMA; 1618 } 1619 1620 printRelocation(Obj->getFileName(), *RelCur, SectionAddr + Offset, 1621 Is64Bits); 1622 ++RelCur; 1623 } 1624 } 1625 1626 Index += Size; 1627 } 1628 } 1629 } 1630 StringSet<> MissingDisasmSymbolSet = 1631 set_difference(DisasmSymbolSet, FoundDisasmSymbolSet); 1632 for (StringRef Sym : MissingDisasmSymbolSet.keys()) 1633 reportWarning("failed to disassemble missing symbol " + Sym, FileName); 1634 } 1635 1636 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1637 const Target *TheTarget = getTarget(Obj); 1638 1639 // Package up features to be passed to target/subtarget 1640 SubtargetFeatures Features = Obj->getFeatures(); 1641 if (!MAttrs.empty()) 1642 for (unsigned I = 0; I != MAttrs.size(); ++I) 1643 Features.AddFeature(MAttrs[I]); 1644 1645 std::unique_ptr<const MCRegisterInfo> MRI( 1646 TheTarget->createMCRegInfo(TripleName)); 1647 if (!MRI) 1648 reportError(Obj->getFileName(), 1649 "no register info for target " + TripleName); 1650 1651 // Set up disassembler. 1652 MCTargetOptions MCOptions; 1653 std::unique_ptr<const MCAsmInfo> AsmInfo( 1654 TheTarget->createMCAsmInfo(*MRI, TripleName, MCOptions)); 1655 if (!AsmInfo) 1656 reportError(Obj->getFileName(), 1657 "no assembly info for target " + TripleName); 1658 std::unique_ptr<const MCSubtargetInfo> STI( 1659 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1660 if (!STI) 1661 reportError(Obj->getFileName(), 1662 "no subtarget info for target " + TripleName); 1663 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1664 if (!MII) 1665 reportError(Obj->getFileName(), 1666 "no instruction info for target " + TripleName); 1667 MCObjectFileInfo MOFI; 1668 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1669 // FIXME: for now initialize MCObjectFileInfo with default values 1670 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1671 1672 std::unique_ptr<MCDisassembler> DisAsm( 1673 TheTarget->createMCDisassembler(*STI, Ctx)); 1674 if (!DisAsm) 1675 reportError(Obj->getFileName(), "no disassembler for target " + TripleName); 1676 1677 // If we have an ARM object file, we need a second disassembler, because 1678 // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode. 1679 // We use mapping symbols to switch between the two assemblers, where 1680 // appropriate. 1681 std::unique_ptr<MCDisassembler> SecondaryDisAsm; 1682 std::unique_ptr<const MCSubtargetInfo> SecondarySTI; 1683 if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) { 1684 if (STI->checkFeatures("+thumb-mode")) 1685 Features.AddFeature("-thumb-mode"); 1686 else 1687 Features.AddFeature("+thumb-mode"); 1688 SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU, 1689 Features.getString())); 1690 SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx)); 1691 } 1692 1693 std::unique_ptr<const MCInstrAnalysis> MIA( 1694 TheTarget->createMCInstrAnalysis(MII.get())); 1695 1696 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1697 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1698 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1699 if (!IP) 1700 reportError(Obj->getFileName(), 1701 "no instruction printer for target " + TripleName); 1702 IP->setPrintImmHex(PrintImmHex); 1703 IP->setPrintBranchImmAsAddress(true); 1704 1705 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1706 SourcePrinter SP(Obj, TheTarget->getName()); 1707 1708 for (StringRef Opt : DisassemblerOptions) 1709 if (!IP->applyTargetSpecificCLOption(Opt)) 1710 reportError(Obj->getFileName(), 1711 "Unrecognized disassembler option: " + Opt); 1712 1713 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(), 1714 MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP, 1715 SP, InlineRelocs); 1716 } 1717 1718 void printRelocations(const ObjectFile *Obj) { 1719 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1720 "%08" PRIx64; 1721 // Regular objdump doesn't print relocations in non-relocatable object 1722 // files. 1723 if (!Obj->isRelocatableObject()) 1724 return; 1725 1726 // Build a mapping from relocation target to a vector of relocation 1727 // sections. Usually, there is an only one relocation section for 1728 // each relocated section. 1729 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec; 1730 uint64_t Ndx; 1731 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Ndx)) { 1732 if (Section.relocation_begin() == Section.relocation_end()) 1733 continue; 1734 Expected<section_iterator> SecOrErr = Section.getRelocatedSection(); 1735 if (!SecOrErr) 1736 reportError(Obj->getFileName(), 1737 "section (" + Twine(Ndx) + 1738 "): unable to get a relocation target: " + 1739 toString(SecOrErr.takeError())); 1740 SecToRelSec[**SecOrErr].push_back(Section); 1741 } 1742 1743 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) { 1744 StringRef SecName = unwrapOrError(P.first.getName(), Obj->getFileName()); 1745 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1746 uint32_t OffsetPadding = (Obj->getBytesInAddress() > 4 ? 16 : 8); 1747 uint32_t TypePadding = 24; 1748 outs() << left_justify("OFFSET", OffsetPadding) << " " 1749 << left_justify("TYPE", TypePadding) << " " 1750 << "VALUE\n"; 1751 1752 for (SectionRef Section : P.second) { 1753 for (const RelocationRef &Reloc : Section.relocations()) { 1754 uint64_t Address = Reloc.getOffset(); 1755 SmallString<32> RelocName; 1756 SmallString<32> ValueStr; 1757 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1758 continue; 1759 Reloc.getTypeName(RelocName); 1760 if (Error E = getRelocationValueString(Reloc, ValueStr)) 1761 reportError(std::move(E), Obj->getFileName()); 1762 1763 outs() << format(Fmt.data(), Address) << " " 1764 << left_justify(RelocName, TypePadding) << " " << ValueStr 1765 << "\n"; 1766 } 1767 } 1768 outs() << "\n"; 1769 } 1770 } 1771 1772 void printDynamicRelocations(const ObjectFile *Obj) { 1773 // For the moment, this option is for ELF only 1774 if (!Obj->isELF()) 1775 return; 1776 1777 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1778 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1779 reportError(Obj->getFileName(), "not a dynamic object"); 1780 return; 1781 } 1782 1783 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1784 if (DynRelSec.empty()) 1785 return; 1786 1787 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1788 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1789 for (const SectionRef &Section : DynRelSec) 1790 for (const RelocationRef &Reloc : Section.relocations()) { 1791 uint64_t Address = Reloc.getOffset(); 1792 SmallString<32> RelocName; 1793 SmallString<32> ValueStr; 1794 Reloc.getTypeName(RelocName); 1795 if (Error E = getRelocationValueString(Reloc, ValueStr)) 1796 reportError(std::move(E), Obj->getFileName()); 1797 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1798 << ValueStr << "\n"; 1799 } 1800 } 1801 1802 // Returns true if we need to show LMA column when dumping section headers. We 1803 // show it only when the platform is ELF and either we have at least one section 1804 // whose VMA and LMA are different and/or when --show-lma flag is used. 1805 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1806 if (!Obj->isELF()) 1807 return false; 1808 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1809 if (S.getAddress() != getELFSectionLMA(S)) 1810 return true; 1811 return ShowLMA; 1812 } 1813 1814 static size_t getMaxSectionNameWidth(const ObjectFile *Obj) { 1815 // Default column width for names is 13 even if no names are that long. 1816 size_t MaxWidth = 13; 1817 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1818 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 1819 MaxWidth = std::max(MaxWidth, Name.size()); 1820 } 1821 return MaxWidth; 1822 } 1823 1824 void printSectionHeaders(const ObjectFile *Obj) { 1825 size_t NameWidth = getMaxSectionNameWidth(Obj); 1826 size_t AddressWidth = 2 * Obj->getBytesInAddress(); 1827 bool HasLMAColumn = shouldDisplayLMA(Obj); 1828 if (HasLMAColumn) 1829 outs() << "Sections:\n" 1830 "Idx " 1831 << left_justify("Name", NameWidth) << " Size " 1832 << left_justify("VMA", AddressWidth) << " " 1833 << left_justify("LMA", AddressWidth) << " Type\n"; 1834 else 1835 outs() << "Sections:\n" 1836 "Idx " 1837 << left_justify("Name", NameWidth) << " Size " 1838 << left_justify("VMA", AddressWidth) << " Type\n"; 1839 1840 uint64_t Idx; 1841 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Idx)) { 1842 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 1843 uint64_t VMA = Section.getAddress(); 1844 if (shouldAdjustVA(Section)) 1845 VMA += AdjustVMA; 1846 1847 uint64_t Size = Section.getSize(); 1848 1849 std::string Type = Section.isText() ? "TEXT" : ""; 1850 if (Section.isData()) 1851 Type += Type.empty() ? "DATA" : " DATA"; 1852 if (Section.isBSS()) 1853 Type += Type.empty() ? "BSS" : " BSS"; 1854 1855 if (HasLMAColumn) 1856 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 1857 Name.str().c_str(), Size) 1858 << format_hex_no_prefix(VMA, AddressWidth) << " " 1859 << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth) 1860 << " " << Type << "\n"; 1861 else 1862 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 1863 Name.str().c_str(), Size) 1864 << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n"; 1865 } 1866 outs() << "\n"; 1867 } 1868 1869 void printSectionContents(const ObjectFile *Obj) { 1870 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1871 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 1872 uint64_t BaseAddr = Section.getAddress(); 1873 uint64_t Size = Section.getSize(); 1874 if (!Size) 1875 continue; 1876 1877 outs() << "Contents of section " << Name << ":\n"; 1878 if (Section.isBSS()) { 1879 outs() << format("<skipping contents of bss section at [%04" PRIx64 1880 ", %04" PRIx64 ")>\n", 1881 BaseAddr, BaseAddr + Size); 1882 continue; 1883 } 1884 1885 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName()); 1886 1887 // Dump out the content as hex and printable ascii characters. 1888 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1889 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1890 // Dump line of hex. 1891 for (std::size_t I = 0; I < 16; ++I) { 1892 if (I != 0 && I % 4 == 0) 1893 outs() << ' '; 1894 if (Addr + I < End) 1895 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1896 << hexdigit(Contents[Addr + I] & 0xF, true); 1897 else 1898 outs() << " "; 1899 } 1900 // Print ascii. 1901 outs() << " "; 1902 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1903 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1904 outs() << Contents[Addr + I]; 1905 else 1906 outs() << "."; 1907 } 1908 outs() << "\n"; 1909 } 1910 } 1911 } 1912 1913 void printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1914 StringRef ArchitectureName, bool DumpDynamic) { 1915 if (O->isCOFF() && !DumpDynamic) { 1916 outs() << "SYMBOL TABLE:\n"; 1917 printCOFFSymbolTable(cast<const COFFObjectFile>(O)); 1918 return; 1919 } 1920 1921 const StringRef FileName = O->getFileName(); 1922 1923 if (!DumpDynamic) { 1924 outs() << "SYMBOL TABLE:\n"; 1925 for (auto I = O->symbol_begin(); I != O->symbol_end(); ++I) 1926 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 1927 return; 1928 } 1929 1930 outs() << "DYNAMIC SYMBOL TABLE:\n"; 1931 if (!O->isELF()) { 1932 reportWarning( 1933 "this operation is not currently supported for this file format", 1934 FileName); 1935 return; 1936 } 1937 1938 const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(O); 1939 for (auto I = ELF->getDynamicSymbolIterators().begin(); 1940 I != ELF->getDynamicSymbolIterators().end(); ++I) 1941 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 1942 } 1943 1944 void printSymbol(const ObjectFile *O, const SymbolRef &Symbol, 1945 StringRef FileName, StringRef ArchiveName, 1946 StringRef ArchitectureName, bool DumpDynamic) { 1947 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(O); 1948 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName, ArchiveName, 1949 ArchitectureName); 1950 if ((Address < StartAddress) || (Address > StopAddress)) 1951 return; 1952 SymbolRef::Type Type = 1953 unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName); 1954 uint32_t Flags = 1955 unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName); 1956 1957 // Don't ask a Mach-O STAB symbol for its section unless you know that 1958 // STAB symbol's section field refers to a valid section index. Otherwise 1959 // the symbol may error trying to load a section that does not exist. 1960 bool IsSTAB = false; 1961 if (MachO) { 1962 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1963 uint8_t NType = 1964 (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type 1965 : MachO->getSymbolTableEntry(SymDRI).n_type); 1966 if (NType & MachO::N_STAB) 1967 IsSTAB = true; 1968 } 1969 section_iterator Section = IsSTAB 1970 ? O->section_end() 1971 : unwrapOrError(Symbol.getSection(), FileName, 1972 ArchiveName, ArchitectureName); 1973 1974 StringRef Name; 1975 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 1976 if (Expected<StringRef> NameOrErr = Section->getName()) 1977 Name = *NameOrErr; 1978 else 1979 consumeError(NameOrErr.takeError()); 1980 1981 } else { 1982 Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName, 1983 ArchitectureName); 1984 } 1985 1986 bool Global = Flags & SymbolRef::SF_Global; 1987 bool Weak = Flags & SymbolRef::SF_Weak; 1988 bool Absolute = Flags & SymbolRef::SF_Absolute; 1989 bool Common = Flags & SymbolRef::SF_Common; 1990 bool Hidden = Flags & SymbolRef::SF_Hidden; 1991 1992 char GlobLoc = ' '; 1993 if ((Section != O->section_end() || Absolute) && !Weak) 1994 GlobLoc = Global ? 'g' : 'l'; 1995 char IFunc = ' '; 1996 if (O->isELF()) { 1997 if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC) 1998 IFunc = 'i'; 1999 if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE) 2000 GlobLoc = 'u'; 2001 } 2002 2003 char Debug = ' '; 2004 if (DumpDynamic) 2005 Debug = 'D'; 2006 else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2007 Debug = 'd'; 2008 2009 char FileFunc = ' '; 2010 if (Type == SymbolRef::ST_File) 2011 FileFunc = 'f'; 2012 else if (Type == SymbolRef::ST_Function) 2013 FileFunc = 'F'; 2014 else if (Type == SymbolRef::ST_Data) 2015 FileFunc = 'O'; 2016 2017 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2018 2019 outs() << format(Fmt, Address) << " " 2020 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2021 << (Weak ? 'w' : ' ') // Weak? 2022 << ' ' // Constructor. Not supported yet. 2023 << ' ' // Warning. Not supported yet. 2024 << IFunc // Indirect reference to another symbol. 2025 << Debug // Debugging (d) or dynamic (D) symbol. 2026 << FileFunc // Name of function (F), file (f) or object (O). 2027 << ' '; 2028 if (Absolute) { 2029 outs() << "*ABS*"; 2030 } else if (Common) { 2031 outs() << "*COM*"; 2032 } else if (Section == O->section_end()) { 2033 outs() << "*UND*"; 2034 } else { 2035 if (MachO) { 2036 DataRefImpl DR = Section->getRawDataRefImpl(); 2037 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 2038 outs() << SegmentName << ","; 2039 } 2040 StringRef SectionName = unwrapOrError(Section->getName(), FileName); 2041 outs() << SectionName; 2042 } 2043 2044 if (Common || O->isELF()) { 2045 uint64_t Val = 2046 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2047 outs() << '\t' << format(Fmt, Val); 2048 } 2049 2050 if (O->isELF()) { 2051 uint8_t Other = ELFSymbolRef(Symbol).getOther(); 2052 switch (Other) { 2053 case ELF::STV_DEFAULT: 2054 break; 2055 case ELF::STV_INTERNAL: 2056 outs() << " .internal"; 2057 break; 2058 case ELF::STV_HIDDEN: 2059 outs() << " .hidden"; 2060 break; 2061 case ELF::STV_PROTECTED: 2062 outs() << " .protected"; 2063 break; 2064 default: 2065 outs() << format(" 0x%02x", Other); 2066 break; 2067 } 2068 } else if (Hidden) { 2069 outs() << " .hidden"; 2070 } 2071 2072 if (Demangle) 2073 outs() << ' ' << demangle(std::string(Name)) << '\n'; 2074 else 2075 outs() << ' ' << Name << '\n'; 2076 } 2077 2078 static void printUnwindInfo(const ObjectFile *O) { 2079 outs() << "Unwind info:\n\n"; 2080 2081 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 2082 printCOFFUnwindInfo(Coff); 2083 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 2084 printMachOUnwindInfo(MachO); 2085 else 2086 // TODO: Extract DWARF dump tool to objdump. 2087 WithColor::error(errs(), ToolName) 2088 << "This operation is only currently supported " 2089 "for COFF and MachO object files.\n"; 2090 } 2091 2092 /// Dump the raw contents of the __clangast section so the output can be piped 2093 /// into llvm-bcanalyzer. 2094 void printRawClangAST(const ObjectFile *Obj) { 2095 if (outs().is_displayed()) { 2096 WithColor::error(errs(), ToolName) 2097 << "The -raw-clang-ast option will dump the raw binary contents of " 2098 "the clang ast section.\n" 2099 "Please redirect the output to a file or another program such as " 2100 "llvm-bcanalyzer.\n"; 2101 return; 2102 } 2103 2104 StringRef ClangASTSectionName("__clangast"); 2105 if (Obj->isCOFF()) { 2106 ClangASTSectionName = "clangast"; 2107 } 2108 2109 Optional<object::SectionRef> ClangASTSection; 2110 for (auto Sec : ToolSectionFilter(*Obj)) { 2111 StringRef Name; 2112 if (Expected<StringRef> NameOrErr = Sec.getName()) 2113 Name = *NameOrErr; 2114 else 2115 consumeError(NameOrErr.takeError()); 2116 2117 if (Name == ClangASTSectionName) { 2118 ClangASTSection = Sec; 2119 break; 2120 } 2121 } 2122 if (!ClangASTSection) 2123 return; 2124 2125 StringRef ClangASTContents = unwrapOrError( 2126 ClangASTSection.getValue().getContents(), Obj->getFileName()); 2127 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2128 } 2129 2130 static void printFaultMaps(const ObjectFile *Obj) { 2131 StringRef FaultMapSectionName; 2132 2133 if (Obj->isELF()) { 2134 FaultMapSectionName = ".llvm_faultmaps"; 2135 } else if (Obj->isMachO()) { 2136 FaultMapSectionName = "__llvm_faultmaps"; 2137 } else { 2138 WithColor::error(errs(), ToolName) 2139 << "This operation is only currently supported " 2140 "for ELF and Mach-O executable files.\n"; 2141 return; 2142 } 2143 2144 Optional<object::SectionRef> FaultMapSection; 2145 2146 for (auto Sec : ToolSectionFilter(*Obj)) { 2147 StringRef Name; 2148 if (Expected<StringRef> NameOrErr = Sec.getName()) 2149 Name = *NameOrErr; 2150 else 2151 consumeError(NameOrErr.takeError()); 2152 2153 if (Name == FaultMapSectionName) { 2154 FaultMapSection = Sec; 2155 break; 2156 } 2157 } 2158 2159 outs() << "FaultMap table:\n"; 2160 2161 if (!FaultMapSection.hasValue()) { 2162 outs() << "<not found>\n"; 2163 return; 2164 } 2165 2166 StringRef FaultMapContents = 2167 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName()); 2168 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2169 FaultMapContents.bytes_end()); 2170 2171 outs() << FMP; 2172 } 2173 2174 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 2175 if (O->isELF()) { 2176 printELFFileHeader(O); 2177 printELFDynamicSection(O); 2178 printELFSymbolVersionInfo(O); 2179 return; 2180 } 2181 if (O->isCOFF()) 2182 return printCOFFFileHeader(O); 2183 if (O->isWasm()) 2184 return printWasmFileHeader(O); 2185 if (O->isMachO()) { 2186 printMachOFileHeader(O); 2187 if (!OnlyFirst) 2188 printMachOLoadCommands(O); 2189 return; 2190 } 2191 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2192 } 2193 2194 static void printFileHeaders(const ObjectFile *O) { 2195 if (!O->isELF() && !O->isCOFF()) 2196 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2197 2198 Triple::ArchType AT = O->getArch(); 2199 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2200 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 2201 2202 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2203 outs() << "start address: " 2204 << "0x" << format(Fmt.data(), Address) << "\n\n"; 2205 } 2206 2207 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2208 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2209 if (!ModeOrErr) { 2210 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 2211 consumeError(ModeOrErr.takeError()); 2212 return; 2213 } 2214 sys::fs::perms Mode = ModeOrErr.get(); 2215 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2216 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2217 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2218 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2219 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2220 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2221 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2222 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2223 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2224 2225 outs() << " "; 2226 2227 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 2228 unwrapOrError(C.getGID(), Filename), 2229 unwrapOrError(C.getRawSize(), Filename)); 2230 2231 StringRef RawLastModified = C.getRawLastModified(); 2232 unsigned Seconds; 2233 if (RawLastModified.getAsInteger(10, Seconds)) 2234 outs() << "(date: \"" << RawLastModified 2235 << "\" contains non-decimal chars) "; 2236 else { 2237 // Since ctime(3) returns a 26 character string of the form: 2238 // "Sun Sep 16 01:03:52 1973\n\0" 2239 // just print 24 characters. 2240 time_t t = Seconds; 2241 outs() << format("%.24s ", ctime(&t)); 2242 } 2243 2244 StringRef Name = ""; 2245 Expected<StringRef> NameOrErr = C.getName(); 2246 if (!NameOrErr) { 2247 consumeError(NameOrErr.takeError()); 2248 Name = unwrapOrError(C.getRawName(), Filename); 2249 } else { 2250 Name = NameOrErr.get(); 2251 } 2252 outs() << Name << "\n"; 2253 } 2254 2255 // For ELF only now. 2256 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) { 2257 if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) { 2258 if (Elf->getEType() != ELF::ET_REL) 2259 return true; 2260 } 2261 return false; 2262 } 2263 2264 static void checkForInvalidStartStopAddress(ObjectFile *Obj, 2265 uint64_t Start, uint64_t Stop) { 2266 if (!shouldWarnForInvalidStartStopAddress(Obj)) 2267 return; 2268 2269 for (const SectionRef &Section : Obj->sections()) 2270 if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) { 2271 uint64_t BaseAddr = Section.getAddress(); 2272 uint64_t Size = Section.getSize(); 2273 if ((Start < BaseAddr + Size) && Stop > BaseAddr) 2274 return; 2275 } 2276 2277 if (StartAddress.getNumOccurrences() == 0) 2278 reportWarning("no section has address less than 0x" + 2279 Twine::utohexstr(Stop) + " specified by --stop-address", 2280 Obj->getFileName()); 2281 else if (StopAddress.getNumOccurrences() == 0) 2282 reportWarning("no section has address greater than or equal to 0x" + 2283 Twine::utohexstr(Start) + " specified by --start-address", 2284 Obj->getFileName()); 2285 else 2286 reportWarning("no section overlaps the range [0x" + 2287 Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) + 2288 ") specified by --start-address/--stop-address", 2289 Obj->getFileName()); 2290 } 2291 2292 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 2293 const Archive::Child *C = nullptr) { 2294 // Avoid other output when using a raw option. 2295 if (!RawClangAST) { 2296 outs() << '\n'; 2297 if (A) 2298 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 2299 else 2300 outs() << O->getFileName(); 2301 outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n\n"; 2302 } 2303 2304 if (StartAddress.getNumOccurrences() || StopAddress.getNumOccurrences()) 2305 checkForInvalidStartStopAddress(O, StartAddress, StopAddress); 2306 2307 // Note: the order here matches GNU objdump for compatability. 2308 StringRef ArchiveName = A ? A->getFileName() : ""; 2309 if (ArchiveHeaders && !MachOOpt && C) 2310 printArchiveChild(ArchiveName, *C); 2311 if (FileHeaders) 2312 printFileHeaders(O); 2313 if (PrivateHeaders || FirstPrivateHeader) 2314 printPrivateFileHeaders(O, FirstPrivateHeader); 2315 if (SectionHeaders) 2316 printSectionHeaders(O); 2317 if (SymbolTable) 2318 printSymbolTable(O, ArchiveName); 2319 if (DynamicSymbolTable) 2320 printSymbolTable(O, ArchiveName, /*ArchitectureName=*/"", 2321 /*DumpDynamic=*/true); 2322 if (DwarfDumpType != DIDT_Null) { 2323 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2324 // Dump the complete DWARF structure. 2325 DIDumpOptions DumpOpts; 2326 DumpOpts.DumpType = DwarfDumpType; 2327 DICtx->dump(outs(), DumpOpts); 2328 } 2329 if (Relocations && !Disassemble) 2330 printRelocations(O); 2331 if (DynamicRelocations) 2332 printDynamicRelocations(O); 2333 if (SectionContents) 2334 printSectionContents(O); 2335 if (Disassemble) 2336 disassembleObject(O, Relocations); 2337 if (UnwindInfo) 2338 printUnwindInfo(O); 2339 2340 // Mach-O specific options: 2341 if (ExportsTrie) 2342 printExportsTrie(O); 2343 if (Rebase) 2344 printRebaseTable(O); 2345 if (Bind) 2346 printBindTable(O); 2347 if (LazyBind) 2348 printLazyBindTable(O); 2349 if (WeakBind) 2350 printWeakBindTable(O); 2351 2352 // Other special sections: 2353 if (RawClangAST) 2354 printRawClangAST(O); 2355 if (FaultMapSection) 2356 printFaultMaps(O); 2357 } 2358 2359 static void dumpObject(const COFFImportFile *I, const Archive *A, 2360 const Archive::Child *C = nullptr) { 2361 StringRef ArchiveName = A ? A->getFileName() : ""; 2362 2363 // Avoid other output when using a raw option. 2364 if (!RawClangAST) 2365 outs() << '\n' 2366 << ArchiveName << "(" << I->getFileName() << ")" 2367 << ":\tfile format COFF-import-file" 2368 << "\n\n"; 2369 2370 if (ArchiveHeaders && !MachOOpt && C) 2371 printArchiveChild(ArchiveName, *C); 2372 if (SymbolTable) 2373 printCOFFSymbolTable(I); 2374 } 2375 2376 /// Dump each object file in \a a; 2377 static void dumpArchive(const Archive *A) { 2378 Error Err = Error::success(); 2379 unsigned I = -1; 2380 for (auto &C : A->children(Err)) { 2381 ++I; 2382 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2383 if (!ChildOrErr) { 2384 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2385 reportError(std::move(E), getFileNameForError(C, I), A->getFileName()); 2386 continue; 2387 } 2388 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2389 dumpObject(O, A, &C); 2390 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2391 dumpObject(I, A, &C); 2392 else 2393 reportError(errorCodeToError(object_error::invalid_file_type), 2394 A->getFileName()); 2395 } 2396 if (Err) 2397 reportError(std::move(Err), A->getFileName()); 2398 } 2399 2400 /// Open file and figure out how to dump it. 2401 static void dumpInput(StringRef file) { 2402 // If we are using the Mach-O specific object file parser, then let it parse 2403 // the file and process the command line options. So the -arch flags can 2404 // be used to select specific slices, etc. 2405 if (MachOOpt) { 2406 parseInputMachO(file); 2407 return; 2408 } 2409 2410 // Attempt to open the binary. 2411 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2412 Binary &Binary = *OBinary.getBinary(); 2413 2414 if (Archive *A = dyn_cast<Archive>(&Binary)) 2415 dumpArchive(A); 2416 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2417 dumpObject(O); 2418 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2419 parseInputMachO(UB); 2420 else 2421 reportError(errorCodeToError(object_error::invalid_file_type), file); 2422 } 2423 } // namespace llvm 2424 2425 int main(int argc, char **argv) { 2426 using namespace llvm; 2427 InitLLVM X(argc, argv); 2428 const cl::OptionCategory *OptionFilters[] = {&ObjdumpCat, &MachOCat}; 2429 cl::HideUnrelatedOptions(OptionFilters); 2430 2431 // Initialize targets and assembly printers/parsers. 2432 InitializeAllTargetInfos(); 2433 InitializeAllTargetMCs(); 2434 InitializeAllDisassemblers(); 2435 2436 // Register the target printer for --version. 2437 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2438 2439 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n", nullptr, 2440 /*EnvVar=*/nullptr, 2441 /*LongOptionsUseDoubleDash=*/true); 2442 2443 if (StartAddress >= StopAddress) 2444 reportCmdLineError("start address should be less than stop address"); 2445 2446 ToolName = argv[0]; 2447 2448 // Defaults to a.out if no filenames specified. 2449 if (InputFilenames.empty()) 2450 InputFilenames.push_back("a.out"); 2451 2452 if (AllHeaders) 2453 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2454 SectionHeaders = SymbolTable = true; 2455 2456 if (DisassembleAll || PrintSource || PrintLines || 2457 !DisassembleSymbols.empty()) 2458 Disassemble = true; 2459 2460 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null && 2461 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST && 2462 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable && 2463 !DynamicSymbolTable && !UnwindInfo && !FaultMapSection && 2464 !(MachOOpt && 2465 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie || 2466 FirstPrivateHeader || IndirectSymbols || InfoPlist || LazyBind || 2467 LinkOptHints || ObjcMetaData || Rebase || UniversalHeaders || 2468 WeakBind || !FilterSections.empty()))) { 2469 cl::PrintHelpMessage(); 2470 return 2; 2471 } 2472 2473 DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end()); 2474 2475 llvm::for_each(InputFilenames, dumpInput); 2476 2477 warnOnNoMatchForSections(); 2478 2479 return EXIT_SUCCESS; 2480 } 2481