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::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 87 Disassembled("d", cl::desc("Alias for --disassemble"), 88 cl::aliasopt(Disassemble)); 89 90 cl::opt<bool> 91 llvm::DisassembleAll("disassemble-all", 92 cl::desc("Display assembler mnemonics for the machine instructions")); 93 static cl::alias 94 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 95 cl::aliasopt(DisassembleAll)); 96 97 cl::opt<bool> llvm::Demangle("demangle", cl::desc("Demangle symbols names"), 98 cl::init(false)); 99 100 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 101 cl::aliasopt(llvm::Demangle)); 102 103 static cl::list<std::string> 104 DisassembleFunctions("df", 105 cl::CommaSeparated, 106 cl::desc("List of functions to disassemble")); 107 static StringSet<> DisasmFuncsSet; 108 109 cl::opt<bool> 110 llvm::Relocations("reloc", 111 cl::desc("Display the relocation entries in the file")); 112 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 113 cl::NotHidden, 114 cl::aliasopt(llvm::Relocations)); 115 116 cl::opt<bool> 117 llvm::DynamicRelocations("dynamic-reloc", 118 cl::desc("Display the dynamic relocation entries in the file")); 119 static cl::alias 120 DynamicRelocationsd("R", cl::desc("Alias for --dynamic-reloc"), 121 cl::aliasopt(DynamicRelocations)); 122 123 cl::opt<bool> 124 llvm::SectionContents("full-contents", 125 cl::desc("Display the content of each section")); 126 static cl::alias SectionContentsShort("s", 127 cl::desc("Alias for --full-contents"), 128 cl::aliasopt(SectionContents)); 129 130 cl::opt<bool> llvm::SymbolTable("syms", cl::desc("Display the symbol table")); 131 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 132 cl::NotHidden, 133 cl::aliasopt(llvm::SymbolTable)); 134 135 cl::opt<bool> 136 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 137 138 cl::opt<bool> 139 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 140 141 cl::opt<bool> 142 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 143 144 cl::opt<bool> 145 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 146 147 cl::opt<bool> 148 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 149 150 cl::opt<bool> 151 llvm::RawClangAST("raw-clang-ast", 152 cl::desc("Dump the raw binary contents of the clang AST section")); 153 154 static cl::opt<bool> 155 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 156 static cl::alias 157 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 158 159 cl::opt<std::string> 160 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 161 "see -version for available targets")); 162 163 cl::opt<std::string> 164 llvm::MCPU("mcpu", 165 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 166 cl::value_desc("cpu-name"), 167 cl::init("")); 168 169 cl::opt<std::string> 170 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 171 "see -version for available targets")); 172 173 cl::opt<bool> 174 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 175 "headers for each section.")); 176 static cl::alias 177 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 178 cl::aliasopt(SectionHeaders)); 179 static cl::alias 180 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 181 cl::aliasopt(SectionHeaders)); 182 183 cl::list<std::string> 184 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 185 "With -macho dump segment,section")); 186 cl::alias 187 static FilterSectionsj("j", cl::desc("Alias for --section"), 188 cl::aliasopt(llvm::FilterSections)); 189 190 cl::list<std::string> 191 llvm::MAttrs("mattr", 192 cl::CommaSeparated, 193 cl::desc("Target specific attributes"), 194 cl::value_desc("a1,+a2,-a3,...")); 195 196 cl::opt<bool> 197 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 198 "instructions, do not print " 199 "the instruction bytes.")); 200 cl::opt<bool> 201 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 202 203 cl::opt<bool> 204 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 205 206 static cl::alias 207 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 208 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 220 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 221 cl::aliasopt(PrivateHeaders)); 222 223 cl::opt<bool> llvm::FileHeaders( 224 "file-headers", 225 cl::desc("Display the contents of the overall file header")); 226 227 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 228 cl::aliasopt(FileHeaders)); 229 230 cl::opt<bool> 231 llvm::ArchiveHeaders("archive-headers", 232 cl::desc("Display archive header information")); 233 234 cl::alias 235 ArchiveHeadersShort("a", cl::desc("Alias for --archive-headers"), 236 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"), 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::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::aliasopt(DisassembleZeroes)); 278 279 static StringRef ToolName; 280 281 namespace { 282 struct SectionSymbol { 283 SectionSymbol(uint64_t Address, StringRef Name, uint8_t Type) 284 : Address(Address), Name(Name), Type(Type) {} 285 286 bool operator<(const SectionSymbol &Other) const { 287 return std::tie(Address, Name, Type) < 288 std::tie(Other.Address, Other.Name, Other.Type); 289 } 290 291 uint64_t Address; 292 StringRef Name; 293 uint8_t Type; 294 }; 295 296 typedef std::vector<SectionSymbol> SectionSymbolsTy; 297 298 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 299 300 class SectionFilterIterator { 301 public: 302 SectionFilterIterator(FilterPredicate P, 303 llvm::object::section_iterator const &I, 304 llvm::object::section_iterator const &E) 305 : Predicate(std::move(P)), Iterator(I), End(E) { 306 ScanPredicate(); 307 } 308 const llvm::object::SectionRef &operator*() const { return *Iterator; } 309 SectionFilterIterator &operator++() { 310 ++Iterator; 311 ScanPredicate(); 312 return *this; 313 } 314 bool operator!=(SectionFilterIterator const &Other) const { 315 return Iterator != Other.Iterator; 316 } 317 318 private: 319 void ScanPredicate() { 320 while (Iterator != End && !Predicate(*Iterator)) { 321 ++Iterator; 322 } 323 } 324 FilterPredicate Predicate; 325 llvm::object::section_iterator Iterator; 326 llvm::object::section_iterator End; 327 }; 328 329 class SectionFilter { 330 public: 331 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 332 : Predicate(std::move(P)), Object(O) {} 333 SectionFilterIterator begin() { 334 return SectionFilterIterator(Predicate, Object.section_begin(), 335 Object.section_end()); 336 } 337 SectionFilterIterator end() { 338 return SectionFilterIterator(Predicate, Object.section_end(), 339 Object.section_end()); 340 } 341 342 private: 343 FilterPredicate Predicate; 344 llvm::object::ObjectFile const &Object; 345 }; 346 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 347 return SectionFilter( 348 [](llvm::object::SectionRef const &S) { 349 if (FilterSections.empty()) 350 return true; 351 llvm::StringRef String; 352 std::error_code error = S.getName(String); 353 if (error) 354 return false; 355 return is_contained(FilterSections, String); 356 }, 357 O); 358 } 359 } 360 361 void llvm::error(std::error_code EC) { 362 if (!EC) 363 return; 364 WithColor::error(errs(), ToolName) 365 << "reading file: " << EC.message() << ".\n"; 366 errs().flush(); 367 exit(1); 368 } 369 370 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 371 WithColor::error(errs(), ToolName) << Message << ".\n"; 372 errs().flush(); 373 exit(1); 374 } 375 376 void llvm::warn(StringRef Message) { 377 WithColor::warning(errs(), ToolName) << Message << ".\n"; 378 errs().flush(); 379 } 380 381 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 382 Twine Message) { 383 WithColor::error(errs(), ToolName) 384 << "'" << File << "': " << Message << ".\n"; 385 exit(1); 386 } 387 388 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 389 std::error_code EC) { 390 assert(EC); 391 WithColor::error(errs(), ToolName) 392 << "'" << File << "': " << EC.message() << ".\n"; 393 exit(1); 394 } 395 396 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 397 llvm::Error E) { 398 assert(E); 399 std::string Buf; 400 raw_string_ostream OS(Buf); 401 logAllUnhandledErrors(std::move(E), OS); 402 OS.flush(); 403 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 404 exit(1); 405 } 406 407 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 408 StringRef FileName, 409 llvm::Error E, 410 StringRef ArchitectureName) { 411 assert(E); 412 WithColor::error(errs(), ToolName); 413 if (ArchiveName != "") 414 errs() << ArchiveName << "(" << FileName << ")"; 415 else 416 errs() << "'" << FileName << "'"; 417 if (!ArchitectureName.empty()) 418 errs() << " (for architecture " << ArchitectureName << ")"; 419 std::string Buf; 420 raw_string_ostream OS(Buf); 421 logAllUnhandledErrors(std::move(E), OS); 422 OS.flush(); 423 errs() << ": " << Buf; 424 exit(1); 425 } 426 427 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 428 const object::Archive::Child &C, 429 llvm::Error E, 430 StringRef ArchitectureName) { 431 Expected<StringRef> NameOrErr = C.getName(); 432 // TODO: if we have a error getting the name then it would be nice to print 433 // the index of which archive member this is and or its offset in the 434 // archive instead of "???" as the name. 435 if (!NameOrErr) { 436 consumeError(NameOrErr.takeError()); 437 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 438 } else 439 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 440 ArchitectureName); 441 } 442 443 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 444 // Figure out the target triple. 445 llvm::Triple TheTriple("unknown-unknown-unknown"); 446 if (TripleName.empty()) { 447 if (Obj) 448 TheTriple = Obj->makeTriple(); 449 } else { 450 TheTriple.setTriple(Triple::normalize(TripleName)); 451 452 // Use the triple, but also try to combine with ARM build attributes. 453 if (Obj) { 454 auto Arch = Obj->getArch(); 455 if (Arch == Triple::arm || Arch == Triple::armeb) 456 Obj->setARMSubArch(TheTriple); 457 } 458 } 459 460 // Get the target specific parser. 461 std::string Error; 462 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 463 Error); 464 if (!TheTarget) { 465 if (Obj) 466 report_error(Obj->getFileName(), "can't find target: " + Error); 467 else 468 error("can't find target: " + Error); 469 } 470 471 // Update the triple name and return the found target. 472 TripleName = TheTriple.getTriple(); 473 return TheTarget; 474 } 475 476 bool llvm::isRelocAddressLess(RelocationRef A, RelocationRef B) { 477 return A.getOffset() < B.getOffset(); 478 } 479 480 template <class ELFT> 481 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 482 const RelocationRef &RelRef, 483 SmallVectorImpl<char> &Result) { 484 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 485 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 486 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 487 488 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 489 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 490 auto SecOrErr = EF.getSection(Rel.d.a); 491 if (!SecOrErr) 492 return errorToErrorCode(SecOrErr.takeError()); 493 494 int64_t Addend = 0; 495 // If there is no Symbol associated with the relocation, we set the undef 496 // boolean value to 'true'. This will prevent us from calling functions that 497 // requires the relocation to be associated with a symbol. 498 // 499 // In SHT_REL case we would need to read the addend from section data. 500 // GNU objdump does not do that and we just follow for simplicity. 501 bool Undef = false; 502 if ((*SecOrErr)->sh_type == ELF::SHT_RELA) { 503 const Elf_Rela *ERela = Obj->getRela(Rel); 504 Addend = ERela->r_addend; 505 Undef = ERela->getSymbol(false) == 0; 506 } else if ((*SecOrErr)->sh_type != ELF::SHT_REL) { 507 return object_error::parse_failed; 508 } 509 510 // Default scheme is to print Target, as well as "+ <addend>" for nonzero 511 // addend. Should be acceptable for all normal purposes. 512 std::string FmtBuf; 513 raw_string_ostream Fmt(FmtBuf); 514 515 if (!Undef) { 516 symbol_iterator SI = RelRef.getSymbol(); 517 const Elf_Sym *Sym = Obj->getSymbol(SI->getRawDataRefImpl()); 518 if (Sym->getType() == ELF::STT_SECTION) { 519 Expected<section_iterator> SymSI = SI->getSection(); 520 if (!SymSI) 521 return errorToErrorCode(SymSI.takeError()); 522 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 523 auto SecName = EF.getSectionName(SymSec); 524 if (!SecName) 525 return errorToErrorCode(SecName.takeError()); 526 Fmt << *SecName; 527 } else { 528 Expected<StringRef> SymName = SI->getName(); 529 if (!SymName) 530 return errorToErrorCode(SymName.takeError()); 531 if (Demangle) 532 Fmt << demangle(*SymName); 533 else 534 Fmt << *SymName; 535 } 536 } else { 537 Fmt << "*ABS*"; 538 } 539 540 if (Addend != 0) 541 Fmt << (Addend < 0 ? "" : "+") << Addend; 542 Fmt.flush(); 543 Result.append(FmtBuf.begin(), FmtBuf.end()); 544 return std::error_code(); 545 } 546 547 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 548 const RelocationRef &Rel, 549 SmallVectorImpl<char> &Result) { 550 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 551 return getRelocationValueString(ELF32LE, Rel, Result); 552 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 553 return getRelocationValueString(ELF64LE, Rel, Result); 554 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 555 return getRelocationValueString(ELF32BE, Rel, Result); 556 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 557 return getRelocationValueString(ELF64BE, Rel, Result); 558 } 559 560 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 561 const RelocationRef &Rel, 562 SmallVectorImpl<char> &Result) { 563 symbol_iterator SymI = Rel.getSymbol(); 564 Expected<StringRef> SymNameOrErr = SymI->getName(); 565 if (!SymNameOrErr) 566 return errorToErrorCode(SymNameOrErr.takeError()); 567 StringRef SymName = *SymNameOrErr; 568 Result.append(SymName.begin(), SymName.end()); 569 return std::error_code(); 570 } 571 572 static void printRelocationTargetName(const MachOObjectFile *O, 573 const MachO::any_relocation_info &RE, 574 raw_string_ostream &Fmt) { 575 // Target of a scattered relocation is an address. In the interest of 576 // generating pretty output, scan through the symbol table looking for a 577 // symbol that aligns with that address. If we find one, print it. 578 // Otherwise, we just print the hex address of the target. 579 if (O->isRelocationScattered(RE)) { 580 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 581 582 for (const SymbolRef &Symbol : O->symbols()) { 583 Expected<uint64_t> Addr = Symbol.getAddress(); 584 if (!Addr) 585 report_error(O->getFileName(), Addr.takeError()); 586 if (*Addr != Val) 587 continue; 588 Expected<StringRef> Name = Symbol.getName(); 589 if (!Name) 590 report_error(O->getFileName(), Name.takeError()); 591 Fmt << *Name; 592 return; 593 } 594 595 // If we couldn't find a symbol that this relocation refers to, try 596 // to find a section beginning instead. 597 for (const SectionRef &Section : ToolSectionFilter(*O)) { 598 std::error_code ec; 599 600 StringRef Name; 601 uint64_t Addr = Section.getAddress(); 602 if (Addr != Val) 603 continue; 604 if ((ec = Section.getName(Name))) 605 report_error(O->getFileName(), ec); 606 Fmt << Name; 607 return; 608 } 609 610 Fmt << format("0x%x", Val); 611 return; 612 } 613 614 StringRef S; 615 bool isExtern = O->getPlainRelocationExternal(RE); 616 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 617 618 if (O->getAnyRelocationType(RE) == MachO::ARM64_RELOC_ADDEND) { 619 Fmt << format("0x%0" PRIx64, Val); 620 return; 621 } 622 623 if (isExtern) { 624 symbol_iterator SI = O->symbol_begin(); 625 advance(SI, Val); 626 Expected<StringRef> SOrErr = SI->getName(); 627 if (!SOrErr) 628 report_error(O->getFileName(), SOrErr.takeError()); 629 S = *SOrErr; 630 } else { 631 section_iterator SI = O->section_begin(); 632 // Adjust for the fact that sections are 1-indexed. 633 if (Val == 0) { 634 Fmt << "0 (?,?)"; 635 return; 636 } 637 uint32_t I = Val - 1; 638 while (I != 0 && SI != O->section_end()) { 639 --I; 640 advance(SI, 1); 641 } 642 if (SI == O->section_end()) 643 Fmt << Val << " (?,?)"; 644 else 645 SI->getName(S); 646 } 647 648 Fmt << S; 649 } 650 651 static std::error_code getRelocationValueString(const WasmObjectFile *Obj, 652 const RelocationRef &RelRef, 653 SmallVectorImpl<char> &Result) { 654 const wasm::WasmRelocation& Rel = Obj->getWasmRelocation(RelRef); 655 symbol_iterator SI = RelRef.getSymbol(); 656 std::string FmtBuf; 657 raw_string_ostream Fmt(FmtBuf); 658 if (SI == Obj->symbol_end()) { 659 // Not all wasm relocations have symbols associated with them. 660 // In particular R_WEBASSEMBLY_TYPE_INDEX_LEB. 661 Fmt << Rel.Index; 662 } else { 663 Expected<StringRef> SymNameOrErr = SI->getName(); 664 if (!SymNameOrErr) 665 return errorToErrorCode(SymNameOrErr.takeError()); 666 StringRef SymName = *SymNameOrErr; 667 Result.append(SymName.begin(), SymName.end()); 668 } 669 Fmt << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 670 Fmt.flush(); 671 Result.append(FmtBuf.begin(), FmtBuf.end()); 672 return std::error_code(); 673 } 674 675 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 676 const RelocationRef &RelRef, 677 SmallVectorImpl<char> &Result) { 678 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 679 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 680 681 unsigned Arch = Obj->getArch(); 682 683 std::string FmtBuf; 684 raw_string_ostream Fmt(FmtBuf); 685 unsigned Type = Obj->getAnyRelocationType(RE); 686 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 687 688 // Determine any addends that should be displayed with the relocation. 689 // These require decoding the relocation type, which is triple-specific. 690 691 // X86_64 has entirely custom relocation types. 692 if (Arch == Triple::x86_64) { 693 switch (Type) { 694 case MachO::X86_64_RELOC_GOT_LOAD: 695 case MachO::X86_64_RELOC_GOT: { 696 printRelocationTargetName(Obj, RE, Fmt); 697 Fmt << "@GOT"; 698 if (IsPCRel) 699 Fmt << "PCREL"; 700 break; 701 } 702 case MachO::X86_64_RELOC_SUBTRACTOR: { 703 DataRefImpl RelNext = Rel; 704 Obj->moveRelocationNext(RelNext); 705 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 706 707 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 708 // X86_64_RELOC_UNSIGNED. 709 // NOTE: Scattered relocations don't exist on x86_64. 710 unsigned RType = Obj->getAnyRelocationType(RENext); 711 if (RType != MachO::X86_64_RELOC_UNSIGNED) 712 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 713 "X86_64_RELOC_SUBTRACTOR."); 714 715 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 716 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 717 printRelocationTargetName(Obj, RENext, Fmt); 718 Fmt << "-"; 719 printRelocationTargetName(Obj, RE, Fmt); 720 break; 721 } 722 case MachO::X86_64_RELOC_TLV: 723 printRelocationTargetName(Obj, RE, Fmt); 724 Fmt << "@TLV"; 725 if (IsPCRel) 726 Fmt << "P"; 727 break; 728 case MachO::X86_64_RELOC_SIGNED_1: 729 printRelocationTargetName(Obj, RE, Fmt); 730 Fmt << "-1"; 731 break; 732 case MachO::X86_64_RELOC_SIGNED_2: 733 printRelocationTargetName(Obj, RE, Fmt); 734 Fmt << "-2"; 735 break; 736 case MachO::X86_64_RELOC_SIGNED_4: 737 printRelocationTargetName(Obj, RE, Fmt); 738 Fmt << "-4"; 739 break; 740 default: 741 printRelocationTargetName(Obj, RE, Fmt); 742 break; 743 } 744 // X86 and ARM share some relocation types in common. 745 } else if (Arch == Triple::x86 || Arch == Triple::arm || 746 Arch == Triple::ppc) { 747 // Generic relocation types... 748 switch (Type) { 749 case MachO::GENERIC_RELOC_PAIR: // prints no info 750 return std::error_code(); 751 case MachO::GENERIC_RELOC_SECTDIFF: { 752 DataRefImpl RelNext = Rel; 753 Obj->moveRelocationNext(RelNext); 754 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 755 756 // X86 sect diff's must be followed by a relocation of type 757 // GENERIC_RELOC_PAIR. 758 unsigned RType = Obj->getAnyRelocationType(RENext); 759 760 if (RType != MachO::GENERIC_RELOC_PAIR) 761 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 762 "GENERIC_RELOC_SECTDIFF."); 763 764 printRelocationTargetName(Obj, RE, Fmt); 765 Fmt << "-"; 766 printRelocationTargetName(Obj, RENext, Fmt); 767 break; 768 } 769 } 770 771 if (Arch == Triple::x86 || Arch == Triple::ppc) { 772 switch (Type) { 773 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 774 DataRefImpl RelNext = Rel; 775 Obj->moveRelocationNext(RelNext); 776 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 777 778 // X86 sect diff's must be followed by a relocation of type 779 // GENERIC_RELOC_PAIR. 780 unsigned RType = Obj->getAnyRelocationType(RENext); 781 if (RType != MachO::GENERIC_RELOC_PAIR) 782 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 783 "GENERIC_RELOC_LOCAL_SECTDIFF."); 784 785 printRelocationTargetName(Obj, RE, Fmt); 786 Fmt << "-"; 787 printRelocationTargetName(Obj, RENext, Fmt); 788 break; 789 } 790 case MachO::GENERIC_RELOC_TLV: { 791 printRelocationTargetName(Obj, RE, Fmt); 792 Fmt << "@TLV"; 793 if (IsPCRel) 794 Fmt << "P"; 795 break; 796 } 797 default: 798 printRelocationTargetName(Obj, RE, Fmt); 799 } 800 } else { // ARM-specific relocations 801 switch (Type) { 802 case MachO::ARM_RELOC_HALF: 803 case MachO::ARM_RELOC_HALF_SECTDIFF: { 804 // Half relocations steal a bit from the length field to encode 805 // whether this is an upper16 or a lower16 relocation. 806 bool isUpper = (Obj->getAnyRelocationLength(RE) & 0x1) == 1; 807 808 if (isUpper) 809 Fmt << ":upper16:("; 810 else 811 Fmt << ":lower16:("; 812 printRelocationTargetName(Obj, RE, Fmt); 813 814 DataRefImpl RelNext = Rel; 815 Obj->moveRelocationNext(RelNext); 816 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 817 818 // ARM half relocs must be followed by a relocation of type 819 // ARM_RELOC_PAIR. 820 unsigned RType = Obj->getAnyRelocationType(RENext); 821 if (RType != MachO::ARM_RELOC_PAIR) 822 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 823 "ARM_RELOC_HALF"); 824 825 // NOTE: The half of the target virtual address is stashed in the 826 // address field of the secondary relocation, but we can't reverse 827 // engineer the constant offset from it without decoding the movw/movt 828 // instruction to find the other half in its immediate field. 829 830 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 831 // symbol/section pointer of the follow-on relocation. 832 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 833 Fmt << "-"; 834 printRelocationTargetName(Obj, RENext, Fmt); 835 } 836 837 Fmt << ")"; 838 break; 839 } 840 default: { printRelocationTargetName(Obj, RE, Fmt); } 841 } 842 } 843 } else 844 printRelocationTargetName(Obj, RE, Fmt); 845 846 Fmt.flush(); 847 Result.append(FmtBuf.begin(), FmtBuf.end()); 848 return std::error_code(); 849 } 850 851 static std::error_code getRelocationValueString(const RelocationRef &Rel, 852 SmallVectorImpl<char> &Result) { 853 const ObjectFile *Obj = Rel.getObject(); 854 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 855 return getRelocationValueString(ELF, Rel, Result); 856 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 857 return getRelocationValueString(COFF, Rel, Result); 858 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 859 return getRelocationValueString(Wasm, Rel, Result); 860 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 861 return getRelocationValueString(MachO, Rel, Result); 862 llvm_unreachable("unknown object file format"); 863 } 864 865 /// Indicates whether this relocation should hidden when listing 866 /// relocations, usually because it is the trailing part of a multipart 867 /// relocation that will be printed as part of the leading relocation. 868 static bool getHidden(RelocationRef RelRef) { 869 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 870 if (!MachO) 871 return false; 872 873 unsigned Arch = MachO->getArch(); 874 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 875 uint64_t Type = MachO->getRelocationType(Rel); 876 877 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 878 // is always hidden. 879 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 880 return Type == MachO::GENERIC_RELOC_PAIR; 881 882 if (Arch == Triple::x86_64) { 883 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 884 // an X86_64_RELOC_SUBTRACTOR. 885 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 886 DataRefImpl RelPrev = Rel; 887 RelPrev.d.a--; 888 uint64_t PrevType = MachO->getRelocationType(RelPrev); 889 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 890 return true; 891 } 892 } 893 894 return false; 895 } 896 897 namespace { 898 class SourcePrinter { 899 protected: 900 DILineInfo OldLineInfo; 901 const ObjectFile *Obj = nullptr; 902 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 903 // File name to file contents of source 904 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 905 // Mark the line endings of the cached source 906 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 907 908 private: 909 bool cacheSource(const DILineInfo& LineInfoFile); 910 911 public: 912 SourcePrinter() = default; 913 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 914 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 915 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 916 DefaultArch); 917 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 918 } 919 virtual ~SourcePrinter() = default; 920 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 921 StringRef Delimiter = "; "); 922 }; 923 924 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 925 std::unique_ptr<MemoryBuffer> Buffer; 926 if (LineInfo.Source) { 927 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 928 } else { 929 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 930 if (!BufferOrError) 931 return false; 932 Buffer = std::move(*BufferOrError); 933 } 934 // Chomp the file to get lines 935 size_t BufferSize = Buffer->getBufferSize(); 936 const char *BufferStart = Buffer->getBufferStart(); 937 for (const char *Start = BufferStart, *End = BufferStart; 938 End < BufferStart + BufferSize; End++) 939 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 940 (*End == '\r' && *(End + 1) == '\n')) { 941 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 942 if (*End == '\r') 943 End++; 944 Start = End + 1; 945 } 946 SourceCache[LineInfo.FileName] = std::move(Buffer); 947 return true; 948 } 949 950 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 951 StringRef Delimiter) { 952 if (!Symbolizer) 953 return; 954 DILineInfo LineInfo = DILineInfo(); 955 auto ExpectecLineInfo = 956 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 957 if (!ExpectecLineInfo) 958 consumeError(ExpectecLineInfo.takeError()); 959 else 960 LineInfo = *ExpectecLineInfo; 961 962 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 963 LineInfo.Line == 0) 964 return; 965 966 if (PrintLines) 967 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 968 if (PrintSource) { 969 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 970 if (!cacheSource(LineInfo)) 971 return; 972 auto FileBuffer = SourceCache.find(LineInfo.FileName); 973 if (FileBuffer != SourceCache.end()) { 974 auto LineBuffer = LineCache.find(LineInfo.FileName); 975 if (LineBuffer != LineCache.end()) { 976 if (LineInfo.Line > LineBuffer->second.size()) 977 return; 978 // Vector begins at 0, line numbers are non-zero 979 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 980 << "\n"; 981 } 982 } 983 } 984 OldLineInfo = LineInfo; 985 } 986 987 static bool isArmElf(const ObjectFile *Obj) { 988 return (Obj->isELF() && 989 (Obj->getArch() == Triple::aarch64 || 990 Obj->getArch() == Triple::aarch64_be || 991 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 992 Obj->getArch() == Triple::thumb || 993 Obj->getArch() == Triple::thumbeb)); 994 } 995 996 class PrettyPrinter { 997 public: 998 virtual ~PrettyPrinter() = default; 999 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 1000 ArrayRef<uint8_t> Bytes, uint64_t Address, 1001 raw_ostream &OS, StringRef Annot, 1002 MCSubtargetInfo const &STI, SourcePrinter *SP, 1003 std::vector<RelocationRef> *Rels = nullptr) { 1004 if (SP && (PrintSource || PrintLines)) 1005 SP->printSourceLine(OS, Address); 1006 if (!NoLeadingAddr) 1007 OS << format("%8" PRIx64 ":", Address); 1008 if (!NoShowRawInsn) { 1009 OS << "\t"; 1010 dumpBytes(Bytes, OS); 1011 } 1012 if (MI) 1013 IP.printInst(MI, OS, "", STI); 1014 else 1015 OS << " <unknown>"; 1016 } 1017 }; 1018 PrettyPrinter PrettyPrinterInst; 1019 class HexagonPrettyPrinter : public PrettyPrinter { 1020 public: 1021 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 1022 raw_ostream &OS) { 1023 uint32_t opcode = 1024 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 1025 if (!NoLeadingAddr) 1026 OS << format("%8" PRIx64 ":", Address); 1027 if (!NoShowRawInsn) { 1028 OS << "\t"; 1029 dumpBytes(Bytes.slice(0, 4), OS); 1030 OS << format("%08" PRIx32, opcode); 1031 } 1032 } 1033 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1034 uint64_t Address, raw_ostream &OS, StringRef Annot, 1035 MCSubtargetInfo const &STI, SourcePrinter *SP, 1036 std::vector<RelocationRef> *Rels) override { 1037 if (SP && (PrintSource || PrintLines)) 1038 SP->printSourceLine(OS, Address, ""); 1039 if (!MI) { 1040 printLead(Bytes, Address, OS); 1041 OS << " <unknown>"; 1042 return; 1043 } 1044 std::string Buffer; 1045 { 1046 raw_string_ostream TempStream(Buffer); 1047 IP.printInst(MI, TempStream, "", STI); 1048 } 1049 StringRef Contents(Buffer); 1050 // Split off bundle attributes 1051 auto PacketBundle = Contents.rsplit('\n'); 1052 // Split off first instruction from the rest 1053 auto HeadTail = PacketBundle.first.split('\n'); 1054 auto Preamble = " { "; 1055 auto Separator = ""; 1056 StringRef Fmt = "\t\t\t%08" PRIx64 ": "; 1057 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 1058 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 1059 1060 // Hexagon's packets require relocations to be inline rather than 1061 // clustered at the end of the packet. 1062 auto PrintReloc = [&]() -> void { 1063 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address)) { 1064 if (RelCur->getOffset() == Address) { 1065 SmallString<16> Name; 1066 SmallString<32> Val; 1067 RelCur->getTypeName(Name); 1068 error(getRelocationValueString(*RelCur, Val)); 1069 OS << Separator << format(Fmt.data(), Address) << Name << "\t" << Val 1070 << "\n"; 1071 return; 1072 } 1073 ++RelCur; 1074 } 1075 }; 1076 1077 while (!HeadTail.first.empty()) { 1078 OS << Separator; 1079 Separator = "\n"; 1080 if (SP && (PrintSource || PrintLines)) 1081 SP->printSourceLine(OS, Address, ""); 1082 printLead(Bytes, Address, OS); 1083 OS << Preamble; 1084 Preamble = " "; 1085 StringRef Inst; 1086 auto Duplex = HeadTail.first.split('\v'); 1087 if (!Duplex.second.empty()) { 1088 OS << Duplex.first; 1089 OS << "; "; 1090 Inst = Duplex.second; 1091 } 1092 else 1093 Inst = HeadTail.first; 1094 OS << Inst; 1095 HeadTail = HeadTail.second.split('\n'); 1096 if (HeadTail.first.empty()) 1097 OS << " } " << PacketBundle.second; 1098 PrintReloc(); 1099 Bytes = Bytes.slice(4); 1100 Address += 4; 1101 } 1102 } 1103 }; 1104 HexagonPrettyPrinter HexagonPrettyPrinterInst; 1105 1106 class AMDGCNPrettyPrinter : public PrettyPrinter { 1107 public: 1108 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1109 uint64_t Address, raw_ostream &OS, StringRef Annot, 1110 MCSubtargetInfo const &STI, SourcePrinter *SP, 1111 std::vector<RelocationRef> *Rels) override { 1112 if (SP && (PrintSource || PrintLines)) 1113 SP->printSourceLine(OS, Address); 1114 1115 typedef support::ulittle32_t U32; 1116 1117 if (MI) { 1118 SmallString<40> InstStr; 1119 raw_svector_ostream IS(InstStr); 1120 1121 IP.printInst(MI, IS, "", STI); 1122 1123 OS << left_justify(IS.str(), 60); 1124 } else { 1125 // an unrecognized encoding - this is probably data so represent it 1126 // using the .long directive, or .byte directive if fewer than 4 bytes 1127 // remaining 1128 if (Bytes.size() >= 4) { 1129 OS << format("\t.long 0x%08" PRIx32 " ", 1130 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 1131 OS.indent(42); 1132 } else { 1133 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 1134 for (unsigned int i = 1; i < Bytes.size(); i++) 1135 OS << format(", 0x%02" PRIx8, Bytes[i]); 1136 OS.indent(55 - (6 * Bytes.size())); 1137 } 1138 } 1139 1140 OS << format("// %012" PRIX64 ": ", Address); 1141 if (Bytes.size() >=4) { 1142 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 1143 Bytes.size() / sizeof(U32))) 1144 // D should be explicitly casted to uint32_t here as it is passed 1145 // by format to snprintf as vararg. 1146 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 1147 } else { 1148 for (unsigned int i = 0; i < Bytes.size(); i++) 1149 OS << format("%02" PRIX8 " ", Bytes[i]); 1150 } 1151 1152 if (!Annot.empty()) 1153 OS << "// " << Annot; 1154 } 1155 }; 1156 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 1157 1158 class BPFPrettyPrinter : public PrettyPrinter { 1159 public: 1160 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1161 uint64_t Address, raw_ostream &OS, StringRef Annot, 1162 MCSubtargetInfo const &STI, SourcePrinter *SP, 1163 std::vector<RelocationRef> *Rels) override { 1164 if (SP && (PrintSource || PrintLines)) 1165 SP->printSourceLine(OS, Address); 1166 if (!NoLeadingAddr) 1167 OS << format("%8" PRId64 ":", Address / 8); 1168 if (!NoShowRawInsn) { 1169 OS << "\t"; 1170 dumpBytes(Bytes, OS); 1171 } 1172 if (MI) 1173 IP.printInst(MI, OS, "", STI); 1174 else 1175 OS << " <unknown>"; 1176 } 1177 }; 1178 BPFPrettyPrinter BPFPrettyPrinterInst; 1179 1180 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 1181 switch(Triple.getArch()) { 1182 default: 1183 return PrettyPrinterInst; 1184 case Triple::hexagon: 1185 return HexagonPrettyPrinterInst; 1186 case Triple::amdgcn: 1187 return AMDGCNPrettyPrinterInst; 1188 case Triple::bpfel: 1189 case Triple::bpfeb: 1190 return BPFPrettyPrinterInst; 1191 } 1192 } 1193 } 1194 1195 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1196 assert(Obj->isELF()); 1197 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1198 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1199 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1200 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1201 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1202 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1203 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1204 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1205 llvm_unreachable("Unsupported binary format"); 1206 } 1207 1208 template <class ELFT> static void 1209 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1210 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1211 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1212 uint8_t SymbolType = Symbol.getELFType(); 1213 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1214 continue; 1215 1216 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1217 if (!AddressOrErr) 1218 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1219 1220 Expected<StringRef> Name = Symbol.getName(); 1221 if (!Name) 1222 report_error(Obj->getFileName(), Name.takeError()); 1223 if (Name->empty()) 1224 continue; 1225 1226 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1227 if (!SectionOrErr) 1228 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1229 section_iterator SecI = *SectionOrErr; 1230 if (SecI == Obj->section_end()) 1231 continue; 1232 1233 AllSymbols[*SecI].emplace_back(*AddressOrErr, *Name, SymbolType); 1234 } 1235 } 1236 1237 static void 1238 addDynamicElfSymbols(const ObjectFile *Obj, 1239 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1240 assert(Obj->isELF()); 1241 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1242 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1243 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1244 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1245 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1246 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1247 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1248 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1249 else 1250 llvm_unreachable("Unsupported binary format"); 1251 } 1252 1253 static void addPltEntries(const ObjectFile *Obj, 1254 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 1255 StringSaver &Saver) { 1256 Optional<SectionRef> Plt = None; 1257 for (const SectionRef &Section : Obj->sections()) { 1258 StringRef Name; 1259 if (Section.getName(Name)) 1260 continue; 1261 if (Name == ".plt") 1262 Plt = Section; 1263 } 1264 if (!Plt) 1265 return; 1266 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1267 for (auto PltEntry : ElfObj->getPltAddresses()) { 1268 SymbolRef Symbol(PltEntry.first, ElfObj); 1269 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1270 1271 Expected<StringRef> NameOrErr = Symbol.getName(); 1272 if (!NameOrErr) 1273 report_error(Obj->getFileName(), NameOrErr.takeError()); 1274 if (NameOrErr->empty()) 1275 continue; 1276 StringRef Name = Saver.save((*NameOrErr + "@plt").str()); 1277 1278 AllSymbols[*Plt].emplace_back(PltEntry.second, Name, SymbolType); 1279 } 1280 } 1281 } 1282 1283 // Normally the disassembly output will skip blocks of zeroes. This function 1284 // returns the number of zero bytes that can be skipped when dumping the 1285 // disassembly of the instructions in Buf. 1286 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 1287 // When -z or --disassemble-zeroes are given we always dissasemble them. 1288 if (DisassembleZeroes) 1289 return 0; 1290 1291 // Find the number of leading zeroes. 1292 size_t N = 0; 1293 while (N < Buf.size() && !Buf[N]) 1294 ++N; 1295 1296 // We may want to skip blocks of zero bytes, but unless we see 1297 // at least 8 of them in a row. 1298 if (N < 8) 1299 return 0; 1300 1301 // We skip zeroes in multiples of 4 because do not want to truncate an 1302 // instruction if it starts with a zero byte. 1303 return N & ~0x3; 1304 } 1305 1306 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1307 if (StartAddress > StopAddress) 1308 error("Start address should be less than stop address"); 1309 1310 const Target *TheTarget = getTarget(Obj); 1311 1312 // Package up features to be passed to target/subtarget 1313 SubtargetFeatures Features = Obj->getFeatures(); 1314 if (!MAttrs.empty()) 1315 for (unsigned I = 0; I != MAttrs.size(); ++I) 1316 Features.AddFeature(MAttrs[I]); 1317 1318 std::unique_ptr<const MCRegisterInfo> MRI( 1319 TheTarget->createMCRegInfo(TripleName)); 1320 if (!MRI) 1321 report_error(Obj->getFileName(), "no register info for target " + 1322 TripleName); 1323 1324 // Set up disassembler. 1325 std::unique_ptr<const MCAsmInfo> AsmInfo( 1326 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1327 if (!AsmInfo) 1328 report_error(Obj->getFileName(), "no assembly info for target " + 1329 TripleName); 1330 std::unique_ptr<const MCSubtargetInfo> STI( 1331 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1332 if (!STI) 1333 report_error(Obj->getFileName(), "no subtarget info for target " + 1334 TripleName); 1335 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1336 if (!MII) 1337 report_error(Obj->getFileName(), "no instruction info for target " + 1338 TripleName); 1339 MCObjectFileInfo MOFI; 1340 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1341 // FIXME: for now initialize MCObjectFileInfo with default values 1342 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1343 1344 std::unique_ptr<MCDisassembler> DisAsm( 1345 TheTarget->createMCDisassembler(*STI, Ctx)); 1346 if (!DisAsm) 1347 report_error(Obj->getFileName(), "no disassembler for target " + 1348 TripleName); 1349 1350 std::unique_ptr<const MCInstrAnalysis> MIA( 1351 TheTarget->createMCInstrAnalysis(MII.get())); 1352 1353 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1354 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1355 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1356 if (!IP) 1357 report_error(Obj->getFileName(), "no instruction printer for target " + 1358 TripleName); 1359 IP->setPrintImmHex(PrintImmHex); 1360 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1361 1362 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1363 "\t\t\t%08" PRIx64 ": "; 1364 1365 SourcePrinter SP(Obj, TheTarget->getName()); 1366 1367 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1368 // in RelocSecs contain the relocations for section S. 1369 std::error_code EC; 1370 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1371 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1372 section_iterator Sec2 = Section.getRelocatedSection(); 1373 if (Sec2 != Obj->section_end()) 1374 SectionRelocMap[*Sec2].push_back(Section); 1375 } 1376 1377 // Create a mapping from virtual address to symbol name. This is used to 1378 // pretty print the symbols while disassembling. 1379 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1380 SectionSymbolsTy AbsoluteSymbols; 1381 for (const SymbolRef &Symbol : Obj->symbols()) { 1382 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1383 if (!AddressOrErr) 1384 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1385 uint64_t Address = *AddressOrErr; 1386 1387 Expected<StringRef> Name = Symbol.getName(); 1388 if (!Name) 1389 report_error(Obj->getFileName(), Name.takeError()); 1390 if (Name->empty()) 1391 continue; 1392 1393 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1394 if (!SectionOrErr) 1395 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1396 1397 uint8_t SymbolType = ELF::STT_NOTYPE; 1398 if (Obj->isELF()) 1399 SymbolType = getElfSymbolType(Obj, Symbol); 1400 1401 section_iterator SecI = *SectionOrErr; 1402 if (SecI != Obj->section_end()) 1403 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1404 else 1405 AbsoluteSymbols.emplace_back(Address, *Name, SymbolType); 1406 1407 1408 } 1409 if (AllSymbols.empty() && Obj->isELF()) 1410 addDynamicElfSymbols(Obj, AllSymbols); 1411 1412 BumpPtrAllocator A; 1413 StringSaver Saver(A); 1414 addPltEntries(Obj, AllSymbols, Saver); 1415 1416 // Create a mapping from virtual address to section. 1417 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1418 for (SectionRef Sec : Obj->sections()) 1419 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1420 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1421 1422 // Linked executables (.exe and .dll files) typically don't include a real 1423 // symbol table but they might contain an export table. 1424 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1425 for (const auto &ExportEntry : COFFObj->export_directories()) { 1426 StringRef Name; 1427 error(ExportEntry.getSymbolName(Name)); 1428 if (Name.empty()) 1429 continue; 1430 uint32_t RVA; 1431 error(ExportEntry.getExportRVA(RVA)); 1432 1433 uint64_t VA = COFFObj->getImageBase() + RVA; 1434 auto Sec = std::upper_bound( 1435 SectionAddresses.begin(), SectionAddresses.end(), VA, 1436 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1437 return LHS < RHS.first; 1438 }); 1439 if (Sec != SectionAddresses.begin()) 1440 --Sec; 1441 else 1442 Sec = SectionAddresses.end(); 1443 1444 if (Sec != SectionAddresses.end()) 1445 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1446 else 1447 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1448 } 1449 } 1450 1451 // Sort all the symbols, this allows us to use a simple binary search to find 1452 // a symbol near an address. 1453 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1454 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1455 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1456 1457 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1458 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1459 continue; 1460 1461 uint64_t SectionAddr = Section.getAddress(); 1462 uint64_t SectSize = Section.getSize(); 1463 if (!SectSize) 1464 continue; 1465 1466 // Get the list of all the symbols in this section. 1467 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1468 std::vector<uint64_t> DataMappingSymsAddr; 1469 std::vector<uint64_t> TextMappingSymsAddr; 1470 if (isArmElf(Obj)) { 1471 for (const auto &Symb : Symbols) { 1472 uint64_t Address = Symb.Address; 1473 StringRef Name = Symb.Name; 1474 if (Name.startswith("$d")) 1475 DataMappingSymsAddr.push_back(Address - SectionAddr); 1476 if (Name.startswith("$x")) 1477 TextMappingSymsAddr.push_back(Address - SectionAddr); 1478 if (Name.startswith("$a")) 1479 TextMappingSymsAddr.push_back(Address - SectionAddr); 1480 if (Name.startswith("$t")) 1481 TextMappingSymsAddr.push_back(Address - SectionAddr); 1482 } 1483 } 1484 1485 llvm::sort(DataMappingSymsAddr); 1486 llvm::sort(TextMappingSymsAddr); 1487 1488 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1489 // AMDGPU disassembler uses symbolizer for printing labels 1490 std::unique_ptr<MCRelocationInfo> RelInfo( 1491 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1492 if (RelInfo) { 1493 std::unique_ptr<MCSymbolizer> Symbolizer( 1494 TheTarget->createMCSymbolizer( 1495 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1496 DisAsm->setSymbolizer(std::move(Symbolizer)); 1497 } 1498 } 1499 1500 // Make a list of all the relocations for this section. 1501 std::vector<RelocationRef> Rels; 1502 if (InlineRelocs) { 1503 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1504 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1505 Rels.push_back(Reloc); 1506 } 1507 } 1508 } 1509 1510 // Sort relocations by address. 1511 llvm::sort(Rels, isRelocAddressLess); 1512 1513 StringRef SegmentName = ""; 1514 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1515 DataRefImpl DR = Section.getRawDataRefImpl(); 1516 SegmentName = MachO->getSectionFinalSegmentName(DR); 1517 } 1518 StringRef SectionName; 1519 error(Section.getName(SectionName)); 1520 1521 // If the section has no symbol at the start, just insert a dummy one. 1522 if (Symbols.empty() || Symbols[0].Address != 0) { 1523 Symbols.insert( 1524 Symbols.begin(), 1525 SectionSymbol(SectionAddr, SectionName, 1526 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1527 } 1528 1529 SmallString<40> Comments; 1530 raw_svector_ostream CommentStream(Comments); 1531 1532 StringRef BytesStr; 1533 error(Section.getContents(BytesStr)); 1534 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1535 BytesStr.size()); 1536 1537 uint64_t Size; 1538 uint64_t Index; 1539 bool PrintedSection = false; 1540 1541 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1542 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1543 // Disassemble symbol by symbol. 1544 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1545 uint64_t Start = Symbols[SI].Address - SectionAddr; 1546 // The end is either the section end or the beginning of the next 1547 // symbol. 1548 uint64_t End = (SI == SE - 1) 1549 ? SectSize 1550 : Symbols[SI + 1].Address - SectionAddr; 1551 // Don't try to disassemble beyond the end of section contents. 1552 if (End > SectSize) 1553 End = SectSize; 1554 // If this symbol has the same address as the next symbol, then skip it. 1555 if (Start >= End) 1556 continue; 1557 1558 // Check if we need to skip symbol 1559 // Skip if the symbol's data is not between StartAddress and StopAddress 1560 if (End + SectionAddr < StartAddress || 1561 Start + SectionAddr > StopAddress) { 1562 continue; 1563 } 1564 1565 /// Skip if user requested specific symbols and this is not in the list 1566 if (!DisasmFuncsSet.empty() && !DisasmFuncsSet.count(Symbols[SI].Name)) 1567 continue; 1568 1569 if (!PrintedSection) { 1570 PrintedSection = true; 1571 outs() << "Disassembly of section "; 1572 if (!SegmentName.empty()) 1573 outs() << SegmentName << ","; 1574 outs() << SectionName << ':'; 1575 } 1576 1577 // Stop disassembly at the stop address specified 1578 if (End + SectionAddr > StopAddress) 1579 End = StopAddress - SectionAddr; 1580 1581 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1582 if (Symbols[SI].Type == ELF::STT_AMDGPU_HSA_KERNEL) { 1583 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1584 Start += 256; 1585 } 1586 if (SI == SE - 1 || 1587 Symbols[SI + 1].Type == ELF::STT_AMDGPU_HSA_KERNEL) { 1588 // cut trailing zeroes at the end of kernel 1589 // cut up to 256 bytes 1590 const uint64_t EndAlign = 256; 1591 const auto Limit = End - (std::min)(EndAlign, End - Start); 1592 while (End > Limit && 1593 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1594 End -= 4; 1595 } 1596 } 1597 1598 outs() << '\n'; 1599 if (!NoLeadingAddr) 1600 outs() << format("%016" PRIx64 " ", SectionAddr + Start); 1601 1602 StringRef SymbolName = Symbols[SI].Name; 1603 if (Demangle) 1604 outs() << demangle(SymbolName) << ":\n"; 1605 else 1606 outs() << SymbolName << ":\n"; 1607 1608 // Don't print raw contents of a virtual section. A virtual section 1609 // doesn't have any contents in the file. 1610 if (Section.isVirtual()) { 1611 outs() << "...\n"; 1612 continue; 1613 } 1614 1615 #ifndef NDEBUG 1616 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1617 #else 1618 raw_ostream &DebugOut = nulls(); 1619 #endif 1620 1621 // Some targets (like WebAssembly) have a special prelude at the start 1622 // of each symbol. 1623 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1624 SectionAddr + Start, DebugOut, CommentStream); 1625 Start += Size; 1626 1627 for (Index = Start; Index < End; Index += Size) { 1628 MCInst Inst; 1629 1630 if (Index + SectionAddr < StartAddress || 1631 Index + SectionAddr > StopAddress) { 1632 // skip byte by byte till StartAddress is reached 1633 Size = 1; 1634 continue; 1635 } 1636 // AArch64 ELF binaries can interleave data and text in the 1637 // same section. We rely on the markers introduced to 1638 // understand what we need to dump. If the data marker is within a 1639 // function, it is denoted as a word/short etc 1640 if (isArmElf(Obj) && Symbols[SI].Type != ELF::STT_OBJECT && 1641 !DisassembleAll) { 1642 uint64_t Stride = 0; 1643 1644 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1645 DataMappingSymsAddr.end(), Index); 1646 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1647 // Switch to data. 1648 while (Index < End) { 1649 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1650 outs() << "\t"; 1651 if (Index + 4 <= End) { 1652 Stride = 4; 1653 dumpBytes(Bytes.slice(Index, 4), outs()); 1654 outs() << "\t.word\t"; 1655 uint32_t Data = 0; 1656 if (Obj->isLittleEndian()) { 1657 const auto Word = 1658 reinterpret_cast<const support::ulittle32_t *>( 1659 Bytes.data() + Index); 1660 Data = *Word; 1661 } else { 1662 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1663 Bytes.data() + Index); 1664 Data = *Word; 1665 } 1666 outs() << "0x" << format("%08" PRIx32, Data); 1667 } else if (Index + 2 <= End) { 1668 Stride = 2; 1669 dumpBytes(Bytes.slice(Index, 2), outs()); 1670 outs() << "\t\t.short\t"; 1671 uint16_t Data = 0; 1672 if (Obj->isLittleEndian()) { 1673 const auto Short = 1674 reinterpret_cast<const support::ulittle16_t *>( 1675 Bytes.data() + Index); 1676 Data = *Short; 1677 } else { 1678 const auto Short = 1679 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1680 Index); 1681 Data = *Short; 1682 } 1683 outs() << "0x" << format("%04" PRIx16, Data); 1684 } else { 1685 Stride = 1; 1686 dumpBytes(Bytes.slice(Index, 1), outs()); 1687 outs() << "\t\t.byte\t"; 1688 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1689 } 1690 Index += Stride; 1691 outs() << "\n"; 1692 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1693 TextMappingSymsAddr.end(), Index); 1694 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1695 break; 1696 } 1697 } 1698 } 1699 1700 // If there is a data symbol inside an ELF text section and we are only 1701 // disassembling text (applicable all architectures), 1702 // we are in a situation where we must print the data and not 1703 // disassemble it. 1704 if (Obj->isELF() && Symbols[SI].Type == ELF::STT_OBJECT && 1705 !DisassembleAll && Section.isText()) { 1706 // print out data up to 8 bytes at a time in hex and ascii 1707 uint8_t AsciiData[9] = {'\0'}; 1708 uint8_t Byte; 1709 int NumBytes = 0; 1710 1711 for (Index = Start; Index < End; Index += 1) { 1712 if (((SectionAddr + Index) < StartAddress) || 1713 ((SectionAddr + Index) > StopAddress)) 1714 continue; 1715 if (NumBytes == 0) { 1716 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1717 outs() << "\t"; 1718 } 1719 Byte = Bytes.slice(Index)[0]; 1720 outs() << format(" %02x", Byte); 1721 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1722 1723 uint8_t IndentOffset = 0; 1724 NumBytes++; 1725 if (Index == End - 1 || NumBytes > 8) { 1726 // Indent the space for less than 8 bytes data. 1727 // 2 spaces for byte and one for space between bytes 1728 IndentOffset = 3 * (8 - NumBytes); 1729 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1730 AsciiData[Excess] = '\0'; 1731 NumBytes = 8; 1732 } 1733 if (NumBytes == 8) { 1734 AsciiData[8] = '\0'; 1735 outs() << std::string(IndentOffset, ' ') << " "; 1736 outs() << reinterpret_cast<char *>(AsciiData); 1737 outs() << '\n'; 1738 NumBytes = 0; 1739 } 1740 } 1741 } 1742 if (Index >= End) 1743 break; 1744 1745 if (size_t N = 1746 countSkippableZeroBytes(Bytes.slice(Index, End - Index))) { 1747 outs() << "\t\t..." << '\n'; 1748 Index += N; 1749 if (Index >= End) 1750 break; 1751 } 1752 1753 // Disassemble a real instruction or a data when disassemble all is 1754 // provided 1755 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1756 SectionAddr + Index, DebugOut, 1757 CommentStream); 1758 if (Size == 0) 1759 Size = 1; 1760 1761 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1762 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1763 *STI, &SP, &Rels); 1764 outs() << CommentStream.str(); 1765 Comments.clear(); 1766 1767 // Try to resolve the target of a call, tail call, etc. to a specific 1768 // symbol. 1769 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1770 MIA->isConditionalBranch(Inst))) { 1771 uint64_t Target; 1772 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1773 // In a relocatable object, the target's section must reside in 1774 // the same section as the call instruction or it is accessed 1775 // through a relocation. 1776 // 1777 // In a non-relocatable object, the target may be in any section. 1778 // 1779 // N.B. We don't walk the relocations in the relocatable case yet. 1780 auto *TargetSectionSymbols = &Symbols; 1781 if (!Obj->isRelocatableObject()) { 1782 auto SectionAddress = std::upper_bound( 1783 SectionAddresses.begin(), SectionAddresses.end(), Target, 1784 [](uint64_t LHS, 1785 const std::pair<uint64_t, SectionRef> &RHS) { 1786 return LHS < RHS.first; 1787 }); 1788 if (SectionAddress != SectionAddresses.begin()) { 1789 --SectionAddress; 1790 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1791 } else { 1792 TargetSectionSymbols = &AbsoluteSymbols; 1793 } 1794 } 1795 1796 // Find the first symbol in the section whose offset is less than 1797 // or equal to the target. If there isn't a section that contains 1798 // the target, find the nearest preceding absolute symbol. 1799 auto TargetSym = std::upper_bound( 1800 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1801 Target, [](uint64_t LHS, const SectionSymbol &RHS) { 1802 return LHS < RHS.Address; 1803 }); 1804 if (TargetSym == TargetSectionSymbols->begin()) { 1805 TargetSectionSymbols = &AbsoluteSymbols; 1806 TargetSym = std::upper_bound( 1807 AbsoluteSymbols.begin(), AbsoluteSymbols.end(), Target, 1808 [](uint64_t LHS, const SectionSymbol &RHS) { 1809 return LHS < RHS.Address; 1810 }); 1811 } 1812 if (TargetSym != TargetSectionSymbols->begin()) { 1813 --TargetSym; 1814 outs() << " <" << TargetSym->Name; 1815 uint64_t Disp = Target - TargetSym->Address; 1816 if (Disp) 1817 outs() << "+0x" << Twine::utohexstr(Disp); 1818 outs() << '>'; 1819 } 1820 } 1821 } 1822 outs() << "\n"; 1823 1824 // Hexagon does this in pretty printer 1825 if (Obj->getArch() != Triple::hexagon) 1826 // Print relocation for instruction. 1827 while (RelCur != RelEnd) { 1828 uint64_t Addr = RelCur->getOffset(); 1829 SmallString<16> Name; 1830 SmallString<32> Val; 1831 1832 // If this relocation is hidden, skip it. 1833 if (getHidden(*RelCur) || ((SectionAddr + Addr) < StartAddress)) { 1834 ++RelCur; 1835 continue; 1836 } 1837 1838 // Stop when rel_cur's address is past the current instruction. 1839 if (Addr >= Index + Size) 1840 break; 1841 RelCur->getTypeName(Name); 1842 error(getRelocationValueString(*RelCur, Val)); 1843 outs() << format(Fmt.data(), SectionAddr + Addr) << Name << "\t" 1844 << Val << "\n"; 1845 ++RelCur; 1846 } 1847 } 1848 } 1849 } 1850 } 1851 1852 void llvm::printRelocations(const ObjectFile *Obj) { 1853 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1854 "%08" PRIx64; 1855 // Regular objdump doesn't print relocations in non-relocatable object 1856 // files. 1857 if (!Obj->isRelocatableObject()) 1858 return; 1859 1860 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1861 if (Section.relocation_begin() == Section.relocation_end()) 1862 continue; 1863 StringRef SecName; 1864 error(Section.getName(SecName)); 1865 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1866 for (const RelocationRef &Reloc : Section.relocations()) { 1867 uint64_t Address = Reloc.getOffset(); 1868 SmallString<32> RelocName; 1869 SmallString<32> ValueStr; 1870 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1871 continue; 1872 Reloc.getTypeName(RelocName); 1873 error(getRelocationValueString(Reloc, ValueStr)); 1874 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1875 << ValueStr << "\n"; 1876 } 1877 outs() << "\n"; 1878 } 1879 } 1880 1881 void llvm::printDynamicRelocations(const ObjectFile *Obj) { 1882 // For the moment, this option is for ELF only 1883 if (!Obj->isELF()) 1884 return; 1885 1886 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1887 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1888 error("not a dynamic object"); 1889 return; 1890 } 1891 1892 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1893 if (DynRelSec.empty()) 1894 return; 1895 1896 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1897 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1898 for (const SectionRef &Section : DynRelSec) { 1899 if (Section.relocation_begin() == Section.relocation_end()) 1900 continue; 1901 for (const RelocationRef &Reloc : Section.relocations()) { 1902 uint64_t Address = Reloc.getOffset(); 1903 SmallString<32> RelocName; 1904 SmallString<32> ValueStr; 1905 Reloc.getTypeName(RelocName); 1906 error(getRelocationValueString(Reloc, ValueStr)); 1907 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1908 << ValueStr << "\n"; 1909 } 1910 } 1911 } 1912 1913 void llvm::printSectionHeaders(const ObjectFile *Obj) { 1914 outs() << "Sections:\n" 1915 "Idx Name Size Address Type\n"; 1916 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1917 StringRef Name; 1918 error(Section.getName(Name)); 1919 uint64_t Address = Section.getAddress(); 1920 uint64_t Size = Section.getSize(); 1921 bool Text = Section.isText(); 1922 bool Data = Section.isData(); 1923 bool BSS = Section.isBSS(); 1924 std::string Type = (std::string(Text ? "TEXT " : "") + 1925 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1926 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1927 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1928 Address, Type.c_str()); 1929 } 1930 outs() << "\n"; 1931 } 1932 1933 void llvm::printSectionContents(const ObjectFile *Obj) { 1934 std::error_code EC; 1935 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1936 StringRef Name; 1937 StringRef Contents; 1938 error(Section.getName(Name)); 1939 uint64_t BaseAddr = Section.getAddress(); 1940 uint64_t Size = Section.getSize(); 1941 if (!Size) 1942 continue; 1943 1944 outs() << "Contents of section " << Name << ":\n"; 1945 if (Section.isBSS()) { 1946 outs() << format("<skipping contents of bss section at [%04" PRIx64 1947 ", %04" PRIx64 ")>\n", 1948 BaseAddr, BaseAddr + Size); 1949 continue; 1950 } 1951 1952 error(Section.getContents(Contents)); 1953 1954 // Dump out the content as hex and printable ascii characters. 1955 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1956 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1957 // Dump line of hex. 1958 for (std::size_t I = 0; I < 16; ++I) { 1959 if (I != 0 && I % 4 == 0) 1960 outs() << ' '; 1961 if (Addr + I < End) 1962 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1963 << hexdigit(Contents[Addr + I] & 0xF, true); 1964 else 1965 outs() << " "; 1966 } 1967 // Print ascii. 1968 outs() << " "; 1969 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1970 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1971 outs() << Contents[Addr + I]; 1972 else 1973 outs() << "."; 1974 } 1975 outs() << "\n"; 1976 } 1977 } 1978 } 1979 1980 void llvm::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1981 StringRef ArchitectureName) { 1982 outs() << "SYMBOL TABLE:\n"; 1983 1984 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1985 printCOFFSymbolTable(Coff); 1986 return; 1987 } 1988 1989 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1990 // Skip printing the special zero symbol when dumping an ELF file. 1991 // This makes the output consistent with the GNU objdump. 1992 if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O)) 1993 continue; 1994 1995 const SymbolRef &Symbol = *I; 1996 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1997 if (!AddressOrError) 1998 report_error(ArchiveName, O->getFileName(), AddressOrError.takeError(), 1999 ArchitectureName); 2000 uint64_t Address = *AddressOrError; 2001 if ((Address < StartAddress) || (Address > StopAddress)) 2002 continue; 2003 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 2004 if (!TypeOrError) 2005 report_error(ArchiveName, O->getFileName(), TypeOrError.takeError(), 2006 ArchitectureName); 2007 SymbolRef::Type Type = *TypeOrError; 2008 uint32_t Flags = Symbol.getFlags(); 2009 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 2010 if (!SectionOrErr) 2011 report_error(ArchiveName, O->getFileName(), SectionOrErr.takeError(), 2012 ArchitectureName); 2013 section_iterator Section = *SectionOrErr; 2014 StringRef Name; 2015 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 2016 Section->getName(Name); 2017 } else { 2018 Expected<StringRef> NameOrErr = Symbol.getName(); 2019 if (!NameOrErr) 2020 report_error(ArchiveName, O->getFileName(), NameOrErr.takeError(), 2021 ArchitectureName); 2022 Name = *NameOrErr; 2023 } 2024 2025 bool Global = Flags & SymbolRef::SF_Global; 2026 bool Weak = Flags & SymbolRef::SF_Weak; 2027 bool Absolute = Flags & SymbolRef::SF_Absolute; 2028 bool Common = Flags & SymbolRef::SF_Common; 2029 bool Hidden = Flags & SymbolRef::SF_Hidden; 2030 2031 char GlobLoc = ' '; 2032 if (Type != SymbolRef::ST_Unknown) 2033 GlobLoc = Global ? 'g' : 'l'; 2034 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2035 ? 'd' : ' '; 2036 char FileFunc = ' '; 2037 if (Type == SymbolRef::ST_File) 2038 FileFunc = 'f'; 2039 else if (Type == SymbolRef::ST_Function) 2040 FileFunc = 'F'; 2041 else if (Type == SymbolRef::ST_Data) 2042 FileFunc = 'O'; 2043 2044 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 2045 "%08" PRIx64; 2046 2047 outs() << format(Fmt, Address) << " " 2048 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2049 << (Weak ? 'w' : ' ') // Weak? 2050 << ' ' // Constructor. Not supported yet. 2051 << ' ' // Warning. Not supported yet. 2052 << ' ' // Indirect reference to another symbol. 2053 << Debug // Debugging (d) or dynamic (D) symbol. 2054 << FileFunc // Name of function (F), file (f) or object (O). 2055 << ' '; 2056 if (Absolute) { 2057 outs() << "*ABS*"; 2058 } else if (Common) { 2059 outs() << "*COM*"; 2060 } else if (Section == O->section_end()) { 2061 outs() << "*UND*"; 2062 } else { 2063 if (const MachOObjectFile *MachO = 2064 dyn_cast<const MachOObjectFile>(O)) { 2065 DataRefImpl DR = Section->getRawDataRefImpl(); 2066 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 2067 outs() << SegmentName << ","; 2068 } 2069 StringRef SectionName; 2070 error(Section->getName(SectionName)); 2071 outs() << SectionName; 2072 } 2073 2074 outs() << '\t'; 2075 if (Common || isa<ELFObjectFileBase>(O)) { 2076 uint64_t Val = 2077 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2078 outs() << format("\t %08" PRIx64 " ", Val); 2079 } 2080 2081 if (Hidden) 2082 outs() << ".hidden "; 2083 2084 if (Demangle) 2085 outs() << demangle(Name) << '\n'; 2086 else 2087 outs() << Name << '\n'; 2088 } 2089 } 2090 2091 static void printUnwindInfo(const ObjectFile *O) { 2092 outs() << "Unwind info:\n\n"; 2093 2094 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 2095 printCOFFUnwindInfo(Coff); 2096 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 2097 printMachOUnwindInfo(MachO); 2098 else 2099 // TODO: Extract DWARF dump tool to objdump. 2100 WithColor::error(errs(), ToolName) 2101 << "This operation is only currently supported " 2102 "for COFF and MachO object files.\n"; 2103 } 2104 2105 void llvm::printExportsTrie(const ObjectFile *o) { 2106 outs() << "Exports trie:\n"; 2107 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2108 printMachOExportsTrie(MachO); 2109 else 2110 WithColor::error(errs(), ToolName) 2111 << "This operation is only currently supported " 2112 "for Mach-O executable files.\n"; 2113 } 2114 2115 void llvm::printRebaseTable(ObjectFile *o) { 2116 outs() << "Rebase table:\n"; 2117 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2118 printMachORebaseTable(MachO); 2119 else 2120 WithColor::error(errs(), ToolName) 2121 << "This operation is only currently supported " 2122 "for Mach-O executable files.\n"; 2123 } 2124 2125 void llvm::printBindTable(ObjectFile *o) { 2126 outs() << "Bind table:\n"; 2127 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2128 printMachOBindTable(MachO); 2129 else 2130 WithColor::error(errs(), ToolName) 2131 << "This operation is only currently supported " 2132 "for Mach-O executable files.\n"; 2133 } 2134 2135 void llvm::printLazyBindTable(ObjectFile *o) { 2136 outs() << "Lazy bind table:\n"; 2137 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2138 printMachOLazyBindTable(MachO); 2139 else 2140 WithColor::error(errs(), ToolName) 2141 << "This operation is only currently supported " 2142 "for Mach-O executable files.\n"; 2143 } 2144 2145 void llvm::printWeakBindTable(ObjectFile *o) { 2146 outs() << "Weak bind table:\n"; 2147 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2148 printMachOWeakBindTable(MachO); 2149 else 2150 WithColor::error(errs(), ToolName) 2151 << "This operation is only currently supported " 2152 "for Mach-O executable files.\n"; 2153 } 2154 2155 /// Dump the raw contents of the __clangast section so the output can be piped 2156 /// into llvm-bcanalyzer. 2157 void llvm::printRawClangAST(const ObjectFile *Obj) { 2158 if (outs().is_displayed()) { 2159 WithColor::error(errs(), ToolName) 2160 << "The -raw-clang-ast option will dump the raw binary contents of " 2161 "the clang ast section.\n" 2162 "Please redirect the output to a file or another program such as " 2163 "llvm-bcanalyzer.\n"; 2164 return; 2165 } 2166 2167 StringRef ClangASTSectionName("__clangast"); 2168 if (isa<COFFObjectFile>(Obj)) { 2169 ClangASTSectionName = "clangast"; 2170 } 2171 2172 Optional<object::SectionRef> ClangASTSection; 2173 for (auto Sec : ToolSectionFilter(*Obj)) { 2174 StringRef Name; 2175 Sec.getName(Name); 2176 if (Name == ClangASTSectionName) { 2177 ClangASTSection = Sec; 2178 break; 2179 } 2180 } 2181 if (!ClangASTSection) 2182 return; 2183 2184 StringRef ClangASTContents; 2185 error(ClangASTSection.getValue().getContents(ClangASTContents)); 2186 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2187 } 2188 2189 static void printFaultMaps(const ObjectFile *Obj) { 2190 StringRef FaultMapSectionName; 2191 2192 if (isa<ELFObjectFileBase>(Obj)) { 2193 FaultMapSectionName = ".llvm_faultmaps"; 2194 } else if (isa<MachOObjectFile>(Obj)) { 2195 FaultMapSectionName = "__llvm_faultmaps"; 2196 } else { 2197 WithColor::error(errs(), ToolName) 2198 << "This operation is only currently supported " 2199 "for ELF and Mach-O executable files.\n"; 2200 return; 2201 } 2202 2203 Optional<object::SectionRef> FaultMapSection; 2204 2205 for (auto Sec : ToolSectionFilter(*Obj)) { 2206 StringRef Name; 2207 Sec.getName(Name); 2208 if (Name == FaultMapSectionName) { 2209 FaultMapSection = Sec; 2210 break; 2211 } 2212 } 2213 2214 outs() << "FaultMap table:\n"; 2215 2216 if (!FaultMapSection.hasValue()) { 2217 outs() << "<not found>\n"; 2218 return; 2219 } 2220 2221 StringRef FaultMapContents; 2222 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2223 2224 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2225 FaultMapContents.bytes_end()); 2226 2227 outs() << FMP; 2228 } 2229 2230 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 2231 if (O->isELF()) { 2232 printELFFileHeader(O); 2233 return printELFDynamicSection(O); 2234 } 2235 if (O->isCOFF()) 2236 return printCOFFFileHeader(O); 2237 if (O->isWasm()) 2238 return printWasmFileHeader(O); 2239 if (O->isMachO()) { 2240 printMachOFileHeader(O); 2241 if (!OnlyFirst) 2242 printMachOLoadCommands(O); 2243 return; 2244 } 2245 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 2246 } 2247 2248 static void printFileHeaders(const ObjectFile *O) { 2249 if (!O->isELF() && !O->isCOFF()) 2250 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 2251 2252 Triple::ArchType AT = O->getArch(); 2253 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2254 Expected<uint64_t> StartAddrOrErr = O->getStartAddress(); 2255 if (!StartAddrOrErr) 2256 report_error(O->getFileName(), StartAddrOrErr.takeError()); 2257 2258 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2259 uint64_t Address = StartAddrOrErr.get(); 2260 outs() << "start address: " 2261 << "0x" << format(Fmt.data(), Address) << "\n\n"; 2262 } 2263 2264 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2265 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2266 if (!ModeOrErr) { 2267 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 2268 consumeError(ModeOrErr.takeError()); 2269 return; 2270 } 2271 sys::fs::perms Mode = ModeOrErr.get(); 2272 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2273 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2274 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2275 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2276 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2277 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2278 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2279 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2280 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2281 2282 outs() << " "; 2283 2284 Expected<unsigned> UIDOrErr = C.getUID(); 2285 if (!UIDOrErr) 2286 report_error(Filename, UIDOrErr.takeError()); 2287 unsigned UID = UIDOrErr.get(); 2288 outs() << format("%d/", UID); 2289 2290 Expected<unsigned> GIDOrErr = C.getGID(); 2291 if (!GIDOrErr) 2292 report_error(Filename, GIDOrErr.takeError()); 2293 unsigned GID = GIDOrErr.get(); 2294 outs() << format("%-d ", GID); 2295 2296 Expected<uint64_t> Size = C.getRawSize(); 2297 if (!Size) 2298 report_error(Filename, Size.takeError()); 2299 outs() << format("%6" PRId64, Size.get()) << " "; 2300 2301 StringRef RawLastModified = C.getRawLastModified(); 2302 unsigned Seconds; 2303 if (RawLastModified.getAsInteger(10, Seconds)) 2304 outs() << "(date: \"" << RawLastModified 2305 << "\" contains non-decimal chars) "; 2306 else { 2307 // Since ctime(3) returns a 26 character string of the form: 2308 // "Sun Sep 16 01:03:52 1973\n\0" 2309 // just print 24 characters. 2310 time_t t = Seconds; 2311 outs() << format("%.24s ", ctime(&t)); 2312 } 2313 2314 StringRef Name = ""; 2315 Expected<StringRef> NameOrErr = C.getName(); 2316 if (!NameOrErr) { 2317 consumeError(NameOrErr.takeError()); 2318 Expected<StringRef> RawNameOrErr = C.getRawName(); 2319 if (!RawNameOrErr) 2320 report_error(Filename, NameOrErr.takeError()); 2321 Name = RawNameOrErr.get(); 2322 } else { 2323 Name = NameOrErr.get(); 2324 } 2325 outs() << Name << "\n"; 2326 } 2327 2328 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 2329 const Archive::Child *C = nullptr) { 2330 // Avoid other output when using a raw option. 2331 if (!RawClangAST) { 2332 outs() << '\n'; 2333 if (A) 2334 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 2335 else 2336 outs() << O->getFileName(); 2337 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 2338 } 2339 2340 StringRef ArchiveName = A ? A->getFileName() : ""; 2341 if (FileHeaders) 2342 printFileHeaders(O); 2343 if (ArchiveHeaders && !MachOOpt && C) 2344 printArchiveChild(ArchiveName, *C); 2345 if (Disassemble) 2346 disassembleObject(O, Relocations); 2347 if (Relocations && !Disassemble) 2348 printRelocations(O); 2349 if (DynamicRelocations) 2350 printDynamicRelocations(O); 2351 if (SectionHeaders) 2352 printSectionHeaders(O); 2353 if (SectionContents) 2354 printSectionContents(O); 2355 if (SymbolTable) 2356 printSymbolTable(O, ArchiveName); 2357 if (UnwindInfo) 2358 printUnwindInfo(O); 2359 if (PrivateHeaders || FirstPrivateHeader) 2360 printPrivateFileHeaders(O, FirstPrivateHeader); 2361 if (ExportsTrie) 2362 printExportsTrie(O); 2363 if (Rebase) 2364 printRebaseTable(O); 2365 if (Bind) 2366 printBindTable(O); 2367 if (LazyBind) 2368 printLazyBindTable(O); 2369 if (WeakBind) 2370 printWeakBindTable(O); 2371 if (RawClangAST) 2372 printRawClangAST(O); 2373 if (PrintFaultMaps) 2374 printFaultMaps(O); 2375 if (DwarfDumpType != DIDT_Null) { 2376 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2377 // Dump the complete DWARF structure. 2378 DIDumpOptions DumpOpts; 2379 DumpOpts.DumpType = DwarfDumpType; 2380 DICtx->dump(outs(), DumpOpts); 2381 } 2382 } 2383 2384 static void dumpObject(const COFFImportFile *I, const Archive *A, 2385 const Archive::Child *C = nullptr) { 2386 StringRef ArchiveName = A ? A->getFileName() : ""; 2387 2388 // Avoid other output when using a raw option. 2389 if (!RawClangAST) 2390 outs() << '\n' 2391 << ArchiveName << "(" << I->getFileName() << ")" 2392 << ":\tfile format COFF-import-file" 2393 << "\n\n"; 2394 2395 if (ArchiveHeaders && !MachOOpt && C) 2396 printArchiveChild(ArchiveName, *C); 2397 if (SymbolTable) 2398 printCOFFSymbolTable(I); 2399 } 2400 2401 /// Dump each object file in \a a; 2402 static void dumpArchive(const Archive *A) { 2403 Error Err = Error::success(); 2404 for (auto &C : A->children(Err)) { 2405 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2406 if (!ChildOrErr) { 2407 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2408 report_error(A->getFileName(), C, std::move(E)); 2409 continue; 2410 } 2411 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2412 dumpObject(O, A, &C); 2413 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2414 dumpObject(I, A, &C); 2415 else 2416 report_error(A->getFileName(), object_error::invalid_file_type); 2417 } 2418 if (Err) 2419 report_error(A->getFileName(), std::move(Err)); 2420 } 2421 2422 /// Open file and figure out how to dump it. 2423 static void dumpInput(StringRef file) { 2424 // If we are using the Mach-O specific object file parser, then let it parse 2425 // the file and process the command line options. So the -arch flags can 2426 // be used to select specific slices, etc. 2427 if (MachOOpt) { 2428 parseInputMachO(file); 2429 return; 2430 } 2431 2432 // Attempt to open the binary. 2433 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2434 if (!BinaryOrErr) 2435 report_error(file, BinaryOrErr.takeError()); 2436 Binary &Binary = *BinaryOrErr.get().getBinary(); 2437 2438 if (Archive *A = dyn_cast<Archive>(&Binary)) 2439 dumpArchive(A); 2440 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2441 dumpObject(O); 2442 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2443 parseInputMachO(UB); 2444 else 2445 report_error(file, object_error::invalid_file_type); 2446 } 2447 2448 int main(int argc, char **argv) { 2449 InitLLVM X(argc, argv); 2450 2451 // Initialize targets and assembly printers/parsers. 2452 llvm::InitializeAllTargetInfos(); 2453 llvm::InitializeAllTargetMCs(); 2454 llvm::InitializeAllDisassemblers(); 2455 2456 // Register the target printer for --version. 2457 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2458 2459 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2460 2461 ToolName = argv[0]; 2462 2463 // Defaults to a.out if no filenames specified. 2464 if (InputFilenames.empty()) 2465 InputFilenames.push_back("a.out"); 2466 2467 if (AllHeaders) 2468 FileHeaders = PrivateHeaders = Relocations = SectionHeaders = SymbolTable = 2469 true; 2470 2471 if (DisassembleAll || PrintSource || PrintLines) 2472 Disassemble = true; 2473 2474 if (!Disassemble 2475 && !Relocations 2476 && !DynamicRelocations 2477 && !SectionHeaders 2478 && !SectionContents 2479 && !SymbolTable 2480 && !UnwindInfo 2481 && !PrivateHeaders 2482 && !FileHeaders 2483 && !FirstPrivateHeader 2484 && !ExportsTrie 2485 && !Rebase 2486 && !Bind 2487 && !LazyBind 2488 && !WeakBind 2489 && !RawClangAST 2490 && !(UniversalHeaders && MachOOpt) 2491 && !ArchiveHeaders 2492 && !(IndirectSymbols && MachOOpt) 2493 && !(DataInCode && MachOOpt) 2494 && !(LinkOptHints && MachOOpt) 2495 && !(InfoPlist && MachOOpt) 2496 && !(DylibsUsed && MachOOpt) 2497 && !(DylibId && MachOOpt) 2498 && !(ObjcMetaData && MachOOpt) 2499 && !(!FilterSections.empty() && MachOOpt) 2500 && !PrintFaultMaps 2501 && DwarfDumpType == DIDT_Null) { 2502 cl::PrintHelpMessage(); 2503 return 2; 2504 } 2505 2506 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2507 DisassembleFunctions.end()); 2508 2509 llvm::for_each(InputFilenames, dumpInput); 2510 2511 return EXIT_SUCCESS; 2512 } 2513