1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This program is a utility that works like binutils "objdump", that is, it 10 // dumps out a plethora of information about an object file depending on the 11 // flags. 12 // 13 // The flags and output of this program should be near identical to those of 14 // binutils objdump. 15 // 16 //===----------------------------------------------------------------------===// 17 18 #include "llvm-objdump.h" 19 #include "llvm/ADT/Optional.h" 20 #include "llvm/ADT/STLExtras.h" 21 #include "llvm/ADT/StringExtras.h" 22 #include "llvm/ADT/StringSet.h" 23 #include "llvm/ADT/Triple.h" 24 #include "llvm/CodeGen/FaultMaps.h" 25 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 26 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 27 #include "llvm/Demangle/Demangle.h" 28 #include "llvm/MC/MCAsmInfo.h" 29 #include "llvm/MC/MCContext.h" 30 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 31 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 32 #include "llvm/MC/MCInst.h" 33 #include "llvm/MC/MCInstPrinter.h" 34 #include "llvm/MC/MCInstrAnalysis.h" 35 #include "llvm/MC/MCInstrInfo.h" 36 #include "llvm/MC/MCObjectFileInfo.h" 37 #include "llvm/MC/MCRegisterInfo.h" 38 #include "llvm/MC/MCSubtargetInfo.h" 39 #include "llvm/Object/Archive.h" 40 #include "llvm/Object/COFF.h" 41 #include "llvm/Object/COFFImportFile.h" 42 #include "llvm/Object/ELFObjectFile.h" 43 #include "llvm/Object/MachO.h" 44 #include "llvm/Object/MachOUniversal.h" 45 #include "llvm/Object/ObjectFile.h" 46 #include "llvm/Object/Wasm.h" 47 #include "llvm/Support/Casting.h" 48 #include "llvm/Support/CommandLine.h" 49 #include "llvm/Support/Debug.h" 50 #include "llvm/Support/Errc.h" 51 #include "llvm/Support/FileSystem.h" 52 #include "llvm/Support/Format.h" 53 #include "llvm/Support/GraphWriter.h" 54 #include "llvm/Support/Host.h" 55 #include "llvm/Support/InitLLVM.h" 56 #include "llvm/Support/MemoryBuffer.h" 57 #include "llvm/Support/SourceMgr.h" 58 #include "llvm/Support/StringSaver.h" 59 #include "llvm/Support/TargetRegistry.h" 60 #include "llvm/Support/TargetSelect.h" 61 #include "llvm/Support/WithColor.h" 62 #include "llvm/Support/raw_ostream.h" 63 #include <algorithm> 64 #include <cctype> 65 #include <cstring> 66 #include <system_error> 67 #include <unordered_map> 68 #include <utility> 69 70 using namespace llvm; 71 using namespace object; 72 73 cl::opt<unsigned long long> AdjustVMA( 74 "adjust-vma", 75 cl::desc("Increase the displayed address by the specified offset"), 76 cl::value_desc("offset"), cl::init(0)); 77 78 cl::opt<bool> 79 llvm::AllHeaders("all-headers", 80 cl::desc("Display all available header information")); 81 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 82 cl::NotHidden, cl::aliasopt(AllHeaders)); 83 84 static cl::list<std::string> 85 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 86 87 cl::opt<bool> 88 llvm::Disassemble("disassemble", 89 cl::desc("Display assembler mnemonics for the machine instructions")); 90 static cl::alias Disassembled("d", cl::desc("Alias for --disassemble"), 91 cl::NotHidden, cl::aliasopt(Disassemble)); 92 93 cl::opt<bool> 94 llvm::DisassembleAll("disassemble-all", 95 cl::desc("Display assembler mnemonics for the machine instructions")); 96 static cl::alias DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 97 cl::NotHidden, cl::aliasopt(DisassembleAll)); 98 99 cl::opt<bool> llvm::Demangle("demangle", cl::desc("Demangle symbols names"), 100 cl::init(false)); 101 102 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 103 cl::NotHidden, cl::aliasopt(llvm::Demangle)); 104 105 static cl::list<std::string> 106 DisassembleFunctions("df", 107 cl::CommaSeparated, 108 cl::desc("List of functions to disassemble")); 109 static StringSet<> DisasmFuncsSet; 110 111 cl::opt<bool> 112 llvm::Relocations("reloc", 113 cl::desc("Display the relocation entries in the file")); 114 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 115 cl::NotHidden, 116 cl::aliasopt(llvm::Relocations)); 117 118 cl::opt<bool> 119 llvm::DynamicRelocations("dynamic-reloc", 120 cl::desc("Display the dynamic relocation entries in the file")); 121 static cl::alias DynamicRelocationsd("R", cl::desc("Alias for --dynamic-reloc"), 122 cl::NotHidden, 123 cl::aliasopt(DynamicRelocations)); 124 125 cl::opt<bool> 126 llvm::SectionContents("full-contents", 127 cl::desc("Display the content of each section")); 128 static cl::alias SectionContentsShort("s", 129 cl::desc("Alias for --full-contents"), 130 cl::NotHidden, 131 cl::aliasopt(SectionContents)); 132 133 cl::opt<bool> llvm::SymbolTable("syms", cl::desc("Display the symbol table")); 134 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 135 cl::NotHidden, 136 cl::aliasopt(llvm::SymbolTable)); 137 138 cl::opt<bool> 139 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 140 141 cl::opt<bool> 142 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 143 144 cl::opt<bool> 145 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 146 147 cl::opt<bool> 148 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 149 150 cl::opt<bool> 151 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 152 153 cl::opt<bool> 154 llvm::RawClangAST("raw-clang-ast", 155 cl::desc("Dump the raw binary contents of the clang AST section")); 156 157 static cl::opt<bool> 158 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 159 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 160 cl::aliasopt(MachOOpt)); 161 162 cl::opt<std::string> 163 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 164 "see -version for available targets")); 165 166 cl::opt<std::string> 167 llvm::MCPU("mcpu", 168 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 169 cl::value_desc("cpu-name"), 170 cl::init("")); 171 172 cl::opt<std::string> 173 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 174 "see -version for available targets")); 175 176 cl::opt<bool> 177 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 178 "headers for each section.")); 179 static cl::alias SectionHeadersShort("headers", 180 cl::desc("Alias for --section-headers"), 181 cl::NotHidden, 182 cl::aliasopt(SectionHeaders)); 183 static cl::alias SectionHeadersShorter("h", 184 cl::desc("Alias for --section-headers"), 185 cl::NotHidden, 186 cl::aliasopt(SectionHeaders)); 187 188 static cl::opt<bool> 189 ShowLMA("show-lma", 190 cl::desc("Display LMA column when dumping ELF section headers")); 191 192 cl::list<std::string> 193 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 194 "With -macho dump segment,section")); 195 cl::alias static FilterSectionsj("j", cl::desc("Alias for --section"), 196 cl::NotHidden, 197 cl::aliasopt(llvm::FilterSections)); 198 199 cl::list<std::string> 200 llvm::MAttrs("mattr", 201 cl::CommaSeparated, 202 cl::desc("Target specific attributes"), 203 cl::value_desc("a1,+a2,-a3,...")); 204 205 cl::opt<bool> 206 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 207 "instructions, do not print " 208 "the instruction bytes.")); 209 cl::opt<bool> 210 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 211 212 cl::opt<bool> 213 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 214 215 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 216 cl::NotHidden, cl::aliasopt(UnwindInfo)); 217 218 cl::opt<bool> 219 llvm::PrivateHeaders("private-headers", 220 cl::desc("Display format specific file headers")); 221 222 cl::opt<bool> 223 llvm::FirstPrivateHeader("private-header", 224 cl::desc("Display only the first format specific file " 225 "header")); 226 227 static cl::alias PrivateHeadersShort("p", 228 cl::desc("Alias for --private-headers"), 229 cl::NotHidden, 230 cl::aliasopt(PrivateHeaders)); 231 232 cl::opt<bool> llvm::FileHeaders( 233 "file-headers", 234 cl::desc("Display the contents of the overall file header")); 235 236 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 237 cl::NotHidden, cl::aliasopt(FileHeaders)); 238 239 cl::opt<bool> 240 llvm::ArchiveHeaders("archive-headers", 241 cl::desc("Display archive header information")); 242 243 cl::alias ArchiveHeadersShort("a", cl::desc("Alias for --archive-headers"), 244 cl::NotHidden, cl::aliasopt(ArchiveHeaders)); 245 246 cl::opt<bool> 247 llvm::PrintImmHex("print-imm-hex", 248 cl::desc("Use hex format for immediate values")); 249 250 cl::opt<bool> PrintFaultMaps("fault-map-section", 251 cl::desc("Display contents of faultmap section")); 252 253 cl::opt<DIDumpType> llvm::DwarfDumpType( 254 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 255 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 256 257 cl::opt<bool> PrintSource( 258 "source", 259 cl::desc( 260 "Display source inlined with disassembly. Implies disassemble object")); 261 262 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), cl::NotHidden, 263 cl::aliasopt(PrintSource)); 264 265 cl::opt<bool> PrintLines("line-numbers", 266 cl::desc("Display source line numbers with " 267 "disassembly. Implies disassemble object")); 268 269 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 270 cl::NotHidden, cl::aliasopt(PrintLines)); 271 272 cl::opt<unsigned long long> 273 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 274 cl::value_desc("address"), cl::init(0)); 275 cl::opt<unsigned long long> 276 StopAddress("stop-address", 277 cl::desc("Stop disassembly at address"), 278 cl::value_desc("address"), cl::init(UINT64_MAX)); 279 280 cl::opt<bool> DisassembleZeroes( 281 "disassemble-zeroes", 282 cl::desc("Do not skip blocks of zeroes when disassembling")); 283 cl::alias DisassembleZeroesShort("z", 284 cl::desc("Alias for --disassemble-zeroes"), 285 cl::NotHidden, 286 cl::aliasopt(DisassembleZeroes)); 287 288 static StringRef ToolName; 289 290 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 291 292 SectionFilter llvm::ToolSectionFilter(llvm::object::ObjectFile const &O) { 293 return SectionFilter( 294 [](llvm::object::SectionRef const &S) { 295 if (FilterSections.empty()) 296 return true; 297 llvm::StringRef String; 298 std::error_code error = S.getName(String); 299 if (error) 300 return false; 301 return is_contained(FilterSections, String); 302 }, 303 O); 304 } 305 306 void llvm::error(std::error_code EC) { 307 if (!EC) 308 return; 309 WithColor::error(errs(), ToolName) 310 << "reading file: " << EC.message() << ".\n"; 311 errs().flush(); 312 exit(1); 313 } 314 315 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 316 WithColor::error(errs(), ToolName) << Message << ".\n"; 317 errs().flush(); 318 exit(1); 319 } 320 321 void llvm::warn(StringRef Message) { 322 WithColor::warning(errs(), ToolName) << Message << ".\n"; 323 errs().flush(); 324 } 325 326 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 327 Twine Message) { 328 WithColor::error(errs(), ToolName) 329 << "'" << File << "': " << Message << ".\n"; 330 exit(1); 331 } 332 333 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 334 std::error_code EC) { 335 assert(EC); 336 WithColor::error(errs(), ToolName) 337 << "'" << File << "': " << EC.message() << ".\n"; 338 exit(1); 339 } 340 341 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 342 llvm::Error E) { 343 assert(E); 344 std::string Buf; 345 raw_string_ostream OS(Buf); 346 logAllUnhandledErrors(std::move(E), OS); 347 OS.flush(); 348 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 349 exit(1); 350 } 351 352 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 353 StringRef FileName, 354 llvm::Error E, 355 StringRef ArchitectureName) { 356 assert(E); 357 WithColor::error(errs(), ToolName); 358 if (ArchiveName != "") 359 errs() << ArchiveName << "(" << FileName << ")"; 360 else 361 errs() << "'" << FileName << "'"; 362 if (!ArchitectureName.empty()) 363 errs() << " (for architecture " << ArchitectureName << ")"; 364 std::string Buf; 365 raw_string_ostream OS(Buf); 366 logAllUnhandledErrors(std::move(E), OS); 367 OS.flush(); 368 errs() << ": " << Buf; 369 exit(1); 370 } 371 372 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 373 const object::Archive::Child &C, 374 llvm::Error E, 375 StringRef ArchitectureName) { 376 Expected<StringRef> NameOrErr = C.getName(); 377 // TODO: if we have a error getting the name then it would be nice to print 378 // the index of which archive member this is and or its offset in the 379 // archive instead of "???" as the name. 380 if (!NameOrErr) { 381 consumeError(NameOrErr.takeError()); 382 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 383 } else 384 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 385 ArchitectureName); 386 } 387 388 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 389 // Figure out the target triple. 390 llvm::Triple TheTriple("unknown-unknown-unknown"); 391 if (TripleName.empty()) { 392 if (Obj) 393 TheTriple = Obj->makeTriple(); 394 } else { 395 TheTriple.setTriple(Triple::normalize(TripleName)); 396 397 // Use the triple, but also try to combine with ARM build attributes. 398 if (Obj) { 399 auto Arch = Obj->getArch(); 400 if (Arch == Triple::arm || Arch == Triple::armeb) 401 Obj->setARMSubArch(TheTriple); 402 } 403 } 404 405 // Get the target specific parser. 406 std::string Error; 407 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 408 Error); 409 if (!TheTarget) { 410 if (Obj) 411 report_error(Obj->getFileName(), "can't find target: " + Error); 412 else 413 error("can't find target: " + Error); 414 } 415 416 // Update the triple name and return the found target. 417 TripleName = TheTriple.getTriple(); 418 return TheTarget; 419 } 420 421 bool llvm::isRelocAddressLess(RelocationRef A, RelocationRef B) { 422 return A.getOffset() < B.getOffset(); 423 } 424 425 static std::error_code getRelocationValueString(const RelocationRef &Rel, 426 SmallVectorImpl<char> &Result) { 427 const ObjectFile *Obj = Rel.getObject(); 428 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 429 return getELFRelocationValueString(ELF, Rel, Result); 430 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 431 return getCOFFRelocationValueString(COFF, Rel, Result); 432 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 433 return getWasmRelocationValueString(Wasm, Rel, Result); 434 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 435 return getMachORelocationValueString(MachO, Rel, Result); 436 llvm_unreachable("unknown object file format"); 437 } 438 439 /// Indicates whether this relocation should hidden when listing 440 /// relocations, usually because it is the trailing part of a multipart 441 /// relocation that will be printed as part of the leading relocation. 442 static bool getHidden(RelocationRef RelRef) { 443 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 444 if (!MachO) 445 return false; 446 447 unsigned Arch = MachO->getArch(); 448 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 449 uint64_t Type = MachO->getRelocationType(Rel); 450 451 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 452 // is always hidden. 453 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 454 return Type == MachO::GENERIC_RELOC_PAIR; 455 456 if (Arch == Triple::x86_64) { 457 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 458 // an X86_64_RELOC_SUBTRACTOR. 459 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 460 DataRefImpl RelPrev = Rel; 461 RelPrev.d.a--; 462 uint64_t PrevType = MachO->getRelocationType(RelPrev); 463 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 464 return true; 465 } 466 } 467 468 return false; 469 } 470 471 namespace { 472 class SourcePrinter { 473 protected: 474 DILineInfo OldLineInfo; 475 const ObjectFile *Obj = nullptr; 476 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 477 // File name to file contents of source 478 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 479 // Mark the line endings of the cached source 480 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 481 482 private: 483 bool cacheSource(const DILineInfo& LineInfoFile); 484 485 public: 486 SourcePrinter() = default; 487 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 488 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 489 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 490 DefaultArch); 491 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 492 } 493 virtual ~SourcePrinter() = default; 494 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 495 StringRef Delimiter = "; "); 496 }; 497 498 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 499 std::unique_ptr<MemoryBuffer> Buffer; 500 if (LineInfo.Source) { 501 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 502 } else { 503 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 504 if (!BufferOrError) 505 return false; 506 Buffer = std::move(*BufferOrError); 507 } 508 // Chomp the file to get lines 509 size_t BufferSize = Buffer->getBufferSize(); 510 const char *BufferStart = Buffer->getBufferStart(); 511 for (const char *Start = BufferStart, *End = BufferStart; 512 End < BufferStart + BufferSize; End++) 513 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 514 (*End == '\r' && *(End + 1) == '\n')) { 515 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 516 if (*End == '\r') 517 End++; 518 Start = End + 1; 519 } 520 SourceCache[LineInfo.FileName] = std::move(Buffer); 521 return true; 522 } 523 524 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 525 StringRef Delimiter) { 526 if (!Symbolizer) 527 return; 528 DILineInfo LineInfo = DILineInfo(); 529 auto ExpectecLineInfo = 530 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 531 if (!ExpectecLineInfo) 532 consumeError(ExpectecLineInfo.takeError()); 533 else 534 LineInfo = *ExpectecLineInfo; 535 536 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 537 LineInfo.Line == 0) 538 return; 539 540 if (PrintLines) 541 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 542 if (PrintSource) { 543 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 544 if (!cacheSource(LineInfo)) 545 return; 546 auto FileBuffer = SourceCache.find(LineInfo.FileName); 547 if (FileBuffer != SourceCache.end()) { 548 auto LineBuffer = LineCache.find(LineInfo.FileName); 549 if (LineBuffer != LineCache.end()) { 550 if (LineInfo.Line > LineBuffer->second.size()) 551 return; 552 // Vector begins at 0, line numbers are non-zero 553 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 554 << "\n"; 555 } 556 } 557 } 558 OldLineInfo = LineInfo; 559 } 560 561 static bool isArmElf(const ObjectFile *Obj) { 562 return (Obj->isELF() && 563 (Obj->getArch() == Triple::aarch64 || 564 Obj->getArch() == Triple::aarch64_be || 565 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 566 Obj->getArch() == Triple::thumb || 567 Obj->getArch() == Triple::thumbeb)); 568 } 569 570 static void printRelocation(const RelocationRef &Rel, uint64_t Address, 571 uint8_t AddrSize) { 572 StringRef Fmt = 573 AddrSize > 4 ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 574 SmallString<16> Name; 575 SmallString<32> Val; 576 Rel.getTypeName(Name); 577 error(getRelocationValueString(Rel, Val)); 578 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n"; 579 } 580 581 class PrettyPrinter { 582 public: 583 virtual ~PrettyPrinter() = default; 584 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 585 ArrayRef<uint8_t> Bytes, uint64_t Address, 586 raw_ostream &OS, StringRef Annot, 587 MCSubtargetInfo const &STI, SourcePrinter *SP, 588 std::vector<RelocationRef> *Rels = nullptr) { 589 if (SP && (PrintSource || PrintLines)) 590 SP->printSourceLine(OS, Address); 591 if (!NoLeadingAddr) 592 OS << format("%8" PRIx64 ":", Address); 593 if (!NoShowRawInsn) { 594 OS << "\t"; 595 dumpBytes(Bytes, OS); 596 } 597 if (MI) 598 IP.printInst(MI, OS, "", STI); 599 else 600 OS << " <unknown>"; 601 } 602 }; 603 PrettyPrinter PrettyPrinterInst; 604 class HexagonPrettyPrinter : public PrettyPrinter { 605 public: 606 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 607 raw_ostream &OS) { 608 uint32_t opcode = 609 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 610 if (!NoLeadingAddr) 611 OS << format("%8" PRIx64 ":", Address); 612 if (!NoShowRawInsn) { 613 OS << "\t"; 614 dumpBytes(Bytes.slice(0, 4), OS); 615 OS << format("%08" PRIx32, opcode); 616 } 617 } 618 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 619 uint64_t Address, raw_ostream &OS, StringRef Annot, 620 MCSubtargetInfo const &STI, SourcePrinter *SP, 621 std::vector<RelocationRef> *Rels) override { 622 if (SP && (PrintSource || PrintLines)) 623 SP->printSourceLine(OS, Address, ""); 624 if (!MI) { 625 printLead(Bytes, Address, OS); 626 OS << " <unknown>"; 627 return; 628 } 629 std::string Buffer; 630 { 631 raw_string_ostream TempStream(Buffer); 632 IP.printInst(MI, TempStream, "", STI); 633 } 634 StringRef Contents(Buffer); 635 // Split off bundle attributes 636 auto PacketBundle = Contents.rsplit('\n'); 637 // Split off first instruction from the rest 638 auto HeadTail = PacketBundle.first.split('\n'); 639 auto Preamble = " { "; 640 auto Separator = ""; 641 642 // Hexagon's packets require relocations to be inline rather than 643 // clustered at the end of the packet. 644 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 645 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 646 auto PrintReloc = [&]() -> void { 647 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address)) { 648 if (RelCur->getOffset() == Address) { 649 printRelocation(*RelCur, Address, 4); 650 return; 651 } 652 ++RelCur; 653 } 654 }; 655 656 while (!HeadTail.first.empty()) { 657 OS << Separator; 658 Separator = "\n"; 659 if (SP && (PrintSource || PrintLines)) 660 SP->printSourceLine(OS, Address, ""); 661 printLead(Bytes, Address, OS); 662 OS << Preamble; 663 Preamble = " "; 664 StringRef Inst; 665 auto Duplex = HeadTail.first.split('\v'); 666 if (!Duplex.second.empty()) { 667 OS << Duplex.first; 668 OS << "; "; 669 Inst = Duplex.second; 670 } 671 else 672 Inst = HeadTail.first; 673 OS << Inst; 674 HeadTail = HeadTail.second.split('\n'); 675 if (HeadTail.first.empty()) 676 OS << " } " << PacketBundle.second; 677 PrintReloc(); 678 Bytes = Bytes.slice(4); 679 Address += 4; 680 } 681 } 682 }; 683 HexagonPrettyPrinter HexagonPrettyPrinterInst; 684 685 class AMDGCNPrettyPrinter : public PrettyPrinter { 686 public: 687 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 688 uint64_t Address, raw_ostream &OS, StringRef Annot, 689 MCSubtargetInfo const &STI, SourcePrinter *SP, 690 std::vector<RelocationRef> *Rels) override { 691 if (SP && (PrintSource || PrintLines)) 692 SP->printSourceLine(OS, Address); 693 694 typedef support::ulittle32_t U32; 695 696 if (MI) { 697 SmallString<40> InstStr; 698 raw_svector_ostream IS(InstStr); 699 700 IP.printInst(MI, IS, "", STI); 701 702 OS << left_justify(IS.str(), 60); 703 } else { 704 // an unrecognized encoding - this is probably data so represent it 705 // using the .long directive, or .byte directive if fewer than 4 bytes 706 // remaining 707 if (Bytes.size() >= 4) { 708 OS << format("\t.long 0x%08" PRIx32 " ", 709 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 710 OS.indent(42); 711 } else { 712 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 713 for (unsigned int i = 1; i < Bytes.size(); i++) 714 OS << format(", 0x%02" PRIx8, Bytes[i]); 715 OS.indent(55 - (6 * Bytes.size())); 716 } 717 } 718 719 OS << format("// %012" PRIX64 ": ", Address); 720 if (Bytes.size() >=4) { 721 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 722 Bytes.size() / sizeof(U32))) 723 // D should be explicitly casted to uint32_t here as it is passed 724 // by format to snprintf as vararg. 725 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 726 } else { 727 for (unsigned int i = 0; i < Bytes.size(); i++) 728 OS << format("%02" PRIX8 " ", Bytes[i]); 729 } 730 731 if (!Annot.empty()) 732 OS << "// " << Annot; 733 } 734 }; 735 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 736 737 class BPFPrettyPrinter : public PrettyPrinter { 738 public: 739 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 740 uint64_t Address, raw_ostream &OS, StringRef Annot, 741 MCSubtargetInfo const &STI, SourcePrinter *SP, 742 std::vector<RelocationRef> *Rels) override { 743 if (SP && (PrintSource || PrintLines)) 744 SP->printSourceLine(OS, Address); 745 if (!NoLeadingAddr) 746 OS << format("%8" PRId64 ":", Address / 8); 747 if (!NoShowRawInsn) { 748 OS << "\t"; 749 dumpBytes(Bytes, OS); 750 } 751 if (MI) 752 IP.printInst(MI, OS, "", STI); 753 else 754 OS << " <unknown>"; 755 } 756 }; 757 BPFPrettyPrinter BPFPrettyPrinterInst; 758 759 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 760 switch(Triple.getArch()) { 761 default: 762 return PrettyPrinterInst; 763 case Triple::hexagon: 764 return HexagonPrettyPrinterInst; 765 case Triple::amdgcn: 766 return AMDGCNPrettyPrinterInst; 767 case Triple::bpfel: 768 case Triple::bpfeb: 769 return BPFPrettyPrinterInst; 770 } 771 } 772 } 773 774 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 775 assert(Obj->isELF()); 776 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 777 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 778 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 779 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 780 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 781 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 782 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 783 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 784 llvm_unreachable("Unsupported binary format"); 785 } 786 787 template <class ELFT> static void 788 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 789 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 790 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 791 uint8_t SymbolType = Symbol.getELFType(); 792 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 793 continue; 794 795 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 796 if (!AddressOrErr) 797 report_error(Obj->getFileName(), AddressOrErr.takeError()); 798 799 Expected<StringRef> Name = Symbol.getName(); 800 if (!Name) 801 report_error(Obj->getFileName(), Name.takeError()); 802 if (Name->empty()) 803 continue; 804 805 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 806 if (!SectionOrErr) 807 report_error(Obj->getFileName(), SectionOrErr.takeError()); 808 section_iterator SecI = *SectionOrErr; 809 if (SecI == Obj->section_end()) 810 continue; 811 812 AllSymbols[*SecI].emplace_back(*AddressOrErr, *Name, SymbolType); 813 } 814 } 815 816 static void 817 addDynamicElfSymbols(const ObjectFile *Obj, 818 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 819 assert(Obj->isELF()); 820 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 821 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 822 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 823 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 824 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 825 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 826 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 827 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 828 else 829 llvm_unreachable("Unsupported binary format"); 830 } 831 832 static void addPltEntries(const ObjectFile *Obj, 833 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 834 StringSaver &Saver) { 835 Optional<SectionRef> Plt = None; 836 for (const SectionRef &Section : Obj->sections()) { 837 StringRef Name; 838 if (Section.getName(Name)) 839 continue; 840 if (Name == ".plt") 841 Plt = Section; 842 } 843 if (!Plt) 844 return; 845 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 846 for (auto PltEntry : ElfObj->getPltAddresses()) { 847 SymbolRef Symbol(PltEntry.first, ElfObj); 848 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 849 850 Expected<StringRef> NameOrErr = Symbol.getName(); 851 if (!NameOrErr) 852 report_error(Obj->getFileName(), NameOrErr.takeError()); 853 if (NameOrErr->empty()) 854 continue; 855 StringRef Name = Saver.save((*NameOrErr + "@plt").str()); 856 857 AllSymbols[*Plt].emplace_back(PltEntry.second, Name, SymbolType); 858 } 859 } 860 } 861 862 // Normally the disassembly output will skip blocks of zeroes. This function 863 // returns the number of zero bytes that can be skipped when dumping the 864 // disassembly of the instructions in Buf. 865 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 866 // Find the number of leading zeroes. 867 size_t N = 0; 868 while (N < Buf.size() && !Buf[N]) 869 ++N; 870 871 // We may want to skip blocks of zero bytes, but unless we see 872 // at least 8 of them in a row. 873 if (N < 8) 874 return 0; 875 876 // We skip zeroes in multiples of 4 because do not want to truncate an 877 // instruction if it starts with a zero byte. 878 return N & ~0x3; 879 } 880 881 // Returns a map from sections to their relocations. 882 static std::map<SectionRef, std::vector<RelocationRef>> 883 getRelocsMap(llvm::object::ObjectFile const &Obj) { 884 std::map<SectionRef, std::vector<RelocationRef>> Ret; 885 for (const SectionRef &Section : ToolSectionFilter(Obj)) { 886 section_iterator RelSec = Section.getRelocatedSection(); 887 if (RelSec == Obj.section_end()) 888 continue; 889 std::vector<RelocationRef> &V = Ret[*RelSec]; 890 for (const RelocationRef &R : Section.relocations()) 891 V.push_back(R); 892 // Sort relocations by address. 893 llvm::sort(V, isRelocAddressLess); 894 } 895 return Ret; 896 } 897 898 // Used for --adjust-vma to check if address should be adjusted by the 899 // specified value for a given section. 900 // For ELF we do not adjust non-allocatable sections like debug ones, 901 // because they are not loadable. 902 // TODO: implement for other file formats. 903 static bool shouldAdjustVA(const SectionRef &Section) { 904 const ObjectFile *Obj = Section.getObject(); 905 if (isa<object::ELFObjectFileBase>(Obj)) 906 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 907 return false; 908 } 909 910 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 911 MCContext &Ctx, MCDisassembler *DisAsm, 912 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 913 const MCSubtargetInfo *STI, PrettyPrinter &PIP, 914 SourcePrinter &SP, bool InlineRelocs) { 915 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 916 if (InlineRelocs) 917 RelocMap = getRelocsMap(*Obj); 918 919 // Create a mapping from virtual address to symbol name. This is used to 920 // pretty print the symbols while disassembling. 921 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 922 SectionSymbolsTy AbsoluteSymbols; 923 for (const SymbolRef &Symbol : Obj->symbols()) { 924 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 925 if (!AddressOrErr) 926 report_error(Obj->getFileName(), AddressOrErr.takeError()); 927 uint64_t Address = *AddressOrErr; 928 929 Expected<StringRef> Name = Symbol.getName(); 930 if (!Name) 931 report_error(Obj->getFileName(), Name.takeError()); 932 if (Name->empty()) 933 continue; 934 935 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 936 if (!SectionOrErr) 937 report_error(Obj->getFileName(), SectionOrErr.takeError()); 938 939 uint8_t SymbolType = ELF::STT_NOTYPE; 940 if (Obj->isELF()) { 941 SymbolType = getElfSymbolType(Obj, Symbol); 942 if (SymbolType == ELF::STT_SECTION) 943 continue; 944 } 945 946 section_iterator SecI = *SectionOrErr; 947 if (SecI != Obj->section_end()) 948 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 949 else 950 AbsoluteSymbols.emplace_back(Address, *Name, SymbolType); 951 952 953 } 954 if (AllSymbols.empty() && Obj->isELF()) 955 addDynamicElfSymbols(Obj, AllSymbols); 956 957 BumpPtrAllocator A; 958 StringSaver Saver(A); 959 addPltEntries(Obj, AllSymbols, Saver); 960 961 // Create a mapping from virtual address to section. 962 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 963 for (SectionRef Sec : Obj->sections()) 964 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 965 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 966 967 // Linked executables (.exe and .dll files) typically don't include a real 968 // symbol table but they might contain an export table. 969 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 970 for (const auto &ExportEntry : COFFObj->export_directories()) { 971 StringRef Name; 972 error(ExportEntry.getSymbolName(Name)); 973 if (Name.empty()) 974 continue; 975 uint32_t RVA; 976 error(ExportEntry.getExportRVA(RVA)); 977 978 uint64_t VA = COFFObj->getImageBase() + RVA; 979 auto Sec = std::upper_bound( 980 SectionAddresses.begin(), SectionAddresses.end(), VA, 981 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 982 return LHS < RHS.first; 983 }); 984 if (Sec != SectionAddresses.begin()) 985 --Sec; 986 else 987 Sec = SectionAddresses.end(); 988 989 if (Sec != SectionAddresses.end()) 990 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 991 else 992 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 993 } 994 } 995 996 // Sort all the symbols, this allows us to use a simple binary search to find 997 // a symbol near an address. 998 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 999 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1000 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1001 1002 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1003 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1004 continue; 1005 1006 uint64_t SectionAddr = Section.getAddress(); 1007 uint64_t SectSize = Section.getSize(); 1008 if (!SectSize) 1009 continue; 1010 1011 // Get the list of all the symbols in this section. 1012 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1013 std::vector<uint64_t> DataMappingSymsAddr; 1014 std::vector<uint64_t> TextMappingSymsAddr; 1015 if (isArmElf(Obj)) { 1016 for (const auto &Symb : Symbols) { 1017 uint64_t Address = std::get<0>(Symb); 1018 StringRef Name = std::get<1>(Symb); 1019 if (Name.startswith("$d")) 1020 DataMappingSymsAddr.push_back(Address - SectionAddr); 1021 if (Name.startswith("$x")) 1022 TextMappingSymsAddr.push_back(Address - SectionAddr); 1023 if (Name.startswith("$a")) 1024 TextMappingSymsAddr.push_back(Address - SectionAddr); 1025 if (Name.startswith("$t")) 1026 TextMappingSymsAddr.push_back(Address - SectionAddr); 1027 } 1028 } 1029 1030 llvm::sort(DataMappingSymsAddr); 1031 llvm::sort(TextMappingSymsAddr); 1032 1033 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1034 // AMDGPU disassembler uses symbolizer for printing labels 1035 std::unique_ptr<MCRelocationInfo> RelInfo( 1036 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1037 if (RelInfo) { 1038 std::unique_ptr<MCSymbolizer> Symbolizer( 1039 TheTarget->createMCSymbolizer( 1040 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1041 DisAsm->setSymbolizer(std::move(Symbolizer)); 1042 } 1043 } 1044 1045 StringRef SegmentName = ""; 1046 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1047 DataRefImpl DR = Section.getRawDataRefImpl(); 1048 SegmentName = MachO->getSectionFinalSegmentName(DR); 1049 } 1050 StringRef SectionName; 1051 error(Section.getName(SectionName)); 1052 1053 // If the section has no symbol at the start, just insert a dummy one. 1054 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1055 Symbols.insert( 1056 Symbols.begin(), 1057 std::make_tuple(SectionAddr, SectionName, 1058 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1059 } 1060 1061 SmallString<40> Comments; 1062 raw_svector_ostream CommentStream(Comments); 1063 1064 StringRef BytesStr; 1065 error(Section.getContents(BytesStr)); 1066 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1067 BytesStr.size()); 1068 1069 uint64_t VMAAdjustment = 0; 1070 if (shouldAdjustVA(Section)) 1071 VMAAdjustment = AdjustVMA; 1072 1073 uint64_t Size; 1074 uint64_t Index; 1075 bool PrintedSection = false; 1076 std::vector<RelocationRef> Rels = RelocMap[Section]; 1077 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1078 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1079 // Disassemble symbol by symbol. 1080 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1081 uint64_t Start = std::get<0>(Symbols[SI]) - SectionAddr; 1082 // The end is either the section end or the beginning of the next 1083 // symbol. 1084 uint64_t End = (SI == SE - 1) 1085 ? SectSize 1086 : std::get<0>(Symbols[SI + 1]) - SectionAddr; 1087 // Don't try to disassemble beyond the end of section contents. 1088 if (End > SectSize) 1089 End = SectSize; 1090 // If this symbol has the same address as the next symbol, then skip it. 1091 if (Start >= End) 1092 continue; 1093 1094 // Check if we need to skip symbol 1095 // Skip if the symbol's data is not between StartAddress and StopAddress 1096 if (End + SectionAddr < StartAddress || 1097 Start + SectionAddr > StopAddress) { 1098 continue; 1099 } 1100 1101 /// Skip if user requested specific symbols and this is not in the list 1102 if (!DisasmFuncsSet.empty() && 1103 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1104 continue; 1105 1106 if (!PrintedSection) { 1107 PrintedSection = true; 1108 outs() << "Disassembly of section "; 1109 if (!SegmentName.empty()) 1110 outs() << SegmentName << ","; 1111 outs() << SectionName << ':'; 1112 } 1113 1114 // Stop disassembly at the stop address specified 1115 if (End + SectionAddr > StopAddress) 1116 End = StopAddress - SectionAddr; 1117 1118 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1119 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1120 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1121 Start += 256; 1122 } 1123 if (SI == SE - 1 || 1124 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1125 // cut trailing zeroes at the end of kernel 1126 // cut up to 256 bytes 1127 const uint64_t EndAlign = 256; 1128 const auto Limit = End - (std::min)(EndAlign, End - Start); 1129 while (End > Limit && 1130 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1131 End -= 4; 1132 } 1133 } 1134 1135 outs() << '\n'; 1136 if (!NoLeadingAddr) 1137 outs() << format("%016" PRIx64 " ", 1138 SectionAddr + Start + VMAAdjustment); 1139 1140 StringRef SymbolName = std::get<1>(Symbols[SI]); 1141 if (Demangle) 1142 outs() << demangle(SymbolName) << ":\n"; 1143 else 1144 outs() << SymbolName << ":\n"; 1145 1146 // Don't print raw contents of a virtual section. A virtual section 1147 // doesn't have any contents in the file. 1148 if (Section.isVirtual()) { 1149 outs() << "...\n"; 1150 continue; 1151 } 1152 1153 #ifndef NDEBUG 1154 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1155 #else 1156 raw_ostream &DebugOut = nulls(); 1157 #endif 1158 1159 // Some targets (like WebAssembly) have a special prelude at the start 1160 // of each symbol. 1161 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1162 SectionAddr + Start, DebugOut, CommentStream); 1163 Start += Size; 1164 1165 for (Index = Start; Index < End; Index += Size) { 1166 MCInst Inst; 1167 1168 if (Index + SectionAddr < StartAddress || 1169 Index + SectionAddr > StopAddress) { 1170 // skip byte by byte till StartAddress is reached 1171 Size = 1; 1172 continue; 1173 } 1174 // AArch64 ELF binaries can interleave data and text in the 1175 // same section. We rely on the markers introduced to 1176 // understand what we need to dump. If the data marker is within a 1177 // function, it is denoted as a word/short etc 1178 if (isArmElf(Obj) && std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1179 !DisassembleAll) { 1180 uint64_t Stride = 0; 1181 1182 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1183 DataMappingSymsAddr.end(), Index); 1184 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1185 // Switch to data. 1186 while (Index < End) { 1187 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1188 outs() << "\t"; 1189 if (Index + 4 <= End) { 1190 Stride = 4; 1191 dumpBytes(Bytes.slice(Index, 4), outs()); 1192 outs() << "\t.word\t"; 1193 uint32_t Data = 0; 1194 if (Obj->isLittleEndian()) { 1195 const auto Word = 1196 reinterpret_cast<const support::ulittle32_t *>( 1197 Bytes.data() + Index); 1198 Data = *Word; 1199 } else { 1200 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1201 Bytes.data() + Index); 1202 Data = *Word; 1203 } 1204 outs() << "0x" << format("%08" PRIx32, Data); 1205 } else if (Index + 2 <= End) { 1206 Stride = 2; 1207 dumpBytes(Bytes.slice(Index, 2), outs()); 1208 outs() << "\t\t.short\t"; 1209 uint16_t Data = 0; 1210 if (Obj->isLittleEndian()) { 1211 const auto Short = 1212 reinterpret_cast<const support::ulittle16_t *>( 1213 Bytes.data() + Index); 1214 Data = *Short; 1215 } else { 1216 const auto Short = 1217 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1218 Index); 1219 Data = *Short; 1220 } 1221 outs() << "0x" << format("%04" PRIx16, Data); 1222 } else { 1223 Stride = 1; 1224 dumpBytes(Bytes.slice(Index, 1), outs()); 1225 outs() << "\t\t.byte\t"; 1226 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1227 } 1228 Index += Stride; 1229 outs() << "\n"; 1230 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1231 TextMappingSymsAddr.end(), Index); 1232 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1233 break; 1234 } 1235 } 1236 } 1237 1238 // If there is a data symbol inside an ELF text section and we are only 1239 // disassembling text (applicable all architectures), 1240 // we are in a situation where we must print the data and not 1241 // disassemble it. 1242 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1243 !DisassembleAll && Section.isText()) { 1244 // print out data up to 8 bytes at a time in hex and ascii 1245 uint8_t AsciiData[9] = {'\0'}; 1246 uint8_t Byte; 1247 int NumBytes = 0; 1248 1249 for (Index = Start; Index < End; Index += 1) { 1250 if (((SectionAddr + Index) < StartAddress) || 1251 ((SectionAddr + Index) > StopAddress)) 1252 continue; 1253 if (NumBytes == 0) { 1254 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1255 outs() << "\t"; 1256 } 1257 Byte = Bytes.slice(Index)[0]; 1258 outs() << format(" %02x", Byte); 1259 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1260 1261 uint8_t IndentOffset = 0; 1262 NumBytes++; 1263 if (Index == End - 1 || NumBytes > 8) { 1264 // Indent the space for less than 8 bytes data. 1265 // 2 spaces for byte and one for space between bytes 1266 IndentOffset = 3 * (8 - NumBytes); 1267 for (int Excess = NumBytes; Excess < 8; Excess++) 1268 AsciiData[Excess] = '\0'; 1269 NumBytes = 8; 1270 } 1271 if (NumBytes == 8) { 1272 AsciiData[8] = '\0'; 1273 outs() << std::string(IndentOffset, ' ') << " "; 1274 outs() << reinterpret_cast<char *>(AsciiData); 1275 outs() << '\n'; 1276 NumBytes = 0; 1277 } 1278 } 1279 } 1280 if (Index >= End) 1281 break; 1282 1283 // When -z or --disassemble-zeroes are given we always dissasemble them. 1284 // Otherwise we might want to skip zero bytes we see. 1285 if (!DisassembleZeroes) { 1286 uint64_t MaxOffset = End - Index; 1287 // For -reloc: print zero blocks patched by relocations, so that 1288 // relocations can be shown in the dump. 1289 if (RelCur != RelEnd) 1290 MaxOffset = RelCur->getOffset() - Index; 1291 1292 if (size_t N = 1293 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1294 outs() << "\t\t..." << '\n'; 1295 Index += N; 1296 if (Index >= End) 1297 break; 1298 } 1299 } 1300 1301 // Disassemble a real instruction or a data when disassemble all is 1302 // provided 1303 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1304 SectionAddr + Index, DebugOut, 1305 CommentStream); 1306 if (Size == 0) 1307 Size = 1; 1308 1309 PIP.printInst( 1310 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 1311 SectionAddr + Index + VMAAdjustment, outs(), "", *STI, &SP, &Rels); 1312 outs() << CommentStream.str(); 1313 Comments.clear(); 1314 1315 // Try to resolve the target of a call, tail call, etc. to a specific 1316 // symbol. 1317 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1318 MIA->isConditionalBranch(Inst))) { 1319 uint64_t Target; 1320 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1321 // In a relocatable object, the target's section must reside in 1322 // the same section as the call instruction or it is accessed 1323 // through a relocation. 1324 // 1325 // In a non-relocatable object, the target may be in any section. 1326 // 1327 // N.B. We don't walk the relocations in the relocatable case yet. 1328 auto *TargetSectionSymbols = &Symbols; 1329 if (!Obj->isRelocatableObject()) { 1330 auto SectionAddress = std::upper_bound( 1331 SectionAddresses.begin(), SectionAddresses.end(), Target, 1332 [](uint64_t LHS, 1333 const std::pair<uint64_t, SectionRef> &RHS) { 1334 return LHS < RHS.first; 1335 }); 1336 if (SectionAddress != SectionAddresses.begin()) { 1337 --SectionAddress; 1338 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1339 } else { 1340 TargetSectionSymbols = &AbsoluteSymbols; 1341 } 1342 } 1343 1344 // Find the first symbol in the section whose offset is less than 1345 // or equal to the target. If there isn't a section that contains 1346 // the target, find the nearest preceding absolute symbol. 1347 auto TargetSym = std::upper_bound( 1348 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1349 Target, [](uint64_t LHS, 1350 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1351 return LHS < std::get<0>(RHS); 1352 }); 1353 if (TargetSym == TargetSectionSymbols->begin()) { 1354 TargetSectionSymbols = &AbsoluteSymbols; 1355 TargetSym = std::upper_bound( 1356 AbsoluteSymbols.begin(), AbsoluteSymbols.end(), 1357 Target, [](uint64_t LHS, 1358 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1359 return LHS < std::get<0>(RHS); 1360 }); 1361 } 1362 if (TargetSym != TargetSectionSymbols->begin()) { 1363 --TargetSym; 1364 uint64_t TargetAddress = std::get<0>(*TargetSym); 1365 StringRef TargetName = std::get<1>(*TargetSym); 1366 outs() << " <" << TargetName; 1367 uint64_t Disp = Target - TargetAddress; 1368 if (Disp) 1369 outs() << "+0x" << Twine::utohexstr(Disp); 1370 outs() << '>'; 1371 } 1372 } 1373 } 1374 outs() << "\n"; 1375 1376 // Hexagon does this in pretty printer 1377 if (Obj->getArch() != Triple::hexagon) { 1378 // Print relocation for instruction. 1379 while (RelCur != RelEnd) { 1380 uint64_t Offset = RelCur->getOffset(); 1381 // If this relocation is hidden, skip it. 1382 if (getHidden(*RelCur) || ((SectionAddr + Offset) < StartAddress)) { 1383 ++RelCur; 1384 continue; 1385 } 1386 1387 // Stop when RelCur's offset is past the current instruction. 1388 if (Offset >= Index + Size) 1389 break; 1390 1391 // When --adjust-vma is used, update the address printed. 1392 if (RelCur->getSymbol() != Obj->symbol_end()) { 1393 Expected<section_iterator> SymSI = 1394 RelCur->getSymbol()->getSection(); 1395 if (SymSI && *SymSI != Obj->section_end() && 1396 (shouldAdjustVA(**SymSI))) 1397 Offset += AdjustVMA; 1398 } 1399 1400 printRelocation(*RelCur, SectionAddr + Offset, 1401 Obj->getBytesInAddress()); 1402 ++RelCur; 1403 } 1404 } 1405 } 1406 } 1407 } 1408 } 1409 1410 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1411 if (StartAddress > StopAddress) 1412 error("Start address should be less than stop address"); 1413 1414 const Target *TheTarget = getTarget(Obj); 1415 1416 // Package up features to be passed to target/subtarget 1417 SubtargetFeatures Features = Obj->getFeatures(); 1418 if (!MAttrs.empty()) 1419 for (unsigned I = 0; I != MAttrs.size(); ++I) 1420 Features.AddFeature(MAttrs[I]); 1421 1422 std::unique_ptr<const MCRegisterInfo> MRI( 1423 TheTarget->createMCRegInfo(TripleName)); 1424 if (!MRI) 1425 report_error(Obj->getFileName(), 1426 "no register info for target " + TripleName); 1427 1428 // Set up disassembler. 1429 std::unique_ptr<const MCAsmInfo> AsmInfo( 1430 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1431 if (!AsmInfo) 1432 report_error(Obj->getFileName(), 1433 "no assembly info for target " + TripleName); 1434 std::unique_ptr<const MCSubtargetInfo> STI( 1435 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1436 if (!STI) 1437 report_error(Obj->getFileName(), 1438 "no subtarget info for target " + TripleName); 1439 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1440 if (!MII) 1441 report_error(Obj->getFileName(), 1442 "no instruction info for target " + TripleName); 1443 MCObjectFileInfo MOFI; 1444 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1445 // FIXME: for now initialize MCObjectFileInfo with default values 1446 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1447 1448 std::unique_ptr<MCDisassembler> DisAsm( 1449 TheTarget->createMCDisassembler(*STI, Ctx)); 1450 if (!DisAsm) 1451 report_error(Obj->getFileName(), 1452 "no disassembler for target " + TripleName); 1453 1454 std::unique_ptr<const MCInstrAnalysis> MIA( 1455 TheTarget->createMCInstrAnalysis(MII.get())); 1456 1457 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1458 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1459 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1460 if (!IP) 1461 report_error(Obj->getFileName(), 1462 "no instruction printer for target " + TripleName); 1463 IP->setPrintImmHex(PrintImmHex); 1464 1465 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1466 SourcePrinter SP(Obj, TheTarget->getName()); 1467 1468 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), MIA.get(), IP.get(), 1469 STI.get(), PIP, SP, InlineRelocs); 1470 } 1471 1472 void llvm::printRelocations(const ObjectFile *Obj) { 1473 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1474 "%08" PRIx64; 1475 // Regular objdump doesn't print relocations in non-relocatable object 1476 // files. 1477 if (!Obj->isRelocatableObject()) 1478 return; 1479 1480 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1481 if (Section.relocation_begin() == Section.relocation_end()) 1482 continue; 1483 StringRef SecName; 1484 error(Section.getName(SecName)); 1485 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1486 for (const RelocationRef &Reloc : Section.relocations()) { 1487 uint64_t Address = Reloc.getOffset(); 1488 SmallString<32> RelocName; 1489 SmallString<32> ValueStr; 1490 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1491 continue; 1492 Reloc.getTypeName(RelocName); 1493 error(getRelocationValueString(Reloc, ValueStr)); 1494 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1495 << ValueStr << "\n"; 1496 } 1497 outs() << "\n"; 1498 } 1499 } 1500 1501 void llvm::printDynamicRelocations(const ObjectFile *Obj) { 1502 // For the moment, this option is for ELF only 1503 if (!Obj->isELF()) 1504 return; 1505 1506 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1507 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1508 error("not a dynamic object"); 1509 return; 1510 } 1511 1512 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1513 if (DynRelSec.empty()) 1514 return; 1515 1516 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1517 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1518 for (const SectionRef &Section : DynRelSec) { 1519 if (Section.relocation_begin() == Section.relocation_end()) 1520 continue; 1521 for (const RelocationRef &Reloc : Section.relocations()) { 1522 uint64_t Address = Reloc.getOffset(); 1523 SmallString<32> RelocName; 1524 SmallString<32> ValueStr; 1525 Reloc.getTypeName(RelocName); 1526 error(getRelocationValueString(Reloc, ValueStr)); 1527 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1528 << ValueStr << "\n"; 1529 } 1530 } 1531 } 1532 1533 // Returns true if we need to show LMA column when dumping section headers. We 1534 // show it only when the platform is ELF and either we have at least one section 1535 // whose VMA and LMA are different and/or when --show-lma flag is used. 1536 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1537 if (!Obj->isELF()) 1538 return false; 1539 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1540 if (S.getAddress() != getELFSectionLMA(S)) 1541 return true; 1542 return ShowLMA; 1543 } 1544 1545 void llvm::printSectionHeaders(const ObjectFile *Obj) { 1546 bool HasLMAColumn = shouldDisplayLMA(Obj); 1547 if (HasLMAColumn) 1548 outs() << "Sections:\n" 1549 "Idx Name Size VMA LMA " 1550 "Type\n"; 1551 else 1552 outs() << "Sections:\n" 1553 "Idx Name Size VMA Type\n"; 1554 1555 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1556 StringRef Name; 1557 error(Section.getName(Name)); 1558 uint64_t VMA = Section.getAddress(); 1559 if (shouldAdjustVA(Section)) 1560 VMA += AdjustVMA; 1561 1562 uint64_t Size = Section.getSize(); 1563 bool Text = Section.isText(); 1564 bool Data = Section.isData(); 1565 bool BSS = Section.isBSS(); 1566 std::string Type = (std::string(Text ? "TEXT " : "") + 1567 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1568 1569 if (HasLMAColumn) 1570 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %016" PRIx64 1571 " %s\n", 1572 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1573 VMA, getELFSectionLMA(Section), Type.c_str()); 1574 else 1575 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1576 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1577 VMA, Type.c_str()); 1578 } 1579 outs() << "\n"; 1580 } 1581 1582 void llvm::printSectionContents(const ObjectFile *Obj) { 1583 std::error_code EC; 1584 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1585 StringRef Name; 1586 StringRef Contents; 1587 error(Section.getName(Name)); 1588 uint64_t BaseAddr = Section.getAddress(); 1589 uint64_t Size = Section.getSize(); 1590 if (!Size) 1591 continue; 1592 1593 outs() << "Contents of section " << Name << ":\n"; 1594 if (Section.isBSS()) { 1595 outs() << format("<skipping contents of bss section at [%04" PRIx64 1596 ", %04" PRIx64 ")>\n", 1597 BaseAddr, BaseAddr + Size); 1598 continue; 1599 } 1600 1601 error(Section.getContents(Contents)); 1602 1603 // Dump out the content as hex and printable ascii characters. 1604 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1605 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1606 // Dump line of hex. 1607 for (std::size_t I = 0; I < 16; ++I) { 1608 if (I != 0 && I % 4 == 0) 1609 outs() << ' '; 1610 if (Addr + I < End) 1611 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1612 << hexdigit(Contents[Addr + I] & 0xF, true); 1613 else 1614 outs() << " "; 1615 } 1616 // Print ascii. 1617 outs() << " "; 1618 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1619 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1620 outs() << Contents[Addr + I]; 1621 else 1622 outs() << "."; 1623 } 1624 outs() << "\n"; 1625 } 1626 } 1627 } 1628 1629 void llvm::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1630 StringRef ArchitectureName) { 1631 outs() << "SYMBOL TABLE:\n"; 1632 1633 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1634 printCOFFSymbolTable(Coff); 1635 return; 1636 } 1637 1638 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1639 // Skip printing the special zero symbol when dumping an ELF file. 1640 // This makes the output consistent with the GNU objdump. 1641 if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O)) 1642 continue; 1643 1644 const SymbolRef &Symbol = *I; 1645 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1646 if (!AddressOrError) 1647 report_error(ArchiveName, O->getFileName(), AddressOrError.takeError(), 1648 ArchitectureName); 1649 uint64_t Address = *AddressOrError; 1650 if ((Address < StartAddress) || (Address > StopAddress)) 1651 continue; 1652 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1653 if (!TypeOrError) 1654 report_error(ArchiveName, O->getFileName(), TypeOrError.takeError(), 1655 ArchitectureName); 1656 SymbolRef::Type Type = *TypeOrError; 1657 uint32_t Flags = Symbol.getFlags(); 1658 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1659 if (!SectionOrErr) 1660 report_error(ArchiveName, O->getFileName(), SectionOrErr.takeError(), 1661 ArchitectureName); 1662 section_iterator Section = *SectionOrErr; 1663 StringRef Name; 1664 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 1665 Section->getName(Name); 1666 } else { 1667 Expected<StringRef> NameOrErr = Symbol.getName(); 1668 if (!NameOrErr) 1669 report_error(ArchiveName, O->getFileName(), NameOrErr.takeError(), 1670 ArchitectureName); 1671 Name = *NameOrErr; 1672 } 1673 1674 bool Global = Flags & SymbolRef::SF_Global; 1675 bool Weak = Flags & SymbolRef::SF_Weak; 1676 bool Absolute = Flags & SymbolRef::SF_Absolute; 1677 bool Common = Flags & SymbolRef::SF_Common; 1678 bool Hidden = Flags & SymbolRef::SF_Hidden; 1679 1680 char GlobLoc = ' '; 1681 if (Type != SymbolRef::ST_Unknown) 1682 GlobLoc = Global ? 'g' : 'l'; 1683 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1684 ? 'd' : ' '; 1685 char FileFunc = ' '; 1686 if (Type == SymbolRef::ST_File) 1687 FileFunc = 'f'; 1688 else if (Type == SymbolRef::ST_Function) 1689 FileFunc = 'F'; 1690 else if (Type == SymbolRef::ST_Data) 1691 FileFunc = 'O'; 1692 1693 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1694 "%08" PRIx64; 1695 1696 outs() << format(Fmt, Address) << " " 1697 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1698 << (Weak ? 'w' : ' ') // Weak? 1699 << ' ' // Constructor. Not supported yet. 1700 << ' ' // Warning. Not supported yet. 1701 << ' ' // Indirect reference to another symbol. 1702 << Debug // Debugging (d) or dynamic (D) symbol. 1703 << FileFunc // Name of function (F), file (f) or object (O). 1704 << ' '; 1705 if (Absolute) { 1706 outs() << "*ABS*"; 1707 } else if (Common) { 1708 outs() << "*COM*"; 1709 } else if (Section == O->section_end()) { 1710 outs() << "*UND*"; 1711 } else { 1712 if (const MachOObjectFile *MachO = 1713 dyn_cast<const MachOObjectFile>(O)) { 1714 DataRefImpl DR = Section->getRawDataRefImpl(); 1715 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1716 outs() << SegmentName << ","; 1717 } 1718 StringRef SectionName; 1719 error(Section->getName(SectionName)); 1720 outs() << SectionName; 1721 } 1722 1723 outs() << '\t'; 1724 if (Common || isa<ELFObjectFileBase>(O)) { 1725 uint64_t Val = 1726 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1727 outs() << format("\t %08" PRIx64 " ", Val); 1728 } 1729 1730 if (Hidden) 1731 outs() << ".hidden "; 1732 1733 if (Demangle) 1734 outs() << demangle(Name) << '\n'; 1735 else 1736 outs() << Name << '\n'; 1737 } 1738 } 1739 1740 static void printUnwindInfo(const ObjectFile *O) { 1741 outs() << "Unwind info:\n\n"; 1742 1743 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1744 printCOFFUnwindInfo(Coff); 1745 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1746 printMachOUnwindInfo(MachO); 1747 else 1748 // TODO: Extract DWARF dump tool to objdump. 1749 WithColor::error(errs(), ToolName) 1750 << "This operation is only currently supported " 1751 "for COFF and MachO object files.\n"; 1752 } 1753 1754 void llvm::printExportsTrie(const ObjectFile *o) { 1755 outs() << "Exports trie:\n"; 1756 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1757 printMachOExportsTrie(MachO); 1758 else 1759 WithColor::error(errs(), ToolName) 1760 << "This operation is only currently supported " 1761 "for Mach-O executable files.\n"; 1762 } 1763 1764 void llvm::printRebaseTable(ObjectFile *o) { 1765 outs() << "Rebase table:\n"; 1766 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1767 printMachORebaseTable(MachO); 1768 else 1769 WithColor::error(errs(), ToolName) 1770 << "This operation is only currently supported " 1771 "for Mach-O executable files.\n"; 1772 } 1773 1774 void llvm::printBindTable(ObjectFile *o) { 1775 outs() << "Bind table:\n"; 1776 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1777 printMachOBindTable(MachO); 1778 else 1779 WithColor::error(errs(), ToolName) 1780 << "This operation is only currently supported " 1781 "for Mach-O executable files.\n"; 1782 } 1783 1784 void llvm::printLazyBindTable(ObjectFile *o) { 1785 outs() << "Lazy bind table:\n"; 1786 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1787 printMachOLazyBindTable(MachO); 1788 else 1789 WithColor::error(errs(), ToolName) 1790 << "This operation is only currently supported " 1791 "for Mach-O executable files.\n"; 1792 } 1793 1794 void llvm::printWeakBindTable(ObjectFile *o) { 1795 outs() << "Weak bind table:\n"; 1796 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1797 printMachOWeakBindTable(MachO); 1798 else 1799 WithColor::error(errs(), ToolName) 1800 << "This operation is only currently supported " 1801 "for Mach-O executable files.\n"; 1802 } 1803 1804 /// Dump the raw contents of the __clangast section so the output can be piped 1805 /// into llvm-bcanalyzer. 1806 void llvm::printRawClangAST(const ObjectFile *Obj) { 1807 if (outs().is_displayed()) { 1808 WithColor::error(errs(), ToolName) 1809 << "The -raw-clang-ast option will dump the raw binary contents of " 1810 "the clang ast section.\n" 1811 "Please redirect the output to a file or another program such as " 1812 "llvm-bcanalyzer.\n"; 1813 return; 1814 } 1815 1816 StringRef ClangASTSectionName("__clangast"); 1817 if (isa<COFFObjectFile>(Obj)) { 1818 ClangASTSectionName = "clangast"; 1819 } 1820 1821 Optional<object::SectionRef> ClangASTSection; 1822 for (auto Sec : ToolSectionFilter(*Obj)) { 1823 StringRef Name; 1824 Sec.getName(Name); 1825 if (Name == ClangASTSectionName) { 1826 ClangASTSection = Sec; 1827 break; 1828 } 1829 } 1830 if (!ClangASTSection) 1831 return; 1832 1833 StringRef ClangASTContents; 1834 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1835 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1836 } 1837 1838 static void printFaultMaps(const ObjectFile *Obj) { 1839 StringRef FaultMapSectionName; 1840 1841 if (isa<ELFObjectFileBase>(Obj)) { 1842 FaultMapSectionName = ".llvm_faultmaps"; 1843 } else if (isa<MachOObjectFile>(Obj)) { 1844 FaultMapSectionName = "__llvm_faultmaps"; 1845 } else { 1846 WithColor::error(errs(), ToolName) 1847 << "This operation is only currently supported " 1848 "for ELF and Mach-O executable files.\n"; 1849 return; 1850 } 1851 1852 Optional<object::SectionRef> FaultMapSection; 1853 1854 for (auto Sec : ToolSectionFilter(*Obj)) { 1855 StringRef Name; 1856 Sec.getName(Name); 1857 if (Name == FaultMapSectionName) { 1858 FaultMapSection = Sec; 1859 break; 1860 } 1861 } 1862 1863 outs() << "FaultMap table:\n"; 1864 1865 if (!FaultMapSection.hasValue()) { 1866 outs() << "<not found>\n"; 1867 return; 1868 } 1869 1870 StringRef FaultMapContents; 1871 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1872 1873 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1874 FaultMapContents.bytes_end()); 1875 1876 outs() << FMP; 1877 } 1878 1879 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1880 if (O->isELF()) { 1881 printELFFileHeader(O); 1882 return printELFDynamicSection(O); 1883 } 1884 if (O->isCOFF()) 1885 return printCOFFFileHeader(O); 1886 if (O->isWasm()) 1887 return printWasmFileHeader(O); 1888 if (O->isMachO()) { 1889 printMachOFileHeader(O); 1890 if (!OnlyFirst) 1891 printMachOLoadCommands(O); 1892 return; 1893 } 1894 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1895 } 1896 1897 static void printFileHeaders(const ObjectFile *O) { 1898 if (!O->isELF() && !O->isCOFF()) 1899 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1900 1901 Triple::ArchType AT = O->getArch(); 1902 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1903 Expected<uint64_t> StartAddrOrErr = O->getStartAddress(); 1904 if (!StartAddrOrErr) 1905 report_error(O->getFileName(), StartAddrOrErr.takeError()); 1906 1907 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1908 uint64_t Address = StartAddrOrErr.get(); 1909 outs() << "start address: " 1910 << "0x" << format(Fmt.data(), Address) << "\n\n"; 1911 } 1912 1913 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 1914 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 1915 if (!ModeOrErr) { 1916 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 1917 consumeError(ModeOrErr.takeError()); 1918 return; 1919 } 1920 sys::fs::perms Mode = ModeOrErr.get(); 1921 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1922 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1923 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1924 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1925 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1926 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1927 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1928 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1929 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1930 1931 outs() << " "; 1932 1933 Expected<unsigned> UIDOrErr = C.getUID(); 1934 if (!UIDOrErr) 1935 report_error(Filename, UIDOrErr.takeError()); 1936 unsigned UID = UIDOrErr.get(); 1937 outs() << format("%d/", UID); 1938 1939 Expected<unsigned> GIDOrErr = C.getGID(); 1940 if (!GIDOrErr) 1941 report_error(Filename, GIDOrErr.takeError()); 1942 unsigned GID = GIDOrErr.get(); 1943 outs() << format("%-d ", GID); 1944 1945 Expected<uint64_t> Size = C.getRawSize(); 1946 if (!Size) 1947 report_error(Filename, Size.takeError()); 1948 outs() << format("%6" PRId64, Size.get()) << " "; 1949 1950 StringRef RawLastModified = C.getRawLastModified(); 1951 unsigned Seconds; 1952 if (RawLastModified.getAsInteger(10, Seconds)) 1953 outs() << "(date: \"" << RawLastModified 1954 << "\" contains non-decimal chars) "; 1955 else { 1956 // Since ctime(3) returns a 26 character string of the form: 1957 // "Sun Sep 16 01:03:52 1973\n\0" 1958 // just print 24 characters. 1959 time_t t = Seconds; 1960 outs() << format("%.24s ", ctime(&t)); 1961 } 1962 1963 StringRef Name = ""; 1964 Expected<StringRef> NameOrErr = C.getName(); 1965 if (!NameOrErr) { 1966 consumeError(NameOrErr.takeError()); 1967 Expected<StringRef> RawNameOrErr = C.getRawName(); 1968 if (!RawNameOrErr) 1969 report_error(Filename, NameOrErr.takeError()); 1970 Name = RawNameOrErr.get(); 1971 } else { 1972 Name = NameOrErr.get(); 1973 } 1974 outs() << Name << "\n"; 1975 } 1976 1977 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 1978 const Archive::Child *C = nullptr) { 1979 // Avoid other output when using a raw option. 1980 if (!RawClangAST) { 1981 outs() << '\n'; 1982 if (A) 1983 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 1984 else 1985 outs() << O->getFileName(); 1986 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 1987 } 1988 1989 StringRef ArchiveName = A ? A->getFileName() : ""; 1990 if (FileHeaders) 1991 printFileHeaders(O); 1992 if (ArchiveHeaders && !MachOOpt && C) 1993 printArchiveChild(ArchiveName, *C); 1994 if (Disassemble) 1995 disassembleObject(O, Relocations); 1996 if (Relocations && !Disassemble) 1997 printRelocations(O); 1998 if (DynamicRelocations) 1999 printDynamicRelocations(O); 2000 if (SectionHeaders) 2001 printSectionHeaders(O); 2002 if (SectionContents) 2003 printSectionContents(O); 2004 if (SymbolTable) 2005 printSymbolTable(O, ArchiveName); 2006 if (UnwindInfo) 2007 printUnwindInfo(O); 2008 if (PrivateHeaders || FirstPrivateHeader) 2009 printPrivateFileHeaders(O, FirstPrivateHeader); 2010 if (ExportsTrie) 2011 printExportsTrie(O); 2012 if (Rebase) 2013 printRebaseTable(O); 2014 if (Bind) 2015 printBindTable(O); 2016 if (LazyBind) 2017 printLazyBindTable(O); 2018 if (WeakBind) 2019 printWeakBindTable(O); 2020 if (RawClangAST) 2021 printRawClangAST(O); 2022 if (PrintFaultMaps) 2023 printFaultMaps(O); 2024 if (DwarfDumpType != DIDT_Null) { 2025 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2026 // Dump the complete DWARF structure. 2027 DIDumpOptions DumpOpts; 2028 DumpOpts.DumpType = DwarfDumpType; 2029 DICtx->dump(outs(), DumpOpts); 2030 } 2031 } 2032 2033 static void dumpObject(const COFFImportFile *I, const Archive *A, 2034 const Archive::Child *C = nullptr) { 2035 StringRef ArchiveName = A ? A->getFileName() : ""; 2036 2037 // Avoid other output when using a raw option. 2038 if (!RawClangAST) 2039 outs() << '\n' 2040 << ArchiveName << "(" << I->getFileName() << ")" 2041 << ":\tfile format COFF-import-file" 2042 << "\n\n"; 2043 2044 if (ArchiveHeaders && !MachOOpt && C) 2045 printArchiveChild(ArchiveName, *C); 2046 if (SymbolTable) 2047 printCOFFSymbolTable(I); 2048 } 2049 2050 /// Dump each object file in \a a; 2051 static void dumpArchive(const Archive *A) { 2052 Error Err = Error::success(); 2053 for (auto &C : A->children(Err)) { 2054 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2055 if (!ChildOrErr) { 2056 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2057 report_error(A->getFileName(), C, std::move(E)); 2058 continue; 2059 } 2060 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2061 dumpObject(O, A, &C); 2062 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2063 dumpObject(I, A, &C); 2064 else 2065 report_error(A->getFileName(), object_error::invalid_file_type); 2066 } 2067 if (Err) 2068 report_error(A->getFileName(), std::move(Err)); 2069 } 2070 2071 /// Open file and figure out how to dump it. 2072 static void dumpInput(StringRef file) { 2073 // If we are using the Mach-O specific object file parser, then let it parse 2074 // the file and process the command line options. So the -arch flags can 2075 // be used to select specific slices, etc. 2076 if (MachOOpt) { 2077 parseInputMachO(file); 2078 return; 2079 } 2080 2081 // Attempt to open the binary. 2082 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2083 if (!BinaryOrErr) 2084 report_error(file, BinaryOrErr.takeError()); 2085 Binary &Binary = *BinaryOrErr.get().getBinary(); 2086 2087 if (Archive *A = dyn_cast<Archive>(&Binary)) 2088 dumpArchive(A); 2089 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2090 dumpObject(O); 2091 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2092 parseInputMachO(UB); 2093 else 2094 report_error(file, object_error::invalid_file_type); 2095 } 2096 2097 int main(int argc, char **argv) { 2098 InitLLVM X(argc, argv); 2099 2100 // Initialize targets and assembly printers/parsers. 2101 llvm::InitializeAllTargetInfos(); 2102 llvm::InitializeAllTargetMCs(); 2103 llvm::InitializeAllDisassemblers(); 2104 2105 // Register the target printer for --version. 2106 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2107 2108 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2109 2110 ToolName = argv[0]; 2111 2112 // Defaults to a.out if no filenames specified. 2113 if (InputFilenames.empty()) 2114 InputFilenames.push_back("a.out"); 2115 2116 if (AllHeaders) 2117 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2118 SectionHeaders = SymbolTable = true; 2119 2120 if (DisassembleAll || PrintSource || PrintLines) 2121 Disassemble = true; 2122 2123 if (!Disassemble 2124 && !Relocations 2125 && !DynamicRelocations 2126 && !SectionHeaders 2127 && !SectionContents 2128 && !SymbolTable 2129 && !UnwindInfo 2130 && !PrivateHeaders 2131 && !FileHeaders 2132 && !FirstPrivateHeader 2133 && !ExportsTrie 2134 && !Rebase 2135 && !Bind 2136 && !LazyBind 2137 && !WeakBind 2138 && !RawClangAST 2139 && !(UniversalHeaders && MachOOpt) 2140 && !ArchiveHeaders 2141 && !(IndirectSymbols && MachOOpt) 2142 && !(DataInCode && MachOOpt) 2143 && !(LinkOptHints && MachOOpt) 2144 && !(InfoPlist && MachOOpt) 2145 && !(DylibsUsed && MachOOpt) 2146 && !(DylibId && MachOOpt) 2147 && !(ObjcMetaData && MachOOpt) 2148 && !(!FilterSections.empty() && MachOOpt) 2149 && !PrintFaultMaps 2150 && DwarfDumpType == DIDT_Null) { 2151 cl::PrintHelpMessage(); 2152 return 2; 2153 } 2154 2155 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2156 DisassembleFunctions.end()); 2157 2158 llvm::for_each(InputFilenames, dumpInput); 2159 2160 return EXIT_SUCCESS; 2161 } 2162