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::NotHidden, 132 cl::aliasopt(llvm::SymbolTable)); 133 134 cl::opt<bool> 135 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 136 137 cl::opt<bool> 138 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 139 140 cl::opt<bool> 141 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 142 143 cl::opt<bool> 144 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 145 146 cl::opt<bool> 147 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 148 149 cl::opt<bool> 150 llvm::RawClangAST("raw-clang-ast", 151 cl::desc("Dump the raw binary contents of the clang AST section")); 152 153 static cl::opt<bool> 154 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 155 static cl::alias 156 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 157 158 cl::opt<std::string> 159 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 160 "see -version for available targets")); 161 162 cl::opt<std::string> 163 llvm::MCPU("mcpu", 164 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 165 cl::value_desc("cpu-name"), 166 cl::init("")); 167 168 cl::opt<std::string> 169 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 170 "see -version for available targets")); 171 172 cl::opt<bool> 173 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 174 "headers for each section.")); 175 static cl::alias 176 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 177 cl::aliasopt(SectionHeaders)); 178 static cl::alias 179 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 180 cl::aliasopt(SectionHeaders)); 181 182 cl::list<std::string> 183 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 184 "With -macho dump segment,section")); 185 cl::alias 186 static FilterSectionsj("j", cl::desc("Alias for --section"), 187 cl::aliasopt(llvm::FilterSections)); 188 189 cl::list<std::string> 190 llvm::MAttrs("mattr", 191 cl::CommaSeparated, 192 cl::desc("Target specific attributes"), 193 cl::value_desc("a1,+a2,-a3,...")); 194 195 cl::opt<bool> 196 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 197 "instructions, do not print " 198 "the instruction bytes.")); 199 cl::opt<bool> 200 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 201 202 cl::opt<bool> 203 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 204 205 static cl::alias 206 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 207 cl::aliasopt(UnwindInfo)); 208 209 cl::opt<bool> 210 llvm::PrivateHeaders("private-headers", 211 cl::desc("Display format specific file headers")); 212 213 cl::opt<bool> 214 llvm::FirstPrivateHeader("private-header", 215 cl::desc("Display only the first format specific file " 216 "header")); 217 218 static cl::alias 219 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 220 cl::aliasopt(PrivateHeaders)); 221 222 cl::opt<bool> llvm::FileHeaders( 223 "file-headers", 224 cl::desc("Display the contents of the overall file header")); 225 226 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 227 cl::aliasopt(FileHeaders)); 228 229 cl::opt<bool> 230 llvm::ArchiveHeaders("archive-headers", 231 cl::desc("Display archive header information")); 232 233 cl::alias 234 ArchiveHeadersShort("a", cl::desc("Alias for --archive-headers"), 235 cl::aliasopt(ArchiveHeaders)); 236 237 cl::opt<bool> 238 llvm::PrintImmHex("print-imm-hex", 239 cl::desc("Use hex format for immediate values")); 240 241 cl::opt<bool> PrintFaultMaps("fault-map-section", 242 cl::desc("Display contents of faultmap section")); 243 244 cl::opt<DIDumpType> llvm::DwarfDumpType( 245 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 246 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 247 248 cl::opt<bool> PrintSource( 249 "source", 250 cl::desc( 251 "Display source inlined with disassembly. Implies disassemble object")); 252 253 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 254 cl::aliasopt(PrintSource)); 255 256 cl::opt<bool> PrintLines("line-numbers", 257 cl::desc("Display source line numbers with " 258 "disassembly. Implies disassemble object")); 259 260 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 261 cl::aliasopt(PrintLines)); 262 263 cl::opt<unsigned long long> 264 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 265 cl::value_desc("address"), cl::init(0)); 266 cl::opt<unsigned long long> 267 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 268 cl::value_desc("address"), cl::init(UINT64_MAX)); 269 static StringRef ToolName; 270 271 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 272 273 namespace { 274 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 275 276 class SectionFilterIterator { 277 public: 278 SectionFilterIterator(FilterPredicate P, 279 llvm::object::section_iterator const &I, 280 llvm::object::section_iterator const &E) 281 : Predicate(std::move(P)), Iterator(I), End(E) { 282 ScanPredicate(); 283 } 284 const llvm::object::SectionRef &operator*() const { return *Iterator; } 285 SectionFilterIterator &operator++() { 286 ++Iterator; 287 ScanPredicate(); 288 return *this; 289 } 290 bool operator!=(SectionFilterIterator const &Other) const { 291 return Iterator != Other.Iterator; 292 } 293 294 private: 295 void ScanPredicate() { 296 while (Iterator != End && !Predicate(*Iterator)) { 297 ++Iterator; 298 } 299 } 300 FilterPredicate Predicate; 301 llvm::object::section_iterator Iterator; 302 llvm::object::section_iterator End; 303 }; 304 305 class SectionFilter { 306 public: 307 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 308 : Predicate(std::move(P)), Object(O) {} 309 SectionFilterIterator begin() { 310 return SectionFilterIterator(Predicate, Object.section_begin(), 311 Object.section_end()); 312 } 313 SectionFilterIterator end() { 314 return SectionFilterIterator(Predicate, Object.section_end(), 315 Object.section_end()); 316 } 317 318 private: 319 FilterPredicate Predicate; 320 llvm::object::ObjectFile const &Object; 321 }; 322 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 323 return SectionFilter( 324 [](llvm::object::SectionRef const &S) { 325 if (FilterSections.empty()) 326 return true; 327 llvm::StringRef String; 328 std::error_code error = S.getName(String); 329 if (error) 330 return false; 331 return is_contained(FilterSections, String); 332 }, 333 O); 334 } 335 } 336 337 void llvm::error(std::error_code EC) { 338 if (!EC) 339 return; 340 341 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 342 errs().flush(); 343 exit(1); 344 } 345 346 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 347 errs() << ToolName << ": " << Message << ".\n"; 348 errs().flush(); 349 exit(1); 350 } 351 352 void llvm::warn(StringRef Message) { 353 errs() << ToolName << ": warning: " << Message << ".\n"; 354 errs().flush(); 355 } 356 357 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 358 Twine Message) { 359 errs() << ToolName << ": '" << File << "': " << Message << ".\n"; 360 exit(1); 361 } 362 363 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 364 std::error_code EC) { 365 assert(EC); 366 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 367 exit(1); 368 } 369 370 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 371 llvm::Error E) { 372 assert(E); 373 std::string Buf; 374 raw_string_ostream OS(Buf); 375 logAllUnhandledErrors(std::move(E), OS); 376 OS.flush(); 377 errs() << ToolName << ": '" << File << "': " << Buf; 378 exit(1); 379 } 380 381 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 382 StringRef FileName, 383 llvm::Error E, 384 StringRef ArchitectureName) { 385 assert(E); 386 errs() << ToolName << ": "; 387 if (ArchiveName != "") 388 errs() << ArchiveName << "(" << FileName << ")"; 389 else 390 errs() << "'" << FileName << "'"; 391 if (!ArchitectureName.empty()) 392 errs() << " (for architecture " << ArchitectureName << ")"; 393 std::string Buf; 394 raw_string_ostream OS(Buf); 395 logAllUnhandledErrors(std::move(E), OS); 396 OS.flush(); 397 errs() << ": " << Buf; 398 exit(1); 399 } 400 401 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 402 const object::Archive::Child &C, 403 llvm::Error E, 404 StringRef ArchitectureName) { 405 Expected<StringRef> NameOrErr = C.getName(); 406 // TODO: if we have a error getting the name then it would be nice to print 407 // the index of which archive member this is and or its offset in the 408 // archive instead of "???" as the name. 409 if (!NameOrErr) { 410 consumeError(NameOrErr.takeError()); 411 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 412 } else 413 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 414 ArchitectureName); 415 } 416 417 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 418 // Figure out the target triple. 419 llvm::Triple TheTriple("unknown-unknown-unknown"); 420 if (TripleName.empty()) { 421 if (Obj) { 422 TheTriple = Obj->makeTriple(); 423 } 424 } else { 425 TheTriple.setTriple(Triple::normalize(TripleName)); 426 427 // Use the triple, but also try to combine with ARM build attributes. 428 if (Obj) { 429 auto Arch = Obj->getArch(); 430 if (Arch == Triple::arm || Arch == Triple::armeb) { 431 Obj->setARMSubArch(TheTriple); 432 } 433 } 434 } 435 436 // Get the target specific parser. 437 std::string Error; 438 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 439 Error); 440 if (!TheTarget) { 441 if (Obj) 442 report_error(Obj->getFileName(), "can't find target: " + Error); 443 else 444 error("can't find target: " + Error); 445 } 446 447 // Update the triple name and return the found target. 448 TripleName = TheTriple.getTriple(); 449 return TheTarget; 450 } 451 452 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 453 return a.getOffset() < b.getOffset(); 454 } 455 456 template <class ELFT> 457 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 458 const RelocationRef &RelRef, 459 SmallVectorImpl<char> &Result) { 460 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 461 462 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 463 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 464 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 465 466 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 467 468 auto SecOrErr = EF.getSection(Rel.d.a); 469 if (!SecOrErr) 470 return errorToErrorCode(SecOrErr.takeError()); 471 const Elf_Shdr *Sec = *SecOrErr; 472 auto SymTabOrErr = EF.getSection(Sec->sh_link); 473 if (!SymTabOrErr) 474 return errorToErrorCode(SymTabOrErr.takeError()); 475 const Elf_Shdr *SymTab = *SymTabOrErr; 476 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 477 SymTab->sh_type == ELF::SHT_DYNSYM); 478 auto StrTabSec = EF.getSection(SymTab->sh_link); 479 if (!StrTabSec) 480 return errorToErrorCode(StrTabSec.takeError()); 481 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 482 if (!StrTabOrErr) 483 return errorToErrorCode(StrTabOrErr.takeError()); 484 StringRef StrTab = *StrTabOrErr; 485 int64_t addend = 0; 486 // If there is no Symbol associated with the relocation, we set the undef 487 // boolean value to 'true'. This will prevent us from calling functions that 488 // requires the relocation to be associated with a symbol. 489 bool undef = false; 490 switch (Sec->sh_type) { 491 default: 492 return object_error::parse_failed; 493 case ELF::SHT_REL: { 494 // TODO: Read implicit addend from section data. 495 break; 496 } 497 case ELF::SHT_RELA: { 498 const Elf_Rela *ERela = Obj->getRela(Rel); 499 addend = ERela->r_addend; 500 undef = ERela->getSymbol(false) == 0; 501 break; 502 } 503 } 504 StringRef Target; 505 if (!undef) { 506 symbol_iterator SI = RelRef.getSymbol(); 507 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 508 if (symb->getType() == ELF::STT_SECTION) { 509 Expected<section_iterator> SymSI = SI->getSection(); 510 if (!SymSI) 511 return errorToErrorCode(SymSI.takeError()); 512 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 513 auto SecName = EF.getSectionName(SymSec); 514 if (!SecName) 515 return errorToErrorCode(SecName.takeError()); 516 Target = *SecName; 517 } else { 518 Expected<StringRef> SymName = symb->getName(StrTab); 519 if (!SymName) 520 return errorToErrorCode(SymName.takeError()); 521 Target = *SymName; 522 } 523 } else 524 Target = "*ABS*"; 525 526 // Default scheme is to print Target, as well as "+ <addend>" for nonzero 527 // addend. Should be acceptable for all normal purposes. 528 std::string fmtbuf; 529 raw_string_ostream fmt(fmtbuf); 530 fmt << Target; 531 if (addend != 0) 532 fmt << (addend < 0 ? "" : "+") << addend; 533 fmt.flush(); 534 Result.append(fmtbuf.begin(), fmtbuf.end()); 535 return std::error_code(); 536 } 537 538 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 539 const RelocationRef &Rel, 540 SmallVectorImpl<char> &Result) { 541 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 542 return getRelocationValueString(ELF32LE, Rel, Result); 543 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 544 return getRelocationValueString(ELF64LE, Rel, Result); 545 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 546 return getRelocationValueString(ELF32BE, Rel, Result); 547 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 548 return getRelocationValueString(ELF64BE, Rel, Result); 549 } 550 551 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 552 const RelocationRef &Rel, 553 SmallVectorImpl<char> &Result) { 554 symbol_iterator SymI = Rel.getSymbol(); 555 Expected<StringRef> SymNameOrErr = SymI->getName(); 556 if (!SymNameOrErr) 557 return errorToErrorCode(SymNameOrErr.takeError()); 558 StringRef SymName = *SymNameOrErr; 559 Result.append(SymName.begin(), SymName.end()); 560 return std::error_code(); 561 } 562 563 static void printRelocationTargetName(const MachOObjectFile *O, 564 const MachO::any_relocation_info &RE, 565 raw_string_ostream &fmt) { 566 bool IsScattered = O->isRelocationScattered(RE); 567 568 // Target of a scattered relocation is an address. In the interest of 569 // generating pretty output, scan through the symbol table looking for a 570 // symbol that aligns with that address. If we find one, print it. 571 // Otherwise, we just print the hex address of the target. 572 if (IsScattered) { 573 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 574 575 for (const SymbolRef &Symbol : O->symbols()) { 576 std::error_code ec; 577 Expected<uint64_t> Addr = Symbol.getAddress(); 578 if (!Addr) 579 report_error(O->getFileName(), Addr.takeError()); 580 if (*Addr != Val) 581 continue; 582 Expected<StringRef> Name = Symbol.getName(); 583 if (!Name) 584 report_error(O->getFileName(), Name.takeError()); 585 fmt << *Name; 586 return; 587 } 588 589 // If we couldn't find a symbol that this relocation refers to, try 590 // to find a section beginning instead. 591 for (const SectionRef &Section : ToolSectionFilter(*O)) { 592 std::error_code ec; 593 594 StringRef Name; 595 uint64_t Addr = Section.getAddress(); 596 if (Addr != Val) 597 continue; 598 if ((ec = Section.getName(Name))) 599 report_error(O->getFileName(), ec); 600 fmt << Name; 601 return; 602 } 603 604 fmt << format("0x%x", Val); 605 return; 606 } 607 608 StringRef S; 609 bool isExtern = O->getPlainRelocationExternal(RE); 610 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 611 612 if (O->getAnyRelocationType(RE) == MachO::ARM64_RELOC_ADDEND) { 613 fmt << format("0x%0" PRIx64, Val); 614 return; 615 } else if (isExtern) { 616 symbol_iterator SI = O->symbol_begin(); 617 advance(SI, Val); 618 Expected<StringRef> SOrErr = SI->getName(); 619 if (!SOrErr) 620 report_error(O->getFileName(), SOrErr.takeError()); 621 S = *SOrErr; 622 } else { 623 section_iterator SI = O->section_begin(); 624 // Adjust for the fact that sections are 1-indexed. 625 if (Val == 0) { 626 fmt << "0 (?,?)"; 627 return; 628 } 629 uint32_t i = Val - 1; 630 while (i != 0 && SI != O->section_end()) { 631 i--; 632 advance(SI, 1); 633 } 634 if (SI == O->section_end()) 635 fmt << Val << " (?,?)"; 636 else 637 SI->getName(S); 638 } 639 640 fmt << S; 641 } 642 643 static std::error_code getRelocationValueString(const WasmObjectFile *Obj, 644 const RelocationRef &RelRef, 645 SmallVectorImpl<char> &Result) { 646 const wasm::WasmRelocation& Rel = Obj->getWasmRelocation(RelRef); 647 symbol_iterator SI = RelRef.getSymbol(); 648 std::string fmtbuf; 649 raw_string_ostream fmt(fmtbuf); 650 if (SI == Obj->symbol_end()) { 651 // Not all wasm relocations have symbols associated with them. 652 // In particular R_WEBASSEMBLY_TYPE_INDEX_LEB. 653 fmt << Rel.Index; 654 } else { 655 Expected<StringRef> SymNameOrErr = SI->getName(); 656 if (!SymNameOrErr) 657 return errorToErrorCode(SymNameOrErr.takeError()); 658 StringRef SymName = *SymNameOrErr; 659 Result.append(SymName.begin(), SymName.end()); 660 } 661 fmt << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 662 fmt.flush(); 663 Result.append(fmtbuf.begin(), fmtbuf.end()); 664 return std::error_code(); 665 } 666 667 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 668 const RelocationRef &RelRef, 669 SmallVectorImpl<char> &Result) { 670 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 671 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 672 673 unsigned Arch = Obj->getArch(); 674 675 std::string fmtbuf; 676 raw_string_ostream fmt(fmtbuf); 677 unsigned Type = Obj->getAnyRelocationType(RE); 678 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 679 680 // Determine any addends that should be displayed with the relocation. 681 // These require decoding the relocation type, which is triple-specific. 682 683 // X86_64 has entirely custom relocation types. 684 if (Arch == Triple::x86_64) { 685 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 686 687 switch (Type) { 688 case MachO::X86_64_RELOC_GOT_LOAD: 689 case MachO::X86_64_RELOC_GOT: { 690 printRelocationTargetName(Obj, RE, fmt); 691 fmt << "@GOT"; 692 if (isPCRel) 693 fmt << "PCREL"; 694 break; 695 } 696 case MachO::X86_64_RELOC_SUBTRACTOR: { 697 DataRefImpl RelNext = Rel; 698 Obj->moveRelocationNext(RelNext); 699 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 700 701 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 702 // X86_64_RELOC_UNSIGNED. 703 // NOTE: Scattered relocations don't exist on x86_64. 704 unsigned RType = Obj->getAnyRelocationType(RENext); 705 if (RType != MachO::X86_64_RELOC_UNSIGNED) 706 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 707 "X86_64_RELOC_SUBTRACTOR."); 708 709 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 710 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 711 printRelocationTargetName(Obj, RENext, fmt); 712 fmt << "-"; 713 printRelocationTargetName(Obj, RE, fmt); 714 break; 715 } 716 case MachO::X86_64_RELOC_TLV: 717 printRelocationTargetName(Obj, RE, fmt); 718 fmt << "@TLV"; 719 if (isPCRel) 720 fmt << "P"; 721 break; 722 case MachO::X86_64_RELOC_SIGNED_1: 723 printRelocationTargetName(Obj, RE, fmt); 724 fmt << "-1"; 725 break; 726 case MachO::X86_64_RELOC_SIGNED_2: 727 printRelocationTargetName(Obj, RE, fmt); 728 fmt << "-2"; 729 break; 730 case MachO::X86_64_RELOC_SIGNED_4: 731 printRelocationTargetName(Obj, RE, fmt); 732 fmt << "-4"; 733 break; 734 default: 735 printRelocationTargetName(Obj, RE, fmt); 736 break; 737 } 738 // X86 and ARM share some relocation types in common. 739 } else if (Arch == Triple::x86 || Arch == Triple::arm || 740 Arch == Triple::ppc) { 741 // Generic relocation types... 742 switch (Type) { 743 case MachO::GENERIC_RELOC_PAIR: // prints no info 744 return std::error_code(); 745 case MachO::GENERIC_RELOC_SECTDIFF: { 746 DataRefImpl RelNext = Rel; 747 Obj->moveRelocationNext(RelNext); 748 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 749 750 // X86 sect diff's must be followed by a relocation of type 751 // GENERIC_RELOC_PAIR. 752 unsigned RType = Obj->getAnyRelocationType(RENext); 753 754 if (RType != MachO::GENERIC_RELOC_PAIR) 755 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 756 "GENERIC_RELOC_SECTDIFF."); 757 758 printRelocationTargetName(Obj, RE, fmt); 759 fmt << "-"; 760 printRelocationTargetName(Obj, RENext, fmt); 761 break; 762 } 763 } 764 765 if (Arch == Triple::x86 || Arch == Triple::ppc) { 766 switch (Type) { 767 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 768 DataRefImpl RelNext = Rel; 769 Obj->moveRelocationNext(RelNext); 770 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 771 772 // X86 sect diff's must be followed by a relocation of type 773 // GENERIC_RELOC_PAIR. 774 unsigned RType = Obj->getAnyRelocationType(RENext); 775 if (RType != MachO::GENERIC_RELOC_PAIR) 776 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 777 "GENERIC_RELOC_LOCAL_SECTDIFF."); 778 779 printRelocationTargetName(Obj, RE, fmt); 780 fmt << "-"; 781 printRelocationTargetName(Obj, RENext, fmt); 782 break; 783 } 784 case MachO::GENERIC_RELOC_TLV: { 785 printRelocationTargetName(Obj, RE, fmt); 786 fmt << "@TLV"; 787 if (IsPCRel) 788 fmt << "P"; 789 break; 790 } 791 default: 792 printRelocationTargetName(Obj, RE, fmt); 793 } 794 } else { // ARM-specific relocations 795 switch (Type) { 796 case MachO::ARM_RELOC_HALF: 797 case MachO::ARM_RELOC_HALF_SECTDIFF: { 798 // Half relocations steal a bit from the length field to encode 799 // whether this is an upper16 or a lower16 relocation. 800 bool isUpper = (Obj->getAnyRelocationLength(RE) & 0x1) == 1; 801 802 if (isUpper) 803 fmt << ":upper16:("; 804 else 805 fmt << ":lower16:("; 806 printRelocationTargetName(Obj, RE, fmt); 807 808 DataRefImpl RelNext = Rel; 809 Obj->moveRelocationNext(RelNext); 810 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 811 812 // ARM half relocs must be followed by a relocation of type 813 // ARM_RELOC_PAIR. 814 unsigned RType = Obj->getAnyRelocationType(RENext); 815 if (RType != MachO::ARM_RELOC_PAIR) 816 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 817 "ARM_RELOC_HALF"); 818 819 // NOTE: The half of the target virtual address is stashed in the 820 // address field of the secondary relocation, but we can't reverse 821 // engineer the constant offset from it without decoding the movw/movt 822 // instruction to find the other half in its immediate field. 823 824 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 825 // symbol/section pointer of the follow-on relocation. 826 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 827 fmt << "-"; 828 printRelocationTargetName(Obj, RENext, fmt); 829 } 830 831 fmt << ")"; 832 break; 833 } 834 default: { printRelocationTargetName(Obj, RE, fmt); } 835 } 836 } 837 } else 838 printRelocationTargetName(Obj, RE, fmt); 839 840 fmt.flush(); 841 Result.append(fmtbuf.begin(), fmtbuf.end()); 842 return std::error_code(); 843 } 844 845 static std::error_code getRelocationValueString(const RelocationRef &Rel, 846 SmallVectorImpl<char> &Result) { 847 const ObjectFile *Obj = Rel.getObject(); 848 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 849 return getRelocationValueString(ELF, Rel, Result); 850 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 851 return getRelocationValueString(COFF, Rel, Result); 852 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 853 return getRelocationValueString(Wasm, Rel, Result); 854 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 855 return getRelocationValueString(MachO, Rel, Result); 856 llvm_unreachable("unknown object file format"); 857 } 858 859 /// Indicates whether this relocation should hidden when listing 860 /// relocations, usually because it is the trailing part of a multipart 861 /// relocation that will be printed as part of the leading relocation. 862 static bool getHidden(RelocationRef RelRef) { 863 const ObjectFile *Obj = RelRef.getObject(); 864 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 865 if (!MachO) 866 return false; 867 868 unsigned Arch = MachO->getArch(); 869 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 870 uint64_t Type = MachO->getRelocationType(Rel); 871 872 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 873 // is always hidden. 874 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 875 if (Type == MachO::GENERIC_RELOC_PAIR) 876 return true; 877 } else if (Arch == Triple::x86_64) { 878 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 879 // an X86_64_RELOC_SUBTRACTOR. 880 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 881 DataRefImpl RelPrev = Rel; 882 RelPrev.d.a--; 883 uint64_t PrevType = MachO->getRelocationType(RelPrev); 884 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 885 return true; 886 } 887 } 888 889 return false; 890 } 891 892 namespace { 893 class SourcePrinter { 894 protected: 895 DILineInfo OldLineInfo; 896 const ObjectFile *Obj = nullptr; 897 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 898 // File name to file contents of source 899 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 900 // Mark the line endings of the cached source 901 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 902 903 private: 904 bool cacheSource(const DILineInfo& LineInfoFile); 905 906 public: 907 SourcePrinter() = default; 908 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 909 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 910 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 911 DefaultArch); 912 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 913 } 914 virtual ~SourcePrinter() = default; 915 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 916 StringRef Delimiter = "; "); 917 }; 918 919 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 920 std::unique_ptr<MemoryBuffer> Buffer; 921 if (LineInfo.Source) { 922 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 923 } else { 924 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 925 if (!BufferOrError) 926 return false; 927 Buffer = std::move(*BufferOrError); 928 } 929 // Chomp the file to get lines 930 size_t BufferSize = Buffer->getBufferSize(); 931 const char *BufferStart = Buffer->getBufferStart(); 932 for (const char *Start = BufferStart, *End = BufferStart; 933 End < BufferStart + BufferSize; End++) 934 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 935 (*End == '\r' && *(End + 1) == '\n')) { 936 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 937 if (*End == '\r') 938 End++; 939 Start = End + 1; 940 } 941 SourceCache[LineInfo.FileName] = std::move(Buffer); 942 return true; 943 } 944 945 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 946 StringRef Delimiter) { 947 if (!Symbolizer) 948 return; 949 DILineInfo LineInfo = DILineInfo(); 950 auto ExpectecLineInfo = 951 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 952 if (!ExpectecLineInfo) 953 consumeError(ExpectecLineInfo.takeError()); 954 else 955 LineInfo = *ExpectecLineInfo; 956 957 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 958 LineInfo.Line == 0) 959 return; 960 961 if (PrintLines) 962 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 963 if (PrintSource) { 964 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 965 if (!cacheSource(LineInfo)) 966 return; 967 auto FileBuffer = SourceCache.find(LineInfo.FileName); 968 if (FileBuffer != SourceCache.end()) { 969 auto LineBuffer = LineCache.find(LineInfo.FileName); 970 if (LineBuffer != LineCache.end()) { 971 if (LineInfo.Line > LineBuffer->second.size()) 972 return; 973 // Vector begins at 0, line numbers are non-zero 974 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 975 << "\n"; 976 } 977 } 978 } 979 OldLineInfo = LineInfo; 980 } 981 982 static bool isArmElf(const ObjectFile *Obj) { 983 return (Obj->isELF() && 984 (Obj->getArch() == Triple::aarch64 || 985 Obj->getArch() == Triple::aarch64_be || 986 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 987 Obj->getArch() == Triple::thumb || 988 Obj->getArch() == Triple::thumbeb)); 989 } 990 991 class PrettyPrinter { 992 public: 993 virtual ~PrettyPrinter() = default; 994 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 995 ArrayRef<uint8_t> Bytes, uint64_t Address, 996 raw_ostream &OS, StringRef Annot, 997 MCSubtargetInfo const &STI, SourcePrinter *SP, 998 std::vector<RelocationRef> *Rels = nullptr) { 999 if (SP && (PrintSource || PrintLines)) 1000 SP->printSourceLine(OS, Address); 1001 if (!NoLeadingAddr) 1002 OS << format("%8" PRIx64 ":", Address); 1003 if (!NoShowRawInsn) { 1004 OS << "\t"; 1005 dumpBytes(Bytes, OS); 1006 } 1007 if (MI) 1008 IP.printInst(MI, OS, "", STI); 1009 else 1010 OS << " <unknown>"; 1011 } 1012 }; 1013 PrettyPrinter PrettyPrinterInst; 1014 class HexagonPrettyPrinter : public PrettyPrinter { 1015 public: 1016 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 1017 raw_ostream &OS) { 1018 uint32_t opcode = 1019 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 1020 if (!NoLeadingAddr) 1021 OS << format("%8" PRIx64 ":", Address); 1022 if (!NoShowRawInsn) { 1023 OS << "\t"; 1024 dumpBytes(Bytes.slice(0, 4), OS); 1025 OS << format("%08" PRIx32, opcode); 1026 } 1027 } 1028 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1029 uint64_t Address, raw_ostream &OS, StringRef Annot, 1030 MCSubtargetInfo const &STI, SourcePrinter *SP, 1031 std::vector<RelocationRef> *Rels) override { 1032 if (SP && (PrintSource || PrintLines)) 1033 SP->printSourceLine(OS, Address, ""); 1034 if (!MI) { 1035 printLead(Bytes, Address, OS); 1036 OS << " <unknown>"; 1037 return; 1038 } 1039 std::string Buffer; 1040 { 1041 raw_string_ostream TempStream(Buffer); 1042 IP.printInst(MI, TempStream, "", STI); 1043 } 1044 StringRef Contents(Buffer); 1045 // Split off bundle attributes 1046 auto PacketBundle = Contents.rsplit('\n'); 1047 // Split off first instruction from the rest 1048 auto HeadTail = PacketBundle.first.split('\n'); 1049 auto Preamble = " { "; 1050 auto Separator = ""; 1051 StringRef Fmt = "\t\t\t%08" PRIx64 ": "; 1052 std::vector<RelocationRef>::const_iterator rel_cur = Rels->begin(); 1053 std::vector<RelocationRef>::const_iterator rel_end = Rels->end(); 1054 1055 // Hexagon's packets require relocations to be inline rather than 1056 // clustered at the end of the packet. 1057 auto PrintReloc = [&]() -> void { 1058 while ((rel_cur != rel_end) && (rel_cur->getOffset() <= Address)) { 1059 if (rel_cur->getOffset() == Address) { 1060 SmallString<16> name; 1061 SmallString<32> val; 1062 rel_cur->getTypeName(name); 1063 error(getRelocationValueString(*rel_cur, val)); 1064 OS << Separator << format(Fmt.data(), Address) << name << "\t" << val 1065 << "\n"; 1066 return; 1067 } 1068 rel_cur++; 1069 } 1070 }; 1071 1072 while(!HeadTail.first.empty()) { 1073 OS << Separator; 1074 Separator = "\n"; 1075 if (SP && (PrintSource || PrintLines)) 1076 SP->printSourceLine(OS, Address, ""); 1077 printLead(Bytes, Address, OS); 1078 OS << Preamble; 1079 Preamble = " "; 1080 StringRef Inst; 1081 auto Duplex = HeadTail.first.split('\v'); 1082 if(!Duplex.second.empty()){ 1083 OS << Duplex.first; 1084 OS << "; "; 1085 Inst = Duplex.second; 1086 } 1087 else 1088 Inst = HeadTail.first; 1089 OS << Inst; 1090 HeadTail = HeadTail.second.split('\n'); 1091 if (HeadTail.first.empty()) 1092 OS << " } " << PacketBundle.second; 1093 PrintReloc(); 1094 Bytes = Bytes.slice(4); 1095 Address += 4; 1096 } 1097 } 1098 }; 1099 HexagonPrettyPrinter HexagonPrettyPrinterInst; 1100 1101 class AMDGCNPrettyPrinter : public PrettyPrinter { 1102 public: 1103 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1104 uint64_t Address, raw_ostream &OS, StringRef Annot, 1105 MCSubtargetInfo const &STI, SourcePrinter *SP, 1106 std::vector<RelocationRef> *Rels) override { 1107 if (SP && (PrintSource || PrintLines)) 1108 SP->printSourceLine(OS, Address); 1109 1110 typedef support::ulittle32_t U32; 1111 1112 if (MI) { 1113 SmallString<40> InstStr; 1114 raw_svector_ostream IS(InstStr); 1115 1116 IP.printInst(MI, IS, "", STI); 1117 1118 OS << left_justify(IS.str(), 60); 1119 } else { 1120 // an unrecognized encoding - this is probably data so represent it 1121 // using the .long directive, or .byte directive if fewer than 4 bytes 1122 // remaining 1123 if (Bytes.size() >= 4) { 1124 OS << format("\t.long 0x%08" PRIx32 " ", 1125 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 1126 OS.indent(42); 1127 } else { 1128 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 1129 for (unsigned int i = 1; i < Bytes.size(); i++) 1130 OS << format(", 0x%02" PRIx8, Bytes[i]); 1131 OS.indent(55 - (6 * Bytes.size())); 1132 } 1133 } 1134 1135 OS << format("// %012" PRIX64 ": ", Address); 1136 if (Bytes.size() >=4) { 1137 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 1138 Bytes.size() / sizeof(U32))) 1139 // D should be explicitly casted to uint32_t here as it is passed 1140 // by format to snprintf as vararg. 1141 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 1142 } else { 1143 for (unsigned int i = 0; i < Bytes.size(); i++) 1144 OS << format("%02" PRIX8 " ", Bytes[i]); 1145 } 1146 1147 if (!Annot.empty()) 1148 OS << "// " << Annot; 1149 } 1150 }; 1151 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 1152 1153 class BPFPrettyPrinter : public PrettyPrinter { 1154 public: 1155 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1156 uint64_t Address, raw_ostream &OS, StringRef Annot, 1157 MCSubtargetInfo const &STI, SourcePrinter *SP, 1158 std::vector<RelocationRef> *Rels) override { 1159 if (SP && (PrintSource || PrintLines)) 1160 SP->printSourceLine(OS, Address); 1161 if (!NoLeadingAddr) 1162 OS << format("%8" PRId64 ":", Address / 8); 1163 if (!NoShowRawInsn) { 1164 OS << "\t"; 1165 dumpBytes(Bytes, OS); 1166 } 1167 if (MI) 1168 IP.printInst(MI, OS, "", STI); 1169 else 1170 OS << " <unknown>"; 1171 } 1172 }; 1173 BPFPrettyPrinter BPFPrettyPrinterInst; 1174 1175 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 1176 switch(Triple.getArch()) { 1177 default: 1178 return PrettyPrinterInst; 1179 case Triple::hexagon: 1180 return HexagonPrettyPrinterInst; 1181 case Triple::amdgcn: 1182 return AMDGCNPrettyPrinterInst; 1183 case Triple::bpfel: 1184 case Triple::bpfeb: 1185 return BPFPrettyPrinterInst; 1186 } 1187 } 1188 } 1189 1190 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1191 assert(Obj->isELF()); 1192 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1193 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1194 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1195 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1196 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1197 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1198 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1199 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1200 llvm_unreachable("Unsupported binary format"); 1201 } 1202 1203 template <class ELFT> static void 1204 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1205 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1206 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1207 uint8_t SymbolType = Symbol.getELFType(); 1208 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1209 continue; 1210 1211 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1212 if (!AddressOrErr) 1213 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1214 uint64_t Address = *AddressOrErr; 1215 1216 Expected<StringRef> Name = Symbol.getName(); 1217 if (!Name) 1218 report_error(Obj->getFileName(), Name.takeError()); 1219 if (Name->empty()) 1220 continue; 1221 1222 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1223 if (!SectionOrErr) 1224 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1225 section_iterator SecI = *SectionOrErr; 1226 if (SecI == Obj->section_end()) 1227 continue; 1228 1229 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1230 } 1231 } 1232 1233 static void 1234 addDynamicElfSymbols(const ObjectFile *Obj, 1235 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1236 assert(Obj->isELF()); 1237 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1238 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1239 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1240 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1241 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1242 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1243 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1244 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1245 else 1246 llvm_unreachable("Unsupported binary format"); 1247 } 1248 1249 static void addPltEntries(const ObjectFile *Obj, 1250 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 1251 StringSaver &Saver) { 1252 Optional<SectionRef> Plt = None; 1253 for (const SectionRef &Section : Obj->sections()) { 1254 StringRef Name; 1255 if (Section.getName(Name)) 1256 continue; 1257 if (Name == ".plt") 1258 Plt = Section; 1259 } 1260 if (!Plt) 1261 return; 1262 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1263 for (auto PltEntry : ElfObj->getPltAddresses()) { 1264 SymbolRef Symbol(PltEntry.first, ElfObj); 1265 1266 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1267 1268 Expected<StringRef> NameOrErr = Symbol.getName(); 1269 if (!NameOrErr) 1270 report_error(Obj->getFileName(), NameOrErr.takeError()); 1271 if (NameOrErr->empty()) 1272 continue; 1273 StringRef Name = Saver.save((*NameOrErr + "@plt").str()); 1274 1275 AllSymbols[*Plt].emplace_back(PltEntry.second, Name, SymbolType); 1276 } 1277 } 1278 } 1279 1280 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1281 if (StartAddress > StopAddress) 1282 error("Start address should be less than stop address"); 1283 1284 const Target *TheTarget = getTarget(Obj); 1285 1286 // Package up features to be passed to target/subtarget 1287 SubtargetFeatures Features = Obj->getFeatures(); 1288 if (MAttrs.size()) { 1289 for (unsigned i = 0; i != MAttrs.size(); ++i) 1290 Features.AddFeature(MAttrs[i]); 1291 } 1292 1293 std::unique_ptr<const MCRegisterInfo> MRI( 1294 TheTarget->createMCRegInfo(TripleName)); 1295 if (!MRI) 1296 report_error(Obj->getFileName(), "no register info for target " + 1297 TripleName); 1298 1299 // Set up disassembler. 1300 std::unique_ptr<const MCAsmInfo> AsmInfo( 1301 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1302 if (!AsmInfo) 1303 report_error(Obj->getFileName(), "no assembly info for target " + 1304 TripleName); 1305 std::unique_ptr<const MCSubtargetInfo> STI( 1306 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1307 if (!STI) 1308 report_error(Obj->getFileName(), "no subtarget info for target " + 1309 TripleName); 1310 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1311 if (!MII) 1312 report_error(Obj->getFileName(), "no instruction info for target " + 1313 TripleName); 1314 MCObjectFileInfo MOFI; 1315 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1316 // FIXME: for now initialize MCObjectFileInfo with default values 1317 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1318 1319 std::unique_ptr<MCDisassembler> DisAsm( 1320 TheTarget->createMCDisassembler(*STI, Ctx)); 1321 if (!DisAsm) 1322 report_error(Obj->getFileName(), "no disassembler for target " + 1323 TripleName); 1324 1325 std::unique_ptr<const MCInstrAnalysis> MIA( 1326 TheTarget->createMCInstrAnalysis(MII.get())); 1327 1328 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1329 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1330 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1331 if (!IP) 1332 report_error(Obj->getFileName(), "no instruction printer for target " + 1333 TripleName); 1334 IP->setPrintImmHex(PrintImmHex); 1335 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1336 1337 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1338 "\t\t\t%08" PRIx64 ": "; 1339 1340 SourcePrinter SP(Obj, TheTarget->getName()); 1341 1342 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1343 // in RelocSecs contain the relocations for section S. 1344 std::error_code EC; 1345 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1346 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1347 section_iterator Sec2 = Section.getRelocatedSection(); 1348 if (Sec2 != Obj->section_end()) 1349 SectionRelocMap[*Sec2].push_back(Section); 1350 } 1351 1352 // Create a mapping from virtual address to symbol name. This is used to 1353 // pretty print the symbols while disassembling. 1354 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1355 SectionSymbolsTy AbsoluteSymbols; 1356 for (const SymbolRef &Symbol : Obj->symbols()) { 1357 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1358 if (!AddressOrErr) 1359 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1360 uint64_t Address = *AddressOrErr; 1361 1362 Expected<StringRef> Name = Symbol.getName(); 1363 if (!Name) 1364 report_error(Obj->getFileName(), Name.takeError()); 1365 if (Name->empty()) 1366 continue; 1367 1368 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1369 if (!SectionOrErr) 1370 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1371 1372 uint8_t SymbolType = ELF::STT_NOTYPE; 1373 if (Obj->isELF()) 1374 SymbolType = getElfSymbolType(Obj, Symbol); 1375 1376 section_iterator SecI = *SectionOrErr; 1377 if (SecI != Obj->section_end()) 1378 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1379 else 1380 AbsoluteSymbols.emplace_back(Address, *Name, SymbolType); 1381 1382 1383 } 1384 if (AllSymbols.empty() && Obj->isELF()) 1385 addDynamicElfSymbols(Obj, AllSymbols); 1386 1387 BumpPtrAllocator A; 1388 StringSaver Saver(A); 1389 addPltEntries(Obj, AllSymbols, Saver); 1390 1391 // Create a mapping from virtual address to section. 1392 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1393 for (SectionRef Sec : Obj->sections()) 1394 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1395 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1396 1397 // Linked executables (.exe and .dll files) typically don't include a real 1398 // symbol table but they might contain an export table. 1399 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1400 for (const auto &ExportEntry : COFFObj->export_directories()) { 1401 StringRef Name; 1402 error(ExportEntry.getSymbolName(Name)); 1403 if (Name.empty()) 1404 continue; 1405 uint32_t RVA; 1406 error(ExportEntry.getExportRVA(RVA)); 1407 1408 uint64_t VA = COFFObj->getImageBase() + RVA; 1409 auto Sec = std::upper_bound( 1410 SectionAddresses.begin(), SectionAddresses.end(), VA, 1411 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1412 return LHS < RHS.first; 1413 }); 1414 if (Sec != SectionAddresses.begin()) 1415 --Sec; 1416 else 1417 Sec = SectionAddresses.end(); 1418 1419 if (Sec != SectionAddresses.end()) 1420 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1421 else 1422 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1423 } 1424 } 1425 1426 // Sort all the symbols, this allows us to use a simple binary search to find 1427 // a symbol near an address. 1428 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1429 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1430 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1431 1432 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1433 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1434 continue; 1435 1436 uint64_t SectionAddr = Section.getAddress(); 1437 uint64_t SectSize = Section.getSize(); 1438 if (!SectSize) 1439 continue; 1440 1441 // Get the list of all the symbols in this section. 1442 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1443 std::vector<uint64_t> DataMappingSymsAddr; 1444 std::vector<uint64_t> TextMappingSymsAddr; 1445 if (isArmElf(Obj)) { 1446 for (const auto &Symb : Symbols) { 1447 uint64_t Address = std::get<0>(Symb); 1448 StringRef Name = std::get<1>(Symb); 1449 if (Name.startswith("$d")) 1450 DataMappingSymsAddr.push_back(Address - SectionAddr); 1451 if (Name.startswith("$x")) 1452 TextMappingSymsAddr.push_back(Address - SectionAddr); 1453 if (Name.startswith("$a")) 1454 TextMappingSymsAddr.push_back(Address - SectionAddr); 1455 if (Name.startswith("$t")) 1456 TextMappingSymsAddr.push_back(Address - SectionAddr); 1457 } 1458 } 1459 1460 llvm::sort(DataMappingSymsAddr); 1461 llvm::sort(TextMappingSymsAddr); 1462 1463 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1464 // AMDGPU disassembler uses symbolizer for printing labels 1465 std::unique_ptr<MCRelocationInfo> RelInfo( 1466 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1467 if (RelInfo) { 1468 std::unique_ptr<MCSymbolizer> Symbolizer( 1469 TheTarget->createMCSymbolizer( 1470 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1471 DisAsm->setSymbolizer(std::move(Symbolizer)); 1472 } 1473 } 1474 1475 // Make a list of all the relocations for this section. 1476 std::vector<RelocationRef> Rels; 1477 if (InlineRelocs) { 1478 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1479 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1480 Rels.push_back(Reloc); 1481 } 1482 } 1483 } 1484 1485 // Sort relocations by address. 1486 llvm::sort(Rels, RelocAddressLess); 1487 1488 StringRef SegmentName = ""; 1489 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1490 DataRefImpl DR = Section.getRawDataRefImpl(); 1491 SegmentName = MachO->getSectionFinalSegmentName(DR); 1492 } 1493 StringRef SectionName; 1494 error(Section.getName(SectionName)); 1495 1496 // If the section has no symbol at the start, just insert a dummy one. 1497 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1498 Symbols.insert( 1499 Symbols.begin(), 1500 std::make_tuple(SectionAddr, SectionName, 1501 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1502 } 1503 1504 SmallString<40> Comments; 1505 raw_svector_ostream CommentStream(Comments); 1506 1507 StringRef BytesStr; 1508 error(Section.getContents(BytesStr)); 1509 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1510 BytesStr.size()); 1511 1512 uint64_t Size; 1513 uint64_t Index; 1514 bool PrintedSection = false; 1515 1516 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1517 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1518 // Disassemble symbol by symbol. 1519 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1520 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1521 // The end is either the section end or the beginning of the next 1522 // symbol. 1523 uint64_t End = 1524 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1525 // Don't try to disassemble beyond the end of section contents. 1526 if (End > SectSize) 1527 End = SectSize; 1528 // If this symbol has the same address as the next symbol, then skip it. 1529 if (Start >= End) 1530 continue; 1531 1532 // Check if we need to skip symbol 1533 // Skip if the symbol's data is not between StartAddress and StopAddress 1534 if (End + SectionAddr < StartAddress || 1535 Start + SectionAddr > StopAddress) { 1536 continue; 1537 } 1538 1539 /// Skip if user requested specific symbols and this is not in the list 1540 if (!DisasmFuncsSet.empty() && 1541 !DisasmFuncsSet.count(std::get<1>(Symbols[si]))) 1542 continue; 1543 1544 if (!PrintedSection) { 1545 PrintedSection = true; 1546 outs() << "Disassembly of section "; 1547 if (!SegmentName.empty()) 1548 outs() << SegmentName << ","; 1549 outs() << SectionName << ':'; 1550 } 1551 1552 // Stop disassembly at the stop address specified 1553 if (End + SectionAddr > StopAddress) 1554 End = StopAddress - SectionAddr; 1555 1556 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1557 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1558 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1559 Start += 256; 1560 } 1561 if (si == se - 1 || 1562 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1563 // cut trailing zeroes at the end of kernel 1564 // cut up to 256 bytes 1565 const uint64_t EndAlign = 256; 1566 const auto Limit = End - (std::min)(EndAlign, End - Start); 1567 while (End > Limit && 1568 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1569 End -= 4; 1570 } 1571 } 1572 1573 auto PrintSymbol = [](StringRef Name) { 1574 outs() << '\n' << Name << ":\n"; 1575 }; 1576 StringRef SymbolName = std::get<1>(Symbols[si]); 1577 if (Demangle) { 1578 char *DemangledSymbol = nullptr; 1579 size_t Size = 0; 1580 int Status = -1; 1581 if (SymbolName.startswith("_Z")) 1582 DemangledSymbol = itaniumDemangle(SymbolName.data(), DemangledSymbol, 1583 &Size, &Status); 1584 else if (SymbolName.startswith("?")) 1585 DemangledSymbol = microsoftDemangle(SymbolName.data(), 1586 DemangledSymbol, &Size, &Status); 1587 1588 if (Status == 0 && DemangledSymbol) 1589 PrintSymbol(StringRef(DemangledSymbol)); 1590 else 1591 PrintSymbol(SymbolName); 1592 1593 if (DemangledSymbol) 1594 free(DemangledSymbol); 1595 } else 1596 PrintSymbol(SymbolName); 1597 1598 // Don't print raw contents of a virtual section. A virtual section 1599 // doesn't have any contents in the file. 1600 if (Section.isVirtual()) { 1601 outs() << "...\n"; 1602 continue; 1603 } 1604 1605 #ifndef NDEBUG 1606 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1607 #else 1608 raw_ostream &DebugOut = nulls(); 1609 #endif 1610 1611 for (Index = Start; Index < End; Index += Size) { 1612 MCInst Inst; 1613 1614 if (Index + SectionAddr < StartAddress || 1615 Index + SectionAddr > StopAddress) { 1616 // skip byte by byte till StartAddress is reached 1617 Size = 1; 1618 continue; 1619 } 1620 // AArch64 ELF binaries can interleave data and text in the 1621 // same section. We rely on the markers introduced to 1622 // understand what we need to dump. If the data marker is within a 1623 // function, it is denoted as a word/short etc 1624 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1625 !DisassembleAll) { 1626 uint64_t Stride = 0; 1627 1628 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1629 DataMappingSymsAddr.end(), Index); 1630 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1631 // Switch to data. 1632 while (Index < End) { 1633 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1634 outs() << "\t"; 1635 if (Index + 4 <= End) { 1636 Stride = 4; 1637 dumpBytes(Bytes.slice(Index, 4), outs()); 1638 outs() << "\t.word\t"; 1639 uint32_t Data = 0; 1640 if (Obj->isLittleEndian()) { 1641 const auto Word = 1642 reinterpret_cast<const support::ulittle32_t *>( 1643 Bytes.data() + Index); 1644 Data = *Word; 1645 } else { 1646 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1647 Bytes.data() + Index); 1648 Data = *Word; 1649 } 1650 outs() << "0x" << format("%08" PRIx32, Data); 1651 } else if (Index + 2 <= End) { 1652 Stride = 2; 1653 dumpBytes(Bytes.slice(Index, 2), outs()); 1654 outs() << "\t\t.short\t"; 1655 uint16_t Data = 0; 1656 if (Obj->isLittleEndian()) { 1657 const auto Short = 1658 reinterpret_cast<const support::ulittle16_t *>( 1659 Bytes.data() + Index); 1660 Data = *Short; 1661 } else { 1662 const auto Short = 1663 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1664 Index); 1665 Data = *Short; 1666 } 1667 outs() << "0x" << format("%04" PRIx16, Data); 1668 } else { 1669 Stride = 1; 1670 dumpBytes(Bytes.slice(Index, 1), outs()); 1671 outs() << "\t\t.byte\t"; 1672 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1673 } 1674 Index += Stride; 1675 outs() << "\n"; 1676 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1677 TextMappingSymsAddr.end(), Index); 1678 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1679 break; 1680 } 1681 } 1682 } 1683 1684 // If there is a data symbol inside an ELF text section and we are only 1685 // disassembling text (applicable all architectures), 1686 // we are in a situation where we must print the data and not 1687 // disassemble it. 1688 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1689 !DisassembleAll && Section.isText()) { 1690 // print out data up to 8 bytes at a time in hex and ascii 1691 uint8_t AsciiData[9] = {'\0'}; 1692 uint8_t Byte; 1693 int NumBytes = 0; 1694 1695 for (Index = Start; Index < End; Index += 1) { 1696 if (((SectionAddr + Index) < StartAddress) || 1697 ((SectionAddr + Index) > StopAddress)) 1698 continue; 1699 if (NumBytes == 0) { 1700 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1701 outs() << "\t"; 1702 } 1703 Byte = Bytes.slice(Index)[0]; 1704 outs() << format(" %02x", Byte); 1705 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1706 1707 uint8_t IndentOffset = 0; 1708 NumBytes++; 1709 if (Index == End - 1 || NumBytes > 8) { 1710 // Indent the space for less than 8 bytes data. 1711 // 2 spaces for byte and one for space between bytes 1712 IndentOffset = 3 * (8 - NumBytes); 1713 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1714 AsciiData[Excess] = '\0'; 1715 NumBytes = 8; 1716 } 1717 if (NumBytes == 8) { 1718 AsciiData[8] = '\0'; 1719 outs() << std::string(IndentOffset, ' ') << " "; 1720 outs() << reinterpret_cast<char *>(AsciiData); 1721 outs() << '\n'; 1722 NumBytes = 0; 1723 } 1724 } 1725 } 1726 if (Index >= End) 1727 break; 1728 1729 // Disassemble a real instruction or a data when disassemble all is 1730 // provided 1731 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1732 SectionAddr + Index, DebugOut, 1733 CommentStream); 1734 if (Size == 0) 1735 Size = 1; 1736 1737 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1738 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1739 *STI, &SP, &Rels); 1740 outs() << CommentStream.str(); 1741 Comments.clear(); 1742 1743 // Try to resolve the target of a call, tail call, etc. to a specific 1744 // symbol. 1745 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1746 MIA->isConditionalBranch(Inst))) { 1747 uint64_t Target; 1748 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1749 // In a relocatable object, the target's section must reside in 1750 // the same section as the call instruction or it is accessed 1751 // through a relocation. 1752 // 1753 // In a non-relocatable object, the target may be in any section. 1754 // 1755 // N.B. We don't walk the relocations in the relocatable case yet. 1756 auto *TargetSectionSymbols = &Symbols; 1757 if (!Obj->isRelocatableObject()) { 1758 auto SectionAddress = std::upper_bound( 1759 SectionAddresses.begin(), SectionAddresses.end(), Target, 1760 [](uint64_t LHS, 1761 const std::pair<uint64_t, SectionRef> &RHS) { 1762 return LHS < RHS.first; 1763 }); 1764 if (SectionAddress != SectionAddresses.begin()) { 1765 --SectionAddress; 1766 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1767 } else { 1768 TargetSectionSymbols = &AbsoluteSymbols; 1769 } 1770 } 1771 1772 // Find the first symbol in the section whose offset is less than 1773 // or equal to the target. If there isn't a section that contains 1774 // the target, find the nearest preceding absolute symbol. 1775 auto TargetSym = std::upper_bound( 1776 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1777 Target, [](uint64_t LHS, 1778 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1779 return LHS < std::get<0>(RHS); 1780 }); 1781 if (TargetSym == TargetSectionSymbols->begin()) { 1782 TargetSectionSymbols = &AbsoluteSymbols; 1783 TargetSym = std::upper_bound( 1784 AbsoluteSymbols.begin(), AbsoluteSymbols.end(), 1785 Target, [](uint64_t LHS, 1786 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1787 return LHS < std::get<0>(RHS); 1788 }); 1789 } 1790 if (TargetSym != TargetSectionSymbols->begin()) { 1791 --TargetSym; 1792 uint64_t TargetAddress = std::get<0>(*TargetSym); 1793 StringRef TargetName = std::get<1>(*TargetSym); 1794 outs() << " <" << TargetName; 1795 uint64_t Disp = Target - TargetAddress; 1796 if (Disp) 1797 outs() << "+0x" << Twine::utohexstr(Disp); 1798 outs() << '>'; 1799 } 1800 } 1801 } 1802 outs() << "\n"; 1803 1804 // Hexagon does this in pretty printer 1805 if (Obj->getArch() != Triple::hexagon) 1806 // Print relocation for instruction. 1807 while (rel_cur != rel_end) { 1808 bool hidden = getHidden(*rel_cur); 1809 uint64_t addr = rel_cur->getOffset(); 1810 SmallString<16> name; 1811 SmallString<32> val; 1812 1813 // If this relocation is hidden, skip it. 1814 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1815 ++rel_cur; 1816 continue; 1817 } 1818 1819 // Stop when rel_cur's address is past the current instruction. 1820 if (addr >= Index + Size) break; 1821 rel_cur->getTypeName(name); 1822 error(getRelocationValueString(*rel_cur, val)); 1823 outs() << format(Fmt.data(), SectionAddr + addr) << name 1824 << "\t" << val << "\n"; 1825 ++rel_cur; 1826 } 1827 } 1828 } 1829 } 1830 } 1831 1832 void llvm::PrintRelocations(const ObjectFile *Obj) { 1833 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1834 "%08" PRIx64; 1835 // Regular objdump doesn't print relocations in non-relocatable object 1836 // files. 1837 if (!Obj->isRelocatableObject()) 1838 return; 1839 1840 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1841 if (Section.relocation_begin() == Section.relocation_end()) 1842 continue; 1843 StringRef secname; 1844 error(Section.getName(secname)); 1845 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1846 for (const RelocationRef &Reloc : Section.relocations()) { 1847 bool hidden = getHidden(Reloc); 1848 uint64_t address = Reloc.getOffset(); 1849 SmallString<32> relocname; 1850 SmallString<32> valuestr; 1851 if (address < StartAddress || address > StopAddress || hidden) 1852 continue; 1853 Reloc.getTypeName(relocname); 1854 error(getRelocationValueString(Reloc, valuestr)); 1855 outs() << format(Fmt.data(), address) << " " << relocname << " " 1856 << valuestr << "\n"; 1857 } 1858 outs() << "\n"; 1859 } 1860 } 1861 1862 void llvm::PrintDynamicRelocations(const ObjectFile *Obj) { 1863 1864 // For the moment, this option is for ELF only 1865 if (!Obj->isELF()) 1866 return; 1867 1868 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1869 1870 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1871 error("not a dynamic object"); 1872 return; 1873 } 1874 1875 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1876 1877 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1878 if (DynRelSec.empty()) 1879 return; 1880 1881 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1882 for (const SectionRef &Section : DynRelSec) { 1883 if (Section.relocation_begin() == Section.relocation_end()) 1884 continue; 1885 for (const RelocationRef &Reloc : Section.relocations()) { 1886 uint64_t address = Reloc.getOffset(); 1887 SmallString<32> relocname; 1888 SmallString<32> valuestr; 1889 Reloc.getTypeName(relocname); 1890 error(getRelocationValueString(Reloc, valuestr)); 1891 outs() << format(Fmt.data(), address) << " " << relocname << " " 1892 << valuestr << "\n"; 1893 } 1894 } 1895 } 1896 1897 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1898 outs() << "Sections:\n" 1899 "Idx Name Size Address Type\n"; 1900 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1901 StringRef Name; 1902 error(Section.getName(Name)); 1903 uint64_t Address = Section.getAddress(); 1904 uint64_t Size = Section.getSize(); 1905 bool Text = Section.isText(); 1906 bool Data = Section.isData(); 1907 bool BSS = Section.isBSS(); 1908 std::string Type = (std::string(Text ? "TEXT " : "") + 1909 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1910 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1911 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1912 Address, Type.c_str()); 1913 } 1914 } 1915 1916 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1917 std::error_code EC; 1918 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1919 StringRef Name; 1920 StringRef Contents; 1921 error(Section.getName(Name)); 1922 uint64_t BaseAddr = Section.getAddress(); 1923 uint64_t Size = Section.getSize(); 1924 if (!Size) 1925 continue; 1926 1927 outs() << "Contents of section " << Name << ":\n"; 1928 if (Section.isBSS()) { 1929 outs() << format("<skipping contents of bss section at [%04" PRIx64 1930 ", %04" PRIx64 ")>\n", 1931 BaseAddr, BaseAddr + Size); 1932 continue; 1933 } 1934 1935 error(Section.getContents(Contents)); 1936 1937 // Dump out the content as hex and printable ascii characters. 1938 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1939 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1940 // Dump line of hex. 1941 for (std::size_t i = 0; i < 16; ++i) { 1942 if (i != 0 && i % 4 == 0) 1943 outs() << ' '; 1944 if (addr + i < end) 1945 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1946 << hexdigit(Contents[addr + i] & 0xF, true); 1947 else 1948 outs() << " "; 1949 } 1950 // Print ascii. 1951 outs() << " "; 1952 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1953 if (isPrint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1954 outs() << Contents[addr + i]; 1955 else 1956 outs() << "."; 1957 } 1958 outs() << "\n"; 1959 } 1960 } 1961 } 1962 1963 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1964 StringRef ArchitectureName) { 1965 outs() << "SYMBOL TABLE:\n"; 1966 1967 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1968 printCOFFSymbolTable(coff); 1969 return; 1970 } 1971 for (const SymbolRef &Symbol : o->symbols()) { 1972 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1973 if (!AddressOrError) 1974 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1975 ArchitectureName); 1976 uint64_t Address = *AddressOrError; 1977 if ((Address < StartAddress) || (Address > StopAddress)) 1978 continue; 1979 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1980 if (!TypeOrError) 1981 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1982 ArchitectureName); 1983 SymbolRef::Type Type = *TypeOrError; 1984 uint32_t Flags = Symbol.getFlags(); 1985 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1986 if (!SectionOrErr) 1987 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1988 ArchitectureName); 1989 section_iterator Section = *SectionOrErr; 1990 StringRef Name; 1991 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1992 Section->getName(Name); 1993 } else { 1994 Expected<StringRef> NameOrErr = Symbol.getName(); 1995 if (!NameOrErr) 1996 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1997 ArchitectureName); 1998 Name = *NameOrErr; 1999 } 2000 2001 bool Global = Flags & SymbolRef::SF_Global; 2002 bool Weak = Flags & SymbolRef::SF_Weak; 2003 bool Absolute = Flags & SymbolRef::SF_Absolute; 2004 bool Common = Flags & SymbolRef::SF_Common; 2005 bool Hidden = Flags & SymbolRef::SF_Hidden; 2006 2007 char GlobLoc = ' '; 2008 if (Type != SymbolRef::ST_Unknown) 2009 GlobLoc = Global ? 'g' : 'l'; 2010 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2011 ? 'd' : ' '; 2012 char FileFunc = ' '; 2013 if (Type == SymbolRef::ST_File) 2014 FileFunc = 'f'; 2015 else if (Type == SymbolRef::ST_Function) 2016 FileFunc = 'F'; 2017 else if (Type == SymbolRef::ST_Data) 2018 FileFunc = 'O'; 2019 2020 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 2021 "%08" PRIx64; 2022 2023 outs() << format(Fmt, Address) << " " 2024 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2025 << (Weak ? 'w' : ' ') // Weak? 2026 << ' ' // Constructor. Not supported yet. 2027 << ' ' // Warning. Not supported yet. 2028 << ' ' // Indirect reference to another symbol. 2029 << Debug // Debugging (d) or dynamic (D) symbol. 2030 << FileFunc // Name of function (F), file (f) or object (O). 2031 << ' '; 2032 if (Absolute) { 2033 outs() << "*ABS*"; 2034 } else if (Common) { 2035 outs() << "*COM*"; 2036 } else if (Section == o->section_end()) { 2037 outs() << "*UND*"; 2038 } else { 2039 if (const MachOObjectFile *MachO = 2040 dyn_cast<const MachOObjectFile>(o)) { 2041 DataRefImpl DR = Section->getRawDataRefImpl(); 2042 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 2043 outs() << SegmentName << ","; 2044 } 2045 StringRef SectionName; 2046 error(Section->getName(SectionName)); 2047 outs() << SectionName; 2048 } 2049 2050 outs() << '\t'; 2051 if (Common || isa<ELFObjectFileBase>(o)) { 2052 uint64_t Val = 2053 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2054 outs() << format("\t %08" PRIx64 " ", Val); 2055 } 2056 2057 if (Hidden) { 2058 outs() << ".hidden "; 2059 } 2060 outs() << Name 2061 << '\n'; 2062 } 2063 } 2064 2065 static void PrintUnwindInfo(const ObjectFile *o) { 2066 outs() << "Unwind info:\n\n"; 2067 2068 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 2069 printCOFFUnwindInfo(coff); 2070 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2071 printMachOUnwindInfo(MachO); 2072 else { 2073 // TODO: Extract DWARF dump tool to objdump. 2074 errs() << "This operation is only currently supported " 2075 "for COFF and MachO object files.\n"; 2076 return; 2077 } 2078 } 2079 2080 void llvm::printExportsTrie(const ObjectFile *o) { 2081 outs() << "Exports trie:\n"; 2082 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2083 printMachOExportsTrie(MachO); 2084 else { 2085 errs() << "This operation is only currently supported " 2086 "for Mach-O executable files.\n"; 2087 return; 2088 } 2089 } 2090 2091 void llvm::printRebaseTable(ObjectFile *o) { 2092 outs() << "Rebase table:\n"; 2093 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2094 printMachORebaseTable(MachO); 2095 else { 2096 errs() << "This operation is only currently supported " 2097 "for Mach-O executable files.\n"; 2098 return; 2099 } 2100 } 2101 2102 void llvm::printBindTable(ObjectFile *o) { 2103 outs() << "Bind table:\n"; 2104 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2105 printMachOBindTable(MachO); 2106 else { 2107 errs() << "This operation is only currently supported " 2108 "for Mach-O executable files.\n"; 2109 return; 2110 } 2111 } 2112 2113 void llvm::printLazyBindTable(ObjectFile *o) { 2114 outs() << "Lazy bind table:\n"; 2115 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2116 printMachOLazyBindTable(MachO); 2117 else { 2118 errs() << "This operation is only currently supported " 2119 "for Mach-O executable files.\n"; 2120 return; 2121 } 2122 } 2123 2124 void llvm::printWeakBindTable(ObjectFile *o) { 2125 outs() << "Weak bind table:\n"; 2126 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2127 printMachOWeakBindTable(MachO); 2128 else { 2129 errs() << "This operation is only currently supported " 2130 "for Mach-O executable files.\n"; 2131 return; 2132 } 2133 } 2134 2135 /// Dump the raw contents of the __clangast section so the output can be piped 2136 /// into llvm-bcanalyzer. 2137 void llvm::printRawClangAST(const ObjectFile *Obj) { 2138 if (outs().is_displayed()) { 2139 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 2140 "the clang ast section.\n" 2141 "Please redirect the output to a file or another program such as " 2142 "llvm-bcanalyzer.\n"; 2143 return; 2144 } 2145 2146 StringRef ClangASTSectionName("__clangast"); 2147 if (isa<COFFObjectFile>(Obj)) { 2148 ClangASTSectionName = "clangast"; 2149 } 2150 2151 Optional<object::SectionRef> ClangASTSection; 2152 for (auto Sec : ToolSectionFilter(*Obj)) { 2153 StringRef Name; 2154 Sec.getName(Name); 2155 if (Name == ClangASTSectionName) { 2156 ClangASTSection = Sec; 2157 break; 2158 } 2159 } 2160 if (!ClangASTSection) 2161 return; 2162 2163 StringRef ClangASTContents; 2164 error(ClangASTSection.getValue().getContents(ClangASTContents)); 2165 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2166 } 2167 2168 static void printFaultMaps(const ObjectFile *Obj) { 2169 const char *FaultMapSectionName = nullptr; 2170 2171 if (isa<ELFObjectFileBase>(Obj)) { 2172 FaultMapSectionName = ".llvm_faultmaps"; 2173 } else if (isa<MachOObjectFile>(Obj)) { 2174 FaultMapSectionName = "__llvm_faultmaps"; 2175 } else { 2176 errs() << "This operation is only currently supported " 2177 "for ELF and Mach-O executable files.\n"; 2178 return; 2179 } 2180 2181 Optional<object::SectionRef> FaultMapSection; 2182 2183 for (auto Sec : ToolSectionFilter(*Obj)) { 2184 StringRef Name; 2185 Sec.getName(Name); 2186 if (Name == FaultMapSectionName) { 2187 FaultMapSection = Sec; 2188 break; 2189 } 2190 } 2191 2192 outs() << "FaultMap table:\n"; 2193 2194 if (!FaultMapSection.hasValue()) { 2195 outs() << "<not found>\n"; 2196 return; 2197 } 2198 2199 StringRef FaultMapContents; 2200 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2201 2202 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2203 FaultMapContents.bytes_end()); 2204 2205 outs() << FMP; 2206 } 2207 2208 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2209 if (o->isELF()) { 2210 printELFFileHeader(o); 2211 return printELFDynamicSection(o); 2212 } 2213 if (o->isCOFF()) 2214 return printCOFFFileHeader(o); 2215 if (o->isWasm()) 2216 return printWasmFileHeader(o); 2217 if (o->isMachO()) { 2218 printMachOFileHeader(o); 2219 if (!onlyFirst) 2220 printMachOLoadCommands(o); 2221 return; 2222 } 2223 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2224 } 2225 2226 static void printFileHeaders(const ObjectFile *o) { 2227 if (!o->isELF() && !o->isCOFF()) 2228 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2229 2230 Triple::ArchType AT = o->getArch(); 2231 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2232 Expected<uint64_t> StartAddrOrErr = o->getStartAddress(); 2233 if (!StartAddrOrErr) 2234 report_error(o->getFileName(), StartAddrOrErr.takeError()); 2235 2236 StringRef Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2237 uint64_t Address = StartAddrOrErr.get(); 2238 outs() << "start address: " 2239 << "0x" << format(Fmt.data(), Address) 2240 << "\n"; 2241 } 2242 2243 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2244 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2245 if (!ModeOrErr) { 2246 errs() << "ill-formed archive entry.\n"; 2247 consumeError(ModeOrErr.takeError()); 2248 return; 2249 } 2250 sys::fs::perms Mode = ModeOrErr.get(); 2251 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2252 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2253 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2254 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2255 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2256 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2257 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2258 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2259 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2260 2261 outs() << " "; 2262 2263 Expected<unsigned> UIDOrErr = C.getUID(); 2264 if (!UIDOrErr) 2265 report_error(Filename, UIDOrErr.takeError()); 2266 unsigned UID = UIDOrErr.get(); 2267 outs() << format("%d/", UID); 2268 2269 Expected<unsigned> GIDOrErr = C.getGID(); 2270 if (!GIDOrErr) 2271 report_error(Filename, GIDOrErr.takeError()); 2272 unsigned GID = GIDOrErr.get(); 2273 outs() << format("%-d ", GID); 2274 2275 Expected<uint64_t> Size = C.getRawSize(); 2276 if (!Size) 2277 report_error(Filename, Size.takeError()); 2278 outs() << format("%6" PRId64, Size.get()) << " "; 2279 2280 StringRef RawLastModified = C.getRawLastModified(); 2281 unsigned Seconds; 2282 if (RawLastModified.getAsInteger(10, Seconds)) 2283 outs() << "(date: \"" << RawLastModified 2284 << "\" contains non-decimal chars) "; 2285 else { 2286 // Since ctime(3) returns a 26 character string of the form: 2287 // "Sun Sep 16 01:03:52 1973\n\0" 2288 // just print 24 characters. 2289 time_t t = Seconds; 2290 outs() << format("%.24s ", ctime(&t)); 2291 } 2292 2293 StringRef Name = ""; 2294 Expected<StringRef> NameOrErr = C.getName(); 2295 if (!NameOrErr) { 2296 consumeError(NameOrErr.takeError()); 2297 Expected<StringRef> RawNameOrErr = C.getRawName(); 2298 if (!RawNameOrErr) 2299 report_error(Filename, NameOrErr.takeError()); 2300 Name = RawNameOrErr.get(); 2301 } else { 2302 Name = NameOrErr.get(); 2303 } 2304 outs() << Name << "\n"; 2305 } 2306 2307 static void DumpObject(ObjectFile *o, const Archive *a = nullptr, 2308 const Archive::Child *c = nullptr) { 2309 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2310 // Avoid other output when using a raw option. 2311 if (!RawClangAST) { 2312 outs() << '\n'; 2313 if (a) 2314 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2315 else 2316 outs() << o->getFileName(); 2317 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2318 } 2319 2320 if (ArchiveHeaders && !MachOOpt && c) 2321 printArchiveChild(ArchiveName, *c); 2322 if (Disassemble) 2323 DisassembleObject(o, Relocations); 2324 if (Relocations && !Disassemble) 2325 PrintRelocations(o); 2326 if (DynamicRelocations) 2327 PrintDynamicRelocations(o); 2328 if (SectionHeaders) 2329 PrintSectionHeaders(o); 2330 if (SectionContents) 2331 PrintSectionContents(o); 2332 if (SymbolTable) 2333 PrintSymbolTable(o, ArchiveName); 2334 if (UnwindInfo) 2335 PrintUnwindInfo(o); 2336 if (PrivateHeaders || FirstPrivateHeader) 2337 printPrivateFileHeaders(o, FirstPrivateHeader); 2338 if (FileHeaders) 2339 printFileHeaders(o); 2340 if (ExportsTrie) 2341 printExportsTrie(o); 2342 if (Rebase) 2343 printRebaseTable(o); 2344 if (Bind) 2345 printBindTable(o); 2346 if (LazyBind) 2347 printLazyBindTable(o); 2348 if (WeakBind) 2349 printWeakBindTable(o); 2350 if (RawClangAST) 2351 printRawClangAST(o); 2352 if (PrintFaultMaps) 2353 printFaultMaps(o); 2354 if (DwarfDumpType != DIDT_Null) { 2355 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*o); 2356 // Dump the complete DWARF structure. 2357 DIDumpOptions DumpOpts; 2358 DumpOpts.DumpType = DwarfDumpType; 2359 DICtx->dump(outs(), DumpOpts); 2360 } 2361 } 2362 2363 static void DumpObject(const COFFImportFile *I, const Archive *A, 2364 const Archive::Child *C = nullptr) { 2365 StringRef ArchiveName = A ? A->getFileName() : ""; 2366 2367 // Avoid other output when using a raw option. 2368 if (!RawClangAST) 2369 outs() << '\n' 2370 << ArchiveName << "(" << I->getFileName() << ")" 2371 << ":\tfile format COFF-import-file" 2372 << "\n\n"; 2373 2374 if (ArchiveHeaders && !MachOOpt && C) 2375 printArchiveChild(ArchiveName, *C); 2376 if (SymbolTable) 2377 printCOFFSymbolTable(I); 2378 } 2379 2380 /// Dump each object file in \a a; 2381 static void DumpArchive(const Archive *a) { 2382 Error Err = Error::success(); 2383 for (auto &C : a->children(Err)) { 2384 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2385 if (!ChildOrErr) { 2386 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2387 report_error(a->getFileName(), C, std::move(E)); 2388 continue; 2389 } 2390 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2391 DumpObject(o, a, &C); 2392 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2393 DumpObject(I, a, &C); 2394 else 2395 report_error(a->getFileName(), object_error::invalid_file_type); 2396 } 2397 if (Err) 2398 report_error(a->getFileName(), std::move(Err)); 2399 } 2400 2401 /// Open file and figure out how to dump it. 2402 static void DumpInput(StringRef file) { 2403 2404 // If we are using the Mach-O specific object file parser, then let it parse 2405 // the file and process the command line options. So the -arch flags can 2406 // be used to select specific slices, etc. 2407 if (MachOOpt) { 2408 ParseInputMachO(file); 2409 return; 2410 } 2411 2412 // Attempt to open the binary. 2413 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2414 if (!BinaryOrErr) 2415 report_error(file, BinaryOrErr.takeError()); 2416 Binary &Binary = *BinaryOrErr.get().getBinary(); 2417 2418 if (Archive *a = dyn_cast<Archive>(&Binary)) 2419 DumpArchive(a); 2420 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2421 DumpObject(o); 2422 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2423 ParseInputMachO(UB); 2424 else 2425 report_error(file, object_error::invalid_file_type); 2426 } 2427 2428 int main(int argc, char **argv) { 2429 InitLLVM X(argc, argv); 2430 2431 // Initialize targets and assembly printers/parsers. 2432 llvm::InitializeAllTargetInfos(); 2433 llvm::InitializeAllTargetMCs(); 2434 llvm::InitializeAllDisassemblers(); 2435 2436 // Register the target printer for --version. 2437 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2438 2439 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2440 2441 ToolName = argv[0]; 2442 2443 // Defaults to a.out if no filenames specified. 2444 if (InputFilenames.size() == 0) 2445 InputFilenames.push_back("a.out"); 2446 2447 if (AllHeaders) 2448 PrivateHeaders = Relocations = SectionHeaders = SymbolTable = true; 2449 2450 if (DisassembleAll || PrintSource || PrintLines) 2451 Disassemble = true; 2452 2453 if (!Disassemble 2454 && !Relocations 2455 && !DynamicRelocations 2456 && !SectionHeaders 2457 && !SectionContents 2458 && !SymbolTable 2459 && !UnwindInfo 2460 && !PrivateHeaders 2461 && !FileHeaders 2462 && !FirstPrivateHeader 2463 && !ExportsTrie 2464 && !Rebase 2465 && !Bind 2466 && !LazyBind 2467 && !WeakBind 2468 && !RawClangAST 2469 && !(UniversalHeaders && MachOOpt) 2470 && !ArchiveHeaders 2471 && !(IndirectSymbols && MachOOpt) 2472 && !(DataInCode && MachOOpt) 2473 && !(LinkOptHints && MachOOpt) 2474 && !(InfoPlist && MachOOpt) 2475 && !(DylibsUsed && MachOOpt) 2476 && !(DylibId && MachOOpt) 2477 && !(ObjcMetaData && MachOOpt) 2478 && !(FilterSections.size() != 0 && MachOOpt) 2479 && !PrintFaultMaps 2480 && DwarfDumpType == DIDT_Null) { 2481 cl::PrintHelpMessage(); 2482 return 2; 2483 } 2484 2485 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2486 DisassembleFunctions.end()); 2487 2488 llvm::for_each(InputFilenames, DumpInput); 2489 2490 return EXIT_SUCCESS; 2491 } 2492