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/raw_ostream.h" 63 #include <algorithm> 64 #include <cctype> 65 #include <cstring> 66 #include <system_error> 67 #include <unordered_map> 68 #include <utility> 69 70 using namespace llvm; 71 using namespace object; 72 73 cl::opt<bool> 74 llvm::AllHeaders("all-headers", 75 cl::desc("Display all available header information")); 76 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 77 cl::aliasopt(AllHeaders)); 78 79 static cl::list<std::string> 80 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 81 82 cl::opt<bool> 83 llvm::Disassemble("disassemble", 84 cl::desc("Display assembler mnemonics for the machine instructions")); 85 static cl::alias 86 Disassembled("d", cl::desc("Alias for --disassemble"), 87 cl::aliasopt(Disassemble)); 88 89 cl::opt<bool> 90 llvm::DisassembleAll("disassemble-all", 91 cl::desc("Display assembler mnemonics for the machine instructions")); 92 static cl::alias 93 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 94 cl::aliasopt(DisassembleAll)); 95 96 cl::opt<bool> llvm::Demangle("demangle", cl::desc("Demangle symbols names"), 97 cl::init(false)); 98 99 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 100 cl::aliasopt(llvm::Demangle)); 101 102 static cl::list<std::string> 103 DisassembleFunctions("df", 104 cl::CommaSeparated, 105 cl::desc("List of functions to disassemble")); 106 static StringSet<> DisasmFuncsSet; 107 108 cl::opt<bool> 109 llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 110 111 cl::opt<bool> 112 llvm::DynamicRelocations("dynamic-reloc", 113 cl::desc("Display the dynamic relocation entries in the file")); 114 static cl::alias 115 DynamicRelocationsd("R", cl::desc("Alias for --dynamic-reloc"), 116 cl::aliasopt(DynamicRelocations)); 117 118 cl::opt<bool> 119 llvm::SectionContents("full-contents", 120 cl::desc("Display the content of each section")); 121 static cl::alias SectionContentsShort("s", 122 cl::desc("Alias for --full-contents"), 123 cl::aliasopt(SectionContents)); 124 125 cl::opt<bool> 126 llvm::SymbolTable("t", cl::desc("Display the symbol table")); 127 128 cl::opt<bool> 129 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 130 131 cl::opt<bool> 132 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 133 134 cl::opt<bool> 135 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 136 137 cl::opt<bool> 138 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 139 140 cl::opt<bool> 141 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 142 143 cl::opt<bool> 144 llvm::RawClangAST("raw-clang-ast", 145 cl::desc("Dump the raw binary contents of the clang AST section")); 146 147 static cl::opt<bool> 148 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 149 static cl::alias 150 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 151 152 cl::opt<std::string> 153 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 154 "see -version for available targets")); 155 156 cl::opt<std::string> 157 llvm::MCPU("mcpu", 158 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 159 cl::value_desc("cpu-name"), 160 cl::init("")); 161 162 cl::opt<std::string> 163 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 164 "see -version for available targets")); 165 166 cl::opt<bool> 167 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 168 "headers for each section.")); 169 static cl::alias 170 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 171 cl::aliasopt(SectionHeaders)); 172 static cl::alias 173 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 174 cl::aliasopt(SectionHeaders)); 175 176 cl::list<std::string> 177 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 178 "With -macho dump segment,section")); 179 cl::alias 180 static FilterSectionsj("j", cl::desc("Alias for --section"), 181 cl::aliasopt(llvm::FilterSections)); 182 183 cl::list<std::string> 184 llvm::MAttrs("mattr", 185 cl::CommaSeparated, 186 cl::desc("Target specific attributes"), 187 cl::value_desc("a1,+a2,-a3,...")); 188 189 cl::opt<bool> 190 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 191 "instructions, do not print " 192 "the instruction bytes.")); 193 cl::opt<bool> 194 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 195 196 cl::opt<bool> 197 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 198 199 static cl::alias 200 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 201 cl::aliasopt(UnwindInfo)); 202 203 cl::opt<bool> 204 llvm::PrivateHeaders("private-headers", 205 cl::desc("Display format specific file headers")); 206 207 cl::opt<bool> 208 llvm::FirstPrivateHeader("private-header", 209 cl::desc("Display only the first format specific file " 210 "header")); 211 212 static cl::alias 213 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 214 cl::aliasopt(PrivateHeaders)); 215 216 cl::opt<bool> llvm::FileHeaders( 217 "file-headers", 218 cl::desc("Display the contents of the overall file header")); 219 220 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 221 cl::aliasopt(FileHeaders)); 222 223 cl::opt<bool> 224 llvm::ArchiveHeaders("archive-headers", 225 cl::desc("Display archive header information")); 226 227 cl::alias 228 ArchiveHeadersShort("a", cl::desc("Alias for --archive-headers"), 229 cl::aliasopt(ArchiveHeaders)); 230 231 cl::opt<bool> 232 llvm::PrintImmHex("print-imm-hex", 233 cl::desc("Use hex format for immediate values")); 234 235 cl::opt<bool> PrintFaultMaps("fault-map-section", 236 cl::desc("Display contents of faultmap section")); 237 238 cl::opt<DIDumpType> llvm::DwarfDumpType( 239 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 240 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 241 242 cl::opt<bool> PrintSource( 243 "source", 244 cl::desc( 245 "Display source inlined with disassembly. Implies disassemble object")); 246 247 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 248 cl::aliasopt(PrintSource)); 249 250 cl::opt<bool> PrintLines("line-numbers", 251 cl::desc("Display source line numbers with " 252 "disassembly. Implies disassemble object")); 253 254 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 255 cl::aliasopt(PrintLines)); 256 257 cl::opt<unsigned long long> 258 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 259 cl::value_desc("address"), cl::init(0)); 260 cl::opt<unsigned long long> 261 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 262 cl::value_desc("address"), cl::init(UINT64_MAX)); 263 static StringRef ToolName; 264 265 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 266 267 namespace { 268 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 269 270 class SectionFilterIterator { 271 public: 272 SectionFilterIterator(FilterPredicate P, 273 llvm::object::section_iterator const &I, 274 llvm::object::section_iterator const &E) 275 : Predicate(std::move(P)), Iterator(I), End(E) { 276 ScanPredicate(); 277 } 278 const llvm::object::SectionRef &operator*() const { return *Iterator; } 279 SectionFilterIterator &operator++() { 280 ++Iterator; 281 ScanPredicate(); 282 return *this; 283 } 284 bool operator!=(SectionFilterIterator const &Other) const { 285 return Iterator != Other.Iterator; 286 } 287 288 private: 289 void ScanPredicate() { 290 while (Iterator != End && !Predicate(*Iterator)) { 291 ++Iterator; 292 } 293 } 294 FilterPredicate Predicate; 295 llvm::object::section_iterator Iterator; 296 llvm::object::section_iterator End; 297 }; 298 299 class SectionFilter { 300 public: 301 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 302 : Predicate(std::move(P)), Object(O) {} 303 SectionFilterIterator begin() { 304 return SectionFilterIterator(Predicate, Object.section_begin(), 305 Object.section_end()); 306 } 307 SectionFilterIterator end() { 308 return SectionFilterIterator(Predicate, Object.section_end(), 309 Object.section_end()); 310 } 311 312 private: 313 FilterPredicate Predicate; 314 llvm::object::ObjectFile const &Object; 315 }; 316 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 317 return SectionFilter( 318 [](llvm::object::SectionRef const &S) { 319 if (FilterSections.empty()) 320 return true; 321 llvm::StringRef String; 322 std::error_code error = S.getName(String); 323 if (error) 324 return false; 325 return is_contained(FilterSections, String); 326 }, 327 O); 328 } 329 } 330 331 void llvm::error(std::error_code EC) { 332 if (!EC) 333 return; 334 335 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 336 errs().flush(); 337 exit(1); 338 } 339 340 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 341 errs() << ToolName << ": " << Message << ".\n"; 342 errs().flush(); 343 exit(1); 344 } 345 346 void llvm::warn(StringRef Message) { 347 errs() << ToolName << ": warning: " << Message << ".\n"; 348 errs().flush(); 349 } 350 351 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 352 Twine Message) { 353 errs() << ToolName << ": '" << File << "': " << Message << ".\n"; 354 exit(1); 355 } 356 357 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 358 std::error_code EC) { 359 assert(EC); 360 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 361 exit(1); 362 } 363 364 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 365 llvm::Error E) { 366 assert(E); 367 std::string Buf; 368 raw_string_ostream OS(Buf); 369 logAllUnhandledErrors(std::move(E), OS, ""); 370 OS.flush(); 371 errs() << ToolName << ": '" << File << "': " << Buf; 372 exit(1); 373 } 374 375 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 376 StringRef FileName, 377 llvm::Error E, 378 StringRef ArchitectureName) { 379 assert(E); 380 errs() << ToolName << ": "; 381 if (ArchiveName != "") 382 errs() << ArchiveName << "(" << FileName << ")"; 383 else 384 errs() << "'" << FileName << "'"; 385 if (!ArchitectureName.empty()) 386 errs() << " (for architecture " << ArchitectureName << ")"; 387 std::string Buf; 388 raw_string_ostream OS(Buf); 389 logAllUnhandledErrors(std::move(E), OS, ""); 390 OS.flush(); 391 errs() << ": " << Buf; 392 exit(1); 393 } 394 395 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 396 const object::Archive::Child &C, 397 llvm::Error E, 398 StringRef ArchitectureName) { 399 Expected<StringRef> NameOrErr = C.getName(); 400 // TODO: if we have a error getting the name then it would be nice to print 401 // the index of which archive member this is and or its offset in the 402 // archive instead of "???" as the name. 403 if (!NameOrErr) { 404 consumeError(NameOrErr.takeError()); 405 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 406 } else 407 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 408 ArchitectureName); 409 } 410 411 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 412 // Figure out the target triple. 413 llvm::Triple TheTriple("unknown-unknown-unknown"); 414 if (TripleName.empty()) { 415 if (Obj) { 416 TheTriple = Obj->makeTriple(); 417 } 418 } else { 419 TheTriple.setTriple(Triple::normalize(TripleName)); 420 421 // Use the triple, but also try to combine with ARM build attributes. 422 if (Obj) { 423 auto Arch = Obj->getArch(); 424 if (Arch == Triple::arm || Arch == Triple::armeb) { 425 Obj->setARMSubArch(TheTriple); 426 } 427 } 428 } 429 430 // Get the target specific parser. 431 std::string Error; 432 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 433 Error); 434 if (!TheTarget) { 435 if (Obj) 436 report_error(Obj->getFileName(), "can't find target: " + Error); 437 else 438 error("can't find target: " + Error); 439 } 440 441 // Update the triple name and return the found target. 442 TripleName = TheTriple.getTriple(); 443 return TheTarget; 444 } 445 446 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 447 return a.getOffset() < b.getOffset(); 448 } 449 450 template <class ELFT> 451 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 452 const RelocationRef &RelRef, 453 SmallVectorImpl<char> &Result) { 454 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 455 456 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 457 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 458 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 459 460 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 461 462 auto SecOrErr = EF.getSection(Rel.d.a); 463 if (!SecOrErr) 464 return errorToErrorCode(SecOrErr.takeError()); 465 const Elf_Shdr *Sec = *SecOrErr; 466 auto SymTabOrErr = EF.getSection(Sec->sh_link); 467 if (!SymTabOrErr) 468 return errorToErrorCode(SymTabOrErr.takeError()); 469 const Elf_Shdr *SymTab = *SymTabOrErr; 470 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 471 SymTab->sh_type == ELF::SHT_DYNSYM); 472 auto StrTabSec = EF.getSection(SymTab->sh_link); 473 if (!StrTabSec) 474 return errorToErrorCode(StrTabSec.takeError()); 475 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 476 if (!StrTabOrErr) 477 return errorToErrorCode(StrTabOrErr.takeError()); 478 StringRef StrTab = *StrTabOrErr; 479 int64_t addend = 0; 480 // If there is no Symbol associated with the relocation, we set the undef 481 // boolean value to 'true'. This will prevent us from calling functions that 482 // requires the relocation to be associated with a symbol. 483 bool undef = false; 484 switch (Sec->sh_type) { 485 default: 486 return object_error::parse_failed; 487 case ELF::SHT_REL: { 488 // TODO: Read implicit addend from section data. 489 break; 490 } 491 case ELF::SHT_RELA: { 492 const Elf_Rela *ERela = Obj->getRela(Rel); 493 addend = ERela->r_addend; 494 undef = ERela->getSymbol(false) == 0; 495 break; 496 } 497 } 498 StringRef Target; 499 if (!undef) { 500 symbol_iterator SI = RelRef.getSymbol(); 501 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 502 if (symb->getType() == ELF::STT_SECTION) { 503 Expected<section_iterator> SymSI = SI->getSection(); 504 if (!SymSI) 505 return errorToErrorCode(SymSI.takeError()); 506 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 507 auto SecName = EF.getSectionName(SymSec); 508 if (!SecName) 509 return errorToErrorCode(SecName.takeError()); 510 Target = *SecName; 511 } else { 512 Expected<StringRef> SymName = symb->getName(StrTab); 513 if (!SymName) 514 return errorToErrorCode(SymName.takeError()); 515 Target = *SymName; 516 } 517 } else 518 Target = "*ABS*"; 519 520 // Default scheme is to print Target, as well as "+ <addend>" for nonzero 521 // addend. Should be acceptable for all normal purposes. 522 std::string fmtbuf; 523 raw_string_ostream fmt(fmtbuf); 524 fmt << Target; 525 if (addend != 0) 526 fmt << (addend < 0 ? "" : "+") << addend; 527 fmt.flush(); 528 Result.append(fmtbuf.begin(), fmtbuf.end()); 529 return std::error_code(); 530 } 531 532 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 533 const RelocationRef &Rel, 534 SmallVectorImpl<char> &Result) { 535 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 536 return getRelocationValueString(ELF32LE, Rel, Result); 537 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 538 return getRelocationValueString(ELF64LE, Rel, Result); 539 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 540 return getRelocationValueString(ELF32BE, Rel, Result); 541 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 542 return getRelocationValueString(ELF64BE, Rel, Result); 543 } 544 545 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 546 const RelocationRef &Rel, 547 SmallVectorImpl<char> &Result) { 548 symbol_iterator SymI = Rel.getSymbol(); 549 Expected<StringRef> SymNameOrErr = SymI->getName(); 550 if (!SymNameOrErr) 551 return errorToErrorCode(SymNameOrErr.takeError()); 552 StringRef SymName = *SymNameOrErr; 553 Result.append(SymName.begin(), SymName.end()); 554 return std::error_code(); 555 } 556 557 static void printRelocationTargetName(const MachOObjectFile *O, 558 const MachO::any_relocation_info &RE, 559 raw_string_ostream &fmt) { 560 bool IsScattered = O->isRelocationScattered(RE); 561 562 // Target of a scattered relocation is an address. In the interest of 563 // generating pretty output, scan through the symbol table looking for a 564 // symbol that aligns with that address. If we find one, print it. 565 // Otherwise, we just print the hex address of the target. 566 if (IsScattered) { 567 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 568 569 for (const SymbolRef &Symbol : O->symbols()) { 570 std::error_code ec; 571 Expected<uint64_t> Addr = Symbol.getAddress(); 572 if (!Addr) 573 report_error(O->getFileName(), Addr.takeError()); 574 if (*Addr != Val) 575 continue; 576 Expected<StringRef> Name = Symbol.getName(); 577 if (!Name) 578 report_error(O->getFileName(), Name.takeError()); 579 fmt << *Name; 580 return; 581 } 582 583 // If we couldn't find a symbol that this relocation refers to, try 584 // to find a section beginning instead. 585 for (const SectionRef &Section : ToolSectionFilter(*O)) { 586 std::error_code ec; 587 588 StringRef Name; 589 uint64_t Addr = Section.getAddress(); 590 if (Addr != Val) 591 continue; 592 if ((ec = Section.getName(Name))) 593 report_error(O->getFileName(), ec); 594 fmt << Name; 595 return; 596 } 597 598 fmt << format("0x%x", Val); 599 return; 600 } 601 602 StringRef S; 603 bool isExtern = O->getPlainRelocationExternal(RE); 604 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 605 606 if (O->getAnyRelocationType(RE) == MachO::ARM64_RELOC_ADDEND) { 607 fmt << format("0x%0" PRIx64, Val); 608 return; 609 } else if (isExtern) { 610 symbol_iterator SI = O->symbol_begin(); 611 advance(SI, Val); 612 Expected<StringRef> SOrErr = SI->getName(); 613 if (!SOrErr) 614 report_error(O->getFileName(), SOrErr.takeError()); 615 S = *SOrErr; 616 } else { 617 section_iterator SI = O->section_begin(); 618 // Adjust for the fact that sections are 1-indexed. 619 if (Val == 0) { 620 fmt << "0 (?,?)"; 621 return; 622 } 623 uint32_t i = Val - 1; 624 while (i != 0 && SI != O->section_end()) { 625 i--; 626 advance(SI, 1); 627 } 628 if (SI == O->section_end()) 629 fmt << Val << " (?,?)"; 630 else 631 SI->getName(S); 632 } 633 634 fmt << S; 635 } 636 637 static std::error_code getRelocationValueString(const WasmObjectFile *Obj, 638 const RelocationRef &RelRef, 639 SmallVectorImpl<char> &Result) { 640 const wasm::WasmRelocation& Rel = Obj->getWasmRelocation(RelRef); 641 symbol_iterator SI = RelRef.getSymbol(); 642 std::string fmtbuf; 643 raw_string_ostream fmt(fmtbuf); 644 if (SI == Obj->symbol_end()) { 645 // Not all wasm relocations have symbols associated with them. 646 // In particular R_WEBASSEMBLY_TYPE_INDEX_LEB. 647 fmt << Rel.Index; 648 } else { 649 Expected<StringRef> SymNameOrErr = SI->getName(); 650 if (!SymNameOrErr) 651 return errorToErrorCode(SymNameOrErr.takeError()); 652 StringRef SymName = *SymNameOrErr; 653 Result.append(SymName.begin(), SymName.end()); 654 } 655 fmt << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 656 fmt.flush(); 657 Result.append(fmtbuf.begin(), fmtbuf.end()); 658 return std::error_code(); 659 } 660 661 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 662 const RelocationRef &RelRef, 663 SmallVectorImpl<char> &Result) { 664 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 665 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 666 667 unsigned Arch = Obj->getArch(); 668 669 std::string fmtbuf; 670 raw_string_ostream fmt(fmtbuf); 671 unsigned Type = Obj->getAnyRelocationType(RE); 672 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 673 674 // Determine any addends that should be displayed with the relocation. 675 // These require decoding the relocation type, which is triple-specific. 676 677 // X86_64 has entirely custom relocation types. 678 if (Arch == Triple::x86_64) { 679 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 680 681 switch (Type) { 682 case MachO::X86_64_RELOC_GOT_LOAD: 683 case MachO::X86_64_RELOC_GOT: { 684 printRelocationTargetName(Obj, RE, fmt); 685 fmt << "@GOT"; 686 if (isPCRel) 687 fmt << "PCREL"; 688 break; 689 } 690 case MachO::X86_64_RELOC_SUBTRACTOR: { 691 DataRefImpl RelNext = Rel; 692 Obj->moveRelocationNext(RelNext); 693 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 694 695 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 696 // X86_64_RELOC_UNSIGNED. 697 // NOTE: Scattered relocations don't exist on x86_64. 698 unsigned RType = Obj->getAnyRelocationType(RENext); 699 if (RType != MachO::X86_64_RELOC_UNSIGNED) 700 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 701 "X86_64_RELOC_SUBTRACTOR."); 702 703 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 704 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 705 printRelocationTargetName(Obj, RENext, fmt); 706 fmt << "-"; 707 printRelocationTargetName(Obj, RE, fmt); 708 break; 709 } 710 case MachO::X86_64_RELOC_TLV: 711 printRelocationTargetName(Obj, RE, fmt); 712 fmt << "@TLV"; 713 if (isPCRel) 714 fmt << "P"; 715 break; 716 case MachO::X86_64_RELOC_SIGNED_1: 717 printRelocationTargetName(Obj, RE, fmt); 718 fmt << "-1"; 719 break; 720 case MachO::X86_64_RELOC_SIGNED_2: 721 printRelocationTargetName(Obj, RE, fmt); 722 fmt << "-2"; 723 break; 724 case MachO::X86_64_RELOC_SIGNED_4: 725 printRelocationTargetName(Obj, RE, fmt); 726 fmt << "-4"; 727 break; 728 default: 729 printRelocationTargetName(Obj, RE, fmt); 730 break; 731 } 732 // X86 and ARM share some relocation types in common. 733 } else if (Arch == Triple::x86 || Arch == Triple::arm || 734 Arch == Triple::ppc) { 735 // Generic relocation types... 736 switch (Type) { 737 case MachO::GENERIC_RELOC_PAIR: // prints no info 738 return std::error_code(); 739 case MachO::GENERIC_RELOC_SECTDIFF: { 740 DataRefImpl RelNext = Rel; 741 Obj->moveRelocationNext(RelNext); 742 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 743 744 // X86 sect diff's must be followed by a relocation of type 745 // GENERIC_RELOC_PAIR. 746 unsigned RType = Obj->getAnyRelocationType(RENext); 747 748 if (RType != MachO::GENERIC_RELOC_PAIR) 749 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 750 "GENERIC_RELOC_SECTDIFF."); 751 752 printRelocationTargetName(Obj, RE, fmt); 753 fmt << "-"; 754 printRelocationTargetName(Obj, RENext, fmt); 755 break; 756 } 757 } 758 759 if (Arch == Triple::x86 || Arch == Triple::ppc) { 760 switch (Type) { 761 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 762 DataRefImpl RelNext = Rel; 763 Obj->moveRelocationNext(RelNext); 764 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 765 766 // X86 sect diff's must be followed by a relocation of type 767 // GENERIC_RELOC_PAIR. 768 unsigned RType = Obj->getAnyRelocationType(RENext); 769 if (RType != MachO::GENERIC_RELOC_PAIR) 770 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 771 "GENERIC_RELOC_LOCAL_SECTDIFF."); 772 773 printRelocationTargetName(Obj, RE, fmt); 774 fmt << "-"; 775 printRelocationTargetName(Obj, RENext, fmt); 776 break; 777 } 778 case MachO::GENERIC_RELOC_TLV: { 779 printRelocationTargetName(Obj, RE, fmt); 780 fmt << "@TLV"; 781 if (IsPCRel) 782 fmt << "P"; 783 break; 784 } 785 default: 786 printRelocationTargetName(Obj, RE, fmt); 787 } 788 } else { // ARM-specific relocations 789 switch (Type) { 790 case MachO::ARM_RELOC_HALF: 791 case MachO::ARM_RELOC_HALF_SECTDIFF: { 792 // Half relocations steal a bit from the length field to encode 793 // whether this is an upper16 or a lower16 relocation. 794 bool isUpper = (Obj->getAnyRelocationLength(RE) & 0x1) == 1; 795 796 if (isUpper) 797 fmt << ":upper16:("; 798 else 799 fmt << ":lower16:("; 800 printRelocationTargetName(Obj, RE, fmt); 801 802 DataRefImpl RelNext = Rel; 803 Obj->moveRelocationNext(RelNext); 804 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 805 806 // ARM half relocs must be followed by a relocation of type 807 // ARM_RELOC_PAIR. 808 unsigned RType = Obj->getAnyRelocationType(RENext); 809 if (RType != MachO::ARM_RELOC_PAIR) 810 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 811 "ARM_RELOC_HALF"); 812 813 // NOTE: The half of the target virtual address is stashed in the 814 // address field of the secondary relocation, but we can't reverse 815 // engineer the constant offset from it without decoding the movw/movt 816 // instruction to find the other half in its immediate field. 817 818 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 819 // symbol/section pointer of the follow-on relocation. 820 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 821 fmt << "-"; 822 printRelocationTargetName(Obj, RENext, fmt); 823 } 824 825 fmt << ")"; 826 break; 827 } 828 default: { printRelocationTargetName(Obj, RE, fmt); } 829 } 830 } 831 } else 832 printRelocationTargetName(Obj, RE, fmt); 833 834 fmt.flush(); 835 Result.append(fmtbuf.begin(), fmtbuf.end()); 836 return std::error_code(); 837 } 838 839 static std::error_code getRelocationValueString(const RelocationRef &Rel, 840 SmallVectorImpl<char> &Result) { 841 const ObjectFile *Obj = Rel.getObject(); 842 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 843 return getRelocationValueString(ELF, Rel, Result); 844 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 845 return getRelocationValueString(COFF, Rel, Result); 846 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 847 return getRelocationValueString(Wasm, Rel, Result); 848 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 849 return getRelocationValueString(MachO, Rel, Result); 850 llvm_unreachable("unknown object file format"); 851 } 852 853 /// Indicates whether this relocation should hidden when listing 854 /// relocations, usually because it is the trailing part of a multipart 855 /// relocation that will be printed as part of the leading relocation. 856 static bool getHidden(RelocationRef RelRef) { 857 const ObjectFile *Obj = RelRef.getObject(); 858 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 859 if (!MachO) 860 return false; 861 862 unsigned Arch = MachO->getArch(); 863 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 864 uint64_t Type = MachO->getRelocationType(Rel); 865 866 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 867 // is always hidden. 868 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 869 if (Type == MachO::GENERIC_RELOC_PAIR) 870 return true; 871 } else if (Arch == Triple::x86_64) { 872 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 873 // an X86_64_RELOC_SUBTRACTOR. 874 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 875 DataRefImpl RelPrev = Rel; 876 RelPrev.d.a--; 877 uint64_t PrevType = MachO->getRelocationType(RelPrev); 878 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 879 return true; 880 } 881 } 882 883 return false; 884 } 885 886 namespace { 887 class SourcePrinter { 888 protected: 889 DILineInfo OldLineInfo; 890 const ObjectFile *Obj = nullptr; 891 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 892 // File name to file contents of source 893 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 894 // Mark the line endings of the cached source 895 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 896 897 private: 898 bool cacheSource(const DILineInfo& LineInfoFile); 899 900 public: 901 SourcePrinter() = default; 902 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 903 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 904 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 905 DefaultArch); 906 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 907 } 908 virtual ~SourcePrinter() = default; 909 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 910 StringRef Delimiter = "; "); 911 }; 912 913 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 914 std::unique_ptr<MemoryBuffer> Buffer; 915 if (LineInfo.Source) { 916 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 917 } else { 918 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 919 if (!BufferOrError) 920 return false; 921 Buffer = std::move(*BufferOrError); 922 } 923 // Chomp the file to get lines 924 size_t BufferSize = Buffer->getBufferSize(); 925 const char *BufferStart = Buffer->getBufferStart(); 926 for (const char *Start = BufferStart, *End = BufferStart; 927 End < BufferStart + BufferSize; End++) 928 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 929 (*End == '\r' && *(End + 1) == '\n')) { 930 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 931 if (*End == '\r') 932 End++; 933 Start = End + 1; 934 } 935 SourceCache[LineInfo.FileName] = std::move(Buffer); 936 return true; 937 } 938 939 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 940 StringRef Delimiter) { 941 if (!Symbolizer) 942 return; 943 DILineInfo LineInfo = DILineInfo(); 944 auto ExpectecLineInfo = 945 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 946 if (!ExpectecLineInfo) 947 consumeError(ExpectecLineInfo.takeError()); 948 else 949 LineInfo = *ExpectecLineInfo; 950 951 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 952 LineInfo.Line == 0) 953 return; 954 955 if (PrintLines) 956 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 957 if (PrintSource) { 958 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 959 if (!cacheSource(LineInfo)) 960 return; 961 auto FileBuffer = SourceCache.find(LineInfo.FileName); 962 if (FileBuffer != SourceCache.end()) { 963 auto LineBuffer = LineCache.find(LineInfo.FileName); 964 if (LineBuffer != LineCache.end()) { 965 if (LineInfo.Line > LineBuffer->second.size()) 966 return; 967 // Vector begins at 0, line numbers are non-zero 968 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 969 << "\n"; 970 } 971 } 972 } 973 OldLineInfo = LineInfo; 974 } 975 976 static bool isArmElf(const ObjectFile *Obj) { 977 return (Obj->isELF() && 978 (Obj->getArch() == Triple::aarch64 || 979 Obj->getArch() == Triple::aarch64_be || 980 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 981 Obj->getArch() == Triple::thumb || 982 Obj->getArch() == Triple::thumbeb)); 983 } 984 985 class PrettyPrinter { 986 public: 987 virtual ~PrettyPrinter() = default; 988 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 989 ArrayRef<uint8_t> Bytes, uint64_t Address, 990 raw_ostream &OS, StringRef Annot, 991 MCSubtargetInfo const &STI, SourcePrinter *SP, 992 std::vector<RelocationRef> *Rels = nullptr) { 993 if (SP && (PrintSource || PrintLines)) 994 SP->printSourceLine(OS, Address); 995 if (!NoLeadingAddr) 996 OS << format("%8" PRIx64 ":", Address); 997 if (!NoShowRawInsn) { 998 OS << "\t"; 999 dumpBytes(Bytes, OS); 1000 } 1001 if (MI) 1002 IP.printInst(MI, OS, "", STI); 1003 else 1004 OS << " <unknown>"; 1005 } 1006 }; 1007 PrettyPrinter PrettyPrinterInst; 1008 class HexagonPrettyPrinter : public PrettyPrinter { 1009 public: 1010 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 1011 raw_ostream &OS) { 1012 uint32_t opcode = 1013 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 1014 if (!NoLeadingAddr) 1015 OS << format("%8" PRIx64 ":", Address); 1016 if (!NoShowRawInsn) { 1017 OS << "\t"; 1018 dumpBytes(Bytes.slice(0, 4), OS); 1019 OS << format("%08" PRIx32, opcode); 1020 } 1021 } 1022 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1023 uint64_t Address, raw_ostream &OS, StringRef Annot, 1024 MCSubtargetInfo const &STI, SourcePrinter *SP, 1025 std::vector<RelocationRef> *Rels) override { 1026 if (SP && (PrintSource || PrintLines)) 1027 SP->printSourceLine(OS, Address, ""); 1028 if (!MI) { 1029 printLead(Bytes, Address, OS); 1030 OS << " <unknown>"; 1031 return; 1032 } 1033 std::string Buffer; 1034 { 1035 raw_string_ostream TempStream(Buffer); 1036 IP.printInst(MI, TempStream, "", STI); 1037 } 1038 StringRef Contents(Buffer); 1039 // Split off bundle attributes 1040 auto PacketBundle = Contents.rsplit('\n'); 1041 // Split off first instruction from the rest 1042 auto HeadTail = PacketBundle.first.split('\n'); 1043 auto Preamble = " { "; 1044 auto Separator = ""; 1045 StringRef Fmt = "\t\t\t%08" PRIx64 ": "; 1046 std::vector<RelocationRef>::const_iterator rel_cur = Rels->begin(); 1047 std::vector<RelocationRef>::const_iterator rel_end = Rels->end(); 1048 1049 // Hexagon's packets require relocations to be inline rather than 1050 // clustered at the end of the packet. 1051 auto PrintReloc = [&]() -> void { 1052 while ((rel_cur != rel_end) && (rel_cur->getOffset() <= Address)) { 1053 if (rel_cur->getOffset() == Address) { 1054 SmallString<16> name; 1055 SmallString<32> val; 1056 rel_cur->getTypeName(name); 1057 error(getRelocationValueString(*rel_cur, val)); 1058 OS << Separator << format(Fmt.data(), Address) << name << "\t" << val 1059 << "\n"; 1060 return; 1061 } 1062 rel_cur++; 1063 } 1064 }; 1065 1066 while(!HeadTail.first.empty()) { 1067 OS << Separator; 1068 Separator = "\n"; 1069 if (SP && (PrintSource || PrintLines)) 1070 SP->printSourceLine(OS, Address, ""); 1071 printLead(Bytes, Address, OS); 1072 OS << Preamble; 1073 Preamble = " "; 1074 StringRef Inst; 1075 auto Duplex = HeadTail.first.split('\v'); 1076 if(!Duplex.second.empty()){ 1077 OS << Duplex.first; 1078 OS << "; "; 1079 Inst = Duplex.second; 1080 } 1081 else 1082 Inst = HeadTail.first; 1083 OS << Inst; 1084 HeadTail = HeadTail.second.split('\n'); 1085 if (HeadTail.first.empty()) 1086 OS << " } " << PacketBundle.second; 1087 PrintReloc(); 1088 Bytes = Bytes.slice(4); 1089 Address += 4; 1090 } 1091 } 1092 }; 1093 HexagonPrettyPrinter HexagonPrettyPrinterInst; 1094 1095 class AMDGCNPrettyPrinter : public PrettyPrinter { 1096 public: 1097 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1098 uint64_t Address, raw_ostream &OS, StringRef Annot, 1099 MCSubtargetInfo const &STI, SourcePrinter *SP, 1100 std::vector<RelocationRef> *Rels) override { 1101 if (SP && (PrintSource || PrintLines)) 1102 SP->printSourceLine(OS, Address); 1103 1104 typedef support::ulittle32_t U32; 1105 1106 if (MI) { 1107 SmallString<40> InstStr; 1108 raw_svector_ostream IS(InstStr); 1109 1110 IP.printInst(MI, IS, "", STI); 1111 1112 OS << left_justify(IS.str(), 60); 1113 } else { 1114 // an unrecognized encoding - this is probably data so represent it 1115 // using the .long directive, or .byte directive if fewer than 4 bytes 1116 // remaining 1117 if (Bytes.size() >= 4) { 1118 OS << format("\t.long 0x%08" PRIx32 " ", 1119 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 1120 OS.indent(42); 1121 } else { 1122 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 1123 for (unsigned int i = 1; i < Bytes.size(); i++) 1124 OS << format(", 0x%02" PRIx8, Bytes[i]); 1125 OS.indent(55 - (6 * Bytes.size())); 1126 } 1127 } 1128 1129 OS << format("// %012" PRIX64 ": ", Address); 1130 if (Bytes.size() >=4) { 1131 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 1132 Bytes.size() / sizeof(U32))) 1133 // D should be explicitly casted to uint32_t here as it is passed 1134 // by format to snprintf as vararg. 1135 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 1136 } else { 1137 for (unsigned int i = 0; i < Bytes.size(); i++) 1138 OS << format("%02" PRIX8 " ", Bytes[i]); 1139 } 1140 1141 if (!Annot.empty()) 1142 OS << "// " << Annot; 1143 } 1144 }; 1145 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 1146 1147 class BPFPrettyPrinter : public PrettyPrinter { 1148 public: 1149 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1150 uint64_t Address, raw_ostream &OS, StringRef Annot, 1151 MCSubtargetInfo const &STI, SourcePrinter *SP, 1152 std::vector<RelocationRef> *Rels) override { 1153 if (SP && (PrintSource || PrintLines)) 1154 SP->printSourceLine(OS, Address); 1155 if (!NoLeadingAddr) 1156 OS << format("%8" PRId64 ":", Address / 8); 1157 if (!NoShowRawInsn) { 1158 OS << "\t"; 1159 dumpBytes(Bytes, OS); 1160 } 1161 if (MI) 1162 IP.printInst(MI, OS, "", STI); 1163 else 1164 OS << " <unknown>"; 1165 } 1166 }; 1167 BPFPrettyPrinter BPFPrettyPrinterInst; 1168 1169 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 1170 switch(Triple.getArch()) { 1171 default: 1172 return PrettyPrinterInst; 1173 case Triple::hexagon: 1174 return HexagonPrettyPrinterInst; 1175 case Triple::amdgcn: 1176 return AMDGCNPrettyPrinterInst; 1177 case Triple::bpfel: 1178 case Triple::bpfeb: 1179 return BPFPrettyPrinterInst; 1180 } 1181 } 1182 } 1183 1184 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1185 assert(Obj->isELF()); 1186 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1187 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1188 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1189 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1190 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1191 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1192 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1193 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1194 llvm_unreachable("Unsupported binary format"); 1195 } 1196 1197 template <class ELFT> static void 1198 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1199 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1200 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1201 uint8_t SymbolType = Symbol.getELFType(); 1202 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1203 continue; 1204 1205 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1206 if (!AddressOrErr) 1207 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1208 uint64_t Address = *AddressOrErr; 1209 1210 Expected<StringRef> Name = Symbol.getName(); 1211 if (!Name) 1212 report_error(Obj->getFileName(), Name.takeError()); 1213 if (Name->empty()) 1214 continue; 1215 1216 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1217 if (!SectionOrErr) 1218 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1219 section_iterator SecI = *SectionOrErr; 1220 if (SecI == Obj->section_end()) 1221 continue; 1222 1223 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1224 } 1225 } 1226 1227 static void 1228 addDynamicElfSymbols(const ObjectFile *Obj, 1229 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1230 assert(Obj->isELF()); 1231 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1232 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1233 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1234 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1235 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1236 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1237 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1238 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1239 else 1240 llvm_unreachable("Unsupported binary format"); 1241 } 1242 1243 static void addPltEntries(const ObjectFile *Obj, 1244 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 1245 StringSaver &Saver) { 1246 Optional<SectionRef> Plt = None; 1247 for (const SectionRef &Section : Obj->sections()) { 1248 StringRef Name; 1249 if (Section.getName(Name)) 1250 continue; 1251 if (Name == ".plt") 1252 Plt = Section; 1253 } 1254 if (!Plt) 1255 return; 1256 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1257 for (auto PltEntry : ElfObj->getPltAddresses()) { 1258 SymbolRef Symbol(PltEntry.first, ElfObj); 1259 1260 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1261 1262 Expected<StringRef> NameOrErr = Symbol.getName(); 1263 if (!NameOrErr) 1264 report_error(Obj->getFileName(), NameOrErr.takeError()); 1265 if (NameOrErr->empty()) 1266 continue; 1267 StringRef Name = Saver.save((*NameOrErr + "@plt").str()); 1268 1269 AllSymbols[*Plt].emplace_back(PltEntry.second, Name, SymbolType); 1270 } 1271 } 1272 } 1273 1274 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1275 if (StartAddress > StopAddress) 1276 error("Start address should be less than stop address"); 1277 1278 const Target *TheTarget = getTarget(Obj); 1279 1280 // Package up features to be passed to target/subtarget 1281 SubtargetFeatures Features = Obj->getFeatures(); 1282 if (MAttrs.size()) { 1283 for (unsigned i = 0; i != MAttrs.size(); ++i) 1284 Features.AddFeature(MAttrs[i]); 1285 } 1286 1287 std::unique_ptr<const MCRegisterInfo> MRI( 1288 TheTarget->createMCRegInfo(TripleName)); 1289 if (!MRI) 1290 report_error(Obj->getFileName(), "no register info for target " + 1291 TripleName); 1292 1293 // Set up disassembler. 1294 std::unique_ptr<const MCAsmInfo> AsmInfo( 1295 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1296 if (!AsmInfo) 1297 report_error(Obj->getFileName(), "no assembly info for target " + 1298 TripleName); 1299 std::unique_ptr<const MCSubtargetInfo> STI( 1300 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1301 if (!STI) 1302 report_error(Obj->getFileName(), "no subtarget info for target " + 1303 TripleName); 1304 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1305 if (!MII) 1306 report_error(Obj->getFileName(), "no instruction info for target " + 1307 TripleName); 1308 MCObjectFileInfo MOFI; 1309 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1310 // FIXME: for now initialize MCObjectFileInfo with default values 1311 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1312 1313 std::unique_ptr<MCDisassembler> DisAsm( 1314 TheTarget->createMCDisassembler(*STI, Ctx)); 1315 if (!DisAsm) 1316 report_error(Obj->getFileName(), "no disassembler for target " + 1317 TripleName); 1318 1319 std::unique_ptr<const MCInstrAnalysis> MIA( 1320 TheTarget->createMCInstrAnalysis(MII.get())); 1321 1322 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1323 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1324 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1325 if (!IP) 1326 report_error(Obj->getFileName(), "no instruction printer for target " + 1327 TripleName); 1328 IP->setPrintImmHex(PrintImmHex); 1329 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1330 1331 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1332 "\t\t\t%08" PRIx64 ": "; 1333 1334 SourcePrinter SP(Obj, TheTarget->getName()); 1335 1336 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1337 // in RelocSecs contain the relocations for section S. 1338 std::error_code EC; 1339 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1340 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1341 section_iterator Sec2 = Section.getRelocatedSection(); 1342 if (Sec2 != Obj->section_end()) 1343 SectionRelocMap[*Sec2].push_back(Section); 1344 } 1345 1346 // Create a mapping from virtual address to symbol name. This is used to 1347 // pretty print the symbols while disassembling. 1348 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1349 SectionSymbolsTy AbsoluteSymbols; 1350 for (const SymbolRef &Symbol : Obj->symbols()) { 1351 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1352 if (!AddressOrErr) 1353 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1354 uint64_t Address = *AddressOrErr; 1355 1356 Expected<StringRef> Name = Symbol.getName(); 1357 if (!Name) 1358 report_error(Obj->getFileName(), Name.takeError()); 1359 if (Name->empty()) 1360 continue; 1361 1362 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1363 if (!SectionOrErr) 1364 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1365 1366 uint8_t SymbolType = ELF::STT_NOTYPE; 1367 if (Obj->isELF()) 1368 SymbolType = getElfSymbolType(Obj, Symbol); 1369 1370 section_iterator SecI = *SectionOrErr; 1371 if (SecI != Obj->section_end()) 1372 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1373 else 1374 AbsoluteSymbols.emplace_back(Address, *Name, SymbolType); 1375 1376 1377 } 1378 if (AllSymbols.empty() && Obj->isELF()) 1379 addDynamicElfSymbols(Obj, AllSymbols); 1380 1381 BumpPtrAllocator A; 1382 StringSaver Saver(A); 1383 addPltEntries(Obj, AllSymbols, Saver); 1384 1385 // Create a mapping from virtual address to section. 1386 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1387 for (SectionRef Sec : Obj->sections()) 1388 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1389 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1390 1391 // Linked executables (.exe and .dll files) typically don't include a real 1392 // symbol table but they might contain an export table. 1393 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1394 for (const auto &ExportEntry : COFFObj->export_directories()) { 1395 StringRef Name; 1396 error(ExportEntry.getSymbolName(Name)); 1397 if (Name.empty()) 1398 continue; 1399 uint32_t RVA; 1400 error(ExportEntry.getExportRVA(RVA)); 1401 1402 uint64_t VA = COFFObj->getImageBase() + RVA; 1403 auto Sec = std::upper_bound( 1404 SectionAddresses.begin(), SectionAddresses.end(), VA, 1405 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1406 return LHS < RHS.first; 1407 }); 1408 if (Sec != SectionAddresses.begin()) 1409 --Sec; 1410 else 1411 Sec = SectionAddresses.end(); 1412 1413 if (Sec != SectionAddresses.end()) 1414 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1415 else 1416 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1417 } 1418 } 1419 1420 // Sort all the symbols, this allows us to use a simple binary search to find 1421 // a symbol near an address. 1422 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1423 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1424 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1425 1426 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1427 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1428 continue; 1429 1430 uint64_t SectionAddr = Section.getAddress(); 1431 uint64_t SectSize = Section.getSize(); 1432 if (!SectSize) 1433 continue; 1434 1435 // Get the list of all the symbols in this section. 1436 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1437 std::vector<uint64_t> DataMappingSymsAddr; 1438 std::vector<uint64_t> TextMappingSymsAddr; 1439 if (isArmElf(Obj)) { 1440 for (const auto &Symb : Symbols) { 1441 uint64_t Address = std::get<0>(Symb); 1442 StringRef Name = std::get<1>(Symb); 1443 if (Name.startswith("$d")) 1444 DataMappingSymsAddr.push_back(Address - SectionAddr); 1445 if (Name.startswith("$x")) 1446 TextMappingSymsAddr.push_back(Address - SectionAddr); 1447 if (Name.startswith("$a")) 1448 TextMappingSymsAddr.push_back(Address - SectionAddr); 1449 if (Name.startswith("$t")) 1450 TextMappingSymsAddr.push_back(Address - SectionAddr); 1451 } 1452 } 1453 1454 llvm::sort(DataMappingSymsAddr); 1455 llvm::sort(TextMappingSymsAddr); 1456 1457 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1458 // AMDGPU disassembler uses symbolizer for printing labels 1459 std::unique_ptr<MCRelocationInfo> RelInfo( 1460 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1461 if (RelInfo) { 1462 std::unique_ptr<MCSymbolizer> Symbolizer( 1463 TheTarget->createMCSymbolizer( 1464 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1465 DisAsm->setSymbolizer(std::move(Symbolizer)); 1466 } 1467 } 1468 1469 // Make a list of all the relocations for this section. 1470 std::vector<RelocationRef> Rels; 1471 if (InlineRelocs) { 1472 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1473 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1474 Rels.push_back(Reloc); 1475 } 1476 } 1477 } 1478 1479 // Sort relocations by address. 1480 llvm::sort(Rels, RelocAddressLess); 1481 1482 StringRef SegmentName = ""; 1483 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1484 DataRefImpl DR = Section.getRawDataRefImpl(); 1485 SegmentName = MachO->getSectionFinalSegmentName(DR); 1486 } 1487 StringRef SectionName; 1488 error(Section.getName(SectionName)); 1489 1490 // If the section has no symbol at the start, just insert a dummy one. 1491 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1492 Symbols.insert( 1493 Symbols.begin(), 1494 std::make_tuple(SectionAddr, SectionName, 1495 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1496 } 1497 1498 SmallString<40> Comments; 1499 raw_svector_ostream CommentStream(Comments); 1500 1501 StringRef BytesStr; 1502 error(Section.getContents(BytesStr)); 1503 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1504 BytesStr.size()); 1505 1506 uint64_t Size; 1507 uint64_t Index; 1508 bool PrintedSection = false; 1509 1510 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1511 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1512 // Disassemble symbol by symbol. 1513 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1514 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1515 // The end is either the section end or the beginning of the next 1516 // symbol. 1517 uint64_t End = 1518 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1519 // Don't try to disassemble beyond the end of section contents. 1520 if (End > SectSize) 1521 End = SectSize; 1522 // If this symbol has the same address as the next symbol, then skip it. 1523 if (Start >= End) 1524 continue; 1525 1526 // Check if we need to skip symbol 1527 // Skip if the symbol's data is not between StartAddress and StopAddress 1528 if (End + SectionAddr < StartAddress || 1529 Start + SectionAddr > StopAddress) { 1530 continue; 1531 } 1532 1533 /// Skip if user requested specific symbols and this is not in the list 1534 if (!DisasmFuncsSet.empty() && 1535 !DisasmFuncsSet.count(std::get<1>(Symbols[si]))) 1536 continue; 1537 1538 if (!PrintedSection) { 1539 PrintedSection = true; 1540 outs() << "Disassembly of section "; 1541 if (!SegmentName.empty()) 1542 outs() << SegmentName << ","; 1543 outs() << SectionName << ':'; 1544 } 1545 1546 // Stop disassembly at the stop address specified 1547 if (End + SectionAddr > StopAddress) 1548 End = StopAddress - SectionAddr; 1549 1550 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1551 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1552 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1553 Start += 256; 1554 } 1555 if (si == se - 1 || 1556 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1557 // cut trailing zeroes at the end of kernel 1558 // cut up to 256 bytes 1559 const uint64_t EndAlign = 256; 1560 const auto Limit = End - (std::min)(EndAlign, End - Start); 1561 while (End > Limit && 1562 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1563 End -= 4; 1564 } 1565 } 1566 1567 auto PrintSymbol = [](StringRef Name) { 1568 outs() << '\n' << Name << ":\n"; 1569 }; 1570 StringRef SymbolName = std::get<1>(Symbols[si]); 1571 if (Demangle) { 1572 char *DemangledSymbol = nullptr; 1573 size_t Size = 0; 1574 int Status = -1; 1575 if (SymbolName.startswith("_Z")) 1576 DemangledSymbol = itaniumDemangle(SymbolName.data(), DemangledSymbol, 1577 &Size, &Status); 1578 else if (SymbolName.startswith("?")) 1579 DemangledSymbol = microsoftDemangle(SymbolName.data(), 1580 DemangledSymbol, &Size, &Status); 1581 1582 if (Status == 0 && DemangledSymbol) 1583 PrintSymbol(StringRef(DemangledSymbol)); 1584 else 1585 PrintSymbol(SymbolName); 1586 1587 if (DemangledSymbol) 1588 free(DemangledSymbol); 1589 } else 1590 PrintSymbol(SymbolName); 1591 1592 // Don't print raw contents of a virtual section. A virtual section 1593 // doesn't have any contents in the file. 1594 if (Section.isVirtual()) { 1595 outs() << "...\n"; 1596 continue; 1597 } 1598 1599 #ifndef NDEBUG 1600 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1601 #else 1602 raw_ostream &DebugOut = nulls(); 1603 #endif 1604 1605 for (Index = Start; Index < End; Index += Size) { 1606 MCInst Inst; 1607 1608 if (Index + SectionAddr < StartAddress || 1609 Index + SectionAddr > StopAddress) { 1610 // skip byte by byte till StartAddress is reached 1611 Size = 1; 1612 continue; 1613 } 1614 // AArch64 ELF binaries can interleave data and text in the 1615 // same section. We rely on the markers introduced to 1616 // understand what we need to dump. If the data marker is within a 1617 // function, it is denoted as a word/short etc 1618 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1619 !DisassembleAll) { 1620 uint64_t Stride = 0; 1621 1622 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1623 DataMappingSymsAddr.end(), Index); 1624 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1625 // Switch to data. 1626 while (Index < End) { 1627 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1628 outs() << "\t"; 1629 if (Index + 4 <= End) { 1630 Stride = 4; 1631 dumpBytes(Bytes.slice(Index, 4), outs()); 1632 outs() << "\t.word\t"; 1633 uint32_t Data = 0; 1634 if (Obj->isLittleEndian()) { 1635 const auto Word = 1636 reinterpret_cast<const support::ulittle32_t *>( 1637 Bytes.data() + Index); 1638 Data = *Word; 1639 } else { 1640 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1641 Bytes.data() + Index); 1642 Data = *Word; 1643 } 1644 outs() << "0x" << format("%08" PRIx32, Data); 1645 } else if (Index + 2 <= End) { 1646 Stride = 2; 1647 dumpBytes(Bytes.slice(Index, 2), outs()); 1648 outs() << "\t\t.short\t"; 1649 uint16_t Data = 0; 1650 if (Obj->isLittleEndian()) { 1651 const auto Short = 1652 reinterpret_cast<const support::ulittle16_t *>( 1653 Bytes.data() + Index); 1654 Data = *Short; 1655 } else { 1656 const auto Short = 1657 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1658 Index); 1659 Data = *Short; 1660 } 1661 outs() << "0x" << format("%04" PRIx16, Data); 1662 } else { 1663 Stride = 1; 1664 dumpBytes(Bytes.slice(Index, 1), outs()); 1665 outs() << "\t\t.byte\t"; 1666 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1667 } 1668 Index += Stride; 1669 outs() << "\n"; 1670 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1671 TextMappingSymsAddr.end(), Index); 1672 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1673 break; 1674 } 1675 } 1676 } 1677 1678 // If there is a data symbol inside an ELF text section and we are only 1679 // disassembling text (applicable all architectures), 1680 // we are in a situation where we must print the data and not 1681 // disassemble it. 1682 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1683 !DisassembleAll && Section.isText()) { 1684 // print out data up to 8 bytes at a time in hex and ascii 1685 uint8_t AsciiData[9] = {'\0'}; 1686 uint8_t Byte; 1687 int NumBytes = 0; 1688 1689 for (Index = Start; Index < End; Index += 1) { 1690 if (((SectionAddr + Index) < StartAddress) || 1691 ((SectionAddr + Index) > StopAddress)) 1692 continue; 1693 if (NumBytes == 0) { 1694 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1695 outs() << "\t"; 1696 } 1697 Byte = Bytes.slice(Index)[0]; 1698 outs() << format(" %02x", Byte); 1699 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1700 1701 uint8_t IndentOffset = 0; 1702 NumBytes++; 1703 if (Index == End - 1 || NumBytes > 8) { 1704 // Indent the space for less than 8 bytes data. 1705 // 2 spaces for byte and one for space between bytes 1706 IndentOffset = 3 * (8 - NumBytes); 1707 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1708 AsciiData[Excess] = '\0'; 1709 NumBytes = 8; 1710 } 1711 if (NumBytes == 8) { 1712 AsciiData[8] = '\0'; 1713 outs() << std::string(IndentOffset, ' ') << " "; 1714 outs() << reinterpret_cast<char *>(AsciiData); 1715 outs() << '\n'; 1716 NumBytes = 0; 1717 } 1718 } 1719 } 1720 if (Index >= End) 1721 break; 1722 1723 // Disassemble a real instruction or a data when disassemble all is 1724 // provided 1725 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1726 SectionAddr + Index, DebugOut, 1727 CommentStream); 1728 if (Size == 0) 1729 Size = 1; 1730 1731 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1732 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1733 *STI, &SP, &Rels); 1734 outs() << CommentStream.str(); 1735 Comments.clear(); 1736 1737 // Try to resolve the target of a call, tail call, etc. to a specific 1738 // symbol. 1739 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1740 MIA->isConditionalBranch(Inst))) { 1741 uint64_t Target; 1742 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1743 // In a relocatable object, the target's section must reside in 1744 // the same section as the call instruction or it is accessed 1745 // through a relocation. 1746 // 1747 // In a non-relocatable object, the target may be in any section. 1748 // 1749 // N.B. We don't walk the relocations in the relocatable case yet. 1750 auto *TargetSectionSymbols = &Symbols; 1751 if (!Obj->isRelocatableObject()) { 1752 auto SectionAddress = std::upper_bound( 1753 SectionAddresses.begin(), SectionAddresses.end(), Target, 1754 [](uint64_t LHS, 1755 const std::pair<uint64_t, SectionRef> &RHS) { 1756 return LHS < RHS.first; 1757 }); 1758 if (SectionAddress != SectionAddresses.begin()) { 1759 --SectionAddress; 1760 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1761 } else { 1762 TargetSectionSymbols = &AbsoluteSymbols; 1763 } 1764 } 1765 1766 // Find the first symbol in the section whose offset is less than 1767 // or equal to the target. If there isn't a section that contains 1768 // the target, find the nearest preceding absolute symbol. 1769 auto TargetSym = std::upper_bound( 1770 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1771 Target, [](uint64_t LHS, 1772 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1773 return LHS < std::get<0>(RHS); 1774 }); 1775 if (TargetSym == TargetSectionSymbols->begin()) { 1776 TargetSectionSymbols = &AbsoluteSymbols; 1777 TargetSym = std::upper_bound( 1778 AbsoluteSymbols.begin(), AbsoluteSymbols.end(), 1779 Target, [](uint64_t LHS, 1780 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1781 return LHS < std::get<0>(RHS); 1782 }); 1783 } 1784 if (TargetSym != TargetSectionSymbols->begin()) { 1785 --TargetSym; 1786 uint64_t TargetAddress = std::get<0>(*TargetSym); 1787 StringRef TargetName = std::get<1>(*TargetSym); 1788 outs() << " <" << TargetName; 1789 uint64_t Disp = Target - TargetAddress; 1790 if (Disp) 1791 outs() << "+0x" << Twine::utohexstr(Disp); 1792 outs() << '>'; 1793 } 1794 } 1795 } 1796 outs() << "\n"; 1797 1798 // Hexagon does this in pretty printer 1799 if (Obj->getArch() != Triple::hexagon) 1800 // Print relocation for instruction. 1801 while (rel_cur != rel_end) { 1802 bool hidden = getHidden(*rel_cur); 1803 uint64_t addr = rel_cur->getOffset(); 1804 SmallString<16> name; 1805 SmallString<32> val; 1806 1807 // If this relocation is hidden, skip it. 1808 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1809 ++rel_cur; 1810 continue; 1811 } 1812 1813 // Stop when rel_cur's address is past the current instruction. 1814 if (addr >= Index + Size) break; 1815 rel_cur->getTypeName(name); 1816 error(getRelocationValueString(*rel_cur, val)); 1817 outs() << format(Fmt.data(), SectionAddr + addr) << name 1818 << "\t" << val << "\n"; 1819 ++rel_cur; 1820 } 1821 } 1822 } 1823 } 1824 } 1825 1826 void llvm::PrintRelocations(const ObjectFile *Obj) { 1827 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1828 "%08" PRIx64; 1829 // Regular objdump doesn't print relocations in non-relocatable object 1830 // files. 1831 if (!Obj->isRelocatableObject()) 1832 return; 1833 1834 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1835 if (Section.relocation_begin() == Section.relocation_end()) 1836 continue; 1837 StringRef secname; 1838 error(Section.getName(secname)); 1839 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1840 for (const RelocationRef &Reloc : Section.relocations()) { 1841 bool hidden = getHidden(Reloc); 1842 uint64_t address = Reloc.getOffset(); 1843 SmallString<32> relocname; 1844 SmallString<32> valuestr; 1845 if (address < StartAddress || address > StopAddress || hidden) 1846 continue; 1847 Reloc.getTypeName(relocname); 1848 error(getRelocationValueString(Reloc, valuestr)); 1849 outs() << format(Fmt.data(), address) << " " << relocname << " " 1850 << valuestr << "\n"; 1851 } 1852 outs() << "\n"; 1853 } 1854 } 1855 1856 void llvm::PrintDynamicRelocations(const ObjectFile *Obj) { 1857 1858 // For the moment, this option is for ELF only 1859 if (!Obj->isELF()) 1860 return; 1861 1862 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1863 1864 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1865 error("not a dynamic object"); 1866 return; 1867 } 1868 1869 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1870 1871 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1872 if (DynRelSec.empty()) 1873 return; 1874 1875 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1876 for (const SectionRef &Section : DynRelSec) { 1877 if (Section.relocation_begin() == Section.relocation_end()) 1878 continue; 1879 for (const RelocationRef &Reloc : Section.relocations()) { 1880 uint64_t address = Reloc.getOffset(); 1881 SmallString<32> relocname; 1882 SmallString<32> valuestr; 1883 Reloc.getTypeName(relocname); 1884 error(getRelocationValueString(Reloc, valuestr)); 1885 outs() << format(Fmt.data(), address) << " " << relocname << " " 1886 << valuestr << "\n"; 1887 } 1888 } 1889 } 1890 1891 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1892 outs() << "Sections:\n" 1893 "Idx Name Size Address Type\n"; 1894 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1895 StringRef Name; 1896 error(Section.getName(Name)); 1897 uint64_t Address = Section.getAddress(); 1898 uint64_t Size = Section.getSize(); 1899 bool Text = Section.isText(); 1900 bool Data = Section.isData(); 1901 bool BSS = Section.isBSS(); 1902 std::string Type = (std::string(Text ? "TEXT " : "") + 1903 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1904 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1905 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1906 Address, Type.c_str()); 1907 } 1908 } 1909 1910 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1911 std::error_code EC; 1912 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1913 StringRef Name; 1914 StringRef Contents; 1915 error(Section.getName(Name)); 1916 uint64_t BaseAddr = Section.getAddress(); 1917 uint64_t Size = Section.getSize(); 1918 if (!Size) 1919 continue; 1920 1921 outs() << "Contents of section " << Name << ":\n"; 1922 if (Section.isBSS()) { 1923 outs() << format("<skipping contents of bss section at [%04" PRIx64 1924 ", %04" PRIx64 ")>\n", 1925 BaseAddr, BaseAddr + Size); 1926 continue; 1927 } 1928 1929 error(Section.getContents(Contents)); 1930 1931 // Dump out the content as hex and printable ascii characters. 1932 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1933 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1934 // Dump line of hex. 1935 for (std::size_t i = 0; i < 16; ++i) { 1936 if (i != 0 && i % 4 == 0) 1937 outs() << ' '; 1938 if (addr + i < end) 1939 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1940 << hexdigit(Contents[addr + i] & 0xF, true); 1941 else 1942 outs() << " "; 1943 } 1944 // Print ascii. 1945 outs() << " "; 1946 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1947 if (isPrint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1948 outs() << Contents[addr + i]; 1949 else 1950 outs() << "."; 1951 } 1952 outs() << "\n"; 1953 } 1954 } 1955 } 1956 1957 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1958 StringRef ArchitectureName) { 1959 outs() << "SYMBOL TABLE:\n"; 1960 1961 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1962 printCOFFSymbolTable(coff); 1963 return; 1964 } 1965 for (const SymbolRef &Symbol : o->symbols()) { 1966 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1967 if (!AddressOrError) 1968 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1969 ArchitectureName); 1970 uint64_t Address = *AddressOrError; 1971 if ((Address < StartAddress) || (Address > StopAddress)) 1972 continue; 1973 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1974 if (!TypeOrError) 1975 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1976 ArchitectureName); 1977 SymbolRef::Type Type = *TypeOrError; 1978 uint32_t Flags = Symbol.getFlags(); 1979 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1980 if (!SectionOrErr) 1981 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1982 ArchitectureName); 1983 section_iterator Section = *SectionOrErr; 1984 StringRef Name; 1985 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1986 Section->getName(Name); 1987 } else { 1988 Expected<StringRef> NameOrErr = Symbol.getName(); 1989 if (!NameOrErr) 1990 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1991 ArchitectureName); 1992 Name = *NameOrErr; 1993 } 1994 1995 bool Global = Flags & SymbolRef::SF_Global; 1996 bool Weak = Flags & SymbolRef::SF_Weak; 1997 bool Absolute = Flags & SymbolRef::SF_Absolute; 1998 bool Common = Flags & SymbolRef::SF_Common; 1999 bool Hidden = Flags & SymbolRef::SF_Hidden; 2000 2001 char GlobLoc = ' '; 2002 if (Type != SymbolRef::ST_Unknown) 2003 GlobLoc = Global ? 'g' : 'l'; 2004 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2005 ? 'd' : ' '; 2006 char FileFunc = ' '; 2007 if (Type == SymbolRef::ST_File) 2008 FileFunc = 'f'; 2009 else if (Type == SymbolRef::ST_Function) 2010 FileFunc = 'F'; 2011 2012 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 2013 "%08" PRIx64; 2014 2015 outs() << format(Fmt, Address) << " " 2016 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2017 << (Weak ? 'w' : ' ') // Weak? 2018 << ' ' // Constructor. Not supported yet. 2019 << ' ' // Warning. Not supported yet. 2020 << ' ' // Indirect reference to another symbol. 2021 << Debug // Debugging (d) or dynamic (D) symbol. 2022 << FileFunc // Name of function (F), file (f) or object (O). 2023 << ' '; 2024 if (Absolute) { 2025 outs() << "*ABS*"; 2026 } else if (Common) { 2027 outs() << "*COM*"; 2028 } else if (Section == o->section_end()) { 2029 outs() << "*UND*"; 2030 } else { 2031 if (const MachOObjectFile *MachO = 2032 dyn_cast<const MachOObjectFile>(o)) { 2033 DataRefImpl DR = Section->getRawDataRefImpl(); 2034 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 2035 outs() << SegmentName << ","; 2036 } 2037 StringRef SectionName; 2038 error(Section->getName(SectionName)); 2039 outs() << SectionName; 2040 } 2041 2042 outs() << '\t'; 2043 if (Common || isa<ELFObjectFileBase>(o)) { 2044 uint64_t Val = 2045 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2046 outs() << format("\t %08" PRIx64 " ", Val); 2047 } 2048 2049 if (Hidden) { 2050 outs() << ".hidden "; 2051 } 2052 outs() << Name 2053 << '\n'; 2054 } 2055 } 2056 2057 static void PrintUnwindInfo(const ObjectFile *o) { 2058 outs() << "Unwind info:\n\n"; 2059 2060 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 2061 printCOFFUnwindInfo(coff); 2062 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2063 printMachOUnwindInfo(MachO); 2064 else { 2065 // TODO: Extract DWARF dump tool to objdump. 2066 errs() << "This operation is only currently supported " 2067 "for COFF and MachO object files.\n"; 2068 return; 2069 } 2070 } 2071 2072 void llvm::printExportsTrie(const ObjectFile *o) { 2073 outs() << "Exports trie:\n"; 2074 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2075 printMachOExportsTrie(MachO); 2076 else { 2077 errs() << "This operation is only currently supported " 2078 "for Mach-O executable files.\n"; 2079 return; 2080 } 2081 } 2082 2083 void llvm::printRebaseTable(ObjectFile *o) { 2084 outs() << "Rebase table:\n"; 2085 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2086 printMachORebaseTable(MachO); 2087 else { 2088 errs() << "This operation is only currently supported " 2089 "for Mach-O executable files.\n"; 2090 return; 2091 } 2092 } 2093 2094 void llvm::printBindTable(ObjectFile *o) { 2095 outs() << "Bind table:\n"; 2096 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2097 printMachOBindTable(MachO); 2098 else { 2099 errs() << "This operation is only currently supported " 2100 "for Mach-O executable files.\n"; 2101 return; 2102 } 2103 } 2104 2105 void llvm::printLazyBindTable(ObjectFile *o) { 2106 outs() << "Lazy bind table:\n"; 2107 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2108 printMachOLazyBindTable(MachO); 2109 else { 2110 errs() << "This operation is only currently supported " 2111 "for Mach-O executable files.\n"; 2112 return; 2113 } 2114 } 2115 2116 void llvm::printWeakBindTable(ObjectFile *o) { 2117 outs() << "Weak bind table:\n"; 2118 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2119 printMachOWeakBindTable(MachO); 2120 else { 2121 errs() << "This operation is only currently supported " 2122 "for Mach-O executable files.\n"; 2123 return; 2124 } 2125 } 2126 2127 /// Dump the raw contents of the __clangast section so the output can be piped 2128 /// into llvm-bcanalyzer. 2129 void llvm::printRawClangAST(const ObjectFile *Obj) { 2130 if (outs().is_displayed()) { 2131 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 2132 "the clang ast section.\n" 2133 "Please redirect the output to a file or another program such as " 2134 "llvm-bcanalyzer.\n"; 2135 return; 2136 } 2137 2138 StringRef ClangASTSectionName("__clangast"); 2139 if (isa<COFFObjectFile>(Obj)) { 2140 ClangASTSectionName = "clangast"; 2141 } 2142 2143 Optional<object::SectionRef> ClangASTSection; 2144 for (auto Sec : ToolSectionFilter(*Obj)) { 2145 StringRef Name; 2146 Sec.getName(Name); 2147 if (Name == ClangASTSectionName) { 2148 ClangASTSection = Sec; 2149 break; 2150 } 2151 } 2152 if (!ClangASTSection) 2153 return; 2154 2155 StringRef ClangASTContents; 2156 error(ClangASTSection.getValue().getContents(ClangASTContents)); 2157 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2158 } 2159 2160 static void printFaultMaps(const ObjectFile *Obj) { 2161 const char *FaultMapSectionName = nullptr; 2162 2163 if (isa<ELFObjectFileBase>(Obj)) { 2164 FaultMapSectionName = ".llvm_faultmaps"; 2165 } else if (isa<MachOObjectFile>(Obj)) { 2166 FaultMapSectionName = "__llvm_faultmaps"; 2167 } else { 2168 errs() << "This operation is only currently supported " 2169 "for ELF and Mach-O executable files.\n"; 2170 return; 2171 } 2172 2173 Optional<object::SectionRef> FaultMapSection; 2174 2175 for (auto Sec : ToolSectionFilter(*Obj)) { 2176 StringRef Name; 2177 Sec.getName(Name); 2178 if (Name == FaultMapSectionName) { 2179 FaultMapSection = Sec; 2180 break; 2181 } 2182 } 2183 2184 outs() << "FaultMap table:\n"; 2185 2186 if (!FaultMapSection.hasValue()) { 2187 outs() << "<not found>\n"; 2188 return; 2189 } 2190 2191 StringRef FaultMapContents; 2192 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2193 2194 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2195 FaultMapContents.bytes_end()); 2196 2197 outs() << FMP; 2198 } 2199 2200 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2201 if (o->isELF()) { 2202 printELFFileHeader(o); 2203 return printELFDynamicSection(o); 2204 } 2205 if (o->isCOFF()) 2206 return printCOFFFileHeader(o); 2207 if (o->isWasm()) 2208 return printWasmFileHeader(o); 2209 if (o->isMachO()) { 2210 printMachOFileHeader(o); 2211 if (!onlyFirst) 2212 printMachOLoadCommands(o); 2213 return; 2214 } 2215 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2216 } 2217 2218 static void printFileHeaders(const ObjectFile *o) { 2219 if (!o->isELF() && !o->isCOFF()) 2220 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2221 2222 Triple::ArchType AT = o->getArch(); 2223 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2224 Expected<uint64_t> StartAddrOrErr = o->getStartAddress(); 2225 if (!StartAddrOrErr) 2226 report_error(o->getFileName(), StartAddrOrErr.takeError()); 2227 2228 StringRef Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2229 uint64_t Address = StartAddrOrErr.get(); 2230 outs() << "start address: " 2231 << "0x" << format(Fmt.data(), Address) 2232 << "\n"; 2233 } 2234 2235 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2236 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2237 if (!ModeOrErr) { 2238 errs() << "ill-formed archive entry.\n"; 2239 consumeError(ModeOrErr.takeError()); 2240 return; 2241 } 2242 sys::fs::perms Mode = ModeOrErr.get(); 2243 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2244 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2245 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2246 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2247 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2248 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2249 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2250 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2251 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2252 2253 outs() << " "; 2254 2255 Expected<unsigned> UIDOrErr = C.getUID(); 2256 if (!UIDOrErr) 2257 report_error(Filename, UIDOrErr.takeError()); 2258 unsigned UID = UIDOrErr.get(); 2259 outs() << format("%d/", UID); 2260 2261 Expected<unsigned> GIDOrErr = C.getGID(); 2262 if (!GIDOrErr) 2263 report_error(Filename, GIDOrErr.takeError()); 2264 unsigned GID = GIDOrErr.get(); 2265 outs() << format("%-d ", GID); 2266 2267 Expected<uint64_t> Size = C.getRawSize(); 2268 if (!Size) 2269 report_error(Filename, Size.takeError()); 2270 outs() << format("%6" PRId64, Size.get()) << " "; 2271 2272 StringRef RawLastModified = C.getRawLastModified(); 2273 unsigned Seconds; 2274 if (RawLastModified.getAsInteger(10, Seconds)) 2275 outs() << "(date: \"" << RawLastModified 2276 << "\" contains non-decimal chars) "; 2277 else { 2278 // Since ctime(3) returns a 26 character string of the form: 2279 // "Sun Sep 16 01:03:52 1973\n\0" 2280 // just print 24 characters. 2281 time_t t = Seconds; 2282 outs() << format("%.24s ", ctime(&t)); 2283 } 2284 2285 StringRef Name = ""; 2286 Expected<StringRef> NameOrErr = C.getName(); 2287 if (!NameOrErr) { 2288 consumeError(NameOrErr.takeError()); 2289 Expected<StringRef> RawNameOrErr = C.getRawName(); 2290 if (!RawNameOrErr) 2291 report_error(Filename, NameOrErr.takeError()); 2292 Name = RawNameOrErr.get(); 2293 } else { 2294 Name = NameOrErr.get(); 2295 } 2296 outs() << Name << "\n"; 2297 } 2298 2299 static void DumpObject(ObjectFile *o, const Archive *a = nullptr, 2300 const Archive::Child *c = nullptr) { 2301 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2302 // Avoid other output when using a raw option. 2303 if (!RawClangAST) { 2304 outs() << '\n'; 2305 if (a) 2306 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2307 else 2308 outs() << o->getFileName(); 2309 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2310 } 2311 2312 if (ArchiveHeaders && !MachOOpt) 2313 printArchiveChild(a->getFileName(), *c); 2314 if (Disassemble) 2315 DisassembleObject(o, Relocations); 2316 if (Relocations && !Disassemble) 2317 PrintRelocations(o); 2318 if (DynamicRelocations) 2319 PrintDynamicRelocations(o); 2320 if (SectionHeaders) 2321 PrintSectionHeaders(o); 2322 if (SectionContents) 2323 PrintSectionContents(o); 2324 if (SymbolTable) 2325 PrintSymbolTable(o, ArchiveName); 2326 if (UnwindInfo) 2327 PrintUnwindInfo(o); 2328 if (PrivateHeaders || FirstPrivateHeader) 2329 printPrivateFileHeaders(o, FirstPrivateHeader); 2330 if (FileHeaders) 2331 printFileHeaders(o); 2332 if (ExportsTrie) 2333 printExportsTrie(o); 2334 if (Rebase) 2335 printRebaseTable(o); 2336 if (Bind) 2337 printBindTable(o); 2338 if (LazyBind) 2339 printLazyBindTable(o); 2340 if (WeakBind) 2341 printWeakBindTable(o); 2342 if (RawClangAST) 2343 printRawClangAST(o); 2344 if (PrintFaultMaps) 2345 printFaultMaps(o); 2346 if (DwarfDumpType != DIDT_Null) { 2347 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*o); 2348 // Dump the complete DWARF structure. 2349 DIDumpOptions DumpOpts; 2350 DumpOpts.DumpType = DwarfDumpType; 2351 DICtx->dump(outs(), DumpOpts); 2352 } 2353 } 2354 2355 static void DumpObject(const COFFImportFile *I, const Archive *A, 2356 const Archive::Child *C = nullptr) { 2357 StringRef ArchiveName = A ? A->getFileName() : ""; 2358 2359 // Avoid other output when using a raw option. 2360 if (!RawClangAST) 2361 outs() << '\n' 2362 << ArchiveName << "(" << I->getFileName() << ")" 2363 << ":\tfile format COFF-import-file" 2364 << "\n\n"; 2365 2366 if (ArchiveHeaders && !MachOOpt) 2367 printArchiveChild(A->getFileName(), *C); 2368 if (SymbolTable) 2369 printCOFFSymbolTable(I); 2370 } 2371 2372 /// Dump each object file in \a a; 2373 static void DumpArchive(const Archive *a) { 2374 Error Err = Error::success(); 2375 for (auto &C : a->children(Err)) { 2376 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2377 if (!ChildOrErr) { 2378 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2379 report_error(a->getFileName(), C, std::move(E)); 2380 continue; 2381 } 2382 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2383 DumpObject(o, a, &C); 2384 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2385 DumpObject(I, a, &C); 2386 else 2387 report_error(a->getFileName(), object_error::invalid_file_type); 2388 } 2389 if (Err) 2390 report_error(a->getFileName(), std::move(Err)); 2391 } 2392 2393 /// Open file and figure out how to dump it. 2394 static void DumpInput(StringRef file) { 2395 2396 // If we are using the Mach-O specific object file parser, then let it parse 2397 // the file and process the command line options. So the -arch flags can 2398 // be used to select specific slices, etc. 2399 if (MachOOpt) { 2400 ParseInputMachO(file); 2401 return; 2402 } 2403 2404 // Attempt to open the binary. 2405 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2406 if (!BinaryOrErr) 2407 report_error(file, BinaryOrErr.takeError()); 2408 Binary &Binary = *BinaryOrErr.get().getBinary(); 2409 2410 if (Archive *a = dyn_cast<Archive>(&Binary)) 2411 DumpArchive(a); 2412 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2413 DumpObject(o); 2414 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2415 ParseInputMachO(UB); 2416 else 2417 report_error(file, object_error::invalid_file_type); 2418 } 2419 2420 int main(int argc, char **argv) { 2421 InitLLVM X(argc, argv); 2422 2423 // Initialize targets and assembly printers/parsers. 2424 llvm::InitializeAllTargetInfos(); 2425 llvm::InitializeAllTargetMCs(); 2426 llvm::InitializeAllDisassemblers(); 2427 2428 // Register the target printer for --version. 2429 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2430 2431 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2432 2433 ToolName = argv[0]; 2434 2435 // Defaults to a.out if no filenames specified. 2436 if (InputFilenames.size() == 0) 2437 InputFilenames.push_back("a.out"); 2438 2439 if (AllHeaders) 2440 PrivateHeaders = Relocations = SectionHeaders = SymbolTable = true; 2441 2442 if (DisassembleAll || PrintSource || PrintLines) 2443 Disassemble = true; 2444 2445 if (!Disassemble 2446 && !Relocations 2447 && !DynamicRelocations 2448 && !SectionHeaders 2449 && !SectionContents 2450 && !SymbolTable 2451 && !UnwindInfo 2452 && !PrivateHeaders 2453 && !FileHeaders 2454 && !FirstPrivateHeader 2455 && !ExportsTrie 2456 && !Rebase 2457 && !Bind 2458 && !LazyBind 2459 && !WeakBind 2460 && !RawClangAST 2461 && !(UniversalHeaders && MachOOpt) 2462 && !ArchiveHeaders 2463 && !(IndirectSymbols && MachOOpt) 2464 && !(DataInCode && MachOOpt) 2465 && !(LinkOptHints && MachOOpt) 2466 && !(InfoPlist && MachOOpt) 2467 && !(DylibsUsed && MachOOpt) 2468 && !(DylibId && MachOOpt) 2469 && !(ObjcMetaData && MachOOpt) 2470 && !(FilterSections.size() != 0 && MachOOpt) 2471 && !PrintFaultMaps 2472 && DwarfDumpType == DIDT_Null) { 2473 cl::PrintHelpMessage(); 2474 return 2; 2475 } 2476 2477 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2478 DisassembleFunctions.end()); 2479 2480 llvm::for_each(InputFilenames, DumpInput); 2481 2482 return EXIT_SUCCESS; 2483 } 2484