1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // This program is a utility that works like binutils "objdump", that is, it 11 // dumps out a plethora of information about an object file depending on the 12 // flags. 13 // 14 // The flags and output of this program should be near identical to those of 15 // binutils objdump. 16 // 17 //===----------------------------------------------------------------------===// 18 19 #include "llvm-objdump.h" 20 #include "llvm/ADT/Optional.h" 21 #include "llvm/ADT/STLExtras.h" 22 #include "llvm/ADT/StringExtras.h" 23 #include "llvm/ADT/StringSet.h" 24 #include "llvm/ADT/Triple.h" 25 #include "llvm/CodeGen/FaultMaps.h" 26 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 27 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 28 #include "llvm/Demangle/Demangle.h" 29 #include "llvm/MC/MCAsmInfo.h" 30 #include "llvm/MC/MCContext.h" 31 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 32 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 33 #include "llvm/MC/MCInst.h" 34 #include "llvm/MC/MCInstPrinter.h" 35 #include "llvm/MC/MCInstrAnalysis.h" 36 #include "llvm/MC/MCInstrInfo.h" 37 #include "llvm/MC/MCObjectFileInfo.h" 38 #include "llvm/MC/MCRegisterInfo.h" 39 #include "llvm/MC/MCSubtargetInfo.h" 40 #include "llvm/Object/Archive.h" 41 #include "llvm/Object/COFF.h" 42 #include "llvm/Object/COFFImportFile.h" 43 #include "llvm/Object/ELFObjectFile.h" 44 #include "llvm/Object/MachO.h" 45 #include "llvm/Object/MachOUniversal.h" 46 #include "llvm/Object/ObjectFile.h" 47 #include "llvm/Object/Wasm.h" 48 #include "llvm/Support/Casting.h" 49 #include "llvm/Support/CommandLine.h" 50 #include "llvm/Support/Debug.h" 51 #include "llvm/Support/Errc.h" 52 #include "llvm/Support/FileSystem.h" 53 #include "llvm/Support/Format.h" 54 #include "llvm/Support/GraphWriter.h" 55 #include "llvm/Support/Host.h" 56 #include "llvm/Support/InitLLVM.h" 57 #include "llvm/Support/MemoryBuffer.h" 58 #include "llvm/Support/SourceMgr.h" 59 #include "llvm/Support/StringSaver.h" 60 #include "llvm/Support/TargetRegistry.h" 61 #include "llvm/Support/TargetSelect.h" 62 #include "llvm/Support/WithColor.h" 63 #include "llvm/Support/raw_ostream.h" 64 #include <algorithm> 65 #include <cctype> 66 #include <cstring> 67 #include <system_error> 68 #include <unordered_map> 69 #include <utility> 70 71 using namespace llvm; 72 using namespace object; 73 74 cl::opt<bool> 75 llvm::AllHeaders("all-headers", 76 cl::desc("Display all available header information")); 77 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 78 cl::aliasopt(AllHeaders)); 79 80 static cl::list<std::string> 81 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 82 83 cl::opt<bool> 84 llvm::Disassemble("disassemble", 85 cl::desc("Display assembler mnemonics for the machine instructions")); 86 static cl::alias 87 Disassembled("d", cl::desc("Alias for --disassemble"), 88 cl::aliasopt(Disassemble)); 89 90 cl::opt<bool> 91 llvm::DisassembleAll("disassemble-all", 92 cl::desc("Display assembler mnemonics for the machine instructions")); 93 static cl::alias 94 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 95 cl::aliasopt(DisassembleAll)); 96 97 cl::opt<bool> llvm::Demangle("demangle", cl::desc("Demangle symbols names"), 98 cl::init(false)); 99 100 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 101 cl::aliasopt(llvm::Demangle)); 102 103 static cl::list<std::string> 104 DisassembleFunctions("df", 105 cl::CommaSeparated, 106 cl::desc("List of functions to disassemble")); 107 static StringSet<> DisasmFuncsSet; 108 109 cl::opt<bool> 110 llvm::Relocations("reloc", 111 cl::desc("Display the relocation entries in the file")); 112 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 113 cl::NotHidden, 114 cl::aliasopt(llvm::Relocations)); 115 116 cl::opt<bool> 117 llvm::DynamicRelocations("dynamic-reloc", 118 cl::desc("Display the dynamic relocation entries in the file")); 119 static cl::alias 120 DynamicRelocationsd("R", cl::desc("Alias for --dynamic-reloc"), 121 cl::aliasopt(DynamicRelocations)); 122 123 cl::opt<bool> 124 llvm::SectionContents("full-contents", 125 cl::desc("Display the content of each section")); 126 static cl::alias SectionContentsShort("s", 127 cl::desc("Alias for --full-contents"), 128 cl::aliasopt(SectionContents)); 129 130 cl::opt<bool> llvm::SymbolTable("syms", cl::desc("Display the symbol table")); 131 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 132 cl::NotHidden, 133 cl::aliasopt(llvm::SymbolTable)); 134 135 cl::opt<bool> 136 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 137 138 cl::opt<bool> 139 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 140 141 cl::opt<bool> 142 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 143 144 cl::opt<bool> 145 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 146 147 cl::opt<bool> 148 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 149 150 cl::opt<bool> 151 llvm::RawClangAST("raw-clang-ast", 152 cl::desc("Dump the raw binary contents of the clang AST section")); 153 154 static cl::opt<bool> 155 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 156 static cl::alias 157 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 158 159 cl::opt<std::string> 160 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 161 "see -version for available targets")); 162 163 cl::opt<std::string> 164 llvm::MCPU("mcpu", 165 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 166 cl::value_desc("cpu-name"), 167 cl::init("")); 168 169 cl::opt<std::string> 170 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 171 "see -version for available targets")); 172 173 cl::opt<bool> 174 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 175 "headers for each section.")); 176 static cl::alias 177 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 178 cl::aliasopt(SectionHeaders)); 179 static cl::alias 180 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 181 cl::aliasopt(SectionHeaders)); 182 183 cl::list<std::string> 184 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 185 "With -macho dump segment,section")); 186 cl::alias 187 static FilterSectionsj("j", cl::desc("Alias for --section"), 188 cl::aliasopt(llvm::FilterSections)); 189 190 cl::list<std::string> 191 llvm::MAttrs("mattr", 192 cl::CommaSeparated, 193 cl::desc("Target specific attributes"), 194 cl::value_desc("a1,+a2,-a3,...")); 195 196 cl::opt<bool> 197 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 198 "instructions, do not print " 199 "the instruction bytes.")); 200 cl::opt<bool> 201 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 202 203 cl::opt<bool> 204 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 205 206 static cl::alias 207 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 208 cl::aliasopt(UnwindInfo)); 209 210 cl::opt<bool> 211 llvm::PrivateHeaders("private-headers", 212 cl::desc("Display format specific file headers")); 213 214 cl::opt<bool> 215 llvm::FirstPrivateHeader("private-header", 216 cl::desc("Display only the first format specific file " 217 "header")); 218 219 static cl::alias 220 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 221 cl::aliasopt(PrivateHeaders)); 222 223 cl::opt<bool> llvm::FileHeaders( 224 "file-headers", 225 cl::desc("Display the contents of the overall file header")); 226 227 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 228 cl::aliasopt(FileHeaders)); 229 230 cl::opt<bool> 231 llvm::ArchiveHeaders("archive-headers", 232 cl::desc("Display archive header information")); 233 234 cl::alias 235 ArchiveHeadersShort("a", cl::desc("Alias for --archive-headers"), 236 cl::aliasopt(ArchiveHeaders)); 237 238 cl::opt<bool> 239 llvm::PrintImmHex("print-imm-hex", 240 cl::desc("Use hex format for immediate values")); 241 242 cl::opt<bool> PrintFaultMaps("fault-map-section", 243 cl::desc("Display contents of faultmap section")); 244 245 cl::opt<DIDumpType> llvm::DwarfDumpType( 246 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 247 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 248 249 cl::opt<bool> PrintSource( 250 "source", 251 cl::desc( 252 "Display source inlined with disassembly. Implies disassemble object")); 253 254 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 255 cl::aliasopt(PrintSource)); 256 257 cl::opt<bool> PrintLines("line-numbers", 258 cl::desc("Display source line numbers with " 259 "disassembly. Implies disassemble object")); 260 261 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 262 cl::aliasopt(PrintLines)); 263 264 cl::opt<unsigned long long> 265 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 266 cl::value_desc("address"), cl::init(0)); 267 cl::opt<unsigned long long> 268 StopAddress("stop-address", 269 cl::desc("Do not skip blocks of zeroes when disassembling"), 270 cl::value_desc("address"), cl::init(UINT64_MAX)); 271 272 cl::opt<bool> DisassembleZeroes("disassemble-zeroes", 273 cl::desc("Do not skip blocks of zeroes when " 274 "disassembling the blocks of zeroes")); 275 cl::alias DisassembleZeroesShort("z", 276 cl::desc("Alias for --disassemble-zeroes"), 277 cl::aliasopt(DisassembleZeroes)); 278 279 static StringRef ToolName; 280 281 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 282 283 namespace { 284 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 285 286 class SectionFilterIterator { 287 public: 288 SectionFilterIterator(FilterPredicate P, 289 llvm::object::section_iterator const &I, 290 llvm::object::section_iterator const &E) 291 : Predicate(std::move(P)), Iterator(I), End(E) { 292 ScanPredicate(); 293 } 294 const llvm::object::SectionRef &operator*() const { return *Iterator; } 295 SectionFilterIterator &operator++() { 296 ++Iterator; 297 ScanPredicate(); 298 return *this; 299 } 300 bool operator!=(SectionFilterIterator const &Other) const { 301 return Iterator != Other.Iterator; 302 } 303 304 private: 305 void ScanPredicate() { 306 while (Iterator != End && !Predicate(*Iterator)) { 307 ++Iterator; 308 } 309 } 310 FilterPredicate Predicate; 311 llvm::object::section_iterator Iterator; 312 llvm::object::section_iterator End; 313 }; 314 315 class SectionFilter { 316 public: 317 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 318 : Predicate(std::move(P)), Object(O) {} 319 SectionFilterIterator begin() { 320 return SectionFilterIterator(Predicate, Object.section_begin(), 321 Object.section_end()); 322 } 323 SectionFilterIterator end() { 324 return SectionFilterIterator(Predicate, Object.section_end(), 325 Object.section_end()); 326 } 327 328 private: 329 FilterPredicate Predicate; 330 llvm::object::ObjectFile const &Object; 331 }; 332 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 333 return SectionFilter( 334 [](llvm::object::SectionRef const &S) { 335 if (FilterSections.empty()) 336 return true; 337 llvm::StringRef String; 338 std::error_code error = S.getName(String); 339 if (error) 340 return false; 341 return is_contained(FilterSections, String); 342 }, 343 O); 344 } 345 } 346 347 void llvm::error(std::error_code EC) { 348 if (!EC) 349 return; 350 WithColor::error(errs(), ToolName) 351 << "reading file: " << EC.message() << ".\n"; 352 errs().flush(); 353 exit(1); 354 } 355 356 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 357 WithColor::error(errs(), ToolName) << Message << ".\n"; 358 errs().flush(); 359 exit(1); 360 } 361 362 void llvm::warn(StringRef Message) { 363 WithColor::warning(errs(), ToolName) << Message << ".\n"; 364 errs().flush(); 365 } 366 367 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 368 Twine Message) { 369 WithColor::error(errs(), ToolName) 370 << "'" << File << "': " << Message << ".\n"; 371 exit(1); 372 } 373 374 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 375 std::error_code EC) { 376 assert(EC); 377 WithColor::error(errs(), ToolName) 378 << "'" << File << "': " << EC.message() << ".\n"; 379 exit(1); 380 } 381 382 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 383 llvm::Error E) { 384 assert(E); 385 std::string Buf; 386 raw_string_ostream OS(Buf); 387 logAllUnhandledErrors(std::move(E), OS); 388 OS.flush(); 389 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 390 exit(1); 391 } 392 393 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 394 StringRef FileName, 395 llvm::Error E, 396 StringRef ArchitectureName) { 397 assert(E); 398 WithColor::error(errs(), ToolName); 399 if (ArchiveName != "") 400 errs() << ArchiveName << "(" << FileName << ")"; 401 else 402 errs() << "'" << FileName << "'"; 403 if (!ArchitectureName.empty()) 404 errs() << " (for architecture " << ArchitectureName << ")"; 405 std::string Buf; 406 raw_string_ostream OS(Buf); 407 logAllUnhandledErrors(std::move(E), OS); 408 OS.flush(); 409 errs() << ": " << Buf; 410 exit(1); 411 } 412 413 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 414 const object::Archive::Child &C, 415 llvm::Error E, 416 StringRef ArchitectureName) { 417 Expected<StringRef> NameOrErr = C.getName(); 418 // TODO: if we have a error getting the name then it would be nice to print 419 // the index of which archive member this is and or its offset in the 420 // archive instead of "???" as the name. 421 if (!NameOrErr) { 422 consumeError(NameOrErr.takeError()); 423 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 424 } else 425 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 426 ArchitectureName); 427 } 428 429 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 430 // Figure out the target triple. 431 llvm::Triple TheTriple("unknown-unknown-unknown"); 432 if (TripleName.empty()) { 433 if (Obj) 434 TheTriple = Obj->makeTriple(); 435 } else { 436 TheTriple.setTriple(Triple::normalize(TripleName)); 437 438 // Use the triple, but also try to combine with ARM build attributes. 439 if (Obj) { 440 auto Arch = Obj->getArch(); 441 if (Arch == Triple::arm || Arch == Triple::armeb) 442 Obj->setARMSubArch(TheTriple); 443 } 444 } 445 446 // Get the target specific parser. 447 std::string Error; 448 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 449 Error); 450 if (!TheTarget) { 451 if (Obj) 452 report_error(Obj->getFileName(), "can't find target: " + Error); 453 else 454 error("can't find target: " + Error); 455 } 456 457 // Update the triple name and return the found target. 458 TripleName = TheTriple.getTriple(); 459 return TheTarget; 460 } 461 462 bool llvm::isRelocAddressLess(RelocationRef A, RelocationRef B) { 463 return A.getOffset() < B.getOffset(); 464 } 465 466 static std::string demangle(StringRef Name) { 467 char *Demangled = nullptr; 468 if (Name.startswith("_Z")) 469 Demangled = itaniumDemangle(Name.data(), Demangled, nullptr, nullptr); 470 else if (Name.startswith("?")) 471 Demangled = microsoftDemangle(Name.data(), Demangled, nullptr, nullptr); 472 473 if (!Demangled) 474 return Name; 475 476 std::string Ret = Demangled; 477 free(Demangled); 478 return Ret; 479 } 480 481 template <class ELFT> 482 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 483 const RelocationRef &RelRef, 484 SmallVectorImpl<char> &Result) { 485 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 486 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 487 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 488 489 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 490 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 491 auto SecOrErr = EF.getSection(Rel.d.a); 492 if (!SecOrErr) 493 return errorToErrorCode(SecOrErr.takeError()); 494 const Elf_Shdr *Sec = *SecOrErr; 495 auto SymTabOrErr = EF.getSection(Sec->sh_link); 496 if (!SymTabOrErr) 497 return errorToErrorCode(SymTabOrErr.takeError()); 498 const Elf_Shdr *SymTab = *SymTabOrErr; 499 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 500 SymTab->sh_type == ELF::SHT_DYNSYM); 501 auto StrTabSec = EF.getSection(SymTab->sh_link); 502 if (!StrTabSec) 503 return errorToErrorCode(StrTabSec.takeError()); 504 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 505 if (!StrTabOrErr) 506 return errorToErrorCode(StrTabOrErr.takeError()); 507 StringRef StrTab = *StrTabOrErr; 508 int64_t Addend = 0; 509 // If there is no Symbol associated with the relocation, we set the undef 510 // boolean value to 'true'. This will prevent us from calling functions that 511 // requires the relocation to be associated with a symbol. 512 bool Undef = false; 513 switch (Sec->sh_type) { 514 default: 515 return object_error::parse_failed; 516 case ELF::SHT_REL: { 517 // TODO: Read implicit addend from section data. 518 break; 519 } 520 case ELF::SHT_RELA: { 521 const Elf_Rela *ERela = Obj->getRela(Rel); 522 Addend = ERela->r_addend; 523 Undef = ERela->getSymbol(false) == 0; 524 break; 525 } 526 } 527 std::string Target; 528 if (!Undef) { 529 symbol_iterator SI = RelRef.getSymbol(); 530 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 531 if (symb->getType() == ELF::STT_SECTION) { 532 Expected<section_iterator> SymSI = SI->getSection(); 533 if (!SymSI) 534 return errorToErrorCode(SymSI.takeError()); 535 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 536 auto SecName = EF.getSectionName(SymSec); 537 if (!SecName) 538 return errorToErrorCode(SecName.takeError()); 539 Target = *SecName; 540 } else { 541 Expected<StringRef> SymName = symb->getName(StrTab); 542 if (!SymName) 543 return errorToErrorCode(SymName.takeError()); 544 if (Demangle) 545 Target = demangle(*SymName); 546 else 547 Target = *SymName; 548 } 549 } else 550 Target = "*ABS*"; 551 552 // Default scheme is to print Target, as well as "+ <addend>" for nonzero 553 // addend. Should be acceptable for all normal purposes. 554 std::string FmtBuf; 555 raw_string_ostream Fmt(FmtBuf); 556 Fmt << Target; 557 if (Addend != 0) 558 Fmt << (Addend < 0 ? "" : "+") << Addend; 559 Fmt.flush(); 560 Result.append(FmtBuf.begin(), FmtBuf.end()); 561 return std::error_code(); 562 } 563 564 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 565 const RelocationRef &Rel, 566 SmallVectorImpl<char> &Result) { 567 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 568 return getRelocationValueString(ELF32LE, Rel, Result); 569 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 570 return getRelocationValueString(ELF64LE, Rel, Result); 571 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 572 return getRelocationValueString(ELF32BE, Rel, Result); 573 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 574 return getRelocationValueString(ELF64BE, Rel, Result); 575 } 576 577 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 578 const RelocationRef &Rel, 579 SmallVectorImpl<char> &Result) { 580 symbol_iterator SymI = Rel.getSymbol(); 581 Expected<StringRef> SymNameOrErr = SymI->getName(); 582 if (!SymNameOrErr) 583 return errorToErrorCode(SymNameOrErr.takeError()); 584 StringRef SymName = *SymNameOrErr; 585 Result.append(SymName.begin(), SymName.end()); 586 return std::error_code(); 587 } 588 589 static void printRelocationTargetName(const MachOObjectFile *O, 590 const MachO::any_relocation_info &RE, 591 raw_string_ostream &Fmt) { 592 // Target of a scattered relocation is an address. In the interest of 593 // generating pretty output, scan through the symbol table looking for a 594 // symbol that aligns with that address. If we find one, print it. 595 // Otherwise, we just print the hex address of the target. 596 if (O->isRelocationScattered(RE)) { 597 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 598 599 for (const SymbolRef &Symbol : O->symbols()) { 600 Expected<uint64_t> Addr = Symbol.getAddress(); 601 if (!Addr) 602 report_error(O->getFileName(), Addr.takeError()); 603 if (*Addr != Val) 604 continue; 605 Expected<StringRef> Name = Symbol.getName(); 606 if (!Name) 607 report_error(O->getFileName(), Name.takeError()); 608 Fmt << *Name; 609 return; 610 } 611 612 // If we couldn't find a symbol that this relocation refers to, try 613 // to find a section beginning instead. 614 for (const SectionRef &Section : ToolSectionFilter(*O)) { 615 std::error_code ec; 616 617 StringRef Name; 618 uint64_t Addr = Section.getAddress(); 619 if (Addr != Val) 620 continue; 621 if ((ec = Section.getName(Name))) 622 report_error(O->getFileName(), ec); 623 Fmt << Name; 624 return; 625 } 626 627 Fmt << format("0x%x", Val); 628 return; 629 } 630 631 StringRef S; 632 bool isExtern = O->getPlainRelocationExternal(RE); 633 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 634 635 if (O->getAnyRelocationType(RE) == MachO::ARM64_RELOC_ADDEND) { 636 Fmt << format("0x%0" PRIx64, Val); 637 return; 638 } 639 640 if (isExtern) { 641 symbol_iterator SI = O->symbol_begin(); 642 advance(SI, Val); 643 Expected<StringRef> SOrErr = SI->getName(); 644 if (!SOrErr) 645 report_error(O->getFileName(), SOrErr.takeError()); 646 S = *SOrErr; 647 } else { 648 section_iterator SI = O->section_begin(); 649 // Adjust for the fact that sections are 1-indexed. 650 if (Val == 0) { 651 Fmt << "0 (?,?)"; 652 return; 653 } 654 uint32_t I = Val - 1; 655 while (I != 0 && SI != O->section_end()) { 656 --I; 657 advance(SI, 1); 658 } 659 if (SI == O->section_end()) 660 Fmt << Val << " (?,?)"; 661 else 662 SI->getName(S); 663 } 664 665 Fmt << S; 666 } 667 668 static std::error_code getRelocationValueString(const WasmObjectFile *Obj, 669 const RelocationRef &RelRef, 670 SmallVectorImpl<char> &Result) { 671 const wasm::WasmRelocation& Rel = Obj->getWasmRelocation(RelRef); 672 symbol_iterator SI = RelRef.getSymbol(); 673 std::string FmtBuf; 674 raw_string_ostream Fmt(FmtBuf); 675 if (SI == Obj->symbol_end()) { 676 // Not all wasm relocations have symbols associated with them. 677 // In particular R_WEBASSEMBLY_TYPE_INDEX_LEB. 678 Fmt << Rel.Index; 679 } else { 680 Expected<StringRef> SymNameOrErr = SI->getName(); 681 if (!SymNameOrErr) 682 return errorToErrorCode(SymNameOrErr.takeError()); 683 StringRef SymName = *SymNameOrErr; 684 Result.append(SymName.begin(), SymName.end()); 685 } 686 Fmt << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 687 Fmt.flush(); 688 Result.append(FmtBuf.begin(), FmtBuf.end()); 689 return std::error_code(); 690 } 691 692 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 693 const RelocationRef &RelRef, 694 SmallVectorImpl<char> &Result) { 695 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 696 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 697 698 unsigned Arch = Obj->getArch(); 699 700 std::string FmtBuf; 701 raw_string_ostream Fmt(FmtBuf); 702 unsigned Type = Obj->getAnyRelocationType(RE); 703 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 704 705 // Determine any addends that should be displayed with the relocation. 706 // These require decoding the relocation type, which is triple-specific. 707 708 // X86_64 has entirely custom relocation types. 709 if (Arch == Triple::x86_64) { 710 switch (Type) { 711 case MachO::X86_64_RELOC_GOT_LOAD: 712 case MachO::X86_64_RELOC_GOT: { 713 printRelocationTargetName(Obj, RE, Fmt); 714 Fmt << "@GOT"; 715 if (IsPCRel) 716 Fmt << "PCREL"; 717 break; 718 } 719 case MachO::X86_64_RELOC_SUBTRACTOR: { 720 DataRefImpl RelNext = Rel; 721 Obj->moveRelocationNext(RelNext); 722 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 723 724 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 725 // X86_64_RELOC_UNSIGNED. 726 // NOTE: Scattered relocations don't exist on x86_64. 727 unsigned RType = Obj->getAnyRelocationType(RENext); 728 if (RType != MachO::X86_64_RELOC_UNSIGNED) 729 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 730 "X86_64_RELOC_SUBTRACTOR."); 731 732 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 733 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 734 printRelocationTargetName(Obj, RENext, Fmt); 735 Fmt << "-"; 736 printRelocationTargetName(Obj, RE, Fmt); 737 break; 738 } 739 case MachO::X86_64_RELOC_TLV: 740 printRelocationTargetName(Obj, RE, Fmt); 741 Fmt << "@TLV"; 742 if (IsPCRel) 743 Fmt << "P"; 744 break; 745 case MachO::X86_64_RELOC_SIGNED_1: 746 printRelocationTargetName(Obj, RE, Fmt); 747 Fmt << "-1"; 748 break; 749 case MachO::X86_64_RELOC_SIGNED_2: 750 printRelocationTargetName(Obj, RE, Fmt); 751 Fmt << "-2"; 752 break; 753 case MachO::X86_64_RELOC_SIGNED_4: 754 printRelocationTargetName(Obj, RE, Fmt); 755 Fmt << "-4"; 756 break; 757 default: 758 printRelocationTargetName(Obj, RE, Fmt); 759 break; 760 } 761 // X86 and ARM share some relocation types in common. 762 } else if (Arch == Triple::x86 || Arch == Triple::arm || 763 Arch == Triple::ppc) { 764 // Generic relocation types... 765 switch (Type) { 766 case MachO::GENERIC_RELOC_PAIR: // prints no info 767 return std::error_code(); 768 case MachO::GENERIC_RELOC_SECTDIFF: { 769 DataRefImpl RelNext = Rel; 770 Obj->moveRelocationNext(RelNext); 771 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 772 773 // X86 sect diff's must be followed by a relocation of type 774 // GENERIC_RELOC_PAIR. 775 unsigned RType = Obj->getAnyRelocationType(RENext); 776 777 if (RType != MachO::GENERIC_RELOC_PAIR) 778 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 779 "GENERIC_RELOC_SECTDIFF."); 780 781 printRelocationTargetName(Obj, RE, Fmt); 782 Fmt << "-"; 783 printRelocationTargetName(Obj, RENext, Fmt); 784 break; 785 } 786 } 787 788 if (Arch == Triple::x86 || Arch == Triple::ppc) { 789 switch (Type) { 790 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 791 DataRefImpl RelNext = Rel; 792 Obj->moveRelocationNext(RelNext); 793 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 794 795 // X86 sect diff's must be followed by a relocation of type 796 // GENERIC_RELOC_PAIR. 797 unsigned RType = Obj->getAnyRelocationType(RENext); 798 if (RType != MachO::GENERIC_RELOC_PAIR) 799 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 800 "GENERIC_RELOC_LOCAL_SECTDIFF."); 801 802 printRelocationTargetName(Obj, RE, Fmt); 803 Fmt << "-"; 804 printRelocationTargetName(Obj, RENext, Fmt); 805 break; 806 } 807 case MachO::GENERIC_RELOC_TLV: { 808 printRelocationTargetName(Obj, RE, Fmt); 809 Fmt << "@TLV"; 810 if (IsPCRel) 811 Fmt << "P"; 812 break; 813 } 814 default: 815 printRelocationTargetName(Obj, RE, Fmt); 816 } 817 } else { // ARM-specific relocations 818 switch (Type) { 819 case MachO::ARM_RELOC_HALF: 820 case MachO::ARM_RELOC_HALF_SECTDIFF: { 821 // Half relocations steal a bit from the length field to encode 822 // whether this is an upper16 or a lower16 relocation. 823 bool isUpper = (Obj->getAnyRelocationLength(RE) & 0x1) == 1; 824 825 if (isUpper) 826 Fmt << ":upper16:("; 827 else 828 Fmt << ":lower16:("; 829 printRelocationTargetName(Obj, RE, Fmt); 830 831 DataRefImpl RelNext = Rel; 832 Obj->moveRelocationNext(RelNext); 833 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 834 835 // ARM half relocs must be followed by a relocation of type 836 // ARM_RELOC_PAIR. 837 unsigned RType = Obj->getAnyRelocationType(RENext); 838 if (RType != MachO::ARM_RELOC_PAIR) 839 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 840 "ARM_RELOC_HALF"); 841 842 // NOTE: The half of the target virtual address is stashed in the 843 // address field of the secondary relocation, but we can't reverse 844 // engineer the constant offset from it without decoding the movw/movt 845 // instruction to find the other half in its immediate field. 846 847 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 848 // symbol/section pointer of the follow-on relocation. 849 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 850 Fmt << "-"; 851 printRelocationTargetName(Obj, RENext, Fmt); 852 } 853 854 Fmt << ")"; 855 break; 856 } 857 default: { printRelocationTargetName(Obj, RE, Fmt); } 858 } 859 } 860 } else 861 printRelocationTargetName(Obj, RE, Fmt); 862 863 Fmt.flush(); 864 Result.append(FmtBuf.begin(), FmtBuf.end()); 865 return std::error_code(); 866 } 867 868 static std::error_code getRelocationValueString(const RelocationRef &Rel, 869 SmallVectorImpl<char> &Result) { 870 const ObjectFile *Obj = Rel.getObject(); 871 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 872 return getRelocationValueString(ELF, Rel, Result); 873 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 874 return getRelocationValueString(COFF, Rel, Result); 875 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 876 return getRelocationValueString(Wasm, Rel, Result); 877 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 878 return getRelocationValueString(MachO, Rel, Result); 879 llvm_unreachable("unknown object file format"); 880 } 881 882 /// Indicates whether this relocation should hidden when listing 883 /// relocations, usually because it is the trailing part of a multipart 884 /// relocation that will be printed as part of the leading relocation. 885 static bool getHidden(RelocationRef RelRef) { 886 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 887 if (!MachO) 888 return false; 889 890 unsigned Arch = MachO->getArch(); 891 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 892 uint64_t Type = MachO->getRelocationType(Rel); 893 894 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 895 // is always hidden. 896 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 897 return Type == MachO::GENERIC_RELOC_PAIR; 898 899 if (Arch == Triple::x86_64) { 900 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 901 // an X86_64_RELOC_SUBTRACTOR. 902 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 903 DataRefImpl RelPrev = Rel; 904 RelPrev.d.a--; 905 uint64_t PrevType = MachO->getRelocationType(RelPrev); 906 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 907 return true; 908 } 909 } 910 911 return false; 912 } 913 914 namespace { 915 class SourcePrinter { 916 protected: 917 DILineInfo OldLineInfo; 918 const ObjectFile *Obj = nullptr; 919 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 920 // File name to file contents of source 921 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 922 // Mark the line endings of the cached source 923 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 924 925 private: 926 bool cacheSource(const DILineInfo& LineInfoFile); 927 928 public: 929 SourcePrinter() = default; 930 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 931 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 932 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 933 DefaultArch); 934 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 935 } 936 virtual ~SourcePrinter() = default; 937 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 938 StringRef Delimiter = "; "); 939 }; 940 941 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 942 std::unique_ptr<MemoryBuffer> Buffer; 943 if (LineInfo.Source) { 944 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 945 } else { 946 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 947 if (!BufferOrError) 948 return false; 949 Buffer = std::move(*BufferOrError); 950 } 951 // Chomp the file to get lines 952 size_t BufferSize = Buffer->getBufferSize(); 953 const char *BufferStart = Buffer->getBufferStart(); 954 for (const char *Start = BufferStart, *End = BufferStart; 955 End < BufferStart + BufferSize; End++) 956 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 957 (*End == '\r' && *(End + 1) == '\n')) { 958 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 959 if (*End == '\r') 960 End++; 961 Start = End + 1; 962 } 963 SourceCache[LineInfo.FileName] = std::move(Buffer); 964 return true; 965 } 966 967 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 968 StringRef Delimiter) { 969 if (!Symbolizer) 970 return; 971 DILineInfo LineInfo = DILineInfo(); 972 auto ExpectecLineInfo = 973 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 974 if (!ExpectecLineInfo) 975 consumeError(ExpectecLineInfo.takeError()); 976 else 977 LineInfo = *ExpectecLineInfo; 978 979 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 980 LineInfo.Line == 0) 981 return; 982 983 if (PrintLines) 984 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 985 if (PrintSource) { 986 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 987 if (!cacheSource(LineInfo)) 988 return; 989 auto FileBuffer = SourceCache.find(LineInfo.FileName); 990 if (FileBuffer != SourceCache.end()) { 991 auto LineBuffer = LineCache.find(LineInfo.FileName); 992 if (LineBuffer != LineCache.end()) { 993 if (LineInfo.Line > LineBuffer->second.size()) 994 return; 995 // Vector begins at 0, line numbers are non-zero 996 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 997 << "\n"; 998 } 999 } 1000 } 1001 OldLineInfo = LineInfo; 1002 } 1003 1004 static bool isArmElf(const ObjectFile *Obj) { 1005 return (Obj->isELF() && 1006 (Obj->getArch() == Triple::aarch64 || 1007 Obj->getArch() == Triple::aarch64_be || 1008 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 1009 Obj->getArch() == Triple::thumb || 1010 Obj->getArch() == Triple::thumbeb)); 1011 } 1012 1013 class PrettyPrinter { 1014 public: 1015 virtual ~PrettyPrinter() = default; 1016 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 1017 ArrayRef<uint8_t> Bytes, uint64_t Address, 1018 raw_ostream &OS, StringRef Annot, 1019 MCSubtargetInfo const &STI, SourcePrinter *SP, 1020 std::vector<RelocationRef> *Rels = nullptr) { 1021 if (SP && (PrintSource || PrintLines)) 1022 SP->printSourceLine(OS, Address); 1023 if (!NoLeadingAddr) 1024 OS << format("%8" PRIx64 ":", Address); 1025 if (!NoShowRawInsn) { 1026 OS << "\t"; 1027 dumpBytes(Bytes, OS); 1028 } 1029 if (MI) 1030 IP.printInst(MI, OS, "", STI); 1031 else 1032 OS << " <unknown>"; 1033 } 1034 }; 1035 PrettyPrinter PrettyPrinterInst; 1036 class HexagonPrettyPrinter : public PrettyPrinter { 1037 public: 1038 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 1039 raw_ostream &OS) { 1040 uint32_t opcode = 1041 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 1042 if (!NoLeadingAddr) 1043 OS << format("%8" PRIx64 ":", Address); 1044 if (!NoShowRawInsn) { 1045 OS << "\t"; 1046 dumpBytes(Bytes.slice(0, 4), OS); 1047 OS << format("%08" PRIx32, opcode); 1048 } 1049 } 1050 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1051 uint64_t Address, raw_ostream &OS, StringRef Annot, 1052 MCSubtargetInfo const &STI, SourcePrinter *SP, 1053 std::vector<RelocationRef> *Rels) override { 1054 if (SP && (PrintSource || PrintLines)) 1055 SP->printSourceLine(OS, Address, ""); 1056 if (!MI) { 1057 printLead(Bytes, Address, OS); 1058 OS << " <unknown>"; 1059 return; 1060 } 1061 std::string Buffer; 1062 { 1063 raw_string_ostream TempStream(Buffer); 1064 IP.printInst(MI, TempStream, "", STI); 1065 } 1066 StringRef Contents(Buffer); 1067 // Split off bundle attributes 1068 auto PacketBundle = Contents.rsplit('\n'); 1069 // Split off first instruction from the rest 1070 auto HeadTail = PacketBundle.first.split('\n'); 1071 auto Preamble = " { "; 1072 auto Separator = ""; 1073 StringRef Fmt = "\t\t\t%08" PRIx64 ": "; 1074 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 1075 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 1076 1077 // Hexagon's packets require relocations to be inline rather than 1078 // clustered at the end of the packet. 1079 auto PrintReloc = [&]() -> void { 1080 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address)) { 1081 if (RelCur->getOffset() == Address) { 1082 SmallString<16> Name; 1083 SmallString<32> Val; 1084 RelCur->getTypeName(Name); 1085 error(getRelocationValueString(*RelCur, Val)); 1086 OS << Separator << format(Fmt.data(), Address) << Name << "\t" << Val 1087 << "\n"; 1088 return; 1089 } 1090 ++RelCur; 1091 } 1092 }; 1093 1094 while (!HeadTail.first.empty()) { 1095 OS << Separator; 1096 Separator = "\n"; 1097 if (SP && (PrintSource || PrintLines)) 1098 SP->printSourceLine(OS, Address, ""); 1099 printLead(Bytes, Address, OS); 1100 OS << Preamble; 1101 Preamble = " "; 1102 StringRef Inst; 1103 auto Duplex = HeadTail.first.split('\v'); 1104 if (!Duplex.second.empty()) { 1105 OS << Duplex.first; 1106 OS << "; "; 1107 Inst = Duplex.second; 1108 } 1109 else 1110 Inst = HeadTail.first; 1111 OS << Inst; 1112 HeadTail = HeadTail.second.split('\n'); 1113 if (HeadTail.first.empty()) 1114 OS << " } " << PacketBundle.second; 1115 PrintReloc(); 1116 Bytes = Bytes.slice(4); 1117 Address += 4; 1118 } 1119 } 1120 }; 1121 HexagonPrettyPrinter HexagonPrettyPrinterInst; 1122 1123 class AMDGCNPrettyPrinter : public PrettyPrinter { 1124 public: 1125 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1126 uint64_t Address, raw_ostream &OS, StringRef Annot, 1127 MCSubtargetInfo const &STI, SourcePrinter *SP, 1128 std::vector<RelocationRef> *Rels) override { 1129 if (SP && (PrintSource || PrintLines)) 1130 SP->printSourceLine(OS, Address); 1131 1132 typedef support::ulittle32_t U32; 1133 1134 if (MI) { 1135 SmallString<40> InstStr; 1136 raw_svector_ostream IS(InstStr); 1137 1138 IP.printInst(MI, IS, "", STI); 1139 1140 OS << left_justify(IS.str(), 60); 1141 } else { 1142 // an unrecognized encoding - this is probably data so represent it 1143 // using the .long directive, or .byte directive if fewer than 4 bytes 1144 // remaining 1145 if (Bytes.size() >= 4) { 1146 OS << format("\t.long 0x%08" PRIx32 " ", 1147 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 1148 OS.indent(42); 1149 } else { 1150 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 1151 for (unsigned int i = 1; i < Bytes.size(); i++) 1152 OS << format(", 0x%02" PRIx8, Bytes[i]); 1153 OS.indent(55 - (6 * Bytes.size())); 1154 } 1155 } 1156 1157 OS << format("// %012" PRIX64 ": ", Address); 1158 if (Bytes.size() >=4) { 1159 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 1160 Bytes.size() / sizeof(U32))) 1161 // D should be explicitly casted to uint32_t here as it is passed 1162 // by format to snprintf as vararg. 1163 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 1164 } else { 1165 for (unsigned int i = 0; i < Bytes.size(); i++) 1166 OS << format("%02" PRIX8 " ", Bytes[i]); 1167 } 1168 1169 if (!Annot.empty()) 1170 OS << "// " << Annot; 1171 } 1172 }; 1173 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 1174 1175 class BPFPrettyPrinter : public PrettyPrinter { 1176 public: 1177 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1178 uint64_t Address, raw_ostream &OS, StringRef Annot, 1179 MCSubtargetInfo const &STI, SourcePrinter *SP, 1180 std::vector<RelocationRef> *Rels) override { 1181 if (SP && (PrintSource || PrintLines)) 1182 SP->printSourceLine(OS, Address); 1183 if (!NoLeadingAddr) 1184 OS << format("%8" PRId64 ":", Address / 8); 1185 if (!NoShowRawInsn) { 1186 OS << "\t"; 1187 dumpBytes(Bytes, OS); 1188 } 1189 if (MI) 1190 IP.printInst(MI, OS, "", STI); 1191 else 1192 OS << " <unknown>"; 1193 } 1194 }; 1195 BPFPrettyPrinter BPFPrettyPrinterInst; 1196 1197 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 1198 switch(Triple.getArch()) { 1199 default: 1200 return PrettyPrinterInst; 1201 case Triple::hexagon: 1202 return HexagonPrettyPrinterInst; 1203 case Triple::amdgcn: 1204 return AMDGCNPrettyPrinterInst; 1205 case Triple::bpfel: 1206 case Triple::bpfeb: 1207 return BPFPrettyPrinterInst; 1208 } 1209 } 1210 } 1211 1212 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1213 assert(Obj->isELF()); 1214 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1215 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1216 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1217 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1218 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1219 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1220 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1221 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1222 llvm_unreachable("Unsupported binary format"); 1223 } 1224 1225 template <class ELFT> static void 1226 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1227 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1228 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1229 uint8_t SymbolType = Symbol.getELFType(); 1230 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1231 continue; 1232 1233 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1234 if (!AddressOrErr) 1235 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1236 1237 Expected<StringRef> Name = Symbol.getName(); 1238 if (!Name) 1239 report_error(Obj->getFileName(), Name.takeError()); 1240 if (Name->empty()) 1241 continue; 1242 1243 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1244 if (!SectionOrErr) 1245 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1246 section_iterator SecI = *SectionOrErr; 1247 if (SecI == Obj->section_end()) 1248 continue; 1249 1250 AllSymbols[*SecI].emplace_back(*AddressOrErr, *Name, SymbolType); 1251 } 1252 } 1253 1254 static void 1255 addDynamicElfSymbols(const ObjectFile *Obj, 1256 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1257 assert(Obj->isELF()); 1258 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1259 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1260 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1261 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1262 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1263 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1264 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1265 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1266 else 1267 llvm_unreachable("Unsupported binary format"); 1268 } 1269 1270 static void addPltEntries(const ObjectFile *Obj, 1271 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 1272 StringSaver &Saver) { 1273 Optional<SectionRef> Plt = None; 1274 for (const SectionRef &Section : Obj->sections()) { 1275 StringRef Name; 1276 if (Section.getName(Name)) 1277 continue; 1278 if (Name == ".plt") 1279 Plt = Section; 1280 } 1281 if (!Plt) 1282 return; 1283 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1284 for (auto PltEntry : ElfObj->getPltAddresses()) { 1285 SymbolRef Symbol(PltEntry.first, ElfObj); 1286 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1287 1288 Expected<StringRef> NameOrErr = Symbol.getName(); 1289 if (!NameOrErr) 1290 report_error(Obj->getFileName(), NameOrErr.takeError()); 1291 if (NameOrErr->empty()) 1292 continue; 1293 StringRef Name = Saver.save((*NameOrErr + "@plt").str()); 1294 1295 AllSymbols[*Plt].emplace_back(PltEntry.second, Name, SymbolType); 1296 } 1297 } 1298 } 1299 1300 // Normally the disassembly output will skip blocks of zeroes. This function 1301 // returns the number of zero bytes that can be skipped when dumping the 1302 // disassembly of the instructions in Buf. 1303 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 1304 // When -z or --disassemble-zeroes are given we always dissasemble them. 1305 if (DisassembleZeroes) 1306 return 0; 1307 1308 // Find the number of leading zeroes. 1309 size_t N = 0; 1310 while (N < Buf.size() && !Buf[N]) 1311 ++N; 1312 1313 // We may want to skip blocks of zero bytes, but unless we see 1314 // at least 8 of them in a row. 1315 if (N < 8) 1316 return 0; 1317 1318 // We skip zeroes in multiples of 4 because do not want to truncate an 1319 // instruction if it starts with a zero byte. 1320 return N & ~0x3; 1321 } 1322 1323 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1324 if (StartAddress > StopAddress) 1325 error("Start address should be less than stop address"); 1326 1327 const Target *TheTarget = getTarget(Obj); 1328 1329 // Package up features to be passed to target/subtarget 1330 SubtargetFeatures Features = Obj->getFeatures(); 1331 if (!MAttrs.empty()) 1332 for (unsigned I = 0; I != MAttrs.size(); ++I) 1333 Features.AddFeature(MAttrs[I]); 1334 1335 std::unique_ptr<const MCRegisterInfo> MRI( 1336 TheTarget->createMCRegInfo(TripleName)); 1337 if (!MRI) 1338 report_error(Obj->getFileName(), "no register info for target " + 1339 TripleName); 1340 1341 // Set up disassembler. 1342 std::unique_ptr<const MCAsmInfo> AsmInfo( 1343 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1344 if (!AsmInfo) 1345 report_error(Obj->getFileName(), "no assembly info for target " + 1346 TripleName); 1347 std::unique_ptr<const MCSubtargetInfo> STI( 1348 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1349 if (!STI) 1350 report_error(Obj->getFileName(), "no subtarget info for target " + 1351 TripleName); 1352 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1353 if (!MII) 1354 report_error(Obj->getFileName(), "no instruction info for target " + 1355 TripleName); 1356 MCObjectFileInfo MOFI; 1357 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1358 // FIXME: for now initialize MCObjectFileInfo with default values 1359 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1360 1361 std::unique_ptr<MCDisassembler> DisAsm( 1362 TheTarget->createMCDisassembler(*STI, Ctx)); 1363 if (!DisAsm) 1364 report_error(Obj->getFileName(), "no disassembler for target " + 1365 TripleName); 1366 1367 std::unique_ptr<const MCInstrAnalysis> MIA( 1368 TheTarget->createMCInstrAnalysis(MII.get())); 1369 1370 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1371 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1372 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1373 if (!IP) 1374 report_error(Obj->getFileName(), "no instruction printer for target " + 1375 TripleName); 1376 IP->setPrintImmHex(PrintImmHex); 1377 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1378 1379 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1380 "\t\t\t%08" PRIx64 ": "; 1381 1382 SourcePrinter SP(Obj, TheTarget->getName()); 1383 1384 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1385 // in RelocSecs contain the relocations for section S. 1386 std::error_code EC; 1387 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1388 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1389 section_iterator Sec2 = Section.getRelocatedSection(); 1390 if (Sec2 != Obj->section_end()) 1391 SectionRelocMap[*Sec2].push_back(Section); 1392 } 1393 1394 // Create a mapping from virtual address to symbol name. This is used to 1395 // pretty print the symbols while disassembling. 1396 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1397 SectionSymbolsTy AbsoluteSymbols; 1398 for (const SymbolRef &Symbol : Obj->symbols()) { 1399 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1400 if (!AddressOrErr) 1401 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1402 uint64_t Address = *AddressOrErr; 1403 1404 Expected<StringRef> Name = Symbol.getName(); 1405 if (!Name) 1406 report_error(Obj->getFileName(), Name.takeError()); 1407 if (Name->empty()) 1408 continue; 1409 1410 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1411 if (!SectionOrErr) 1412 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1413 1414 uint8_t SymbolType = ELF::STT_NOTYPE; 1415 if (Obj->isELF()) 1416 SymbolType = getElfSymbolType(Obj, Symbol); 1417 1418 section_iterator SecI = *SectionOrErr; 1419 if (SecI != Obj->section_end()) 1420 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1421 else 1422 AbsoluteSymbols.emplace_back(Address, *Name, SymbolType); 1423 1424 1425 } 1426 if (AllSymbols.empty() && Obj->isELF()) 1427 addDynamicElfSymbols(Obj, AllSymbols); 1428 1429 BumpPtrAllocator A; 1430 StringSaver Saver(A); 1431 addPltEntries(Obj, AllSymbols, Saver); 1432 1433 // Create a mapping from virtual address to section. 1434 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1435 for (SectionRef Sec : Obj->sections()) 1436 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1437 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1438 1439 // Linked executables (.exe and .dll files) typically don't include a real 1440 // symbol table but they might contain an export table. 1441 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1442 for (const auto &ExportEntry : COFFObj->export_directories()) { 1443 StringRef Name; 1444 error(ExportEntry.getSymbolName(Name)); 1445 if (Name.empty()) 1446 continue; 1447 uint32_t RVA; 1448 error(ExportEntry.getExportRVA(RVA)); 1449 1450 uint64_t VA = COFFObj->getImageBase() + RVA; 1451 auto Sec = std::upper_bound( 1452 SectionAddresses.begin(), SectionAddresses.end(), VA, 1453 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1454 return LHS < RHS.first; 1455 }); 1456 if (Sec != SectionAddresses.begin()) 1457 --Sec; 1458 else 1459 Sec = SectionAddresses.end(); 1460 1461 if (Sec != SectionAddresses.end()) 1462 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1463 else 1464 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1465 } 1466 } 1467 1468 // Sort all the symbols, this allows us to use a simple binary search to find 1469 // a symbol near an address. 1470 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1471 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1472 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1473 1474 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1475 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1476 continue; 1477 1478 uint64_t SectionAddr = Section.getAddress(); 1479 uint64_t SectSize = Section.getSize(); 1480 if (!SectSize) 1481 continue; 1482 1483 // Get the list of all the symbols in this section. 1484 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1485 std::vector<uint64_t> DataMappingSymsAddr; 1486 std::vector<uint64_t> TextMappingSymsAddr; 1487 if (isArmElf(Obj)) { 1488 for (const auto &Symb : Symbols) { 1489 uint64_t Address = std::get<0>(Symb); 1490 StringRef Name = std::get<1>(Symb); 1491 if (Name.startswith("$d")) 1492 DataMappingSymsAddr.push_back(Address - SectionAddr); 1493 if (Name.startswith("$x")) 1494 TextMappingSymsAddr.push_back(Address - SectionAddr); 1495 if (Name.startswith("$a")) 1496 TextMappingSymsAddr.push_back(Address - SectionAddr); 1497 if (Name.startswith("$t")) 1498 TextMappingSymsAddr.push_back(Address - SectionAddr); 1499 } 1500 } 1501 1502 llvm::sort(DataMappingSymsAddr); 1503 llvm::sort(TextMappingSymsAddr); 1504 1505 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1506 // AMDGPU disassembler uses symbolizer for printing labels 1507 std::unique_ptr<MCRelocationInfo> RelInfo( 1508 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1509 if (RelInfo) { 1510 std::unique_ptr<MCSymbolizer> Symbolizer( 1511 TheTarget->createMCSymbolizer( 1512 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1513 DisAsm->setSymbolizer(std::move(Symbolizer)); 1514 } 1515 } 1516 1517 // Make a list of all the relocations for this section. 1518 std::vector<RelocationRef> Rels; 1519 if (InlineRelocs) { 1520 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1521 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1522 Rels.push_back(Reloc); 1523 } 1524 } 1525 } 1526 1527 // Sort relocations by address. 1528 llvm::sort(Rels, isRelocAddressLess); 1529 1530 StringRef SegmentName = ""; 1531 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1532 DataRefImpl DR = Section.getRawDataRefImpl(); 1533 SegmentName = MachO->getSectionFinalSegmentName(DR); 1534 } 1535 StringRef SectionName; 1536 error(Section.getName(SectionName)); 1537 1538 // If the section has no symbol at the start, just insert a dummy one. 1539 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1540 Symbols.insert( 1541 Symbols.begin(), 1542 std::make_tuple(SectionAddr, SectionName, 1543 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1544 } 1545 1546 SmallString<40> Comments; 1547 raw_svector_ostream CommentStream(Comments); 1548 1549 StringRef BytesStr; 1550 error(Section.getContents(BytesStr)); 1551 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1552 BytesStr.size()); 1553 1554 uint64_t Size; 1555 uint64_t Index; 1556 bool PrintedSection = false; 1557 1558 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1559 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1560 // Disassemble symbol by symbol. 1561 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1562 uint64_t Start = std::get<0>(Symbols[SI]) - SectionAddr; 1563 // The end is either the section end or the beginning of the next 1564 // symbol. 1565 uint64_t End = (SI == SE - 1) 1566 ? SectSize 1567 : std::get<0>(Symbols[SI + 1]) - SectionAddr; 1568 // Don't try to disassemble beyond the end of section contents. 1569 if (End > SectSize) 1570 End = SectSize; 1571 // If this symbol has the same address as the next symbol, then skip it. 1572 if (Start >= End) 1573 continue; 1574 1575 // Check if we need to skip symbol 1576 // Skip if the symbol's data is not between StartAddress and StopAddress 1577 if (End + SectionAddr < StartAddress || 1578 Start + SectionAddr > StopAddress) { 1579 continue; 1580 } 1581 1582 /// Skip if user requested specific symbols and this is not in the list 1583 if (!DisasmFuncsSet.empty() && 1584 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1585 continue; 1586 1587 if (!PrintedSection) { 1588 PrintedSection = true; 1589 outs() << "Disassembly of section "; 1590 if (!SegmentName.empty()) 1591 outs() << SegmentName << ","; 1592 outs() << SectionName << ':'; 1593 } 1594 1595 // Stop disassembly at the stop address specified 1596 if (End + SectionAddr > StopAddress) 1597 End = StopAddress - SectionAddr; 1598 1599 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1600 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1601 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1602 Start += 256; 1603 } 1604 if (SI == SE - 1 || 1605 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1606 // cut trailing zeroes at the end of kernel 1607 // cut up to 256 bytes 1608 const uint64_t EndAlign = 256; 1609 const auto Limit = End - (std::min)(EndAlign, End - Start); 1610 while (End > Limit && 1611 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1612 End -= 4; 1613 } 1614 } 1615 1616 outs() << '\n'; 1617 if (!NoLeadingAddr) 1618 outs() << format("%016" PRIx64 " ", SectionAddr + Start); 1619 1620 StringRef SymbolName = std::get<1>(Symbols[SI]); 1621 if (Demangle) 1622 outs() << demangle(SymbolName) << ":\n"; 1623 else 1624 outs() << SymbolName << ":\n"; 1625 1626 // Don't print raw contents of a virtual section. A virtual section 1627 // doesn't have any contents in the file. 1628 if (Section.isVirtual()) { 1629 outs() << "...\n"; 1630 continue; 1631 } 1632 1633 #ifndef NDEBUG 1634 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1635 #else 1636 raw_ostream &DebugOut = nulls(); 1637 #endif 1638 1639 for (Index = Start; Index < End; Index += Size) { 1640 MCInst Inst; 1641 1642 if (Index + SectionAddr < StartAddress || 1643 Index + SectionAddr > StopAddress) { 1644 // skip byte by byte till StartAddress is reached 1645 Size = 1; 1646 continue; 1647 } 1648 // AArch64 ELF binaries can interleave data and text in the 1649 // same section. We rely on the markers introduced to 1650 // understand what we need to dump. If the data marker is within a 1651 // function, it is denoted as a word/short etc 1652 if (isArmElf(Obj) && std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1653 !DisassembleAll) { 1654 uint64_t Stride = 0; 1655 1656 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1657 DataMappingSymsAddr.end(), Index); 1658 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1659 // Switch to data. 1660 while (Index < End) { 1661 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1662 outs() << "\t"; 1663 if (Index + 4 <= End) { 1664 Stride = 4; 1665 dumpBytes(Bytes.slice(Index, 4), outs()); 1666 outs() << "\t.word\t"; 1667 uint32_t Data = 0; 1668 if (Obj->isLittleEndian()) { 1669 const auto Word = 1670 reinterpret_cast<const support::ulittle32_t *>( 1671 Bytes.data() + Index); 1672 Data = *Word; 1673 } else { 1674 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1675 Bytes.data() + Index); 1676 Data = *Word; 1677 } 1678 outs() << "0x" << format("%08" PRIx32, Data); 1679 } else if (Index + 2 <= End) { 1680 Stride = 2; 1681 dumpBytes(Bytes.slice(Index, 2), outs()); 1682 outs() << "\t\t.short\t"; 1683 uint16_t Data = 0; 1684 if (Obj->isLittleEndian()) { 1685 const auto Short = 1686 reinterpret_cast<const support::ulittle16_t *>( 1687 Bytes.data() + Index); 1688 Data = *Short; 1689 } else { 1690 const auto Short = 1691 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1692 Index); 1693 Data = *Short; 1694 } 1695 outs() << "0x" << format("%04" PRIx16, Data); 1696 } else { 1697 Stride = 1; 1698 dumpBytes(Bytes.slice(Index, 1), outs()); 1699 outs() << "\t\t.byte\t"; 1700 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1701 } 1702 Index += Stride; 1703 outs() << "\n"; 1704 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1705 TextMappingSymsAddr.end(), Index); 1706 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1707 break; 1708 } 1709 } 1710 } 1711 1712 // If there is a data symbol inside an ELF text section and we are only 1713 // disassembling text (applicable all architectures), 1714 // we are in a situation where we must print the data and not 1715 // disassemble it. 1716 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1717 !DisassembleAll && Section.isText()) { 1718 // print out data up to 8 bytes at a time in hex and ascii 1719 uint8_t AsciiData[9] = {'\0'}; 1720 uint8_t Byte; 1721 int NumBytes = 0; 1722 1723 for (Index = Start; Index < End; Index += 1) { 1724 if (((SectionAddr + Index) < StartAddress) || 1725 ((SectionAddr + Index) > StopAddress)) 1726 continue; 1727 if (NumBytes == 0) { 1728 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1729 outs() << "\t"; 1730 } 1731 Byte = Bytes.slice(Index)[0]; 1732 outs() << format(" %02x", Byte); 1733 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1734 1735 uint8_t IndentOffset = 0; 1736 NumBytes++; 1737 if (Index == End - 1 || NumBytes > 8) { 1738 // Indent the space for less than 8 bytes data. 1739 // 2 spaces for byte and one for space between bytes 1740 IndentOffset = 3 * (8 - NumBytes); 1741 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1742 AsciiData[Excess] = '\0'; 1743 NumBytes = 8; 1744 } 1745 if (NumBytes == 8) { 1746 AsciiData[8] = '\0'; 1747 outs() << std::string(IndentOffset, ' ') << " "; 1748 outs() << reinterpret_cast<char *>(AsciiData); 1749 outs() << '\n'; 1750 NumBytes = 0; 1751 } 1752 } 1753 } 1754 if (Index >= End) 1755 break; 1756 1757 if (size_t N = 1758 countSkippableZeroBytes(Bytes.slice(Index, End - Index))) { 1759 outs() << "\t\t..." << '\n'; 1760 Index += N; 1761 if (Index >= End) 1762 break; 1763 } 1764 1765 // Disassemble a real instruction or a data when disassemble all is 1766 // provided 1767 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1768 SectionAddr + Index, DebugOut, 1769 CommentStream); 1770 if (Size == 0) 1771 Size = 1; 1772 1773 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1774 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1775 *STI, &SP, &Rels); 1776 outs() << CommentStream.str(); 1777 Comments.clear(); 1778 1779 // Try to resolve the target of a call, tail call, etc. to a specific 1780 // symbol. 1781 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1782 MIA->isConditionalBranch(Inst))) { 1783 uint64_t Target; 1784 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1785 // In a relocatable object, the target's section must reside in 1786 // the same section as the call instruction or it is accessed 1787 // through a relocation. 1788 // 1789 // In a non-relocatable object, the target may be in any section. 1790 // 1791 // N.B. We don't walk the relocations in the relocatable case yet. 1792 auto *TargetSectionSymbols = &Symbols; 1793 if (!Obj->isRelocatableObject()) { 1794 auto SectionAddress = std::upper_bound( 1795 SectionAddresses.begin(), SectionAddresses.end(), Target, 1796 [](uint64_t LHS, 1797 const std::pair<uint64_t, SectionRef> &RHS) { 1798 return LHS < RHS.first; 1799 }); 1800 if (SectionAddress != SectionAddresses.begin()) { 1801 --SectionAddress; 1802 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1803 } else { 1804 TargetSectionSymbols = &AbsoluteSymbols; 1805 } 1806 } 1807 1808 // Find the first symbol in the section whose offset is less than 1809 // or equal to the target. If there isn't a section that contains 1810 // the target, find the nearest preceding absolute symbol. 1811 auto TargetSym = std::upper_bound( 1812 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1813 Target, [](uint64_t LHS, 1814 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1815 return LHS < std::get<0>(RHS); 1816 }); 1817 if (TargetSym == TargetSectionSymbols->begin()) { 1818 TargetSectionSymbols = &AbsoluteSymbols; 1819 TargetSym = std::upper_bound( 1820 AbsoluteSymbols.begin(), AbsoluteSymbols.end(), 1821 Target, [](uint64_t LHS, 1822 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1823 return LHS < std::get<0>(RHS); 1824 }); 1825 } 1826 if (TargetSym != TargetSectionSymbols->begin()) { 1827 --TargetSym; 1828 uint64_t TargetAddress = std::get<0>(*TargetSym); 1829 StringRef TargetName = std::get<1>(*TargetSym); 1830 outs() << " <" << TargetName; 1831 uint64_t Disp = Target - TargetAddress; 1832 if (Disp) 1833 outs() << "+0x" << Twine::utohexstr(Disp); 1834 outs() << '>'; 1835 } 1836 } 1837 } 1838 outs() << "\n"; 1839 1840 // Hexagon does this in pretty printer 1841 if (Obj->getArch() != Triple::hexagon) 1842 // Print relocation for instruction. 1843 while (RelCur != RelEnd) { 1844 uint64_t Addr = RelCur->getOffset(); 1845 SmallString<16> Name; 1846 SmallString<32> Val; 1847 1848 // If this relocation is hidden, skip it. 1849 if (getHidden(*RelCur) || ((SectionAddr + Addr) < StartAddress)) { 1850 ++RelCur; 1851 continue; 1852 } 1853 1854 // Stop when rel_cur's address is past the current instruction. 1855 if (Addr >= Index + Size) 1856 break; 1857 RelCur->getTypeName(Name); 1858 error(getRelocationValueString(*RelCur, Val)); 1859 outs() << format(Fmt.data(), SectionAddr + Addr) << Name << "\t" 1860 << Val << "\n"; 1861 ++RelCur; 1862 } 1863 } 1864 } 1865 } 1866 } 1867 1868 void llvm::printRelocations(const ObjectFile *Obj) { 1869 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1870 "%08" PRIx64; 1871 // Regular objdump doesn't print relocations in non-relocatable object 1872 // files. 1873 if (!Obj->isRelocatableObject()) 1874 return; 1875 1876 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1877 if (Section.relocation_begin() == Section.relocation_end()) 1878 continue; 1879 StringRef SecName; 1880 error(Section.getName(SecName)); 1881 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1882 for (const RelocationRef &Reloc : Section.relocations()) { 1883 uint64_t Address = Reloc.getOffset(); 1884 SmallString<32> RelocName; 1885 SmallString<32> ValueStr; 1886 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1887 continue; 1888 Reloc.getTypeName(RelocName); 1889 error(getRelocationValueString(Reloc, ValueStr)); 1890 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1891 << ValueStr << "\n"; 1892 } 1893 outs() << "\n"; 1894 } 1895 } 1896 1897 void llvm::printDynamicRelocations(const ObjectFile *Obj) { 1898 // For the moment, this option is for ELF only 1899 if (!Obj->isELF()) 1900 return; 1901 1902 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1903 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1904 error("not a dynamic object"); 1905 return; 1906 } 1907 1908 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1909 if (DynRelSec.empty()) 1910 return; 1911 1912 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1913 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1914 for (const SectionRef &Section : DynRelSec) { 1915 if (Section.relocation_begin() == Section.relocation_end()) 1916 continue; 1917 for (const RelocationRef &Reloc : Section.relocations()) { 1918 uint64_t Address = Reloc.getOffset(); 1919 SmallString<32> RelocName; 1920 SmallString<32> ValueStr; 1921 Reloc.getTypeName(RelocName); 1922 error(getRelocationValueString(Reloc, ValueStr)); 1923 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1924 << ValueStr << "\n"; 1925 } 1926 } 1927 } 1928 1929 void llvm::printSectionHeaders(const ObjectFile *Obj) { 1930 outs() << "Sections:\n" 1931 "Idx Name Size Address Type\n"; 1932 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1933 StringRef Name; 1934 error(Section.getName(Name)); 1935 uint64_t Address = Section.getAddress(); 1936 uint64_t Size = Section.getSize(); 1937 bool Text = Section.isText(); 1938 bool Data = Section.isData(); 1939 bool BSS = Section.isBSS(); 1940 std::string Type = (std::string(Text ? "TEXT " : "") + 1941 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1942 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1943 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1944 Address, Type.c_str()); 1945 } 1946 outs() << "\n"; 1947 } 1948 1949 void llvm::printSectionContents(const ObjectFile *Obj) { 1950 std::error_code EC; 1951 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1952 StringRef Name; 1953 StringRef Contents; 1954 error(Section.getName(Name)); 1955 uint64_t BaseAddr = Section.getAddress(); 1956 uint64_t Size = Section.getSize(); 1957 if (!Size) 1958 continue; 1959 1960 outs() << "Contents of section " << Name << ":\n"; 1961 if (Section.isBSS()) { 1962 outs() << format("<skipping contents of bss section at [%04" PRIx64 1963 ", %04" PRIx64 ")>\n", 1964 BaseAddr, BaseAddr + Size); 1965 continue; 1966 } 1967 1968 error(Section.getContents(Contents)); 1969 1970 // Dump out the content as hex and printable ascii characters. 1971 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1972 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1973 // Dump line of hex. 1974 for (std::size_t I = 0; I < 16; ++I) { 1975 if (I != 0 && I % 4 == 0) 1976 outs() << ' '; 1977 if (Addr + I < End) 1978 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1979 << hexdigit(Contents[Addr + I] & 0xF, true); 1980 else 1981 outs() << " "; 1982 } 1983 // Print ascii. 1984 outs() << " "; 1985 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1986 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1987 outs() << Contents[Addr + I]; 1988 else 1989 outs() << "."; 1990 } 1991 outs() << "\n"; 1992 } 1993 } 1994 } 1995 1996 void llvm::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1997 StringRef ArchitectureName) { 1998 outs() << "SYMBOL TABLE:\n"; 1999 2000 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 2001 printCOFFSymbolTable(Coff); 2002 return; 2003 } 2004 2005 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 2006 // Skip printing the special zero symbol when dumping an ELF file. 2007 // This makes the output consistent with the GNU objdump. 2008 if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O)) 2009 continue; 2010 2011 const SymbolRef &Symbol = *I; 2012 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 2013 if (!AddressOrError) 2014 report_error(ArchiveName, O->getFileName(), AddressOrError.takeError(), 2015 ArchitectureName); 2016 uint64_t Address = *AddressOrError; 2017 if ((Address < StartAddress) || (Address > StopAddress)) 2018 continue; 2019 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 2020 if (!TypeOrError) 2021 report_error(ArchiveName, O->getFileName(), TypeOrError.takeError(), 2022 ArchitectureName); 2023 SymbolRef::Type Type = *TypeOrError; 2024 uint32_t Flags = Symbol.getFlags(); 2025 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 2026 if (!SectionOrErr) 2027 report_error(ArchiveName, O->getFileName(), SectionOrErr.takeError(), 2028 ArchitectureName); 2029 section_iterator Section = *SectionOrErr; 2030 StringRef Name; 2031 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 2032 Section->getName(Name); 2033 } else { 2034 Expected<StringRef> NameOrErr = Symbol.getName(); 2035 if (!NameOrErr) 2036 report_error(ArchiveName, O->getFileName(), NameOrErr.takeError(), 2037 ArchitectureName); 2038 Name = *NameOrErr; 2039 } 2040 2041 bool Global = Flags & SymbolRef::SF_Global; 2042 bool Weak = Flags & SymbolRef::SF_Weak; 2043 bool Absolute = Flags & SymbolRef::SF_Absolute; 2044 bool Common = Flags & SymbolRef::SF_Common; 2045 bool Hidden = Flags & SymbolRef::SF_Hidden; 2046 2047 char GlobLoc = ' '; 2048 if (Type != SymbolRef::ST_Unknown) 2049 GlobLoc = Global ? 'g' : 'l'; 2050 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2051 ? 'd' : ' '; 2052 char FileFunc = ' '; 2053 if (Type == SymbolRef::ST_File) 2054 FileFunc = 'f'; 2055 else if (Type == SymbolRef::ST_Function) 2056 FileFunc = 'F'; 2057 else if (Type == SymbolRef::ST_Data) 2058 FileFunc = 'O'; 2059 2060 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 2061 "%08" PRIx64; 2062 2063 outs() << format(Fmt, Address) << " " 2064 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2065 << (Weak ? 'w' : ' ') // Weak? 2066 << ' ' // Constructor. Not supported yet. 2067 << ' ' // Warning. Not supported yet. 2068 << ' ' // Indirect reference to another symbol. 2069 << Debug // Debugging (d) or dynamic (D) symbol. 2070 << FileFunc // Name of function (F), file (f) or object (O). 2071 << ' '; 2072 if (Absolute) { 2073 outs() << "*ABS*"; 2074 } else if (Common) { 2075 outs() << "*COM*"; 2076 } else if (Section == O->section_end()) { 2077 outs() << "*UND*"; 2078 } else { 2079 if (const MachOObjectFile *MachO = 2080 dyn_cast<const MachOObjectFile>(O)) { 2081 DataRefImpl DR = Section->getRawDataRefImpl(); 2082 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 2083 outs() << SegmentName << ","; 2084 } 2085 StringRef SectionName; 2086 error(Section->getName(SectionName)); 2087 outs() << SectionName; 2088 } 2089 2090 outs() << '\t'; 2091 if (Common || isa<ELFObjectFileBase>(O)) { 2092 uint64_t Val = 2093 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2094 outs() << format("\t %08" PRIx64 " ", Val); 2095 } 2096 2097 if (Hidden) 2098 outs() << ".hidden "; 2099 2100 if (Demangle) 2101 outs() << demangle(Name) << '\n'; 2102 else 2103 outs() << Name << '\n'; 2104 } 2105 } 2106 2107 static void printUnwindInfo(const ObjectFile *O) { 2108 outs() << "Unwind info:\n\n"; 2109 2110 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 2111 printCOFFUnwindInfo(Coff); 2112 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 2113 printMachOUnwindInfo(MachO); 2114 else 2115 // TODO: Extract DWARF dump tool to objdump. 2116 WithColor::error(errs(), ToolName) 2117 << "This operation is only currently supported " 2118 "for COFF and MachO object files.\n"; 2119 } 2120 2121 void llvm::printExportsTrie(const ObjectFile *o) { 2122 outs() << "Exports trie:\n"; 2123 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2124 printMachOExportsTrie(MachO); 2125 else 2126 WithColor::error(errs(), ToolName) 2127 << "This operation is only currently supported " 2128 "for Mach-O executable files.\n"; 2129 } 2130 2131 void llvm::printRebaseTable(ObjectFile *o) { 2132 outs() << "Rebase table:\n"; 2133 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2134 printMachORebaseTable(MachO); 2135 else 2136 WithColor::error(errs(), ToolName) 2137 << "This operation is only currently supported " 2138 "for Mach-O executable files.\n"; 2139 } 2140 2141 void llvm::printBindTable(ObjectFile *o) { 2142 outs() << "Bind table:\n"; 2143 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2144 printMachOBindTable(MachO); 2145 else 2146 WithColor::error(errs(), ToolName) 2147 << "This operation is only currently supported " 2148 "for Mach-O executable files.\n"; 2149 } 2150 2151 void llvm::printLazyBindTable(ObjectFile *o) { 2152 outs() << "Lazy bind table:\n"; 2153 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2154 printMachOLazyBindTable(MachO); 2155 else 2156 WithColor::error(errs(), ToolName) 2157 << "This operation is only currently supported " 2158 "for Mach-O executable files.\n"; 2159 } 2160 2161 void llvm::printWeakBindTable(ObjectFile *o) { 2162 outs() << "Weak bind table:\n"; 2163 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2164 printMachOWeakBindTable(MachO); 2165 else 2166 WithColor::error(errs(), ToolName) 2167 << "This operation is only currently supported " 2168 "for Mach-O executable files.\n"; 2169 } 2170 2171 /// Dump the raw contents of the __clangast section so the output can be piped 2172 /// into llvm-bcanalyzer. 2173 void llvm::printRawClangAST(const ObjectFile *Obj) { 2174 if (outs().is_displayed()) { 2175 WithColor::error(errs(), ToolName) 2176 << "The -raw-clang-ast option will dump the raw binary contents of " 2177 "the clang ast section.\n" 2178 "Please redirect the output to a file or another program such as " 2179 "llvm-bcanalyzer.\n"; 2180 return; 2181 } 2182 2183 StringRef ClangASTSectionName("__clangast"); 2184 if (isa<COFFObjectFile>(Obj)) { 2185 ClangASTSectionName = "clangast"; 2186 } 2187 2188 Optional<object::SectionRef> ClangASTSection; 2189 for (auto Sec : ToolSectionFilter(*Obj)) { 2190 StringRef Name; 2191 Sec.getName(Name); 2192 if (Name == ClangASTSectionName) { 2193 ClangASTSection = Sec; 2194 break; 2195 } 2196 } 2197 if (!ClangASTSection) 2198 return; 2199 2200 StringRef ClangASTContents; 2201 error(ClangASTSection.getValue().getContents(ClangASTContents)); 2202 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2203 } 2204 2205 static void printFaultMaps(const ObjectFile *Obj) { 2206 StringRef FaultMapSectionName; 2207 2208 if (isa<ELFObjectFileBase>(Obj)) { 2209 FaultMapSectionName = ".llvm_faultmaps"; 2210 } else if (isa<MachOObjectFile>(Obj)) { 2211 FaultMapSectionName = "__llvm_faultmaps"; 2212 } else { 2213 WithColor::error(errs(), ToolName) 2214 << "This operation is only currently supported " 2215 "for ELF and Mach-O executable files.\n"; 2216 return; 2217 } 2218 2219 Optional<object::SectionRef> FaultMapSection; 2220 2221 for (auto Sec : ToolSectionFilter(*Obj)) { 2222 StringRef Name; 2223 Sec.getName(Name); 2224 if (Name == FaultMapSectionName) { 2225 FaultMapSection = Sec; 2226 break; 2227 } 2228 } 2229 2230 outs() << "FaultMap table:\n"; 2231 2232 if (!FaultMapSection.hasValue()) { 2233 outs() << "<not found>\n"; 2234 return; 2235 } 2236 2237 StringRef FaultMapContents; 2238 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2239 2240 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2241 FaultMapContents.bytes_end()); 2242 2243 outs() << FMP; 2244 } 2245 2246 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 2247 if (O->isELF()) { 2248 printELFFileHeader(O); 2249 return printELFDynamicSection(O); 2250 } 2251 if (O->isCOFF()) 2252 return printCOFFFileHeader(O); 2253 if (O->isWasm()) 2254 return printWasmFileHeader(O); 2255 if (O->isMachO()) { 2256 printMachOFileHeader(O); 2257 if (!OnlyFirst) 2258 printMachOLoadCommands(O); 2259 return; 2260 } 2261 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 2262 } 2263 2264 static void printFileHeaders(const ObjectFile *O) { 2265 if (!O->isELF() && !O->isCOFF()) 2266 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 2267 2268 Triple::ArchType AT = O->getArch(); 2269 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2270 Expected<uint64_t> StartAddrOrErr = O->getStartAddress(); 2271 if (!StartAddrOrErr) 2272 report_error(O->getFileName(), StartAddrOrErr.takeError()); 2273 2274 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2275 uint64_t Address = StartAddrOrErr.get(); 2276 outs() << "start address: " 2277 << "0x" << format(Fmt.data(), Address) << "\n\n"; 2278 } 2279 2280 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2281 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2282 if (!ModeOrErr) { 2283 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 2284 consumeError(ModeOrErr.takeError()); 2285 return; 2286 } 2287 sys::fs::perms Mode = ModeOrErr.get(); 2288 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2289 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2290 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2291 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2292 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2293 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2294 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2295 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2296 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2297 2298 outs() << " "; 2299 2300 Expected<unsigned> UIDOrErr = C.getUID(); 2301 if (!UIDOrErr) 2302 report_error(Filename, UIDOrErr.takeError()); 2303 unsigned UID = UIDOrErr.get(); 2304 outs() << format("%d/", UID); 2305 2306 Expected<unsigned> GIDOrErr = C.getGID(); 2307 if (!GIDOrErr) 2308 report_error(Filename, GIDOrErr.takeError()); 2309 unsigned GID = GIDOrErr.get(); 2310 outs() << format("%-d ", GID); 2311 2312 Expected<uint64_t> Size = C.getRawSize(); 2313 if (!Size) 2314 report_error(Filename, Size.takeError()); 2315 outs() << format("%6" PRId64, Size.get()) << " "; 2316 2317 StringRef RawLastModified = C.getRawLastModified(); 2318 unsigned Seconds; 2319 if (RawLastModified.getAsInteger(10, Seconds)) 2320 outs() << "(date: \"" << RawLastModified 2321 << "\" contains non-decimal chars) "; 2322 else { 2323 // Since ctime(3) returns a 26 character string of the form: 2324 // "Sun Sep 16 01:03:52 1973\n\0" 2325 // just print 24 characters. 2326 time_t t = Seconds; 2327 outs() << format("%.24s ", ctime(&t)); 2328 } 2329 2330 StringRef Name = ""; 2331 Expected<StringRef> NameOrErr = C.getName(); 2332 if (!NameOrErr) { 2333 consumeError(NameOrErr.takeError()); 2334 Expected<StringRef> RawNameOrErr = C.getRawName(); 2335 if (!RawNameOrErr) 2336 report_error(Filename, NameOrErr.takeError()); 2337 Name = RawNameOrErr.get(); 2338 } else { 2339 Name = NameOrErr.get(); 2340 } 2341 outs() << Name << "\n"; 2342 } 2343 2344 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 2345 const Archive::Child *C = nullptr) { 2346 // Avoid other output when using a raw option. 2347 if (!RawClangAST) { 2348 outs() << '\n'; 2349 if (A) 2350 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 2351 else 2352 outs() << O->getFileName(); 2353 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 2354 } 2355 2356 StringRef ArchiveName = A ? A->getFileName() : ""; 2357 if (FileHeaders) 2358 printFileHeaders(O); 2359 if (ArchiveHeaders && !MachOOpt && C) 2360 printArchiveChild(ArchiveName, *C); 2361 if (Disassemble) 2362 disassembleObject(O, Relocations); 2363 if (Relocations && !Disassemble) 2364 printRelocations(O); 2365 if (DynamicRelocations) 2366 printDynamicRelocations(O); 2367 if (SectionHeaders) 2368 printSectionHeaders(O); 2369 if (SectionContents) 2370 printSectionContents(O); 2371 if (SymbolTable) 2372 printSymbolTable(O, ArchiveName); 2373 if (UnwindInfo) 2374 printUnwindInfo(O); 2375 if (PrivateHeaders || FirstPrivateHeader) 2376 printPrivateFileHeaders(O, FirstPrivateHeader); 2377 if (ExportsTrie) 2378 printExportsTrie(O); 2379 if (Rebase) 2380 printRebaseTable(O); 2381 if (Bind) 2382 printBindTable(O); 2383 if (LazyBind) 2384 printLazyBindTable(O); 2385 if (WeakBind) 2386 printWeakBindTable(O); 2387 if (RawClangAST) 2388 printRawClangAST(O); 2389 if (PrintFaultMaps) 2390 printFaultMaps(O); 2391 if (DwarfDumpType != DIDT_Null) { 2392 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2393 // Dump the complete DWARF structure. 2394 DIDumpOptions DumpOpts; 2395 DumpOpts.DumpType = DwarfDumpType; 2396 DICtx->dump(outs(), DumpOpts); 2397 } 2398 } 2399 2400 static void dumpObject(const COFFImportFile *I, const Archive *A, 2401 const Archive::Child *C = nullptr) { 2402 StringRef ArchiveName = A ? A->getFileName() : ""; 2403 2404 // Avoid other output when using a raw option. 2405 if (!RawClangAST) 2406 outs() << '\n' 2407 << ArchiveName << "(" << I->getFileName() << ")" 2408 << ":\tfile format COFF-import-file" 2409 << "\n\n"; 2410 2411 if (ArchiveHeaders && !MachOOpt && C) 2412 printArchiveChild(ArchiveName, *C); 2413 if (SymbolTable) 2414 printCOFFSymbolTable(I); 2415 } 2416 2417 /// Dump each object file in \a a; 2418 static void dumpArchive(const Archive *A) { 2419 Error Err = Error::success(); 2420 for (auto &C : A->children(Err)) { 2421 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2422 if (!ChildOrErr) { 2423 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2424 report_error(A->getFileName(), C, std::move(E)); 2425 continue; 2426 } 2427 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2428 dumpObject(O, A, &C); 2429 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2430 dumpObject(I, A, &C); 2431 else 2432 report_error(A->getFileName(), object_error::invalid_file_type); 2433 } 2434 if (Err) 2435 report_error(A->getFileName(), std::move(Err)); 2436 } 2437 2438 /// Open file and figure out how to dump it. 2439 static void dumpInput(StringRef file) { 2440 // If we are using the Mach-O specific object file parser, then let it parse 2441 // the file and process the command line options. So the -arch flags can 2442 // be used to select specific slices, etc. 2443 if (MachOOpt) { 2444 parseInputMachO(file); 2445 return; 2446 } 2447 2448 // Attempt to open the binary. 2449 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2450 if (!BinaryOrErr) 2451 report_error(file, BinaryOrErr.takeError()); 2452 Binary &Binary = *BinaryOrErr.get().getBinary(); 2453 2454 if (Archive *A = dyn_cast<Archive>(&Binary)) 2455 dumpArchive(A); 2456 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2457 dumpObject(O); 2458 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2459 parseInputMachO(UB); 2460 else 2461 report_error(file, object_error::invalid_file_type); 2462 } 2463 2464 int main(int argc, char **argv) { 2465 InitLLVM X(argc, argv); 2466 2467 // Initialize targets and assembly printers/parsers. 2468 llvm::InitializeAllTargetInfos(); 2469 llvm::InitializeAllTargetMCs(); 2470 llvm::InitializeAllDisassemblers(); 2471 2472 // Register the target printer for --version. 2473 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2474 2475 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2476 2477 ToolName = argv[0]; 2478 2479 // Defaults to a.out if no filenames specified. 2480 if (InputFilenames.empty()) 2481 InputFilenames.push_back("a.out"); 2482 2483 if (AllHeaders) 2484 FileHeaders = PrivateHeaders = Relocations = SectionHeaders = SymbolTable = 2485 true; 2486 2487 if (DisassembleAll || PrintSource || PrintLines) 2488 Disassemble = true; 2489 2490 if (!Disassemble 2491 && !Relocations 2492 && !DynamicRelocations 2493 && !SectionHeaders 2494 && !SectionContents 2495 && !SymbolTable 2496 && !UnwindInfo 2497 && !PrivateHeaders 2498 && !FileHeaders 2499 && !FirstPrivateHeader 2500 && !ExportsTrie 2501 && !Rebase 2502 && !Bind 2503 && !LazyBind 2504 && !WeakBind 2505 && !RawClangAST 2506 && !(UniversalHeaders && MachOOpt) 2507 && !ArchiveHeaders 2508 && !(IndirectSymbols && MachOOpt) 2509 && !(DataInCode && MachOOpt) 2510 && !(LinkOptHints && MachOOpt) 2511 && !(InfoPlist && MachOOpt) 2512 && !(DylibsUsed && MachOOpt) 2513 && !(DylibId && MachOOpt) 2514 && !(ObjcMetaData && MachOOpt) 2515 && !(!FilterSections.empty() && MachOOpt) 2516 && !PrintFaultMaps 2517 && DwarfDumpType == DIDT_Null) { 2518 cl::PrintHelpMessage(); 2519 return 2; 2520 } 2521 2522 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2523 DisassembleFunctions.end()); 2524 2525 llvm::for_each(InputFilenames, dumpInput); 2526 2527 return EXIT_SUCCESS; 2528 } 2529