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<bool> 74 llvm::AllHeaders("all-headers", 75 cl::desc("Display all available header information")); 76 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 77 cl::NotHidden, cl::aliasopt(AllHeaders)); 78 79 static cl::list<std::string> 80 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 81 82 cl::opt<bool> 83 llvm::Disassemble("disassemble", 84 cl::desc("Display assembler mnemonics for the machine instructions")); 85 static cl::alias Disassembled("d", cl::desc("Alias for --disassemble"), 86 cl::NotHidden, cl::aliasopt(Disassemble)); 87 88 cl::opt<bool> 89 llvm::DisassembleAll("disassemble-all", 90 cl::desc("Display assembler mnemonics for the machine instructions")); 91 static cl::alias DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 92 cl::NotHidden, cl::aliasopt(DisassembleAll)); 93 94 cl::opt<bool> llvm::Demangle("demangle", cl::desc("Demangle symbols names"), 95 cl::init(false)); 96 97 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 98 cl::NotHidden, cl::aliasopt(llvm::Demangle)); 99 100 static cl::list<std::string> 101 DisassembleFunctions("df", 102 cl::CommaSeparated, 103 cl::desc("List of functions to disassemble")); 104 static StringSet<> DisasmFuncsSet; 105 106 cl::opt<bool> 107 llvm::Relocations("reloc", 108 cl::desc("Display the relocation entries in the file")); 109 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 110 cl::NotHidden, 111 cl::aliasopt(llvm::Relocations)); 112 113 cl::opt<bool> 114 llvm::DynamicRelocations("dynamic-reloc", 115 cl::desc("Display the dynamic relocation entries in the file")); 116 static cl::alias DynamicRelocationsd("R", cl::desc("Alias for --dynamic-reloc"), 117 cl::NotHidden, 118 cl::aliasopt(DynamicRelocations)); 119 120 cl::opt<bool> 121 llvm::SectionContents("full-contents", 122 cl::desc("Display the content of each section")); 123 static cl::alias SectionContentsShort("s", 124 cl::desc("Alias for --full-contents"), 125 cl::NotHidden, 126 cl::aliasopt(SectionContents)); 127 128 cl::opt<bool> llvm::SymbolTable("syms", cl::desc("Display the symbol table")); 129 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 130 cl::NotHidden, 131 cl::aliasopt(llvm::SymbolTable)); 132 133 cl::opt<bool> 134 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 135 136 cl::opt<bool> 137 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 138 139 cl::opt<bool> 140 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 141 142 cl::opt<bool> 143 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 144 145 cl::opt<bool> 146 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 147 148 cl::opt<bool> 149 llvm::RawClangAST("raw-clang-ast", 150 cl::desc("Dump the raw binary contents of the clang AST section")); 151 152 static cl::opt<bool> 153 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 154 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 155 cl::aliasopt(MachOOpt)); 156 157 cl::opt<std::string> 158 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 159 "see -version for available targets")); 160 161 cl::opt<std::string> 162 llvm::MCPU("mcpu", 163 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 164 cl::value_desc("cpu-name"), 165 cl::init("")); 166 167 cl::opt<std::string> 168 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 169 "see -version for available targets")); 170 171 cl::opt<bool> 172 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 173 "headers for each section.")); 174 static cl::alias SectionHeadersShort("headers", 175 cl::desc("Alias for --section-headers"), 176 cl::NotHidden, 177 cl::aliasopt(SectionHeaders)); 178 static cl::alias SectionHeadersShorter("h", 179 cl::desc("Alias for --section-headers"), 180 cl::NotHidden, 181 cl::aliasopt(SectionHeaders)); 182 183 cl::list<std::string> 184 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 185 "With -macho dump segment,section")); 186 cl::alias static FilterSectionsj("j", cl::desc("Alias for --section"), 187 cl::NotHidden, 188 cl::aliasopt(llvm::FilterSections)); 189 190 cl::list<std::string> 191 llvm::MAttrs("mattr", 192 cl::CommaSeparated, 193 cl::desc("Target specific attributes"), 194 cl::value_desc("a1,+a2,-a3,...")); 195 196 cl::opt<bool> 197 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 198 "instructions, do not print " 199 "the instruction bytes.")); 200 cl::opt<bool> 201 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 202 203 cl::opt<bool> 204 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 205 206 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 207 cl::NotHidden, cl::aliasopt(UnwindInfo)); 208 209 cl::opt<bool> 210 llvm::PrivateHeaders("private-headers", 211 cl::desc("Display format specific file headers")); 212 213 cl::opt<bool> 214 llvm::FirstPrivateHeader("private-header", 215 cl::desc("Display only the first format specific file " 216 "header")); 217 218 static cl::alias PrivateHeadersShort("p", 219 cl::desc("Alias for --private-headers"), 220 cl::NotHidden, 221 cl::aliasopt(PrivateHeaders)); 222 223 cl::opt<bool> llvm::FileHeaders( 224 "file-headers", 225 cl::desc("Display the contents of the overall file header")); 226 227 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 228 cl::NotHidden, cl::aliasopt(FileHeaders)); 229 230 cl::opt<bool> 231 llvm::ArchiveHeaders("archive-headers", 232 cl::desc("Display archive header information")); 233 234 cl::alias ArchiveHeadersShort("a", cl::desc("Alias for --archive-headers"), 235 cl::NotHidden, cl::aliasopt(ArchiveHeaders)); 236 237 cl::opt<bool> 238 llvm::PrintImmHex("print-imm-hex", 239 cl::desc("Use hex format for immediate values")); 240 241 cl::opt<bool> PrintFaultMaps("fault-map-section", 242 cl::desc("Display contents of faultmap section")); 243 244 cl::opt<DIDumpType> llvm::DwarfDumpType( 245 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 246 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 247 248 cl::opt<bool> PrintSource( 249 "source", 250 cl::desc( 251 "Display source inlined with disassembly. Implies disassemble object")); 252 253 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), cl::NotHidden, 254 cl::aliasopt(PrintSource)); 255 256 cl::opt<bool> PrintLines("line-numbers", 257 cl::desc("Display source line numbers with " 258 "disassembly. Implies disassemble object")); 259 260 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 261 cl::NotHidden, cl::aliasopt(PrintLines)); 262 263 cl::opt<unsigned long long> 264 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 265 cl::value_desc("address"), cl::init(0)); 266 cl::opt<unsigned long long> 267 StopAddress("stop-address", 268 cl::desc("Stop disassembly at address"), 269 cl::value_desc("address"), cl::init(UINT64_MAX)); 270 271 cl::opt<bool> DisassembleZeroes( 272 "disassemble-zeroes", 273 cl::desc("Do not skip blocks of zeroes when disassembling")); 274 cl::alias DisassembleZeroesShort("z", 275 cl::desc("Alias for --disassemble-zeroes"), 276 cl::NotHidden, 277 cl::aliasopt(DisassembleZeroes)); 278 279 static StringRef ToolName; 280 281 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 282 283 SectionFilter llvm::ToolSectionFilter(llvm::object::ObjectFile const &O) { 284 return SectionFilter( 285 [](llvm::object::SectionRef const &S) { 286 if (FilterSections.empty()) 287 return true; 288 llvm::StringRef String; 289 std::error_code error = S.getName(String); 290 if (error) 291 return false; 292 return is_contained(FilterSections, String); 293 }, 294 O); 295 } 296 297 void llvm::error(std::error_code EC) { 298 if (!EC) 299 return; 300 WithColor::error(errs(), ToolName) 301 << "reading file: " << EC.message() << ".\n"; 302 errs().flush(); 303 exit(1); 304 } 305 306 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 307 WithColor::error(errs(), ToolName) << Message << ".\n"; 308 errs().flush(); 309 exit(1); 310 } 311 312 void llvm::warn(StringRef Message) { 313 WithColor::warning(errs(), ToolName) << Message << ".\n"; 314 errs().flush(); 315 } 316 317 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 318 Twine Message) { 319 WithColor::error(errs(), ToolName) 320 << "'" << File << "': " << Message << ".\n"; 321 exit(1); 322 } 323 324 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 325 std::error_code EC) { 326 assert(EC); 327 WithColor::error(errs(), ToolName) 328 << "'" << File << "': " << EC.message() << ".\n"; 329 exit(1); 330 } 331 332 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 333 llvm::Error E) { 334 assert(E); 335 std::string Buf; 336 raw_string_ostream OS(Buf); 337 logAllUnhandledErrors(std::move(E), OS); 338 OS.flush(); 339 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 340 exit(1); 341 } 342 343 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 344 StringRef FileName, 345 llvm::Error E, 346 StringRef ArchitectureName) { 347 assert(E); 348 WithColor::error(errs(), ToolName); 349 if (ArchiveName != "") 350 errs() << ArchiveName << "(" << FileName << ")"; 351 else 352 errs() << "'" << FileName << "'"; 353 if (!ArchitectureName.empty()) 354 errs() << " (for architecture " << ArchitectureName << ")"; 355 std::string Buf; 356 raw_string_ostream OS(Buf); 357 logAllUnhandledErrors(std::move(E), OS); 358 OS.flush(); 359 errs() << ": " << Buf; 360 exit(1); 361 } 362 363 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 364 const object::Archive::Child &C, 365 llvm::Error E, 366 StringRef ArchitectureName) { 367 Expected<StringRef> NameOrErr = C.getName(); 368 // TODO: if we have a error getting the name then it would be nice to print 369 // the index of which archive member this is and or its offset in the 370 // archive instead of "???" as the name. 371 if (!NameOrErr) { 372 consumeError(NameOrErr.takeError()); 373 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 374 } else 375 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 376 ArchitectureName); 377 } 378 379 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 380 // Figure out the target triple. 381 llvm::Triple TheTriple("unknown-unknown-unknown"); 382 if (TripleName.empty()) { 383 if (Obj) 384 TheTriple = Obj->makeTriple(); 385 } else { 386 TheTriple.setTriple(Triple::normalize(TripleName)); 387 388 // Use the triple, but also try to combine with ARM build attributes. 389 if (Obj) { 390 auto Arch = Obj->getArch(); 391 if (Arch == Triple::arm || Arch == Triple::armeb) 392 Obj->setARMSubArch(TheTriple); 393 } 394 } 395 396 // Get the target specific parser. 397 std::string Error; 398 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 399 Error); 400 if (!TheTarget) { 401 if (Obj) 402 report_error(Obj->getFileName(), "can't find target: " + Error); 403 else 404 error("can't find target: " + Error); 405 } 406 407 // Update the triple name and return the found target. 408 TripleName = TheTriple.getTriple(); 409 return TheTarget; 410 } 411 412 bool llvm::isRelocAddressLess(RelocationRef A, RelocationRef B) { 413 return A.getOffset() < B.getOffset(); 414 } 415 416 static std::error_code getRelocationValueString(const RelocationRef &Rel, 417 SmallVectorImpl<char> &Result) { 418 const ObjectFile *Obj = Rel.getObject(); 419 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 420 return getELFRelocationValueString(ELF, Rel, Result); 421 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 422 return getCOFFRelocationValueString(COFF, Rel, Result); 423 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 424 return getWasmRelocationValueString(Wasm, Rel, Result); 425 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 426 return getMachORelocationValueString(MachO, Rel, Result); 427 llvm_unreachable("unknown object file format"); 428 } 429 430 /// Indicates whether this relocation should hidden when listing 431 /// relocations, usually because it is the trailing part of a multipart 432 /// relocation that will be printed as part of the leading relocation. 433 static bool getHidden(RelocationRef RelRef) { 434 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 435 if (!MachO) 436 return false; 437 438 unsigned Arch = MachO->getArch(); 439 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 440 uint64_t Type = MachO->getRelocationType(Rel); 441 442 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 443 // is always hidden. 444 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 445 return Type == MachO::GENERIC_RELOC_PAIR; 446 447 if (Arch == Triple::x86_64) { 448 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 449 // an X86_64_RELOC_SUBTRACTOR. 450 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 451 DataRefImpl RelPrev = Rel; 452 RelPrev.d.a--; 453 uint64_t PrevType = MachO->getRelocationType(RelPrev); 454 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 455 return true; 456 } 457 } 458 459 return false; 460 } 461 462 namespace { 463 class SourcePrinter { 464 protected: 465 DILineInfo OldLineInfo; 466 const ObjectFile *Obj = nullptr; 467 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 468 // File name to file contents of source 469 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 470 // Mark the line endings of the cached source 471 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 472 473 private: 474 bool cacheSource(const DILineInfo& LineInfoFile); 475 476 public: 477 SourcePrinter() = default; 478 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 479 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 480 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 481 DefaultArch); 482 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 483 } 484 virtual ~SourcePrinter() = default; 485 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 486 StringRef Delimiter = "; "); 487 }; 488 489 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 490 std::unique_ptr<MemoryBuffer> Buffer; 491 if (LineInfo.Source) { 492 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 493 } else { 494 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 495 if (!BufferOrError) 496 return false; 497 Buffer = std::move(*BufferOrError); 498 } 499 // Chomp the file to get lines 500 size_t BufferSize = Buffer->getBufferSize(); 501 const char *BufferStart = Buffer->getBufferStart(); 502 for (const char *Start = BufferStart, *End = BufferStart; 503 End < BufferStart + BufferSize; End++) 504 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 505 (*End == '\r' && *(End + 1) == '\n')) { 506 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 507 if (*End == '\r') 508 End++; 509 Start = End + 1; 510 } 511 SourceCache[LineInfo.FileName] = std::move(Buffer); 512 return true; 513 } 514 515 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 516 StringRef Delimiter) { 517 if (!Symbolizer) 518 return; 519 DILineInfo LineInfo = DILineInfo(); 520 auto ExpectecLineInfo = 521 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 522 if (!ExpectecLineInfo) 523 consumeError(ExpectecLineInfo.takeError()); 524 else 525 LineInfo = *ExpectecLineInfo; 526 527 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 528 LineInfo.Line == 0) 529 return; 530 531 if (PrintLines) 532 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 533 if (PrintSource) { 534 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 535 if (!cacheSource(LineInfo)) 536 return; 537 auto FileBuffer = SourceCache.find(LineInfo.FileName); 538 if (FileBuffer != SourceCache.end()) { 539 auto LineBuffer = LineCache.find(LineInfo.FileName); 540 if (LineBuffer != LineCache.end()) { 541 if (LineInfo.Line > LineBuffer->second.size()) 542 return; 543 // Vector begins at 0, line numbers are non-zero 544 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 545 << "\n"; 546 } 547 } 548 } 549 OldLineInfo = LineInfo; 550 } 551 552 static bool isArmElf(const ObjectFile *Obj) { 553 return (Obj->isELF() && 554 (Obj->getArch() == Triple::aarch64 || 555 Obj->getArch() == Triple::aarch64_be || 556 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 557 Obj->getArch() == Triple::thumb || 558 Obj->getArch() == Triple::thumbeb)); 559 } 560 561 class PrettyPrinter { 562 public: 563 virtual ~PrettyPrinter() = default; 564 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 565 ArrayRef<uint8_t> Bytes, uint64_t Address, 566 raw_ostream &OS, StringRef Annot, 567 MCSubtargetInfo const &STI, SourcePrinter *SP, 568 std::vector<RelocationRef> *Rels = nullptr) { 569 if (SP && (PrintSource || PrintLines)) 570 SP->printSourceLine(OS, Address); 571 if (!NoLeadingAddr) 572 OS << format("%8" PRIx64 ":", Address); 573 if (!NoShowRawInsn) { 574 OS << "\t"; 575 dumpBytes(Bytes, OS); 576 } 577 if (MI) 578 IP.printInst(MI, OS, "", STI); 579 else 580 OS << " <unknown>"; 581 } 582 }; 583 PrettyPrinter PrettyPrinterInst; 584 class HexagonPrettyPrinter : public PrettyPrinter { 585 public: 586 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 587 raw_ostream &OS) { 588 uint32_t opcode = 589 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 590 if (!NoLeadingAddr) 591 OS << format("%8" PRIx64 ":", Address); 592 if (!NoShowRawInsn) { 593 OS << "\t"; 594 dumpBytes(Bytes.slice(0, 4), OS); 595 OS << format("%08" PRIx32, opcode); 596 } 597 } 598 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 599 uint64_t Address, raw_ostream &OS, StringRef Annot, 600 MCSubtargetInfo const &STI, SourcePrinter *SP, 601 std::vector<RelocationRef> *Rels) override { 602 if (SP && (PrintSource || PrintLines)) 603 SP->printSourceLine(OS, Address, ""); 604 if (!MI) { 605 printLead(Bytes, Address, OS); 606 OS << " <unknown>"; 607 return; 608 } 609 std::string Buffer; 610 { 611 raw_string_ostream TempStream(Buffer); 612 IP.printInst(MI, TempStream, "", STI); 613 } 614 StringRef Contents(Buffer); 615 // Split off bundle attributes 616 auto PacketBundle = Contents.rsplit('\n'); 617 // Split off first instruction from the rest 618 auto HeadTail = PacketBundle.first.split('\n'); 619 auto Preamble = " { "; 620 auto Separator = ""; 621 StringRef Fmt = "\t\t\t%08" PRIx64 ": "; 622 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 623 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 624 625 // Hexagon's packets require relocations to be inline rather than 626 // clustered at the end of the packet. 627 auto PrintReloc = [&]() -> void { 628 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address)) { 629 if (RelCur->getOffset() == Address) { 630 SmallString<16> Name; 631 SmallString<32> Val; 632 RelCur->getTypeName(Name); 633 error(getRelocationValueString(*RelCur, Val)); 634 OS << Separator << format(Fmt.data(), Address) << Name << "\t" << Val 635 << "\n"; 636 return; 637 } 638 ++RelCur; 639 } 640 }; 641 642 while (!HeadTail.first.empty()) { 643 OS << Separator; 644 Separator = "\n"; 645 if (SP && (PrintSource || PrintLines)) 646 SP->printSourceLine(OS, Address, ""); 647 printLead(Bytes, Address, OS); 648 OS << Preamble; 649 Preamble = " "; 650 StringRef Inst; 651 auto Duplex = HeadTail.first.split('\v'); 652 if (!Duplex.second.empty()) { 653 OS << Duplex.first; 654 OS << "; "; 655 Inst = Duplex.second; 656 } 657 else 658 Inst = HeadTail.first; 659 OS << Inst; 660 HeadTail = HeadTail.second.split('\n'); 661 if (HeadTail.first.empty()) 662 OS << " } " << PacketBundle.second; 663 PrintReloc(); 664 Bytes = Bytes.slice(4); 665 Address += 4; 666 } 667 } 668 }; 669 HexagonPrettyPrinter HexagonPrettyPrinterInst; 670 671 class AMDGCNPrettyPrinter : public PrettyPrinter { 672 public: 673 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 674 uint64_t Address, raw_ostream &OS, StringRef Annot, 675 MCSubtargetInfo const &STI, SourcePrinter *SP, 676 std::vector<RelocationRef> *Rels) override { 677 if (SP && (PrintSource || PrintLines)) 678 SP->printSourceLine(OS, Address); 679 680 typedef support::ulittle32_t U32; 681 682 if (MI) { 683 SmallString<40> InstStr; 684 raw_svector_ostream IS(InstStr); 685 686 IP.printInst(MI, IS, "", STI); 687 688 OS << left_justify(IS.str(), 60); 689 } else { 690 // an unrecognized encoding - this is probably data so represent it 691 // using the .long directive, or .byte directive if fewer than 4 bytes 692 // remaining 693 if (Bytes.size() >= 4) { 694 OS << format("\t.long 0x%08" PRIx32 " ", 695 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 696 OS.indent(42); 697 } else { 698 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 699 for (unsigned int i = 1; i < Bytes.size(); i++) 700 OS << format(", 0x%02" PRIx8, Bytes[i]); 701 OS.indent(55 - (6 * Bytes.size())); 702 } 703 } 704 705 OS << format("// %012" PRIX64 ": ", Address); 706 if (Bytes.size() >=4) { 707 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 708 Bytes.size() / sizeof(U32))) 709 // D should be explicitly casted to uint32_t here as it is passed 710 // by format to snprintf as vararg. 711 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 712 } else { 713 for (unsigned int i = 0; i < Bytes.size(); i++) 714 OS << format("%02" PRIX8 " ", Bytes[i]); 715 } 716 717 if (!Annot.empty()) 718 OS << "// " << Annot; 719 } 720 }; 721 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 722 723 class BPFPrettyPrinter : public PrettyPrinter { 724 public: 725 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 726 uint64_t Address, raw_ostream &OS, StringRef Annot, 727 MCSubtargetInfo const &STI, SourcePrinter *SP, 728 std::vector<RelocationRef> *Rels) override { 729 if (SP && (PrintSource || PrintLines)) 730 SP->printSourceLine(OS, Address); 731 if (!NoLeadingAddr) 732 OS << format("%8" PRId64 ":", Address / 8); 733 if (!NoShowRawInsn) { 734 OS << "\t"; 735 dumpBytes(Bytes, OS); 736 } 737 if (MI) 738 IP.printInst(MI, OS, "", STI); 739 else 740 OS << " <unknown>"; 741 } 742 }; 743 BPFPrettyPrinter BPFPrettyPrinterInst; 744 745 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 746 switch(Triple.getArch()) { 747 default: 748 return PrettyPrinterInst; 749 case Triple::hexagon: 750 return HexagonPrettyPrinterInst; 751 case Triple::amdgcn: 752 return AMDGCNPrettyPrinterInst; 753 case Triple::bpfel: 754 case Triple::bpfeb: 755 return BPFPrettyPrinterInst; 756 } 757 } 758 } 759 760 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 761 assert(Obj->isELF()); 762 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 763 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 764 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 765 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 766 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 767 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 768 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 769 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 770 llvm_unreachable("Unsupported binary format"); 771 } 772 773 template <class ELFT> static void 774 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 775 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 776 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 777 uint8_t SymbolType = Symbol.getELFType(); 778 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 779 continue; 780 781 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 782 if (!AddressOrErr) 783 report_error(Obj->getFileName(), AddressOrErr.takeError()); 784 785 Expected<StringRef> Name = Symbol.getName(); 786 if (!Name) 787 report_error(Obj->getFileName(), Name.takeError()); 788 if (Name->empty()) 789 continue; 790 791 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 792 if (!SectionOrErr) 793 report_error(Obj->getFileName(), SectionOrErr.takeError()); 794 section_iterator SecI = *SectionOrErr; 795 if (SecI == Obj->section_end()) 796 continue; 797 798 AllSymbols[*SecI].emplace_back(*AddressOrErr, *Name, SymbolType); 799 } 800 } 801 802 static void 803 addDynamicElfSymbols(const ObjectFile *Obj, 804 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 805 assert(Obj->isELF()); 806 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 807 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 808 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 809 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 810 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 811 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 812 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 813 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 814 else 815 llvm_unreachable("Unsupported binary format"); 816 } 817 818 static void addPltEntries(const ObjectFile *Obj, 819 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 820 StringSaver &Saver) { 821 Optional<SectionRef> Plt = None; 822 for (const SectionRef &Section : Obj->sections()) { 823 StringRef Name; 824 if (Section.getName(Name)) 825 continue; 826 if (Name == ".plt") 827 Plt = Section; 828 } 829 if (!Plt) 830 return; 831 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 832 for (auto PltEntry : ElfObj->getPltAddresses()) { 833 SymbolRef Symbol(PltEntry.first, ElfObj); 834 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 835 836 Expected<StringRef> NameOrErr = Symbol.getName(); 837 if (!NameOrErr) 838 report_error(Obj->getFileName(), NameOrErr.takeError()); 839 if (NameOrErr->empty()) 840 continue; 841 StringRef Name = Saver.save((*NameOrErr + "@plt").str()); 842 843 AllSymbols[*Plt].emplace_back(PltEntry.second, Name, SymbolType); 844 } 845 } 846 } 847 848 // Normally the disassembly output will skip blocks of zeroes. This function 849 // returns the number of zero bytes that can be skipped when dumping the 850 // disassembly of the instructions in Buf. 851 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 852 // When -z or --disassemble-zeroes are given we always dissasemble them. 853 if (DisassembleZeroes) 854 return 0; 855 856 // Find the number of leading zeroes. 857 size_t N = 0; 858 while (N < Buf.size() && !Buf[N]) 859 ++N; 860 861 // We may want to skip blocks of zero bytes, but unless we see 862 // at least 8 of them in a row. 863 if (N < 8) 864 return 0; 865 866 // We skip zeroes in multiples of 4 because do not want to truncate an 867 // instruction if it starts with a zero byte. 868 return N & ~0x3; 869 } 870 871 // Returns a map from sections to their relocations. 872 static std::map<SectionRef, std::vector<RelocationRef>> 873 getRelocsMap(llvm::object::ObjectFile const &Obj) { 874 std::map<SectionRef, std::vector<RelocationRef>> Ret; 875 for (const SectionRef &Section : ToolSectionFilter(Obj)) { 876 section_iterator RelSec = Section.getRelocatedSection(); 877 if (RelSec == Obj.section_end()) 878 continue; 879 std::vector<RelocationRef> &V = Ret[*RelSec]; 880 for (const RelocationRef &R : Section.relocations()) 881 V.push_back(R); 882 // Sort relocations by address. 883 llvm::sort(V, isRelocAddressLess); 884 } 885 return Ret; 886 } 887 888 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 889 MCContext &Ctx, MCDisassembler *DisAsm, 890 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 891 const MCSubtargetInfo *STI, PrettyPrinter &PIP, 892 SourcePrinter &SP, bool InlineRelocs) { 893 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " 894 : "\t\t\t%08" PRIx64 ": "; 895 896 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 897 if (InlineRelocs) 898 RelocMap = getRelocsMap(*Obj); 899 900 // Create a mapping from virtual address to symbol name. This is used to 901 // pretty print the symbols while disassembling. 902 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 903 SectionSymbolsTy AbsoluteSymbols; 904 for (const SymbolRef &Symbol : Obj->symbols()) { 905 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 906 if (!AddressOrErr) 907 report_error(Obj->getFileName(), AddressOrErr.takeError()); 908 uint64_t Address = *AddressOrErr; 909 910 Expected<StringRef> Name = Symbol.getName(); 911 if (!Name) 912 report_error(Obj->getFileName(), Name.takeError()); 913 if (Name->empty()) 914 continue; 915 916 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 917 if (!SectionOrErr) 918 report_error(Obj->getFileName(), SectionOrErr.takeError()); 919 920 uint8_t SymbolType = ELF::STT_NOTYPE; 921 if (Obj->isELF()) 922 SymbolType = getElfSymbolType(Obj, Symbol); 923 924 section_iterator SecI = *SectionOrErr; 925 if (SecI != Obj->section_end()) 926 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 927 else 928 AbsoluteSymbols.emplace_back(Address, *Name, SymbolType); 929 930 931 } 932 if (AllSymbols.empty() && Obj->isELF()) 933 addDynamicElfSymbols(Obj, AllSymbols); 934 935 BumpPtrAllocator A; 936 StringSaver Saver(A); 937 addPltEntries(Obj, AllSymbols, Saver); 938 939 // Create a mapping from virtual address to section. 940 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 941 for (SectionRef Sec : Obj->sections()) 942 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 943 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 944 945 // Linked executables (.exe and .dll files) typically don't include a real 946 // symbol table but they might contain an export table. 947 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 948 for (const auto &ExportEntry : COFFObj->export_directories()) { 949 StringRef Name; 950 error(ExportEntry.getSymbolName(Name)); 951 if (Name.empty()) 952 continue; 953 uint32_t RVA; 954 error(ExportEntry.getExportRVA(RVA)); 955 956 uint64_t VA = COFFObj->getImageBase() + RVA; 957 auto Sec = std::upper_bound( 958 SectionAddresses.begin(), SectionAddresses.end(), VA, 959 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 960 return LHS < RHS.first; 961 }); 962 if (Sec != SectionAddresses.begin()) 963 --Sec; 964 else 965 Sec = SectionAddresses.end(); 966 967 if (Sec != SectionAddresses.end()) 968 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 969 else 970 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 971 } 972 } 973 974 // Sort all the symbols, this allows us to use a simple binary search to find 975 // a symbol near an address. 976 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 977 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 978 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 979 980 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 981 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 982 continue; 983 984 uint64_t SectionAddr = Section.getAddress(); 985 uint64_t SectSize = Section.getSize(); 986 if (!SectSize) 987 continue; 988 989 // Get the list of all the symbols in this section. 990 SectionSymbolsTy &Symbols = AllSymbols[Section]; 991 std::vector<uint64_t> DataMappingSymsAddr; 992 std::vector<uint64_t> TextMappingSymsAddr; 993 if (isArmElf(Obj)) { 994 for (const auto &Symb : Symbols) { 995 uint64_t Address = std::get<0>(Symb); 996 StringRef Name = std::get<1>(Symb); 997 if (Name.startswith("$d")) 998 DataMappingSymsAddr.push_back(Address - SectionAddr); 999 if (Name.startswith("$x")) 1000 TextMappingSymsAddr.push_back(Address - SectionAddr); 1001 if (Name.startswith("$a")) 1002 TextMappingSymsAddr.push_back(Address - SectionAddr); 1003 if (Name.startswith("$t")) 1004 TextMappingSymsAddr.push_back(Address - SectionAddr); 1005 } 1006 } 1007 1008 llvm::sort(DataMappingSymsAddr); 1009 llvm::sort(TextMappingSymsAddr); 1010 1011 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1012 // AMDGPU disassembler uses symbolizer for printing labels 1013 std::unique_ptr<MCRelocationInfo> RelInfo( 1014 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1015 if (RelInfo) { 1016 std::unique_ptr<MCSymbolizer> Symbolizer( 1017 TheTarget->createMCSymbolizer( 1018 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1019 DisAsm->setSymbolizer(std::move(Symbolizer)); 1020 } 1021 } 1022 1023 StringRef SegmentName = ""; 1024 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1025 DataRefImpl DR = Section.getRawDataRefImpl(); 1026 SegmentName = MachO->getSectionFinalSegmentName(DR); 1027 } 1028 StringRef SectionName; 1029 error(Section.getName(SectionName)); 1030 1031 // If the section has no symbol at the start, just insert a dummy one. 1032 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1033 Symbols.insert( 1034 Symbols.begin(), 1035 std::make_tuple(SectionAddr, SectionName, 1036 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1037 } 1038 1039 SmallString<40> Comments; 1040 raw_svector_ostream CommentStream(Comments); 1041 1042 StringRef BytesStr; 1043 error(Section.getContents(BytesStr)); 1044 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1045 BytesStr.size()); 1046 1047 uint64_t Size; 1048 uint64_t Index; 1049 bool PrintedSection = false; 1050 1051 std::vector<RelocationRef> Rels = RelocMap[Section]; 1052 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1053 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1054 // Disassemble symbol by symbol. 1055 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1056 uint64_t Start = std::get<0>(Symbols[SI]) - SectionAddr; 1057 // The end is either the section end or the beginning of the next 1058 // symbol. 1059 uint64_t End = (SI == SE - 1) 1060 ? SectSize 1061 : std::get<0>(Symbols[SI + 1]) - SectionAddr; 1062 // Don't try to disassemble beyond the end of section contents. 1063 if (End > SectSize) 1064 End = SectSize; 1065 // If this symbol has the same address as the next symbol, then skip it. 1066 if (Start >= End) 1067 continue; 1068 1069 // Check if we need to skip symbol 1070 // Skip if the symbol's data is not between StartAddress and StopAddress 1071 if (End + SectionAddr < StartAddress || 1072 Start + SectionAddr > StopAddress) { 1073 continue; 1074 } 1075 1076 /// Skip if user requested specific symbols and this is not in the list 1077 if (!DisasmFuncsSet.empty() && 1078 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1079 continue; 1080 1081 if (!PrintedSection) { 1082 PrintedSection = true; 1083 outs() << "Disassembly of section "; 1084 if (!SegmentName.empty()) 1085 outs() << SegmentName << ","; 1086 outs() << SectionName << ':'; 1087 } 1088 1089 // Stop disassembly at the stop address specified 1090 if (End + SectionAddr > StopAddress) 1091 End = StopAddress - SectionAddr; 1092 1093 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1094 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1095 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1096 Start += 256; 1097 } 1098 if (SI == SE - 1 || 1099 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1100 // cut trailing zeroes at the end of kernel 1101 // cut up to 256 bytes 1102 const uint64_t EndAlign = 256; 1103 const auto Limit = End - (std::min)(EndAlign, End - Start); 1104 while (End > Limit && 1105 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1106 End -= 4; 1107 } 1108 } 1109 1110 outs() << '\n'; 1111 if (!NoLeadingAddr) 1112 outs() << format("%016" PRIx64 " ", SectionAddr + Start); 1113 1114 StringRef SymbolName = std::get<1>(Symbols[SI]); 1115 if (Demangle) 1116 outs() << demangle(SymbolName) << ":\n"; 1117 else 1118 outs() << SymbolName << ":\n"; 1119 1120 // Don't print raw contents of a virtual section. A virtual section 1121 // doesn't have any contents in the file. 1122 if (Section.isVirtual()) { 1123 outs() << "...\n"; 1124 continue; 1125 } 1126 1127 #ifndef NDEBUG 1128 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1129 #else 1130 raw_ostream &DebugOut = nulls(); 1131 #endif 1132 1133 // Some targets (like WebAssembly) have a special prelude at the start 1134 // of each symbol. 1135 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1136 SectionAddr + Start, DebugOut, CommentStream); 1137 Start += Size; 1138 1139 for (Index = Start; Index < End; Index += Size) { 1140 MCInst Inst; 1141 1142 if (Index + SectionAddr < StartAddress || 1143 Index + SectionAddr > StopAddress) { 1144 // skip byte by byte till StartAddress is reached 1145 Size = 1; 1146 continue; 1147 } 1148 // AArch64 ELF binaries can interleave data and text in the 1149 // same section. We rely on the markers introduced to 1150 // understand what we need to dump. If the data marker is within a 1151 // function, it is denoted as a word/short etc 1152 if (isArmElf(Obj) && std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1153 !DisassembleAll) { 1154 uint64_t Stride = 0; 1155 1156 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1157 DataMappingSymsAddr.end(), Index); 1158 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1159 // Switch to data. 1160 while (Index < End) { 1161 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1162 outs() << "\t"; 1163 if (Index + 4 <= End) { 1164 Stride = 4; 1165 dumpBytes(Bytes.slice(Index, 4), outs()); 1166 outs() << "\t.word\t"; 1167 uint32_t Data = 0; 1168 if (Obj->isLittleEndian()) { 1169 const auto Word = 1170 reinterpret_cast<const support::ulittle32_t *>( 1171 Bytes.data() + Index); 1172 Data = *Word; 1173 } else { 1174 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1175 Bytes.data() + Index); 1176 Data = *Word; 1177 } 1178 outs() << "0x" << format("%08" PRIx32, Data); 1179 } else if (Index + 2 <= End) { 1180 Stride = 2; 1181 dumpBytes(Bytes.slice(Index, 2), outs()); 1182 outs() << "\t\t.short\t"; 1183 uint16_t Data = 0; 1184 if (Obj->isLittleEndian()) { 1185 const auto Short = 1186 reinterpret_cast<const support::ulittle16_t *>( 1187 Bytes.data() + Index); 1188 Data = *Short; 1189 } else { 1190 const auto Short = 1191 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1192 Index); 1193 Data = *Short; 1194 } 1195 outs() << "0x" << format("%04" PRIx16, Data); 1196 } else { 1197 Stride = 1; 1198 dumpBytes(Bytes.slice(Index, 1), outs()); 1199 outs() << "\t\t.byte\t"; 1200 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1201 } 1202 Index += Stride; 1203 outs() << "\n"; 1204 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1205 TextMappingSymsAddr.end(), Index); 1206 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1207 break; 1208 } 1209 } 1210 } 1211 1212 // If there is a data symbol inside an ELF text section and we are only 1213 // disassembling text (applicable all architectures), 1214 // we are in a situation where we must print the data and not 1215 // disassemble it. 1216 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1217 !DisassembleAll && Section.isText()) { 1218 // print out data up to 8 bytes at a time in hex and ascii 1219 uint8_t AsciiData[9] = {'\0'}; 1220 uint8_t Byte; 1221 int NumBytes = 0; 1222 1223 for (Index = Start; Index < End; Index += 1) { 1224 if (((SectionAddr + Index) < StartAddress) || 1225 ((SectionAddr + Index) > StopAddress)) 1226 continue; 1227 if (NumBytes == 0) { 1228 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1229 outs() << "\t"; 1230 } 1231 Byte = Bytes.slice(Index)[0]; 1232 outs() << format(" %02x", Byte); 1233 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1234 1235 uint8_t IndentOffset = 0; 1236 NumBytes++; 1237 if (Index == End - 1 || NumBytes > 8) { 1238 // Indent the space for less than 8 bytes data. 1239 // 2 spaces for byte and one for space between bytes 1240 IndentOffset = 3 * (8 - NumBytes); 1241 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1242 AsciiData[Excess] = '\0'; 1243 NumBytes = 8; 1244 } 1245 if (NumBytes == 8) { 1246 AsciiData[8] = '\0'; 1247 outs() << std::string(IndentOffset, ' ') << " "; 1248 outs() << reinterpret_cast<char *>(AsciiData); 1249 outs() << '\n'; 1250 NumBytes = 0; 1251 } 1252 } 1253 } 1254 if (Index >= End) 1255 break; 1256 1257 if (size_t N = 1258 countSkippableZeroBytes(Bytes.slice(Index, End - Index))) { 1259 outs() << "\t\t..." << '\n'; 1260 Index += N; 1261 if (Index >= End) 1262 break; 1263 } 1264 1265 // Disassemble a real instruction or a data when disassemble all is 1266 // provided 1267 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1268 SectionAddr + Index, DebugOut, 1269 CommentStream); 1270 if (Size == 0) 1271 Size = 1; 1272 1273 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1274 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1275 *STI, &SP, &Rels); 1276 outs() << CommentStream.str(); 1277 Comments.clear(); 1278 1279 // Try to resolve the target of a call, tail call, etc. to a specific 1280 // symbol. 1281 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1282 MIA->isConditionalBranch(Inst))) { 1283 uint64_t Target; 1284 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1285 // In a relocatable object, the target's section must reside in 1286 // the same section as the call instruction or it is accessed 1287 // through a relocation. 1288 // 1289 // In a non-relocatable object, the target may be in any section. 1290 // 1291 // N.B. We don't walk the relocations in the relocatable case yet. 1292 auto *TargetSectionSymbols = &Symbols; 1293 if (!Obj->isRelocatableObject()) { 1294 auto SectionAddress = std::upper_bound( 1295 SectionAddresses.begin(), SectionAddresses.end(), Target, 1296 [](uint64_t LHS, 1297 const std::pair<uint64_t, SectionRef> &RHS) { 1298 return LHS < RHS.first; 1299 }); 1300 if (SectionAddress != SectionAddresses.begin()) { 1301 --SectionAddress; 1302 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1303 } else { 1304 TargetSectionSymbols = &AbsoluteSymbols; 1305 } 1306 } 1307 1308 // Find the first symbol in the section whose offset is less than 1309 // or equal to the target. If there isn't a section that contains 1310 // the target, find the nearest preceding absolute symbol. 1311 auto TargetSym = std::upper_bound( 1312 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1313 Target, [](uint64_t LHS, 1314 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1315 return LHS < std::get<0>(RHS); 1316 }); 1317 if (TargetSym == TargetSectionSymbols->begin()) { 1318 TargetSectionSymbols = &AbsoluteSymbols; 1319 TargetSym = std::upper_bound( 1320 AbsoluteSymbols.begin(), AbsoluteSymbols.end(), 1321 Target, [](uint64_t LHS, 1322 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1323 return LHS < std::get<0>(RHS); 1324 }); 1325 } 1326 if (TargetSym != TargetSectionSymbols->begin()) { 1327 --TargetSym; 1328 uint64_t TargetAddress = std::get<0>(*TargetSym); 1329 StringRef TargetName = std::get<1>(*TargetSym); 1330 outs() << " <" << TargetName; 1331 uint64_t Disp = Target - TargetAddress; 1332 if (Disp) 1333 outs() << "+0x" << Twine::utohexstr(Disp); 1334 outs() << '>'; 1335 } 1336 } 1337 } 1338 outs() << "\n"; 1339 1340 // Hexagon does this in pretty printer 1341 if (Obj->getArch() != Triple::hexagon) 1342 // Print relocation for instruction. 1343 while (RelCur != RelEnd) { 1344 uint64_t Addr = RelCur->getOffset(); 1345 SmallString<16> Name; 1346 SmallString<32> Val; 1347 1348 // If this relocation is hidden, skip it. 1349 if (getHidden(*RelCur) || ((SectionAddr + Addr) < StartAddress)) { 1350 ++RelCur; 1351 continue; 1352 } 1353 1354 // Stop when rel_cur's address is past the current instruction. 1355 if (Addr >= Index + Size) 1356 break; 1357 RelCur->getTypeName(Name); 1358 error(getRelocationValueString(*RelCur, Val)); 1359 outs() << format(Fmt.data(), SectionAddr + Addr) << Name << "\t" 1360 << Val << "\n"; 1361 ++RelCur; 1362 } 1363 } 1364 } 1365 } 1366 } 1367 1368 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1369 if (StartAddress > StopAddress) 1370 error("Start address should be less than stop address"); 1371 1372 const Target *TheTarget = getTarget(Obj); 1373 1374 // Package up features to be passed to target/subtarget 1375 SubtargetFeatures Features = Obj->getFeatures(); 1376 if (!MAttrs.empty()) 1377 for (unsigned I = 0; I != MAttrs.size(); ++I) 1378 Features.AddFeature(MAttrs[I]); 1379 1380 std::unique_ptr<const MCRegisterInfo> MRI( 1381 TheTarget->createMCRegInfo(TripleName)); 1382 if (!MRI) 1383 report_error(Obj->getFileName(), 1384 "no register info for target " + TripleName); 1385 1386 // Set up disassembler. 1387 std::unique_ptr<const MCAsmInfo> AsmInfo( 1388 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1389 if (!AsmInfo) 1390 report_error(Obj->getFileName(), 1391 "no assembly info for target " + TripleName); 1392 std::unique_ptr<const MCSubtargetInfo> STI( 1393 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1394 if (!STI) 1395 report_error(Obj->getFileName(), 1396 "no subtarget info for target " + TripleName); 1397 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1398 if (!MII) 1399 report_error(Obj->getFileName(), 1400 "no instruction info for target " + TripleName); 1401 MCObjectFileInfo MOFI; 1402 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1403 // FIXME: for now initialize MCObjectFileInfo with default values 1404 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1405 1406 std::unique_ptr<MCDisassembler> DisAsm( 1407 TheTarget->createMCDisassembler(*STI, Ctx)); 1408 if (!DisAsm) 1409 report_error(Obj->getFileName(), 1410 "no disassembler for target " + TripleName); 1411 1412 std::unique_ptr<const MCInstrAnalysis> MIA( 1413 TheTarget->createMCInstrAnalysis(MII.get())); 1414 1415 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1416 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1417 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1418 if (!IP) 1419 report_error(Obj->getFileName(), 1420 "no instruction printer for target " + TripleName); 1421 IP->setPrintImmHex(PrintImmHex); 1422 1423 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1424 SourcePrinter SP(Obj, TheTarget->getName()); 1425 1426 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), MIA.get(), IP.get(), 1427 STI.get(), PIP, SP, InlineRelocs); 1428 } 1429 1430 void llvm::printRelocations(const ObjectFile *Obj) { 1431 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1432 "%08" PRIx64; 1433 // Regular objdump doesn't print relocations in non-relocatable object 1434 // files. 1435 if (!Obj->isRelocatableObject()) 1436 return; 1437 1438 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1439 if (Section.relocation_begin() == Section.relocation_end()) 1440 continue; 1441 StringRef SecName; 1442 error(Section.getName(SecName)); 1443 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1444 for (const RelocationRef &Reloc : Section.relocations()) { 1445 uint64_t Address = Reloc.getOffset(); 1446 SmallString<32> RelocName; 1447 SmallString<32> ValueStr; 1448 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1449 continue; 1450 Reloc.getTypeName(RelocName); 1451 error(getRelocationValueString(Reloc, ValueStr)); 1452 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1453 << ValueStr << "\n"; 1454 } 1455 outs() << "\n"; 1456 } 1457 } 1458 1459 void llvm::printDynamicRelocations(const ObjectFile *Obj) { 1460 // For the moment, this option is for ELF only 1461 if (!Obj->isELF()) 1462 return; 1463 1464 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1465 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1466 error("not a dynamic object"); 1467 return; 1468 } 1469 1470 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1471 if (DynRelSec.empty()) 1472 return; 1473 1474 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1475 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1476 for (const SectionRef &Section : DynRelSec) { 1477 if (Section.relocation_begin() == Section.relocation_end()) 1478 continue; 1479 for (const RelocationRef &Reloc : Section.relocations()) { 1480 uint64_t Address = Reloc.getOffset(); 1481 SmallString<32> RelocName; 1482 SmallString<32> ValueStr; 1483 Reloc.getTypeName(RelocName); 1484 error(getRelocationValueString(Reloc, ValueStr)); 1485 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1486 << ValueStr << "\n"; 1487 } 1488 } 1489 } 1490 1491 void llvm::printSectionHeaders(const ObjectFile *Obj) { 1492 outs() << "Sections:\n" 1493 "Idx Name Size Address Type\n"; 1494 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1495 StringRef Name; 1496 error(Section.getName(Name)); 1497 uint64_t Address = Section.getAddress(); 1498 uint64_t Size = Section.getSize(); 1499 bool Text = Section.isText(); 1500 bool Data = Section.isData(); 1501 bool BSS = Section.isBSS(); 1502 std::string Type = (std::string(Text ? "TEXT " : "") + 1503 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1504 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1505 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1506 Address, Type.c_str()); 1507 } 1508 outs() << "\n"; 1509 } 1510 1511 void llvm::printSectionContents(const ObjectFile *Obj) { 1512 std::error_code EC; 1513 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1514 StringRef Name; 1515 StringRef Contents; 1516 error(Section.getName(Name)); 1517 uint64_t BaseAddr = Section.getAddress(); 1518 uint64_t Size = Section.getSize(); 1519 if (!Size) 1520 continue; 1521 1522 outs() << "Contents of section " << Name << ":\n"; 1523 if (Section.isBSS()) { 1524 outs() << format("<skipping contents of bss section at [%04" PRIx64 1525 ", %04" PRIx64 ")>\n", 1526 BaseAddr, BaseAddr + Size); 1527 continue; 1528 } 1529 1530 error(Section.getContents(Contents)); 1531 1532 // Dump out the content as hex and printable ascii characters. 1533 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1534 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1535 // Dump line of hex. 1536 for (std::size_t I = 0; I < 16; ++I) { 1537 if (I != 0 && I % 4 == 0) 1538 outs() << ' '; 1539 if (Addr + I < End) 1540 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1541 << hexdigit(Contents[Addr + I] & 0xF, true); 1542 else 1543 outs() << " "; 1544 } 1545 // Print ascii. 1546 outs() << " "; 1547 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1548 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1549 outs() << Contents[Addr + I]; 1550 else 1551 outs() << "."; 1552 } 1553 outs() << "\n"; 1554 } 1555 } 1556 } 1557 1558 void llvm::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1559 StringRef ArchitectureName) { 1560 outs() << "SYMBOL TABLE:\n"; 1561 1562 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1563 printCOFFSymbolTable(Coff); 1564 return; 1565 } 1566 1567 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1568 // Skip printing the special zero symbol when dumping an ELF file. 1569 // This makes the output consistent with the GNU objdump. 1570 if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O)) 1571 continue; 1572 1573 const SymbolRef &Symbol = *I; 1574 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1575 if (!AddressOrError) 1576 report_error(ArchiveName, O->getFileName(), AddressOrError.takeError(), 1577 ArchitectureName); 1578 uint64_t Address = *AddressOrError; 1579 if ((Address < StartAddress) || (Address > StopAddress)) 1580 continue; 1581 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1582 if (!TypeOrError) 1583 report_error(ArchiveName, O->getFileName(), TypeOrError.takeError(), 1584 ArchitectureName); 1585 SymbolRef::Type Type = *TypeOrError; 1586 uint32_t Flags = Symbol.getFlags(); 1587 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1588 if (!SectionOrErr) 1589 report_error(ArchiveName, O->getFileName(), SectionOrErr.takeError(), 1590 ArchitectureName); 1591 section_iterator Section = *SectionOrErr; 1592 StringRef Name; 1593 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 1594 Section->getName(Name); 1595 } else { 1596 Expected<StringRef> NameOrErr = Symbol.getName(); 1597 if (!NameOrErr) 1598 report_error(ArchiveName, O->getFileName(), NameOrErr.takeError(), 1599 ArchitectureName); 1600 Name = *NameOrErr; 1601 } 1602 1603 bool Global = Flags & SymbolRef::SF_Global; 1604 bool Weak = Flags & SymbolRef::SF_Weak; 1605 bool Absolute = Flags & SymbolRef::SF_Absolute; 1606 bool Common = Flags & SymbolRef::SF_Common; 1607 bool Hidden = Flags & SymbolRef::SF_Hidden; 1608 1609 char GlobLoc = ' '; 1610 if (Type != SymbolRef::ST_Unknown) 1611 GlobLoc = Global ? 'g' : 'l'; 1612 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1613 ? 'd' : ' '; 1614 char FileFunc = ' '; 1615 if (Type == SymbolRef::ST_File) 1616 FileFunc = 'f'; 1617 else if (Type == SymbolRef::ST_Function) 1618 FileFunc = 'F'; 1619 else if (Type == SymbolRef::ST_Data) 1620 FileFunc = 'O'; 1621 1622 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1623 "%08" PRIx64; 1624 1625 outs() << format(Fmt, Address) << " " 1626 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1627 << (Weak ? 'w' : ' ') // Weak? 1628 << ' ' // Constructor. Not supported yet. 1629 << ' ' // Warning. Not supported yet. 1630 << ' ' // Indirect reference to another symbol. 1631 << Debug // Debugging (d) or dynamic (D) symbol. 1632 << FileFunc // Name of function (F), file (f) or object (O). 1633 << ' '; 1634 if (Absolute) { 1635 outs() << "*ABS*"; 1636 } else if (Common) { 1637 outs() << "*COM*"; 1638 } else if (Section == O->section_end()) { 1639 outs() << "*UND*"; 1640 } else { 1641 if (const MachOObjectFile *MachO = 1642 dyn_cast<const MachOObjectFile>(O)) { 1643 DataRefImpl DR = Section->getRawDataRefImpl(); 1644 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1645 outs() << SegmentName << ","; 1646 } 1647 StringRef SectionName; 1648 error(Section->getName(SectionName)); 1649 outs() << SectionName; 1650 } 1651 1652 outs() << '\t'; 1653 if (Common || isa<ELFObjectFileBase>(O)) { 1654 uint64_t Val = 1655 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1656 outs() << format("\t %08" PRIx64 " ", Val); 1657 } 1658 1659 if (Hidden) 1660 outs() << ".hidden "; 1661 1662 if (Demangle) 1663 outs() << demangle(Name) << '\n'; 1664 else 1665 outs() << Name << '\n'; 1666 } 1667 } 1668 1669 static void printUnwindInfo(const ObjectFile *O) { 1670 outs() << "Unwind info:\n\n"; 1671 1672 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1673 printCOFFUnwindInfo(Coff); 1674 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1675 printMachOUnwindInfo(MachO); 1676 else 1677 // TODO: Extract DWARF dump tool to objdump. 1678 WithColor::error(errs(), ToolName) 1679 << "This operation is only currently supported " 1680 "for COFF and MachO object files.\n"; 1681 } 1682 1683 void llvm::printExportsTrie(const ObjectFile *o) { 1684 outs() << "Exports trie:\n"; 1685 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1686 printMachOExportsTrie(MachO); 1687 else 1688 WithColor::error(errs(), ToolName) 1689 << "This operation is only currently supported " 1690 "for Mach-O executable files.\n"; 1691 } 1692 1693 void llvm::printRebaseTable(ObjectFile *o) { 1694 outs() << "Rebase table:\n"; 1695 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1696 printMachORebaseTable(MachO); 1697 else 1698 WithColor::error(errs(), ToolName) 1699 << "This operation is only currently supported " 1700 "for Mach-O executable files.\n"; 1701 } 1702 1703 void llvm::printBindTable(ObjectFile *o) { 1704 outs() << "Bind table:\n"; 1705 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1706 printMachOBindTable(MachO); 1707 else 1708 WithColor::error(errs(), ToolName) 1709 << "This operation is only currently supported " 1710 "for Mach-O executable files.\n"; 1711 } 1712 1713 void llvm::printLazyBindTable(ObjectFile *o) { 1714 outs() << "Lazy bind table:\n"; 1715 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1716 printMachOLazyBindTable(MachO); 1717 else 1718 WithColor::error(errs(), ToolName) 1719 << "This operation is only currently supported " 1720 "for Mach-O executable files.\n"; 1721 } 1722 1723 void llvm::printWeakBindTable(ObjectFile *o) { 1724 outs() << "Weak bind table:\n"; 1725 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1726 printMachOWeakBindTable(MachO); 1727 else 1728 WithColor::error(errs(), ToolName) 1729 << "This operation is only currently supported " 1730 "for Mach-O executable files.\n"; 1731 } 1732 1733 /// Dump the raw contents of the __clangast section so the output can be piped 1734 /// into llvm-bcanalyzer. 1735 void llvm::printRawClangAST(const ObjectFile *Obj) { 1736 if (outs().is_displayed()) { 1737 WithColor::error(errs(), ToolName) 1738 << "The -raw-clang-ast option will dump the raw binary contents of " 1739 "the clang ast section.\n" 1740 "Please redirect the output to a file or another program such as " 1741 "llvm-bcanalyzer.\n"; 1742 return; 1743 } 1744 1745 StringRef ClangASTSectionName("__clangast"); 1746 if (isa<COFFObjectFile>(Obj)) { 1747 ClangASTSectionName = "clangast"; 1748 } 1749 1750 Optional<object::SectionRef> ClangASTSection; 1751 for (auto Sec : ToolSectionFilter(*Obj)) { 1752 StringRef Name; 1753 Sec.getName(Name); 1754 if (Name == ClangASTSectionName) { 1755 ClangASTSection = Sec; 1756 break; 1757 } 1758 } 1759 if (!ClangASTSection) 1760 return; 1761 1762 StringRef ClangASTContents; 1763 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1764 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1765 } 1766 1767 static void printFaultMaps(const ObjectFile *Obj) { 1768 StringRef FaultMapSectionName; 1769 1770 if (isa<ELFObjectFileBase>(Obj)) { 1771 FaultMapSectionName = ".llvm_faultmaps"; 1772 } else if (isa<MachOObjectFile>(Obj)) { 1773 FaultMapSectionName = "__llvm_faultmaps"; 1774 } else { 1775 WithColor::error(errs(), ToolName) 1776 << "This operation is only currently supported " 1777 "for ELF and Mach-O executable files.\n"; 1778 return; 1779 } 1780 1781 Optional<object::SectionRef> FaultMapSection; 1782 1783 for (auto Sec : ToolSectionFilter(*Obj)) { 1784 StringRef Name; 1785 Sec.getName(Name); 1786 if (Name == FaultMapSectionName) { 1787 FaultMapSection = Sec; 1788 break; 1789 } 1790 } 1791 1792 outs() << "FaultMap table:\n"; 1793 1794 if (!FaultMapSection.hasValue()) { 1795 outs() << "<not found>\n"; 1796 return; 1797 } 1798 1799 StringRef FaultMapContents; 1800 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1801 1802 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1803 FaultMapContents.bytes_end()); 1804 1805 outs() << FMP; 1806 } 1807 1808 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1809 if (O->isELF()) { 1810 printELFFileHeader(O); 1811 return printELFDynamicSection(O); 1812 } 1813 if (O->isCOFF()) 1814 return printCOFFFileHeader(O); 1815 if (O->isWasm()) 1816 return printWasmFileHeader(O); 1817 if (O->isMachO()) { 1818 printMachOFileHeader(O); 1819 if (!OnlyFirst) 1820 printMachOLoadCommands(O); 1821 return; 1822 } 1823 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1824 } 1825 1826 static void printFileHeaders(const ObjectFile *O) { 1827 if (!O->isELF() && !O->isCOFF()) 1828 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1829 1830 Triple::ArchType AT = O->getArch(); 1831 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1832 Expected<uint64_t> StartAddrOrErr = O->getStartAddress(); 1833 if (!StartAddrOrErr) 1834 report_error(O->getFileName(), StartAddrOrErr.takeError()); 1835 1836 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1837 uint64_t Address = StartAddrOrErr.get(); 1838 outs() << "start address: " 1839 << "0x" << format(Fmt.data(), Address) << "\n\n"; 1840 } 1841 1842 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 1843 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 1844 if (!ModeOrErr) { 1845 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 1846 consumeError(ModeOrErr.takeError()); 1847 return; 1848 } 1849 sys::fs::perms Mode = ModeOrErr.get(); 1850 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1851 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1852 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1853 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1854 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1855 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1856 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1857 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1858 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1859 1860 outs() << " "; 1861 1862 Expected<unsigned> UIDOrErr = C.getUID(); 1863 if (!UIDOrErr) 1864 report_error(Filename, UIDOrErr.takeError()); 1865 unsigned UID = UIDOrErr.get(); 1866 outs() << format("%d/", UID); 1867 1868 Expected<unsigned> GIDOrErr = C.getGID(); 1869 if (!GIDOrErr) 1870 report_error(Filename, GIDOrErr.takeError()); 1871 unsigned GID = GIDOrErr.get(); 1872 outs() << format("%-d ", GID); 1873 1874 Expected<uint64_t> Size = C.getRawSize(); 1875 if (!Size) 1876 report_error(Filename, Size.takeError()); 1877 outs() << format("%6" PRId64, Size.get()) << " "; 1878 1879 StringRef RawLastModified = C.getRawLastModified(); 1880 unsigned Seconds; 1881 if (RawLastModified.getAsInteger(10, Seconds)) 1882 outs() << "(date: \"" << RawLastModified 1883 << "\" contains non-decimal chars) "; 1884 else { 1885 // Since ctime(3) returns a 26 character string of the form: 1886 // "Sun Sep 16 01:03:52 1973\n\0" 1887 // just print 24 characters. 1888 time_t t = Seconds; 1889 outs() << format("%.24s ", ctime(&t)); 1890 } 1891 1892 StringRef Name = ""; 1893 Expected<StringRef> NameOrErr = C.getName(); 1894 if (!NameOrErr) { 1895 consumeError(NameOrErr.takeError()); 1896 Expected<StringRef> RawNameOrErr = C.getRawName(); 1897 if (!RawNameOrErr) 1898 report_error(Filename, NameOrErr.takeError()); 1899 Name = RawNameOrErr.get(); 1900 } else { 1901 Name = NameOrErr.get(); 1902 } 1903 outs() << Name << "\n"; 1904 } 1905 1906 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 1907 const Archive::Child *C = nullptr) { 1908 // Avoid other output when using a raw option. 1909 if (!RawClangAST) { 1910 outs() << '\n'; 1911 if (A) 1912 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 1913 else 1914 outs() << O->getFileName(); 1915 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 1916 } 1917 1918 StringRef ArchiveName = A ? A->getFileName() : ""; 1919 if (FileHeaders) 1920 printFileHeaders(O); 1921 if (ArchiveHeaders && !MachOOpt && C) 1922 printArchiveChild(ArchiveName, *C); 1923 if (Disassemble) 1924 disassembleObject(O, Relocations); 1925 if (Relocations && !Disassemble) 1926 printRelocations(O); 1927 if (DynamicRelocations) 1928 printDynamicRelocations(O); 1929 if (SectionHeaders) 1930 printSectionHeaders(O); 1931 if (SectionContents) 1932 printSectionContents(O); 1933 if (SymbolTable) 1934 printSymbolTable(O, ArchiveName); 1935 if (UnwindInfo) 1936 printUnwindInfo(O); 1937 if (PrivateHeaders || FirstPrivateHeader) 1938 printPrivateFileHeaders(O, FirstPrivateHeader); 1939 if (ExportsTrie) 1940 printExportsTrie(O); 1941 if (Rebase) 1942 printRebaseTable(O); 1943 if (Bind) 1944 printBindTable(O); 1945 if (LazyBind) 1946 printLazyBindTable(O); 1947 if (WeakBind) 1948 printWeakBindTable(O); 1949 if (RawClangAST) 1950 printRawClangAST(O); 1951 if (PrintFaultMaps) 1952 printFaultMaps(O); 1953 if (DwarfDumpType != DIDT_Null) { 1954 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 1955 // Dump the complete DWARF structure. 1956 DIDumpOptions DumpOpts; 1957 DumpOpts.DumpType = DwarfDumpType; 1958 DICtx->dump(outs(), DumpOpts); 1959 } 1960 } 1961 1962 static void dumpObject(const COFFImportFile *I, const Archive *A, 1963 const Archive::Child *C = nullptr) { 1964 StringRef ArchiveName = A ? A->getFileName() : ""; 1965 1966 // Avoid other output when using a raw option. 1967 if (!RawClangAST) 1968 outs() << '\n' 1969 << ArchiveName << "(" << I->getFileName() << ")" 1970 << ":\tfile format COFF-import-file" 1971 << "\n\n"; 1972 1973 if (ArchiveHeaders && !MachOOpt && C) 1974 printArchiveChild(ArchiveName, *C); 1975 if (SymbolTable) 1976 printCOFFSymbolTable(I); 1977 } 1978 1979 /// Dump each object file in \a a; 1980 static void dumpArchive(const Archive *A) { 1981 Error Err = Error::success(); 1982 for (auto &C : A->children(Err)) { 1983 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1984 if (!ChildOrErr) { 1985 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 1986 report_error(A->getFileName(), C, std::move(E)); 1987 continue; 1988 } 1989 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 1990 dumpObject(O, A, &C); 1991 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 1992 dumpObject(I, A, &C); 1993 else 1994 report_error(A->getFileName(), object_error::invalid_file_type); 1995 } 1996 if (Err) 1997 report_error(A->getFileName(), std::move(Err)); 1998 } 1999 2000 /// Open file and figure out how to dump it. 2001 static void dumpInput(StringRef file) { 2002 // If we are using the Mach-O specific object file parser, then let it parse 2003 // the file and process the command line options. So the -arch flags can 2004 // be used to select specific slices, etc. 2005 if (MachOOpt) { 2006 parseInputMachO(file); 2007 return; 2008 } 2009 2010 // Attempt to open the binary. 2011 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2012 if (!BinaryOrErr) 2013 report_error(file, BinaryOrErr.takeError()); 2014 Binary &Binary = *BinaryOrErr.get().getBinary(); 2015 2016 if (Archive *A = dyn_cast<Archive>(&Binary)) 2017 dumpArchive(A); 2018 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2019 dumpObject(O); 2020 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2021 parseInputMachO(UB); 2022 else 2023 report_error(file, object_error::invalid_file_type); 2024 } 2025 2026 int main(int argc, char **argv) { 2027 InitLLVM X(argc, argv); 2028 2029 // Initialize targets and assembly printers/parsers. 2030 llvm::InitializeAllTargetInfos(); 2031 llvm::InitializeAllTargetMCs(); 2032 llvm::InitializeAllDisassemblers(); 2033 2034 // Register the target printer for --version. 2035 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2036 2037 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2038 2039 ToolName = argv[0]; 2040 2041 // Defaults to a.out if no filenames specified. 2042 if (InputFilenames.empty()) 2043 InputFilenames.push_back("a.out"); 2044 2045 if (AllHeaders) 2046 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2047 SectionHeaders = SymbolTable = true; 2048 2049 if (DisassembleAll || PrintSource || PrintLines) 2050 Disassemble = true; 2051 2052 if (!Disassemble 2053 && !Relocations 2054 && !DynamicRelocations 2055 && !SectionHeaders 2056 && !SectionContents 2057 && !SymbolTable 2058 && !UnwindInfo 2059 && !PrivateHeaders 2060 && !FileHeaders 2061 && !FirstPrivateHeader 2062 && !ExportsTrie 2063 && !Rebase 2064 && !Bind 2065 && !LazyBind 2066 && !WeakBind 2067 && !RawClangAST 2068 && !(UniversalHeaders && MachOOpt) 2069 && !ArchiveHeaders 2070 && !(IndirectSymbols && MachOOpt) 2071 && !(DataInCode && MachOOpt) 2072 && !(LinkOptHints && MachOOpt) 2073 && !(InfoPlist && MachOOpt) 2074 && !(DylibsUsed && MachOOpt) 2075 && !(DylibId && MachOOpt) 2076 && !(ObjcMetaData && MachOOpt) 2077 && !(!FilterSections.empty() && MachOOpt) 2078 && !PrintFaultMaps 2079 && DwarfDumpType == DIDT_Null) { 2080 cl::PrintHelpMessage(); 2081 return 2; 2082 } 2083 2084 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2085 DisassembleFunctions.end()); 2086 2087 llvm::for_each(InputFilenames, dumpInput); 2088 2089 return EXIT_SUCCESS; 2090 } 2091