1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // This program is a utility that works like binutils "objdump", that is, it 11 // dumps out a plethora of information about an object file depending on the 12 // flags. 13 // 14 // The flags and output of this program should be near identical to those of 15 // binutils objdump. 16 // 17 //===----------------------------------------------------------------------===// 18 19 #include "llvm-objdump.h" 20 #include "llvm/ADT/Optional.h" 21 #include "llvm/ADT/STLExtras.h" 22 #include "llvm/ADT/StringExtras.h" 23 #include "llvm/ADT/StringSet.h" 24 #include "llvm/ADT/Triple.h" 25 #include "llvm/CodeGen/FaultMaps.h" 26 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 27 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 28 #include "llvm/Demangle/Demangle.h" 29 #include "llvm/MC/MCAsmInfo.h" 30 #include "llvm/MC/MCContext.h" 31 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 32 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 33 #include "llvm/MC/MCInst.h" 34 #include "llvm/MC/MCInstPrinter.h" 35 #include "llvm/MC/MCInstrAnalysis.h" 36 #include "llvm/MC/MCInstrInfo.h" 37 #include "llvm/MC/MCObjectFileInfo.h" 38 #include "llvm/MC/MCRegisterInfo.h" 39 #include "llvm/MC/MCSubtargetInfo.h" 40 #include "llvm/Object/Archive.h" 41 #include "llvm/Object/COFF.h" 42 #include "llvm/Object/COFFImportFile.h" 43 #include "llvm/Object/ELFObjectFile.h" 44 #include "llvm/Object/MachO.h" 45 #include "llvm/Object/MachOUniversal.h" 46 #include "llvm/Object/ObjectFile.h" 47 #include "llvm/Object/Wasm.h" 48 #include "llvm/Support/Casting.h" 49 #include "llvm/Support/CommandLine.h" 50 #include "llvm/Support/Debug.h" 51 #include "llvm/Support/Errc.h" 52 #include "llvm/Support/FileSystem.h" 53 #include "llvm/Support/Format.h" 54 #include "llvm/Support/GraphWriter.h" 55 #include "llvm/Support/Host.h" 56 #include "llvm/Support/InitLLVM.h" 57 #include "llvm/Support/MemoryBuffer.h" 58 #include "llvm/Support/SourceMgr.h" 59 #include "llvm/Support/StringSaver.h" 60 #include "llvm/Support/TargetRegistry.h" 61 #include "llvm/Support/TargetSelect.h" 62 #include "llvm/Support/WithColor.h" 63 #include "llvm/Support/raw_ostream.h" 64 #include <algorithm> 65 #include <cctype> 66 #include <cstring> 67 #include <system_error> 68 #include <unordered_map> 69 #include <utility> 70 71 using namespace llvm; 72 using namespace object; 73 74 cl::opt<bool> 75 llvm::AllHeaders("all-headers", 76 cl::desc("Display all available header information")); 77 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 78 cl::NotHidden, cl::aliasopt(AllHeaders)); 79 80 static cl::list<std::string> 81 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 82 83 cl::opt<bool> 84 llvm::Disassemble("disassemble", 85 cl::desc("Display assembler mnemonics for the machine instructions")); 86 static cl::alias Disassembled("d", cl::desc("Alias for --disassemble"), 87 cl::NotHidden, cl::aliasopt(Disassemble)); 88 89 cl::opt<bool> 90 llvm::DisassembleAll("disassemble-all", 91 cl::desc("Display assembler mnemonics for the machine instructions")); 92 static cl::alias DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 93 cl::NotHidden, cl::aliasopt(DisassembleAll)); 94 95 cl::opt<bool> llvm::Demangle("demangle", cl::desc("Demangle symbols names"), 96 cl::init(false)); 97 98 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 99 cl::NotHidden, cl::aliasopt(llvm::Demangle)); 100 101 static cl::list<std::string> 102 DisassembleFunctions("df", 103 cl::CommaSeparated, 104 cl::desc("List of functions to disassemble")); 105 static StringSet<> DisasmFuncsSet; 106 107 cl::opt<bool> 108 llvm::Relocations("reloc", 109 cl::desc("Display the relocation entries in the file")); 110 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 111 cl::NotHidden, 112 cl::aliasopt(llvm::Relocations)); 113 114 cl::opt<bool> 115 llvm::DynamicRelocations("dynamic-reloc", 116 cl::desc("Display the dynamic relocation entries in the file")); 117 static cl::alias DynamicRelocationsd("R", cl::desc("Alias for --dynamic-reloc"), 118 cl::NotHidden, 119 cl::aliasopt(DynamicRelocations)); 120 121 cl::opt<bool> 122 llvm::SectionContents("full-contents", 123 cl::desc("Display the content of each section")); 124 static cl::alias SectionContentsShort("s", 125 cl::desc("Alias for --full-contents"), 126 cl::NotHidden, 127 cl::aliasopt(SectionContents)); 128 129 cl::opt<bool> llvm::SymbolTable("syms", cl::desc("Display the symbol table")); 130 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 131 cl::NotHidden, 132 cl::aliasopt(llvm::SymbolTable)); 133 134 cl::opt<bool> 135 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 136 137 cl::opt<bool> 138 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 139 140 cl::opt<bool> 141 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 142 143 cl::opt<bool> 144 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 145 146 cl::opt<bool> 147 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 148 149 cl::opt<bool> 150 llvm::RawClangAST("raw-clang-ast", 151 cl::desc("Dump the raw binary contents of the clang AST section")); 152 153 static cl::opt<bool> 154 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 155 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 156 cl::aliasopt(MachOOpt)); 157 158 cl::opt<std::string> 159 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 160 "see -version for available targets")); 161 162 cl::opt<std::string> 163 llvm::MCPU("mcpu", 164 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 165 cl::value_desc("cpu-name"), 166 cl::init("")); 167 168 cl::opt<std::string> 169 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 170 "see -version for available targets")); 171 172 cl::opt<bool> 173 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 174 "headers for each section.")); 175 static cl::alias SectionHeadersShort("headers", 176 cl::desc("Alias for --section-headers"), 177 cl::NotHidden, 178 cl::aliasopt(SectionHeaders)); 179 static cl::alias SectionHeadersShorter("h", 180 cl::desc("Alias for --section-headers"), 181 cl::NotHidden, 182 cl::aliasopt(SectionHeaders)); 183 184 cl::list<std::string> 185 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 186 "With -macho dump segment,section")); 187 cl::alias static FilterSectionsj("j", cl::desc("Alias for --section"), 188 cl::NotHidden, 189 cl::aliasopt(llvm::FilterSections)); 190 191 cl::list<std::string> 192 llvm::MAttrs("mattr", 193 cl::CommaSeparated, 194 cl::desc("Target specific attributes"), 195 cl::value_desc("a1,+a2,-a3,...")); 196 197 cl::opt<bool> 198 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 199 "instructions, do not print " 200 "the instruction bytes.")); 201 cl::opt<bool> 202 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 203 204 cl::opt<bool> 205 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 206 207 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 208 cl::NotHidden, cl::aliasopt(UnwindInfo)); 209 210 cl::opt<bool> 211 llvm::PrivateHeaders("private-headers", 212 cl::desc("Display format specific file headers")); 213 214 cl::opt<bool> 215 llvm::FirstPrivateHeader("private-header", 216 cl::desc("Display only the first format specific file " 217 "header")); 218 219 static cl::alias PrivateHeadersShort("p", 220 cl::desc("Alias for --private-headers"), 221 cl::NotHidden, 222 cl::aliasopt(PrivateHeaders)); 223 224 cl::opt<bool> llvm::FileHeaders( 225 "file-headers", 226 cl::desc("Display the contents of the overall file header")); 227 228 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 229 cl::NotHidden, cl::aliasopt(FileHeaders)); 230 231 cl::opt<bool> 232 llvm::ArchiveHeaders("archive-headers", 233 cl::desc("Display archive header information")); 234 235 cl::alias ArchiveHeadersShort("a", cl::desc("Alias for --archive-headers"), 236 cl::NotHidden, cl::aliasopt(ArchiveHeaders)); 237 238 cl::opt<bool> 239 llvm::PrintImmHex("print-imm-hex", 240 cl::desc("Use hex format for immediate values")); 241 242 cl::opt<bool> PrintFaultMaps("fault-map-section", 243 cl::desc("Display contents of faultmap section")); 244 245 cl::opt<DIDumpType> llvm::DwarfDumpType( 246 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 247 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 248 249 cl::opt<bool> PrintSource( 250 "source", 251 cl::desc( 252 "Display source inlined with disassembly. Implies disassemble object")); 253 254 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), cl::NotHidden, 255 cl::aliasopt(PrintSource)); 256 257 cl::opt<bool> PrintLines("line-numbers", 258 cl::desc("Display source line numbers with " 259 "disassembly. Implies disassemble object")); 260 261 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 262 cl::NotHidden, cl::aliasopt(PrintLines)); 263 264 cl::opt<unsigned long long> 265 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 266 cl::value_desc("address"), cl::init(0)); 267 cl::opt<unsigned long long> 268 StopAddress("stop-address", 269 cl::desc("Stop disassembly at address"), 270 cl::value_desc("address"), cl::init(UINT64_MAX)); 271 272 cl::opt<bool> DisassembleZeroes( 273 "disassemble-zeroes", 274 cl::desc("Do not skip blocks of zeroes when disassembling")); 275 cl::alias DisassembleZeroesShort("z", 276 cl::desc("Alias for --disassemble-zeroes"), 277 cl::NotHidden, 278 cl::aliasopt(DisassembleZeroes)); 279 280 static StringRef ToolName; 281 282 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 283 284 SectionFilter llvm::ToolSectionFilter(llvm::object::ObjectFile const &O) { 285 return SectionFilter( 286 [](llvm::object::SectionRef const &S) { 287 if (FilterSections.empty()) 288 return true; 289 llvm::StringRef String; 290 std::error_code error = S.getName(String); 291 if (error) 292 return false; 293 return is_contained(FilterSections, String); 294 }, 295 O); 296 } 297 298 void llvm::error(std::error_code EC) { 299 if (!EC) 300 return; 301 WithColor::error(errs(), ToolName) 302 << "reading file: " << EC.message() << ".\n"; 303 errs().flush(); 304 exit(1); 305 } 306 307 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 308 WithColor::error(errs(), ToolName) << Message << ".\n"; 309 errs().flush(); 310 exit(1); 311 } 312 313 void llvm::warn(StringRef Message) { 314 WithColor::warning(errs(), ToolName) << Message << ".\n"; 315 errs().flush(); 316 } 317 318 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 319 Twine Message) { 320 WithColor::error(errs(), ToolName) 321 << "'" << File << "': " << Message << ".\n"; 322 exit(1); 323 } 324 325 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 326 std::error_code EC) { 327 assert(EC); 328 WithColor::error(errs(), ToolName) 329 << "'" << File << "': " << EC.message() << ".\n"; 330 exit(1); 331 } 332 333 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 334 llvm::Error E) { 335 assert(E); 336 std::string Buf; 337 raw_string_ostream OS(Buf); 338 logAllUnhandledErrors(std::move(E), OS); 339 OS.flush(); 340 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 341 exit(1); 342 } 343 344 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 345 StringRef FileName, 346 llvm::Error E, 347 StringRef ArchitectureName) { 348 assert(E); 349 WithColor::error(errs(), ToolName); 350 if (ArchiveName != "") 351 errs() << ArchiveName << "(" << FileName << ")"; 352 else 353 errs() << "'" << FileName << "'"; 354 if (!ArchitectureName.empty()) 355 errs() << " (for architecture " << ArchitectureName << ")"; 356 std::string Buf; 357 raw_string_ostream OS(Buf); 358 logAllUnhandledErrors(std::move(E), OS); 359 OS.flush(); 360 errs() << ": " << Buf; 361 exit(1); 362 } 363 364 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 365 const object::Archive::Child &C, 366 llvm::Error E, 367 StringRef ArchitectureName) { 368 Expected<StringRef> NameOrErr = C.getName(); 369 // TODO: if we have a error getting the name then it would be nice to print 370 // the index of which archive member this is and or its offset in the 371 // archive instead of "???" as the name. 372 if (!NameOrErr) { 373 consumeError(NameOrErr.takeError()); 374 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 375 } else 376 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 377 ArchitectureName); 378 } 379 380 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 381 // Figure out the target triple. 382 llvm::Triple TheTriple("unknown-unknown-unknown"); 383 if (TripleName.empty()) { 384 if (Obj) 385 TheTriple = Obj->makeTriple(); 386 } else { 387 TheTriple.setTriple(Triple::normalize(TripleName)); 388 389 // Use the triple, but also try to combine with ARM build attributes. 390 if (Obj) { 391 auto Arch = Obj->getArch(); 392 if (Arch == Triple::arm || Arch == Triple::armeb) 393 Obj->setARMSubArch(TheTriple); 394 } 395 } 396 397 // Get the target specific parser. 398 std::string Error; 399 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 400 Error); 401 if (!TheTarget) { 402 if (Obj) 403 report_error(Obj->getFileName(), "can't find target: " + Error); 404 else 405 error("can't find target: " + Error); 406 } 407 408 // Update the triple name and return the found target. 409 TripleName = TheTriple.getTriple(); 410 return TheTarget; 411 } 412 413 bool llvm::isRelocAddressLess(RelocationRef A, RelocationRef B) { 414 return A.getOffset() < B.getOffset(); 415 } 416 417 static std::error_code getRelocationValueString(const RelocationRef &Rel, 418 SmallVectorImpl<char> &Result) { 419 const ObjectFile *Obj = Rel.getObject(); 420 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 421 return getELFRelocationValueString(ELF, Rel, Result); 422 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 423 return getCOFFRelocationValueString(COFF, Rel, Result); 424 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 425 return getWasmRelocationValueString(Wasm, Rel, Result); 426 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 427 return getMachORelocationValueString(MachO, Rel, Result); 428 llvm_unreachable("unknown object file format"); 429 } 430 431 /// Indicates whether this relocation should hidden when listing 432 /// relocations, usually because it is the trailing part of a multipart 433 /// relocation that will be printed as part of the leading relocation. 434 static bool getHidden(RelocationRef RelRef) { 435 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 436 if (!MachO) 437 return false; 438 439 unsigned Arch = MachO->getArch(); 440 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 441 uint64_t Type = MachO->getRelocationType(Rel); 442 443 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 444 // is always hidden. 445 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 446 return Type == MachO::GENERIC_RELOC_PAIR; 447 448 if (Arch == Triple::x86_64) { 449 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 450 // an X86_64_RELOC_SUBTRACTOR. 451 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 452 DataRefImpl RelPrev = Rel; 453 RelPrev.d.a--; 454 uint64_t PrevType = MachO->getRelocationType(RelPrev); 455 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 456 return true; 457 } 458 } 459 460 return false; 461 } 462 463 namespace { 464 class SourcePrinter { 465 protected: 466 DILineInfo OldLineInfo; 467 const ObjectFile *Obj = nullptr; 468 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 469 // File name to file contents of source 470 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 471 // Mark the line endings of the cached source 472 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 473 474 private: 475 bool cacheSource(const DILineInfo& LineInfoFile); 476 477 public: 478 SourcePrinter() = default; 479 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 480 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 481 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 482 DefaultArch); 483 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 484 } 485 virtual ~SourcePrinter() = default; 486 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 487 StringRef Delimiter = "; "); 488 }; 489 490 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 491 std::unique_ptr<MemoryBuffer> Buffer; 492 if (LineInfo.Source) { 493 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 494 } else { 495 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 496 if (!BufferOrError) 497 return false; 498 Buffer = std::move(*BufferOrError); 499 } 500 // Chomp the file to get lines 501 size_t BufferSize = Buffer->getBufferSize(); 502 const char *BufferStart = Buffer->getBufferStart(); 503 for (const char *Start = BufferStart, *End = BufferStart; 504 End < BufferStart + BufferSize; End++) 505 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 506 (*End == '\r' && *(End + 1) == '\n')) { 507 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 508 if (*End == '\r') 509 End++; 510 Start = End + 1; 511 } 512 SourceCache[LineInfo.FileName] = std::move(Buffer); 513 return true; 514 } 515 516 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 517 StringRef Delimiter) { 518 if (!Symbolizer) 519 return; 520 DILineInfo LineInfo = DILineInfo(); 521 auto ExpectecLineInfo = 522 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 523 if (!ExpectecLineInfo) 524 consumeError(ExpectecLineInfo.takeError()); 525 else 526 LineInfo = *ExpectecLineInfo; 527 528 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 529 LineInfo.Line == 0) 530 return; 531 532 if (PrintLines) 533 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 534 if (PrintSource) { 535 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 536 if (!cacheSource(LineInfo)) 537 return; 538 auto FileBuffer = SourceCache.find(LineInfo.FileName); 539 if (FileBuffer != SourceCache.end()) { 540 auto LineBuffer = LineCache.find(LineInfo.FileName); 541 if (LineBuffer != LineCache.end()) { 542 if (LineInfo.Line > LineBuffer->second.size()) 543 return; 544 // Vector begins at 0, line numbers are non-zero 545 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 546 << "\n"; 547 } 548 } 549 } 550 OldLineInfo = LineInfo; 551 } 552 553 static bool isArmElf(const ObjectFile *Obj) { 554 return (Obj->isELF() && 555 (Obj->getArch() == Triple::aarch64 || 556 Obj->getArch() == Triple::aarch64_be || 557 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 558 Obj->getArch() == Triple::thumb || 559 Obj->getArch() == Triple::thumbeb)); 560 } 561 562 class PrettyPrinter { 563 public: 564 virtual ~PrettyPrinter() = default; 565 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 566 ArrayRef<uint8_t> Bytes, uint64_t Address, 567 raw_ostream &OS, StringRef Annot, 568 MCSubtargetInfo const &STI, SourcePrinter *SP, 569 std::vector<RelocationRef> *Rels = nullptr) { 570 if (SP && (PrintSource || PrintLines)) 571 SP->printSourceLine(OS, Address); 572 if (!NoLeadingAddr) 573 OS << format("%8" PRIx64 ":", Address); 574 if (!NoShowRawInsn) { 575 OS << "\t"; 576 dumpBytes(Bytes, OS); 577 } 578 if (MI) 579 IP.printInst(MI, OS, "", STI); 580 else 581 OS << " <unknown>"; 582 } 583 }; 584 PrettyPrinter PrettyPrinterInst; 585 class HexagonPrettyPrinter : public PrettyPrinter { 586 public: 587 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 588 raw_ostream &OS) { 589 uint32_t opcode = 590 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 591 if (!NoLeadingAddr) 592 OS << format("%8" PRIx64 ":", Address); 593 if (!NoShowRawInsn) { 594 OS << "\t"; 595 dumpBytes(Bytes.slice(0, 4), OS); 596 OS << format("%08" PRIx32, opcode); 597 } 598 } 599 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 600 uint64_t Address, raw_ostream &OS, StringRef Annot, 601 MCSubtargetInfo const &STI, SourcePrinter *SP, 602 std::vector<RelocationRef> *Rels) override { 603 if (SP && (PrintSource || PrintLines)) 604 SP->printSourceLine(OS, Address, ""); 605 if (!MI) { 606 printLead(Bytes, Address, OS); 607 OS << " <unknown>"; 608 return; 609 } 610 std::string Buffer; 611 { 612 raw_string_ostream TempStream(Buffer); 613 IP.printInst(MI, TempStream, "", STI); 614 } 615 StringRef Contents(Buffer); 616 // Split off bundle attributes 617 auto PacketBundle = Contents.rsplit('\n'); 618 // Split off first instruction from the rest 619 auto HeadTail = PacketBundle.first.split('\n'); 620 auto Preamble = " { "; 621 auto Separator = ""; 622 StringRef Fmt = "\t\t\t%08" PRIx64 ": "; 623 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 624 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 625 626 // Hexagon's packets require relocations to be inline rather than 627 // clustered at the end of the packet. 628 auto PrintReloc = [&]() -> void { 629 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address)) { 630 if (RelCur->getOffset() == Address) { 631 SmallString<16> Name; 632 SmallString<32> Val; 633 RelCur->getTypeName(Name); 634 error(getRelocationValueString(*RelCur, Val)); 635 OS << Separator << format(Fmt.data(), Address) << Name << "\t" << Val 636 << "\n"; 637 return; 638 } 639 ++RelCur; 640 } 641 }; 642 643 while (!HeadTail.first.empty()) { 644 OS << Separator; 645 Separator = "\n"; 646 if (SP && (PrintSource || PrintLines)) 647 SP->printSourceLine(OS, Address, ""); 648 printLead(Bytes, Address, OS); 649 OS << Preamble; 650 Preamble = " "; 651 StringRef Inst; 652 auto Duplex = HeadTail.first.split('\v'); 653 if (!Duplex.second.empty()) { 654 OS << Duplex.first; 655 OS << "; "; 656 Inst = Duplex.second; 657 } 658 else 659 Inst = HeadTail.first; 660 OS << Inst; 661 HeadTail = HeadTail.second.split('\n'); 662 if (HeadTail.first.empty()) 663 OS << " } " << PacketBundle.second; 664 PrintReloc(); 665 Bytes = Bytes.slice(4); 666 Address += 4; 667 } 668 } 669 }; 670 HexagonPrettyPrinter HexagonPrettyPrinterInst; 671 672 class AMDGCNPrettyPrinter : public PrettyPrinter { 673 public: 674 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 675 uint64_t Address, raw_ostream &OS, StringRef Annot, 676 MCSubtargetInfo const &STI, SourcePrinter *SP, 677 std::vector<RelocationRef> *Rels) override { 678 if (SP && (PrintSource || PrintLines)) 679 SP->printSourceLine(OS, Address); 680 681 typedef support::ulittle32_t U32; 682 683 if (MI) { 684 SmallString<40> InstStr; 685 raw_svector_ostream IS(InstStr); 686 687 IP.printInst(MI, IS, "", STI); 688 689 OS << left_justify(IS.str(), 60); 690 } else { 691 // an unrecognized encoding - this is probably data so represent it 692 // using the .long directive, or .byte directive if fewer than 4 bytes 693 // remaining 694 if (Bytes.size() >= 4) { 695 OS << format("\t.long 0x%08" PRIx32 " ", 696 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 697 OS.indent(42); 698 } else { 699 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 700 for (unsigned int i = 1; i < Bytes.size(); i++) 701 OS << format(", 0x%02" PRIx8, Bytes[i]); 702 OS.indent(55 - (6 * Bytes.size())); 703 } 704 } 705 706 OS << format("// %012" PRIX64 ": ", Address); 707 if (Bytes.size() >=4) { 708 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 709 Bytes.size() / sizeof(U32))) 710 // D should be explicitly casted to uint32_t here as it is passed 711 // by format to snprintf as vararg. 712 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 713 } else { 714 for (unsigned int i = 0; i < Bytes.size(); i++) 715 OS << format("%02" PRIX8 " ", Bytes[i]); 716 } 717 718 if (!Annot.empty()) 719 OS << "// " << Annot; 720 } 721 }; 722 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 723 724 class BPFPrettyPrinter : public PrettyPrinter { 725 public: 726 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 727 uint64_t Address, raw_ostream &OS, StringRef Annot, 728 MCSubtargetInfo const &STI, SourcePrinter *SP, 729 std::vector<RelocationRef> *Rels) override { 730 if (SP && (PrintSource || PrintLines)) 731 SP->printSourceLine(OS, Address); 732 if (!NoLeadingAddr) 733 OS << format("%8" PRId64 ":", Address / 8); 734 if (!NoShowRawInsn) { 735 OS << "\t"; 736 dumpBytes(Bytes, OS); 737 } 738 if (MI) 739 IP.printInst(MI, OS, "", STI); 740 else 741 OS << " <unknown>"; 742 } 743 }; 744 BPFPrettyPrinter BPFPrettyPrinterInst; 745 746 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 747 switch(Triple.getArch()) { 748 default: 749 return PrettyPrinterInst; 750 case Triple::hexagon: 751 return HexagonPrettyPrinterInst; 752 case Triple::amdgcn: 753 return AMDGCNPrettyPrinterInst; 754 case Triple::bpfel: 755 case Triple::bpfeb: 756 return BPFPrettyPrinterInst; 757 } 758 } 759 } 760 761 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 762 assert(Obj->isELF()); 763 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 764 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 765 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 766 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 767 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 768 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 769 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 770 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 771 llvm_unreachable("Unsupported binary format"); 772 } 773 774 template <class ELFT> static void 775 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 776 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 777 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 778 uint8_t SymbolType = Symbol.getELFType(); 779 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 780 continue; 781 782 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 783 if (!AddressOrErr) 784 report_error(Obj->getFileName(), AddressOrErr.takeError()); 785 786 Expected<StringRef> Name = Symbol.getName(); 787 if (!Name) 788 report_error(Obj->getFileName(), Name.takeError()); 789 if (Name->empty()) 790 continue; 791 792 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 793 if (!SectionOrErr) 794 report_error(Obj->getFileName(), SectionOrErr.takeError()); 795 section_iterator SecI = *SectionOrErr; 796 if (SecI == Obj->section_end()) 797 continue; 798 799 AllSymbols[*SecI].emplace_back(*AddressOrErr, *Name, SymbolType); 800 } 801 } 802 803 static void 804 addDynamicElfSymbols(const ObjectFile *Obj, 805 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 806 assert(Obj->isELF()); 807 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 808 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 809 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 810 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 811 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 812 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 813 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 814 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 815 else 816 llvm_unreachable("Unsupported binary format"); 817 } 818 819 static void addPltEntries(const ObjectFile *Obj, 820 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 821 StringSaver &Saver) { 822 Optional<SectionRef> Plt = None; 823 for (const SectionRef &Section : Obj->sections()) { 824 StringRef Name; 825 if (Section.getName(Name)) 826 continue; 827 if (Name == ".plt") 828 Plt = Section; 829 } 830 if (!Plt) 831 return; 832 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 833 for (auto PltEntry : ElfObj->getPltAddresses()) { 834 SymbolRef Symbol(PltEntry.first, ElfObj); 835 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 836 837 Expected<StringRef> NameOrErr = Symbol.getName(); 838 if (!NameOrErr) 839 report_error(Obj->getFileName(), NameOrErr.takeError()); 840 if (NameOrErr->empty()) 841 continue; 842 StringRef Name = Saver.save((*NameOrErr + "@plt").str()); 843 844 AllSymbols[*Plt].emplace_back(PltEntry.second, Name, SymbolType); 845 } 846 } 847 } 848 849 // Normally the disassembly output will skip blocks of zeroes. This function 850 // returns the number of zero bytes that can be skipped when dumping the 851 // disassembly of the instructions in Buf. 852 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 853 // When -z or --disassemble-zeroes are given we always dissasemble them. 854 if (DisassembleZeroes) 855 return 0; 856 857 // Find the number of leading zeroes. 858 size_t N = 0; 859 while (N < Buf.size() && !Buf[N]) 860 ++N; 861 862 // We may want to skip blocks of zero bytes, but unless we see 863 // at least 8 of them in a row. 864 if (N < 8) 865 return 0; 866 867 // We skip zeroes in multiples of 4 because do not want to truncate an 868 // instruction if it starts with a zero byte. 869 return N & ~0x3; 870 } 871 872 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 873 if (StartAddress > StopAddress) 874 error("Start address should be less than stop address"); 875 876 const Target *TheTarget = getTarget(Obj); 877 878 // Package up features to be passed to target/subtarget 879 SubtargetFeatures Features = Obj->getFeatures(); 880 if (!MAttrs.empty()) 881 for (unsigned I = 0; I != MAttrs.size(); ++I) 882 Features.AddFeature(MAttrs[I]); 883 884 std::unique_ptr<const MCRegisterInfo> MRI( 885 TheTarget->createMCRegInfo(TripleName)); 886 if (!MRI) 887 report_error(Obj->getFileName(), "no register info for target " + 888 TripleName); 889 890 // Set up disassembler. 891 std::unique_ptr<const MCAsmInfo> AsmInfo( 892 TheTarget->createMCAsmInfo(*MRI, TripleName)); 893 if (!AsmInfo) 894 report_error(Obj->getFileName(), "no assembly info for target " + 895 TripleName); 896 std::unique_ptr<const MCSubtargetInfo> STI( 897 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 898 if (!STI) 899 report_error(Obj->getFileName(), "no subtarget info for target " + 900 TripleName); 901 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 902 if (!MII) 903 report_error(Obj->getFileName(), "no instruction info for target " + 904 TripleName); 905 MCObjectFileInfo MOFI; 906 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 907 // FIXME: for now initialize MCObjectFileInfo with default values 908 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 909 910 std::unique_ptr<MCDisassembler> DisAsm( 911 TheTarget->createMCDisassembler(*STI, Ctx)); 912 if (!DisAsm) 913 report_error(Obj->getFileName(), "no disassembler for target " + 914 TripleName); 915 916 std::unique_ptr<const MCInstrAnalysis> MIA( 917 TheTarget->createMCInstrAnalysis(MII.get())); 918 919 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 920 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 921 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 922 if (!IP) 923 report_error(Obj->getFileName(), "no instruction printer for target " + 924 TripleName); 925 IP->setPrintImmHex(PrintImmHex); 926 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 927 928 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 929 "\t\t\t%08" PRIx64 ": "; 930 931 SourcePrinter SP(Obj, TheTarget->getName()); 932 933 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 934 // in RelocSecs contain the relocations for section S. 935 std::error_code EC; 936 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 937 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 938 section_iterator Sec2 = Section.getRelocatedSection(); 939 if (Sec2 != Obj->section_end()) 940 SectionRelocMap[*Sec2].push_back(Section); 941 } 942 943 // Create a mapping from virtual address to symbol name. This is used to 944 // pretty print the symbols while disassembling. 945 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 946 SectionSymbolsTy AbsoluteSymbols; 947 for (const SymbolRef &Symbol : Obj->symbols()) { 948 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 949 if (!AddressOrErr) 950 report_error(Obj->getFileName(), AddressOrErr.takeError()); 951 uint64_t Address = *AddressOrErr; 952 953 Expected<StringRef> Name = Symbol.getName(); 954 if (!Name) 955 report_error(Obj->getFileName(), Name.takeError()); 956 if (Name->empty()) 957 continue; 958 959 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 960 if (!SectionOrErr) 961 report_error(Obj->getFileName(), SectionOrErr.takeError()); 962 963 uint8_t SymbolType = ELF::STT_NOTYPE; 964 if (Obj->isELF()) 965 SymbolType = getElfSymbolType(Obj, Symbol); 966 967 section_iterator SecI = *SectionOrErr; 968 if (SecI != Obj->section_end()) 969 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 970 else 971 AbsoluteSymbols.emplace_back(Address, *Name, SymbolType); 972 973 974 } 975 if (AllSymbols.empty() && Obj->isELF()) 976 addDynamicElfSymbols(Obj, AllSymbols); 977 978 BumpPtrAllocator A; 979 StringSaver Saver(A); 980 addPltEntries(Obj, AllSymbols, Saver); 981 982 // Create a mapping from virtual address to section. 983 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 984 for (SectionRef Sec : Obj->sections()) 985 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 986 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 987 988 // Linked executables (.exe and .dll files) typically don't include a real 989 // symbol table but they might contain an export table. 990 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 991 for (const auto &ExportEntry : COFFObj->export_directories()) { 992 StringRef Name; 993 error(ExportEntry.getSymbolName(Name)); 994 if (Name.empty()) 995 continue; 996 uint32_t RVA; 997 error(ExportEntry.getExportRVA(RVA)); 998 999 uint64_t VA = COFFObj->getImageBase() + RVA; 1000 auto Sec = std::upper_bound( 1001 SectionAddresses.begin(), SectionAddresses.end(), VA, 1002 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1003 return LHS < RHS.first; 1004 }); 1005 if (Sec != SectionAddresses.begin()) 1006 --Sec; 1007 else 1008 Sec = SectionAddresses.end(); 1009 1010 if (Sec != SectionAddresses.end()) 1011 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1012 else 1013 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1014 } 1015 } 1016 1017 // Sort all the symbols, this allows us to use a simple binary search to find 1018 // a symbol near an address. 1019 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1020 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1021 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1022 1023 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1024 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1025 continue; 1026 1027 uint64_t SectionAddr = Section.getAddress(); 1028 uint64_t SectSize = Section.getSize(); 1029 if (!SectSize) 1030 continue; 1031 1032 // Get the list of all the symbols in this section. 1033 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1034 std::vector<uint64_t> DataMappingSymsAddr; 1035 std::vector<uint64_t> TextMappingSymsAddr; 1036 if (isArmElf(Obj)) { 1037 for (const auto &Symb : Symbols) { 1038 uint64_t Address = std::get<0>(Symb); 1039 StringRef Name = std::get<1>(Symb); 1040 if (Name.startswith("$d")) 1041 DataMappingSymsAddr.push_back(Address - SectionAddr); 1042 if (Name.startswith("$x")) 1043 TextMappingSymsAddr.push_back(Address - SectionAddr); 1044 if (Name.startswith("$a")) 1045 TextMappingSymsAddr.push_back(Address - SectionAddr); 1046 if (Name.startswith("$t")) 1047 TextMappingSymsAddr.push_back(Address - SectionAddr); 1048 } 1049 } 1050 1051 llvm::sort(DataMappingSymsAddr); 1052 llvm::sort(TextMappingSymsAddr); 1053 1054 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1055 // AMDGPU disassembler uses symbolizer for printing labels 1056 std::unique_ptr<MCRelocationInfo> RelInfo( 1057 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1058 if (RelInfo) { 1059 std::unique_ptr<MCSymbolizer> Symbolizer( 1060 TheTarget->createMCSymbolizer( 1061 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1062 DisAsm->setSymbolizer(std::move(Symbolizer)); 1063 } 1064 } 1065 1066 // Make a list of all the relocations for this section. 1067 std::vector<RelocationRef> Rels; 1068 if (InlineRelocs) { 1069 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1070 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1071 Rels.push_back(Reloc); 1072 } 1073 } 1074 } 1075 1076 // Sort relocations by address. 1077 llvm::sort(Rels, isRelocAddressLess); 1078 1079 StringRef SegmentName = ""; 1080 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1081 DataRefImpl DR = Section.getRawDataRefImpl(); 1082 SegmentName = MachO->getSectionFinalSegmentName(DR); 1083 } 1084 StringRef SectionName; 1085 error(Section.getName(SectionName)); 1086 1087 // If the section has no symbol at the start, just insert a dummy one. 1088 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1089 Symbols.insert( 1090 Symbols.begin(), 1091 std::make_tuple(SectionAddr, SectionName, 1092 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1093 } 1094 1095 SmallString<40> Comments; 1096 raw_svector_ostream CommentStream(Comments); 1097 1098 StringRef BytesStr; 1099 error(Section.getContents(BytesStr)); 1100 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1101 BytesStr.size()); 1102 1103 uint64_t Size; 1104 uint64_t Index; 1105 bool PrintedSection = false; 1106 1107 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1108 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1109 // Disassemble symbol by symbol. 1110 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1111 uint64_t Start = std::get<0>(Symbols[SI]) - SectionAddr; 1112 // The end is either the section end or the beginning of the next 1113 // symbol. 1114 uint64_t End = (SI == SE - 1) 1115 ? SectSize 1116 : std::get<0>(Symbols[SI + 1]) - SectionAddr; 1117 // Don't try to disassemble beyond the end of section contents. 1118 if (End > SectSize) 1119 End = SectSize; 1120 // If this symbol has the same address as the next symbol, then skip it. 1121 if (Start >= End) 1122 continue; 1123 1124 // Check if we need to skip symbol 1125 // Skip if the symbol's data is not between StartAddress and StopAddress 1126 if (End + SectionAddr < StartAddress || 1127 Start + SectionAddr > StopAddress) { 1128 continue; 1129 } 1130 1131 /// Skip if user requested specific symbols and this is not in the list 1132 if (!DisasmFuncsSet.empty() && 1133 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1134 continue; 1135 1136 if (!PrintedSection) { 1137 PrintedSection = true; 1138 outs() << "Disassembly of section "; 1139 if (!SegmentName.empty()) 1140 outs() << SegmentName << ","; 1141 outs() << SectionName << ':'; 1142 } 1143 1144 // Stop disassembly at the stop address specified 1145 if (End + SectionAddr > StopAddress) 1146 End = StopAddress - SectionAddr; 1147 1148 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1149 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1150 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1151 Start += 256; 1152 } 1153 if (SI == SE - 1 || 1154 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1155 // cut trailing zeroes at the end of kernel 1156 // cut up to 256 bytes 1157 const uint64_t EndAlign = 256; 1158 const auto Limit = End - (std::min)(EndAlign, End - Start); 1159 while (End > Limit && 1160 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1161 End -= 4; 1162 } 1163 } 1164 1165 outs() << '\n'; 1166 if (!NoLeadingAddr) 1167 outs() << format("%016" PRIx64 " ", SectionAddr + Start); 1168 1169 StringRef SymbolName = std::get<1>(Symbols[SI]); 1170 if (Demangle) 1171 outs() << demangle(SymbolName) << ":\n"; 1172 else 1173 outs() << SymbolName << ":\n"; 1174 1175 // Don't print raw contents of a virtual section. A virtual section 1176 // doesn't have any contents in the file. 1177 if (Section.isVirtual()) { 1178 outs() << "...\n"; 1179 continue; 1180 } 1181 1182 #ifndef NDEBUG 1183 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1184 #else 1185 raw_ostream &DebugOut = nulls(); 1186 #endif 1187 1188 // Some targets (like WebAssembly) have a special prelude at the start 1189 // of each symbol. 1190 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1191 SectionAddr + Start, DebugOut, CommentStream); 1192 Start += Size; 1193 1194 for (Index = Start; Index < End; Index += Size) { 1195 MCInst Inst; 1196 1197 if (Index + SectionAddr < StartAddress || 1198 Index + SectionAddr > StopAddress) { 1199 // skip byte by byte till StartAddress is reached 1200 Size = 1; 1201 continue; 1202 } 1203 // AArch64 ELF binaries can interleave data and text in the 1204 // same section. We rely on the markers introduced to 1205 // understand what we need to dump. If the data marker is within a 1206 // function, it is denoted as a word/short etc 1207 if (isArmElf(Obj) && std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1208 !DisassembleAll) { 1209 uint64_t Stride = 0; 1210 1211 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1212 DataMappingSymsAddr.end(), Index); 1213 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1214 // Switch to data. 1215 while (Index < End) { 1216 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1217 outs() << "\t"; 1218 if (Index + 4 <= End) { 1219 Stride = 4; 1220 dumpBytes(Bytes.slice(Index, 4), outs()); 1221 outs() << "\t.word\t"; 1222 uint32_t Data = 0; 1223 if (Obj->isLittleEndian()) { 1224 const auto Word = 1225 reinterpret_cast<const support::ulittle32_t *>( 1226 Bytes.data() + Index); 1227 Data = *Word; 1228 } else { 1229 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1230 Bytes.data() + Index); 1231 Data = *Word; 1232 } 1233 outs() << "0x" << format("%08" PRIx32, Data); 1234 } else if (Index + 2 <= End) { 1235 Stride = 2; 1236 dumpBytes(Bytes.slice(Index, 2), outs()); 1237 outs() << "\t\t.short\t"; 1238 uint16_t Data = 0; 1239 if (Obj->isLittleEndian()) { 1240 const auto Short = 1241 reinterpret_cast<const support::ulittle16_t *>( 1242 Bytes.data() + Index); 1243 Data = *Short; 1244 } else { 1245 const auto Short = 1246 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1247 Index); 1248 Data = *Short; 1249 } 1250 outs() << "0x" << format("%04" PRIx16, Data); 1251 } else { 1252 Stride = 1; 1253 dumpBytes(Bytes.slice(Index, 1), outs()); 1254 outs() << "\t\t.byte\t"; 1255 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1256 } 1257 Index += Stride; 1258 outs() << "\n"; 1259 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1260 TextMappingSymsAddr.end(), Index); 1261 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1262 break; 1263 } 1264 } 1265 } 1266 1267 // If there is a data symbol inside an ELF text section and we are only 1268 // disassembling text (applicable all architectures), 1269 // we are in a situation where we must print the data and not 1270 // disassemble it. 1271 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1272 !DisassembleAll && Section.isText()) { 1273 // print out data up to 8 bytes at a time in hex and ascii 1274 uint8_t AsciiData[9] = {'\0'}; 1275 uint8_t Byte; 1276 int NumBytes = 0; 1277 1278 for (Index = Start; Index < End; Index += 1) { 1279 if (((SectionAddr + Index) < StartAddress) || 1280 ((SectionAddr + Index) > StopAddress)) 1281 continue; 1282 if (NumBytes == 0) { 1283 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1284 outs() << "\t"; 1285 } 1286 Byte = Bytes.slice(Index)[0]; 1287 outs() << format(" %02x", Byte); 1288 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1289 1290 uint8_t IndentOffset = 0; 1291 NumBytes++; 1292 if (Index == End - 1 || NumBytes > 8) { 1293 // Indent the space for less than 8 bytes data. 1294 // 2 spaces for byte and one for space between bytes 1295 IndentOffset = 3 * (8 - NumBytes); 1296 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1297 AsciiData[Excess] = '\0'; 1298 NumBytes = 8; 1299 } 1300 if (NumBytes == 8) { 1301 AsciiData[8] = '\0'; 1302 outs() << std::string(IndentOffset, ' ') << " "; 1303 outs() << reinterpret_cast<char *>(AsciiData); 1304 outs() << '\n'; 1305 NumBytes = 0; 1306 } 1307 } 1308 } 1309 if (Index >= End) 1310 break; 1311 1312 if (size_t N = 1313 countSkippableZeroBytes(Bytes.slice(Index, End - Index))) { 1314 outs() << "\t\t..." << '\n'; 1315 Index += N; 1316 if (Index >= End) 1317 break; 1318 } 1319 1320 // Disassemble a real instruction or a data when disassemble all is 1321 // provided 1322 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1323 SectionAddr + Index, DebugOut, 1324 CommentStream); 1325 if (Size == 0) 1326 Size = 1; 1327 1328 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1329 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1330 *STI, &SP, &Rels); 1331 outs() << CommentStream.str(); 1332 Comments.clear(); 1333 1334 // Try to resolve the target of a call, tail call, etc. to a specific 1335 // symbol. 1336 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1337 MIA->isConditionalBranch(Inst))) { 1338 uint64_t Target; 1339 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1340 // In a relocatable object, the target's section must reside in 1341 // the same section as the call instruction or it is accessed 1342 // through a relocation. 1343 // 1344 // In a non-relocatable object, the target may be in any section. 1345 // 1346 // N.B. We don't walk the relocations in the relocatable case yet. 1347 auto *TargetSectionSymbols = &Symbols; 1348 if (!Obj->isRelocatableObject()) { 1349 auto SectionAddress = std::upper_bound( 1350 SectionAddresses.begin(), SectionAddresses.end(), Target, 1351 [](uint64_t LHS, 1352 const std::pair<uint64_t, SectionRef> &RHS) { 1353 return LHS < RHS.first; 1354 }); 1355 if (SectionAddress != SectionAddresses.begin()) { 1356 --SectionAddress; 1357 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1358 } else { 1359 TargetSectionSymbols = &AbsoluteSymbols; 1360 } 1361 } 1362 1363 // Find the first symbol in the section whose offset is less than 1364 // or equal to the target. If there isn't a section that contains 1365 // the target, find the nearest preceding absolute symbol. 1366 auto TargetSym = std::upper_bound( 1367 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1368 Target, [](uint64_t LHS, 1369 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1370 return LHS < std::get<0>(RHS); 1371 }); 1372 if (TargetSym == TargetSectionSymbols->begin()) { 1373 TargetSectionSymbols = &AbsoluteSymbols; 1374 TargetSym = std::upper_bound( 1375 AbsoluteSymbols.begin(), AbsoluteSymbols.end(), 1376 Target, [](uint64_t LHS, 1377 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1378 return LHS < std::get<0>(RHS); 1379 }); 1380 } 1381 if (TargetSym != TargetSectionSymbols->begin()) { 1382 --TargetSym; 1383 uint64_t TargetAddress = std::get<0>(*TargetSym); 1384 StringRef TargetName = std::get<1>(*TargetSym); 1385 outs() << " <" << TargetName; 1386 uint64_t Disp = Target - TargetAddress; 1387 if (Disp) 1388 outs() << "+0x" << Twine::utohexstr(Disp); 1389 outs() << '>'; 1390 } 1391 } 1392 } 1393 outs() << "\n"; 1394 1395 // Hexagon does this in pretty printer 1396 if (Obj->getArch() != Triple::hexagon) 1397 // Print relocation for instruction. 1398 while (RelCur != RelEnd) { 1399 uint64_t Addr = RelCur->getOffset(); 1400 SmallString<16> Name; 1401 SmallString<32> Val; 1402 1403 // If this relocation is hidden, skip it. 1404 if (getHidden(*RelCur) || ((SectionAddr + Addr) < StartAddress)) { 1405 ++RelCur; 1406 continue; 1407 } 1408 1409 // Stop when rel_cur's address is past the current instruction. 1410 if (Addr >= Index + Size) 1411 break; 1412 RelCur->getTypeName(Name); 1413 error(getRelocationValueString(*RelCur, Val)); 1414 outs() << format(Fmt.data(), SectionAddr + Addr) << Name << "\t" 1415 << Val << "\n"; 1416 ++RelCur; 1417 } 1418 } 1419 } 1420 } 1421 } 1422 1423 void llvm::printRelocations(const ObjectFile *Obj) { 1424 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1425 "%08" PRIx64; 1426 // Regular objdump doesn't print relocations in non-relocatable object 1427 // files. 1428 if (!Obj->isRelocatableObject()) 1429 return; 1430 1431 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1432 if (Section.relocation_begin() == Section.relocation_end()) 1433 continue; 1434 StringRef SecName; 1435 error(Section.getName(SecName)); 1436 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1437 for (const RelocationRef &Reloc : Section.relocations()) { 1438 uint64_t Address = Reloc.getOffset(); 1439 SmallString<32> RelocName; 1440 SmallString<32> ValueStr; 1441 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1442 continue; 1443 Reloc.getTypeName(RelocName); 1444 error(getRelocationValueString(Reloc, ValueStr)); 1445 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1446 << ValueStr << "\n"; 1447 } 1448 outs() << "\n"; 1449 } 1450 } 1451 1452 void llvm::printDynamicRelocations(const ObjectFile *Obj) { 1453 // For the moment, this option is for ELF only 1454 if (!Obj->isELF()) 1455 return; 1456 1457 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1458 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1459 error("not a dynamic object"); 1460 return; 1461 } 1462 1463 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1464 if (DynRelSec.empty()) 1465 return; 1466 1467 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1468 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1469 for (const SectionRef &Section : DynRelSec) { 1470 if (Section.relocation_begin() == Section.relocation_end()) 1471 continue; 1472 for (const RelocationRef &Reloc : Section.relocations()) { 1473 uint64_t Address = Reloc.getOffset(); 1474 SmallString<32> RelocName; 1475 SmallString<32> ValueStr; 1476 Reloc.getTypeName(RelocName); 1477 error(getRelocationValueString(Reloc, ValueStr)); 1478 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1479 << ValueStr << "\n"; 1480 } 1481 } 1482 } 1483 1484 void llvm::printSectionHeaders(const ObjectFile *Obj) { 1485 outs() << "Sections:\n" 1486 "Idx Name Size Address Type\n"; 1487 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1488 StringRef Name; 1489 error(Section.getName(Name)); 1490 uint64_t Address = Section.getAddress(); 1491 uint64_t Size = Section.getSize(); 1492 bool Text = Section.isText(); 1493 bool Data = Section.isData(); 1494 bool BSS = Section.isBSS(); 1495 std::string Type = (std::string(Text ? "TEXT " : "") + 1496 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1497 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1498 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1499 Address, Type.c_str()); 1500 } 1501 outs() << "\n"; 1502 } 1503 1504 void llvm::printSectionContents(const ObjectFile *Obj) { 1505 std::error_code EC; 1506 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1507 StringRef Name; 1508 StringRef Contents; 1509 error(Section.getName(Name)); 1510 uint64_t BaseAddr = Section.getAddress(); 1511 uint64_t Size = Section.getSize(); 1512 if (!Size) 1513 continue; 1514 1515 outs() << "Contents of section " << Name << ":\n"; 1516 if (Section.isBSS()) { 1517 outs() << format("<skipping contents of bss section at [%04" PRIx64 1518 ", %04" PRIx64 ")>\n", 1519 BaseAddr, BaseAddr + Size); 1520 continue; 1521 } 1522 1523 error(Section.getContents(Contents)); 1524 1525 // Dump out the content as hex and printable ascii characters. 1526 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1527 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1528 // Dump line of hex. 1529 for (std::size_t I = 0; I < 16; ++I) { 1530 if (I != 0 && I % 4 == 0) 1531 outs() << ' '; 1532 if (Addr + I < End) 1533 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1534 << hexdigit(Contents[Addr + I] & 0xF, true); 1535 else 1536 outs() << " "; 1537 } 1538 // Print ascii. 1539 outs() << " "; 1540 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1541 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1542 outs() << Contents[Addr + I]; 1543 else 1544 outs() << "."; 1545 } 1546 outs() << "\n"; 1547 } 1548 } 1549 } 1550 1551 void llvm::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1552 StringRef ArchitectureName) { 1553 outs() << "SYMBOL TABLE:\n"; 1554 1555 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1556 printCOFFSymbolTable(Coff); 1557 return; 1558 } 1559 1560 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1561 // Skip printing the special zero symbol when dumping an ELF file. 1562 // This makes the output consistent with the GNU objdump. 1563 if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O)) 1564 continue; 1565 1566 const SymbolRef &Symbol = *I; 1567 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1568 if (!AddressOrError) 1569 report_error(ArchiveName, O->getFileName(), AddressOrError.takeError(), 1570 ArchitectureName); 1571 uint64_t Address = *AddressOrError; 1572 if ((Address < StartAddress) || (Address > StopAddress)) 1573 continue; 1574 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1575 if (!TypeOrError) 1576 report_error(ArchiveName, O->getFileName(), TypeOrError.takeError(), 1577 ArchitectureName); 1578 SymbolRef::Type Type = *TypeOrError; 1579 uint32_t Flags = Symbol.getFlags(); 1580 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1581 if (!SectionOrErr) 1582 report_error(ArchiveName, O->getFileName(), SectionOrErr.takeError(), 1583 ArchitectureName); 1584 section_iterator Section = *SectionOrErr; 1585 StringRef Name; 1586 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 1587 Section->getName(Name); 1588 } else { 1589 Expected<StringRef> NameOrErr = Symbol.getName(); 1590 if (!NameOrErr) 1591 report_error(ArchiveName, O->getFileName(), NameOrErr.takeError(), 1592 ArchitectureName); 1593 Name = *NameOrErr; 1594 } 1595 1596 bool Global = Flags & SymbolRef::SF_Global; 1597 bool Weak = Flags & SymbolRef::SF_Weak; 1598 bool Absolute = Flags & SymbolRef::SF_Absolute; 1599 bool Common = Flags & SymbolRef::SF_Common; 1600 bool Hidden = Flags & SymbolRef::SF_Hidden; 1601 1602 char GlobLoc = ' '; 1603 if (Type != SymbolRef::ST_Unknown) 1604 GlobLoc = Global ? 'g' : 'l'; 1605 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1606 ? 'd' : ' '; 1607 char FileFunc = ' '; 1608 if (Type == SymbolRef::ST_File) 1609 FileFunc = 'f'; 1610 else if (Type == SymbolRef::ST_Function) 1611 FileFunc = 'F'; 1612 else if (Type == SymbolRef::ST_Data) 1613 FileFunc = 'O'; 1614 1615 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1616 "%08" PRIx64; 1617 1618 outs() << format(Fmt, Address) << " " 1619 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1620 << (Weak ? 'w' : ' ') // Weak? 1621 << ' ' // Constructor. Not supported yet. 1622 << ' ' // Warning. Not supported yet. 1623 << ' ' // Indirect reference to another symbol. 1624 << Debug // Debugging (d) or dynamic (D) symbol. 1625 << FileFunc // Name of function (F), file (f) or object (O). 1626 << ' '; 1627 if (Absolute) { 1628 outs() << "*ABS*"; 1629 } else if (Common) { 1630 outs() << "*COM*"; 1631 } else if (Section == O->section_end()) { 1632 outs() << "*UND*"; 1633 } else { 1634 if (const MachOObjectFile *MachO = 1635 dyn_cast<const MachOObjectFile>(O)) { 1636 DataRefImpl DR = Section->getRawDataRefImpl(); 1637 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1638 outs() << SegmentName << ","; 1639 } 1640 StringRef SectionName; 1641 error(Section->getName(SectionName)); 1642 outs() << SectionName; 1643 } 1644 1645 outs() << '\t'; 1646 if (Common || isa<ELFObjectFileBase>(O)) { 1647 uint64_t Val = 1648 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1649 outs() << format("\t %08" PRIx64 " ", Val); 1650 } 1651 1652 if (Hidden) 1653 outs() << ".hidden "; 1654 1655 if (Demangle) 1656 outs() << demangle(Name) << '\n'; 1657 else 1658 outs() << Name << '\n'; 1659 } 1660 } 1661 1662 static void printUnwindInfo(const ObjectFile *O) { 1663 outs() << "Unwind info:\n\n"; 1664 1665 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1666 printCOFFUnwindInfo(Coff); 1667 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1668 printMachOUnwindInfo(MachO); 1669 else 1670 // TODO: Extract DWARF dump tool to objdump. 1671 WithColor::error(errs(), ToolName) 1672 << "This operation is only currently supported " 1673 "for COFF and MachO object files.\n"; 1674 } 1675 1676 void llvm::printExportsTrie(const ObjectFile *o) { 1677 outs() << "Exports trie:\n"; 1678 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1679 printMachOExportsTrie(MachO); 1680 else 1681 WithColor::error(errs(), ToolName) 1682 << "This operation is only currently supported " 1683 "for Mach-O executable files.\n"; 1684 } 1685 1686 void llvm::printRebaseTable(ObjectFile *o) { 1687 outs() << "Rebase table:\n"; 1688 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1689 printMachORebaseTable(MachO); 1690 else 1691 WithColor::error(errs(), ToolName) 1692 << "This operation is only currently supported " 1693 "for Mach-O executable files.\n"; 1694 } 1695 1696 void llvm::printBindTable(ObjectFile *o) { 1697 outs() << "Bind table:\n"; 1698 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1699 printMachOBindTable(MachO); 1700 else 1701 WithColor::error(errs(), ToolName) 1702 << "This operation is only currently supported " 1703 "for Mach-O executable files.\n"; 1704 } 1705 1706 void llvm::printLazyBindTable(ObjectFile *o) { 1707 outs() << "Lazy bind table:\n"; 1708 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1709 printMachOLazyBindTable(MachO); 1710 else 1711 WithColor::error(errs(), ToolName) 1712 << "This operation is only currently supported " 1713 "for Mach-O executable files.\n"; 1714 } 1715 1716 void llvm::printWeakBindTable(ObjectFile *o) { 1717 outs() << "Weak bind table:\n"; 1718 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1719 printMachOWeakBindTable(MachO); 1720 else 1721 WithColor::error(errs(), ToolName) 1722 << "This operation is only currently supported " 1723 "for Mach-O executable files.\n"; 1724 } 1725 1726 /// Dump the raw contents of the __clangast section so the output can be piped 1727 /// into llvm-bcanalyzer. 1728 void llvm::printRawClangAST(const ObjectFile *Obj) { 1729 if (outs().is_displayed()) { 1730 WithColor::error(errs(), ToolName) 1731 << "The -raw-clang-ast option will dump the raw binary contents of " 1732 "the clang ast section.\n" 1733 "Please redirect the output to a file or another program such as " 1734 "llvm-bcanalyzer.\n"; 1735 return; 1736 } 1737 1738 StringRef ClangASTSectionName("__clangast"); 1739 if (isa<COFFObjectFile>(Obj)) { 1740 ClangASTSectionName = "clangast"; 1741 } 1742 1743 Optional<object::SectionRef> ClangASTSection; 1744 for (auto Sec : ToolSectionFilter(*Obj)) { 1745 StringRef Name; 1746 Sec.getName(Name); 1747 if (Name == ClangASTSectionName) { 1748 ClangASTSection = Sec; 1749 break; 1750 } 1751 } 1752 if (!ClangASTSection) 1753 return; 1754 1755 StringRef ClangASTContents; 1756 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1757 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1758 } 1759 1760 static void printFaultMaps(const ObjectFile *Obj) { 1761 StringRef FaultMapSectionName; 1762 1763 if (isa<ELFObjectFileBase>(Obj)) { 1764 FaultMapSectionName = ".llvm_faultmaps"; 1765 } else if (isa<MachOObjectFile>(Obj)) { 1766 FaultMapSectionName = "__llvm_faultmaps"; 1767 } else { 1768 WithColor::error(errs(), ToolName) 1769 << "This operation is only currently supported " 1770 "for ELF and Mach-O executable files.\n"; 1771 return; 1772 } 1773 1774 Optional<object::SectionRef> FaultMapSection; 1775 1776 for (auto Sec : ToolSectionFilter(*Obj)) { 1777 StringRef Name; 1778 Sec.getName(Name); 1779 if (Name == FaultMapSectionName) { 1780 FaultMapSection = Sec; 1781 break; 1782 } 1783 } 1784 1785 outs() << "FaultMap table:\n"; 1786 1787 if (!FaultMapSection.hasValue()) { 1788 outs() << "<not found>\n"; 1789 return; 1790 } 1791 1792 StringRef FaultMapContents; 1793 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1794 1795 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1796 FaultMapContents.bytes_end()); 1797 1798 outs() << FMP; 1799 } 1800 1801 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1802 if (O->isELF()) { 1803 printELFFileHeader(O); 1804 return printELFDynamicSection(O); 1805 } 1806 if (O->isCOFF()) 1807 return printCOFFFileHeader(O); 1808 if (O->isWasm()) 1809 return printWasmFileHeader(O); 1810 if (O->isMachO()) { 1811 printMachOFileHeader(O); 1812 if (!OnlyFirst) 1813 printMachOLoadCommands(O); 1814 return; 1815 } 1816 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1817 } 1818 1819 static void printFileHeaders(const ObjectFile *O) { 1820 if (!O->isELF() && !O->isCOFF()) 1821 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1822 1823 Triple::ArchType AT = O->getArch(); 1824 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1825 Expected<uint64_t> StartAddrOrErr = O->getStartAddress(); 1826 if (!StartAddrOrErr) 1827 report_error(O->getFileName(), StartAddrOrErr.takeError()); 1828 1829 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1830 uint64_t Address = StartAddrOrErr.get(); 1831 outs() << "start address: " 1832 << "0x" << format(Fmt.data(), Address) << "\n\n"; 1833 } 1834 1835 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 1836 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 1837 if (!ModeOrErr) { 1838 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 1839 consumeError(ModeOrErr.takeError()); 1840 return; 1841 } 1842 sys::fs::perms Mode = ModeOrErr.get(); 1843 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1844 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1845 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1846 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1847 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1848 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1849 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1850 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1851 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1852 1853 outs() << " "; 1854 1855 Expected<unsigned> UIDOrErr = C.getUID(); 1856 if (!UIDOrErr) 1857 report_error(Filename, UIDOrErr.takeError()); 1858 unsigned UID = UIDOrErr.get(); 1859 outs() << format("%d/", UID); 1860 1861 Expected<unsigned> GIDOrErr = C.getGID(); 1862 if (!GIDOrErr) 1863 report_error(Filename, GIDOrErr.takeError()); 1864 unsigned GID = GIDOrErr.get(); 1865 outs() << format("%-d ", GID); 1866 1867 Expected<uint64_t> Size = C.getRawSize(); 1868 if (!Size) 1869 report_error(Filename, Size.takeError()); 1870 outs() << format("%6" PRId64, Size.get()) << " "; 1871 1872 StringRef RawLastModified = C.getRawLastModified(); 1873 unsigned Seconds; 1874 if (RawLastModified.getAsInteger(10, Seconds)) 1875 outs() << "(date: \"" << RawLastModified 1876 << "\" contains non-decimal chars) "; 1877 else { 1878 // Since ctime(3) returns a 26 character string of the form: 1879 // "Sun Sep 16 01:03:52 1973\n\0" 1880 // just print 24 characters. 1881 time_t t = Seconds; 1882 outs() << format("%.24s ", ctime(&t)); 1883 } 1884 1885 StringRef Name = ""; 1886 Expected<StringRef> NameOrErr = C.getName(); 1887 if (!NameOrErr) { 1888 consumeError(NameOrErr.takeError()); 1889 Expected<StringRef> RawNameOrErr = C.getRawName(); 1890 if (!RawNameOrErr) 1891 report_error(Filename, NameOrErr.takeError()); 1892 Name = RawNameOrErr.get(); 1893 } else { 1894 Name = NameOrErr.get(); 1895 } 1896 outs() << Name << "\n"; 1897 } 1898 1899 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 1900 const Archive::Child *C = nullptr) { 1901 // Avoid other output when using a raw option. 1902 if (!RawClangAST) { 1903 outs() << '\n'; 1904 if (A) 1905 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 1906 else 1907 outs() << O->getFileName(); 1908 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 1909 } 1910 1911 StringRef ArchiveName = A ? A->getFileName() : ""; 1912 if (FileHeaders) 1913 printFileHeaders(O); 1914 if (ArchiveHeaders && !MachOOpt && C) 1915 printArchiveChild(ArchiveName, *C); 1916 if (Disassemble) 1917 disassembleObject(O, Relocations); 1918 if (Relocations && !Disassemble) 1919 printRelocations(O); 1920 if (DynamicRelocations) 1921 printDynamicRelocations(O); 1922 if (SectionHeaders) 1923 printSectionHeaders(O); 1924 if (SectionContents) 1925 printSectionContents(O); 1926 if (SymbolTable) 1927 printSymbolTable(O, ArchiveName); 1928 if (UnwindInfo) 1929 printUnwindInfo(O); 1930 if (PrivateHeaders || FirstPrivateHeader) 1931 printPrivateFileHeaders(O, FirstPrivateHeader); 1932 if (ExportsTrie) 1933 printExportsTrie(O); 1934 if (Rebase) 1935 printRebaseTable(O); 1936 if (Bind) 1937 printBindTable(O); 1938 if (LazyBind) 1939 printLazyBindTable(O); 1940 if (WeakBind) 1941 printWeakBindTable(O); 1942 if (RawClangAST) 1943 printRawClangAST(O); 1944 if (PrintFaultMaps) 1945 printFaultMaps(O); 1946 if (DwarfDumpType != DIDT_Null) { 1947 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 1948 // Dump the complete DWARF structure. 1949 DIDumpOptions DumpOpts; 1950 DumpOpts.DumpType = DwarfDumpType; 1951 DICtx->dump(outs(), DumpOpts); 1952 } 1953 } 1954 1955 static void dumpObject(const COFFImportFile *I, const Archive *A, 1956 const Archive::Child *C = nullptr) { 1957 StringRef ArchiveName = A ? A->getFileName() : ""; 1958 1959 // Avoid other output when using a raw option. 1960 if (!RawClangAST) 1961 outs() << '\n' 1962 << ArchiveName << "(" << I->getFileName() << ")" 1963 << ":\tfile format COFF-import-file" 1964 << "\n\n"; 1965 1966 if (ArchiveHeaders && !MachOOpt && C) 1967 printArchiveChild(ArchiveName, *C); 1968 if (SymbolTable) 1969 printCOFFSymbolTable(I); 1970 } 1971 1972 /// Dump each object file in \a a; 1973 static void dumpArchive(const Archive *A) { 1974 Error Err = Error::success(); 1975 for (auto &C : A->children(Err)) { 1976 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1977 if (!ChildOrErr) { 1978 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 1979 report_error(A->getFileName(), C, std::move(E)); 1980 continue; 1981 } 1982 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 1983 dumpObject(O, A, &C); 1984 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 1985 dumpObject(I, A, &C); 1986 else 1987 report_error(A->getFileName(), object_error::invalid_file_type); 1988 } 1989 if (Err) 1990 report_error(A->getFileName(), std::move(Err)); 1991 } 1992 1993 /// Open file and figure out how to dump it. 1994 static void dumpInput(StringRef file) { 1995 // If we are using the Mach-O specific object file parser, then let it parse 1996 // the file and process the command line options. So the -arch flags can 1997 // be used to select specific slices, etc. 1998 if (MachOOpt) { 1999 parseInputMachO(file); 2000 return; 2001 } 2002 2003 // Attempt to open the binary. 2004 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2005 if (!BinaryOrErr) 2006 report_error(file, BinaryOrErr.takeError()); 2007 Binary &Binary = *BinaryOrErr.get().getBinary(); 2008 2009 if (Archive *A = dyn_cast<Archive>(&Binary)) 2010 dumpArchive(A); 2011 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2012 dumpObject(O); 2013 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2014 parseInputMachO(UB); 2015 else 2016 report_error(file, object_error::invalid_file_type); 2017 } 2018 2019 int main(int argc, char **argv) { 2020 InitLLVM X(argc, argv); 2021 2022 // Initialize targets and assembly printers/parsers. 2023 llvm::InitializeAllTargetInfos(); 2024 llvm::InitializeAllTargetMCs(); 2025 llvm::InitializeAllDisassemblers(); 2026 2027 // Register the target printer for --version. 2028 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2029 2030 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2031 2032 ToolName = argv[0]; 2033 2034 // Defaults to a.out if no filenames specified. 2035 if (InputFilenames.empty()) 2036 InputFilenames.push_back("a.out"); 2037 2038 if (AllHeaders) 2039 FileHeaders = PrivateHeaders = Relocations = SectionHeaders = SymbolTable = 2040 true; 2041 2042 if (DisassembleAll || PrintSource || PrintLines) 2043 Disassemble = true; 2044 2045 if (!Disassemble 2046 && !Relocations 2047 && !DynamicRelocations 2048 && !SectionHeaders 2049 && !SectionContents 2050 && !SymbolTable 2051 && !UnwindInfo 2052 && !PrivateHeaders 2053 && !FileHeaders 2054 && !FirstPrivateHeader 2055 && !ExportsTrie 2056 && !Rebase 2057 && !Bind 2058 && !LazyBind 2059 && !WeakBind 2060 && !RawClangAST 2061 && !(UniversalHeaders && MachOOpt) 2062 && !ArchiveHeaders 2063 && !(IndirectSymbols && MachOOpt) 2064 && !(DataInCode && MachOOpt) 2065 && !(LinkOptHints && MachOOpt) 2066 && !(InfoPlist && MachOOpt) 2067 && !(DylibsUsed && MachOOpt) 2068 && !(DylibId && MachOOpt) 2069 && !(ObjcMetaData && MachOOpt) 2070 && !(!FilterSections.empty() && MachOOpt) 2071 && !PrintFaultMaps 2072 && DwarfDumpType == DIDT_Null) { 2073 cl::PrintHelpMessage(); 2074 return 2; 2075 } 2076 2077 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2078 DisassembleFunctions.end()); 2079 2080 llvm::for_each(InputFilenames, dumpInput); 2081 2082 return EXIT_SUCCESS; 2083 } 2084