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 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 894 if (InlineRelocs) 895 RelocMap = getRelocsMap(*Obj); 896 897 // Create a mapping from virtual address to symbol name. This is used to 898 // pretty print the symbols while disassembling. 899 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 900 SectionSymbolsTy AbsoluteSymbols; 901 for (const SymbolRef &Symbol : Obj->symbols()) { 902 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 903 if (!AddressOrErr) 904 report_error(Obj->getFileName(), AddressOrErr.takeError()); 905 uint64_t Address = *AddressOrErr; 906 907 Expected<StringRef> Name = Symbol.getName(); 908 if (!Name) 909 report_error(Obj->getFileName(), Name.takeError()); 910 if (Name->empty()) 911 continue; 912 913 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 914 if (!SectionOrErr) 915 report_error(Obj->getFileName(), SectionOrErr.takeError()); 916 917 uint8_t SymbolType = ELF::STT_NOTYPE; 918 if (Obj->isELF()) 919 SymbolType = getElfSymbolType(Obj, Symbol); 920 921 section_iterator SecI = *SectionOrErr; 922 if (SecI != Obj->section_end()) 923 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 924 else 925 AbsoluteSymbols.emplace_back(Address, *Name, SymbolType); 926 927 928 } 929 if (AllSymbols.empty() && Obj->isELF()) 930 addDynamicElfSymbols(Obj, AllSymbols); 931 932 BumpPtrAllocator A; 933 StringSaver Saver(A); 934 addPltEntries(Obj, AllSymbols, Saver); 935 936 // Create a mapping from virtual address to section. 937 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 938 for (SectionRef Sec : Obj->sections()) 939 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 940 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 941 942 // Linked executables (.exe and .dll files) typically don't include a real 943 // symbol table but they might contain an export table. 944 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 945 for (const auto &ExportEntry : COFFObj->export_directories()) { 946 StringRef Name; 947 error(ExportEntry.getSymbolName(Name)); 948 if (Name.empty()) 949 continue; 950 uint32_t RVA; 951 error(ExportEntry.getExportRVA(RVA)); 952 953 uint64_t VA = COFFObj->getImageBase() + RVA; 954 auto Sec = std::upper_bound( 955 SectionAddresses.begin(), SectionAddresses.end(), VA, 956 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 957 return LHS < RHS.first; 958 }); 959 if (Sec != SectionAddresses.begin()) 960 --Sec; 961 else 962 Sec = SectionAddresses.end(); 963 964 if (Sec != SectionAddresses.end()) 965 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 966 else 967 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 968 } 969 } 970 971 // Sort all the symbols, this allows us to use a simple binary search to find 972 // a symbol near an address. 973 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 974 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 975 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 976 977 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 978 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 979 continue; 980 981 uint64_t SectionAddr = Section.getAddress(); 982 uint64_t SectSize = Section.getSize(); 983 if (!SectSize) 984 continue; 985 986 // Get the list of all the symbols in this section. 987 SectionSymbolsTy &Symbols = AllSymbols[Section]; 988 std::vector<uint64_t> DataMappingSymsAddr; 989 std::vector<uint64_t> TextMappingSymsAddr; 990 if (isArmElf(Obj)) { 991 for (const auto &Symb : Symbols) { 992 uint64_t Address = std::get<0>(Symb); 993 StringRef Name = std::get<1>(Symb); 994 if (Name.startswith("$d")) 995 DataMappingSymsAddr.push_back(Address - SectionAddr); 996 if (Name.startswith("$x")) 997 TextMappingSymsAddr.push_back(Address - SectionAddr); 998 if (Name.startswith("$a")) 999 TextMappingSymsAddr.push_back(Address - SectionAddr); 1000 if (Name.startswith("$t")) 1001 TextMappingSymsAddr.push_back(Address - SectionAddr); 1002 } 1003 } 1004 1005 llvm::sort(DataMappingSymsAddr); 1006 llvm::sort(TextMappingSymsAddr); 1007 1008 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1009 // AMDGPU disassembler uses symbolizer for printing labels 1010 std::unique_ptr<MCRelocationInfo> RelInfo( 1011 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1012 if (RelInfo) { 1013 std::unique_ptr<MCSymbolizer> Symbolizer( 1014 TheTarget->createMCSymbolizer( 1015 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1016 DisAsm->setSymbolizer(std::move(Symbolizer)); 1017 } 1018 } 1019 1020 StringRef SegmentName = ""; 1021 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1022 DataRefImpl DR = Section.getRawDataRefImpl(); 1023 SegmentName = MachO->getSectionFinalSegmentName(DR); 1024 } 1025 StringRef SectionName; 1026 error(Section.getName(SectionName)); 1027 1028 // If the section has no symbol at the start, just insert a dummy one. 1029 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1030 Symbols.insert( 1031 Symbols.begin(), 1032 std::make_tuple(SectionAddr, SectionName, 1033 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1034 } 1035 1036 SmallString<40> Comments; 1037 raw_svector_ostream CommentStream(Comments); 1038 1039 StringRef BytesStr; 1040 error(Section.getContents(BytesStr)); 1041 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1042 BytesStr.size()); 1043 1044 uint64_t Size; 1045 uint64_t Index; 1046 bool PrintedSection = false; 1047 1048 std::vector<RelocationRef> Rels = RelocMap[Section]; 1049 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1050 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1051 // Disassemble symbol by symbol. 1052 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1053 uint64_t Start = std::get<0>(Symbols[SI]) - SectionAddr; 1054 // The end is either the section end or the beginning of the next 1055 // symbol. 1056 uint64_t End = (SI == SE - 1) 1057 ? SectSize 1058 : std::get<0>(Symbols[SI + 1]) - SectionAddr; 1059 // Don't try to disassemble beyond the end of section contents. 1060 if (End > SectSize) 1061 End = SectSize; 1062 // If this symbol has the same address as the next symbol, then skip it. 1063 if (Start >= End) 1064 continue; 1065 1066 // Check if we need to skip symbol 1067 // Skip if the symbol's data is not between StartAddress and StopAddress 1068 if (End + SectionAddr < StartAddress || 1069 Start + SectionAddr > StopAddress) { 1070 continue; 1071 } 1072 1073 /// Skip if user requested specific symbols and this is not in the list 1074 if (!DisasmFuncsSet.empty() && 1075 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1076 continue; 1077 1078 if (!PrintedSection) { 1079 PrintedSection = true; 1080 outs() << "Disassembly of section "; 1081 if (!SegmentName.empty()) 1082 outs() << SegmentName << ","; 1083 outs() << SectionName << ':'; 1084 } 1085 1086 // Stop disassembly at the stop address specified 1087 if (End + SectionAddr > StopAddress) 1088 End = StopAddress - SectionAddr; 1089 1090 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1091 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1092 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1093 Start += 256; 1094 } 1095 if (SI == SE - 1 || 1096 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1097 // cut trailing zeroes at the end of kernel 1098 // cut up to 256 bytes 1099 const uint64_t EndAlign = 256; 1100 const auto Limit = End - (std::min)(EndAlign, End - Start); 1101 while (End > Limit && 1102 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1103 End -= 4; 1104 } 1105 } 1106 1107 outs() << '\n'; 1108 if (!NoLeadingAddr) 1109 outs() << format("%016" PRIx64 " ", SectionAddr + Start); 1110 1111 StringRef SymbolName = std::get<1>(Symbols[SI]); 1112 if (Demangle) 1113 outs() << demangle(SymbolName) << ":\n"; 1114 else 1115 outs() << SymbolName << ":\n"; 1116 1117 // Don't print raw contents of a virtual section. A virtual section 1118 // doesn't have any contents in the file. 1119 if (Section.isVirtual()) { 1120 outs() << "...\n"; 1121 continue; 1122 } 1123 1124 #ifndef NDEBUG 1125 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1126 #else 1127 raw_ostream &DebugOut = nulls(); 1128 #endif 1129 1130 // Some targets (like WebAssembly) have a special prelude at the start 1131 // of each symbol. 1132 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1133 SectionAddr + Start, DebugOut, CommentStream); 1134 Start += Size; 1135 1136 for (Index = Start; Index < End; Index += Size) { 1137 MCInst Inst; 1138 1139 if (Index + SectionAddr < StartAddress || 1140 Index + SectionAddr > StopAddress) { 1141 // skip byte by byte till StartAddress is reached 1142 Size = 1; 1143 continue; 1144 } 1145 // AArch64 ELF binaries can interleave data and text in the 1146 // same section. We rely on the markers introduced to 1147 // understand what we need to dump. If the data marker is within a 1148 // function, it is denoted as a word/short etc 1149 if (isArmElf(Obj) && std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1150 !DisassembleAll) { 1151 uint64_t Stride = 0; 1152 1153 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1154 DataMappingSymsAddr.end(), Index); 1155 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1156 // Switch to data. 1157 while (Index < End) { 1158 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1159 outs() << "\t"; 1160 if (Index + 4 <= End) { 1161 Stride = 4; 1162 dumpBytes(Bytes.slice(Index, 4), outs()); 1163 outs() << "\t.word\t"; 1164 uint32_t Data = 0; 1165 if (Obj->isLittleEndian()) { 1166 const auto Word = 1167 reinterpret_cast<const support::ulittle32_t *>( 1168 Bytes.data() + Index); 1169 Data = *Word; 1170 } else { 1171 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1172 Bytes.data() + Index); 1173 Data = *Word; 1174 } 1175 outs() << "0x" << format("%08" PRIx32, Data); 1176 } else if (Index + 2 <= End) { 1177 Stride = 2; 1178 dumpBytes(Bytes.slice(Index, 2), outs()); 1179 outs() << "\t\t.short\t"; 1180 uint16_t Data = 0; 1181 if (Obj->isLittleEndian()) { 1182 const auto Short = 1183 reinterpret_cast<const support::ulittle16_t *>( 1184 Bytes.data() + Index); 1185 Data = *Short; 1186 } else { 1187 const auto Short = 1188 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1189 Index); 1190 Data = *Short; 1191 } 1192 outs() << "0x" << format("%04" PRIx16, Data); 1193 } else { 1194 Stride = 1; 1195 dumpBytes(Bytes.slice(Index, 1), outs()); 1196 outs() << "\t\t.byte\t"; 1197 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1198 } 1199 Index += Stride; 1200 outs() << "\n"; 1201 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1202 TextMappingSymsAddr.end(), Index); 1203 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1204 break; 1205 } 1206 } 1207 } 1208 1209 // If there is a data symbol inside an ELF text section and we are only 1210 // disassembling text (applicable all architectures), 1211 // we are in a situation where we must print the data and not 1212 // disassemble it. 1213 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1214 !DisassembleAll && Section.isText()) { 1215 // print out data up to 8 bytes at a time in hex and ascii 1216 uint8_t AsciiData[9] = {'\0'}; 1217 uint8_t Byte; 1218 int NumBytes = 0; 1219 1220 for (Index = Start; Index < End; Index += 1) { 1221 if (((SectionAddr + Index) < StartAddress) || 1222 ((SectionAddr + Index) > StopAddress)) 1223 continue; 1224 if (NumBytes == 0) { 1225 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1226 outs() << "\t"; 1227 } 1228 Byte = Bytes.slice(Index)[0]; 1229 outs() << format(" %02x", Byte); 1230 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1231 1232 uint8_t IndentOffset = 0; 1233 NumBytes++; 1234 if (Index == End - 1 || NumBytes > 8) { 1235 // Indent the space for less than 8 bytes data. 1236 // 2 spaces for byte and one for space between bytes 1237 IndentOffset = 3 * (8 - NumBytes); 1238 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1239 AsciiData[Excess] = '\0'; 1240 NumBytes = 8; 1241 } 1242 if (NumBytes == 8) { 1243 AsciiData[8] = '\0'; 1244 outs() << std::string(IndentOffset, ' ') << " "; 1245 outs() << reinterpret_cast<char *>(AsciiData); 1246 outs() << '\n'; 1247 NumBytes = 0; 1248 } 1249 } 1250 } 1251 if (Index >= End) 1252 break; 1253 1254 if (size_t N = 1255 countSkippableZeroBytes(Bytes.slice(Index, End - Index))) { 1256 outs() << "\t\t..." << '\n'; 1257 Index += N; 1258 if (Index >= End) 1259 break; 1260 } 1261 1262 // Disassemble a real instruction or a data when disassemble all is 1263 // provided 1264 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1265 SectionAddr + Index, DebugOut, 1266 CommentStream); 1267 if (Size == 0) 1268 Size = 1; 1269 1270 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1271 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1272 *STI, &SP, &Rels); 1273 outs() << CommentStream.str(); 1274 Comments.clear(); 1275 1276 // Try to resolve the target of a call, tail call, etc. to a specific 1277 // symbol. 1278 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1279 MIA->isConditionalBranch(Inst))) { 1280 uint64_t Target; 1281 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1282 // In a relocatable object, the target's section must reside in 1283 // the same section as the call instruction or it is accessed 1284 // through a relocation. 1285 // 1286 // In a non-relocatable object, the target may be in any section. 1287 // 1288 // N.B. We don't walk the relocations in the relocatable case yet. 1289 auto *TargetSectionSymbols = &Symbols; 1290 if (!Obj->isRelocatableObject()) { 1291 auto SectionAddress = std::upper_bound( 1292 SectionAddresses.begin(), SectionAddresses.end(), Target, 1293 [](uint64_t LHS, 1294 const std::pair<uint64_t, SectionRef> &RHS) { 1295 return LHS < RHS.first; 1296 }); 1297 if (SectionAddress != SectionAddresses.begin()) { 1298 --SectionAddress; 1299 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1300 } else { 1301 TargetSectionSymbols = &AbsoluteSymbols; 1302 } 1303 } 1304 1305 // Find the first symbol in the section whose offset is less than 1306 // or equal to the target. If there isn't a section that contains 1307 // the target, find the nearest preceding absolute symbol. 1308 auto TargetSym = std::upper_bound( 1309 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1310 Target, [](uint64_t LHS, 1311 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1312 return LHS < std::get<0>(RHS); 1313 }); 1314 if (TargetSym == TargetSectionSymbols->begin()) { 1315 TargetSectionSymbols = &AbsoluteSymbols; 1316 TargetSym = std::upper_bound( 1317 AbsoluteSymbols.begin(), AbsoluteSymbols.end(), 1318 Target, [](uint64_t LHS, 1319 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1320 return LHS < std::get<0>(RHS); 1321 }); 1322 } 1323 if (TargetSym != TargetSectionSymbols->begin()) { 1324 --TargetSym; 1325 uint64_t TargetAddress = std::get<0>(*TargetSym); 1326 StringRef TargetName = std::get<1>(*TargetSym); 1327 outs() << " <" << TargetName; 1328 uint64_t Disp = Target - TargetAddress; 1329 if (Disp) 1330 outs() << "+0x" << Twine::utohexstr(Disp); 1331 outs() << '>'; 1332 } 1333 } 1334 } 1335 outs() << "\n"; 1336 1337 // Hexagon does this in pretty printer 1338 if (Obj->getArch() != Triple::hexagon) { 1339 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " 1340 : "\t\t\t%08" PRIx64 1341 ": "; 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 1369 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1370 if (StartAddress > StopAddress) 1371 error("Start address should be less than stop address"); 1372 1373 const Target *TheTarget = getTarget(Obj); 1374 1375 // Package up features to be passed to target/subtarget 1376 SubtargetFeatures Features = Obj->getFeatures(); 1377 if (!MAttrs.empty()) 1378 for (unsigned I = 0; I != MAttrs.size(); ++I) 1379 Features.AddFeature(MAttrs[I]); 1380 1381 std::unique_ptr<const MCRegisterInfo> MRI( 1382 TheTarget->createMCRegInfo(TripleName)); 1383 if (!MRI) 1384 report_error(Obj->getFileName(), 1385 "no register info for target " + TripleName); 1386 1387 // Set up disassembler. 1388 std::unique_ptr<const MCAsmInfo> AsmInfo( 1389 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1390 if (!AsmInfo) 1391 report_error(Obj->getFileName(), 1392 "no assembly info for target " + TripleName); 1393 std::unique_ptr<const MCSubtargetInfo> STI( 1394 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1395 if (!STI) 1396 report_error(Obj->getFileName(), 1397 "no subtarget info for target " + TripleName); 1398 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1399 if (!MII) 1400 report_error(Obj->getFileName(), 1401 "no instruction info for target " + TripleName); 1402 MCObjectFileInfo MOFI; 1403 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1404 // FIXME: for now initialize MCObjectFileInfo with default values 1405 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1406 1407 std::unique_ptr<MCDisassembler> DisAsm( 1408 TheTarget->createMCDisassembler(*STI, Ctx)); 1409 if (!DisAsm) 1410 report_error(Obj->getFileName(), 1411 "no disassembler for target " + TripleName); 1412 1413 std::unique_ptr<const MCInstrAnalysis> MIA( 1414 TheTarget->createMCInstrAnalysis(MII.get())); 1415 1416 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1417 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1418 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1419 if (!IP) 1420 report_error(Obj->getFileName(), 1421 "no instruction printer for target " + TripleName); 1422 IP->setPrintImmHex(PrintImmHex); 1423 1424 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1425 SourcePrinter SP(Obj, TheTarget->getName()); 1426 1427 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), MIA.get(), IP.get(), 1428 STI.get(), PIP, SP, InlineRelocs); 1429 } 1430 1431 void llvm::printRelocations(const ObjectFile *Obj) { 1432 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1433 "%08" PRIx64; 1434 // Regular objdump doesn't print relocations in non-relocatable object 1435 // files. 1436 if (!Obj->isRelocatableObject()) 1437 return; 1438 1439 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1440 if (Section.relocation_begin() == Section.relocation_end()) 1441 continue; 1442 StringRef SecName; 1443 error(Section.getName(SecName)); 1444 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1445 for (const RelocationRef &Reloc : Section.relocations()) { 1446 uint64_t Address = Reloc.getOffset(); 1447 SmallString<32> RelocName; 1448 SmallString<32> ValueStr; 1449 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1450 continue; 1451 Reloc.getTypeName(RelocName); 1452 error(getRelocationValueString(Reloc, ValueStr)); 1453 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1454 << ValueStr << "\n"; 1455 } 1456 outs() << "\n"; 1457 } 1458 } 1459 1460 void llvm::printDynamicRelocations(const ObjectFile *Obj) { 1461 // For the moment, this option is for ELF only 1462 if (!Obj->isELF()) 1463 return; 1464 1465 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1466 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1467 error("not a dynamic object"); 1468 return; 1469 } 1470 1471 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1472 if (DynRelSec.empty()) 1473 return; 1474 1475 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1476 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1477 for (const SectionRef &Section : DynRelSec) { 1478 if (Section.relocation_begin() == Section.relocation_end()) 1479 continue; 1480 for (const RelocationRef &Reloc : Section.relocations()) { 1481 uint64_t Address = Reloc.getOffset(); 1482 SmallString<32> RelocName; 1483 SmallString<32> ValueStr; 1484 Reloc.getTypeName(RelocName); 1485 error(getRelocationValueString(Reloc, ValueStr)); 1486 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1487 << ValueStr << "\n"; 1488 } 1489 } 1490 } 1491 1492 void llvm::printSectionHeaders(const ObjectFile *Obj) { 1493 outs() << "Sections:\n" 1494 "Idx Name Size Address Type\n"; 1495 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1496 StringRef Name; 1497 error(Section.getName(Name)); 1498 uint64_t Address = Section.getAddress(); 1499 uint64_t Size = Section.getSize(); 1500 bool Text = Section.isText(); 1501 bool Data = Section.isData(); 1502 bool BSS = Section.isBSS(); 1503 std::string Type = (std::string(Text ? "TEXT " : "") + 1504 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1505 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1506 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1507 Address, Type.c_str()); 1508 } 1509 outs() << "\n"; 1510 } 1511 1512 void llvm::printSectionContents(const ObjectFile *Obj) { 1513 std::error_code EC; 1514 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1515 StringRef Name; 1516 StringRef Contents; 1517 error(Section.getName(Name)); 1518 uint64_t BaseAddr = Section.getAddress(); 1519 uint64_t Size = Section.getSize(); 1520 if (!Size) 1521 continue; 1522 1523 outs() << "Contents of section " << Name << ":\n"; 1524 if (Section.isBSS()) { 1525 outs() << format("<skipping contents of bss section at [%04" PRIx64 1526 ", %04" PRIx64 ")>\n", 1527 BaseAddr, BaseAddr + Size); 1528 continue; 1529 } 1530 1531 error(Section.getContents(Contents)); 1532 1533 // Dump out the content as hex and printable ascii characters. 1534 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1535 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1536 // Dump line of hex. 1537 for (std::size_t I = 0; I < 16; ++I) { 1538 if (I != 0 && I % 4 == 0) 1539 outs() << ' '; 1540 if (Addr + I < End) 1541 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1542 << hexdigit(Contents[Addr + I] & 0xF, true); 1543 else 1544 outs() << " "; 1545 } 1546 // Print ascii. 1547 outs() << " "; 1548 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1549 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1550 outs() << Contents[Addr + I]; 1551 else 1552 outs() << "."; 1553 } 1554 outs() << "\n"; 1555 } 1556 } 1557 } 1558 1559 void llvm::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1560 StringRef ArchitectureName) { 1561 outs() << "SYMBOL TABLE:\n"; 1562 1563 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1564 printCOFFSymbolTable(Coff); 1565 return; 1566 } 1567 1568 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1569 // Skip printing the special zero symbol when dumping an ELF file. 1570 // This makes the output consistent with the GNU objdump. 1571 if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O)) 1572 continue; 1573 1574 const SymbolRef &Symbol = *I; 1575 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1576 if (!AddressOrError) 1577 report_error(ArchiveName, O->getFileName(), AddressOrError.takeError(), 1578 ArchitectureName); 1579 uint64_t Address = *AddressOrError; 1580 if ((Address < StartAddress) || (Address > StopAddress)) 1581 continue; 1582 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1583 if (!TypeOrError) 1584 report_error(ArchiveName, O->getFileName(), TypeOrError.takeError(), 1585 ArchitectureName); 1586 SymbolRef::Type Type = *TypeOrError; 1587 uint32_t Flags = Symbol.getFlags(); 1588 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1589 if (!SectionOrErr) 1590 report_error(ArchiveName, O->getFileName(), SectionOrErr.takeError(), 1591 ArchitectureName); 1592 section_iterator Section = *SectionOrErr; 1593 StringRef Name; 1594 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 1595 Section->getName(Name); 1596 } else { 1597 Expected<StringRef> NameOrErr = Symbol.getName(); 1598 if (!NameOrErr) 1599 report_error(ArchiveName, O->getFileName(), NameOrErr.takeError(), 1600 ArchitectureName); 1601 Name = *NameOrErr; 1602 } 1603 1604 bool Global = Flags & SymbolRef::SF_Global; 1605 bool Weak = Flags & SymbolRef::SF_Weak; 1606 bool Absolute = Flags & SymbolRef::SF_Absolute; 1607 bool Common = Flags & SymbolRef::SF_Common; 1608 bool Hidden = Flags & SymbolRef::SF_Hidden; 1609 1610 char GlobLoc = ' '; 1611 if (Type != SymbolRef::ST_Unknown) 1612 GlobLoc = Global ? 'g' : 'l'; 1613 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1614 ? 'd' : ' '; 1615 char FileFunc = ' '; 1616 if (Type == SymbolRef::ST_File) 1617 FileFunc = 'f'; 1618 else if (Type == SymbolRef::ST_Function) 1619 FileFunc = 'F'; 1620 else if (Type == SymbolRef::ST_Data) 1621 FileFunc = 'O'; 1622 1623 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1624 "%08" PRIx64; 1625 1626 outs() << format(Fmt, Address) << " " 1627 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1628 << (Weak ? 'w' : ' ') // Weak? 1629 << ' ' // Constructor. Not supported yet. 1630 << ' ' // Warning. Not supported yet. 1631 << ' ' // Indirect reference to another symbol. 1632 << Debug // Debugging (d) or dynamic (D) symbol. 1633 << FileFunc // Name of function (F), file (f) or object (O). 1634 << ' '; 1635 if (Absolute) { 1636 outs() << "*ABS*"; 1637 } else if (Common) { 1638 outs() << "*COM*"; 1639 } else if (Section == O->section_end()) { 1640 outs() << "*UND*"; 1641 } else { 1642 if (const MachOObjectFile *MachO = 1643 dyn_cast<const MachOObjectFile>(O)) { 1644 DataRefImpl DR = Section->getRawDataRefImpl(); 1645 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1646 outs() << SegmentName << ","; 1647 } 1648 StringRef SectionName; 1649 error(Section->getName(SectionName)); 1650 outs() << SectionName; 1651 } 1652 1653 outs() << '\t'; 1654 if (Common || isa<ELFObjectFileBase>(O)) { 1655 uint64_t Val = 1656 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1657 outs() << format("\t %08" PRIx64 " ", Val); 1658 } 1659 1660 if (Hidden) 1661 outs() << ".hidden "; 1662 1663 if (Demangle) 1664 outs() << demangle(Name) << '\n'; 1665 else 1666 outs() << Name << '\n'; 1667 } 1668 } 1669 1670 static void printUnwindInfo(const ObjectFile *O) { 1671 outs() << "Unwind info:\n\n"; 1672 1673 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1674 printCOFFUnwindInfo(Coff); 1675 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1676 printMachOUnwindInfo(MachO); 1677 else 1678 // TODO: Extract DWARF dump tool to objdump. 1679 WithColor::error(errs(), ToolName) 1680 << "This operation is only currently supported " 1681 "for COFF and MachO object files.\n"; 1682 } 1683 1684 void llvm::printExportsTrie(const ObjectFile *o) { 1685 outs() << "Exports trie:\n"; 1686 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1687 printMachOExportsTrie(MachO); 1688 else 1689 WithColor::error(errs(), ToolName) 1690 << "This operation is only currently supported " 1691 "for Mach-O executable files.\n"; 1692 } 1693 1694 void llvm::printRebaseTable(ObjectFile *o) { 1695 outs() << "Rebase table:\n"; 1696 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1697 printMachORebaseTable(MachO); 1698 else 1699 WithColor::error(errs(), ToolName) 1700 << "This operation is only currently supported " 1701 "for Mach-O executable files.\n"; 1702 } 1703 1704 void llvm::printBindTable(ObjectFile *o) { 1705 outs() << "Bind table:\n"; 1706 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1707 printMachOBindTable(MachO); 1708 else 1709 WithColor::error(errs(), ToolName) 1710 << "This operation is only currently supported " 1711 "for Mach-O executable files.\n"; 1712 } 1713 1714 void llvm::printLazyBindTable(ObjectFile *o) { 1715 outs() << "Lazy bind table:\n"; 1716 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1717 printMachOLazyBindTable(MachO); 1718 else 1719 WithColor::error(errs(), ToolName) 1720 << "This operation is only currently supported " 1721 "for Mach-O executable files.\n"; 1722 } 1723 1724 void llvm::printWeakBindTable(ObjectFile *o) { 1725 outs() << "Weak bind table:\n"; 1726 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1727 printMachOWeakBindTable(MachO); 1728 else 1729 WithColor::error(errs(), ToolName) 1730 << "This operation is only currently supported " 1731 "for Mach-O executable files.\n"; 1732 } 1733 1734 /// Dump the raw contents of the __clangast section so the output can be piped 1735 /// into llvm-bcanalyzer. 1736 void llvm::printRawClangAST(const ObjectFile *Obj) { 1737 if (outs().is_displayed()) { 1738 WithColor::error(errs(), ToolName) 1739 << "The -raw-clang-ast option will dump the raw binary contents of " 1740 "the clang ast section.\n" 1741 "Please redirect the output to a file or another program such as " 1742 "llvm-bcanalyzer.\n"; 1743 return; 1744 } 1745 1746 StringRef ClangASTSectionName("__clangast"); 1747 if (isa<COFFObjectFile>(Obj)) { 1748 ClangASTSectionName = "clangast"; 1749 } 1750 1751 Optional<object::SectionRef> ClangASTSection; 1752 for (auto Sec : ToolSectionFilter(*Obj)) { 1753 StringRef Name; 1754 Sec.getName(Name); 1755 if (Name == ClangASTSectionName) { 1756 ClangASTSection = Sec; 1757 break; 1758 } 1759 } 1760 if (!ClangASTSection) 1761 return; 1762 1763 StringRef ClangASTContents; 1764 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1765 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1766 } 1767 1768 static void printFaultMaps(const ObjectFile *Obj) { 1769 StringRef FaultMapSectionName; 1770 1771 if (isa<ELFObjectFileBase>(Obj)) { 1772 FaultMapSectionName = ".llvm_faultmaps"; 1773 } else if (isa<MachOObjectFile>(Obj)) { 1774 FaultMapSectionName = "__llvm_faultmaps"; 1775 } else { 1776 WithColor::error(errs(), ToolName) 1777 << "This operation is only currently supported " 1778 "for ELF and Mach-O executable files.\n"; 1779 return; 1780 } 1781 1782 Optional<object::SectionRef> FaultMapSection; 1783 1784 for (auto Sec : ToolSectionFilter(*Obj)) { 1785 StringRef Name; 1786 Sec.getName(Name); 1787 if (Name == FaultMapSectionName) { 1788 FaultMapSection = Sec; 1789 break; 1790 } 1791 } 1792 1793 outs() << "FaultMap table:\n"; 1794 1795 if (!FaultMapSection.hasValue()) { 1796 outs() << "<not found>\n"; 1797 return; 1798 } 1799 1800 StringRef FaultMapContents; 1801 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1802 1803 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1804 FaultMapContents.bytes_end()); 1805 1806 outs() << FMP; 1807 } 1808 1809 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1810 if (O->isELF()) { 1811 printELFFileHeader(O); 1812 return printELFDynamicSection(O); 1813 } 1814 if (O->isCOFF()) 1815 return printCOFFFileHeader(O); 1816 if (O->isWasm()) 1817 return printWasmFileHeader(O); 1818 if (O->isMachO()) { 1819 printMachOFileHeader(O); 1820 if (!OnlyFirst) 1821 printMachOLoadCommands(O); 1822 return; 1823 } 1824 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1825 } 1826 1827 static void printFileHeaders(const ObjectFile *O) { 1828 if (!O->isELF() && !O->isCOFF()) 1829 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1830 1831 Triple::ArchType AT = O->getArch(); 1832 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1833 Expected<uint64_t> StartAddrOrErr = O->getStartAddress(); 1834 if (!StartAddrOrErr) 1835 report_error(O->getFileName(), StartAddrOrErr.takeError()); 1836 1837 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1838 uint64_t Address = StartAddrOrErr.get(); 1839 outs() << "start address: " 1840 << "0x" << format(Fmt.data(), Address) << "\n\n"; 1841 } 1842 1843 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 1844 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 1845 if (!ModeOrErr) { 1846 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 1847 consumeError(ModeOrErr.takeError()); 1848 return; 1849 } 1850 sys::fs::perms Mode = ModeOrErr.get(); 1851 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1852 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1853 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1854 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1855 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1856 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1857 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1858 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1859 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1860 1861 outs() << " "; 1862 1863 Expected<unsigned> UIDOrErr = C.getUID(); 1864 if (!UIDOrErr) 1865 report_error(Filename, UIDOrErr.takeError()); 1866 unsigned UID = UIDOrErr.get(); 1867 outs() << format("%d/", UID); 1868 1869 Expected<unsigned> GIDOrErr = C.getGID(); 1870 if (!GIDOrErr) 1871 report_error(Filename, GIDOrErr.takeError()); 1872 unsigned GID = GIDOrErr.get(); 1873 outs() << format("%-d ", GID); 1874 1875 Expected<uint64_t> Size = C.getRawSize(); 1876 if (!Size) 1877 report_error(Filename, Size.takeError()); 1878 outs() << format("%6" PRId64, Size.get()) << " "; 1879 1880 StringRef RawLastModified = C.getRawLastModified(); 1881 unsigned Seconds; 1882 if (RawLastModified.getAsInteger(10, Seconds)) 1883 outs() << "(date: \"" << RawLastModified 1884 << "\" contains non-decimal chars) "; 1885 else { 1886 // Since ctime(3) returns a 26 character string of the form: 1887 // "Sun Sep 16 01:03:52 1973\n\0" 1888 // just print 24 characters. 1889 time_t t = Seconds; 1890 outs() << format("%.24s ", ctime(&t)); 1891 } 1892 1893 StringRef Name = ""; 1894 Expected<StringRef> NameOrErr = C.getName(); 1895 if (!NameOrErr) { 1896 consumeError(NameOrErr.takeError()); 1897 Expected<StringRef> RawNameOrErr = C.getRawName(); 1898 if (!RawNameOrErr) 1899 report_error(Filename, NameOrErr.takeError()); 1900 Name = RawNameOrErr.get(); 1901 } else { 1902 Name = NameOrErr.get(); 1903 } 1904 outs() << Name << "\n"; 1905 } 1906 1907 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 1908 const Archive::Child *C = nullptr) { 1909 // Avoid other output when using a raw option. 1910 if (!RawClangAST) { 1911 outs() << '\n'; 1912 if (A) 1913 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 1914 else 1915 outs() << O->getFileName(); 1916 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 1917 } 1918 1919 StringRef ArchiveName = A ? A->getFileName() : ""; 1920 if (FileHeaders) 1921 printFileHeaders(O); 1922 if (ArchiveHeaders && !MachOOpt && C) 1923 printArchiveChild(ArchiveName, *C); 1924 if (Disassemble) 1925 disassembleObject(O, Relocations); 1926 if (Relocations && !Disassemble) 1927 printRelocations(O); 1928 if (DynamicRelocations) 1929 printDynamicRelocations(O); 1930 if (SectionHeaders) 1931 printSectionHeaders(O); 1932 if (SectionContents) 1933 printSectionContents(O); 1934 if (SymbolTable) 1935 printSymbolTable(O, ArchiveName); 1936 if (UnwindInfo) 1937 printUnwindInfo(O); 1938 if (PrivateHeaders || FirstPrivateHeader) 1939 printPrivateFileHeaders(O, FirstPrivateHeader); 1940 if (ExportsTrie) 1941 printExportsTrie(O); 1942 if (Rebase) 1943 printRebaseTable(O); 1944 if (Bind) 1945 printBindTable(O); 1946 if (LazyBind) 1947 printLazyBindTable(O); 1948 if (WeakBind) 1949 printWeakBindTable(O); 1950 if (RawClangAST) 1951 printRawClangAST(O); 1952 if (PrintFaultMaps) 1953 printFaultMaps(O); 1954 if (DwarfDumpType != DIDT_Null) { 1955 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 1956 // Dump the complete DWARF structure. 1957 DIDumpOptions DumpOpts; 1958 DumpOpts.DumpType = DwarfDumpType; 1959 DICtx->dump(outs(), DumpOpts); 1960 } 1961 } 1962 1963 static void dumpObject(const COFFImportFile *I, const Archive *A, 1964 const Archive::Child *C = nullptr) { 1965 StringRef ArchiveName = A ? A->getFileName() : ""; 1966 1967 // Avoid other output when using a raw option. 1968 if (!RawClangAST) 1969 outs() << '\n' 1970 << ArchiveName << "(" << I->getFileName() << ")" 1971 << ":\tfile format COFF-import-file" 1972 << "\n\n"; 1973 1974 if (ArchiveHeaders && !MachOOpt && C) 1975 printArchiveChild(ArchiveName, *C); 1976 if (SymbolTable) 1977 printCOFFSymbolTable(I); 1978 } 1979 1980 /// Dump each object file in \a a; 1981 static void dumpArchive(const Archive *A) { 1982 Error Err = Error::success(); 1983 for (auto &C : A->children(Err)) { 1984 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1985 if (!ChildOrErr) { 1986 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 1987 report_error(A->getFileName(), C, std::move(E)); 1988 continue; 1989 } 1990 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 1991 dumpObject(O, A, &C); 1992 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 1993 dumpObject(I, A, &C); 1994 else 1995 report_error(A->getFileName(), object_error::invalid_file_type); 1996 } 1997 if (Err) 1998 report_error(A->getFileName(), std::move(Err)); 1999 } 2000 2001 /// Open file and figure out how to dump it. 2002 static void dumpInput(StringRef file) { 2003 // If we are using the Mach-O specific object file parser, then let it parse 2004 // the file and process the command line options. So the -arch flags can 2005 // be used to select specific slices, etc. 2006 if (MachOOpt) { 2007 parseInputMachO(file); 2008 return; 2009 } 2010 2011 // Attempt to open the binary. 2012 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2013 if (!BinaryOrErr) 2014 report_error(file, BinaryOrErr.takeError()); 2015 Binary &Binary = *BinaryOrErr.get().getBinary(); 2016 2017 if (Archive *A = dyn_cast<Archive>(&Binary)) 2018 dumpArchive(A); 2019 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2020 dumpObject(O); 2021 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2022 parseInputMachO(UB); 2023 else 2024 report_error(file, object_error::invalid_file_type); 2025 } 2026 2027 int main(int argc, char **argv) { 2028 InitLLVM X(argc, argv); 2029 2030 // Initialize targets and assembly printers/parsers. 2031 llvm::InitializeAllTargetInfos(); 2032 llvm::InitializeAllTargetMCs(); 2033 llvm::InitializeAllDisassemblers(); 2034 2035 // Register the target printer for --version. 2036 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2037 2038 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2039 2040 ToolName = argv[0]; 2041 2042 // Defaults to a.out if no filenames specified. 2043 if (InputFilenames.empty()) 2044 InputFilenames.push_back("a.out"); 2045 2046 if (AllHeaders) 2047 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2048 SectionHeaders = SymbolTable = true; 2049 2050 if (DisassembleAll || PrintSource || PrintLines) 2051 Disassemble = true; 2052 2053 if (!Disassemble 2054 && !Relocations 2055 && !DynamicRelocations 2056 && !SectionHeaders 2057 && !SectionContents 2058 && !SymbolTable 2059 && !UnwindInfo 2060 && !PrivateHeaders 2061 && !FileHeaders 2062 && !FirstPrivateHeader 2063 && !ExportsTrie 2064 && !Rebase 2065 && !Bind 2066 && !LazyBind 2067 && !WeakBind 2068 && !RawClangAST 2069 && !(UniversalHeaders && MachOOpt) 2070 && !ArchiveHeaders 2071 && !(IndirectSymbols && MachOOpt) 2072 && !(DataInCode && MachOOpt) 2073 && !(LinkOptHints && MachOOpt) 2074 && !(InfoPlist && MachOOpt) 2075 && !(DylibsUsed && MachOOpt) 2076 && !(DylibId && MachOOpt) 2077 && !(ObjcMetaData && MachOOpt) 2078 && !(!FilterSections.empty() && MachOOpt) 2079 && !PrintFaultMaps 2080 && DwarfDumpType == DIDT_Null) { 2081 cl::PrintHelpMessage(); 2082 return 2; 2083 } 2084 2085 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2086 DisassembleFunctions.end()); 2087 2088 llvm::for_each(InputFilenames, dumpInput); 2089 2090 return EXIT_SUCCESS; 2091 } 2092