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("reloc", 110 cl::desc("Display the relocation entries in the file")); 111 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 112 cl::NotHidden, 113 cl::aliasopt(llvm::Relocations)); 114 115 cl::opt<bool> 116 llvm::DynamicRelocations("dynamic-reloc", 117 cl::desc("Display the dynamic relocation entries in the file")); 118 static cl::alias 119 DynamicRelocationsd("R", cl::desc("Alias for --dynamic-reloc"), 120 cl::aliasopt(DynamicRelocations)); 121 122 cl::opt<bool> 123 llvm::SectionContents("full-contents", 124 cl::desc("Display the content of each section")); 125 static cl::alias SectionContentsShort("s", 126 cl::desc("Alias for --full-contents"), 127 cl::aliasopt(SectionContents)); 128 129 cl::opt<bool> llvm::SymbolTable("syms", cl::desc("Display the symbol table")); 130 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 131 cl::aliasopt(llvm::SymbolTable)); 132 133 cl::opt<bool> 134 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 135 136 cl::opt<bool> 137 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 138 139 cl::opt<bool> 140 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 141 142 cl::opt<bool> 143 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 144 145 cl::opt<bool> 146 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 147 148 cl::opt<bool> 149 llvm::RawClangAST("raw-clang-ast", 150 cl::desc("Dump the raw binary contents of the clang AST section")); 151 152 static cl::opt<bool> 153 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 154 static cl::alias 155 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 156 157 cl::opt<std::string> 158 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 159 "see -version for available targets")); 160 161 cl::opt<std::string> 162 llvm::MCPU("mcpu", 163 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 164 cl::value_desc("cpu-name"), 165 cl::init("")); 166 167 cl::opt<std::string> 168 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 169 "see -version for available targets")); 170 171 cl::opt<bool> 172 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 173 "headers for each section.")); 174 static cl::alias 175 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 176 cl::aliasopt(SectionHeaders)); 177 static cl::alias 178 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 179 cl::aliasopt(SectionHeaders)); 180 181 cl::list<std::string> 182 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 183 "With -macho dump segment,section")); 184 cl::alias 185 static FilterSectionsj("j", cl::desc("Alias for --section"), 186 cl::aliasopt(llvm::FilterSections)); 187 188 cl::list<std::string> 189 llvm::MAttrs("mattr", 190 cl::CommaSeparated, 191 cl::desc("Target specific attributes"), 192 cl::value_desc("a1,+a2,-a3,...")); 193 194 cl::opt<bool> 195 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 196 "instructions, do not print " 197 "the instruction bytes.")); 198 cl::opt<bool> 199 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 200 201 cl::opt<bool> 202 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 203 204 static cl::alias 205 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 206 cl::aliasopt(UnwindInfo)); 207 208 cl::opt<bool> 209 llvm::PrivateHeaders("private-headers", 210 cl::desc("Display format specific file headers")); 211 212 cl::opt<bool> 213 llvm::FirstPrivateHeader("private-header", 214 cl::desc("Display only the first format specific file " 215 "header")); 216 217 static cl::alias 218 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 219 cl::aliasopt(PrivateHeaders)); 220 221 cl::opt<bool> llvm::FileHeaders( 222 "file-headers", 223 cl::desc("Display the contents of the overall file header")); 224 225 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 226 cl::aliasopt(FileHeaders)); 227 228 cl::opt<bool> 229 llvm::ArchiveHeaders("archive-headers", 230 cl::desc("Display archive header information")); 231 232 cl::alias 233 ArchiveHeadersShort("a", cl::desc("Alias for --archive-headers"), 234 cl::aliasopt(ArchiveHeaders)); 235 236 cl::opt<bool> 237 llvm::PrintImmHex("print-imm-hex", 238 cl::desc("Use hex format for immediate values")); 239 240 cl::opt<bool> PrintFaultMaps("fault-map-section", 241 cl::desc("Display contents of faultmap section")); 242 243 cl::opt<DIDumpType> llvm::DwarfDumpType( 244 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 245 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 246 247 cl::opt<bool> PrintSource( 248 "source", 249 cl::desc( 250 "Display source inlined with disassembly. Implies disassemble object")); 251 252 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 253 cl::aliasopt(PrintSource)); 254 255 cl::opt<bool> PrintLines("line-numbers", 256 cl::desc("Display source line numbers with " 257 "disassembly. Implies disassemble object")); 258 259 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 260 cl::aliasopt(PrintLines)); 261 262 cl::opt<unsigned long long> 263 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 264 cl::value_desc("address"), cl::init(0)); 265 cl::opt<unsigned long long> 266 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 267 cl::value_desc("address"), cl::init(UINT64_MAX)); 268 static StringRef ToolName; 269 270 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 271 272 namespace { 273 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 274 275 class SectionFilterIterator { 276 public: 277 SectionFilterIterator(FilterPredicate P, 278 llvm::object::section_iterator const &I, 279 llvm::object::section_iterator const &E) 280 : Predicate(std::move(P)), Iterator(I), End(E) { 281 ScanPredicate(); 282 } 283 const llvm::object::SectionRef &operator*() const { return *Iterator; } 284 SectionFilterIterator &operator++() { 285 ++Iterator; 286 ScanPredicate(); 287 return *this; 288 } 289 bool operator!=(SectionFilterIterator const &Other) const { 290 return Iterator != Other.Iterator; 291 } 292 293 private: 294 void ScanPredicate() { 295 while (Iterator != End && !Predicate(*Iterator)) { 296 ++Iterator; 297 } 298 } 299 FilterPredicate Predicate; 300 llvm::object::section_iterator Iterator; 301 llvm::object::section_iterator End; 302 }; 303 304 class SectionFilter { 305 public: 306 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 307 : Predicate(std::move(P)), Object(O) {} 308 SectionFilterIterator begin() { 309 return SectionFilterIterator(Predicate, Object.section_begin(), 310 Object.section_end()); 311 } 312 SectionFilterIterator end() { 313 return SectionFilterIterator(Predicate, Object.section_end(), 314 Object.section_end()); 315 } 316 317 private: 318 FilterPredicate Predicate; 319 llvm::object::ObjectFile const &Object; 320 }; 321 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 322 return SectionFilter( 323 [](llvm::object::SectionRef const &S) { 324 if (FilterSections.empty()) 325 return true; 326 llvm::StringRef String; 327 std::error_code error = S.getName(String); 328 if (error) 329 return false; 330 return is_contained(FilterSections, String); 331 }, 332 O); 333 } 334 } 335 336 void llvm::error(std::error_code EC) { 337 if (!EC) 338 return; 339 340 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 341 errs().flush(); 342 exit(1); 343 } 344 345 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 346 errs() << ToolName << ": " << Message << ".\n"; 347 errs().flush(); 348 exit(1); 349 } 350 351 void llvm::warn(StringRef Message) { 352 errs() << ToolName << ": warning: " << Message << ".\n"; 353 errs().flush(); 354 } 355 356 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 357 Twine Message) { 358 errs() << ToolName << ": '" << File << "': " << Message << ".\n"; 359 exit(1); 360 } 361 362 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 363 std::error_code EC) { 364 assert(EC); 365 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 366 exit(1); 367 } 368 369 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 370 llvm::Error E) { 371 assert(E); 372 std::string Buf; 373 raw_string_ostream OS(Buf); 374 logAllUnhandledErrors(std::move(E), OS, ""); 375 OS.flush(); 376 errs() << ToolName << ": '" << File << "': " << Buf; 377 exit(1); 378 } 379 380 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 381 StringRef FileName, 382 llvm::Error E, 383 StringRef ArchitectureName) { 384 assert(E); 385 errs() << ToolName << ": "; 386 if (ArchiveName != "") 387 errs() << ArchiveName << "(" << FileName << ")"; 388 else 389 errs() << "'" << FileName << "'"; 390 if (!ArchitectureName.empty()) 391 errs() << " (for architecture " << ArchitectureName << ")"; 392 std::string Buf; 393 raw_string_ostream OS(Buf); 394 logAllUnhandledErrors(std::move(E), OS, ""); 395 OS.flush(); 396 errs() << ": " << Buf; 397 exit(1); 398 } 399 400 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 401 const object::Archive::Child &C, 402 llvm::Error E, 403 StringRef ArchitectureName) { 404 Expected<StringRef> NameOrErr = C.getName(); 405 // TODO: if we have a error getting the name then it would be nice to print 406 // the index of which archive member this is and or its offset in the 407 // archive instead of "???" as the name. 408 if (!NameOrErr) { 409 consumeError(NameOrErr.takeError()); 410 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 411 } else 412 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 413 ArchitectureName); 414 } 415 416 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 417 // Figure out the target triple. 418 llvm::Triple TheTriple("unknown-unknown-unknown"); 419 if (TripleName.empty()) { 420 if (Obj) { 421 TheTriple = Obj->makeTriple(); 422 } 423 } else { 424 TheTriple.setTriple(Triple::normalize(TripleName)); 425 426 // Use the triple, but also try to combine with ARM build attributes. 427 if (Obj) { 428 auto Arch = Obj->getArch(); 429 if (Arch == Triple::arm || Arch == Triple::armeb) { 430 Obj->setARMSubArch(TheTriple); 431 } 432 } 433 } 434 435 // Get the target specific parser. 436 std::string Error; 437 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 438 Error); 439 if (!TheTarget) { 440 if (Obj) 441 report_error(Obj->getFileName(), "can't find target: " + Error); 442 else 443 error("can't find target: " + Error); 444 } 445 446 // Update the triple name and return the found target. 447 TripleName = TheTriple.getTriple(); 448 return TheTarget; 449 } 450 451 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 452 return a.getOffset() < b.getOffset(); 453 } 454 455 template <class ELFT> 456 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 457 const RelocationRef &RelRef, 458 SmallVectorImpl<char> &Result) { 459 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 460 461 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 462 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 463 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 464 465 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 466 467 auto SecOrErr = EF.getSection(Rel.d.a); 468 if (!SecOrErr) 469 return errorToErrorCode(SecOrErr.takeError()); 470 const Elf_Shdr *Sec = *SecOrErr; 471 auto SymTabOrErr = EF.getSection(Sec->sh_link); 472 if (!SymTabOrErr) 473 return errorToErrorCode(SymTabOrErr.takeError()); 474 const Elf_Shdr *SymTab = *SymTabOrErr; 475 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 476 SymTab->sh_type == ELF::SHT_DYNSYM); 477 auto StrTabSec = EF.getSection(SymTab->sh_link); 478 if (!StrTabSec) 479 return errorToErrorCode(StrTabSec.takeError()); 480 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 481 if (!StrTabOrErr) 482 return errorToErrorCode(StrTabOrErr.takeError()); 483 StringRef StrTab = *StrTabOrErr; 484 int64_t addend = 0; 485 // If there is no Symbol associated with the relocation, we set the undef 486 // boolean value to 'true'. This will prevent us from calling functions that 487 // requires the relocation to be associated with a symbol. 488 bool undef = false; 489 switch (Sec->sh_type) { 490 default: 491 return object_error::parse_failed; 492 case ELF::SHT_REL: { 493 // TODO: Read implicit addend from section data. 494 break; 495 } 496 case ELF::SHT_RELA: { 497 const Elf_Rela *ERela = Obj->getRela(Rel); 498 addend = ERela->r_addend; 499 undef = ERela->getSymbol(false) == 0; 500 break; 501 } 502 } 503 StringRef Target; 504 if (!undef) { 505 symbol_iterator SI = RelRef.getSymbol(); 506 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 507 if (symb->getType() == ELF::STT_SECTION) { 508 Expected<section_iterator> SymSI = SI->getSection(); 509 if (!SymSI) 510 return errorToErrorCode(SymSI.takeError()); 511 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 512 auto SecName = EF.getSectionName(SymSec); 513 if (!SecName) 514 return errorToErrorCode(SecName.takeError()); 515 Target = *SecName; 516 } else { 517 Expected<StringRef> SymName = symb->getName(StrTab); 518 if (!SymName) 519 return errorToErrorCode(SymName.takeError()); 520 Target = *SymName; 521 } 522 } else 523 Target = "*ABS*"; 524 525 // Default scheme is to print Target, as well as "+ <addend>" for nonzero 526 // addend. Should be acceptable for all normal purposes. 527 std::string fmtbuf; 528 raw_string_ostream fmt(fmtbuf); 529 fmt << Target; 530 if (addend != 0) 531 fmt << (addend < 0 ? "" : "+") << addend; 532 fmt.flush(); 533 Result.append(fmtbuf.begin(), fmtbuf.end()); 534 return std::error_code(); 535 } 536 537 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 538 const RelocationRef &Rel, 539 SmallVectorImpl<char> &Result) { 540 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 541 return getRelocationValueString(ELF32LE, Rel, Result); 542 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 543 return getRelocationValueString(ELF64LE, Rel, Result); 544 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 545 return getRelocationValueString(ELF32BE, Rel, Result); 546 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 547 return getRelocationValueString(ELF64BE, Rel, Result); 548 } 549 550 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 551 const RelocationRef &Rel, 552 SmallVectorImpl<char> &Result) { 553 symbol_iterator SymI = Rel.getSymbol(); 554 Expected<StringRef> SymNameOrErr = SymI->getName(); 555 if (!SymNameOrErr) 556 return errorToErrorCode(SymNameOrErr.takeError()); 557 StringRef SymName = *SymNameOrErr; 558 Result.append(SymName.begin(), SymName.end()); 559 return std::error_code(); 560 } 561 562 static void printRelocationTargetName(const MachOObjectFile *O, 563 const MachO::any_relocation_info &RE, 564 raw_string_ostream &fmt) { 565 bool IsScattered = O->isRelocationScattered(RE); 566 567 // Target of a scattered relocation is an address. In the interest of 568 // generating pretty output, scan through the symbol table looking for a 569 // symbol that aligns with that address. If we find one, print it. 570 // Otherwise, we just print the hex address of the target. 571 if (IsScattered) { 572 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 573 574 for (const SymbolRef &Symbol : O->symbols()) { 575 std::error_code ec; 576 Expected<uint64_t> Addr = Symbol.getAddress(); 577 if (!Addr) 578 report_error(O->getFileName(), Addr.takeError()); 579 if (*Addr != Val) 580 continue; 581 Expected<StringRef> Name = Symbol.getName(); 582 if (!Name) 583 report_error(O->getFileName(), Name.takeError()); 584 fmt << *Name; 585 return; 586 } 587 588 // If we couldn't find a symbol that this relocation refers to, try 589 // to find a section beginning instead. 590 for (const SectionRef &Section : ToolSectionFilter(*O)) { 591 std::error_code ec; 592 593 StringRef Name; 594 uint64_t Addr = Section.getAddress(); 595 if (Addr != Val) 596 continue; 597 if ((ec = Section.getName(Name))) 598 report_error(O->getFileName(), ec); 599 fmt << Name; 600 return; 601 } 602 603 fmt << format("0x%x", Val); 604 return; 605 } 606 607 StringRef S; 608 bool isExtern = O->getPlainRelocationExternal(RE); 609 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 610 611 if (O->getAnyRelocationType(RE) == MachO::ARM64_RELOC_ADDEND) { 612 fmt << format("0x%0" PRIx64, Val); 613 return; 614 } else if (isExtern) { 615 symbol_iterator SI = O->symbol_begin(); 616 advance(SI, Val); 617 Expected<StringRef> SOrErr = SI->getName(); 618 if (!SOrErr) 619 report_error(O->getFileName(), SOrErr.takeError()); 620 S = *SOrErr; 621 } else { 622 section_iterator SI = O->section_begin(); 623 // Adjust for the fact that sections are 1-indexed. 624 if (Val == 0) { 625 fmt << "0 (?,?)"; 626 return; 627 } 628 uint32_t i = Val - 1; 629 while (i != 0 && SI != O->section_end()) { 630 i--; 631 advance(SI, 1); 632 } 633 if (SI == O->section_end()) 634 fmt << Val << " (?,?)"; 635 else 636 SI->getName(S); 637 } 638 639 fmt << S; 640 } 641 642 static std::error_code getRelocationValueString(const WasmObjectFile *Obj, 643 const RelocationRef &RelRef, 644 SmallVectorImpl<char> &Result) { 645 const wasm::WasmRelocation& Rel = Obj->getWasmRelocation(RelRef); 646 symbol_iterator SI = RelRef.getSymbol(); 647 std::string fmtbuf; 648 raw_string_ostream fmt(fmtbuf); 649 if (SI == Obj->symbol_end()) { 650 // Not all wasm relocations have symbols associated with them. 651 // In particular R_WEBASSEMBLY_TYPE_INDEX_LEB. 652 fmt << Rel.Index; 653 } else { 654 Expected<StringRef> SymNameOrErr = SI->getName(); 655 if (!SymNameOrErr) 656 return errorToErrorCode(SymNameOrErr.takeError()); 657 StringRef SymName = *SymNameOrErr; 658 Result.append(SymName.begin(), SymName.end()); 659 } 660 fmt << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 661 fmt.flush(); 662 Result.append(fmtbuf.begin(), fmtbuf.end()); 663 return std::error_code(); 664 } 665 666 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 667 const RelocationRef &RelRef, 668 SmallVectorImpl<char> &Result) { 669 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 670 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 671 672 unsigned Arch = Obj->getArch(); 673 674 std::string fmtbuf; 675 raw_string_ostream fmt(fmtbuf); 676 unsigned Type = Obj->getAnyRelocationType(RE); 677 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 678 679 // Determine any addends that should be displayed with the relocation. 680 // These require decoding the relocation type, which is triple-specific. 681 682 // X86_64 has entirely custom relocation types. 683 if (Arch == Triple::x86_64) { 684 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 685 686 switch (Type) { 687 case MachO::X86_64_RELOC_GOT_LOAD: 688 case MachO::X86_64_RELOC_GOT: { 689 printRelocationTargetName(Obj, RE, fmt); 690 fmt << "@GOT"; 691 if (isPCRel) 692 fmt << "PCREL"; 693 break; 694 } 695 case MachO::X86_64_RELOC_SUBTRACTOR: { 696 DataRefImpl RelNext = Rel; 697 Obj->moveRelocationNext(RelNext); 698 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 699 700 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 701 // X86_64_RELOC_UNSIGNED. 702 // NOTE: Scattered relocations don't exist on x86_64. 703 unsigned RType = Obj->getAnyRelocationType(RENext); 704 if (RType != MachO::X86_64_RELOC_UNSIGNED) 705 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 706 "X86_64_RELOC_SUBTRACTOR."); 707 708 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 709 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 710 printRelocationTargetName(Obj, RENext, fmt); 711 fmt << "-"; 712 printRelocationTargetName(Obj, RE, fmt); 713 break; 714 } 715 case MachO::X86_64_RELOC_TLV: 716 printRelocationTargetName(Obj, RE, fmt); 717 fmt << "@TLV"; 718 if (isPCRel) 719 fmt << "P"; 720 break; 721 case MachO::X86_64_RELOC_SIGNED_1: 722 printRelocationTargetName(Obj, RE, fmt); 723 fmt << "-1"; 724 break; 725 case MachO::X86_64_RELOC_SIGNED_2: 726 printRelocationTargetName(Obj, RE, fmt); 727 fmt << "-2"; 728 break; 729 case MachO::X86_64_RELOC_SIGNED_4: 730 printRelocationTargetName(Obj, RE, fmt); 731 fmt << "-4"; 732 break; 733 default: 734 printRelocationTargetName(Obj, RE, fmt); 735 break; 736 } 737 // X86 and ARM share some relocation types in common. 738 } else if (Arch == Triple::x86 || Arch == Triple::arm || 739 Arch == Triple::ppc) { 740 // Generic relocation types... 741 switch (Type) { 742 case MachO::GENERIC_RELOC_PAIR: // prints no info 743 return std::error_code(); 744 case MachO::GENERIC_RELOC_SECTDIFF: { 745 DataRefImpl RelNext = Rel; 746 Obj->moveRelocationNext(RelNext); 747 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 748 749 // X86 sect diff's must be followed by a relocation of type 750 // GENERIC_RELOC_PAIR. 751 unsigned RType = Obj->getAnyRelocationType(RENext); 752 753 if (RType != MachO::GENERIC_RELOC_PAIR) 754 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 755 "GENERIC_RELOC_SECTDIFF."); 756 757 printRelocationTargetName(Obj, RE, fmt); 758 fmt << "-"; 759 printRelocationTargetName(Obj, RENext, fmt); 760 break; 761 } 762 } 763 764 if (Arch == Triple::x86 || Arch == Triple::ppc) { 765 switch (Type) { 766 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 767 DataRefImpl RelNext = Rel; 768 Obj->moveRelocationNext(RelNext); 769 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 770 771 // X86 sect diff's must be followed by a relocation of type 772 // GENERIC_RELOC_PAIR. 773 unsigned RType = Obj->getAnyRelocationType(RENext); 774 if (RType != MachO::GENERIC_RELOC_PAIR) 775 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 776 "GENERIC_RELOC_LOCAL_SECTDIFF."); 777 778 printRelocationTargetName(Obj, RE, fmt); 779 fmt << "-"; 780 printRelocationTargetName(Obj, RENext, fmt); 781 break; 782 } 783 case MachO::GENERIC_RELOC_TLV: { 784 printRelocationTargetName(Obj, RE, fmt); 785 fmt << "@TLV"; 786 if (IsPCRel) 787 fmt << "P"; 788 break; 789 } 790 default: 791 printRelocationTargetName(Obj, RE, fmt); 792 } 793 } else { // ARM-specific relocations 794 switch (Type) { 795 case MachO::ARM_RELOC_HALF: 796 case MachO::ARM_RELOC_HALF_SECTDIFF: { 797 // Half relocations steal a bit from the length field to encode 798 // whether this is an upper16 or a lower16 relocation. 799 bool isUpper = (Obj->getAnyRelocationLength(RE) & 0x1) == 1; 800 801 if (isUpper) 802 fmt << ":upper16:("; 803 else 804 fmt << ":lower16:("; 805 printRelocationTargetName(Obj, RE, fmt); 806 807 DataRefImpl RelNext = Rel; 808 Obj->moveRelocationNext(RelNext); 809 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 810 811 // ARM half relocs must be followed by a relocation of type 812 // ARM_RELOC_PAIR. 813 unsigned RType = Obj->getAnyRelocationType(RENext); 814 if (RType != MachO::ARM_RELOC_PAIR) 815 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 816 "ARM_RELOC_HALF"); 817 818 // NOTE: The half of the target virtual address is stashed in the 819 // address field of the secondary relocation, but we can't reverse 820 // engineer the constant offset from it without decoding the movw/movt 821 // instruction to find the other half in its immediate field. 822 823 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 824 // symbol/section pointer of the follow-on relocation. 825 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 826 fmt << "-"; 827 printRelocationTargetName(Obj, RENext, fmt); 828 } 829 830 fmt << ")"; 831 break; 832 } 833 default: { printRelocationTargetName(Obj, RE, fmt); } 834 } 835 } 836 } else 837 printRelocationTargetName(Obj, RE, fmt); 838 839 fmt.flush(); 840 Result.append(fmtbuf.begin(), fmtbuf.end()); 841 return std::error_code(); 842 } 843 844 static std::error_code getRelocationValueString(const RelocationRef &Rel, 845 SmallVectorImpl<char> &Result) { 846 const ObjectFile *Obj = Rel.getObject(); 847 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 848 return getRelocationValueString(ELF, Rel, Result); 849 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 850 return getRelocationValueString(COFF, Rel, Result); 851 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 852 return getRelocationValueString(Wasm, Rel, Result); 853 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 854 return getRelocationValueString(MachO, Rel, Result); 855 llvm_unreachable("unknown object file format"); 856 } 857 858 /// Indicates whether this relocation should hidden when listing 859 /// relocations, usually because it is the trailing part of a multipart 860 /// relocation that will be printed as part of the leading relocation. 861 static bool getHidden(RelocationRef RelRef) { 862 const ObjectFile *Obj = RelRef.getObject(); 863 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 864 if (!MachO) 865 return false; 866 867 unsigned Arch = MachO->getArch(); 868 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 869 uint64_t Type = MachO->getRelocationType(Rel); 870 871 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 872 // is always hidden. 873 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 874 if (Type == MachO::GENERIC_RELOC_PAIR) 875 return true; 876 } else if (Arch == Triple::x86_64) { 877 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 878 // an X86_64_RELOC_SUBTRACTOR. 879 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 880 DataRefImpl RelPrev = Rel; 881 RelPrev.d.a--; 882 uint64_t PrevType = MachO->getRelocationType(RelPrev); 883 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 884 return true; 885 } 886 } 887 888 return false; 889 } 890 891 namespace { 892 class SourcePrinter { 893 protected: 894 DILineInfo OldLineInfo; 895 const ObjectFile *Obj = nullptr; 896 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 897 // File name to file contents of source 898 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 899 // Mark the line endings of the cached source 900 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 901 902 private: 903 bool cacheSource(const DILineInfo& LineInfoFile); 904 905 public: 906 SourcePrinter() = default; 907 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 908 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 909 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 910 DefaultArch); 911 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 912 } 913 virtual ~SourcePrinter() = default; 914 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 915 StringRef Delimiter = "; "); 916 }; 917 918 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 919 std::unique_ptr<MemoryBuffer> Buffer; 920 if (LineInfo.Source) { 921 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 922 } else { 923 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 924 if (!BufferOrError) 925 return false; 926 Buffer = std::move(*BufferOrError); 927 } 928 // Chomp the file to get lines 929 size_t BufferSize = Buffer->getBufferSize(); 930 const char *BufferStart = Buffer->getBufferStart(); 931 for (const char *Start = BufferStart, *End = BufferStart; 932 End < BufferStart + BufferSize; End++) 933 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 934 (*End == '\r' && *(End + 1) == '\n')) { 935 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 936 if (*End == '\r') 937 End++; 938 Start = End + 1; 939 } 940 SourceCache[LineInfo.FileName] = std::move(Buffer); 941 return true; 942 } 943 944 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 945 StringRef Delimiter) { 946 if (!Symbolizer) 947 return; 948 DILineInfo LineInfo = DILineInfo(); 949 auto ExpectecLineInfo = 950 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 951 if (!ExpectecLineInfo) 952 consumeError(ExpectecLineInfo.takeError()); 953 else 954 LineInfo = *ExpectecLineInfo; 955 956 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 957 LineInfo.Line == 0) 958 return; 959 960 if (PrintLines) 961 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 962 if (PrintSource) { 963 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 964 if (!cacheSource(LineInfo)) 965 return; 966 auto FileBuffer = SourceCache.find(LineInfo.FileName); 967 if (FileBuffer != SourceCache.end()) { 968 auto LineBuffer = LineCache.find(LineInfo.FileName); 969 if (LineBuffer != LineCache.end()) { 970 if (LineInfo.Line > LineBuffer->second.size()) 971 return; 972 // Vector begins at 0, line numbers are non-zero 973 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 974 << "\n"; 975 } 976 } 977 } 978 OldLineInfo = LineInfo; 979 } 980 981 static bool isArmElf(const ObjectFile *Obj) { 982 return (Obj->isELF() && 983 (Obj->getArch() == Triple::aarch64 || 984 Obj->getArch() == Triple::aarch64_be || 985 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 986 Obj->getArch() == Triple::thumb || 987 Obj->getArch() == Triple::thumbeb)); 988 } 989 990 class PrettyPrinter { 991 public: 992 virtual ~PrettyPrinter() = default; 993 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 994 ArrayRef<uint8_t> Bytes, uint64_t Address, 995 raw_ostream &OS, StringRef Annot, 996 MCSubtargetInfo const &STI, SourcePrinter *SP, 997 std::vector<RelocationRef> *Rels = nullptr) { 998 if (SP && (PrintSource || PrintLines)) 999 SP->printSourceLine(OS, Address); 1000 if (!NoLeadingAddr) 1001 OS << format("%8" PRIx64 ":", Address); 1002 if (!NoShowRawInsn) { 1003 OS << "\t"; 1004 dumpBytes(Bytes, OS); 1005 } 1006 if (MI) 1007 IP.printInst(MI, OS, "", STI); 1008 else 1009 OS << " <unknown>"; 1010 } 1011 }; 1012 PrettyPrinter PrettyPrinterInst; 1013 class HexagonPrettyPrinter : public PrettyPrinter { 1014 public: 1015 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 1016 raw_ostream &OS) { 1017 uint32_t opcode = 1018 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 1019 if (!NoLeadingAddr) 1020 OS << format("%8" PRIx64 ":", Address); 1021 if (!NoShowRawInsn) { 1022 OS << "\t"; 1023 dumpBytes(Bytes.slice(0, 4), OS); 1024 OS << format("%08" PRIx32, opcode); 1025 } 1026 } 1027 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1028 uint64_t Address, raw_ostream &OS, StringRef Annot, 1029 MCSubtargetInfo const &STI, SourcePrinter *SP, 1030 std::vector<RelocationRef> *Rels) override { 1031 if (SP && (PrintSource || PrintLines)) 1032 SP->printSourceLine(OS, Address, ""); 1033 if (!MI) { 1034 printLead(Bytes, Address, OS); 1035 OS << " <unknown>"; 1036 return; 1037 } 1038 std::string Buffer; 1039 { 1040 raw_string_ostream TempStream(Buffer); 1041 IP.printInst(MI, TempStream, "", STI); 1042 } 1043 StringRef Contents(Buffer); 1044 // Split off bundle attributes 1045 auto PacketBundle = Contents.rsplit('\n'); 1046 // Split off first instruction from the rest 1047 auto HeadTail = PacketBundle.first.split('\n'); 1048 auto Preamble = " { "; 1049 auto Separator = ""; 1050 StringRef Fmt = "\t\t\t%08" PRIx64 ": "; 1051 std::vector<RelocationRef>::const_iterator rel_cur = Rels->begin(); 1052 std::vector<RelocationRef>::const_iterator rel_end = Rels->end(); 1053 1054 // Hexagon's packets require relocations to be inline rather than 1055 // clustered at the end of the packet. 1056 auto PrintReloc = [&]() -> void { 1057 while ((rel_cur != rel_end) && (rel_cur->getOffset() <= Address)) { 1058 if (rel_cur->getOffset() == Address) { 1059 SmallString<16> name; 1060 SmallString<32> val; 1061 rel_cur->getTypeName(name); 1062 error(getRelocationValueString(*rel_cur, val)); 1063 OS << Separator << format(Fmt.data(), Address) << name << "\t" << val 1064 << "\n"; 1065 return; 1066 } 1067 rel_cur++; 1068 } 1069 }; 1070 1071 while(!HeadTail.first.empty()) { 1072 OS << Separator; 1073 Separator = "\n"; 1074 if (SP && (PrintSource || PrintLines)) 1075 SP->printSourceLine(OS, Address, ""); 1076 printLead(Bytes, Address, OS); 1077 OS << Preamble; 1078 Preamble = " "; 1079 StringRef Inst; 1080 auto Duplex = HeadTail.first.split('\v'); 1081 if(!Duplex.second.empty()){ 1082 OS << Duplex.first; 1083 OS << "; "; 1084 Inst = Duplex.second; 1085 } 1086 else 1087 Inst = HeadTail.first; 1088 OS << Inst; 1089 HeadTail = HeadTail.second.split('\n'); 1090 if (HeadTail.first.empty()) 1091 OS << " } " << PacketBundle.second; 1092 PrintReloc(); 1093 Bytes = Bytes.slice(4); 1094 Address += 4; 1095 } 1096 } 1097 }; 1098 HexagonPrettyPrinter HexagonPrettyPrinterInst; 1099 1100 class AMDGCNPrettyPrinter : public PrettyPrinter { 1101 public: 1102 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1103 uint64_t Address, raw_ostream &OS, StringRef Annot, 1104 MCSubtargetInfo const &STI, SourcePrinter *SP, 1105 std::vector<RelocationRef> *Rels) override { 1106 if (SP && (PrintSource || PrintLines)) 1107 SP->printSourceLine(OS, Address); 1108 1109 typedef support::ulittle32_t U32; 1110 1111 if (MI) { 1112 SmallString<40> InstStr; 1113 raw_svector_ostream IS(InstStr); 1114 1115 IP.printInst(MI, IS, "", STI); 1116 1117 OS << left_justify(IS.str(), 60); 1118 } else { 1119 // an unrecognized encoding - this is probably data so represent it 1120 // using the .long directive, or .byte directive if fewer than 4 bytes 1121 // remaining 1122 if (Bytes.size() >= 4) { 1123 OS << format("\t.long 0x%08" PRIx32 " ", 1124 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 1125 OS.indent(42); 1126 } else { 1127 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 1128 for (unsigned int i = 1; i < Bytes.size(); i++) 1129 OS << format(", 0x%02" PRIx8, Bytes[i]); 1130 OS.indent(55 - (6 * Bytes.size())); 1131 } 1132 } 1133 1134 OS << format("// %012" PRIX64 ": ", Address); 1135 if (Bytes.size() >=4) { 1136 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 1137 Bytes.size() / sizeof(U32))) 1138 // D should be explicitly casted to uint32_t here as it is passed 1139 // by format to snprintf as vararg. 1140 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 1141 } else { 1142 for (unsigned int i = 0; i < Bytes.size(); i++) 1143 OS << format("%02" PRIX8 " ", Bytes[i]); 1144 } 1145 1146 if (!Annot.empty()) 1147 OS << "// " << Annot; 1148 } 1149 }; 1150 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 1151 1152 class BPFPrettyPrinter : public PrettyPrinter { 1153 public: 1154 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1155 uint64_t Address, raw_ostream &OS, StringRef Annot, 1156 MCSubtargetInfo const &STI, SourcePrinter *SP, 1157 std::vector<RelocationRef> *Rels) override { 1158 if (SP && (PrintSource || PrintLines)) 1159 SP->printSourceLine(OS, Address); 1160 if (!NoLeadingAddr) 1161 OS << format("%8" PRId64 ":", Address / 8); 1162 if (!NoShowRawInsn) { 1163 OS << "\t"; 1164 dumpBytes(Bytes, OS); 1165 } 1166 if (MI) 1167 IP.printInst(MI, OS, "", STI); 1168 else 1169 OS << " <unknown>"; 1170 } 1171 }; 1172 BPFPrettyPrinter BPFPrettyPrinterInst; 1173 1174 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 1175 switch(Triple.getArch()) { 1176 default: 1177 return PrettyPrinterInst; 1178 case Triple::hexagon: 1179 return HexagonPrettyPrinterInst; 1180 case Triple::amdgcn: 1181 return AMDGCNPrettyPrinterInst; 1182 case Triple::bpfel: 1183 case Triple::bpfeb: 1184 return BPFPrettyPrinterInst; 1185 } 1186 } 1187 } 1188 1189 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1190 assert(Obj->isELF()); 1191 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1192 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1193 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1194 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1195 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1196 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1197 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1198 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1199 llvm_unreachable("Unsupported binary format"); 1200 } 1201 1202 template <class ELFT> static void 1203 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1204 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1205 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1206 uint8_t SymbolType = Symbol.getELFType(); 1207 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1208 continue; 1209 1210 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1211 if (!AddressOrErr) 1212 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1213 uint64_t Address = *AddressOrErr; 1214 1215 Expected<StringRef> Name = Symbol.getName(); 1216 if (!Name) 1217 report_error(Obj->getFileName(), Name.takeError()); 1218 if (Name->empty()) 1219 continue; 1220 1221 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1222 if (!SectionOrErr) 1223 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1224 section_iterator SecI = *SectionOrErr; 1225 if (SecI == Obj->section_end()) 1226 continue; 1227 1228 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1229 } 1230 } 1231 1232 static void 1233 addDynamicElfSymbols(const ObjectFile *Obj, 1234 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1235 assert(Obj->isELF()); 1236 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1237 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1238 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1239 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1240 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1241 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1242 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1243 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1244 else 1245 llvm_unreachable("Unsupported binary format"); 1246 } 1247 1248 static void addPltEntries(const ObjectFile *Obj, 1249 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 1250 StringSaver &Saver) { 1251 Optional<SectionRef> Plt = None; 1252 for (const SectionRef &Section : Obj->sections()) { 1253 StringRef Name; 1254 if (Section.getName(Name)) 1255 continue; 1256 if (Name == ".plt") 1257 Plt = Section; 1258 } 1259 if (!Plt) 1260 return; 1261 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1262 for (auto PltEntry : ElfObj->getPltAddresses()) { 1263 SymbolRef Symbol(PltEntry.first, ElfObj); 1264 1265 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1266 1267 Expected<StringRef> NameOrErr = Symbol.getName(); 1268 if (!NameOrErr) 1269 report_error(Obj->getFileName(), NameOrErr.takeError()); 1270 if (NameOrErr->empty()) 1271 continue; 1272 StringRef Name = Saver.save((*NameOrErr + "@plt").str()); 1273 1274 AllSymbols[*Plt].emplace_back(PltEntry.second, Name, SymbolType); 1275 } 1276 } 1277 } 1278 1279 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1280 if (StartAddress > StopAddress) 1281 error("Start address should be less than stop address"); 1282 1283 const Target *TheTarget = getTarget(Obj); 1284 1285 // Package up features to be passed to target/subtarget 1286 SubtargetFeatures Features = Obj->getFeatures(); 1287 if (MAttrs.size()) { 1288 for (unsigned i = 0; i != MAttrs.size(); ++i) 1289 Features.AddFeature(MAttrs[i]); 1290 } 1291 1292 std::unique_ptr<const MCRegisterInfo> MRI( 1293 TheTarget->createMCRegInfo(TripleName)); 1294 if (!MRI) 1295 report_error(Obj->getFileName(), "no register info for target " + 1296 TripleName); 1297 1298 // Set up disassembler. 1299 std::unique_ptr<const MCAsmInfo> AsmInfo( 1300 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1301 if (!AsmInfo) 1302 report_error(Obj->getFileName(), "no assembly info for target " + 1303 TripleName); 1304 std::unique_ptr<const MCSubtargetInfo> STI( 1305 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1306 if (!STI) 1307 report_error(Obj->getFileName(), "no subtarget info for target " + 1308 TripleName); 1309 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1310 if (!MII) 1311 report_error(Obj->getFileName(), "no instruction info for target " + 1312 TripleName); 1313 MCObjectFileInfo MOFI; 1314 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1315 // FIXME: for now initialize MCObjectFileInfo with default values 1316 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1317 1318 std::unique_ptr<MCDisassembler> DisAsm( 1319 TheTarget->createMCDisassembler(*STI, Ctx)); 1320 if (!DisAsm) 1321 report_error(Obj->getFileName(), "no disassembler for target " + 1322 TripleName); 1323 1324 std::unique_ptr<const MCInstrAnalysis> MIA( 1325 TheTarget->createMCInstrAnalysis(MII.get())); 1326 1327 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1328 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1329 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1330 if (!IP) 1331 report_error(Obj->getFileName(), "no instruction printer for target " + 1332 TripleName); 1333 IP->setPrintImmHex(PrintImmHex); 1334 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1335 1336 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1337 "\t\t\t%08" PRIx64 ": "; 1338 1339 SourcePrinter SP(Obj, TheTarget->getName()); 1340 1341 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1342 // in RelocSecs contain the relocations for section S. 1343 std::error_code EC; 1344 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1345 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1346 section_iterator Sec2 = Section.getRelocatedSection(); 1347 if (Sec2 != Obj->section_end()) 1348 SectionRelocMap[*Sec2].push_back(Section); 1349 } 1350 1351 // Create a mapping from virtual address to symbol name. This is used to 1352 // pretty print the symbols while disassembling. 1353 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1354 SectionSymbolsTy AbsoluteSymbols; 1355 for (const SymbolRef &Symbol : Obj->symbols()) { 1356 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1357 if (!AddressOrErr) 1358 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1359 uint64_t Address = *AddressOrErr; 1360 1361 Expected<StringRef> Name = Symbol.getName(); 1362 if (!Name) 1363 report_error(Obj->getFileName(), Name.takeError()); 1364 if (Name->empty()) 1365 continue; 1366 1367 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1368 if (!SectionOrErr) 1369 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1370 1371 uint8_t SymbolType = ELF::STT_NOTYPE; 1372 if (Obj->isELF()) 1373 SymbolType = getElfSymbolType(Obj, Symbol); 1374 1375 section_iterator SecI = *SectionOrErr; 1376 if (SecI != Obj->section_end()) 1377 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1378 else 1379 AbsoluteSymbols.emplace_back(Address, *Name, SymbolType); 1380 1381 1382 } 1383 if (AllSymbols.empty() && Obj->isELF()) 1384 addDynamicElfSymbols(Obj, AllSymbols); 1385 1386 BumpPtrAllocator A; 1387 StringSaver Saver(A); 1388 addPltEntries(Obj, AllSymbols, Saver); 1389 1390 // Create a mapping from virtual address to section. 1391 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1392 for (SectionRef Sec : Obj->sections()) 1393 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1394 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1395 1396 // Linked executables (.exe and .dll files) typically don't include a real 1397 // symbol table but they might contain an export table. 1398 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1399 for (const auto &ExportEntry : COFFObj->export_directories()) { 1400 StringRef Name; 1401 error(ExportEntry.getSymbolName(Name)); 1402 if (Name.empty()) 1403 continue; 1404 uint32_t RVA; 1405 error(ExportEntry.getExportRVA(RVA)); 1406 1407 uint64_t VA = COFFObj->getImageBase() + RVA; 1408 auto Sec = std::upper_bound( 1409 SectionAddresses.begin(), SectionAddresses.end(), VA, 1410 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1411 return LHS < RHS.first; 1412 }); 1413 if (Sec != SectionAddresses.begin()) 1414 --Sec; 1415 else 1416 Sec = SectionAddresses.end(); 1417 1418 if (Sec != SectionAddresses.end()) 1419 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1420 else 1421 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1422 } 1423 } 1424 1425 // Sort all the symbols, this allows us to use a simple binary search to find 1426 // a symbol near an address. 1427 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1428 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1429 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1430 1431 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1432 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1433 continue; 1434 1435 uint64_t SectionAddr = Section.getAddress(); 1436 uint64_t SectSize = Section.getSize(); 1437 if (!SectSize) 1438 continue; 1439 1440 // Get the list of all the symbols in this section. 1441 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1442 std::vector<uint64_t> DataMappingSymsAddr; 1443 std::vector<uint64_t> TextMappingSymsAddr; 1444 if (isArmElf(Obj)) { 1445 for (const auto &Symb : Symbols) { 1446 uint64_t Address = std::get<0>(Symb); 1447 StringRef Name = std::get<1>(Symb); 1448 if (Name.startswith("$d")) 1449 DataMappingSymsAddr.push_back(Address - SectionAddr); 1450 if (Name.startswith("$x")) 1451 TextMappingSymsAddr.push_back(Address - SectionAddr); 1452 if (Name.startswith("$a")) 1453 TextMappingSymsAddr.push_back(Address - SectionAddr); 1454 if (Name.startswith("$t")) 1455 TextMappingSymsAddr.push_back(Address - SectionAddr); 1456 } 1457 } 1458 1459 llvm::sort(DataMappingSymsAddr); 1460 llvm::sort(TextMappingSymsAddr); 1461 1462 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1463 // AMDGPU disassembler uses symbolizer for printing labels 1464 std::unique_ptr<MCRelocationInfo> RelInfo( 1465 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1466 if (RelInfo) { 1467 std::unique_ptr<MCSymbolizer> Symbolizer( 1468 TheTarget->createMCSymbolizer( 1469 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1470 DisAsm->setSymbolizer(std::move(Symbolizer)); 1471 } 1472 } 1473 1474 // Make a list of all the relocations for this section. 1475 std::vector<RelocationRef> Rels; 1476 if (InlineRelocs) { 1477 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1478 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1479 Rels.push_back(Reloc); 1480 } 1481 } 1482 } 1483 1484 // Sort relocations by address. 1485 llvm::sort(Rels, RelocAddressLess); 1486 1487 StringRef SegmentName = ""; 1488 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1489 DataRefImpl DR = Section.getRawDataRefImpl(); 1490 SegmentName = MachO->getSectionFinalSegmentName(DR); 1491 } 1492 StringRef SectionName; 1493 error(Section.getName(SectionName)); 1494 1495 // If the section has no symbol at the start, just insert a dummy one. 1496 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1497 Symbols.insert( 1498 Symbols.begin(), 1499 std::make_tuple(SectionAddr, SectionName, 1500 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1501 } 1502 1503 SmallString<40> Comments; 1504 raw_svector_ostream CommentStream(Comments); 1505 1506 StringRef BytesStr; 1507 error(Section.getContents(BytesStr)); 1508 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1509 BytesStr.size()); 1510 1511 uint64_t Size; 1512 uint64_t Index; 1513 bool PrintedSection = false; 1514 1515 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1516 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1517 // Disassemble symbol by symbol. 1518 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1519 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1520 // The end is either the section end or the beginning of the next 1521 // symbol. 1522 uint64_t End = 1523 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1524 // Don't try to disassemble beyond the end of section contents. 1525 if (End > SectSize) 1526 End = SectSize; 1527 // If this symbol has the same address as the next symbol, then skip it. 1528 if (Start >= End) 1529 continue; 1530 1531 // Check if we need to skip symbol 1532 // Skip if the symbol's data is not between StartAddress and StopAddress 1533 if (End + SectionAddr < StartAddress || 1534 Start + SectionAddr > StopAddress) { 1535 continue; 1536 } 1537 1538 /// Skip if user requested specific symbols and this is not in the list 1539 if (!DisasmFuncsSet.empty() && 1540 !DisasmFuncsSet.count(std::get<1>(Symbols[si]))) 1541 continue; 1542 1543 if (!PrintedSection) { 1544 PrintedSection = true; 1545 outs() << "Disassembly of section "; 1546 if (!SegmentName.empty()) 1547 outs() << SegmentName << ","; 1548 outs() << SectionName << ':'; 1549 } 1550 1551 // Stop disassembly at the stop address specified 1552 if (End + SectionAddr > StopAddress) 1553 End = StopAddress - SectionAddr; 1554 1555 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1556 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1557 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1558 Start += 256; 1559 } 1560 if (si == se - 1 || 1561 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1562 // cut trailing zeroes at the end of kernel 1563 // cut up to 256 bytes 1564 const uint64_t EndAlign = 256; 1565 const auto Limit = End - (std::min)(EndAlign, End - Start); 1566 while (End > Limit && 1567 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1568 End -= 4; 1569 } 1570 } 1571 1572 auto PrintSymbol = [](StringRef Name) { 1573 outs() << '\n' << Name << ":\n"; 1574 }; 1575 StringRef SymbolName = std::get<1>(Symbols[si]); 1576 if (Demangle) { 1577 char *DemangledSymbol = nullptr; 1578 size_t Size = 0; 1579 int Status = -1; 1580 if (SymbolName.startswith("_Z")) 1581 DemangledSymbol = itaniumDemangle(SymbolName.data(), DemangledSymbol, 1582 &Size, &Status); 1583 else if (SymbolName.startswith("?")) 1584 DemangledSymbol = microsoftDemangle(SymbolName.data(), 1585 DemangledSymbol, &Size, &Status); 1586 1587 if (Status == 0 && DemangledSymbol) 1588 PrintSymbol(StringRef(DemangledSymbol)); 1589 else 1590 PrintSymbol(SymbolName); 1591 1592 if (DemangledSymbol) 1593 free(DemangledSymbol); 1594 } else 1595 PrintSymbol(SymbolName); 1596 1597 // Don't print raw contents of a virtual section. A virtual section 1598 // doesn't have any contents in the file. 1599 if (Section.isVirtual()) { 1600 outs() << "...\n"; 1601 continue; 1602 } 1603 1604 #ifndef NDEBUG 1605 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1606 #else 1607 raw_ostream &DebugOut = nulls(); 1608 #endif 1609 1610 for (Index = Start; Index < End; Index += Size) { 1611 MCInst Inst; 1612 1613 if (Index + SectionAddr < StartAddress || 1614 Index + SectionAddr > StopAddress) { 1615 // skip byte by byte till StartAddress is reached 1616 Size = 1; 1617 continue; 1618 } 1619 // AArch64 ELF binaries can interleave data and text in the 1620 // same section. We rely on the markers introduced to 1621 // understand what we need to dump. If the data marker is within a 1622 // function, it is denoted as a word/short etc 1623 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1624 !DisassembleAll) { 1625 uint64_t Stride = 0; 1626 1627 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1628 DataMappingSymsAddr.end(), Index); 1629 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1630 // Switch to data. 1631 while (Index < End) { 1632 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1633 outs() << "\t"; 1634 if (Index + 4 <= End) { 1635 Stride = 4; 1636 dumpBytes(Bytes.slice(Index, 4), outs()); 1637 outs() << "\t.word\t"; 1638 uint32_t Data = 0; 1639 if (Obj->isLittleEndian()) { 1640 const auto Word = 1641 reinterpret_cast<const support::ulittle32_t *>( 1642 Bytes.data() + Index); 1643 Data = *Word; 1644 } else { 1645 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1646 Bytes.data() + Index); 1647 Data = *Word; 1648 } 1649 outs() << "0x" << format("%08" PRIx32, Data); 1650 } else if (Index + 2 <= End) { 1651 Stride = 2; 1652 dumpBytes(Bytes.slice(Index, 2), outs()); 1653 outs() << "\t\t.short\t"; 1654 uint16_t Data = 0; 1655 if (Obj->isLittleEndian()) { 1656 const auto Short = 1657 reinterpret_cast<const support::ulittle16_t *>( 1658 Bytes.data() + Index); 1659 Data = *Short; 1660 } else { 1661 const auto Short = 1662 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1663 Index); 1664 Data = *Short; 1665 } 1666 outs() << "0x" << format("%04" PRIx16, Data); 1667 } else { 1668 Stride = 1; 1669 dumpBytes(Bytes.slice(Index, 1), outs()); 1670 outs() << "\t\t.byte\t"; 1671 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1672 } 1673 Index += Stride; 1674 outs() << "\n"; 1675 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1676 TextMappingSymsAddr.end(), Index); 1677 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1678 break; 1679 } 1680 } 1681 } 1682 1683 // If there is a data symbol inside an ELF text section and we are only 1684 // disassembling text (applicable all architectures), 1685 // we are in a situation where we must print the data and not 1686 // disassemble it. 1687 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1688 !DisassembleAll && Section.isText()) { 1689 // print out data up to 8 bytes at a time in hex and ascii 1690 uint8_t AsciiData[9] = {'\0'}; 1691 uint8_t Byte; 1692 int NumBytes = 0; 1693 1694 for (Index = Start; Index < End; Index += 1) { 1695 if (((SectionAddr + Index) < StartAddress) || 1696 ((SectionAddr + Index) > StopAddress)) 1697 continue; 1698 if (NumBytes == 0) { 1699 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1700 outs() << "\t"; 1701 } 1702 Byte = Bytes.slice(Index)[0]; 1703 outs() << format(" %02x", Byte); 1704 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1705 1706 uint8_t IndentOffset = 0; 1707 NumBytes++; 1708 if (Index == End - 1 || NumBytes > 8) { 1709 // Indent the space for less than 8 bytes data. 1710 // 2 spaces for byte and one for space between bytes 1711 IndentOffset = 3 * (8 - NumBytes); 1712 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1713 AsciiData[Excess] = '\0'; 1714 NumBytes = 8; 1715 } 1716 if (NumBytes == 8) { 1717 AsciiData[8] = '\0'; 1718 outs() << std::string(IndentOffset, ' ') << " "; 1719 outs() << reinterpret_cast<char *>(AsciiData); 1720 outs() << '\n'; 1721 NumBytes = 0; 1722 } 1723 } 1724 } 1725 if (Index >= End) 1726 break; 1727 1728 // Disassemble a real instruction or a data when disassemble all is 1729 // provided 1730 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1731 SectionAddr + Index, DebugOut, 1732 CommentStream); 1733 if (Size == 0) 1734 Size = 1; 1735 1736 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1737 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1738 *STI, &SP, &Rels); 1739 outs() << CommentStream.str(); 1740 Comments.clear(); 1741 1742 // Try to resolve the target of a call, tail call, etc. to a specific 1743 // symbol. 1744 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1745 MIA->isConditionalBranch(Inst))) { 1746 uint64_t Target; 1747 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1748 // In a relocatable object, the target's section must reside in 1749 // the same section as the call instruction or it is accessed 1750 // through a relocation. 1751 // 1752 // In a non-relocatable object, the target may be in any section. 1753 // 1754 // N.B. We don't walk the relocations in the relocatable case yet. 1755 auto *TargetSectionSymbols = &Symbols; 1756 if (!Obj->isRelocatableObject()) { 1757 auto SectionAddress = std::upper_bound( 1758 SectionAddresses.begin(), SectionAddresses.end(), Target, 1759 [](uint64_t LHS, 1760 const std::pair<uint64_t, SectionRef> &RHS) { 1761 return LHS < RHS.first; 1762 }); 1763 if (SectionAddress != SectionAddresses.begin()) { 1764 --SectionAddress; 1765 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1766 } else { 1767 TargetSectionSymbols = &AbsoluteSymbols; 1768 } 1769 } 1770 1771 // Find the first symbol in the section whose offset is less than 1772 // or equal to the target. If there isn't a section that contains 1773 // the target, find the nearest preceding absolute symbol. 1774 auto TargetSym = std::upper_bound( 1775 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1776 Target, [](uint64_t LHS, 1777 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1778 return LHS < std::get<0>(RHS); 1779 }); 1780 if (TargetSym == TargetSectionSymbols->begin()) { 1781 TargetSectionSymbols = &AbsoluteSymbols; 1782 TargetSym = std::upper_bound( 1783 AbsoluteSymbols.begin(), AbsoluteSymbols.end(), 1784 Target, [](uint64_t LHS, 1785 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1786 return LHS < std::get<0>(RHS); 1787 }); 1788 } 1789 if (TargetSym != TargetSectionSymbols->begin()) { 1790 --TargetSym; 1791 uint64_t TargetAddress = std::get<0>(*TargetSym); 1792 StringRef TargetName = std::get<1>(*TargetSym); 1793 outs() << " <" << TargetName; 1794 uint64_t Disp = Target - TargetAddress; 1795 if (Disp) 1796 outs() << "+0x" << Twine::utohexstr(Disp); 1797 outs() << '>'; 1798 } 1799 } 1800 } 1801 outs() << "\n"; 1802 1803 // Hexagon does this in pretty printer 1804 if (Obj->getArch() != Triple::hexagon) 1805 // Print relocation for instruction. 1806 while (rel_cur != rel_end) { 1807 bool hidden = getHidden(*rel_cur); 1808 uint64_t addr = rel_cur->getOffset(); 1809 SmallString<16> name; 1810 SmallString<32> val; 1811 1812 // If this relocation is hidden, skip it. 1813 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1814 ++rel_cur; 1815 continue; 1816 } 1817 1818 // Stop when rel_cur's address is past the current instruction. 1819 if (addr >= Index + Size) break; 1820 rel_cur->getTypeName(name); 1821 error(getRelocationValueString(*rel_cur, val)); 1822 outs() << format(Fmt.data(), SectionAddr + addr) << name 1823 << "\t" << val << "\n"; 1824 ++rel_cur; 1825 } 1826 } 1827 } 1828 } 1829 } 1830 1831 void llvm::PrintRelocations(const ObjectFile *Obj) { 1832 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1833 "%08" PRIx64; 1834 // Regular objdump doesn't print relocations in non-relocatable object 1835 // files. 1836 if (!Obj->isRelocatableObject()) 1837 return; 1838 1839 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1840 if (Section.relocation_begin() == Section.relocation_end()) 1841 continue; 1842 StringRef secname; 1843 error(Section.getName(secname)); 1844 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1845 for (const RelocationRef &Reloc : Section.relocations()) { 1846 bool hidden = getHidden(Reloc); 1847 uint64_t address = Reloc.getOffset(); 1848 SmallString<32> relocname; 1849 SmallString<32> valuestr; 1850 if (address < StartAddress || address > StopAddress || hidden) 1851 continue; 1852 Reloc.getTypeName(relocname); 1853 error(getRelocationValueString(Reloc, valuestr)); 1854 outs() << format(Fmt.data(), address) << " " << relocname << " " 1855 << valuestr << "\n"; 1856 } 1857 outs() << "\n"; 1858 } 1859 } 1860 1861 void llvm::PrintDynamicRelocations(const ObjectFile *Obj) { 1862 1863 // For the moment, this option is for ELF only 1864 if (!Obj->isELF()) 1865 return; 1866 1867 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1868 1869 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1870 error("not a dynamic object"); 1871 return; 1872 } 1873 1874 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1875 1876 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1877 if (DynRelSec.empty()) 1878 return; 1879 1880 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1881 for (const SectionRef &Section : DynRelSec) { 1882 if (Section.relocation_begin() == Section.relocation_end()) 1883 continue; 1884 for (const RelocationRef &Reloc : Section.relocations()) { 1885 uint64_t address = Reloc.getOffset(); 1886 SmallString<32> relocname; 1887 SmallString<32> valuestr; 1888 Reloc.getTypeName(relocname); 1889 error(getRelocationValueString(Reloc, valuestr)); 1890 outs() << format(Fmt.data(), address) << " " << relocname << " " 1891 << valuestr << "\n"; 1892 } 1893 } 1894 } 1895 1896 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1897 outs() << "Sections:\n" 1898 "Idx Name Size Address Type\n"; 1899 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1900 StringRef Name; 1901 error(Section.getName(Name)); 1902 uint64_t Address = Section.getAddress(); 1903 uint64_t Size = Section.getSize(); 1904 bool Text = Section.isText(); 1905 bool Data = Section.isData(); 1906 bool BSS = Section.isBSS(); 1907 std::string Type = (std::string(Text ? "TEXT " : "") + 1908 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1909 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1910 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1911 Address, Type.c_str()); 1912 } 1913 } 1914 1915 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1916 std::error_code EC; 1917 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1918 StringRef Name; 1919 StringRef Contents; 1920 error(Section.getName(Name)); 1921 uint64_t BaseAddr = Section.getAddress(); 1922 uint64_t Size = Section.getSize(); 1923 if (!Size) 1924 continue; 1925 1926 outs() << "Contents of section " << Name << ":\n"; 1927 if (Section.isBSS()) { 1928 outs() << format("<skipping contents of bss section at [%04" PRIx64 1929 ", %04" PRIx64 ")>\n", 1930 BaseAddr, BaseAddr + Size); 1931 continue; 1932 } 1933 1934 error(Section.getContents(Contents)); 1935 1936 // Dump out the content as hex and printable ascii characters. 1937 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1938 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1939 // Dump line of hex. 1940 for (std::size_t i = 0; i < 16; ++i) { 1941 if (i != 0 && i % 4 == 0) 1942 outs() << ' '; 1943 if (addr + i < end) 1944 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1945 << hexdigit(Contents[addr + i] & 0xF, true); 1946 else 1947 outs() << " "; 1948 } 1949 // Print ascii. 1950 outs() << " "; 1951 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1952 if (isPrint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1953 outs() << Contents[addr + i]; 1954 else 1955 outs() << "."; 1956 } 1957 outs() << "\n"; 1958 } 1959 } 1960 } 1961 1962 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1963 StringRef ArchitectureName) { 1964 outs() << "SYMBOL TABLE:\n"; 1965 1966 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1967 printCOFFSymbolTable(coff); 1968 return; 1969 } 1970 for (const SymbolRef &Symbol : o->symbols()) { 1971 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1972 if (!AddressOrError) 1973 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1974 ArchitectureName); 1975 uint64_t Address = *AddressOrError; 1976 if ((Address < StartAddress) || (Address > StopAddress)) 1977 continue; 1978 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1979 if (!TypeOrError) 1980 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1981 ArchitectureName); 1982 SymbolRef::Type Type = *TypeOrError; 1983 uint32_t Flags = Symbol.getFlags(); 1984 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1985 if (!SectionOrErr) 1986 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1987 ArchitectureName); 1988 section_iterator Section = *SectionOrErr; 1989 StringRef Name; 1990 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1991 Section->getName(Name); 1992 } else { 1993 Expected<StringRef> NameOrErr = Symbol.getName(); 1994 if (!NameOrErr) 1995 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1996 ArchitectureName); 1997 Name = *NameOrErr; 1998 } 1999 2000 bool Global = Flags & SymbolRef::SF_Global; 2001 bool Weak = Flags & SymbolRef::SF_Weak; 2002 bool Absolute = Flags & SymbolRef::SF_Absolute; 2003 bool Common = Flags & SymbolRef::SF_Common; 2004 bool Hidden = Flags & SymbolRef::SF_Hidden; 2005 2006 char GlobLoc = ' '; 2007 if (Type != SymbolRef::ST_Unknown) 2008 GlobLoc = Global ? 'g' : 'l'; 2009 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2010 ? 'd' : ' '; 2011 char FileFunc = ' '; 2012 if (Type == SymbolRef::ST_File) 2013 FileFunc = 'f'; 2014 else if (Type == SymbolRef::ST_Function) 2015 FileFunc = 'F'; 2016 2017 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 2018 "%08" PRIx64; 2019 2020 outs() << format(Fmt, Address) << " " 2021 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2022 << (Weak ? 'w' : ' ') // Weak? 2023 << ' ' // Constructor. Not supported yet. 2024 << ' ' // Warning. Not supported yet. 2025 << ' ' // Indirect reference to another symbol. 2026 << Debug // Debugging (d) or dynamic (D) symbol. 2027 << FileFunc // Name of function (F), file (f) or object (O). 2028 << ' '; 2029 if (Absolute) { 2030 outs() << "*ABS*"; 2031 } else if (Common) { 2032 outs() << "*COM*"; 2033 } else if (Section == o->section_end()) { 2034 outs() << "*UND*"; 2035 } else { 2036 if (const MachOObjectFile *MachO = 2037 dyn_cast<const MachOObjectFile>(o)) { 2038 DataRefImpl DR = Section->getRawDataRefImpl(); 2039 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 2040 outs() << SegmentName << ","; 2041 } 2042 StringRef SectionName; 2043 error(Section->getName(SectionName)); 2044 outs() << SectionName; 2045 } 2046 2047 outs() << '\t'; 2048 if (Common || isa<ELFObjectFileBase>(o)) { 2049 uint64_t Val = 2050 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2051 outs() << format("\t %08" PRIx64 " ", Val); 2052 } 2053 2054 if (Hidden) { 2055 outs() << ".hidden "; 2056 } 2057 outs() << Name 2058 << '\n'; 2059 } 2060 } 2061 2062 static void PrintUnwindInfo(const ObjectFile *o) { 2063 outs() << "Unwind info:\n\n"; 2064 2065 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 2066 printCOFFUnwindInfo(coff); 2067 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2068 printMachOUnwindInfo(MachO); 2069 else { 2070 // TODO: Extract DWARF dump tool to objdump. 2071 errs() << "This operation is only currently supported " 2072 "for COFF and MachO object files.\n"; 2073 return; 2074 } 2075 } 2076 2077 void llvm::printExportsTrie(const ObjectFile *o) { 2078 outs() << "Exports trie:\n"; 2079 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2080 printMachOExportsTrie(MachO); 2081 else { 2082 errs() << "This operation is only currently supported " 2083 "for Mach-O executable files.\n"; 2084 return; 2085 } 2086 } 2087 2088 void llvm::printRebaseTable(ObjectFile *o) { 2089 outs() << "Rebase table:\n"; 2090 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2091 printMachORebaseTable(MachO); 2092 else { 2093 errs() << "This operation is only currently supported " 2094 "for Mach-O executable files.\n"; 2095 return; 2096 } 2097 } 2098 2099 void llvm::printBindTable(ObjectFile *o) { 2100 outs() << "Bind table:\n"; 2101 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2102 printMachOBindTable(MachO); 2103 else { 2104 errs() << "This operation is only currently supported " 2105 "for Mach-O executable files.\n"; 2106 return; 2107 } 2108 } 2109 2110 void llvm::printLazyBindTable(ObjectFile *o) { 2111 outs() << "Lazy bind table:\n"; 2112 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2113 printMachOLazyBindTable(MachO); 2114 else { 2115 errs() << "This operation is only currently supported " 2116 "for Mach-O executable files.\n"; 2117 return; 2118 } 2119 } 2120 2121 void llvm::printWeakBindTable(ObjectFile *o) { 2122 outs() << "Weak bind table:\n"; 2123 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2124 printMachOWeakBindTable(MachO); 2125 else { 2126 errs() << "This operation is only currently supported " 2127 "for Mach-O executable files.\n"; 2128 return; 2129 } 2130 } 2131 2132 /// Dump the raw contents of the __clangast section so the output can be piped 2133 /// into llvm-bcanalyzer. 2134 void llvm::printRawClangAST(const ObjectFile *Obj) { 2135 if (outs().is_displayed()) { 2136 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 2137 "the clang ast section.\n" 2138 "Please redirect the output to a file or another program such as " 2139 "llvm-bcanalyzer.\n"; 2140 return; 2141 } 2142 2143 StringRef ClangASTSectionName("__clangast"); 2144 if (isa<COFFObjectFile>(Obj)) { 2145 ClangASTSectionName = "clangast"; 2146 } 2147 2148 Optional<object::SectionRef> ClangASTSection; 2149 for (auto Sec : ToolSectionFilter(*Obj)) { 2150 StringRef Name; 2151 Sec.getName(Name); 2152 if (Name == ClangASTSectionName) { 2153 ClangASTSection = Sec; 2154 break; 2155 } 2156 } 2157 if (!ClangASTSection) 2158 return; 2159 2160 StringRef ClangASTContents; 2161 error(ClangASTSection.getValue().getContents(ClangASTContents)); 2162 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2163 } 2164 2165 static void printFaultMaps(const ObjectFile *Obj) { 2166 const char *FaultMapSectionName = nullptr; 2167 2168 if (isa<ELFObjectFileBase>(Obj)) { 2169 FaultMapSectionName = ".llvm_faultmaps"; 2170 } else if (isa<MachOObjectFile>(Obj)) { 2171 FaultMapSectionName = "__llvm_faultmaps"; 2172 } else { 2173 errs() << "This operation is only currently supported " 2174 "for ELF and Mach-O executable files.\n"; 2175 return; 2176 } 2177 2178 Optional<object::SectionRef> FaultMapSection; 2179 2180 for (auto Sec : ToolSectionFilter(*Obj)) { 2181 StringRef Name; 2182 Sec.getName(Name); 2183 if (Name == FaultMapSectionName) { 2184 FaultMapSection = Sec; 2185 break; 2186 } 2187 } 2188 2189 outs() << "FaultMap table:\n"; 2190 2191 if (!FaultMapSection.hasValue()) { 2192 outs() << "<not found>\n"; 2193 return; 2194 } 2195 2196 StringRef FaultMapContents; 2197 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2198 2199 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2200 FaultMapContents.bytes_end()); 2201 2202 outs() << FMP; 2203 } 2204 2205 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2206 if (o->isELF()) { 2207 printELFFileHeader(o); 2208 return printELFDynamicSection(o); 2209 } 2210 if (o->isCOFF()) 2211 return printCOFFFileHeader(o); 2212 if (o->isWasm()) 2213 return printWasmFileHeader(o); 2214 if (o->isMachO()) { 2215 printMachOFileHeader(o); 2216 if (!onlyFirst) 2217 printMachOLoadCommands(o); 2218 return; 2219 } 2220 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2221 } 2222 2223 static void printFileHeaders(const ObjectFile *o) { 2224 if (!o->isELF() && !o->isCOFF()) 2225 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2226 2227 Triple::ArchType AT = o->getArch(); 2228 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2229 Expected<uint64_t> StartAddrOrErr = o->getStartAddress(); 2230 if (!StartAddrOrErr) 2231 report_error(o->getFileName(), StartAddrOrErr.takeError()); 2232 2233 StringRef Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2234 uint64_t Address = StartAddrOrErr.get(); 2235 outs() << "start address: " 2236 << "0x" << format(Fmt.data(), Address) 2237 << "\n"; 2238 } 2239 2240 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2241 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2242 if (!ModeOrErr) { 2243 errs() << "ill-formed archive entry.\n"; 2244 consumeError(ModeOrErr.takeError()); 2245 return; 2246 } 2247 sys::fs::perms Mode = ModeOrErr.get(); 2248 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2249 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2250 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2251 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2252 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2253 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2254 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2255 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2256 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2257 2258 outs() << " "; 2259 2260 Expected<unsigned> UIDOrErr = C.getUID(); 2261 if (!UIDOrErr) 2262 report_error(Filename, UIDOrErr.takeError()); 2263 unsigned UID = UIDOrErr.get(); 2264 outs() << format("%d/", UID); 2265 2266 Expected<unsigned> GIDOrErr = C.getGID(); 2267 if (!GIDOrErr) 2268 report_error(Filename, GIDOrErr.takeError()); 2269 unsigned GID = GIDOrErr.get(); 2270 outs() << format("%-d ", GID); 2271 2272 Expected<uint64_t> Size = C.getRawSize(); 2273 if (!Size) 2274 report_error(Filename, Size.takeError()); 2275 outs() << format("%6" PRId64, Size.get()) << " "; 2276 2277 StringRef RawLastModified = C.getRawLastModified(); 2278 unsigned Seconds; 2279 if (RawLastModified.getAsInteger(10, Seconds)) 2280 outs() << "(date: \"" << RawLastModified 2281 << "\" contains non-decimal chars) "; 2282 else { 2283 // Since ctime(3) returns a 26 character string of the form: 2284 // "Sun Sep 16 01:03:52 1973\n\0" 2285 // just print 24 characters. 2286 time_t t = Seconds; 2287 outs() << format("%.24s ", ctime(&t)); 2288 } 2289 2290 StringRef Name = ""; 2291 Expected<StringRef> NameOrErr = C.getName(); 2292 if (!NameOrErr) { 2293 consumeError(NameOrErr.takeError()); 2294 Expected<StringRef> RawNameOrErr = C.getRawName(); 2295 if (!RawNameOrErr) 2296 report_error(Filename, NameOrErr.takeError()); 2297 Name = RawNameOrErr.get(); 2298 } else { 2299 Name = NameOrErr.get(); 2300 } 2301 outs() << Name << "\n"; 2302 } 2303 2304 static void DumpObject(ObjectFile *o, const Archive *a = nullptr, 2305 const Archive::Child *c = nullptr) { 2306 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2307 // Avoid other output when using a raw option. 2308 if (!RawClangAST) { 2309 outs() << '\n'; 2310 if (a) 2311 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2312 else 2313 outs() << o->getFileName(); 2314 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2315 } 2316 2317 if (ArchiveHeaders && !MachOOpt && c) 2318 printArchiveChild(ArchiveName, *c); 2319 if (Disassemble) 2320 DisassembleObject(o, Relocations); 2321 if (Relocations && !Disassemble) 2322 PrintRelocations(o); 2323 if (DynamicRelocations) 2324 PrintDynamicRelocations(o); 2325 if (SectionHeaders) 2326 PrintSectionHeaders(o); 2327 if (SectionContents) 2328 PrintSectionContents(o); 2329 if (SymbolTable) 2330 PrintSymbolTable(o, ArchiveName); 2331 if (UnwindInfo) 2332 PrintUnwindInfo(o); 2333 if (PrivateHeaders || FirstPrivateHeader) 2334 printPrivateFileHeaders(o, FirstPrivateHeader); 2335 if (FileHeaders) 2336 printFileHeaders(o); 2337 if (ExportsTrie) 2338 printExportsTrie(o); 2339 if (Rebase) 2340 printRebaseTable(o); 2341 if (Bind) 2342 printBindTable(o); 2343 if (LazyBind) 2344 printLazyBindTable(o); 2345 if (WeakBind) 2346 printWeakBindTable(o); 2347 if (RawClangAST) 2348 printRawClangAST(o); 2349 if (PrintFaultMaps) 2350 printFaultMaps(o); 2351 if (DwarfDumpType != DIDT_Null) { 2352 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*o); 2353 // Dump the complete DWARF structure. 2354 DIDumpOptions DumpOpts; 2355 DumpOpts.DumpType = DwarfDumpType; 2356 DICtx->dump(outs(), DumpOpts); 2357 } 2358 } 2359 2360 static void DumpObject(const COFFImportFile *I, const Archive *A, 2361 const Archive::Child *C = nullptr) { 2362 StringRef ArchiveName = A ? A->getFileName() : ""; 2363 2364 // Avoid other output when using a raw option. 2365 if (!RawClangAST) 2366 outs() << '\n' 2367 << ArchiveName << "(" << I->getFileName() << ")" 2368 << ":\tfile format COFF-import-file" 2369 << "\n\n"; 2370 2371 if (ArchiveHeaders && !MachOOpt && C) 2372 printArchiveChild(ArchiveName, *C); 2373 if (SymbolTable) 2374 printCOFFSymbolTable(I); 2375 } 2376 2377 /// Dump each object file in \a a; 2378 static void DumpArchive(const Archive *a) { 2379 Error Err = Error::success(); 2380 for (auto &C : a->children(Err)) { 2381 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2382 if (!ChildOrErr) { 2383 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2384 report_error(a->getFileName(), C, std::move(E)); 2385 continue; 2386 } 2387 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2388 DumpObject(o, a, &C); 2389 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2390 DumpObject(I, a, &C); 2391 else 2392 report_error(a->getFileName(), object_error::invalid_file_type); 2393 } 2394 if (Err) 2395 report_error(a->getFileName(), std::move(Err)); 2396 } 2397 2398 /// Open file and figure out how to dump it. 2399 static void DumpInput(StringRef file) { 2400 2401 // If we are using the Mach-O specific object file parser, then let it parse 2402 // the file and process the command line options. So the -arch flags can 2403 // be used to select specific slices, etc. 2404 if (MachOOpt) { 2405 ParseInputMachO(file); 2406 return; 2407 } 2408 2409 // Attempt to open the binary. 2410 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2411 if (!BinaryOrErr) 2412 report_error(file, BinaryOrErr.takeError()); 2413 Binary &Binary = *BinaryOrErr.get().getBinary(); 2414 2415 if (Archive *a = dyn_cast<Archive>(&Binary)) 2416 DumpArchive(a); 2417 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2418 DumpObject(o); 2419 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2420 ParseInputMachO(UB); 2421 else 2422 report_error(file, object_error::invalid_file_type); 2423 } 2424 2425 int main(int argc, char **argv) { 2426 InitLLVM X(argc, argv); 2427 2428 // Initialize targets and assembly printers/parsers. 2429 llvm::InitializeAllTargetInfos(); 2430 llvm::InitializeAllTargetMCs(); 2431 llvm::InitializeAllDisassemblers(); 2432 2433 // Register the target printer for --version. 2434 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2435 2436 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2437 2438 ToolName = argv[0]; 2439 2440 // Defaults to a.out if no filenames specified. 2441 if (InputFilenames.size() == 0) 2442 InputFilenames.push_back("a.out"); 2443 2444 if (AllHeaders) 2445 PrivateHeaders = Relocations = SectionHeaders = SymbolTable = true; 2446 2447 if (DisassembleAll || PrintSource || PrintLines) 2448 Disassemble = true; 2449 2450 if (!Disassemble 2451 && !Relocations 2452 && !DynamicRelocations 2453 && !SectionHeaders 2454 && !SectionContents 2455 && !SymbolTable 2456 && !UnwindInfo 2457 && !PrivateHeaders 2458 && !FileHeaders 2459 && !FirstPrivateHeader 2460 && !ExportsTrie 2461 && !Rebase 2462 && !Bind 2463 && !LazyBind 2464 && !WeakBind 2465 && !RawClangAST 2466 && !(UniversalHeaders && MachOOpt) 2467 && !ArchiveHeaders 2468 && !(IndirectSymbols && MachOOpt) 2469 && !(DataInCode && MachOOpt) 2470 && !(LinkOptHints && MachOOpt) 2471 && !(InfoPlist && MachOOpt) 2472 && !(DylibsUsed && MachOOpt) 2473 && !(DylibId && MachOOpt) 2474 && !(ObjcMetaData && MachOOpt) 2475 && !(FilterSections.size() != 0 && MachOOpt) 2476 && !PrintFaultMaps 2477 && DwarfDumpType == DIDT_Null) { 2478 cl::PrintHelpMessage(); 2479 return 2; 2480 } 2481 2482 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2483 DisassembleFunctions.end()); 2484 2485 llvm::for_each(InputFilenames, DumpInput); 2486 2487 return EXIT_SUCCESS; 2488 } 2489