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 "MachODump.h" 21 #include "WasmDump.h" 22 #include "XCOFFDump.h" 23 #include "llvm/ADT/Optional.h" 24 #include "llvm/ADT/STLExtras.h" 25 #include "llvm/ADT/SetOperations.h" 26 #include "llvm/ADT/StringExtras.h" 27 #include "llvm/ADT/StringSet.h" 28 #include "llvm/ADT/Triple.h" 29 #include "llvm/CodeGen/FaultMaps.h" 30 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 31 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 32 #include "llvm/Demangle/Demangle.h" 33 #include "llvm/MC/MCAsmInfo.h" 34 #include "llvm/MC/MCContext.h" 35 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 36 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 37 #include "llvm/MC/MCInst.h" 38 #include "llvm/MC/MCInstPrinter.h" 39 #include "llvm/MC/MCInstrAnalysis.h" 40 #include "llvm/MC/MCInstrInfo.h" 41 #include "llvm/MC/MCObjectFileInfo.h" 42 #include "llvm/MC/MCRegisterInfo.h" 43 #include "llvm/MC/MCSubtargetInfo.h" 44 #include "llvm/MC/MCTargetOptions.h" 45 #include "llvm/Object/Archive.h" 46 #include "llvm/Object/COFF.h" 47 #include "llvm/Object/COFFImportFile.h" 48 #include "llvm/Object/ELFObjectFile.h" 49 #include "llvm/Object/MachO.h" 50 #include "llvm/Object/MachOUniversal.h" 51 #include "llvm/Object/ObjectFile.h" 52 #include "llvm/Object/Wasm.h" 53 #include "llvm/Support/Casting.h" 54 #include "llvm/Support/CommandLine.h" 55 #include "llvm/Support/Debug.h" 56 #include "llvm/Support/Errc.h" 57 #include "llvm/Support/FileSystem.h" 58 #include "llvm/Support/Format.h" 59 #include "llvm/Support/FormatVariadic.h" 60 #include "llvm/Support/GraphWriter.h" 61 #include "llvm/Support/Host.h" 62 #include "llvm/Support/InitLLVM.h" 63 #include "llvm/Support/MemoryBuffer.h" 64 #include "llvm/Support/SourceMgr.h" 65 #include "llvm/Support/StringSaver.h" 66 #include "llvm/Support/TargetRegistry.h" 67 #include "llvm/Support/TargetSelect.h" 68 #include "llvm/Support/WithColor.h" 69 #include "llvm/Support/raw_ostream.h" 70 #include <algorithm> 71 #include <cctype> 72 #include <cstring> 73 #include <system_error> 74 #include <unordered_map> 75 #include <utility> 76 77 using namespace llvm; 78 using namespace llvm::object; 79 using namespace llvm::objdump; 80 81 static cl::OptionCategory ObjdumpCat("llvm-objdump Options"); 82 83 static cl::opt<uint64_t> AdjustVMA( 84 "adjust-vma", 85 cl::desc("Increase the displayed address by the specified offset"), 86 cl::value_desc("offset"), cl::init(0), cl::cat(ObjdumpCat)); 87 88 static cl::opt<bool> 89 AllHeaders("all-headers", 90 cl::desc("Display all available header information"), 91 cl::cat(ObjdumpCat)); 92 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 93 cl::NotHidden, cl::Grouping, 94 cl::aliasopt(AllHeaders)); 95 96 static cl::opt<std::string> 97 ArchName("arch-name", 98 cl::desc("Target arch to disassemble for, " 99 "see -version for available targets"), 100 cl::cat(ObjdumpCat)); 101 102 cl::opt<bool> 103 objdump::ArchiveHeaders("archive-headers", 104 cl::desc("Display archive header information"), 105 cl::cat(ObjdumpCat)); 106 static cl::alias ArchiveHeadersShort("a", 107 cl::desc("Alias for --archive-headers"), 108 cl::NotHidden, cl::Grouping, 109 cl::aliasopt(ArchiveHeaders)); 110 111 cl::opt<bool> objdump::Demangle("demangle", cl::desc("Demangle symbols names"), 112 cl::init(false), cl::cat(ObjdumpCat)); 113 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 114 cl::NotHidden, cl::Grouping, 115 cl::aliasopt(Demangle)); 116 117 cl::opt<bool> objdump::Disassemble( 118 "disassemble", 119 cl::desc("Display assembler mnemonics for the machine instructions"), 120 cl::cat(ObjdumpCat)); 121 static cl::alias DisassembleShort("d", cl::desc("Alias for --disassemble"), 122 cl::NotHidden, cl::Grouping, 123 cl::aliasopt(Disassemble)); 124 125 cl::opt<bool> objdump::DisassembleAll( 126 "disassemble-all", 127 cl::desc("Display assembler mnemonics for the machine instructions"), 128 cl::cat(ObjdumpCat)); 129 static cl::alias DisassembleAllShort("D", 130 cl::desc("Alias for --disassemble-all"), 131 cl::NotHidden, cl::Grouping, 132 cl::aliasopt(DisassembleAll)); 133 134 static cl::opt<bool> 135 SymbolDescription("symbol-description", 136 cl::desc("Add symbol description for disassembly. This " 137 "option is for XCOFF files only"), 138 cl::init(false), cl::cat(ObjdumpCat)); 139 140 static cl::list<std::string> 141 DisassembleSymbols("disassemble-symbols", cl::CommaSeparated, 142 cl::desc("List of symbols to disassemble. " 143 "Accept demangled names when --demangle is " 144 "specified, otherwise accept mangled names"), 145 cl::cat(ObjdumpCat)); 146 147 static cl::opt<bool> DisassembleZeroes( 148 "disassemble-zeroes", 149 cl::desc("Do not skip blocks of zeroes when disassembling"), 150 cl::cat(ObjdumpCat)); 151 static cl::alias 152 DisassembleZeroesShort("z", cl::desc("Alias for --disassemble-zeroes"), 153 cl::NotHidden, cl::Grouping, 154 cl::aliasopt(DisassembleZeroes)); 155 156 static cl::list<std::string> 157 DisassemblerOptions("disassembler-options", 158 cl::desc("Pass target specific disassembler options"), 159 cl::value_desc("options"), cl::CommaSeparated, 160 cl::cat(ObjdumpCat)); 161 static cl::alias 162 DisassemblerOptionsShort("M", cl::desc("Alias for --disassembler-options"), 163 cl::NotHidden, cl::Grouping, cl::Prefix, 164 cl::CommaSeparated, 165 cl::aliasopt(DisassemblerOptions)); 166 167 cl::opt<DIDumpType> objdump::DwarfDumpType( 168 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 169 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame")), 170 cl::cat(ObjdumpCat)); 171 172 static cl::opt<bool> DynamicRelocations( 173 "dynamic-reloc", 174 cl::desc("Display the dynamic relocation entries in the file"), 175 cl::cat(ObjdumpCat)); 176 static cl::alias DynamicRelocationShort("R", 177 cl::desc("Alias for --dynamic-reloc"), 178 cl::NotHidden, cl::Grouping, 179 cl::aliasopt(DynamicRelocations)); 180 181 static cl::opt<bool> 182 FaultMapSection("fault-map-section", 183 cl::desc("Display contents of faultmap section"), 184 cl::cat(ObjdumpCat)); 185 186 static cl::opt<bool> 187 FileHeaders("file-headers", 188 cl::desc("Display the contents of the overall file header"), 189 cl::cat(ObjdumpCat)); 190 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 191 cl::NotHidden, cl::Grouping, 192 cl::aliasopt(FileHeaders)); 193 194 cl::opt<bool> 195 objdump::SectionContents("full-contents", 196 cl::desc("Display the content of each section"), 197 cl::cat(ObjdumpCat)); 198 static cl::alias SectionContentsShort("s", 199 cl::desc("Alias for --full-contents"), 200 cl::NotHidden, cl::Grouping, 201 cl::aliasopt(SectionContents)); 202 203 static cl::list<std::string> InputFilenames(cl::Positional, 204 cl::desc("<input object files>"), 205 cl::ZeroOrMore, 206 cl::cat(ObjdumpCat)); 207 208 static cl::opt<bool> 209 PrintLines("line-numbers", 210 cl::desc("Display source line numbers with " 211 "disassembly. Implies disassemble object"), 212 cl::cat(ObjdumpCat)); 213 static cl::alias PrintLinesShort("l", cl::desc("Alias for --line-numbers"), 214 cl::NotHidden, cl::Grouping, 215 cl::aliasopt(PrintLines)); 216 217 static cl::opt<bool> MachOOpt("macho", 218 cl::desc("Use MachO specific object file parser"), 219 cl::cat(ObjdumpCat)); 220 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 221 cl::Grouping, cl::aliasopt(MachOOpt)); 222 223 cl::opt<std::string> objdump::MCPU( 224 "mcpu", cl::desc("Target a specific cpu type (-mcpu=help for details)"), 225 cl::value_desc("cpu-name"), cl::init(""), cl::cat(ObjdumpCat)); 226 227 cl::list<std::string> objdump::MAttrs("mattr", cl::CommaSeparated, 228 cl::desc("Target specific attributes"), 229 cl::value_desc("a1,+a2,-a3,..."), 230 cl::cat(ObjdumpCat)); 231 232 cl::opt<bool> objdump::NoShowRawInsn( 233 "no-show-raw-insn", 234 cl::desc( 235 "When disassembling instructions, do not print the instruction bytes."), 236 cl::cat(ObjdumpCat)); 237 238 cl::opt<bool> objdump::NoLeadingAddr("no-leading-addr", 239 cl::desc("Print no leading address"), 240 cl::cat(ObjdumpCat)); 241 242 static cl::opt<bool> RawClangAST( 243 "raw-clang-ast", 244 cl::desc("Dump the raw binary contents of the clang AST section"), 245 cl::cat(ObjdumpCat)); 246 247 cl::opt<bool> 248 objdump::Relocations("reloc", 249 cl::desc("Display the relocation entries in the file"), 250 cl::cat(ObjdumpCat)); 251 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 252 cl::NotHidden, cl::Grouping, 253 cl::aliasopt(Relocations)); 254 255 cl::opt<bool> 256 objdump::PrintImmHex("print-imm-hex", 257 cl::desc("Use hex format for immediate values"), 258 cl::cat(ObjdumpCat)); 259 260 cl::opt<bool> 261 objdump::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 objdump::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> objdump::SectionHeaders( 279 "section-headers", 280 cl::desc("Display summaries of the 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> objdump::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 static cl::opt<bool> DynamicSymbolTable( 320 "dynamic-syms", 321 cl::desc("Display the contents of the dynamic symbol table"), 322 cl::cat(ObjdumpCat)); 323 static cl::alias DynamicSymbolTableShort("T", 324 cl::desc("Alias for --dynamic-syms"), 325 cl::NotHidden, cl::Grouping, 326 cl::aliasopt(DynamicSymbolTable)); 327 328 cl::opt<std::string> objdump::TripleName( 329 "triple", 330 cl::desc( 331 "Target triple to disassemble for, see -version for available targets"), 332 cl::cat(ObjdumpCat)); 333 334 cl::opt<bool> objdump::UnwindInfo("unwind-info", 335 cl::desc("Display unwind information"), 336 cl::cat(ObjdumpCat)); 337 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 338 cl::NotHidden, cl::Grouping, 339 cl::aliasopt(UnwindInfo)); 340 341 static cl::opt<bool> 342 Wide("wide", cl::desc("Ignored for compatibility with GNU objdump"), 343 cl::cat(ObjdumpCat)); 344 static cl::alias WideShort("w", cl::Grouping, cl::aliasopt(Wide)); 345 346 static cl::extrahelp 347 HelpResponse("\nPass @FILE as argument to read options from FILE.\n"); 348 349 static StringSet<> DisasmSymbolSet; 350 StringSet<> objdump::FoundSectionSet; 351 static StringRef ToolName; 352 353 namespace { 354 struct FilterResult { 355 // True if the section should not be skipped. 356 bool Keep; 357 358 // True if the index counter should be incremented, even if the section should 359 // be skipped. For example, sections may be skipped if they are not included 360 // in the --section flag, but we still want those to count toward the section 361 // count. 362 bool IncrementIndex; 363 }; 364 } // namespace 365 366 static FilterResult checkSectionFilter(object::SectionRef S) { 367 if (FilterSections.empty()) 368 return {/*Keep=*/true, /*IncrementIndex=*/true}; 369 370 Expected<StringRef> SecNameOrErr = S.getName(); 371 if (!SecNameOrErr) { 372 consumeError(SecNameOrErr.takeError()); 373 return {/*Keep=*/false, /*IncrementIndex=*/false}; 374 } 375 StringRef SecName = *SecNameOrErr; 376 377 // StringSet does not allow empty key so avoid adding sections with 378 // no name (such as the section with index 0) here. 379 if (!SecName.empty()) 380 FoundSectionSet.insert(SecName); 381 382 // Only show the section if it's in the FilterSections list, but always 383 // increment so the indexing is stable. 384 return {/*Keep=*/is_contained(FilterSections, SecName), 385 /*IncrementIndex=*/true}; 386 } 387 388 namespace llvm { 389 390 SectionFilter ToolSectionFilter(object::ObjectFile const &O, uint64_t *Idx) { 391 // Start at UINT64_MAX so that the first index returned after an increment is 392 // zero (after the unsigned wrap). 393 if (Idx) 394 *Idx = UINT64_MAX; 395 return SectionFilter( 396 [Idx](object::SectionRef S) { 397 FilterResult Result = checkSectionFilter(S); 398 if (Idx != nullptr && Result.IncrementIndex) 399 *Idx += 1; 400 return Result.Keep; 401 }, 402 O); 403 } 404 405 std::string getFileNameForError(const object::Archive::Child &C, 406 unsigned Index) { 407 Expected<StringRef> NameOrErr = C.getName(); 408 if (NameOrErr) 409 return std::string(NameOrErr.get()); 410 // If we have an error getting the name then we print the index of the archive 411 // member. Since we are already in an error state, we just ignore this error. 412 consumeError(NameOrErr.takeError()); 413 return "<file index: " + std::to_string(Index) + ">"; 414 } 415 416 void reportWarning(Twine Message, StringRef File) { 417 // Output order between errs() and outs() matters especially for archive 418 // files where the output is per member object. 419 outs().flush(); 420 WithColor::warning(errs(), ToolName) 421 << "'" << File << "': " << Message << "\n"; 422 errs().flush(); 423 } 424 425 LLVM_ATTRIBUTE_NORETURN void reportError(StringRef File, Twine Message) { 426 WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n"; 427 exit(1); 428 } 429 430 LLVM_ATTRIBUTE_NORETURN void reportError(Error E, StringRef FileName, 431 StringRef ArchiveName, 432 StringRef ArchitectureName) { 433 assert(E); 434 WithColor::error(errs(), ToolName); 435 if (ArchiveName != "") 436 errs() << ArchiveName << "(" << FileName << ")"; 437 else 438 errs() << "'" << FileName << "'"; 439 if (!ArchitectureName.empty()) 440 errs() << " (for architecture " << ArchitectureName << ")"; 441 std::string Buf; 442 raw_string_ostream OS(Buf); 443 logAllUnhandledErrors(std::move(E), OS); 444 OS.flush(); 445 errs() << ": " << Buf; 446 exit(1); 447 } 448 449 static void reportCmdLineWarning(Twine Message) { 450 WithColor::warning(errs(), ToolName) << Message << "\n"; 451 } 452 453 LLVM_ATTRIBUTE_NORETURN static void reportCmdLineError(Twine Message) { 454 WithColor::error(errs(), ToolName) << Message << "\n"; 455 exit(1); 456 } 457 458 static void warnOnNoMatchForSections() { 459 SetVector<StringRef> MissingSections; 460 for (StringRef S : FilterSections) { 461 if (FoundSectionSet.count(S)) 462 return; 463 // User may specify a unnamed section. Don't warn for it. 464 if (!S.empty()) 465 MissingSections.insert(S); 466 } 467 468 // Warn only if no section in FilterSections is matched. 469 for (StringRef S : MissingSections) 470 reportCmdLineWarning("section '" + S + 471 "' mentioned in a -j/--section option, but not " 472 "found in any input file"); 473 } 474 475 static const Target *getTarget(const ObjectFile *Obj) { 476 // Figure out the target triple. 477 Triple TheTriple("unknown-unknown-unknown"); 478 if (TripleName.empty()) { 479 TheTriple = Obj->makeTriple(); 480 } else { 481 TheTriple.setTriple(Triple::normalize(TripleName)); 482 auto Arch = Obj->getArch(); 483 if (Arch == Triple::arm || Arch == Triple::armeb) 484 Obj->setARMSubArch(TheTriple); 485 } 486 487 // Get the target specific parser. 488 std::string Error; 489 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 490 Error); 491 if (!TheTarget) 492 reportError(Obj->getFileName(), "can't find target: " + Error); 493 494 // Update the triple name and return the found target. 495 TripleName = TheTriple.getTriple(); 496 return TheTarget; 497 } 498 499 bool isRelocAddressLess(RelocationRef A, RelocationRef B) { 500 return A.getOffset() < B.getOffset(); 501 } 502 503 static Error getRelocationValueString(const RelocationRef &Rel, 504 SmallVectorImpl<char> &Result) { 505 const ObjectFile *Obj = Rel.getObject(); 506 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 507 return getELFRelocationValueString(ELF, Rel, Result); 508 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 509 return getCOFFRelocationValueString(COFF, Rel, Result); 510 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 511 return getWasmRelocationValueString(Wasm, Rel, Result); 512 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 513 return getMachORelocationValueString(MachO, Rel, Result); 514 if (auto *XCOFF = dyn_cast<XCOFFObjectFile>(Obj)) 515 return getXCOFFRelocationValueString(XCOFF, Rel, Result); 516 llvm_unreachable("unknown object file format"); 517 } 518 519 /// Indicates whether this relocation should hidden when listing 520 /// relocations, usually because it is the trailing part of a multipart 521 /// relocation that will be printed as part of the leading relocation. 522 static bool getHidden(RelocationRef RelRef) { 523 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 524 if (!MachO) 525 return false; 526 527 unsigned Arch = MachO->getArch(); 528 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 529 uint64_t Type = MachO->getRelocationType(Rel); 530 531 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 532 // is always hidden. 533 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 534 return Type == MachO::GENERIC_RELOC_PAIR; 535 536 if (Arch == Triple::x86_64) { 537 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 538 // an X86_64_RELOC_SUBTRACTOR. 539 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 540 DataRefImpl RelPrev = Rel; 541 RelPrev.d.a--; 542 uint64_t PrevType = MachO->getRelocationType(RelPrev); 543 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 544 return true; 545 } 546 } 547 548 return false; 549 } 550 551 namespace { 552 class SourcePrinter { 553 protected: 554 DILineInfo OldLineInfo; 555 const ObjectFile *Obj = nullptr; 556 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 557 // File name to file contents of source. 558 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 559 // Mark the line endings of the cached source. 560 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 561 // Keep track of missing sources. 562 StringSet<> MissingSources; 563 // Only emit 'no debug info' warning once. 564 bool WarnedNoDebugInfo; 565 566 private: 567 bool cacheSource(const DILineInfo& LineInfoFile); 568 569 void printLines(raw_ostream &OS, const DILineInfo &LineInfo, 570 StringRef Delimiter); 571 572 void printSources(raw_ostream &OS, const DILineInfo &LineInfo, 573 StringRef ObjectFilename, StringRef Delimiter); 574 575 public: 576 SourcePrinter() = default; 577 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) 578 : Obj(Obj), WarnedNoDebugInfo(false) { 579 symbolize::LLVMSymbolizer::Options SymbolizerOpts; 580 SymbolizerOpts.PrintFunctions = 581 DILineInfoSpecifier::FunctionNameKind::LinkageName; 582 SymbolizerOpts.Demangle = Demangle; 583 SymbolizerOpts.DefaultArch = std::string(DefaultArch); 584 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 585 } 586 virtual ~SourcePrinter() = default; 587 virtual void printSourceLine(raw_ostream &OS, 588 object::SectionedAddress Address, 589 StringRef ObjectFilename, 590 StringRef Delimiter = "; "); 591 }; 592 593 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 594 std::unique_ptr<MemoryBuffer> Buffer; 595 if (LineInfo.Source) { 596 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 597 } else { 598 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 599 if (!BufferOrError) { 600 if (MissingSources.insert(LineInfo.FileName).second) 601 reportWarning("failed to find source " + LineInfo.FileName, 602 Obj->getFileName()); 603 return false; 604 } 605 Buffer = std::move(*BufferOrError); 606 } 607 // Chomp the file to get lines 608 const char *BufferStart = Buffer->getBufferStart(), 609 *BufferEnd = Buffer->getBufferEnd(); 610 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 611 const char *Start = BufferStart; 612 for (const char *I = BufferStart; I != BufferEnd; ++I) 613 if (*I == '\n') { 614 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 615 Start = I + 1; 616 } 617 if (Start < BufferEnd) 618 Lines.emplace_back(Start, BufferEnd - Start); 619 SourceCache[LineInfo.FileName] = std::move(Buffer); 620 return true; 621 } 622 623 void SourcePrinter::printSourceLine(raw_ostream &OS, 624 object::SectionedAddress Address, 625 StringRef ObjectFilename, 626 StringRef Delimiter) { 627 if (!Symbolizer) 628 return; 629 630 DILineInfo LineInfo = DILineInfo(); 631 auto ExpectedLineInfo = Symbolizer->symbolizeCode(*Obj, Address); 632 std::string ErrorMessage; 633 if (!ExpectedLineInfo) 634 ErrorMessage = toString(ExpectedLineInfo.takeError()); 635 else 636 LineInfo = *ExpectedLineInfo; 637 638 if (LineInfo.FileName == DILineInfo::BadString) { 639 if (!WarnedNoDebugInfo) { 640 std::string Warning = 641 "failed to parse debug information for " + ObjectFilename.str(); 642 if (!ErrorMessage.empty()) 643 Warning += ": " + ErrorMessage; 644 reportWarning(Warning, ObjectFilename); 645 WarnedNoDebugInfo = true; 646 } 647 } 648 649 if (PrintLines) 650 printLines(OS, LineInfo, Delimiter); 651 if (PrintSource) 652 printSources(OS, LineInfo, ObjectFilename, Delimiter); 653 OldLineInfo = LineInfo; 654 } 655 656 void SourcePrinter::printLines(raw_ostream &OS, const DILineInfo &LineInfo, 657 StringRef Delimiter) { 658 bool PrintFunctionName = LineInfo.FunctionName != DILineInfo::BadString && 659 LineInfo.FunctionName != OldLineInfo.FunctionName; 660 if (PrintFunctionName) { 661 OS << Delimiter << LineInfo.FunctionName; 662 // If demangling is successful, FunctionName will end with "()". Print it 663 // only if demangling did not run or was unsuccessful. 664 if (!StringRef(LineInfo.FunctionName).endswith("()")) 665 OS << "()"; 666 OS << ":\n"; 667 } 668 if (LineInfo.FileName != DILineInfo::BadString && LineInfo.Line != 0 && 669 (OldLineInfo.Line != LineInfo.Line || 670 OldLineInfo.FileName != LineInfo.FileName || PrintFunctionName)) 671 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 672 } 673 674 void SourcePrinter::printSources(raw_ostream &OS, const DILineInfo &LineInfo, 675 StringRef ObjectFilename, 676 StringRef Delimiter) { 677 if (LineInfo.FileName == DILineInfo::BadString || LineInfo.Line == 0 || 678 (OldLineInfo.Line == LineInfo.Line && 679 OldLineInfo.FileName == LineInfo.FileName)) 680 return; 681 682 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 683 if (!cacheSource(LineInfo)) 684 return; 685 auto LineBuffer = LineCache.find(LineInfo.FileName); 686 if (LineBuffer != LineCache.end()) { 687 if (LineInfo.Line > LineBuffer->second.size()) { 688 reportWarning( 689 formatv( 690 "debug info line number {0} exceeds the number of lines in {1}", 691 LineInfo.Line, LineInfo.FileName), 692 ObjectFilename); 693 return; 694 } 695 // Vector begins at 0, line numbers are non-zero 696 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1] << '\n'; 697 } 698 } 699 700 static bool isAArch64Elf(const ObjectFile *Obj) { 701 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 702 return Elf && Elf->getEMachine() == ELF::EM_AARCH64; 703 } 704 705 static bool isArmElf(const ObjectFile *Obj) { 706 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 707 return Elf && Elf->getEMachine() == ELF::EM_ARM; 708 } 709 710 static bool hasMappingSymbols(const ObjectFile *Obj) { 711 return isArmElf(Obj) || isAArch64Elf(Obj); 712 } 713 714 static void printRelocation(StringRef FileName, const RelocationRef &Rel, 715 uint64_t Address, bool Is64Bits) { 716 StringRef Fmt = Is64Bits ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 717 SmallString<16> Name; 718 SmallString<32> Val; 719 Rel.getTypeName(Name); 720 if (Error E = getRelocationValueString(Rel, Val)) 721 reportError(std::move(E), FileName); 722 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n"; 723 } 724 725 class PrettyPrinter { 726 public: 727 virtual ~PrettyPrinter() = default; 728 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 729 ArrayRef<uint8_t> Bytes, 730 object::SectionedAddress Address, raw_ostream &OS, 731 StringRef Annot, MCSubtargetInfo const &STI, 732 SourcePrinter *SP, StringRef ObjectFilename, 733 std::vector<RelocationRef> *Rels = nullptr) { 734 if (SP && (PrintSource || PrintLines)) 735 SP->printSourceLine(OS, Address, ObjectFilename); 736 737 size_t Start = OS.tell(); 738 if (!NoLeadingAddr) 739 OS << format("%8" PRIx64 ":", Address.Address); 740 if (!NoShowRawInsn) { 741 OS << ' '; 742 dumpBytes(Bytes, OS); 743 } 744 745 // The output of printInst starts with a tab. Print some spaces so that 746 // the tab has 1 column and advances to the target tab stop. 747 unsigned TabStop = NoShowRawInsn ? 16 : 40; 748 unsigned Column = OS.tell() - Start; 749 OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8); 750 751 if (MI) { 752 // See MCInstPrinter::printInst. On targets where a PC relative immediate 753 // is relative to the next instruction and the length of a MCInst is 754 // difficult to measure (x86), this is the address of the next 755 // instruction. 756 uint64_t Addr = 757 Address.Address + (STI.getTargetTriple().isX86() ? Bytes.size() : 0); 758 IP.printInst(MI, Addr, "", STI, OS); 759 } else 760 OS << "\t<unknown>"; 761 } 762 }; 763 PrettyPrinter PrettyPrinterInst; 764 765 class HexagonPrettyPrinter : public PrettyPrinter { 766 public: 767 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 768 raw_ostream &OS) { 769 uint32_t opcode = 770 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 771 if (!NoLeadingAddr) 772 OS << format("%8" PRIx64 ":", Address); 773 if (!NoShowRawInsn) { 774 OS << "\t"; 775 dumpBytes(Bytes.slice(0, 4), OS); 776 OS << format("\t%08" PRIx32, opcode); 777 } 778 } 779 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 780 object::SectionedAddress Address, raw_ostream &OS, 781 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 782 StringRef ObjectFilename, 783 std::vector<RelocationRef> *Rels) override { 784 if (SP && (PrintSource || PrintLines)) 785 SP->printSourceLine(OS, Address, ObjectFilename, ""); 786 if (!MI) { 787 printLead(Bytes, Address.Address, OS); 788 OS << " <unknown>"; 789 return; 790 } 791 std::string Buffer; 792 { 793 raw_string_ostream TempStream(Buffer); 794 IP.printInst(MI, Address.Address, "", STI, TempStream); 795 } 796 StringRef Contents(Buffer); 797 // Split off bundle attributes 798 auto PacketBundle = Contents.rsplit('\n'); 799 // Split off first instruction from the rest 800 auto HeadTail = PacketBundle.first.split('\n'); 801 auto Preamble = " { "; 802 auto Separator = ""; 803 804 // Hexagon's packets require relocations to be inline rather than 805 // clustered at the end of the packet. 806 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 807 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 808 auto PrintReloc = [&]() -> void { 809 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 810 if (RelCur->getOffset() == Address.Address) { 811 printRelocation(ObjectFilename, *RelCur, Address.Address, false); 812 return; 813 } 814 ++RelCur; 815 } 816 }; 817 818 while (!HeadTail.first.empty()) { 819 OS << Separator; 820 Separator = "\n"; 821 if (SP && (PrintSource || PrintLines)) 822 SP->printSourceLine(OS, Address, ObjectFilename, ""); 823 printLead(Bytes, Address.Address, OS); 824 OS << Preamble; 825 Preamble = " "; 826 StringRef Inst; 827 auto Duplex = HeadTail.first.split('\v'); 828 if (!Duplex.second.empty()) { 829 OS << Duplex.first; 830 OS << "; "; 831 Inst = Duplex.second; 832 } 833 else 834 Inst = HeadTail.first; 835 OS << Inst; 836 HeadTail = HeadTail.second.split('\n'); 837 if (HeadTail.first.empty()) 838 OS << " } " << PacketBundle.second; 839 PrintReloc(); 840 Bytes = Bytes.slice(4); 841 Address.Address += 4; 842 } 843 } 844 }; 845 HexagonPrettyPrinter HexagonPrettyPrinterInst; 846 847 class AMDGCNPrettyPrinter : public PrettyPrinter { 848 public: 849 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 850 object::SectionedAddress Address, raw_ostream &OS, 851 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 852 StringRef ObjectFilename, 853 std::vector<RelocationRef> *Rels) override { 854 if (SP && (PrintSource || PrintLines)) 855 SP->printSourceLine(OS, Address, ObjectFilename); 856 857 if (MI) { 858 SmallString<40> InstStr; 859 raw_svector_ostream IS(InstStr); 860 861 IP.printInst(MI, Address.Address, "", STI, IS); 862 863 OS << left_justify(IS.str(), 60); 864 } else { 865 // an unrecognized encoding - this is probably data so represent it 866 // using the .long directive, or .byte directive if fewer than 4 bytes 867 // remaining 868 if (Bytes.size() >= 4) { 869 OS << format("\t.long 0x%08" PRIx32 " ", 870 support::endian::read32<support::little>(Bytes.data())); 871 OS.indent(42); 872 } else { 873 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 874 for (unsigned int i = 1; i < Bytes.size(); i++) 875 OS << format(", 0x%02" PRIx8, Bytes[i]); 876 OS.indent(55 - (6 * Bytes.size())); 877 } 878 } 879 880 OS << format("// %012" PRIX64 ":", Address.Address); 881 if (Bytes.size() >= 4) { 882 // D should be casted to uint32_t here as it is passed by format to 883 // snprintf as vararg. 884 for (uint32_t D : makeArrayRef( 885 reinterpret_cast<const support::little32_t *>(Bytes.data()), 886 Bytes.size() / 4)) 887 OS << format(" %08" PRIX32, D); 888 } else { 889 for (unsigned char B : Bytes) 890 OS << format(" %02" PRIX8, B); 891 } 892 893 if (!Annot.empty()) 894 OS << " // " << Annot; 895 } 896 }; 897 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 898 899 class BPFPrettyPrinter : public PrettyPrinter { 900 public: 901 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 902 object::SectionedAddress Address, raw_ostream &OS, 903 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 904 StringRef ObjectFilename, 905 std::vector<RelocationRef> *Rels) override { 906 if (SP && (PrintSource || PrintLines)) 907 SP->printSourceLine(OS, Address, ObjectFilename); 908 if (!NoLeadingAddr) 909 OS << format("%8" PRId64 ":", Address.Address / 8); 910 if (!NoShowRawInsn) { 911 OS << "\t"; 912 dumpBytes(Bytes, OS); 913 } 914 if (MI) 915 IP.printInst(MI, Address.Address, "", STI, OS); 916 else 917 OS << "\t<unknown>"; 918 } 919 }; 920 BPFPrettyPrinter BPFPrettyPrinterInst; 921 922 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 923 switch(Triple.getArch()) { 924 default: 925 return PrettyPrinterInst; 926 case Triple::hexagon: 927 return HexagonPrettyPrinterInst; 928 case Triple::amdgcn: 929 return AMDGCNPrettyPrinterInst; 930 case Triple::bpfel: 931 case Triple::bpfeb: 932 return BPFPrettyPrinterInst; 933 } 934 } 935 } 936 937 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 938 assert(Obj->isELF()); 939 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 940 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 941 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 942 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 943 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 944 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 945 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 946 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 947 llvm_unreachable("Unsupported binary format"); 948 } 949 950 template <class ELFT> static void 951 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 952 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 953 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 954 uint8_t SymbolType = Symbol.getELFType(); 955 if (SymbolType == ELF::STT_SECTION) 956 continue; 957 958 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 959 // ELFSymbolRef::getAddress() returns size instead of value for common 960 // symbols which is not desirable for disassembly output. Overriding. 961 if (SymbolType == ELF::STT_COMMON) 962 Address = Obj->getSymbol(Symbol.getRawDataRefImpl())->st_value; 963 964 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 965 if (Name.empty()) 966 continue; 967 968 section_iterator SecI = 969 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 970 if (SecI == Obj->section_end()) 971 continue; 972 973 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 974 } 975 } 976 977 static void 978 addDynamicElfSymbols(const ObjectFile *Obj, 979 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 980 assert(Obj->isELF()); 981 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 982 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 983 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 984 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 985 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 986 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 987 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 988 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 989 else 990 llvm_unreachable("Unsupported binary format"); 991 } 992 993 static void addPltEntries(const ObjectFile *Obj, 994 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 995 StringSaver &Saver) { 996 Optional<SectionRef> Plt = None; 997 for (const SectionRef &Section : Obj->sections()) { 998 Expected<StringRef> SecNameOrErr = Section.getName(); 999 if (!SecNameOrErr) { 1000 consumeError(SecNameOrErr.takeError()); 1001 continue; 1002 } 1003 if (*SecNameOrErr == ".plt") 1004 Plt = Section; 1005 } 1006 if (!Plt) 1007 return; 1008 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1009 for (auto PltEntry : ElfObj->getPltAddresses()) { 1010 SymbolRef Symbol(PltEntry.first, ElfObj); 1011 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1012 1013 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 1014 if (!Name.empty()) 1015 AllSymbols[*Plt].emplace_back( 1016 PltEntry.second, Saver.save((Name + "@plt").str()), SymbolType); 1017 } 1018 } 1019 } 1020 1021 // Normally the disassembly output will skip blocks of zeroes. This function 1022 // returns the number of zero bytes that can be skipped when dumping the 1023 // disassembly of the instructions in Buf. 1024 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 1025 // Find the number of leading zeroes. 1026 size_t N = 0; 1027 while (N < Buf.size() && !Buf[N]) 1028 ++N; 1029 1030 // We may want to skip blocks of zero bytes, but unless we see 1031 // at least 8 of them in a row. 1032 if (N < 8) 1033 return 0; 1034 1035 // We skip zeroes in multiples of 4 because do not want to truncate an 1036 // instruction if it starts with a zero byte. 1037 return N & ~0x3; 1038 } 1039 1040 // Returns a map from sections to their relocations. 1041 static std::map<SectionRef, std::vector<RelocationRef>> 1042 getRelocsMap(object::ObjectFile const &Obj) { 1043 std::map<SectionRef, std::vector<RelocationRef>> Ret; 1044 uint64_t I = (uint64_t)-1; 1045 for (SectionRef Sec : Obj.sections()) { 1046 ++I; 1047 Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection(); 1048 if (!RelocatedOrErr) 1049 reportError(Obj.getFileName(), 1050 "section (" + Twine(I) + 1051 "): failed to get a relocated section: " + 1052 toString(RelocatedOrErr.takeError())); 1053 1054 section_iterator Relocated = *RelocatedOrErr; 1055 if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep) 1056 continue; 1057 std::vector<RelocationRef> &V = Ret[*Relocated]; 1058 for (const RelocationRef &R : Sec.relocations()) 1059 V.push_back(R); 1060 // Sort relocations by address. 1061 llvm::stable_sort(V, isRelocAddressLess); 1062 } 1063 return Ret; 1064 } 1065 1066 // Used for --adjust-vma to check if address should be adjusted by the 1067 // specified value for a given section. 1068 // For ELF we do not adjust non-allocatable sections like debug ones, 1069 // because they are not loadable. 1070 // TODO: implement for other file formats. 1071 static bool shouldAdjustVA(const SectionRef &Section) { 1072 const ObjectFile *Obj = Section.getObject(); 1073 if (Obj->isELF()) 1074 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 1075 return false; 1076 } 1077 1078 1079 typedef std::pair<uint64_t, char> MappingSymbolPair; 1080 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols, 1081 uint64_t Address) { 1082 auto It = 1083 partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) { 1084 return Val.first <= Address; 1085 }); 1086 // Return zero for any address before the first mapping symbol; this means 1087 // we should use the default disassembly mode, depending on the target. 1088 if (It == MappingSymbols.begin()) 1089 return '\x00'; 1090 return (It - 1)->second; 1091 } 1092 1093 static uint64_t 1094 dumpARMELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 1095 const ObjectFile *Obj, ArrayRef<uint8_t> Bytes, 1096 ArrayRef<MappingSymbolPair> MappingSymbols) { 1097 support::endianness Endian = 1098 Obj->isLittleEndian() ? support::little : support::big; 1099 while (Index < End) { 1100 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1101 outs() << "\t"; 1102 if (Index + 4 <= End) { 1103 dumpBytes(Bytes.slice(Index, 4), outs()); 1104 outs() << "\t.word\t" 1105 << format_hex( 1106 support::endian::read32(Bytes.data() + Index, Endian), 10); 1107 Index += 4; 1108 } else if (Index + 2 <= End) { 1109 dumpBytes(Bytes.slice(Index, 2), outs()); 1110 outs() << "\t\t.short\t" 1111 << format_hex( 1112 support::endian::read16(Bytes.data() + Index, Endian), 6); 1113 Index += 2; 1114 } else { 1115 dumpBytes(Bytes.slice(Index, 1), outs()); 1116 outs() << "\t\t.byte\t" << format_hex(Bytes[0], 4); 1117 ++Index; 1118 } 1119 outs() << "\n"; 1120 if (getMappingSymbolKind(MappingSymbols, Index) != 'd') 1121 break; 1122 } 1123 return Index; 1124 } 1125 1126 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 1127 ArrayRef<uint8_t> Bytes) { 1128 // print out data up to 8 bytes at a time in hex and ascii 1129 uint8_t AsciiData[9] = {'\0'}; 1130 uint8_t Byte; 1131 int NumBytes = 0; 1132 1133 for (; Index < End; ++Index) { 1134 if (NumBytes == 0) 1135 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1136 Byte = Bytes.slice(Index)[0]; 1137 outs() << format(" %02x", Byte); 1138 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1139 1140 uint8_t IndentOffset = 0; 1141 NumBytes++; 1142 if (Index == End - 1 || NumBytes > 8) { 1143 // Indent the space for less than 8 bytes data. 1144 // 2 spaces for byte and one for space between bytes 1145 IndentOffset = 3 * (8 - NumBytes); 1146 for (int Excess = NumBytes; Excess < 8; Excess++) 1147 AsciiData[Excess] = '\0'; 1148 NumBytes = 8; 1149 } 1150 if (NumBytes == 8) { 1151 AsciiData[8] = '\0'; 1152 outs() << std::string(IndentOffset, ' ') << " "; 1153 outs() << reinterpret_cast<char *>(AsciiData); 1154 outs() << '\n'; 1155 NumBytes = 0; 1156 } 1157 } 1158 } 1159 1160 SymbolInfoTy createSymbolInfo(const ObjectFile *Obj, const SymbolRef &Symbol) { 1161 const StringRef FileName = Obj->getFileName(); 1162 const uint64_t Addr = unwrapOrError(Symbol.getAddress(), FileName); 1163 const StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1164 1165 if (Obj->isXCOFF() && SymbolDescription) { 1166 const auto *XCOFFObj = cast<XCOFFObjectFile>(Obj); 1167 DataRefImpl SymbolDRI = Symbol.getRawDataRefImpl(); 1168 1169 const uint32_t SymbolIndex = XCOFFObj->getSymbolIndex(SymbolDRI.p); 1170 Optional<XCOFF::StorageMappingClass> Smc = 1171 getXCOFFSymbolCsectSMC(XCOFFObj, Symbol); 1172 return SymbolInfoTy(Addr, Name, Smc, SymbolIndex, 1173 isLabel(XCOFFObj, Symbol)); 1174 } else 1175 return SymbolInfoTy(Addr, Name, 1176 Obj->isELF() ? getElfSymbolType(Obj, Symbol) 1177 : (uint8_t)ELF::STT_NOTYPE); 1178 } 1179 1180 SymbolInfoTy createDummySymbolInfo(const ObjectFile *Obj, const uint64_t Addr, 1181 StringRef &Name, uint8_t Type) { 1182 if (Obj->isXCOFF() && SymbolDescription) 1183 return SymbolInfoTy(Addr, Name, None, None, false); 1184 else 1185 return SymbolInfoTy(Addr, Name, Type); 1186 } 1187 1188 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 1189 MCContext &Ctx, MCDisassembler *PrimaryDisAsm, 1190 MCDisassembler *SecondaryDisAsm, 1191 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 1192 const MCSubtargetInfo *PrimarySTI, 1193 const MCSubtargetInfo *SecondarySTI, 1194 PrettyPrinter &PIP, 1195 SourcePrinter &SP, bool InlineRelocs) { 1196 const MCSubtargetInfo *STI = PrimarySTI; 1197 MCDisassembler *DisAsm = PrimaryDisAsm; 1198 bool PrimaryIsThumb = false; 1199 if (isArmElf(Obj)) 1200 PrimaryIsThumb = STI->checkFeatures("+thumb-mode"); 1201 1202 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 1203 if (InlineRelocs) 1204 RelocMap = getRelocsMap(*Obj); 1205 bool Is64Bits = Obj->getBytesInAddress() > 4; 1206 1207 // Create a mapping from virtual address to symbol name. This is used to 1208 // pretty print the symbols while disassembling. 1209 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1210 SectionSymbolsTy AbsoluteSymbols; 1211 const StringRef FileName = Obj->getFileName(); 1212 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj); 1213 for (const SymbolRef &Symbol : Obj->symbols()) { 1214 StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1215 if (Name.empty() && !(Obj->isXCOFF() && SymbolDescription)) 1216 continue; 1217 1218 if (Obj->isELF() && getElfSymbolType(Obj, Symbol) == ELF::STT_SECTION) 1219 continue; 1220 1221 // Don't ask a Mach-O STAB symbol for its section unless you know that 1222 // STAB symbol's section field refers to a valid section index. Otherwise 1223 // the symbol may error trying to load a section that does not exist. 1224 if (MachO) { 1225 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1226 uint8_t NType = (MachO->is64Bit() ? 1227 MachO->getSymbol64TableEntry(SymDRI).n_type: 1228 MachO->getSymbolTableEntry(SymDRI).n_type); 1229 if (NType & MachO::N_STAB) 1230 continue; 1231 } 1232 1233 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1234 if (SecI != Obj->section_end()) 1235 AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol)); 1236 else 1237 AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol)); 1238 } 1239 1240 if (AllSymbols.empty() && Obj->isELF()) 1241 addDynamicElfSymbols(Obj, AllSymbols); 1242 1243 BumpPtrAllocator A; 1244 StringSaver Saver(A); 1245 addPltEntries(Obj, AllSymbols, Saver); 1246 1247 // Create a mapping from virtual address to section. An empty section can 1248 // cause more than one section at the same address. Use a stable sort to 1249 // stabilize the output. 1250 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1251 for (SectionRef Sec : Obj->sections()) 1252 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1253 llvm::stable_sort(SectionAddresses, llvm::less_first()); 1254 1255 // Linked executables (.exe and .dll files) typically don't include a real 1256 // symbol table but they might contain an export table. 1257 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1258 for (const auto &ExportEntry : COFFObj->export_directories()) { 1259 StringRef Name; 1260 if (std::error_code EC = ExportEntry.getSymbolName(Name)) 1261 reportError(errorCodeToError(EC), Obj->getFileName()); 1262 if (Name.empty()) 1263 continue; 1264 1265 uint32_t RVA; 1266 if (std::error_code EC = ExportEntry.getExportRVA(RVA)) 1267 reportError(errorCodeToError(EC), Obj->getFileName()); 1268 1269 uint64_t VA = COFFObj->getImageBase() + RVA; 1270 auto Sec = partition_point( 1271 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) { 1272 return O.first <= VA; 1273 }); 1274 if (Sec != SectionAddresses.begin()) { 1275 --Sec; 1276 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1277 } else 1278 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1279 } 1280 } 1281 1282 // Sort all the symbols, this allows us to use a simple binary search to find 1283 // Multiple symbols can have the same address. Use a stable sort to stabilize 1284 // the output. 1285 StringSet<> FoundDisasmSymbolSet; 1286 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1287 stable_sort(SecSyms.second); 1288 stable_sort(AbsoluteSymbols); 1289 1290 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1291 if (FilterSections.empty() && !DisassembleAll && 1292 (!Section.isText() || Section.isVirtual())) 1293 continue; 1294 1295 uint64_t SectionAddr = Section.getAddress(); 1296 uint64_t SectSize = Section.getSize(); 1297 if (!SectSize) 1298 continue; 1299 1300 // Get the list of all the symbols in this section. 1301 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1302 std::vector<MappingSymbolPair> MappingSymbols; 1303 if (hasMappingSymbols(Obj)) { 1304 for (const auto &Symb : Symbols) { 1305 uint64_t Address = Symb.Addr; 1306 StringRef Name = Symb.Name; 1307 if (Name.startswith("$d")) 1308 MappingSymbols.emplace_back(Address - SectionAddr, 'd'); 1309 if (Name.startswith("$x")) 1310 MappingSymbols.emplace_back(Address - SectionAddr, 'x'); 1311 if (Name.startswith("$a")) 1312 MappingSymbols.emplace_back(Address - SectionAddr, 'a'); 1313 if (Name.startswith("$t")) 1314 MappingSymbols.emplace_back(Address - SectionAddr, 't'); 1315 } 1316 } 1317 1318 llvm::sort(MappingSymbols); 1319 1320 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1321 // AMDGPU disassembler uses symbolizer for printing labels 1322 std::unique_ptr<MCRelocationInfo> RelInfo( 1323 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1324 if (RelInfo) { 1325 std::unique_ptr<MCSymbolizer> Symbolizer( 1326 TheTarget->createMCSymbolizer( 1327 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1328 DisAsm->setSymbolizer(std::move(Symbolizer)); 1329 } 1330 } 1331 1332 StringRef SegmentName = ""; 1333 if (MachO) { 1334 DataRefImpl DR = Section.getRawDataRefImpl(); 1335 SegmentName = MachO->getSectionFinalSegmentName(DR); 1336 } 1337 1338 StringRef SectionName = unwrapOrError(Section.getName(), Obj->getFileName()); 1339 // If the section has no symbol at the start, just insert a dummy one. 1340 if (Symbols.empty() || Symbols[0].Addr != 0) { 1341 Symbols.insert(Symbols.begin(), 1342 createDummySymbolInfo(Obj, SectionAddr, SectionName, 1343 Section.isText() ? ELF::STT_FUNC 1344 : ELF::STT_OBJECT)); 1345 } 1346 1347 SmallString<40> Comments; 1348 raw_svector_ostream CommentStream(Comments); 1349 1350 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef( 1351 unwrapOrError(Section.getContents(), Obj->getFileName())); 1352 1353 uint64_t VMAAdjustment = 0; 1354 if (shouldAdjustVA(Section)) 1355 VMAAdjustment = AdjustVMA; 1356 1357 uint64_t Size; 1358 uint64_t Index; 1359 bool PrintedSection = false; 1360 std::vector<RelocationRef> Rels = RelocMap[Section]; 1361 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1362 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1363 // Disassemble symbol by symbol. 1364 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1365 std::string SymbolName = Symbols[SI].Name.str(); 1366 if (Demangle) 1367 SymbolName = demangle(SymbolName); 1368 1369 // Skip if --disassemble-symbols is not empty and the symbol is not in 1370 // the list. 1371 if (!DisasmSymbolSet.empty() && !DisasmSymbolSet.count(SymbolName)) 1372 continue; 1373 1374 uint64_t Start = Symbols[SI].Addr; 1375 if (Start < SectionAddr || StopAddress <= Start) 1376 continue; 1377 else 1378 FoundDisasmSymbolSet.insert(SymbolName); 1379 1380 // The end is the section end, the beginning of the next symbol, or 1381 // --stop-address. 1382 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress); 1383 if (SI + 1 < SE) 1384 End = std::min(End, Symbols[SI + 1].Addr); 1385 if (Start >= End || End <= StartAddress) 1386 continue; 1387 Start -= SectionAddr; 1388 End -= SectionAddr; 1389 1390 if (!PrintedSection) { 1391 PrintedSection = true; 1392 outs() << "\nDisassembly of section "; 1393 if (!SegmentName.empty()) 1394 outs() << SegmentName << ","; 1395 outs() << SectionName << ":\n"; 1396 } 1397 1398 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1399 if (Symbols[SI].Type == ELF::STT_AMDGPU_HSA_KERNEL) { 1400 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1401 Start += 256; 1402 } 1403 if (SI == SE - 1 || 1404 Symbols[SI + 1].Type == ELF::STT_AMDGPU_HSA_KERNEL) { 1405 // cut trailing zeroes at the end of kernel 1406 // cut up to 256 bytes 1407 const uint64_t EndAlign = 256; 1408 const auto Limit = End - (std::min)(EndAlign, End - Start); 1409 while (End > Limit && 1410 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1411 End -= 4; 1412 } 1413 } 1414 1415 outs() << '\n'; 1416 if (!NoLeadingAddr) 1417 outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ", 1418 SectionAddr + Start + VMAAdjustment); 1419 if (Obj->isXCOFF() && SymbolDescription) { 1420 printXCOFFSymbolDescription(Symbols[SI], SymbolName); 1421 outs() << ":\n"; 1422 } else 1423 outs() << '<' << SymbolName << ">:\n"; 1424 1425 // Don't print raw contents of a virtual section. A virtual section 1426 // doesn't have any contents in the file. 1427 if (Section.isVirtual()) { 1428 outs() << "...\n"; 1429 continue; 1430 } 1431 1432 // Some targets (like WebAssembly) have a special prelude at the start 1433 // of each symbol. 1434 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1435 SectionAddr + Start, CommentStream); 1436 Start += Size; 1437 1438 Index = Start; 1439 if (SectionAddr < StartAddress) 1440 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1441 1442 // If there is a data/common symbol inside an ELF text section and we are 1443 // only disassembling text (applicable all architectures), we are in a 1444 // situation where we must print the data and not disassemble it. 1445 if (Obj->isELF() && !DisassembleAll && Section.isText()) { 1446 uint8_t SymTy = Symbols[SI].Type; 1447 if (SymTy == ELF::STT_OBJECT || SymTy == ELF::STT_COMMON) { 1448 dumpELFData(SectionAddr, Index, End, Bytes); 1449 Index = End; 1450 } 1451 } 1452 1453 bool CheckARMELFData = hasMappingSymbols(Obj) && 1454 Symbols[SI].Type != ELF::STT_OBJECT && 1455 !DisassembleAll; 1456 while (Index < End) { 1457 // ARM and AArch64 ELF binaries can interleave data and text in the 1458 // same section. We rely on the markers introduced to understand what 1459 // we need to dump. If the data marker is within a function, it is 1460 // denoted as a word/short etc. 1461 if (CheckARMELFData && 1462 getMappingSymbolKind(MappingSymbols, Index) == 'd') { 1463 Index = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1464 MappingSymbols); 1465 continue; 1466 } 1467 1468 // When -z or --disassemble-zeroes are given we always dissasemble 1469 // them. Otherwise we might want to skip zero bytes we see. 1470 if (!DisassembleZeroes) { 1471 uint64_t MaxOffset = End - Index; 1472 // For --reloc: print zero blocks patched by relocations, so that 1473 // relocations can be shown in the dump. 1474 if (RelCur != RelEnd) 1475 MaxOffset = RelCur->getOffset() - Index; 1476 1477 if (size_t N = 1478 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1479 outs() << "\t\t..." << '\n'; 1480 Index += N; 1481 continue; 1482 } 1483 } 1484 1485 if (SecondarySTI) { 1486 if (getMappingSymbolKind(MappingSymbols, Index) == 'a') { 1487 STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI; 1488 DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm; 1489 } else if (getMappingSymbolKind(MappingSymbols, Index) == 't') { 1490 STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI; 1491 DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm; 1492 } 1493 } 1494 1495 // Disassemble a real instruction or a data when disassemble all is 1496 // provided 1497 MCInst Inst; 1498 bool Disassembled = DisAsm->getInstruction( 1499 Inst, Size, Bytes.slice(Index), SectionAddr + Index, CommentStream); 1500 if (Size == 0) 1501 Size = 1; 1502 1503 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1504 Bytes.slice(Index, Size), 1505 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, 1506 outs(), "", *STI, &SP, Obj->getFileName(), &Rels); 1507 outs() << CommentStream.str(); 1508 Comments.clear(); 1509 1510 // If disassembly has failed, avoid analysing invalid/incomplete 1511 // instruction information. Otherwise, try to resolve the target of a 1512 // call, tail call, etc. to a specific symbol. 1513 if (Disassembled && MIA && 1514 (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1515 MIA->isConditionalBranch(Inst))) { 1516 uint64_t Target; 1517 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1518 // In a relocatable object, the target's section must reside in 1519 // the same section as the call instruction or it is accessed 1520 // through a relocation. 1521 // 1522 // In a non-relocatable object, the target may be in any section. 1523 // 1524 // N.B. We don't walk the relocations in the relocatable case yet. 1525 auto *TargetSectionSymbols = &Symbols; 1526 if (!Obj->isRelocatableObject()) { 1527 auto It = partition_point( 1528 SectionAddresses, 1529 [=](const std::pair<uint64_t, SectionRef> &O) { 1530 return O.first <= Target; 1531 }); 1532 if (It != SectionAddresses.begin()) { 1533 --It; 1534 TargetSectionSymbols = &AllSymbols[It->second]; 1535 } else { 1536 TargetSectionSymbols = &AbsoluteSymbols; 1537 } 1538 } 1539 1540 // Find the last symbol in the section whose offset is less than 1541 // or equal to the target. If there isn't a section that contains 1542 // the target, find the nearest preceding absolute symbol. 1543 auto TargetSym = partition_point( 1544 *TargetSectionSymbols, 1545 [=](const SymbolInfoTy &O) { 1546 return O.Addr <= Target; 1547 }); 1548 if (TargetSym == TargetSectionSymbols->begin()) { 1549 TargetSectionSymbols = &AbsoluteSymbols; 1550 TargetSym = partition_point( 1551 AbsoluteSymbols, 1552 [=](const SymbolInfoTy &O) { 1553 return O.Addr <= Target; 1554 }); 1555 } 1556 if (TargetSym != TargetSectionSymbols->begin()) { 1557 --TargetSym; 1558 uint64_t TargetAddress = TargetSym->Addr; 1559 std::string TargetName = TargetSym->Name.str(); 1560 if (Demangle) 1561 TargetName = demangle(TargetName); 1562 1563 outs() << " <" << TargetName; 1564 uint64_t Disp = Target - TargetAddress; 1565 if (Disp) 1566 outs() << "+0x" << Twine::utohexstr(Disp); 1567 outs() << '>'; 1568 } 1569 } 1570 } 1571 outs() << "\n"; 1572 1573 // Hexagon does this in pretty printer 1574 if (Obj->getArch() != Triple::hexagon) { 1575 // Print relocation for instruction and data. 1576 while (RelCur != RelEnd) { 1577 uint64_t Offset = RelCur->getOffset(); 1578 // If this relocation is hidden, skip it. 1579 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 1580 ++RelCur; 1581 continue; 1582 } 1583 1584 // Stop when RelCur's offset is past the disassembled 1585 // instruction/data. Note that it's possible the disassembled data 1586 // is not the complete data: we might see the relocation printed in 1587 // the middle of the data, but this matches the binutils objdump 1588 // output. 1589 if (Offset >= Index + Size) 1590 break; 1591 1592 // When --adjust-vma is used, update the address printed. 1593 if (RelCur->getSymbol() != Obj->symbol_end()) { 1594 Expected<section_iterator> SymSI = 1595 RelCur->getSymbol()->getSection(); 1596 if (SymSI && *SymSI != Obj->section_end() && 1597 shouldAdjustVA(**SymSI)) 1598 Offset += AdjustVMA; 1599 } 1600 1601 printRelocation(Obj->getFileName(), *RelCur, SectionAddr + Offset, 1602 Is64Bits); 1603 ++RelCur; 1604 } 1605 } 1606 1607 Index += Size; 1608 } 1609 } 1610 } 1611 StringSet<> MissingDisasmSymbolSet = 1612 set_difference(DisasmSymbolSet, FoundDisasmSymbolSet); 1613 for (StringRef Sym : MissingDisasmSymbolSet.keys()) 1614 reportWarning("failed to disassemble missing symbol " + Sym, FileName); 1615 } 1616 1617 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1618 const Target *TheTarget = getTarget(Obj); 1619 1620 // Package up features to be passed to target/subtarget 1621 SubtargetFeatures Features = Obj->getFeatures(); 1622 if (!MAttrs.empty()) 1623 for (unsigned I = 0; I != MAttrs.size(); ++I) 1624 Features.AddFeature(MAttrs[I]); 1625 1626 std::unique_ptr<const MCRegisterInfo> MRI( 1627 TheTarget->createMCRegInfo(TripleName)); 1628 if (!MRI) 1629 reportError(Obj->getFileName(), 1630 "no register info for target " + TripleName); 1631 1632 // Set up disassembler. 1633 MCTargetOptions MCOptions; 1634 std::unique_ptr<const MCAsmInfo> AsmInfo( 1635 TheTarget->createMCAsmInfo(*MRI, TripleName, MCOptions)); 1636 if (!AsmInfo) 1637 reportError(Obj->getFileName(), 1638 "no assembly info for target " + TripleName); 1639 std::unique_ptr<const MCSubtargetInfo> STI( 1640 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1641 if (!STI) 1642 reportError(Obj->getFileName(), 1643 "no subtarget info for target " + TripleName); 1644 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1645 if (!MII) 1646 reportError(Obj->getFileName(), 1647 "no instruction info for target " + TripleName); 1648 MCObjectFileInfo MOFI; 1649 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1650 // FIXME: for now initialize MCObjectFileInfo with default values 1651 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1652 1653 std::unique_ptr<MCDisassembler> DisAsm( 1654 TheTarget->createMCDisassembler(*STI, Ctx)); 1655 if (!DisAsm) 1656 reportError(Obj->getFileName(), "no disassembler for target " + TripleName); 1657 1658 // If we have an ARM object file, we need a second disassembler, because 1659 // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode. 1660 // We use mapping symbols to switch between the two assemblers, where 1661 // appropriate. 1662 std::unique_ptr<MCDisassembler> SecondaryDisAsm; 1663 std::unique_ptr<const MCSubtargetInfo> SecondarySTI; 1664 if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) { 1665 if (STI->checkFeatures("+thumb-mode")) 1666 Features.AddFeature("-thumb-mode"); 1667 else 1668 Features.AddFeature("+thumb-mode"); 1669 SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU, 1670 Features.getString())); 1671 SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx)); 1672 } 1673 1674 std::unique_ptr<const MCInstrAnalysis> MIA( 1675 TheTarget->createMCInstrAnalysis(MII.get())); 1676 1677 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1678 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1679 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1680 if (!IP) 1681 reportError(Obj->getFileName(), 1682 "no instruction printer for target " + TripleName); 1683 IP->setPrintImmHex(PrintImmHex); 1684 IP->setPrintBranchImmAsAddress(true); 1685 1686 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1687 SourcePrinter SP(Obj, TheTarget->getName()); 1688 1689 for (StringRef Opt : DisassemblerOptions) 1690 if (!IP->applyTargetSpecificCLOption(Opt)) 1691 reportError(Obj->getFileName(), 1692 "Unrecognized disassembler option: " + Opt); 1693 1694 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(), 1695 MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP, 1696 SP, InlineRelocs); 1697 } 1698 1699 void printRelocations(const ObjectFile *Obj) { 1700 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1701 "%08" PRIx64; 1702 // Regular objdump doesn't print relocations in non-relocatable object 1703 // files. 1704 if (!Obj->isRelocatableObject()) 1705 return; 1706 1707 // Build a mapping from relocation target to a vector of relocation 1708 // sections. Usually, there is an only one relocation section for 1709 // each relocated section. 1710 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec; 1711 uint64_t Ndx; 1712 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Ndx)) { 1713 if (Section.relocation_begin() == Section.relocation_end()) 1714 continue; 1715 Expected<section_iterator> SecOrErr = Section.getRelocatedSection(); 1716 if (!SecOrErr) 1717 reportError(Obj->getFileName(), 1718 "section (" + Twine(Ndx) + 1719 "): unable to get a relocation target: " + 1720 toString(SecOrErr.takeError())); 1721 SecToRelSec[**SecOrErr].push_back(Section); 1722 } 1723 1724 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) { 1725 StringRef SecName = unwrapOrError(P.first.getName(), Obj->getFileName()); 1726 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1727 uint32_t OffsetPadding = (Obj->getBytesInAddress() > 4 ? 16 : 8); 1728 uint32_t TypePadding = 24; 1729 outs() << left_justify("OFFSET", OffsetPadding) << " " 1730 << left_justify("TYPE", TypePadding) << " " 1731 << "VALUE\n"; 1732 1733 for (SectionRef Section : P.second) { 1734 for (const RelocationRef &Reloc : Section.relocations()) { 1735 uint64_t Address = Reloc.getOffset(); 1736 SmallString<32> RelocName; 1737 SmallString<32> ValueStr; 1738 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1739 continue; 1740 Reloc.getTypeName(RelocName); 1741 if (Error E = getRelocationValueString(Reloc, ValueStr)) 1742 reportError(std::move(E), Obj->getFileName()); 1743 1744 outs() << format(Fmt.data(), Address) << " " 1745 << left_justify(RelocName, TypePadding) << " " << ValueStr 1746 << "\n"; 1747 } 1748 } 1749 outs() << "\n"; 1750 } 1751 } 1752 1753 void printDynamicRelocations(const ObjectFile *Obj) { 1754 // For the moment, this option is for ELF only 1755 if (!Obj->isELF()) 1756 return; 1757 1758 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1759 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1760 reportError(Obj->getFileName(), "not a dynamic object"); 1761 return; 1762 } 1763 1764 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1765 if (DynRelSec.empty()) 1766 return; 1767 1768 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1769 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1770 for (const SectionRef &Section : DynRelSec) 1771 for (const RelocationRef &Reloc : Section.relocations()) { 1772 uint64_t Address = Reloc.getOffset(); 1773 SmallString<32> RelocName; 1774 SmallString<32> ValueStr; 1775 Reloc.getTypeName(RelocName); 1776 if (Error E = getRelocationValueString(Reloc, ValueStr)) 1777 reportError(std::move(E), Obj->getFileName()); 1778 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1779 << ValueStr << "\n"; 1780 } 1781 } 1782 1783 // Returns true if we need to show LMA column when dumping section headers. We 1784 // show it only when the platform is ELF and either we have at least one section 1785 // whose VMA and LMA are different and/or when --show-lma flag is used. 1786 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1787 if (!Obj->isELF()) 1788 return false; 1789 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1790 if (S.getAddress() != getELFSectionLMA(S)) 1791 return true; 1792 return ShowLMA; 1793 } 1794 1795 static size_t getMaxSectionNameWidth(const ObjectFile *Obj) { 1796 // Default column width for names is 13 even if no names are that long. 1797 size_t MaxWidth = 13; 1798 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1799 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 1800 MaxWidth = std::max(MaxWidth, Name.size()); 1801 } 1802 return MaxWidth; 1803 } 1804 1805 void printSectionHeaders(const ObjectFile *Obj) { 1806 size_t NameWidth = getMaxSectionNameWidth(Obj); 1807 size_t AddressWidth = 2 * Obj->getBytesInAddress(); 1808 bool HasLMAColumn = shouldDisplayLMA(Obj); 1809 if (HasLMAColumn) 1810 outs() << "Sections:\n" 1811 "Idx " 1812 << left_justify("Name", NameWidth) << " Size " 1813 << left_justify("VMA", AddressWidth) << " " 1814 << left_justify("LMA", AddressWidth) << " Type\n"; 1815 else 1816 outs() << "Sections:\n" 1817 "Idx " 1818 << left_justify("Name", NameWidth) << " Size " 1819 << left_justify("VMA", AddressWidth) << " Type\n"; 1820 1821 uint64_t Idx; 1822 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Idx)) { 1823 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 1824 uint64_t VMA = Section.getAddress(); 1825 if (shouldAdjustVA(Section)) 1826 VMA += AdjustVMA; 1827 1828 uint64_t Size = Section.getSize(); 1829 1830 std::string Type = Section.isText() ? "TEXT" : ""; 1831 if (Section.isData()) 1832 Type += Type.empty() ? "DATA" : " DATA"; 1833 if (Section.isBSS()) 1834 Type += Type.empty() ? "BSS" : " BSS"; 1835 1836 if (HasLMAColumn) 1837 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 1838 Name.str().c_str(), Size) 1839 << format_hex_no_prefix(VMA, AddressWidth) << " " 1840 << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth) 1841 << " " << Type << "\n"; 1842 else 1843 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 1844 Name.str().c_str(), Size) 1845 << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n"; 1846 } 1847 outs() << "\n"; 1848 } 1849 1850 void printSectionContents(const ObjectFile *Obj) { 1851 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1852 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 1853 uint64_t BaseAddr = Section.getAddress(); 1854 uint64_t Size = Section.getSize(); 1855 if (!Size) 1856 continue; 1857 1858 outs() << "Contents of section " << Name << ":\n"; 1859 if (Section.isBSS()) { 1860 outs() << format("<skipping contents of bss section at [%04" PRIx64 1861 ", %04" PRIx64 ")>\n", 1862 BaseAddr, BaseAddr + Size); 1863 continue; 1864 } 1865 1866 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName()); 1867 1868 // Dump out the content as hex and printable ascii characters. 1869 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1870 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1871 // Dump line of hex. 1872 for (std::size_t I = 0; I < 16; ++I) { 1873 if (I != 0 && I % 4 == 0) 1874 outs() << ' '; 1875 if (Addr + I < End) 1876 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1877 << hexdigit(Contents[Addr + I] & 0xF, true); 1878 else 1879 outs() << " "; 1880 } 1881 // Print ascii. 1882 outs() << " "; 1883 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1884 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1885 outs() << Contents[Addr + I]; 1886 else 1887 outs() << "."; 1888 } 1889 outs() << "\n"; 1890 } 1891 } 1892 } 1893 1894 void printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1895 StringRef ArchitectureName, bool DumpDynamic) { 1896 if (O->isCOFF() && !DumpDynamic) { 1897 outs() << "SYMBOL TABLE:\n"; 1898 printCOFFSymbolTable(cast<const COFFObjectFile>(O)); 1899 return; 1900 } 1901 1902 const StringRef FileName = O->getFileName(); 1903 1904 if (!DumpDynamic) { 1905 outs() << "SYMBOL TABLE:\n"; 1906 for (auto I = O->symbol_begin(); I != O->symbol_end(); ++I) 1907 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 1908 return; 1909 } 1910 1911 outs() << "DYNAMIC SYMBOL TABLE:\n"; 1912 if (!O->isELF()) { 1913 reportWarning( 1914 "this operation is not currently supported for this file format", 1915 FileName); 1916 return; 1917 } 1918 1919 const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(O); 1920 for (auto I = ELF->getDynamicSymbolIterators().begin(); 1921 I != ELF->getDynamicSymbolIterators().end(); ++I) 1922 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 1923 } 1924 1925 void printSymbol(const ObjectFile *O, const SymbolRef &Symbol, 1926 StringRef FileName, StringRef ArchiveName, 1927 StringRef ArchitectureName, bool DumpDynamic) { 1928 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(O); 1929 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName, ArchiveName, 1930 ArchitectureName); 1931 if ((Address < StartAddress) || (Address > StopAddress)) 1932 return; 1933 SymbolRef::Type Type = 1934 unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName); 1935 uint32_t Flags = 1936 unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName); 1937 1938 // Don't ask a Mach-O STAB symbol for its section unless you know that 1939 // STAB symbol's section field refers to a valid section index. Otherwise 1940 // the symbol may error trying to load a section that does not exist. 1941 bool IsSTAB = false; 1942 if (MachO) { 1943 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1944 uint8_t NType = 1945 (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type 1946 : MachO->getSymbolTableEntry(SymDRI).n_type); 1947 if (NType & MachO::N_STAB) 1948 IsSTAB = true; 1949 } 1950 section_iterator Section = IsSTAB 1951 ? O->section_end() 1952 : unwrapOrError(Symbol.getSection(), FileName, 1953 ArchiveName, ArchitectureName); 1954 1955 StringRef Name; 1956 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 1957 if (Expected<StringRef> NameOrErr = Section->getName()) 1958 Name = *NameOrErr; 1959 else 1960 consumeError(NameOrErr.takeError()); 1961 1962 } else { 1963 Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName, 1964 ArchitectureName); 1965 } 1966 1967 bool Global = Flags & SymbolRef::SF_Global; 1968 bool Weak = Flags & SymbolRef::SF_Weak; 1969 bool Absolute = Flags & SymbolRef::SF_Absolute; 1970 bool Common = Flags & SymbolRef::SF_Common; 1971 bool Hidden = Flags & SymbolRef::SF_Hidden; 1972 1973 char GlobLoc = ' '; 1974 if ((Section != O->section_end() || Absolute) && !Weak) 1975 GlobLoc = Global ? 'g' : 'l'; 1976 char IFunc = ' '; 1977 if (O->isELF()) { 1978 if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC) 1979 IFunc = 'i'; 1980 if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE) 1981 GlobLoc = 'u'; 1982 } 1983 1984 char Debug = ' '; 1985 if (DumpDynamic) 1986 Debug = 'D'; 1987 else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1988 Debug = 'd'; 1989 1990 char FileFunc = ' '; 1991 if (Type == SymbolRef::ST_File) 1992 FileFunc = 'f'; 1993 else if (Type == SymbolRef::ST_Function) 1994 FileFunc = 'F'; 1995 else if (Type == SymbolRef::ST_Data) 1996 FileFunc = 'O'; 1997 1998 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1999 2000 outs() << format(Fmt, Address) << " " 2001 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2002 << (Weak ? 'w' : ' ') // Weak? 2003 << ' ' // Constructor. Not supported yet. 2004 << ' ' // Warning. Not supported yet. 2005 << IFunc // Indirect reference to another symbol. 2006 << Debug // Debugging (d) or dynamic (D) symbol. 2007 << FileFunc // Name of function (F), file (f) or object (O). 2008 << ' '; 2009 if (Absolute) { 2010 outs() << "*ABS*"; 2011 } else if (Common) { 2012 outs() << "*COM*"; 2013 } else if (Section == O->section_end()) { 2014 outs() << "*UND*"; 2015 } else { 2016 if (MachO) { 2017 DataRefImpl DR = Section->getRawDataRefImpl(); 2018 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 2019 outs() << SegmentName << ","; 2020 } 2021 StringRef SectionName = unwrapOrError(Section->getName(), FileName); 2022 outs() << SectionName; 2023 } 2024 2025 if (Common || O->isELF()) { 2026 uint64_t Val = 2027 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2028 outs() << '\t' << format(Fmt, Val); 2029 } 2030 2031 if (O->isELF()) { 2032 uint8_t Other = ELFSymbolRef(Symbol).getOther(); 2033 switch (Other) { 2034 case ELF::STV_DEFAULT: 2035 break; 2036 case ELF::STV_INTERNAL: 2037 outs() << " .internal"; 2038 break; 2039 case ELF::STV_HIDDEN: 2040 outs() << " .hidden"; 2041 break; 2042 case ELF::STV_PROTECTED: 2043 outs() << " .protected"; 2044 break; 2045 default: 2046 outs() << format(" 0x%02x", Other); 2047 break; 2048 } 2049 } else if (Hidden) { 2050 outs() << " .hidden"; 2051 } 2052 2053 if (Demangle) 2054 outs() << ' ' << demangle(std::string(Name)) << '\n'; 2055 else 2056 outs() << ' ' << Name << '\n'; 2057 } 2058 2059 static void printUnwindInfo(const ObjectFile *O) { 2060 outs() << "Unwind info:\n\n"; 2061 2062 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 2063 printCOFFUnwindInfo(Coff); 2064 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 2065 printMachOUnwindInfo(MachO); 2066 else 2067 // TODO: Extract DWARF dump tool to objdump. 2068 WithColor::error(errs(), ToolName) 2069 << "This operation is only currently supported " 2070 "for COFF and MachO object files.\n"; 2071 } 2072 2073 /// Dump the raw contents of the __clangast section so the output can be piped 2074 /// into llvm-bcanalyzer. 2075 void printRawClangAST(const ObjectFile *Obj) { 2076 if (outs().is_displayed()) { 2077 WithColor::error(errs(), ToolName) 2078 << "The -raw-clang-ast option will dump the raw binary contents of " 2079 "the clang ast section.\n" 2080 "Please redirect the output to a file or another program such as " 2081 "llvm-bcanalyzer.\n"; 2082 return; 2083 } 2084 2085 StringRef ClangASTSectionName("__clangast"); 2086 if (Obj->isCOFF()) { 2087 ClangASTSectionName = "clangast"; 2088 } 2089 2090 Optional<object::SectionRef> ClangASTSection; 2091 for (auto Sec : ToolSectionFilter(*Obj)) { 2092 StringRef Name; 2093 if (Expected<StringRef> NameOrErr = Sec.getName()) 2094 Name = *NameOrErr; 2095 else 2096 consumeError(NameOrErr.takeError()); 2097 2098 if (Name == ClangASTSectionName) { 2099 ClangASTSection = Sec; 2100 break; 2101 } 2102 } 2103 if (!ClangASTSection) 2104 return; 2105 2106 StringRef ClangASTContents = unwrapOrError( 2107 ClangASTSection.getValue().getContents(), Obj->getFileName()); 2108 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2109 } 2110 2111 static void printFaultMaps(const ObjectFile *Obj) { 2112 StringRef FaultMapSectionName; 2113 2114 if (Obj->isELF()) { 2115 FaultMapSectionName = ".llvm_faultmaps"; 2116 } else if (Obj->isMachO()) { 2117 FaultMapSectionName = "__llvm_faultmaps"; 2118 } else { 2119 WithColor::error(errs(), ToolName) 2120 << "This operation is only currently supported " 2121 "for ELF and Mach-O executable files.\n"; 2122 return; 2123 } 2124 2125 Optional<object::SectionRef> FaultMapSection; 2126 2127 for (auto Sec : ToolSectionFilter(*Obj)) { 2128 StringRef Name; 2129 if (Expected<StringRef> NameOrErr = Sec.getName()) 2130 Name = *NameOrErr; 2131 else 2132 consumeError(NameOrErr.takeError()); 2133 2134 if (Name == FaultMapSectionName) { 2135 FaultMapSection = Sec; 2136 break; 2137 } 2138 } 2139 2140 outs() << "FaultMap table:\n"; 2141 2142 if (!FaultMapSection.hasValue()) { 2143 outs() << "<not found>\n"; 2144 return; 2145 } 2146 2147 StringRef FaultMapContents = 2148 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName()); 2149 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2150 FaultMapContents.bytes_end()); 2151 2152 outs() << FMP; 2153 } 2154 2155 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 2156 if (O->isELF()) { 2157 printELFFileHeader(O); 2158 printELFDynamicSection(O); 2159 printELFSymbolVersionInfo(O); 2160 return; 2161 } 2162 if (O->isCOFF()) 2163 return printCOFFFileHeader(O); 2164 if (O->isWasm()) 2165 return printWasmFileHeader(O); 2166 if (O->isMachO()) { 2167 printMachOFileHeader(O); 2168 if (!OnlyFirst) 2169 printMachOLoadCommands(O); 2170 return; 2171 } 2172 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2173 } 2174 2175 static void printFileHeaders(const ObjectFile *O) { 2176 if (!O->isELF() && !O->isCOFF()) 2177 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2178 2179 Triple::ArchType AT = O->getArch(); 2180 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2181 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 2182 2183 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2184 outs() << "start address: " 2185 << "0x" << format(Fmt.data(), Address) << "\n\n"; 2186 } 2187 2188 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2189 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2190 if (!ModeOrErr) { 2191 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 2192 consumeError(ModeOrErr.takeError()); 2193 return; 2194 } 2195 sys::fs::perms Mode = ModeOrErr.get(); 2196 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2197 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2198 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2199 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2200 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2201 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2202 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2203 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2204 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2205 2206 outs() << " "; 2207 2208 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 2209 unwrapOrError(C.getGID(), Filename), 2210 unwrapOrError(C.getRawSize(), Filename)); 2211 2212 StringRef RawLastModified = C.getRawLastModified(); 2213 unsigned Seconds; 2214 if (RawLastModified.getAsInteger(10, Seconds)) 2215 outs() << "(date: \"" << RawLastModified 2216 << "\" contains non-decimal chars) "; 2217 else { 2218 // Since ctime(3) returns a 26 character string of the form: 2219 // "Sun Sep 16 01:03:52 1973\n\0" 2220 // just print 24 characters. 2221 time_t t = Seconds; 2222 outs() << format("%.24s ", ctime(&t)); 2223 } 2224 2225 StringRef Name = ""; 2226 Expected<StringRef> NameOrErr = C.getName(); 2227 if (!NameOrErr) { 2228 consumeError(NameOrErr.takeError()); 2229 Name = unwrapOrError(C.getRawName(), Filename); 2230 } else { 2231 Name = NameOrErr.get(); 2232 } 2233 outs() << Name << "\n"; 2234 } 2235 2236 // For ELF only now. 2237 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) { 2238 if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) { 2239 if (Elf->getEType() != ELF::ET_REL) 2240 return true; 2241 } 2242 return false; 2243 } 2244 2245 static void checkForInvalidStartStopAddress(ObjectFile *Obj, 2246 uint64_t Start, uint64_t Stop) { 2247 if (!shouldWarnForInvalidStartStopAddress(Obj)) 2248 return; 2249 2250 for (const SectionRef &Section : Obj->sections()) 2251 if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) { 2252 uint64_t BaseAddr = Section.getAddress(); 2253 uint64_t Size = Section.getSize(); 2254 if ((Start < BaseAddr + Size) && Stop > BaseAddr) 2255 return; 2256 } 2257 2258 if (StartAddress.getNumOccurrences() == 0) 2259 reportWarning("no section has address less than 0x" + 2260 Twine::utohexstr(Stop) + " specified by --stop-address", 2261 Obj->getFileName()); 2262 else if (StopAddress.getNumOccurrences() == 0) 2263 reportWarning("no section has address greater than or equal to 0x" + 2264 Twine::utohexstr(Start) + " specified by --start-address", 2265 Obj->getFileName()); 2266 else 2267 reportWarning("no section overlaps the range [0x" + 2268 Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) + 2269 ") specified by --start-address/--stop-address", 2270 Obj->getFileName()); 2271 } 2272 2273 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 2274 const Archive::Child *C = nullptr) { 2275 // Avoid other output when using a raw option. 2276 if (!RawClangAST) { 2277 outs() << '\n'; 2278 if (A) 2279 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 2280 else 2281 outs() << O->getFileName(); 2282 outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n\n"; 2283 } 2284 2285 if (StartAddress.getNumOccurrences() || StopAddress.getNumOccurrences()) 2286 checkForInvalidStartStopAddress(O, StartAddress, StopAddress); 2287 2288 // Note: the order here matches GNU objdump for compatability. 2289 StringRef ArchiveName = A ? A->getFileName() : ""; 2290 if (ArchiveHeaders && !MachOOpt && C) 2291 printArchiveChild(ArchiveName, *C); 2292 if (FileHeaders) 2293 printFileHeaders(O); 2294 if (PrivateHeaders || FirstPrivateHeader) 2295 printPrivateFileHeaders(O, FirstPrivateHeader); 2296 if (SectionHeaders) 2297 printSectionHeaders(O); 2298 if (SymbolTable) 2299 printSymbolTable(O, ArchiveName); 2300 if (DynamicSymbolTable) 2301 printSymbolTable(O, ArchiveName, /*ArchitectureName=*/"", 2302 /*DumpDynamic=*/true); 2303 if (DwarfDumpType != DIDT_Null) { 2304 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2305 // Dump the complete DWARF structure. 2306 DIDumpOptions DumpOpts; 2307 DumpOpts.DumpType = DwarfDumpType; 2308 DICtx->dump(outs(), DumpOpts); 2309 } 2310 if (Relocations && !Disassemble) 2311 printRelocations(O); 2312 if (DynamicRelocations) 2313 printDynamicRelocations(O); 2314 if (SectionContents) 2315 printSectionContents(O); 2316 if (Disassemble) 2317 disassembleObject(O, Relocations); 2318 if (UnwindInfo) 2319 printUnwindInfo(O); 2320 2321 // Mach-O specific options: 2322 if (ExportsTrie) 2323 printExportsTrie(O); 2324 if (Rebase) 2325 printRebaseTable(O); 2326 if (Bind) 2327 printBindTable(O); 2328 if (LazyBind) 2329 printLazyBindTable(O); 2330 if (WeakBind) 2331 printWeakBindTable(O); 2332 2333 // Other special sections: 2334 if (RawClangAST) 2335 printRawClangAST(O); 2336 if (FaultMapSection) 2337 printFaultMaps(O); 2338 } 2339 2340 static void dumpObject(const COFFImportFile *I, const Archive *A, 2341 const Archive::Child *C = nullptr) { 2342 StringRef ArchiveName = A ? A->getFileName() : ""; 2343 2344 // Avoid other output when using a raw option. 2345 if (!RawClangAST) 2346 outs() << '\n' 2347 << ArchiveName << "(" << I->getFileName() << ")" 2348 << ":\tfile format COFF-import-file" 2349 << "\n\n"; 2350 2351 if (ArchiveHeaders && !MachOOpt && C) 2352 printArchiveChild(ArchiveName, *C); 2353 if (SymbolTable) 2354 printCOFFSymbolTable(I); 2355 } 2356 2357 /// Dump each object file in \a a; 2358 static void dumpArchive(const Archive *A) { 2359 Error Err = Error::success(); 2360 unsigned I = -1; 2361 for (auto &C : A->children(Err)) { 2362 ++I; 2363 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2364 if (!ChildOrErr) { 2365 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2366 reportError(std::move(E), getFileNameForError(C, I), A->getFileName()); 2367 continue; 2368 } 2369 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2370 dumpObject(O, A, &C); 2371 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2372 dumpObject(I, A, &C); 2373 else 2374 reportError(errorCodeToError(object_error::invalid_file_type), 2375 A->getFileName()); 2376 } 2377 if (Err) 2378 reportError(std::move(Err), A->getFileName()); 2379 } 2380 2381 /// Open file and figure out how to dump it. 2382 static void dumpInput(StringRef file) { 2383 // If we are using the Mach-O specific object file parser, then let it parse 2384 // the file and process the command line options. So the -arch flags can 2385 // be used to select specific slices, etc. 2386 if (MachOOpt) { 2387 parseInputMachO(file); 2388 return; 2389 } 2390 2391 // Attempt to open the binary. 2392 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2393 Binary &Binary = *OBinary.getBinary(); 2394 2395 if (Archive *A = dyn_cast<Archive>(&Binary)) 2396 dumpArchive(A); 2397 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2398 dumpObject(O); 2399 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2400 parseInputMachO(UB); 2401 else 2402 reportError(errorCodeToError(object_error::invalid_file_type), file); 2403 } 2404 } // namespace llvm 2405 2406 int main(int argc, char **argv) { 2407 using namespace llvm; 2408 InitLLVM X(argc, argv); 2409 const cl::OptionCategory *OptionFilters[] = {&ObjdumpCat, &MachOCat}; 2410 cl::HideUnrelatedOptions(OptionFilters); 2411 2412 // Initialize targets and assembly printers/parsers. 2413 InitializeAllTargetInfos(); 2414 InitializeAllTargetMCs(); 2415 InitializeAllDisassemblers(); 2416 2417 // Register the target printer for --version. 2418 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2419 2420 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n", nullptr, 2421 /*EnvVar=*/nullptr, 2422 /*LongOptionsUseDoubleDash=*/true); 2423 2424 if (StartAddress >= StopAddress) 2425 reportCmdLineError("start address should be less than stop address"); 2426 2427 ToolName = argv[0]; 2428 2429 // Defaults to a.out if no filenames specified. 2430 if (InputFilenames.empty()) 2431 InputFilenames.push_back("a.out"); 2432 2433 if (AllHeaders) 2434 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2435 SectionHeaders = SymbolTable = true; 2436 2437 if (DisassembleAll || PrintSource || PrintLines || 2438 !DisassembleSymbols.empty()) 2439 Disassemble = true; 2440 2441 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null && 2442 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST && 2443 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable && 2444 !DynamicSymbolTable && !UnwindInfo && !FaultMapSection && 2445 !(MachOOpt && 2446 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie || 2447 FirstPrivateHeader || IndirectSymbols || InfoPlist || LazyBind || 2448 LinkOptHints || ObjcMetaData || Rebase || UniversalHeaders || 2449 WeakBind || !FilterSections.empty()))) { 2450 cl::PrintHelpMessage(); 2451 return 2; 2452 } 2453 2454 DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end()); 2455 2456 llvm::for_each(InputFilenames, dumpInput); 2457 2458 warnOnNoMatchForSections(); 2459 2460 return EXIT_SUCCESS; 2461 } 2462