1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This program is a utility that works like binutils "objdump", that is, it 10 // dumps out a plethora of information about an object file depending on the 11 // flags. 12 // 13 // The flags and output of this program should be near identical to those of 14 // binutils objdump. 15 // 16 //===----------------------------------------------------------------------===// 17 18 #include "llvm-objdump.h" 19 #include "llvm/ADT/Optional.h" 20 #include "llvm/ADT/STLExtras.h" 21 #include "llvm/ADT/StringExtras.h" 22 #include "llvm/ADT/StringSet.h" 23 #include "llvm/ADT/Triple.h" 24 #include "llvm/CodeGen/FaultMaps.h" 25 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 26 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 27 #include "llvm/Demangle/Demangle.h" 28 #include "llvm/MC/MCAsmInfo.h" 29 #include "llvm/MC/MCContext.h" 30 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 31 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 32 #include "llvm/MC/MCInst.h" 33 #include "llvm/MC/MCInstPrinter.h" 34 #include "llvm/MC/MCInstrAnalysis.h" 35 #include "llvm/MC/MCInstrInfo.h" 36 #include "llvm/MC/MCObjectFileInfo.h" 37 #include "llvm/MC/MCRegisterInfo.h" 38 #include "llvm/MC/MCSubtargetInfo.h" 39 #include "llvm/Object/Archive.h" 40 #include "llvm/Object/COFF.h" 41 #include "llvm/Object/COFFImportFile.h" 42 #include "llvm/Object/ELFObjectFile.h" 43 #include "llvm/Object/MachO.h" 44 #include "llvm/Object/MachOUniversal.h" 45 #include "llvm/Object/ObjectFile.h" 46 #include "llvm/Object/Wasm.h" 47 #include "llvm/Support/Casting.h" 48 #include "llvm/Support/CommandLine.h" 49 #include "llvm/Support/Debug.h" 50 #include "llvm/Support/Errc.h" 51 #include "llvm/Support/FileSystem.h" 52 #include "llvm/Support/Format.h" 53 #include "llvm/Support/GraphWriter.h" 54 #include "llvm/Support/Host.h" 55 #include "llvm/Support/InitLLVM.h" 56 #include "llvm/Support/MemoryBuffer.h" 57 #include "llvm/Support/SourceMgr.h" 58 #include "llvm/Support/StringSaver.h" 59 #include "llvm/Support/TargetRegistry.h" 60 #include "llvm/Support/TargetSelect.h" 61 #include "llvm/Support/WithColor.h" 62 #include "llvm/Support/raw_ostream.h" 63 #include <algorithm> 64 #include <cctype> 65 #include <cstring> 66 #include <system_error> 67 #include <unordered_map> 68 #include <utility> 69 70 using namespace llvm; 71 using namespace object; 72 73 cl::opt<unsigned long long> AdjustVMA( 74 "adjust-vma", 75 cl::desc("Increase the displayed address by the specified offset"), 76 cl::value_desc("offset"), cl::init(0)); 77 78 cl::opt<bool> 79 llvm::AllHeaders("all-headers", 80 cl::desc("Display all available header information")); 81 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 82 cl::NotHidden, cl::Grouping, 83 cl::aliasopt(AllHeaders)); 84 85 static cl::list<std::string> 86 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 87 88 cl::opt<bool> 89 llvm::Disassemble("disassemble", 90 cl::desc("Display assembler mnemonics for the machine instructions")); 91 static cl::alias Disassembled("d", cl::desc("Alias for --disassemble"), 92 cl::NotHidden, cl::Grouping, 93 cl::aliasopt(Disassemble)); 94 95 cl::opt<bool> 96 llvm::DisassembleAll("disassemble-all", 97 cl::desc("Display assembler mnemonics for the machine instructions")); 98 static cl::alias DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 99 cl::NotHidden, cl::Grouping, 100 cl::aliasopt(DisassembleAll)); 101 102 cl::opt<bool> llvm::Demangle("demangle", cl::desc("Demangle symbols names"), 103 cl::init(false)); 104 105 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 106 cl::NotHidden, cl::Grouping, 107 cl::aliasopt(llvm::Demangle)); 108 109 static cl::list<std::string> 110 DisassembleFunctions("disassemble-functions", 111 cl::CommaSeparated, 112 cl::desc("List of functions to disassemble")); 113 static StringSet<> DisasmFuncsSet; 114 115 cl::opt<bool> 116 llvm::Relocations("reloc", 117 cl::desc("Display the relocation entries in the file")); 118 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 119 cl::NotHidden, cl::Grouping, 120 cl::aliasopt(llvm::Relocations)); 121 122 cl::opt<bool> 123 llvm::DynamicRelocations("dynamic-reloc", 124 cl::desc("Display the dynamic relocation entries in the file")); 125 static cl::alias DynamicRelocationsd("R", cl::desc("Alias for --dynamic-reloc"), 126 cl::NotHidden, cl::Grouping, 127 cl::aliasopt(DynamicRelocations)); 128 129 cl::opt<bool> 130 llvm::SectionContents("full-contents", 131 cl::desc("Display the content of each section")); 132 static cl::alias SectionContentsShort("s", 133 cl::desc("Alias for --full-contents"), 134 cl::NotHidden, cl::Grouping, 135 cl::aliasopt(SectionContents)); 136 137 cl::opt<bool> llvm::SymbolTable("syms", cl::desc("Display the symbol table")); 138 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 139 cl::NotHidden, cl::Grouping, 140 cl::aliasopt(llvm::SymbolTable)); 141 142 cl::opt<bool> 143 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 144 145 cl::opt<bool> 146 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 147 148 cl::opt<bool> 149 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 150 151 cl::opt<bool> 152 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 153 154 cl::opt<bool> 155 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 156 157 cl::opt<bool> 158 llvm::RawClangAST("raw-clang-ast", 159 cl::desc("Dump the raw binary contents of the clang AST section")); 160 161 static cl::opt<bool> 162 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 163 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 164 cl::Grouping, cl::aliasopt(MachOOpt)); 165 166 cl::opt<std::string> 167 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 168 "see -version for available targets")); 169 170 cl::opt<std::string> 171 llvm::MCPU("mcpu", 172 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 173 cl::value_desc("cpu-name"), 174 cl::init("")); 175 176 cl::opt<std::string> 177 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 178 "see -version for available targets")); 179 180 cl::opt<bool> 181 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 182 "headers for each section.")); 183 static cl::alias SectionHeadersShort("headers", 184 cl::desc("Alias for --section-headers"), 185 cl::NotHidden, 186 cl::aliasopt(SectionHeaders)); 187 static cl::alias SectionHeadersShorter("h", 188 cl::desc("Alias for --section-headers"), 189 cl::NotHidden, cl::Grouping, 190 cl::aliasopt(SectionHeaders)); 191 192 static cl::opt<bool> 193 ShowLMA("show-lma", 194 cl::desc("Display LMA column when dumping ELF section headers")); 195 196 cl::list<std::string> 197 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 198 "With -macho dump segment,section")); 199 cl::alias static FilterSectionsj("j", cl::desc("Alias for --section"), 200 cl::NotHidden, cl::Grouping, cl::Prefix, 201 cl::aliasopt(llvm::FilterSections)); 202 203 cl::list<std::string> 204 llvm::MAttrs("mattr", 205 cl::CommaSeparated, 206 cl::desc("Target specific attributes"), 207 cl::value_desc("a1,+a2,-a3,...")); 208 209 cl::opt<bool> 210 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 211 "instructions, do not print " 212 "the instruction bytes.")); 213 cl::opt<bool> 214 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 215 216 cl::opt<bool> 217 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 218 219 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 220 cl::NotHidden, cl::Grouping, 221 cl::aliasopt(UnwindInfo)); 222 223 cl::opt<bool> 224 llvm::PrivateHeaders("private-headers", 225 cl::desc("Display format specific file headers")); 226 227 cl::opt<bool> 228 llvm::FirstPrivateHeader("private-header", 229 cl::desc("Display only the first format specific file " 230 "header")); 231 232 static cl::alias PrivateHeadersShort("p", 233 cl::desc("Alias for --private-headers"), 234 cl::NotHidden, cl::Grouping, 235 cl::aliasopt(PrivateHeaders)); 236 237 cl::opt<bool> llvm::FileHeaders( 238 "file-headers", 239 cl::desc("Display the contents of the overall file header")); 240 241 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 242 cl::NotHidden, cl::Grouping, 243 cl::aliasopt(FileHeaders)); 244 245 cl::opt<bool> 246 llvm::ArchiveHeaders("archive-headers", 247 cl::desc("Display archive header information")); 248 249 cl::alias ArchiveHeadersShort("a", cl::desc("Alias for --archive-headers"), 250 cl::NotHidden, cl::Grouping, 251 cl::aliasopt(ArchiveHeaders)); 252 253 cl::opt<bool> 254 llvm::PrintImmHex("print-imm-hex", 255 cl::desc("Use hex format for immediate values")); 256 257 cl::opt<bool> PrintFaultMaps("fault-map-section", 258 cl::desc("Display contents of faultmap section")); 259 260 cl::opt<DIDumpType> llvm::DwarfDumpType( 261 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 262 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 263 264 cl::opt<bool> PrintSource( 265 "source", 266 cl::desc( 267 "Display source inlined with disassembly. Implies disassemble object")); 268 269 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), cl::NotHidden, 270 cl::Grouping, cl::aliasopt(PrintSource)); 271 272 cl::opt<bool> PrintLines("line-numbers", 273 cl::desc("Display source line numbers with " 274 "disassembly. Implies disassemble object")); 275 276 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 277 cl::NotHidden, cl::Grouping, 278 cl::aliasopt(PrintLines)); 279 280 cl::opt<unsigned long long> 281 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 282 cl::value_desc("address"), cl::init(0)); 283 cl::opt<unsigned long long> 284 StopAddress("stop-address", 285 cl::desc("Stop disassembly at address"), 286 cl::value_desc("address"), cl::init(UINT64_MAX)); 287 288 cl::opt<bool> DisassembleZeroes( 289 "disassemble-zeroes", 290 cl::desc("Do not skip blocks of zeroes when disassembling")); 291 cl::alias DisassembleZeroesShort("z", 292 cl::desc("Alias for --disassemble-zeroes"), 293 cl::NotHidden, cl::Grouping, 294 cl::aliasopt(DisassembleZeroes)); 295 296 static cl::list<std::string> 297 DisassemblerOptions("disassembler-options", 298 cl::desc("Pass target specific disassembler options"), 299 cl::value_desc("options"), cl::CommaSeparated); 300 static cl::alias 301 DisassemblerOptionsShort("M", cl::desc("Alias for --disassembler-options"), 302 cl::NotHidden, cl::Grouping, cl::Prefix, 303 cl::CommaSeparated, 304 cl::aliasopt(DisassemblerOptions)); 305 306 static StringRef ToolName; 307 308 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 309 310 SectionFilter llvm::ToolSectionFilter(llvm::object::ObjectFile const &O) { 311 return SectionFilter( 312 [](llvm::object::SectionRef const &S) { 313 if (FilterSections.empty()) 314 return true; 315 llvm::StringRef String; 316 std::error_code error = S.getName(String); 317 if (error) 318 return false; 319 return is_contained(FilterSections, String); 320 }, 321 O); 322 } 323 324 void llvm::error(std::error_code EC) { 325 if (!EC) 326 return; 327 WithColor::error(errs(), ToolName) 328 << "reading file: " << EC.message() << ".\n"; 329 errs().flush(); 330 exit(1); 331 } 332 333 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 334 WithColor::error(errs(), ToolName) << Message << ".\n"; 335 errs().flush(); 336 exit(1); 337 } 338 339 void llvm::warn(StringRef Message) { 340 WithColor::warning(errs(), ToolName) << Message << ".\n"; 341 errs().flush(); 342 } 343 344 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 345 Twine Message) { 346 WithColor::error(errs(), ToolName) 347 << "'" << File << "': " << Message << ".\n"; 348 exit(1); 349 } 350 351 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 352 std::error_code EC) { 353 assert(EC); 354 WithColor::error(errs(), ToolName) 355 << "'" << File << "': " << EC.message() << ".\n"; 356 exit(1); 357 } 358 359 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 360 llvm::Error E) { 361 assert(E); 362 std::string Buf; 363 raw_string_ostream OS(Buf); 364 logAllUnhandledErrors(std::move(E), OS); 365 OS.flush(); 366 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 367 exit(1); 368 } 369 370 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 371 StringRef FileName, 372 llvm::Error E, 373 StringRef ArchitectureName) { 374 assert(E); 375 WithColor::error(errs(), ToolName); 376 if (ArchiveName != "") 377 errs() << ArchiveName << "(" << FileName << ")"; 378 else 379 errs() << "'" << FileName << "'"; 380 if (!ArchitectureName.empty()) 381 errs() << " (for architecture " << ArchitectureName << ")"; 382 std::string Buf; 383 raw_string_ostream OS(Buf); 384 logAllUnhandledErrors(std::move(E), OS); 385 OS.flush(); 386 errs() << ": " << Buf; 387 exit(1); 388 } 389 390 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 391 const object::Archive::Child &C, 392 llvm::Error E, 393 StringRef ArchitectureName) { 394 Expected<StringRef> NameOrErr = C.getName(); 395 // TODO: if we have a error getting the name then it would be nice to print 396 // the index of which archive member this is and or its offset in the 397 // archive instead of "???" as the name. 398 if (!NameOrErr) { 399 consumeError(NameOrErr.takeError()); 400 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 401 } else 402 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 403 ArchitectureName); 404 } 405 406 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 407 // Figure out the target triple. 408 llvm::Triple TheTriple("unknown-unknown-unknown"); 409 if (TripleName.empty()) { 410 if (Obj) 411 TheTriple = Obj->makeTriple(); 412 } else { 413 TheTriple.setTriple(Triple::normalize(TripleName)); 414 415 // Use the triple, but also try to combine with ARM build attributes. 416 if (Obj) { 417 auto Arch = Obj->getArch(); 418 if (Arch == Triple::arm || Arch == Triple::armeb) 419 Obj->setARMSubArch(TheTriple); 420 } 421 } 422 423 // Get the target specific parser. 424 std::string Error; 425 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 426 Error); 427 if (!TheTarget) { 428 if (Obj) 429 report_error(Obj->getFileName(), "can't find target: " + Error); 430 else 431 error("can't find target: " + Error); 432 } 433 434 // Update the triple name and return the found target. 435 TripleName = TheTriple.getTriple(); 436 return TheTarget; 437 } 438 439 bool llvm::isRelocAddressLess(RelocationRef A, RelocationRef B) { 440 return A.getOffset() < B.getOffset(); 441 } 442 443 static std::error_code getRelocationValueString(const RelocationRef &Rel, 444 SmallVectorImpl<char> &Result) { 445 const ObjectFile *Obj = Rel.getObject(); 446 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 447 return getELFRelocationValueString(ELF, Rel, Result); 448 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 449 return getCOFFRelocationValueString(COFF, Rel, Result); 450 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 451 return getWasmRelocationValueString(Wasm, Rel, Result); 452 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 453 return getMachORelocationValueString(MachO, Rel, Result); 454 llvm_unreachable("unknown object file format"); 455 } 456 457 /// Indicates whether this relocation should hidden when listing 458 /// relocations, usually because it is the trailing part of a multipart 459 /// relocation that will be printed as part of the leading relocation. 460 static bool getHidden(RelocationRef RelRef) { 461 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 462 if (!MachO) 463 return false; 464 465 unsigned Arch = MachO->getArch(); 466 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 467 uint64_t Type = MachO->getRelocationType(Rel); 468 469 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 470 // is always hidden. 471 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 472 return Type == MachO::GENERIC_RELOC_PAIR; 473 474 if (Arch == Triple::x86_64) { 475 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 476 // an X86_64_RELOC_SUBTRACTOR. 477 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 478 DataRefImpl RelPrev = Rel; 479 RelPrev.d.a--; 480 uint64_t PrevType = MachO->getRelocationType(RelPrev); 481 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 482 return true; 483 } 484 } 485 486 return false; 487 } 488 489 namespace { 490 class SourcePrinter { 491 protected: 492 DILineInfo OldLineInfo; 493 const ObjectFile *Obj = nullptr; 494 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 495 // File name to file contents of source 496 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 497 // Mark the line endings of the cached source 498 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 499 500 private: 501 bool cacheSource(const DILineInfo& LineInfoFile); 502 503 public: 504 SourcePrinter() = default; 505 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 506 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 507 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 508 DefaultArch); 509 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 510 } 511 virtual ~SourcePrinter() = default; 512 virtual void printSourceLine(raw_ostream &OS, 513 object::SectionedAddress Address, 514 StringRef Delimiter = "; "); 515 }; 516 517 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 518 std::unique_ptr<MemoryBuffer> Buffer; 519 if (LineInfo.Source) { 520 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 521 } else { 522 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 523 if (!BufferOrError) 524 return false; 525 Buffer = std::move(*BufferOrError); 526 } 527 // Chomp the file to get lines 528 size_t BufferSize = Buffer->getBufferSize(); 529 const char *BufferStart = Buffer->getBufferStart(); 530 for (const char *Start = BufferStart, *End = BufferStart; 531 End < BufferStart + BufferSize; End++) 532 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 533 (*End == '\r' && *(End + 1) == '\n')) { 534 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 535 if (*End == '\r') 536 End++; 537 Start = End + 1; 538 } 539 SourceCache[LineInfo.FileName] = std::move(Buffer); 540 return true; 541 } 542 543 void SourcePrinter::printSourceLine(raw_ostream &OS, 544 object::SectionedAddress Address, 545 StringRef Delimiter) { 546 if (!Symbolizer) 547 return; 548 DILineInfo LineInfo = DILineInfo(); 549 auto ExpectedLineInfo = 550 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 551 if (!ExpectedLineInfo) 552 consumeError(ExpectedLineInfo.takeError()); 553 else 554 LineInfo = *ExpectedLineInfo; 555 556 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 557 LineInfo.Line == 0) 558 return; 559 560 if (PrintLines) 561 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 562 if (PrintSource) { 563 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 564 if (!cacheSource(LineInfo)) 565 return; 566 auto FileBuffer = SourceCache.find(LineInfo.FileName); 567 if (FileBuffer != SourceCache.end()) { 568 auto LineBuffer = LineCache.find(LineInfo.FileName); 569 if (LineBuffer != LineCache.end()) { 570 if (LineInfo.Line > LineBuffer->second.size()) 571 return; 572 // Vector begins at 0, line numbers are non-zero 573 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 574 << "\n"; 575 } 576 } 577 } 578 OldLineInfo = LineInfo; 579 } 580 581 static bool isArmElf(const ObjectFile *Obj) { 582 return (Obj->isELF() && 583 (Obj->getArch() == Triple::aarch64 || 584 Obj->getArch() == Triple::aarch64_be || 585 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 586 Obj->getArch() == Triple::thumb || 587 Obj->getArch() == Triple::thumbeb)); 588 } 589 590 static void printRelocation(const RelocationRef &Rel, uint64_t Address, 591 uint8_t AddrSize) { 592 StringRef Fmt = 593 AddrSize > 4 ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 594 SmallString<16> Name; 595 SmallString<32> Val; 596 Rel.getTypeName(Name); 597 error(getRelocationValueString(Rel, Val)); 598 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n"; 599 } 600 601 class PrettyPrinter { 602 public: 603 virtual ~PrettyPrinter() = default; 604 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 605 ArrayRef<uint8_t> Bytes, 606 object::SectionedAddress Address, raw_ostream &OS, 607 StringRef Annot, MCSubtargetInfo const &STI, 608 SourcePrinter *SP, 609 std::vector<RelocationRef> *Rels = nullptr) { 610 if (SP && (PrintSource || PrintLines)) 611 SP->printSourceLine(OS, Address); 612 if (!NoLeadingAddr) 613 OS << format("%8" PRIx64 ":", Address.Address); 614 if (!NoShowRawInsn) { 615 OS << "\t"; 616 dumpBytes(Bytes, OS); 617 } 618 if (MI) 619 IP.printInst(MI, OS, "", STI); 620 else 621 OS << " <unknown>"; 622 } 623 }; 624 PrettyPrinter PrettyPrinterInst; 625 class HexagonPrettyPrinter : public PrettyPrinter { 626 public: 627 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 628 raw_ostream &OS) { 629 uint32_t opcode = 630 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 631 if (!NoLeadingAddr) 632 OS << format("%8" PRIx64 ":", Address); 633 if (!NoShowRawInsn) { 634 OS << "\t"; 635 dumpBytes(Bytes.slice(0, 4), OS); 636 OS << format("%08" PRIx32, opcode); 637 } 638 } 639 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 640 object::SectionedAddress Address, raw_ostream &OS, 641 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 642 std::vector<RelocationRef> *Rels) override { 643 if (SP && (PrintSource || PrintLines)) 644 SP->printSourceLine(OS, Address, ""); 645 if (!MI) { 646 printLead(Bytes, Address.Address, OS); 647 OS << " <unknown>"; 648 return; 649 } 650 std::string Buffer; 651 { 652 raw_string_ostream TempStream(Buffer); 653 IP.printInst(MI, TempStream, "", STI); 654 } 655 StringRef Contents(Buffer); 656 // Split off bundle attributes 657 auto PacketBundle = Contents.rsplit('\n'); 658 // Split off first instruction from the rest 659 auto HeadTail = PacketBundle.first.split('\n'); 660 auto Preamble = " { "; 661 auto Separator = ""; 662 663 // Hexagon's packets require relocations to be inline rather than 664 // clustered at the end of the packet. 665 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 666 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 667 auto PrintReloc = [&]() -> void { 668 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 669 if (RelCur->getOffset() == Address.Address) { 670 printRelocation(*RelCur, Address.Address, 4); 671 return; 672 } 673 ++RelCur; 674 } 675 }; 676 677 while (!HeadTail.first.empty()) { 678 OS << Separator; 679 Separator = "\n"; 680 if (SP && (PrintSource || PrintLines)) 681 SP->printSourceLine(OS, Address, ""); 682 printLead(Bytes, Address.Address, OS); 683 OS << Preamble; 684 Preamble = " "; 685 StringRef Inst; 686 auto Duplex = HeadTail.first.split('\v'); 687 if (!Duplex.second.empty()) { 688 OS << Duplex.first; 689 OS << "; "; 690 Inst = Duplex.second; 691 } 692 else 693 Inst = HeadTail.first; 694 OS << Inst; 695 HeadTail = HeadTail.second.split('\n'); 696 if (HeadTail.first.empty()) 697 OS << " } " << PacketBundle.second; 698 PrintReloc(); 699 Bytes = Bytes.slice(4); 700 Address.Address += 4; 701 } 702 } 703 }; 704 HexagonPrettyPrinter HexagonPrettyPrinterInst; 705 706 class AMDGCNPrettyPrinter : public PrettyPrinter { 707 public: 708 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 709 object::SectionedAddress Address, raw_ostream &OS, 710 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 711 std::vector<RelocationRef> *Rels) override { 712 if (SP && (PrintSource || PrintLines)) 713 SP->printSourceLine(OS, Address); 714 715 typedef support::ulittle32_t U32; 716 717 if (MI) { 718 SmallString<40> InstStr; 719 raw_svector_ostream IS(InstStr); 720 721 IP.printInst(MI, IS, "", STI); 722 723 OS << left_justify(IS.str(), 60); 724 } else { 725 // an unrecognized encoding - this is probably data so represent it 726 // using the .long directive, or .byte directive if fewer than 4 bytes 727 // remaining 728 if (Bytes.size() >= 4) { 729 OS << format("\t.long 0x%08" PRIx32 " ", 730 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 731 OS.indent(42); 732 } else { 733 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 734 for (unsigned int i = 1; i < Bytes.size(); i++) 735 OS << format(", 0x%02" PRIx8, Bytes[i]); 736 OS.indent(55 - (6 * Bytes.size())); 737 } 738 } 739 740 OS << format("// %012" PRIX64 ": ", Address.Address); 741 if (Bytes.size() >=4) { 742 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 743 Bytes.size() / sizeof(U32))) 744 // D should be explicitly casted to uint32_t here as it is passed 745 // by format to snprintf as vararg. 746 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 747 } else { 748 for (unsigned int i = 0; i < Bytes.size(); i++) 749 OS << format("%02" PRIX8 " ", Bytes[i]); 750 } 751 752 if (!Annot.empty()) 753 OS << "// " << Annot; 754 } 755 }; 756 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 757 758 class BPFPrettyPrinter : public PrettyPrinter { 759 public: 760 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 761 object::SectionedAddress Address, raw_ostream &OS, 762 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 763 std::vector<RelocationRef> *Rels) override { 764 if (SP && (PrintSource || PrintLines)) 765 SP->printSourceLine(OS, Address); 766 if (!NoLeadingAddr) 767 OS << format("%8" PRId64 ":", Address.Address / 8); 768 if (!NoShowRawInsn) { 769 OS << "\t"; 770 dumpBytes(Bytes, OS); 771 } 772 if (MI) 773 IP.printInst(MI, OS, "", STI); 774 else 775 OS << " <unknown>"; 776 } 777 }; 778 BPFPrettyPrinter BPFPrettyPrinterInst; 779 780 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 781 switch(Triple.getArch()) { 782 default: 783 return PrettyPrinterInst; 784 case Triple::hexagon: 785 return HexagonPrettyPrinterInst; 786 case Triple::amdgcn: 787 return AMDGCNPrettyPrinterInst; 788 case Triple::bpfel: 789 case Triple::bpfeb: 790 return BPFPrettyPrinterInst; 791 } 792 } 793 } 794 795 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 796 assert(Obj->isELF()); 797 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 798 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 799 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 800 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 801 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 802 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 803 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 804 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 805 llvm_unreachable("Unsupported binary format"); 806 } 807 808 template <class ELFT> static void 809 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 810 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 811 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 812 uint8_t SymbolType = Symbol.getELFType(); 813 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 814 continue; 815 816 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 817 if (!AddressOrErr) 818 report_error(Obj->getFileName(), AddressOrErr.takeError()); 819 820 Expected<StringRef> Name = Symbol.getName(); 821 if (!Name) 822 report_error(Obj->getFileName(), Name.takeError()); 823 if (Name->empty()) 824 continue; 825 826 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 827 if (!SectionOrErr) 828 report_error(Obj->getFileName(), SectionOrErr.takeError()); 829 section_iterator SecI = *SectionOrErr; 830 if (SecI == Obj->section_end()) 831 continue; 832 833 AllSymbols[*SecI].emplace_back(*AddressOrErr, *Name, SymbolType); 834 } 835 } 836 837 static void 838 addDynamicElfSymbols(const ObjectFile *Obj, 839 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 840 assert(Obj->isELF()); 841 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 842 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 843 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 844 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 845 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 846 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 847 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 848 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 849 else 850 llvm_unreachable("Unsupported binary format"); 851 } 852 853 static void addPltEntries(const ObjectFile *Obj, 854 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 855 StringSaver &Saver) { 856 Optional<SectionRef> Plt = None; 857 for (const SectionRef &Section : Obj->sections()) { 858 StringRef Name; 859 if (Section.getName(Name)) 860 continue; 861 if (Name == ".plt") 862 Plt = Section; 863 } 864 if (!Plt) 865 return; 866 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 867 for (auto PltEntry : ElfObj->getPltAddresses()) { 868 SymbolRef Symbol(PltEntry.first, ElfObj); 869 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 870 871 Expected<StringRef> NameOrErr = Symbol.getName(); 872 if (!NameOrErr) 873 report_error(Obj->getFileName(), NameOrErr.takeError()); 874 if (NameOrErr->empty()) 875 continue; 876 StringRef Name = Saver.save((*NameOrErr + "@plt").str()); 877 878 AllSymbols[*Plt].emplace_back(PltEntry.second, Name, SymbolType); 879 } 880 } 881 } 882 883 // Normally the disassembly output will skip blocks of zeroes. This function 884 // returns the number of zero bytes that can be skipped when dumping the 885 // disassembly of the instructions in Buf. 886 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 887 // Find the number of leading zeroes. 888 size_t N = 0; 889 while (N < Buf.size() && !Buf[N]) 890 ++N; 891 892 // We may want to skip blocks of zero bytes, but unless we see 893 // at least 8 of them in a row. 894 if (N < 8) 895 return 0; 896 897 // We skip zeroes in multiples of 4 because do not want to truncate an 898 // instruction if it starts with a zero byte. 899 return N & ~0x3; 900 } 901 902 // Returns a map from sections to their relocations. 903 static std::map<SectionRef, std::vector<RelocationRef>> 904 getRelocsMap(llvm::object::ObjectFile const &Obj) { 905 std::map<SectionRef, std::vector<RelocationRef>> Ret; 906 for (const SectionRef &Section : ToolSectionFilter(Obj)) { 907 section_iterator RelSec = Section.getRelocatedSection(); 908 if (RelSec == Obj.section_end()) 909 continue; 910 std::vector<RelocationRef> &V = Ret[*RelSec]; 911 for (const RelocationRef &R : Section.relocations()) 912 V.push_back(R); 913 // Sort relocations by address. 914 llvm::sort(V, isRelocAddressLess); 915 } 916 return Ret; 917 } 918 919 // Used for --adjust-vma to check if address should be adjusted by the 920 // specified value for a given section. 921 // For ELF we do not adjust non-allocatable sections like debug ones, 922 // because they are not loadable. 923 // TODO: implement for other file formats. 924 static bool shouldAdjustVA(const SectionRef &Section) { 925 const ObjectFile *Obj = Section.getObject(); 926 if (isa<object::ELFObjectFileBase>(Obj)) 927 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 928 return false; 929 } 930 931 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 932 MCContext &Ctx, MCDisassembler *DisAsm, 933 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 934 const MCSubtargetInfo *STI, PrettyPrinter &PIP, 935 SourcePrinter &SP, bool InlineRelocs) { 936 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 937 if (InlineRelocs) 938 RelocMap = getRelocsMap(*Obj); 939 940 // Create a mapping from virtual address to symbol name. This is used to 941 // pretty print the symbols while disassembling. 942 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 943 SectionSymbolsTy AbsoluteSymbols; 944 for (const SymbolRef &Symbol : Obj->symbols()) { 945 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 946 if (!AddressOrErr) 947 report_error(Obj->getFileName(), AddressOrErr.takeError()); 948 uint64_t Address = *AddressOrErr; 949 950 Expected<StringRef> Name = Symbol.getName(); 951 if (!Name) 952 report_error(Obj->getFileName(), Name.takeError()); 953 if (Name->empty()) 954 continue; 955 956 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 957 if (!SectionOrErr) 958 report_error(Obj->getFileName(), SectionOrErr.takeError()); 959 960 uint8_t SymbolType = ELF::STT_NOTYPE; 961 if (Obj->isELF()) { 962 SymbolType = getElfSymbolType(Obj, Symbol); 963 if (SymbolType == ELF::STT_SECTION) 964 continue; 965 } 966 967 section_iterator SecI = *SectionOrErr; 968 if (SecI != Obj->section_end()) 969 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 970 else 971 AbsoluteSymbols.emplace_back(Address, *Name, SymbolType); 972 } 973 if (AllSymbols.empty() && Obj->isELF()) 974 addDynamicElfSymbols(Obj, AllSymbols); 975 976 BumpPtrAllocator A; 977 StringSaver Saver(A); 978 addPltEntries(Obj, AllSymbols, Saver); 979 980 // Create a mapping from virtual address to section. 981 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 982 for (SectionRef Sec : Obj->sections()) 983 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 984 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 985 986 // Linked executables (.exe and .dll files) typically don't include a real 987 // symbol table but they might contain an export table. 988 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 989 for (const auto &ExportEntry : COFFObj->export_directories()) { 990 StringRef Name; 991 error(ExportEntry.getSymbolName(Name)); 992 if (Name.empty()) 993 continue; 994 uint32_t RVA; 995 error(ExportEntry.getExportRVA(RVA)); 996 997 uint64_t VA = COFFObj->getImageBase() + RVA; 998 auto Sec = llvm::upper_bound( 999 SectionAddresses, VA, 1000 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1001 return LHS < RHS.first; 1002 }); 1003 if (Sec != SectionAddresses.begin()) { 1004 --Sec; 1005 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1006 } else 1007 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1008 } 1009 } 1010 1011 // Sort all the symbols, this allows us to use a simple binary search to find 1012 // a symbol near an address. 1013 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1014 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1015 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1016 1017 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1018 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1019 continue; 1020 1021 uint64_t SectionAddr = Section.getAddress(); 1022 uint64_t SectSize = Section.getSize(); 1023 if (!SectSize) 1024 continue; 1025 1026 // Get the list of all the symbols in this section. 1027 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1028 std::vector<uint64_t> DataMappingSymsAddr; 1029 std::vector<uint64_t> TextMappingSymsAddr; 1030 if (isArmElf(Obj)) { 1031 for (const auto &Symb : Symbols) { 1032 uint64_t Address = std::get<0>(Symb); 1033 StringRef Name = std::get<1>(Symb); 1034 if (Name.startswith("$d")) 1035 DataMappingSymsAddr.push_back(Address - SectionAddr); 1036 if (Name.startswith("$x")) 1037 TextMappingSymsAddr.push_back(Address - SectionAddr); 1038 if (Name.startswith("$a")) 1039 TextMappingSymsAddr.push_back(Address - SectionAddr); 1040 if (Name.startswith("$t")) 1041 TextMappingSymsAddr.push_back(Address - SectionAddr); 1042 } 1043 } 1044 1045 llvm::sort(DataMappingSymsAddr); 1046 llvm::sort(TextMappingSymsAddr); 1047 1048 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1049 // AMDGPU disassembler uses symbolizer for printing labels 1050 std::unique_ptr<MCRelocationInfo> RelInfo( 1051 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1052 if (RelInfo) { 1053 std::unique_ptr<MCSymbolizer> Symbolizer( 1054 TheTarget->createMCSymbolizer( 1055 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1056 DisAsm->setSymbolizer(std::move(Symbolizer)); 1057 } 1058 } 1059 1060 StringRef SegmentName = ""; 1061 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1062 DataRefImpl DR = Section.getRawDataRefImpl(); 1063 SegmentName = MachO->getSectionFinalSegmentName(DR); 1064 } 1065 StringRef SectionName; 1066 error(Section.getName(SectionName)); 1067 1068 // If the section has no symbol at the start, just insert a dummy one. 1069 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1070 Symbols.insert( 1071 Symbols.begin(), 1072 std::make_tuple(SectionAddr, SectionName, 1073 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1074 } 1075 1076 SmallString<40> Comments; 1077 raw_svector_ostream CommentStream(Comments); 1078 1079 StringRef BytesStr; 1080 error(Section.getContents(BytesStr)); 1081 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef(BytesStr); 1082 1083 uint64_t VMAAdjustment = 0; 1084 if (shouldAdjustVA(Section)) 1085 VMAAdjustment = AdjustVMA; 1086 1087 uint64_t Size; 1088 uint64_t Index; 1089 bool PrintedSection = false; 1090 std::vector<RelocationRef> Rels = RelocMap[Section]; 1091 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1092 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1093 // Disassemble symbol by symbol. 1094 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1095 uint64_t Start = std::get<0>(Symbols[SI]) - SectionAddr; 1096 // The end is either the section end or the beginning of the next 1097 // symbol. 1098 uint64_t End = (SI == SE - 1) 1099 ? SectSize 1100 : std::get<0>(Symbols[SI + 1]) - SectionAddr; 1101 // Don't try to disassemble beyond the end of section contents. 1102 if (End > SectSize) 1103 End = SectSize; 1104 // If this symbol has the same address as the next symbol, then skip it. 1105 if (Start >= End) 1106 continue; 1107 1108 // Check if we need to skip symbol 1109 // Skip if the symbol's data is not between StartAddress and StopAddress 1110 if (End + SectionAddr < StartAddress || 1111 Start + SectionAddr > StopAddress) { 1112 continue; 1113 } 1114 1115 /// Skip if user requested specific symbols and this is not in the list 1116 if (!DisasmFuncsSet.empty() && 1117 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1118 continue; 1119 1120 if (!PrintedSection) { 1121 PrintedSection = true; 1122 outs() << "Disassembly of section "; 1123 if (!SegmentName.empty()) 1124 outs() << SegmentName << ","; 1125 outs() << SectionName << ':'; 1126 } 1127 1128 // Stop disassembly at the stop address specified 1129 if (End + SectionAddr > StopAddress) 1130 End = StopAddress - SectionAddr; 1131 1132 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1133 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1134 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1135 Start += 256; 1136 } 1137 if (SI == SE - 1 || 1138 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1139 // cut trailing zeroes at the end of kernel 1140 // cut up to 256 bytes 1141 const uint64_t EndAlign = 256; 1142 const auto Limit = End - (std::min)(EndAlign, End - Start); 1143 while (End > Limit && 1144 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1145 End -= 4; 1146 } 1147 } 1148 1149 outs() << '\n'; 1150 if (!NoLeadingAddr) 1151 outs() << format("%016" PRIx64 " ", 1152 SectionAddr + Start + VMAAdjustment); 1153 1154 StringRef SymbolName = std::get<1>(Symbols[SI]); 1155 if (Demangle) 1156 outs() << demangle(SymbolName) << ":\n"; 1157 else 1158 outs() << SymbolName << ":\n"; 1159 1160 // Don't print raw contents of a virtual section. A virtual section 1161 // doesn't have any contents in the file. 1162 if (Section.isVirtual()) { 1163 outs() << "...\n"; 1164 continue; 1165 } 1166 1167 #ifndef NDEBUG 1168 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1169 #else 1170 raw_ostream &DebugOut = nulls(); 1171 #endif 1172 1173 // Some targets (like WebAssembly) have a special prelude at the start 1174 // of each symbol. 1175 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1176 SectionAddr + Start, DebugOut, CommentStream); 1177 Start += Size; 1178 1179 for (Index = Start; Index < End; Index += Size) { 1180 MCInst Inst; 1181 1182 if (Index + SectionAddr < StartAddress || 1183 Index + SectionAddr > StopAddress) { 1184 // skip byte by byte till StartAddress is reached 1185 Size = 1; 1186 continue; 1187 } 1188 // AArch64 ELF binaries can interleave data and text in the 1189 // same section. We rely on the markers introduced to 1190 // understand what we need to dump. If the data marker is within a 1191 // function, it is denoted as a word/short etc 1192 if (isArmElf(Obj) && std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1193 !DisassembleAll && 1194 std::binary_search(DataMappingSymsAddr.begin(), 1195 DataMappingSymsAddr.end(), Index)) { 1196 // Switch to data. 1197 support::endianness Endian = 1198 Obj->isLittleEndian() ? support::little : support::big; 1199 while (Index < End) { 1200 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1201 outs() << "\t"; 1202 if (Index + 4 <= End) { 1203 dumpBytes(Bytes.slice(Index, 4), outs()); 1204 outs() << "\t.word\t" 1205 << format_hex(support::endian::read32(Bytes.data() + Index, 1206 Endian), 1207 10); 1208 Index += 4; 1209 } else if (Index + 2 <= End) { 1210 dumpBytes(Bytes.slice(Index, 2), outs()); 1211 outs() << "\t\t.short\t" 1212 << format_hex(support::endian::read16(Bytes.data() + Index, 1213 Endian), 1214 6); 1215 Index += 2; 1216 } else { 1217 dumpBytes(Bytes.slice(Index, 1), outs()); 1218 outs() << "\t\t.byte\t" << format_hex(Bytes[0], 4); 1219 ++Index; 1220 } 1221 outs() << "\n"; 1222 if (std::binary_search(TextMappingSymsAddr.begin(), 1223 TextMappingSymsAddr.end(), Index)) 1224 break; 1225 } 1226 } 1227 1228 // If there is a data symbol inside an ELF text section and we are only 1229 // disassembling text (applicable all architectures), 1230 // we are in a situation where we must print the data and not 1231 // disassemble it. 1232 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1233 !DisassembleAll && Section.isText()) { 1234 // print out data up to 8 bytes at a time in hex and ascii 1235 uint8_t AsciiData[9] = {'\0'}; 1236 uint8_t Byte; 1237 int NumBytes = 0; 1238 1239 for (Index = Start; Index < End; Index += 1) { 1240 if (((SectionAddr + Index) < StartAddress) || 1241 ((SectionAddr + Index) > StopAddress)) 1242 continue; 1243 if (NumBytes == 0) { 1244 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1245 outs() << "\t"; 1246 } 1247 Byte = Bytes.slice(Index)[0]; 1248 outs() << format(" %02x", Byte); 1249 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1250 1251 uint8_t IndentOffset = 0; 1252 NumBytes++; 1253 if (Index == End - 1 || NumBytes > 8) { 1254 // Indent the space for less than 8 bytes data. 1255 // 2 spaces for byte and one for space between bytes 1256 IndentOffset = 3 * (8 - NumBytes); 1257 for (int Excess = NumBytes; Excess < 8; Excess++) 1258 AsciiData[Excess] = '\0'; 1259 NumBytes = 8; 1260 } 1261 if (NumBytes == 8) { 1262 AsciiData[8] = '\0'; 1263 outs() << std::string(IndentOffset, ' ') << " "; 1264 outs() << reinterpret_cast<char *>(AsciiData); 1265 outs() << '\n'; 1266 NumBytes = 0; 1267 } 1268 } 1269 } 1270 if (Index >= End) 1271 break; 1272 1273 // When -z or --disassemble-zeroes are given we always dissasemble them. 1274 // Otherwise we might want to skip zero bytes we see. 1275 if (!DisassembleZeroes) { 1276 uint64_t MaxOffset = End - Index; 1277 // For -reloc: print zero blocks patched by relocations, so that 1278 // relocations can be shown in the dump. 1279 if (RelCur != RelEnd) 1280 MaxOffset = RelCur->getOffset() - Index; 1281 1282 if (size_t N = 1283 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1284 outs() << "\t\t..." << '\n'; 1285 Index += N; 1286 if (Index >= End) 1287 break; 1288 } 1289 } 1290 1291 // Disassemble a real instruction or a data when disassemble all is 1292 // provided 1293 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1294 SectionAddr + Index, DebugOut, 1295 CommentStream); 1296 if (Size == 0) 1297 Size = 1; 1298 1299 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1300 Bytes.slice(Index, Size), 1301 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, 1302 outs(), "", *STI, &SP, &Rels); 1303 outs() << CommentStream.str(); 1304 Comments.clear(); 1305 1306 // Try to resolve the target of a call, tail call, etc. to a specific 1307 // symbol. 1308 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1309 MIA->isConditionalBranch(Inst))) { 1310 uint64_t Target; 1311 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1312 // In a relocatable object, the target's section must reside in 1313 // the same section as the call instruction or it is accessed 1314 // through a relocation. 1315 // 1316 // In a non-relocatable object, the target may be in any section. 1317 // 1318 // N.B. We don't walk the relocations in the relocatable case yet. 1319 auto *TargetSectionSymbols = &Symbols; 1320 if (!Obj->isRelocatableObject()) { 1321 auto SectionAddress = llvm::upper_bound( 1322 SectionAddresses, Target, 1323 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1324 return LHS < RHS.first; 1325 }); 1326 if (SectionAddress != SectionAddresses.begin()) { 1327 --SectionAddress; 1328 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1329 } else { 1330 TargetSectionSymbols = &AbsoluteSymbols; 1331 } 1332 } 1333 1334 // Find the first symbol in the section whose offset is less than 1335 // or equal to the target. If there isn't a section that contains 1336 // the target, find the nearest preceding absolute symbol. 1337 auto TargetSym = llvm::upper_bound( 1338 *TargetSectionSymbols, Target, 1339 [](uint64_t LHS, 1340 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1341 return LHS < std::get<0>(RHS); 1342 }); 1343 if (TargetSym == TargetSectionSymbols->begin()) { 1344 TargetSectionSymbols = &AbsoluteSymbols; 1345 TargetSym = llvm::upper_bound( 1346 AbsoluteSymbols, Target, 1347 [](uint64_t LHS, 1348 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1349 return LHS < std::get<0>(RHS); 1350 }); 1351 } 1352 if (TargetSym != TargetSectionSymbols->begin()) { 1353 --TargetSym; 1354 uint64_t TargetAddress = std::get<0>(*TargetSym); 1355 StringRef TargetName = std::get<1>(*TargetSym); 1356 outs() << " <" << TargetName; 1357 uint64_t Disp = Target - TargetAddress; 1358 if (Disp) 1359 outs() << "+0x" << Twine::utohexstr(Disp); 1360 outs() << '>'; 1361 } 1362 } 1363 } 1364 outs() << "\n"; 1365 1366 // Hexagon does this in pretty printer 1367 if (Obj->getArch() != Triple::hexagon) { 1368 // Print relocation for instruction. 1369 while (RelCur != RelEnd) { 1370 uint64_t Offset = RelCur->getOffset(); 1371 // If this relocation is hidden, skip it. 1372 if (getHidden(*RelCur) || ((SectionAddr + Offset) < StartAddress)) { 1373 ++RelCur; 1374 continue; 1375 } 1376 1377 // Stop when RelCur's offset is past the current instruction. 1378 if (Offset >= Index + Size) 1379 break; 1380 1381 // When --adjust-vma is used, update the address printed. 1382 if (RelCur->getSymbol() != Obj->symbol_end()) { 1383 Expected<section_iterator> SymSI = 1384 RelCur->getSymbol()->getSection(); 1385 if (SymSI && *SymSI != Obj->section_end() && 1386 (shouldAdjustVA(**SymSI))) 1387 Offset += AdjustVMA; 1388 } 1389 1390 printRelocation(*RelCur, SectionAddr + Offset, 1391 Obj->getBytesInAddress()); 1392 ++RelCur; 1393 } 1394 } 1395 } 1396 } 1397 } 1398 } 1399 1400 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1401 if (StartAddress > StopAddress) 1402 error("Start address should be less than stop address"); 1403 1404 const Target *TheTarget = getTarget(Obj); 1405 1406 // Package up features to be passed to target/subtarget 1407 SubtargetFeatures Features = Obj->getFeatures(); 1408 if (!MAttrs.empty()) 1409 for (unsigned I = 0; I != MAttrs.size(); ++I) 1410 Features.AddFeature(MAttrs[I]); 1411 1412 std::unique_ptr<const MCRegisterInfo> MRI( 1413 TheTarget->createMCRegInfo(TripleName)); 1414 if (!MRI) 1415 report_error(Obj->getFileName(), 1416 "no register info for target " + TripleName); 1417 1418 // Set up disassembler. 1419 std::unique_ptr<const MCAsmInfo> AsmInfo( 1420 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1421 if (!AsmInfo) 1422 report_error(Obj->getFileName(), 1423 "no assembly info for target " + TripleName); 1424 std::unique_ptr<const MCSubtargetInfo> STI( 1425 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1426 if (!STI) 1427 report_error(Obj->getFileName(), 1428 "no subtarget info for target " + TripleName); 1429 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1430 if (!MII) 1431 report_error(Obj->getFileName(), 1432 "no instruction info for target " + TripleName); 1433 MCObjectFileInfo MOFI; 1434 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1435 // FIXME: for now initialize MCObjectFileInfo with default values 1436 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1437 1438 std::unique_ptr<MCDisassembler> DisAsm( 1439 TheTarget->createMCDisassembler(*STI, Ctx)); 1440 if (!DisAsm) 1441 report_error(Obj->getFileName(), 1442 "no disassembler for target " + TripleName); 1443 1444 std::unique_ptr<const MCInstrAnalysis> MIA( 1445 TheTarget->createMCInstrAnalysis(MII.get())); 1446 1447 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1448 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1449 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1450 if (!IP) 1451 report_error(Obj->getFileName(), 1452 "no instruction printer for target " + TripleName); 1453 IP->setPrintImmHex(PrintImmHex); 1454 1455 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1456 SourcePrinter SP(Obj, TheTarget->getName()); 1457 1458 for (StringRef Opt : DisassemblerOptions) 1459 if (!IP->applyTargetSpecificCLOption(Opt)) 1460 error("Unrecognized disassembler option: " + Opt); 1461 1462 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), MIA.get(), IP.get(), 1463 STI.get(), PIP, SP, InlineRelocs); 1464 } 1465 1466 void llvm::printRelocations(const ObjectFile *Obj) { 1467 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1468 "%08" PRIx64; 1469 // Regular objdump doesn't print relocations in non-relocatable object 1470 // files. 1471 if (!Obj->isRelocatableObject()) 1472 return; 1473 1474 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1475 if (Section.relocation_begin() == Section.relocation_end()) 1476 continue; 1477 StringRef SecName; 1478 error(Section.getName(SecName)); 1479 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1480 for (const RelocationRef &Reloc : Section.relocations()) { 1481 uint64_t Address = Reloc.getOffset(); 1482 SmallString<32> RelocName; 1483 SmallString<32> ValueStr; 1484 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1485 continue; 1486 Reloc.getTypeName(RelocName); 1487 error(getRelocationValueString(Reloc, ValueStr)); 1488 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1489 << ValueStr << "\n"; 1490 } 1491 outs() << "\n"; 1492 } 1493 } 1494 1495 void llvm::printDynamicRelocations(const ObjectFile *Obj) { 1496 // For the moment, this option is for ELF only 1497 if (!Obj->isELF()) 1498 return; 1499 1500 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1501 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1502 error("not a dynamic object"); 1503 return; 1504 } 1505 1506 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1507 if (DynRelSec.empty()) 1508 return; 1509 1510 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1511 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1512 for (const SectionRef &Section : DynRelSec) { 1513 if (Section.relocation_begin() == Section.relocation_end()) 1514 continue; 1515 for (const RelocationRef &Reloc : Section.relocations()) { 1516 uint64_t Address = Reloc.getOffset(); 1517 SmallString<32> RelocName; 1518 SmallString<32> ValueStr; 1519 Reloc.getTypeName(RelocName); 1520 error(getRelocationValueString(Reloc, ValueStr)); 1521 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1522 << ValueStr << "\n"; 1523 } 1524 } 1525 } 1526 1527 // Returns true if we need to show LMA column when dumping section headers. We 1528 // show it only when the platform is ELF and either we have at least one section 1529 // whose VMA and LMA are different and/or when --show-lma flag is used. 1530 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1531 if (!Obj->isELF()) 1532 return false; 1533 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1534 if (S.getAddress() != getELFSectionLMA(S)) 1535 return true; 1536 return ShowLMA; 1537 } 1538 1539 void llvm::printSectionHeaders(const ObjectFile *Obj) { 1540 bool HasLMAColumn = shouldDisplayLMA(Obj); 1541 if (HasLMAColumn) 1542 outs() << "Sections:\n" 1543 "Idx Name Size VMA LMA " 1544 "Type\n"; 1545 else 1546 outs() << "Sections:\n" 1547 "Idx Name Size VMA Type\n"; 1548 1549 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1550 StringRef Name; 1551 error(Section.getName(Name)); 1552 uint64_t VMA = Section.getAddress(); 1553 if (shouldAdjustVA(Section)) 1554 VMA += AdjustVMA; 1555 1556 uint64_t Size = Section.getSize(); 1557 bool Text = Section.isText(); 1558 bool Data = Section.isData(); 1559 bool BSS = Section.isBSS(); 1560 std::string Type = (std::string(Text ? "TEXT " : "") + 1561 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1562 1563 if (HasLMAColumn) 1564 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %016" PRIx64 1565 " %s\n", 1566 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1567 VMA, getELFSectionLMA(Section), Type.c_str()); 1568 else 1569 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1570 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1571 VMA, Type.c_str()); 1572 } 1573 outs() << "\n"; 1574 } 1575 1576 void llvm::printSectionContents(const ObjectFile *Obj) { 1577 std::error_code EC; 1578 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1579 StringRef Name; 1580 StringRef Contents; 1581 error(Section.getName(Name)); 1582 uint64_t BaseAddr = Section.getAddress(); 1583 uint64_t Size = Section.getSize(); 1584 if (!Size) 1585 continue; 1586 1587 outs() << "Contents of section " << Name << ":\n"; 1588 if (Section.isBSS()) { 1589 outs() << format("<skipping contents of bss section at [%04" PRIx64 1590 ", %04" PRIx64 ")>\n", 1591 BaseAddr, BaseAddr + Size); 1592 continue; 1593 } 1594 1595 error(Section.getContents(Contents)); 1596 1597 // Dump out the content as hex and printable ascii characters. 1598 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1599 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1600 // Dump line of hex. 1601 for (std::size_t I = 0; I < 16; ++I) { 1602 if (I != 0 && I % 4 == 0) 1603 outs() << ' '; 1604 if (Addr + I < End) 1605 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1606 << hexdigit(Contents[Addr + I] & 0xF, true); 1607 else 1608 outs() << " "; 1609 } 1610 // Print ascii. 1611 outs() << " "; 1612 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1613 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1614 outs() << Contents[Addr + I]; 1615 else 1616 outs() << "."; 1617 } 1618 outs() << "\n"; 1619 } 1620 } 1621 } 1622 1623 void llvm::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1624 StringRef ArchitectureName) { 1625 outs() << "SYMBOL TABLE:\n"; 1626 1627 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1628 printCOFFSymbolTable(Coff); 1629 return; 1630 } 1631 1632 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1633 // Skip printing the special zero symbol when dumping an ELF file. 1634 // This makes the output consistent with the GNU objdump. 1635 if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O)) 1636 continue; 1637 1638 const SymbolRef &Symbol = *I; 1639 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1640 if (!AddressOrError) 1641 report_error(ArchiveName, O->getFileName(), AddressOrError.takeError(), 1642 ArchitectureName); 1643 uint64_t Address = *AddressOrError; 1644 if ((Address < StartAddress) || (Address > StopAddress)) 1645 continue; 1646 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1647 if (!TypeOrError) 1648 report_error(ArchiveName, O->getFileName(), TypeOrError.takeError(), 1649 ArchitectureName); 1650 SymbolRef::Type Type = *TypeOrError; 1651 uint32_t Flags = Symbol.getFlags(); 1652 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1653 if (!SectionOrErr) 1654 report_error(ArchiveName, O->getFileName(), SectionOrErr.takeError(), 1655 ArchitectureName); 1656 section_iterator Section = *SectionOrErr; 1657 StringRef Name; 1658 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 1659 Section->getName(Name); 1660 } else { 1661 Expected<StringRef> NameOrErr = Symbol.getName(); 1662 if (!NameOrErr) 1663 report_error(ArchiveName, O->getFileName(), NameOrErr.takeError(), 1664 ArchitectureName); 1665 Name = *NameOrErr; 1666 } 1667 1668 bool Global = Flags & SymbolRef::SF_Global; 1669 bool Weak = Flags & SymbolRef::SF_Weak; 1670 bool Absolute = Flags & SymbolRef::SF_Absolute; 1671 bool Common = Flags & SymbolRef::SF_Common; 1672 bool Hidden = Flags & SymbolRef::SF_Hidden; 1673 1674 char GlobLoc = ' '; 1675 if (Type != SymbolRef::ST_Unknown) 1676 GlobLoc = Global ? 'g' : 'l'; 1677 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1678 ? 'd' : ' '; 1679 char FileFunc = ' '; 1680 if (Type == SymbolRef::ST_File) 1681 FileFunc = 'f'; 1682 else if (Type == SymbolRef::ST_Function) 1683 FileFunc = 'F'; 1684 else if (Type == SymbolRef::ST_Data) 1685 FileFunc = 'O'; 1686 1687 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1688 "%08" PRIx64; 1689 1690 outs() << format(Fmt, Address) << " " 1691 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1692 << (Weak ? 'w' : ' ') // Weak? 1693 << ' ' // Constructor. Not supported yet. 1694 << ' ' // Warning. Not supported yet. 1695 << ' ' // Indirect reference to another symbol. 1696 << Debug // Debugging (d) or dynamic (D) symbol. 1697 << FileFunc // Name of function (F), file (f) or object (O). 1698 << ' '; 1699 if (Absolute) { 1700 outs() << "*ABS*"; 1701 } else if (Common) { 1702 outs() << "*COM*"; 1703 } else if (Section == O->section_end()) { 1704 outs() << "*UND*"; 1705 } else { 1706 if (const MachOObjectFile *MachO = 1707 dyn_cast<const MachOObjectFile>(O)) { 1708 DataRefImpl DR = Section->getRawDataRefImpl(); 1709 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1710 outs() << SegmentName << ","; 1711 } 1712 StringRef SectionName; 1713 error(Section->getName(SectionName)); 1714 outs() << SectionName; 1715 } 1716 1717 outs() << '\t'; 1718 if (Common || isa<ELFObjectFileBase>(O)) { 1719 uint64_t Val = 1720 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1721 outs() << format("\t %08" PRIx64 " ", Val); 1722 } 1723 1724 if (Hidden) 1725 outs() << ".hidden "; 1726 1727 if (Demangle) 1728 outs() << demangle(Name) << '\n'; 1729 else 1730 outs() << Name << '\n'; 1731 } 1732 } 1733 1734 static void printUnwindInfo(const ObjectFile *O) { 1735 outs() << "Unwind info:\n\n"; 1736 1737 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1738 printCOFFUnwindInfo(Coff); 1739 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1740 printMachOUnwindInfo(MachO); 1741 else 1742 // TODO: Extract DWARF dump tool to objdump. 1743 WithColor::error(errs(), ToolName) 1744 << "This operation is only currently supported " 1745 "for COFF and MachO object files.\n"; 1746 } 1747 1748 void llvm::printExportsTrie(const ObjectFile *o) { 1749 outs() << "Exports trie:\n"; 1750 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1751 printMachOExportsTrie(MachO); 1752 else 1753 WithColor::error(errs(), ToolName) 1754 << "This operation is only currently supported " 1755 "for Mach-O executable files.\n"; 1756 } 1757 1758 void llvm::printRebaseTable(ObjectFile *o) { 1759 outs() << "Rebase table:\n"; 1760 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1761 printMachORebaseTable(MachO); 1762 else 1763 WithColor::error(errs(), ToolName) 1764 << "This operation is only currently supported " 1765 "for Mach-O executable files.\n"; 1766 } 1767 1768 void llvm::printBindTable(ObjectFile *o) { 1769 outs() << "Bind table:\n"; 1770 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1771 printMachOBindTable(MachO); 1772 else 1773 WithColor::error(errs(), ToolName) 1774 << "This operation is only currently supported " 1775 "for Mach-O executable files.\n"; 1776 } 1777 1778 void llvm::printLazyBindTable(ObjectFile *o) { 1779 outs() << "Lazy bind table:\n"; 1780 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1781 printMachOLazyBindTable(MachO); 1782 else 1783 WithColor::error(errs(), ToolName) 1784 << "This operation is only currently supported " 1785 "for Mach-O executable files.\n"; 1786 } 1787 1788 void llvm::printWeakBindTable(ObjectFile *o) { 1789 outs() << "Weak bind table:\n"; 1790 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1791 printMachOWeakBindTable(MachO); 1792 else 1793 WithColor::error(errs(), ToolName) 1794 << "This operation is only currently supported " 1795 "for Mach-O executable files.\n"; 1796 } 1797 1798 /// Dump the raw contents of the __clangast section so the output can be piped 1799 /// into llvm-bcanalyzer. 1800 void llvm::printRawClangAST(const ObjectFile *Obj) { 1801 if (outs().is_displayed()) { 1802 WithColor::error(errs(), ToolName) 1803 << "The -raw-clang-ast option will dump the raw binary contents of " 1804 "the clang ast section.\n" 1805 "Please redirect the output to a file or another program such as " 1806 "llvm-bcanalyzer.\n"; 1807 return; 1808 } 1809 1810 StringRef ClangASTSectionName("__clangast"); 1811 if (isa<COFFObjectFile>(Obj)) { 1812 ClangASTSectionName = "clangast"; 1813 } 1814 1815 Optional<object::SectionRef> ClangASTSection; 1816 for (auto Sec : ToolSectionFilter(*Obj)) { 1817 StringRef Name; 1818 Sec.getName(Name); 1819 if (Name == ClangASTSectionName) { 1820 ClangASTSection = Sec; 1821 break; 1822 } 1823 } 1824 if (!ClangASTSection) 1825 return; 1826 1827 StringRef ClangASTContents; 1828 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1829 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1830 } 1831 1832 static void printFaultMaps(const ObjectFile *Obj) { 1833 StringRef FaultMapSectionName; 1834 1835 if (isa<ELFObjectFileBase>(Obj)) { 1836 FaultMapSectionName = ".llvm_faultmaps"; 1837 } else if (isa<MachOObjectFile>(Obj)) { 1838 FaultMapSectionName = "__llvm_faultmaps"; 1839 } else { 1840 WithColor::error(errs(), ToolName) 1841 << "This operation is only currently supported " 1842 "for ELF and Mach-O executable files.\n"; 1843 return; 1844 } 1845 1846 Optional<object::SectionRef> FaultMapSection; 1847 1848 for (auto Sec : ToolSectionFilter(*Obj)) { 1849 StringRef Name; 1850 Sec.getName(Name); 1851 if (Name == FaultMapSectionName) { 1852 FaultMapSection = Sec; 1853 break; 1854 } 1855 } 1856 1857 outs() << "FaultMap table:\n"; 1858 1859 if (!FaultMapSection.hasValue()) { 1860 outs() << "<not found>\n"; 1861 return; 1862 } 1863 1864 StringRef FaultMapContents; 1865 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1866 1867 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1868 FaultMapContents.bytes_end()); 1869 1870 outs() << FMP; 1871 } 1872 1873 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1874 if (O->isELF()) { 1875 printELFFileHeader(O); 1876 printELFDynamicSection(O); 1877 printELFSymbolVersionInfo(O); 1878 return; 1879 } 1880 if (O->isCOFF()) 1881 return printCOFFFileHeader(O); 1882 if (O->isWasm()) 1883 return printWasmFileHeader(O); 1884 if (O->isMachO()) { 1885 printMachOFileHeader(O); 1886 if (!OnlyFirst) 1887 printMachOLoadCommands(O); 1888 return; 1889 } 1890 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1891 } 1892 1893 static void printFileHeaders(const ObjectFile *O) { 1894 if (!O->isELF() && !O->isCOFF()) 1895 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1896 1897 Triple::ArchType AT = O->getArch(); 1898 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1899 Expected<uint64_t> StartAddrOrErr = O->getStartAddress(); 1900 if (!StartAddrOrErr) 1901 report_error(O->getFileName(), StartAddrOrErr.takeError()); 1902 1903 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1904 uint64_t Address = StartAddrOrErr.get(); 1905 outs() << "start address: " 1906 << "0x" << format(Fmt.data(), Address) << "\n\n"; 1907 } 1908 1909 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 1910 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 1911 if (!ModeOrErr) { 1912 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 1913 consumeError(ModeOrErr.takeError()); 1914 return; 1915 } 1916 sys::fs::perms Mode = ModeOrErr.get(); 1917 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1918 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1919 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1920 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1921 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1922 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1923 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1924 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1925 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1926 1927 outs() << " "; 1928 1929 Expected<unsigned> UIDOrErr = C.getUID(); 1930 if (!UIDOrErr) 1931 report_error(Filename, UIDOrErr.takeError()); 1932 unsigned UID = UIDOrErr.get(); 1933 outs() << format("%d/", UID); 1934 1935 Expected<unsigned> GIDOrErr = C.getGID(); 1936 if (!GIDOrErr) 1937 report_error(Filename, GIDOrErr.takeError()); 1938 unsigned GID = GIDOrErr.get(); 1939 outs() << format("%-d ", GID); 1940 1941 Expected<uint64_t> Size = C.getRawSize(); 1942 if (!Size) 1943 report_error(Filename, Size.takeError()); 1944 outs() << format("%6" PRId64, Size.get()) << " "; 1945 1946 StringRef RawLastModified = C.getRawLastModified(); 1947 unsigned Seconds; 1948 if (RawLastModified.getAsInteger(10, Seconds)) 1949 outs() << "(date: \"" << RawLastModified 1950 << "\" contains non-decimal chars) "; 1951 else { 1952 // Since ctime(3) returns a 26 character string of the form: 1953 // "Sun Sep 16 01:03:52 1973\n\0" 1954 // just print 24 characters. 1955 time_t t = Seconds; 1956 outs() << format("%.24s ", ctime(&t)); 1957 } 1958 1959 StringRef Name = ""; 1960 Expected<StringRef> NameOrErr = C.getName(); 1961 if (!NameOrErr) { 1962 consumeError(NameOrErr.takeError()); 1963 Expected<StringRef> RawNameOrErr = C.getRawName(); 1964 if (!RawNameOrErr) 1965 report_error(Filename, NameOrErr.takeError()); 1966 Name = RawNameOrErr.get(); 1967 } else { 1968 Name = NameOrErr.get(); 1969 } 1970 outs() << Name << "\n"; 1971 } 1972 1973 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 1974 const Archive::Child *C = nullptr) { 1975 // Avoid other output when using a raw option. 1976 if (!RawClangAST) { 1977 outs() << '\n'; 1978 if (A) 1979 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 1980 else 1981 outs() << O->getFileName(); 1982 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 1983 } 1984 1985 StringRef ArchiveName = A ? A->getFileName() : ""; 1986 if (FileHeaders) 1987 printFileHeaders(O); 1988 if (ArchiveHeaders && !MachOOpt && C) 1989 printArchiveChild(ArchiveName, *C); 1990 if (Disassemble) 1991 disassembleObject(O, Relocations); 1992 if (Relocations && !Disassemble) 1993 printRelocations(O); 1994 if (DynamicRelocations) 1995 printDynamicRelocations(O); 1996 if (SectionHeaders) 1997 printSectionHeaders(O); 1998 if (SectionContents) 1999 printSectionContents(O); 2000 if (SymbolTable) 2001 printSymbolTable(O, ArchiveName); 2002 if (UnwindInfo) 2003 printUnwindInfo(O); 2004 if (PrivateHeaders || FirstPrivateHeader) 2005 printPrivateFileHeaders(O, FirstPrivateHeader); 2006 if (ExportsTrie) 2007 printExportsTrie(O); 2008 if (Rebase) 2009 printRebaseTable(O); 2010 if (Bind) 2011 printBindTable(O); 2012 if (LazyBind) 2013 printLazyBindTable(O); 2014 if (WeakBind) 2015 printWeakBindTable(O); 2016 if (RawClangAST) 2017 printRawClangAST(O); 2018 if (PrintFaultMaps) 2019 printFaultMaps(O); 2020 if (DwarfDumpType != DIDT_Null) { 2021 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2022 // Dump the complete DWARF structure. 2023 DIDumpOptions DumpOpts; 2024 DumpOpts.DumpType = DwarfDumpType; 2025 DICtx->dump(outs(), DumpOpts); 2026 } 2027 } 2028 2029 static void dumpObject(const COFFImportFile *I, const Archive *A, 2030 const Archive::Child *C = nullptr) { 2031 StringRef ArchiveName = A ? A->getFileName() : ""; 2032 2033 // Avoid other output when using a raw option. 2034 if (!RawClangAST) 2035 outs() << '\n' 2036 << ArchiveName << "(" << I->getFileName() << ")" 2037 << ":\tfile format COFF-import-file" 2038 << "\n\n"; 2039 2040 if (ArchiveHeaders && !MachOOpt && C) 2041 printArchiveChild(ArchiveName, *C); 2042 if (SymbolTable) 2043 printCOFFSymbolTable(I); 2044 } 2045 2046 /// Dump each object file in \a a; 2047 static void dumpArchive(const Archive *A) { 2048 Error Err = Error::success(); 2049 for (auto &C : A->children(Err)) { 2050 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2051 if (!ChildOrErr) { 2052 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2053 report_error(A->getFileName(), C, std::move(E)); 2054 continue; 2055 } 2056 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2057 dumpObject(O, A, &C); 2058 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2059 dumpObject(I, A, &C); 2060 else 2061 report_error(A->getFileName(), object_error::invalid_file_type); 2062 } 2063 if (Err) 2064 report_error(A->getFileName(), std::move(Err)); 2065 } 2066 2067 /// Open file and figure out how to dump it. 2068 static void dumpInput(StringRef file) { 2069 // If we are using the Mach-O specific object file parser, then let it parse 2070 // the file and process the command line options. So the -arch flags can 2071 // be used to select specific slices, etc. 2072 if (MachOOpt) { 2073 parseInputMachO(file); 2074 return; 2075 } 2076 2077 // Attempt to open the binary. 2078 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2079 if (!BinaryOrErr) 2080 report_error(file, BinaryOrErr.takeError()); 2081 Binary &Binary = *BinaryOrErr.get().getBinary(); 2082 2083 if (Archive *A = dyn_cast<Archive>(&Binary)) 2084 dumpArchive(A); 2085 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2086 dumpObject(O); 2087 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2088 parseInputMachO(UB); 2089 else 2090 report_error(file, object_error::invalid_file_type); 2091 } 2092 2093 int main(int argc, char **argv) { 2094 InitLLVM X(argc, argv); 2095 2096 // Initialize targets and assembly printers/parsers. 2097 llvm::InitializeAllTargetInfos(); 2098 llvm::InitializeAllTargetMCs(); 2099 llvm::InitializeAllDisassemblers(); 2100 2101 // Register the target printer for --version. 2102 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2103 2104 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2105 2106 ToolName = argv[0]; 2107 2108 // Defaults to a.out if no filenames specified. 2109 if (InputFilenames.empty()) 2110 InputFilenames.push_back("a.out"); 2111 2112 if (AllHeaders) 2113 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2114 SectionHeaders = SymbolTable = true; 2115 2116 if (DisassembleAll || PrintSource || PrintLines) 2117 Disassemble = true; 2118 2119 if (!Disassemble 2120 && !Relocations 2121 && !DynamicRelocations 2122 && !SectionHeaders 2123 && !SectionContents 2124 && !SymbolTable 2125 && !UnwindInfo 2126 && !PrivateHeaders 2127 && !FileHeaders 2128 && !FirstPrivateHeader 2129 && !ExportsTrie 2130 && !Rebase 2131 && !Bind 2132 && !LazyBind 2133 && !WeakBind 2134 && !RawClangAST 2135 && !(UniversalHeaders && MachOOpt) 2136 && !ArchiveHeaders 2137 && !(IndirectSymbols && MachOOpt) 2138 && !(DataInCode && MachOOpt) 2139 && !(LinkOptHints && MachOOpt) 2140 && !(InfoPlist && MachOOpt) 2141 && !(DylibsUsed && MachOOpt) 2142 && !(DylibId && MachOOpt) 2143 && !(ObjcMetaData && MachOOpt) 2144 && !(!FilterSections.empty() && MachOOpt) 2145 && !PrintFaultMaps 2146 && DwarfDumpType == DIDT_Null) { 2147 cl::PrintHelpMessage(); 2148 return 2; 2149 } 2150 2151 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2152 DisassembleFunctions.end()); 2153 2154 llvm::for_each(InputFilenames, dumpInput); 2155 2156 return EXIT_SUCCESS; 2157 } 2158