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 void llvm::error(Error E) { 334 if (!E) 335 return; 336 WithColor::error(errs(), ToolName) << toString(std::move(E)); 337 exit(1); 338 } 339 340 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 341 WithColor::error(errs(), ToolName) << Message << ".\n"; 342 errs().flush(); 343 exit(1); 344 } 345 346 void llvm::warn(StringRef Message) { 347 WithColor::warning(errs(), ToolName) << Message << ".\n"; 348 errs().flush(); 349 } 350 351 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 352 Twine Message) { 353 WithColor::error(errs(), ToolName) 354 << "'" << File << "': " << Message << ".\n"; 355 exit(1); 356 } 357 358 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 359 std::error_code EC) { 360 assert(EC); 361 WithColor::error(errs(), ToolName) 362 << "'" << File << "': " << EC.message() << ".\n"; 363 exit(1); 364 } 365 366 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(Error E, StringRef File) { 367 assert(E); 368 std::string Buf; 369 raw_string_ostream OS(Buf); 370 logAllUnhandledErrors(std::move(E), OS); 371 OS.flush(); 372 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 373 exit(1); 374 } 375 376 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(Error E, StringRef ArchiveName, 377 StringRef FileName, 378 StringRef ArchitectureName) { 379 assert(E); 380 WithColor::error(errs(), ToolName); 381 if (ArchiveName != "") 382 errs() << ArchiveName << "(" << FileName << ")"; 383 else 384 errs() << "'" << FileName << "'"; 385 if (!ArchitectureName.empty()) 386 errs() << " (for architecture " << ArchitectureName << ")"; 387 std::string Buf; 388 raw_string_ostream OS(Buf); 389 logAllUnhandledErrors(std::move(E), OS); 390 OS.flush(); 391 errs() << ": " << Buf; 392 exit(1); 393 } 394 395 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(Error E, StringRef ArchiveName, 396 const object::Archive::Child &C, 397 StringRef ArchitectureName) { 398 Expected<StringRef> NameOrErr = C.getName(); 399 // TODO: if we have a error getting the name then it would be nice to print 400 // the index of which archive member this is and or its offset in the 401 // archive instead of "???" as the name. 402 if (!NameOrErr) { 403 consumeError(NameOrErr.takeError()); 404 llvm::report_error(std::move(E), ArchiveName, "???", ArchitectureName); 405 } else 406 llvm::report_error(std::move(E), ArchiveName, NameOrErr.get(), 407 ArchitectureName); 408 } 409 410 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 411 // Figure out the target triple. 412 llvm::Triple TheTriple("unknown-unknown-unknown"); 413 if (TripleName.empty()) { 414 if (Obj) 415 TheTriple = Obj->makeTriple(); 416 } else { 417 TheTriple.setTriple(Triple::normalize(TripleName)); 418 419 // Use the triple, but also try to combine with ARM build attributes. 420 if (Obj) { 421 auto Arch = Obj->getArch(); 422 if (Arch == Triple::arm || Arch == Triple::armeb) 423 Obj->setARMSubArch(TheTriple); 424 } 425 } 426 427 // Get the target specific parser. 428 std::string Error; 429 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 430 Error); 431 if (!TheTarget) { 432 if (Obj) 433 report_error(Obj->getFileName(), "can't find target: " + Error); 434 else 435 error("can't find target: " + Error); 436 } 437 438 // Update the triple name and return the found target. 439 TripleName = TheTriple.getTriple(); 440 return TheTarget; 441 } 442 443 bool llvm::isRelocAddressLess(RelocationRef A, RelocationRef B) { 444 return A.getOffset() < B.getOffset(); 445 } 446 447 static Error getRelocationValueString(const RelocationRef &Rel, 448 SmallVectorImpl<char> &Result) { 449 const ObjectFile *Obj = Rel.getObject(); 450 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 451 return getELFRelocationValueString(ELF, Rel, Result); 452 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 453 return getCOFFRelocationValueString(COFF, Rel, Result); 454 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 455 return getWasmRelocationValueString(Wasm, Rel, Result); 456 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 457 return getMachORelocationValueString(MachO, Rel, Result); 458 llvm_unreachable("unknown object file format"); 459 } 460 461 /// Indicates whether this relocation should hidden when listing 462 /// relocations, usually because it is the trailing part of a multipart 463 /// relocation that will be printed as part of the leading relocation. 464 static bool getHidden(RelocationRef RelRef) { 465 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 466 if (!MachO) 467 return false; 468 469 unsigned Arch = MachO->getArch(); 470 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 471 uint64_t Type = MachO->getRelocationType(Rel); 472 473 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 474 // is always hidden. 475 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 476 return Type == MachO::GENERIC_RELOC_PAIR; 477 478 if (Arch == Triple::x86_64) { 479 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 480 // an X86_64_RELOC_SUBTRACTOR. 481 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 482 DataRefImpl RelPrev = Rel; 483 RelPrev.d.a--; 484 uint64_t PrevType = MachO->getRelocationType(RelPrev); 485 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 486 return true; 487 } 488 } 489 490 return false; 491 } 492 493 namespace { 494 class SourcePrinter { 495 protected: 496 DILineInfo OldLineInfo; 497 const ObjectFile *Obj = nullptr; 498 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 499 // File name to file contents of source 500 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 501 // Mark the line endings of the cached source 502 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 503 504 private: 505 bool cacheSource(const DILineInfo& LineInfoFile); 506 507 public: 508 SourcePrinter() = default; 509 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 510 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 511 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 512 DefaultArch); 513 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 514 } 515 virtual ~SourcePrinter() = default; 516 virtual void printSourceLine(raw_ostream &OS, 517 object::SectionedAddress Address, 518 StringRef Delimiter = "; "); 519 }; 520 521 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 522 std::unique_ptr<MemoryBuffer> Buffer; 523 if (LineInfo.Source) { 524 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 525 } else { 526 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 527 if (!BufferOrError) 528 return false; 529 Buffer = std::move(*BufferOrError); 530 } 531 // Chomp the file to get lines 532 const char *BufferStart = Buffer->getBufferStart(), 533 *BufferEnd = Buffer->getBufferEnd(); 534 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 535 const char *Start = BufferStart; 536 for (const char *I = BufferStart; I != BufferEnd; ++I) 537 if (*I == '\n') { 538 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 539 Start = I + 1; 540 } 541 if (Start < BufferEnd) 542 Lines.emplace_back(Start, BufferEnd - Start); 543 SourceCache[LineInfo.FileName] = std::move(Buffer); 544 return true; 545 } 546 547 void SourcePrinter::printSourceLine(raw_ostream &OS, 548 object::SectionedAddress Address, 549 StringRef Delimiter) { 550 if (!Symbolizer) 551 return; 552 DILineInfo LineInfo = DILineInfo(); 553 auto ExpectedLineInfo = 554 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 555 if (!ExpectedLineInfo) 556 consumeError(ExpectedLineInfo.takeError()); 557 else 558 LineInfo = *ExpectedLineInfo; 559 560 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 561 LineInfo.Line == 0) 562 return; 563 564 if (PrintLines) 565 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 566 if (PrintSource) { 567 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 568 if (!cacheSource(LineInfo)) 569 return; 570 auto LineBuffer = LineCache.find(LineInfo.FileName); 571 if (LineBuffer != LineCache.end()) { 572 if (LineInfo.Line > LineBuffer->second.size()) 573 return; 574 // Vector begins at 0, line numbers are non-zero 575 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1] << '\n'; 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 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 817 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 818 if (Name.empty()) 819 continue; 820 821 section_iterator SecI = 822 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 823 if (SecI == Obj->section_end()) 824 continue; 825 826 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 827 } 828 } 829 830 static void 831 addDynamicElfSymbols(const ObjectFile *Obj, 832 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 833 assert(Obj->isELF()); 834 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 835 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 836 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 837 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 838 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 839 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 840 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 841 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 842 else 843 llvm_unreachable("Unsupported binary format"); 844 } 845 846 static void addPltEntries(const ObjectFile *Obj, 847 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 848 StringSaver &Saver) { 849 Optional<SectionRef> Plt = None; 850 for (const SectionRef &Section : Obj->sections()) { 851 StringRef Name; 852 if (Section.getName(Name)) 853 continue; 854 if (Name == ".plt") 855 Plt = Section; 856 } 857 if (!Plt) 858 return; 859 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 860 for (auto PltEntry : ElfObj->getPltAddresses()) { 861 SymbolRef Symbol(PltEntry.first, ElfObj); 862 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 863 864 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 865 if (!Name.empty()) 866 AllSymbols[*Plt].emplace_back( 867 PltEntry.second, Saver.save((Name + "@plt").str()), SymbolType); 868 } 869 } 870 } 871 872 // Normally the disassembly output will skip blocks of zeroes. This function 873 // returns the number of zero bytes that can be skipped when dumping the 874 // disassembly of the instructions in Buf. 875 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 876 // Find the number of leading zeroes. 877 size_t N = 0; 878 while (N < Buf.size() && !Buf[N]) 879 ++N; 880 881 // We may want to skip blocks of zero bytes, but unless we see 882 // at least 8 of them in a row. 883 if (N < 8) 884 return 0; 885 886 // We skip zeroes in multiples of 4 because do not want to truncate an 887 // instruction if it starts with a zero byte. 888 return N & ~0x3; 889 } 890 891 // Returns a map from sections to their relocations. 892 static std::map<SectionRef, std::vector<RelocationRef>> 893 getRelocsMap(llvm::object::ObjectFile const &Obj) { 894 std::map<SectionRef, std::vector<RelocationRef>> Ret; 895 for (const SectionRef &Section : ToolSectionFilter(Obj)) { 896 section_iterator RelSec = Section.getRelocatedSection(); 897 if (RelSec == Obj.section_end()) 898 continue; 899 std::vector<RelocationRef> &V = Ret[*RelSec]; 900 for (const RelocationRef &R : Section.relocations()) 901 V.push_back(R); 902 // Sort relocations by address. 903 llvm::sort(V, isRelocAddressLess); 904 } 905 return Ret; 906 } 907 908 // Used for --adjust-vma to check if address should be adjusted by the 909 // specified value for a given section. 910 // For ELF we do not adjust non-allocatable sections like debug ones, 911 // because they are not loadable. 912 // TODO: implement for other file formats. 913 static bool shouldAdjustVA(const SectionRef &Section) { 914 const ObjectFile *Obj = Section.getObject(); 915 if (isa<object::ELFObjectFileBase>(Obj)) 916 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 917 return false; 918 } 919 920 static uint64_t 921 dumpARMELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 922 const ObjectFile *Obj, ArrayRef<uint8_t> Bytes, 923 const std::vector<uint64_t> &TextMappingSymsAddr) { 924 support::endianness Endian = 925 Obj->isLittleEndian() ? support::little : support::big; 926 while (Index < End) { 927 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 928 outs() << "\t"; 929 if (Index + 4 <= End) { 930 dumpBytes(Bytes.slice(Index, 4), outs()); 931 outs() << "\t.word\t" 932 << format_hex( 933 support::endian::read32(Bytes.data() + Index, Endian), 10); 934 Index += 4; 935 } else if (Index + 2 <= End) { 936 dumpBytes(Bytes.slice(Index, 2), outs()); 937 outs() << "\t\t.short\t" 938 << format_hex( 939 support::endian::read16(Bytes.data() + Index, Endian), 6); 940 Index += 2; 941 } else { 942 dumpBytes(Bytes.slice(Index, 1), outs()); 943 outs() << "\t\t.byte\t" << format_hex(Bytes[0], 4); 944 ++Index; 945 } 946 outs() << "\n"; 947 if (std::binary_search(TextMappingSymsAddr.begin(), 948 TextMappingSymsAddr.end(), Index)) 949 break; 950 } 951 return Index; 952 } 953 954 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 955 ArrayRef<uint8_t> Bytes) { 956 // print out data up to 8 bytes at a time in hex and ascii 957 uint8_t AsciiData[9] = {'\0'}; 958 uint8_t Byte; 959 int NumBytes = 0; 960 961 for (; Index < End; ++Index) { 962 if (NumBytes == 0) { 963 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 964 outs() << "\t"; 965 } 966 Byte = Bytes.slice(Index)[0]; 967 outs() << format(" %02x", Byte); 968 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 969 970 uint8_t IndentOffset = 0; 971 NumBytes++; 972 if (Index == End - 1 || NumBytes > 8) { 973 // Indent the space for less than 8 bytes data. 974 // 2 spaces for byte and one for space between bytes 975 IndentOffset = 3 * (8 - NumBytes); 976 for (int Excess = NumBytes; Excess < 8; Excess++) 977 AsciiData[Excess] = '\0'; 978 NumBytes = 8; 979 } 980 if (NumBytes == 8) { 981 AsciiData[8] = '\0'; 982 outs() << std::string(IndentOffset, ' ') << " "; 983 outs() << reinterpret_cast<char *>(AsciiData); 984 outs() << '\n'; 985 NumBytes = 0; 986 } 987 } 988 } 989 990 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 991 MCContext &Ctx, MCDisassembler *DisAsm, 992 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 993 const MCSubtargetInfo *STI, PrettyPrinter &PIP, 994 SourcePrinter &SP, bool InlineRelocs) { 995 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 996 if (InlineRelocs) 997 RelocMap = getRelocsMap(*Obj); 998 999 // Create a mapping from virtual address to symbol name. This is used to 1000 // pretty print the symbols while disassembling. 1001 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1002 SectionSymbolsTy AbsoluteSymbols; 1003 const StringRef FileName = Obj->getFileName(); 1004 for (const SymbolRef &Symbol : Obj->symbols()) { 1005 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName); 1006 1007 StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1008 if (Name.empty()) 1009 continue; 1010 1011 uint8_t SymbolType = ELF::STT_NOTYPE; 1012 if (Obj->isELF()) { 1013 SymbolType = getElfSymbolType(Obj, Symbol); 1014 if (SymbolType == ELF::STT_SECTION) 1015 continue; 1016 } 1017 1018 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1019 if (SecI != Obj->section_end()) 1020 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 1021 else 1022 AbsoluteSymbols.emplace_back(Address, Name, SymbolType); 1023 } 1024 if (AllSymbols.empty() && Obj->isELF()) 1025 addDynamicElfSymbols(Obj, AllSymbols); 1026 1027 BumpPtrAllocator A; 1028 StringSaver Saver(A); 1029 addPltEntries(Obj, AllSymbols, Saver); 1030 1031 // Create a mapping from virtual address to section. 1032 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1033 for (SectionRef Sec : Obj->sections()) 1034 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1035 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1036 1037 // Linked executables (.exe and .dll files) typically don't include a real 1038 // symbol table but they might contain an export table. 1039 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1040 for (const auto &ExportEntry : COFFObj->export_directories()) { 1041 StringRef Name; 1042 error(ExportEntry.getSymbolName(Name)); 1043 if (Name.empty()) 1044 continue; 1045 uint32_t RVA; 1046 error(ExportEntry.getExportRVA(RVA)); 1047 1048 uint64_t VA = COFFObj->getImageBase() + RVA; 1049 auto Sec = llvm::upper_bound( 1050 SectionAddresses, VA, 1051 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1052 return LHS < RHS.first; 1053 }); 1054 if (Sec != SectionAddresses.begin()) { 1055 --Sec; 1056 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1057 } else 1058 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1059 } 1060 } 1061 1062 // Sort all the symbols, this allows us to use a simple binary search to find 1063 // a symbol near an address. 1064 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1065 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1066 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1067 1068 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1069 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1070 continue; 1071 1072 uint64_t SectionAddr = Section.getAddress(); 1073 uint64_t SectSize = Section.getSize(); 1074 if (!SectSize) 1075 continue; 1076 1077 // Get the list of all the symbols in this section. 1078 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1079 std::vector<uint64_t> DataMappingSymsAddr; 1080 std::vector<uint64_t> TextMappingSymsAddr; 1081 if (isArmElf(Obj)) { 1082 for (const auto &Symb : Symbols) { 1083 uint64_t Address = std::get<0>(Symb); 1084 StringRef Name = std::get<1>(Symb); 1085 if (Name.startswith("$d")) 1086 DataMappingSymsAddr.push_back(Address - SectionAddr); 1087 if (Name.startswith("$x")) 1088 TextMappingSymsAddr.push_back(Address - SectionAddr); 1089 if (Name.startswith("$a")) 1090 TextMappingSymsAddr.push_back(Address - SectionAddr); 1091 if (Name.startswith("$t")) 1092 TextMappingSymsAddr.push_back(Address - SectionAddr); 1093 } 1094 } 1095 1096 llvm::sort(DataMappingSymsAddr); 1097 llvm::sort(TextMappingSymsAddr); 1098 1099 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1100 // AMDGPU disassembler uses symbolizer for printing labels 1101 std::unique_ptr<MCRelocationInfo> RelInfo( 1102 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1103 if (RelInfo) { 1104 std::unique_ptr<MCSymbolizer> Symbolizer( 1105 TheTarget->createMCSymbolizer( 1106 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1107 DisAsm->setSymbolizer(std::move(Symbolizer)); 1108 } 1109 } 1110 1111 StringRef SegmentName = ""; 1112 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1113 DataRefImpl DR = Section.getRawDataRefImpl(); 1114 SegmentName = MachO->getSectionFinalSegmentName(DR); 1115 } 1116 StringRef SectionName; 1117 error(Section.getName(SectionName)); 1118 1119 // If the section has no symbol at the start, just insert a dummy one. 1120 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1121 Symbols.insert( 1122 Symbols.begin(), 1123 std::make_tuple(SectionAddr, SectionName, 1124 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1125 } 1126 1127 SmallString<40> Comments; 1128 raw_svector_ostream CommentStream(Comments); 1129 1130 StringRef BytesStr; 1131 error(Section.getContents(BytesStr)); 1132 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef(BytesStr); 1133 1134 uint64_t VMAAdjustment = 0; 1135 if (shouldAdjustVA(Section)) 1136 VMAAdjustment = AdjustVMA; 1137 1138 uint64_t Size; 1139 uint64_t Index; 1140 bool PrintedSection = false; 1141 std::vector<RelocationRef> Rels = RelocMap[Section]; 1142 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1143 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1144 // Disassemble symbol by symbol. 1145 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1146 uint64_t Start = std::get<0>(Symbols[SI]) - SectionAddr; 1147 // The end is either the section end or the beginning of the next 1148 // symbol. 1149 uint64_t End = (SI == SE - 1) 1150 ? SectSize 1151 : std::get<0>(Symbols[SI + 1]) - SectionAddr; 1152 // Don't try to disassemble beyond the end of section contents. 1153 if (End > SectSize) 1154 End = SectSize; 1155 // If this symbol has the same address as the next symbol, then skip it. 1156 if (Start >= End) 1157 continue; 1158 1159 // Check if we need to skip symbol 1160 // Skip if the symbol's data is not between StartAddress and StopAddress 1161 if (End + SectionAddr <= StartAddress || 1162 Start + SectionAddr >= StopAddress) 1163 continue; 1164 1165 // Stop disassembly at the stop address specified 1166 if (End + SectionAddr > StopAddress) 1167 End = StopAddress - SectionAddr; 1168 1169 /// Skip if user requested specific symbols and this is not in the list 1170 if (!DisasmFuncsSet.empty() && 1171 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1172 continue; 1173 1174 if (!PrintedSection) { 1175 PrintedSection = true; 1176 outs() << "Disassembly of section "; 1177 if (!SegmentName.empty()) 1178 outs() << SegmentName << ","; 1179 outs() << SectionName << ':'; 1180 } 1181 1182 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1183 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1184 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1185 Start += 256; 1186 } 1187 if (SI == SE - 1 || 1188 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1189 // cut trailing zeroes at the end of kernel 1190 // cut up to 256 bytes 1191 const uint64_t EndAlign = 256; 1192 const auto Limit = End - (std::min)(EndAlign, End - Start); 1193 while (End > Limit && 1194 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1195 End -= 4; 1196 } 1197 } 1198 1199 outs() << '\n'; 1200 if (!NoLeadingAddr) 1201 outs() << format("%016" PRIx64 " ", 1202 SectionAddr + Start + VMAAdjustment); 1203 1204 StringRef SymbolName = std::get<1>(Symbols[SI]); 1205 if (Demangle) 1206 outs() << demangle(SymbolName) << ":\n"; 1207 else 1208 outs() << SymbolName << ":\n"; 1209 1210 // Don't print raw contents of a virtual section. A virtual section 1211 // doesn't have any contents in the file. 1212 if (Section.isVirtual()) { 1213 outs() << "...\n"; 1214 continue; 1215 } 1216 1217 #ifndef NDEBUG 1218 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1219 #else 1220 raw_ostream &DebugOut = nulls(); 1221 #endif 1222 1223 // Some targets (like WebAssembly) have a special prelude at the start 1224 // of each symbol. 1225 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1226 SectionAddr + Start, DebugOut, CommentStream); 1227 Start += Size; 1228 1229 Index = Start; 1230 if (SectionAddr < StartAddress) 1231 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1232 1233 // If there is a data symbol inside an ELF text section and we are 1234 // only disassembling text (applicable all architectures), we are in a 1235 // situation where we must print the data and not disassemble it. 1236 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1237 !DisassembleAll && Section.isText()) { 1238 dumpELFData(SectionAddr, Index, End, Bytes); 1239 Index = End; 1240 } 1241 1242 bool CheckARMELFData = isArmElf(Obj) && 1243 std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1244 !DisassembleAll; 1245 while (Index < End) { 1246 // AArch64 ELF binaries can interleave data and text in the same 1247 // section. We rely on the markers introduced to understand what we 1248 // need to dump. If the data marker is within a function, it is 1249 // denoted as a word/short etc. 1250 if (CheckARMELFData && 1251 std::binary_search(DataMappingSymsAddr.begin(), 1252 DataMappingSymsAddr.end(), Index)) { 1253 Index = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1254 TextMappingSymsAddr); 1255 continue; 1256 } 1257 1258 // When -z or --disassemble-zeroes are given we always dissasemble 1259 // them. Otherwise we might want to skip zero bytes we see. 1260 if (!DisassembleZeroes) { 1261 uint64_t MaxOffset = End - Index; 1262 // For -reloc: print zero blocks patched by relocations, so that 1263 // relocations can be shown in the dump. 1264 if (RelCur != RelEnd) 1265 MaxOffset = RelCur->getOffset() - Index; 1266 1267 if (size_t N = 1268 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1269 outs() << "\t\t..." << '\n'; 1270 Index += N; 1271 continue; 1272 } 1273 } 1274 1275 // Disassemble a real instruction or a data when disassemble all is 1276 // provided 1277 MCInst Inst; 1278 bool Disassembled = DisAsm->getInstruction( 1279 Inst, Size, Bytes.slice(Index), SectionAddr + Index, DebugOut, 1280 CommentStream); 1281 if (Size == 0) 1282 Size = 1; 1283 1284 PIP.printInst( 1285 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 1286 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, outs(), 1287 "", *STI, &SP, &Rels); 1288 outs() << CommentStream.str(); 1289 Comments.clear(); 1290 1291 // Try to resolve the target of a call, tail call, etc. to a specific 1292 // symbol. 1293 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1294 MIA->isConditionalBranch(Inst))) { 1295 uint64_t Target; 1296 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1297 // In a relocatable object, the target's section must reside in 1298 // the same section as the call instruction or it is accessed 1299 // through a relocation. 1300 // 1301 // In a non-relocatable object, the target may be in any section. 1302 // 1303 // N.B. We don't walk the relocations in the relocatable case yet. 1304 auto *TargetSectionSymbols = &Symbols; 1305 if (!Obj->isRelocatableObject()) { 1306 auto SectionAddress = llvm::upper_bound( 1307 SectionAddresses, Target, 1308 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1309 return LHS < RHS.first; 1310 }); 1311 if (SectionAddress != SectionAddresses.begin()) { 1312 --SectionAddress; 1313 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1314 } else { 1315 TargetSectionSymbols = &AbsoluteSymbols; 1316 } 1317 } 1318 1319 // Find the first symbol in the section whose offset is less than 1320 // or equal to the target. If there isn't a section that contains 1321 // the target, find the nearest preceding absolute symbol. 1322 auto TargetSym = llvm::upper_bound( 1323 *TargetSectionSymbols, Target, 1324 [](uint64_t LHS, 1325 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1326 return LHS < std::get<0>(RHS); 1327 }); 1328 if (TargetSym == TargetSectionSymbols->begin()) { 1329 TargetSectionSymbols = &AbsoluteSymbols; 1330 TargetSym = llvm::upper_bound( 1331 AbsoluteSymbols, Target, 1332 [](uint64_t LHS, 1333 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1334 return LHS < std::get<0>(RHS); 1335 }); 1336 } 1337 if (TargetSym != TargetSectionSymbols->begin()) { 1338 --TargetSym; 1339 uint64_t TargetAddress = std::get<0>(*TargetSym); 1340 StringRef TargetName = std::get<1>(*TargetSym); 1341 outs() << " <" << TargetName; 1342 uint64_t Disp = Target - TargetAddress; 1343 if (Disp) 1344 outs() << "+0x" << Twine::utohexstr(Disp); 1345 outs() << '>'; 1346 } 1347 } 1348 } 1349 outs() << "\n"; 1350 1351 // Hexagon does this in pretty printer 1352 if (Obj->getArch() != Triple::hexagon) { 1353 // Print relocation for instruction. 1354 while (RelCur != RelEnd) { 1355 uint64_t Offset = RelCur->getOffset(); 1356 // If this relocation is hidden, skip it. 1357 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 1358 ++RelCur; 1359 continue; 1360 } 1361 1362 // Stop when RelCur's offset is past the current instruction. 1363 if (Offset >= Index + Size) 1364 break; 1365 1366 // When --adjust-vma is used, update the address printed. 1367 if (RelCur->getSymbol() != Obj->symbol_end()) { 1368 Expected<section_iterator> SymSI = 1369 RelCur->getSymbol()->getSection(); 1370 if (SymSI && *SymSI != Obj->section_end() && 1371 shouldAdjustVA(**SymSI)) 1372 Offset += AdjustVMA; 1373 } 1374 1375 printRelocation(*RelCur, SectionAddr + Offset, 1376 Obj->getBytesInAddress()); 1377 ++RelCur; 1378 } 1379 } 1380 1381 Index += Size; 1382 } 1383 } 1384 } 1385 } 1386 1387 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1388 if (StartAddress > StopAddress) 1389 error("Start address should be less than stop address"); 1390 1391 const Target *TheTarget = getTarget(Obj); 1392 1393 // Package up features to be passed to target/subtarget 1394 SubtargetFeatures Features = Obj->getFeatures(); 1395 if (!MAttrs.empty()) 1396 for (unsigned I = 0; I != MAttrs.size(); ++I) 1397 Features.AddFeature(MAttrs[I]); 1398 1399 std::unique_ptr<const MCRegisterInfo> MRI( 1400 TheTarget->createMCRegInfo(TripleName)); 1401 if (!MRI) 1402 report_error(Obj->getFileName(), 1403 "no register info for target " + TripleName); 1404 1405 // Set up disassembler. 1406 std::unique_ptr<const MCAsmInfo> AsmInfo( 1407 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1408 if (!AsmInfo) 1409 report_error(Obj->getFileName(), 1410 "no assembly info for target " + TripleName); 1411 std::unique_ptr<const MCSubtargetInfo> STI( 1412 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1413 if (!STI) 1414 report_error(Obj->getFileName(), 1415 "no subtarget info for target " + TripleName); 1416 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1417 if (!MII) 1418 report_error(Obj->getFileName(), 1419 "no instruction info for target " + TripleName); 1420 MCObjectFileInfo MOFI; 1421 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1422 // FIXME: for now initialize MCObjectFileInfo with default values 1423 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1424 1425 std::unique_ptr<MCDisassembler> DisAsm( 1426 TheTarget->createMCDisassembler(*STI, Ctx)); 1427 if (!DisAsm) 1428 report_error(Obj->getFileName(), 1429 "no disassembler for target " + TripleName); 1430 1431 std::unique_ptr<const MCInstrAnalysis> MIA( 1432 TheTarget->createMCInstrAnalysis(MII.get())); 1433 1434 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1435 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1436 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1437 if (!IP) 1438 report_error(Obj->getFileName(), 1439 "no instruction printer for target " + TripleName); 1440 IP->setPrintImmHex(PrintImmHex); 1441 1442 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1443 SourcePrinter SP(Obj, TheTarget->getName()); 1444 1445 for (StringRef Opt : DisassemblerOptions) 1446 if (!IP->applyTargetSpecificCLOption(Opt)) 1447 error("Unrecognized disassembler option: " + Opt); 1448 1449 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), MIA.get(), IP.get(), 1450 STI.get(), PIP, SP, InlineRelocs); 1451 } 1452 1453 void llvm::printRelocations(const ObjectFile *Obj) { 1454 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1455 "%08" PRIx64; 1456 // Regular objdump doesn't print relocations in non-relocatable object 1457 // files. 1458 if (!Obj->isRelocatableObject()) 1459 return; 1460 1461 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1462 if (Section.relocation_begin() == Section.relocation_end()) 1463 continue; 1464 StringRef SecName; 1465 error(Section.getName(SecName)); 1466 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1467 for (const RelocationRef &Reloc : Section.relocations()) { 1468 uint64_t Address = Reloc.getOffset(); 1469 SmallString<32> RelocName; 1470 SmallString<32> ValueStr; 1471 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1472 continue; 1473 Reloc.getTypeName(RelocName); 1474 error(getRelocationValueString(Reloc, ValueStr)); 1475 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1476 << ValueStr << "\n"; 1477 } 1478 outs() << "\n"; 1479 } 1480 } 1481 1482 void llvm::printDynamicRelocations(const ObjectFile *Obj) { 1483 // For the moment, this option is for ELF only 1484 if (!Obj->isELF()) 1485 return; 1486 1487 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1488 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1489 error("not a dynamic object"); 1490 return; 1491 } 1492 1493 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1494 if (DynRelSec.empty()) 1495 return; 1496 1497 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1498 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1499 for (const SectionRef &Section : DynRelSec) { 1500 if (Section.relocation_begin() == Section.relocation_end()) 1501 continue; 1502 for (const RelocationRef &Reloc : Section.relocations()) { 1503 uint64_t Address = Reloc.getOffset(); 1504 SmallString<32> RelocName; 1505 SmallString<32> ValueStr; 1506 Reloc.getTypeName(RelocName); 1507 error(getRelocationValueString(Reloc, ValueStr)); 1508 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1509 << ValueStr << "\n"; 1510 } 1511 } 1512 } 1513 1514 // Returns true if we need to show LMA column when dumping section headers. We 1515 // show it only when the platform is ELF and either we have at least one section 1516 // whose VMA and LMA are different and/or when --show-lma flag is used. 1517 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1518 if (!Obj->isELF()) 1519 return false; 1520 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1521 if (S.getAddress() != getELFSectionLMA(S)) 1522 return true; 1523 return ShowLMA; 1524 } 1525 1526 void llvm::printSectionHeaders(const ObjectFile *Obj) { 1527 bool HasLMAColumn = shouldDisplayLMA(Obj); 1528 if (HasLMAColumn) 1529 outs() << "Sections:\n" 1530 "Idx Name Size VMA LMA " 1531 "Type\n"; 1532 else 1533 outs() << "Sections:\n" 1534 "Idx Name Size VMA Type\n"; 1535 1536 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1537 StringRef Name; 1538 error(Section.getName(Name)); 1539 uint64_t VMA = Section.getAddress(); 1540 if (shouldAdjustVA(Section)) 1541 VMA += AdjustVMA; 1542 1543 uint64_t Size = Section.getSize(); 1544 bool Text = Section.isText(); 1545 bool Data = Section.isData(); 1546 bool BSS = Section.isBSS(); 1547 std::string Type = (std::string(Text ? "TEXT " : "") + 1548 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1549 1550 if (HasLMAColumn) 1551 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %016" PRIx64 1552 " %s\n", 1553 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1554 VMA, getELFSectionLMA(Section), Type.c_str()); 1555 else 1556 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1557 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1558 VMA, Type.c_str()); 1559 } 1560 outs() << "\n"; 1561 } 1562 1563 void llvm::printSectionContents(const ObjectFile *Obj) { 1564 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1565 StringRef Name; 1566 StringRef Contents; 1567 error(Section.getName(Name)); 1568 uint64_t BaseAddr = Section.getAddress(); 1569 uint64_t Size = Section.getSize(); 1570 if (!Size) 1571 continue; 1572 1573 outs() << "Contents of section " << Name << ":\n"; 1574 if (Section.isBSS()) { 1575 outs() << format("<skipping contents of bss section at [%04" PRIx64 1576 ", %04" PRIx64 ")>\n", 1577 BaseAddr, BaseAddr + Size); 1578 continue; 1579 } 1580 1581 error(Section.getContents(Contents)); 1582 1583 // Dump out the content as hex and printable ascii characters. 1584 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1585 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1586 // Dump line of hex. 1587 for (std::size_t I = 0; I < 16; ++I) { 1588 if (I != 0 && I % 4 == 0) 1589 outs() << ' '; 1590 if (Addr + I < End) 1591 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1592 << hexdigit(Contents[Addr + I] & 0xF, true); 1593 else 1594 outs() << " "; 1595 } 1596 // Print ascii. 1597 outs() << " "; 1598 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1599 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1600 outs() << Contents[Addr + I]; 1601 else 1602 outs() << "."; 1603 } 1604 outs() << "\n"; 1605 } 1606 } 1607 } 1608 1609 void llvm::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1610 StringRef ArchitectureName) { 1611 outs() << "SYMBOL TABLE:\n"; 1612 1613 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1614 printCOFFSymbolTable(Coff); 1615 return; 1616 } 1617 1618 const StringRef FileName = O->getFileName(); 1619 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1620 // Skip printing the special zero symbol when dumping an ELF file. 1621 // This makes the output consistent with the GNU objdump. 1622 if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O)) 1623 continue; 1624 1625 const SymbolRef &Symbol = *I; 1626 uint64_t Address = unwrapOrError(Symbol.getAddress(), ArchiveName, FileName, 1627 ArchitectureName); 1628 if ((Address < StartAddress) || (Address > StopAddress)) 1629 continue; 1630 SymbolRef::Type Type = unwrapOrError(Symbol.getType(), ArchiveName, 1631 FileName, ArchitectureName); 1632 uint32_t Flags = Symbol.getFlags(); 1633 section_iterator Section = unwrapOrError(Symbol.getSection(), ArchiveName, 1634 FileName, ArchitectureName); 1635 StringRef Name; 1636 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) 1637 Section->getName(Name); 1638 else 1639 Name = unwrapOrError(Symbol.getName(), ArchiveName, FileName, 1640 ArchitectureName); 1641 1642 bool Global = Flags & SymbolRef::SF_Global; 1643 bool Weak = Flags & SymbolRef::SF_Weak; 1644 bool Absolute = Flags & SymbolRef::SF_Absolute; 1645 bool Common = Flags & SymbolRef::SF_Common; 1646 bool Hidden = Flags & SymbolRef::SF_Hidden; 1647 1648 char GlobLoc = ' '; 1649 if (Type != SymbolRef::ST_Unknown) 1650 GlobLoc = Global ? 'g' : 'l'; 1651 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1652 ? 'd' : ' '; 1653 char FileFunc = ' '; 1654 if (Type == SymbolRef::ST_File) 1655 FileFunc = 'f'; 1656 else if (Type == SymbolRef::ST_Function) 1657 FileFunc = 'F'; 1658 else if (Type == SymbolRef::ST_Data) 1659 FileFunc = 'O'; 1660 1661 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1662 "%08" PRIx64; 1663 1664 outs() << format(Fmt, Address) << " " 1665 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1666 << (Weak ? 'w' : ' ') // Weak? 1667 << ' ' // Constructor. Not supported yet. 1668 << ' ' // Warning. Not supported yet. 1669 << ' ' // Indirect reference to another symbol. 1670 << Debug // Debugging (d) or dynamic (D) symbol. 1671 << FileFunc // Name of function (F), file (f) or object (O). 1672 << ' '; 1673 if (Absolute) { 1674 outs() << "*ABS*"; 1675 } else if (Common) { 1676 outs() << "*COM*"; 1677 } else if (Section == O->section_end()) { 1678 outs() << "*UND*"; 1679 } else { 1680 if (const MachOObjectFile *MachO = 1681 dyn_cast<const MachOObjectFile>(O)) { 1682 DataRefImpl DR = Section->getRawDataRefImpl(); 1683 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1684 outs() << SegmentName << ","; 1685 } 1686 StringRef SectionName; 1687 error(Section->getName(SectionName)); 1688 outs() << SectionName; 1689 } 1690 1691 outs() << '\t'; 1692 if (Common || isa<ELFObjectFileBase>(O)) { 1693 uint64_t Val = 1694 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1695 outs() << format("\t %08" PRIx64 " ", Val); 1696 } 1697 1698 if (Hidden) 1699 outs() << ".hidden "; 1700 1701 if (Demangle) 1702 outs() << demangle(Name) << '\n'; 1703 else 1704 outs() << Name << '\n'; 1705 } 1706 } 1707 1708 static void printUnwindInfo(const ObjectFile *O) { 1709 outs() << "Unwind info:\n\n"; 1710 1711 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1712 printCOFFUnwindInfo(Coff); 1713 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1714 printMachOUnwindInfo(MachO); 1715 else 1716 // TODO: Extract DWARF dump tool to objdump. 1717 WithColor::error(errs(), ToolName) 1718 << "This operation is only currently supported " 1719 "for COFF and MachO object files.\n"; 1720 } 1721 1722 void llvm::printExportsTrie(const ObjectFile *o) { 1723 outs() << "Exports trie:\n"; 1724 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1725 printMachOExportsTrie(MachO); 1726 else 1727 WithColor::error(errs(), ToolName) 1728 << "This operation is only currently supported " 1729 "for Mach-O executable files.\n"; 1730 } 1731 1732 void llvm::printRebaseTable(ObjectFile *o) { 1733 outs() << "Rebase table:\n"; 1734 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1735 printMachORebaseTable(MachO); 1736 else 1737 WithColor::error(errs(), ToolName) 1738 << "This operation is only currently supported " 1739 "for Mach-O executable files.\n"; 1740 } 1741 1742 void llvm::printBindTable(ObjectFile *o) { 1743 outs() << "Bind table:\n"; 1744 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1745 printMachOBindTable(MachO); 1746 else 1747 WithColor::error(errs(), ToolName) 1748 << "This operation is only currently supported " 1749 "for Mach-O executable files.\n"; 1750 } 1751 1752 void llvm::printLazyBindTable(ObjectFile *o) { 1753 outs() << "Lazy bind table:\n"; 1754 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1755 printMachOLazyBindTable(MachO); 1756 else 1757 WithColor::error(errs(), ToolName) 1758 << "This operation is only currently supported " 1759 "for Mach-O executable files.\n"; 1760 } 1761 1762 void llvm::printWeakBindTable(ObjectFile *o) { 1763 outs() << "Weak bind table:\n"; 1764 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1765 printMachOWeakBindTable(MachO); 1766 else 1767 WithColor::error(errs(), ToolName) 1768 << "This operation is only currently supported " 1769 "for Mach-O executable files.\n"; 1770 } 1771 1772 /// Dump the raw contents of the __clangast section so the output can be piped 1773 /// into llvm-bcanalyzer. 1774 void llvm::printRawClangAST(const ObjectFile *Obj) { 1775 if (outs().is_displayed()) { 1776 WithColor::error(errs(), ToolName) 1777 << "The -raw-clang-ast option will dump the raw binary contents of " 1778 "the clang ast section.\n" 1779 "Please redirect the output to a file or another program such as " 1780 "llvm-bcanalyzer.\n"; 1781 return; 1782 } 1783 1784 StringRef ClangASTSectionName("__clangast"); 1785 if (isa<COFFObjectFile>(Obj)) { 1786 ClangASTSectionName = "clangast"; 1787 } 1788 1789 Optional<object::SectionRef> ClangASTSection; 1790 for (auto Sec : ToolSectionFilter(*Obj)) { 1791 StringRef Name; 1792 Sec.getName(Name); 1793 if (Name == ClangASTSectionName) { 1794 ClangASTSection = Sec; 1795 break; 1796 } 1797 } 1798 if (!ClangASTSection) 1799 return; 1800 1801 StringRef ClangASTContents; 1802 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1803 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1804 } 1805 1806 static void printFaultMaps(const ObjectFile *Obj) { 1807 StringRef FaultMapSectionName; 1808 1809 if (isa<ELFObjectFileBase>(Obj)) { 1810 FaultMapSectionName = ".llvm_faultmaps"; 1811 } else if (isa<MachOObjectFile>(Obj)) { 1812 FaultMapSectionName = "__llvm_faultmaps"; 1813 } else { 1814 WithColor::error(errs(), ToolName) 1815 << "This operation is only currently supported " 1816 "for ELF and Mach-O executable files.\n"; 1817 return; 1818 } 1819 1820 Optional<object::SectionRef> FaultMapSection; 1821 1822 for (auto Sec : ToolSectionFilter(*Obj)) { 1823 StringRef Name; 1824 Sec.getName(Name); 1825 if (Name == FaultMapSectionName) { 1826 FaultMapSection = Sec; 1827 break; 1828 } 1829 } 1830 1831 outs() << "FaultMap table:\n"; 1832 1833 if (!FaultMapSection.hasValue()) { 1834 outs() << "<not found>\n"; 1835 return; 1836 } 1837 1838 StringRef FaultMapContents; 1839 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1840 1841 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1842 FaultMapContents.bytes_end()); 1843 1844 outs() << FMP; 1845 } 1846 1847 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1848 if (O->isELF()) { 1849 printELFFileHeader(O); 1850 printELFDynamicSection(O); 1851 printELFSymbolVersionInfo(O); 1852 return; 1853 } 1854 if (O->isCOFF()) 1855 return printCOFFFileHeader(O); 1856 if (O->isWasm()) 1857 return printWasmFileHeader(O); 1858 if (O->isMachO()) { 1859 printMachOFileHeader(O); 1860 if (!OnlyFirst) 1861 printMachOLoadCommands(O); 1862 return; 1863 } 1864 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1865 } 1866 1867 static void printFileHeaders(const ObjectFile *O) { 1868 if (!O->isELF() && !O->isCOFF()) 1869 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1870 1871 Triple::ArchType AT = O->getArch(); 1872 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1873 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 1874 1875 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1876 outs() << "start address: " 1877 << "0x" << format(Fmt.data(), Address) << "\n\n"; 1878 } 1879 1880 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 1881 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 1882 if (!ModeOrErr) { 1883 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 1884 consumeError(ModeOrErr.takeError()); 1885 return; 1886 } 1887 sys::fs::perms Mode = ModeOrErr.get(); 1888 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1889 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1890 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1891 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1892 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1893 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1894 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1895 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1896 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1897 1898 outs() << " "; 1899 1900 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 1901 unwrapOrError(C.getGID(), Filename), 1902 unwrapOrError(C.getRawSize(), Filename)); 1903 1904 StringRef RawLastModified = C.getRawLastModified(); 1905 unsigned Seconds; 1906 if (RawLastModified.getAsInteger(10, Seconds)) 1907 outs() << "(date: \"" << RawLastModified 1908 << "\" contains non-decimal chars) "; 1909 else { 1910 // Since ctime(3) returns a 26 character string of the form: 1911 // "Sun Sep 16 01:03:52 1973\n\0" 1912 // just print 24 characters. 1913 time_t t = Seconds; 1914 outs() << format("%.24s ", ctime(&t)); 1915 } 1916 1917 StringRef Name = ""; 1918 Expected<StringRef> NameOrErr = C.getName(); 1919 if (!NameOrErr) { 1920 consumeError(NameOrErr.takeError()); 1921 Name = unwrapOrError(C.getRawName(), Filename); 1922 } else { 1923 Name = NameOrErr.get(); 1924 } 1925 outs() << Name << "\n"; 1926 } 1927 1928 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 1929 const Archive::Child *C = nullptr) { 1930 // Avoid other output when using a raw option. 1931 if (!RawClangAST) { 1932 outs() << '\n'; 1933 if (A) 1934 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 1935 else 1936 outs() << O->getFileName(); 1937 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 1938 } 1939 1940 StringRef ArchiveName = A ? A->getFileName() : ""; 1941 if (FileHeaders) 1942 printFileHeaders(O); 1943 if (ArchiveHeaders && !MachOOpt && C) 1944 printArchiveChild(ArchiveName, *C); 1945 if (Disassemble) 1946 disassembleObject(O, Relocations); 1947 if (Relocations && !Disassemble) 1948 printRelocations(O); 1949 if (DynamicRelocations) 1950 printDynamicRelocations(O); 1951 if (SectionHeaders) 1952 printSectionHeaders(O); 1953 if (SectionContents) 1954 printSectionContents(O); 1955 if (SymbolTable) 1956 printSymbolTable(O, ArchiveName); 1957 if (UnwindInfo) 1958 printUnwindInfo(O); 1959 if (PrivateHeaders || FirstPrivateHeader) 1960 printPrivateFileHeaders(O, FirstPrivateHeader); 1961 if (ExportsTrie) 1962 printExportsTrie(O); 1963 if (Rebase) 1964 printRebaseTable(O); 1965 if (Bind) 1966 printBindTable(O); 1967 if (LazyBind) 1968 printLazyBindTable(O); 1969 if (WeakBind) 1970 printWeakBindTable(O); 1971 if (RawClangAST) 1972 printRawClangAST(O); 1973 if (PrintFaultMaps) 1974 printFaultMaps(O); 1975 if (DwarfDumpType != DIDT_Null) { 1976 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 1977 // Dump the complete DWARF structure. 1978 DIDumpOptions DumpOpts; 1979 DumpOpts.DumpType = DwarfDumpType; 1980 DICtx->dump(outs(), DumpOpts); 1981 } 1982 } 1983 1984 static void dumpObject(const COFFImportFile *I, const Archive *A, 1985 const Archive::Child *C = nullptr) { 1986 StringRef ArchiveName = A ? A->getFileName() : ""; 1987 1988 // Avoid other output when using a raw option. 1989 if (!RawClangAST) 1990 outs() << '\n' 1991 << ArchiveName << "(" << I->getFileName() << ")" 1992 << ":\tfile format COFF-import-file" 1993 << "\n\n"; 1994 1995 if (ArchiveHeaders && !MachOOpt && C) 1996 printArchiveChild(ArchiveName, *C); 1997 if (SymbolTable) 1998 printCOFFSymbolTable(I); 1999 } 2000 2001 /// Dump each object file in \a a; 2002 static void dumpArchive(const Archive *A) { 2003 Error Err = Error::success(); 2004 for (auto &C : A->children(Err)) { 2005 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2006 if (!ChildOrErr) { 2007 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2008 report_error(std::move(E), A->getFileName(), C); 2009 continue; 2010 } 2011 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2012 dumpObject(O, A, &C); 2013 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2014 dumpObject(I, A, &C); 2015 else 2016 report_error(A->getFileName(), object_error::invalid_file_type); 2017 } 2018 if (Err) 2019 report_error(std::move(Err), A->getFileName()); 2020 } 2021 2022 /// Open file and figure out how to dump it. 2023 static void dumpInput(StringRef file) { 2024 // If we are using the Mach-O specific object file parser, then let it parse 2025 // the file and process the command line options. So the -arch flags can 2026 // be used to select specific slices, etc. 2027 if (MachOOpt) { 2028 parseInputMachO(file); 2029 return; 2030 } 2031 2032 // Attempt to open the binary. 2033 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2034 Binary &Binary = *OBinary.getBinary(); 2035 2036 if (Archive *A = dyn_cast<Archive>(&Binary)) 2037 dumpArchive(A); 2038 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2039 dumpObject(O); 2040 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2041 parseInputMachO(UB); 2042 else 2043 report_error(file, object_error::invalid_file_type); 2044 } 2045 2046 int main(int argc, char **argv) { 2047 InitLLVM X(argc, argv); 2048 2049 // Initialize targets and assembly printers/parsers. 2050 llvm::InitializeAllTargetInfos(); 2051 llvm::InitializeAllTargetMCs(); 2052 llvm::InitializeAllDisassemblers(); 2053 2054 // Register the target printer for --version. 2055 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2056 2057 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2058 2059 ToolName = argv[0]; 2060 2061 // Defaults to a.out if no filenames specified. 2062 if (InputFilenames.empty()) 2063 InputFilenames.push_back("a.out"); 2064 2065 if (AllHeaders) 2066 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2067 SectionHeaders = SymbolTable = true; 2068 2069 if (DisassembleAll || PrintSource || PrintLines) 2070 Disassemble = true; 2071 2072 if (!Disassemble 2073 && !Relocations 2074 && !DynamicRelocations 2075 && !SectionHeaders 2076 && !SectionContents 2077 && !SymbolTable 2078 && !UnwindInfo 2079 && !PrivateHeaders 2080 && !FileHeaders 2081 && !FirstPrivateHeader 2082 && !ExportsTrie 2083 && !Rebase 2084 && !Bind 2085 && !LazyBind 2086 && !WeakBind 2087 && !RawClangAST 2088 && !(UniversalHeaders && MachOOpt) 2089 && !ArchiveHeaders 2090 && !(IndirectSymbols && MachOOpt) 2091 && !(DataInCode && MachOOpt) 2092 && !(LinkOptHints && MachOOpt) 2093 && !(InfoPlist && MachOOpt) 2094 && !(DylibsUsed && MachOOpt) 2095 && !(DylibId && MachOOpt) 2096 && !(ObjcMetaData && MachOOpt) 2097 && !(!FilterSections.empty() && MachOOpt) 2098 && !PrintFaultMaps 2099 && DwarfDumpType == DIDT_Null) { 2100 cl::PrintHelpMessage(); 2101 return 2; 2102 } 2103 2104 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2105 DisassembleFunctions.end()); 2106 2107 llvm::for_each(InputFilenames, dumpInput); 2108 2109 return EXIT_SUCCESS; 2110 } 2111