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 } 436 } else { 437 TheTriple.setTriple(Triple::normalize(TripleName)); 438 439 // Use the triple, but also try to combine with ARM build attributes. 440 if (Obj) { 441 auto Arch = Obj->getArch(); 442 if (Arch == Triple::arm || Arch == Triple::armeb) { 443 Obj->setARMSubArch(TheTriple); 444 } 445 } 446 } 447 448 // Get the target specific parser. 449 std::string Error; 450 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 451 Error); 452 if (!TheTarget) { 453 if (Obj) 454 report_error(Obj->getFileName(), "can't find target: " + Error); 455 else 456 error("can't find target: " + Error); 457 } 458 459 // Update the triple name and return the found target. 460 TripleName = TheTriple.getTriple(); 461 return TheTarget; 462 } 463 464 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 465 return a.getOffset() < b.getOffset(); 466 } 467 468 static std::string demangle(StringRef Name) { 469 char *Demangled = nullptr; 470 if (Name.startswith("_Z")) 471 Demangled = itaniumDemangle(Name.data(), Demangled, nullptr, nullptr); 472 else if (Name.startswith("?")) 473 Demangled = microsoftDemangle(Name.data(), Demangled, nullptr, nullptr); 474 475 if (!Demangled) 476 return Name; 477 478 std::string Ret = Demangled; 479 free(Demangled); 480 return Ret; 481 } 482 483 template <class ELFT> 484 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 485 const RelocationRef &RelRef, 486 SmallVectorImpl<char> &Result) { 487 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 488 489 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 490 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 491 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 492 493 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 494 495 auto SecOrErr = EF.getSection(Rel.d.a); 496 if (!SecOrErr) 497 return errorToErrorCode(SecOrErr.takeError()); 498 const Elf_Shdr *Sec = *SecOrErr; 499 auto SymTabOrErr = EF.getSection(Sec->sh_link); 500 if (!SymTabOrErr) 501 return errorToErrorCode(SymTabOrErr.takeError()); 502 const Elf_Shdr *SymTab = *SymTabOrErr; 503 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 504 SymTab->sh_type == ELF::SHT_DYNSYM); 505 auto StrTabSec = EF.getSection(SymTab->sh_link); 506 if (!StrTabSec) 507 return errorToErrorCode(StrTabSec.takeError()); 508 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 509 if (!StrTabOrErr) 510 return errorToErrorCode(StrTabOrErr.takeError()); 511 StringRef StrTab = *StrTabOrErr; 512 int64_t addend = 0; 513 // If there is no Symbol associated with the relocation, we set the undef 514 // boolean value to 'true'. This will prevent us from calling functions that 515 // requires the relocation to be associated with a symbol. 516 bool undef = false; 517 switch (Sec->sh_type) { 518 default: 519 return object_error::parse_failed; 520 case ELF::SHT_REL: { 521 // TODO: Read implicit addend from section data. 522 break; 523 } 524 case ELF::SHT_RELA: { 525 const Elf_Rela *ERela = Obj->getRela(Rel); 526 addend = ERela->r_addend; 527 undef = ERela->getSymbol(false) == 0; 528 break; 529 } 530 } 531 std::string Target; 532 if (!undef) { 533 symbol_iterator SI = RelRef.getSymbol(); 534 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 535 if (symb->getType() == ELF::STT_SECTION) { 536 Expected<section_iterator> SymSI = SI->getSection(); 537 if (!SymSI) 538 return errorToErrorCode(SymSI.takeError()); 539 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 540 auto SecName = EF.getSectionName(SymSec); 541 if (!SecName) 542 return errorToErrorCode(SecName.takeError()); 543 Target = *SecName; 544 } else { 545 Expected<StringRef> SymName = symb->getName(StrTab); 546 if (!SymName) 547 return errorToErrorCode(SymName.takeError()); 548 if (Demangle) 549 Target = demangle(*SymName); 550 else 551 Target = *SymName; 552 } 553 } else 554 Target = "*ABS*"; 555 556 // Default scheme is to print Target, as well as "+ <addend>" for nonzero 557 // addend. Should be acceptable for all normal purposes. 558 std::string fmtbuf; 559 raw_string_ostream fmt(fmtbuf); 560 fmt << Target; 561 if (addend != 0) 562 fmt << (addend < 0 ? "" : "+") << addend; 563 fmt.flush(); 564 Result.append(fmtbuf.begin(), fmtbuf.end()); 565 return std::error_code(); 566 } 567 568 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 569 const RelocationRef &Rel, 570 SmallVectorImpl<char> &Result) { 571 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 572 return getRelocationValueString(ELF32LE, Rel, Result); 573 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 574 return getRelocationValueString(ELF64LE, Rel, Result); 575 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 576 return getRelocationValueString(ELF32BE, Rel, Result); 577 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 578 return getRelocationValueString(ELF64BE, Rel, Result); 579 } 580 581 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 582 const RelocationRef &Rel, 583 SmallVectorImpl<char> &Result) { 584 symbol_iterator SymI = Rel.getSymbol(); 585 Expected<StringRef> SymNameOrErr = SymI->getName(); 586 if (!SymNameOrErr) 587 return errorToErrorCode(SymNameOrErr.takeError()); 588 StringRef SymName = *SymNameOrErr; 589 Result.append(SymName.begin(), SymName.end()); 590 return std::error_code(); 591 } 592 593 static void printRelocationTargetName(const MachOObjectFile *O, 594 const MachO::any_relocation_info &RE, 595 raw_string_ostream &fmt) { 596 bool IsScattered = O->isRelocationScattered(RE); 597 598 // Target of a scattered relocation is an address. In the interest of 599 // generating pretty output, scan through the symbol table looking for a 600 // symbol that aligns with that address. If we find one, print it. 601 // Otherwise, we just print the hex address of the target. 602 if (IsScattered) { 603 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 604 605 for (const SymbolRef &Symbol : O->symbols()) { 606 std::error_code ec; 607 Expected<uint64_t> Addr = Symbol.getAddress(); 608 if (!Addr) 609 report_error(O->getFileName(), Addr.takeError()); 610 if (*Addr != Val) 611 continue; 612 Expected<StringRef> Name = Symbol.getName(); 613 if (!Name) 614 report_error(O->getFileName(), Name.takeError()); 615 fmt << *Name; 616 return; 617 } 618 619 // If we couldn't find a symbol that this relocation refers to, try 620 // to find a section beginning instead. 621 for (const SectionRef &Section : ToolSectionFilter(*O)) { 622 std::error_code ec; 623 624 StringRef Name; 625 uint64_t Addr = Section.getAddress(); 626 if (Addr != Val) 627 continue; 628 if ((ec = Section.getName(Name))) 629 report_error(O->getFileName(), ec); 630 fmt << Name; 631 return; 632 } 633 634 fmt << format("0x%x", Val); 635 return; 636 } 637 638 StringRef S; 639 bool isExtern = O->getPlainRelocationExternal(RE); 640 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 641 642 if (O->getAnyRelocationType(RE) == MachO::ARM64_RELOC_ADDEND) { 643 fmt << format("0x%0" PRIx64, Val); 644 return; 645 } else if (isExtern) { 646 symbol_iterator SI = O->symbol_begin(); 647 advance(SI, Val); 648 Expected<StringRef> SOrErr = SI->getName(); 649 if (!SOrErr) 650 report_error(O->getFileName(), SOrErr.takeError()); 651 S = *SOrErr; 652 } else { 653 section_iterator SI = O->section_begin(); 654 // Adjust for the fact that sections are 1-indexed. 655 if (Val == 0) { 656 fmt << "0 (?,?)"; 657 return; 658 } 659 uint32_t i = Val - 1; 660 while (i != 0 && SI != O->section_end()) { 661 i--; 662 advance(SI, 1); 663 } 664 if (SI == O->section_end()) 665 fmt << Val << " (?,?)"; 666 else 667 SI->getName(S); 668 } 669 670 fmt << S; 671 } 672 673 static std::error_code getRelocationValueString(const WasmObjectFile *Obj, 674 const RelocationRef &RelRef, 675 SmallVectorImpl<char> &Result) { 676 const wasm::WasmRelocation& Rel = Obj->getWasmRelocation(RelRef); 677 symbol_iterator SI = RelRef.getSymbol(); 678 std::string fmtbuf; 679 raw_string_ostream fmt(fmtbuf); 680 if (SI == Obj->symbol_end()) { 681 // Not all wasm relocations have symbols associated with them. 682 // In particular R_WEBASSEMBLY_TYPE_INDEX_LEB. 683 fmt << Rel.Index; 684 } else { 685 Expected<StringRef> SymNameOrErr = SI->getName(); 686 if (!SymNameOrErr) 687 return errorToErrorCode(SymNameOrErr.takeError()); 688 StringRef SymName = *SymNameOrErr; 689 Result.append(SymName.begin(), SymName.end()); 690 } 691 fmt << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 692 fmt.flush(); 693 Result.append(fmtbuf.begin(), fmtbuf.end()); 694 return std::error_code(); 695 } 696 697 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 698 const RelocationRef &RelRef, 699 SmallVectorImpl<char> &Result) { 700 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 701 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 702 703 unsigned Arch = Obj->getArch(); 704 705 std::string fmtbuf; 706 raw_string_ostream fmt(fmtbuf); 707 unsigned Type = Obj->getAnyRelocationType(RE); 708 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 709 710 // Determine any addends that should be displayed with the relocation. 711 // These require decoding the relocation type, which is triple-specific. 712 713 // X86_64 has entirely custom relocation types. 714 if (Arch == Triple::x86_64) { 715 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 716 717 switch (Type) { 718 case MachO::X86_64_RELOC_GOT_LOAD: 719 case MachO::X86_64_RELOC_GOT: { 720 printRelocationTargetName(Obj, RE, fmt); 721 fmt << "@GOT"; 722 if (isPCRel) 723 fmt << "PCREL"; 724 break; 725 } 726 case MachO::X86_64_RELOC_SUBTRACTOR: { 727 DataRefImpl RelNext = Rel; 728 Obj->moveRelocationNext(RelNext); 729 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 730 731 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 732 // X86_64_RELOC_UNSIGNED. 733 // NOTE: Scattered relocations don't exist on x86_64. 734 unsigned RType = Obj->getAnyRelocationType(RENext); 735 if (RType != MachO::X86_64_RELOC_UNSIGNED) 736 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 737 "X86_64_RELOC_SUBTRACTOR."); 738 739 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 740 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 741 printRelocationTargetName(Obj, RENext, fmt); 742 fmt << "-"; 743 printRelocationTargetName(Obj, RE, fmt); 744 break; 745 } 746 case MachO::X86_64_RELOC_TLV: 747 printRelocationTargetName(Obj, RE, fmt); 748 fmt << "@TLV"; 749 if (isPCRel) 750 fmt << "P"; 751 break; 752 case MachO::X86_64_RELOC_SIGNED_1: 753 printRelocationTargetName(Obj, RE, fmt); 754 fmt << "-1"; 755 break; 756 case MachO::X86_64_RELOC_SIGNED_2: 757 printRelocationTargetName(Obj, RE, fmt); 758 fmt << "-2"; 759 break; 760 case MachO::X86_64_RELOC_SIGNED_4: 761 printRelocationTargetName(Obj, RE, fmt); 762 fmt << "-4"; 763 break; 764 default: 765 printRelocationTargetName(Obj, RE, fmt); 766 break; 767 } 768 // X86 and ARM share some relocation types in common. 769 } else if (Arch == Triple::x86 || Arch == Triple::arm || 770 Arch == Triple::ppc) { 771 // Generic relocation types... 772 switch (Type) { 773 case MachO::GENERIC_RELOC_PAIR: // prints no info 774 return std::error_code(); 775 case MachO::GENERIC_RELOC_SECTDIFF: { 776 DataRefImpl RelNext = Rel; 777 Obj->moveRelocationNext(RelNext); 778 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 779 780 // X86 sect diff's must be followed by a relocation of type 781 // GENERIC_RELOC_PAIR. 782 unsigned RType = Obj->getAnyRelocationType(RENext); 783 784 if (RType != MachO::GENERIC_RELOC_PAIR) 785 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 786 "GENERIC_RELOC_SECTDIFF."); 787 788 printRelocationTargetName(Obj, RE, fmt); 789 fmt << "-"; 790 printRelocationTargetName(Obj, RENext, fmt); 791 break; 792 } 793 } 794 795 if (Arch == Triple::x86 || Arch == Triple::ppc) { 796 switch (Type) { 797 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 798 DataRefImpl RelNext = Rel; 799 Obj->moveRelocationNext(RelNext); 800 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 801 802 // X86 sect diff's must be followed by a relocation of type 803 // GENERIC_RELOC_PAIR. 804 unsigned RType = Obj->getAnyRelocationType(RENext); 805 if (RType != MachO::GENERIC_RELOC_PAIR) 806 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 807 "GENERIC_RELOC_LOCAL_SECTDIFF."); 808 809 printRelocationTargetName(Obj, RE, fmt); 810 fmt << "-"; 811 printRelocationTargetName(Obj, RENext, fmt); 812 break; 813 } 814 case MachO::GENERIC_RELOC_TLV: { 815 printRelocationTargetName(Obj, RE, fmt); 816 fmt << "@TLV"; 817 if (IsPCRel) 818 fmt << "P"; 819 break; 820 } 821 default: 822 printRelocationTargetName(Obj, RE, fmt); 823 } 824 } else { // ARM-specific relocations 825 switch (Type) { 826 case MachO::ARM_RELOC_HALF: 827 case MachO::ARM_RELOC_HALF_SECTDIFF: { 828 // Half relocations steal a bit from the length field to encode 829 // whether this is an upper16 or a lower16 relocation. 830 bool isUpper = (Obj->getAnyRelocationLength(RE) & 0x1) == 1; 831 832 if (isUpper) 833 fmt << ":upper16:("; 834 else 835 fmt << ":lower16:("; 836 printRelocationTargetName(Obj, RE, fmt); 837 838 DataRefImpl RelNext = Rel; 839 Obj->moveRelocationNext(RelNext); 840 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 841 842 // ARM half relocs must be followed by a relocation of type 843 // ARM_RELOC_PAIR. 844 unsigned RType = Obj->getAnyRelocationType(RENext); 845 if (RType != MachO::ARM_RELOC_PAIR) 846 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 847 "ARM_RELOC_HALF"); 848 849 // NOTE: The half of the target virtual address is stashed in the 850 // address field of the secondary relocation, but we can't reverse 851 // engineer the constant offset from it without decoding the movw/movt 852 // instruction to find the other half in its immediate field. 853 854 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 855 // symbol/section pointer of the follow-on relocation. 856 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 857 fmt << "-"; 858 printRelocationTargetName(Obj, RENext, fmt); 859 } 860 861 fmt << ")"; 862 break; 863 } 864 default: { printRelocationTargetName(Obj, RE, fmt); } 865 } 866 } 867 } else 868 printRelocationTargetName(Obj, RE, fmt); 869 870 fmt.flush(); 871 Result.append(fmtbuf.begin(), fmtbuf.end()); 872 return std::error_code(); 873 } 874 875 static std::error_code getRelocationValueString(const RelocationRef &Rel, 876 SmallVectorImpl<char> &Result) { 877 const ObjectFile *Obj = Rel.getObject(); 878 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 879 return getRelocationValueString(ELF, Rel, Result); 880 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 881 return getRelocationValueString(COFF, Rel, Result); 882 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 883 return getRelocationValueString(Wasm, Rel, Result); 884 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 885 return getRelocationValueString(MachO, Rel, Result); 886 llvm_unreachable("unknown object file format"); 887 } 888 889 /// Indicates whether this relocation should hidden when listing 890 /// relocations, usually because it is the trailing part of a multipart 891 /// relocation that will be printed as part of the leading relocation. 892 static bool getHidden(RelocationRef RelRef) { 893 const ObjectFile *Obj = RelRef.getObject(); 894 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 895 if (!MachO) 896 return false; 897 898 unsigned Arch = MachO->getArch(); 899 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 900 uint64_t Type = MachO->getRelocationType(Rel); 901 902 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 903 // is always hidden. 904 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 905 if (Type == MachO::GENERIC_RELOC_PAIR) 906 return true; 907 } else if (Arch == Triple::x86_64) { 908 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 909 // an X86_64_RELOC_SUBTRACTOR. 910 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 911 DataRefImpl RelPrev = Rel; 912 RelPrev.d.a--; 913 uint64_t PrevType = MachO->getRelocationType(RelPrev); 914 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 915 return true; 916 } 917 } 918 919 return false; 920 } 921 922 namespace { 923 class SourcePrinter { 924 protected: 925 DILineInfo OldLineInfo; 926 const ObjectFile *Obj = nullptr; 927 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 928 // File name to file contents of source 929 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 930 // Mark the line endings of the cached source 931 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 932 933 private: 934 bool cacheSource(const DILineInfo& LineInfoFile); 935 936 public: 937 SourcePrinter() = default; 938 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 939 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 940 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 941 DefaultArch); 942 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 943 } 944 virtual ~SourcePrinter() = default; 945 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 946 StringRef Delimiter = "; "); 947 }; 948 949 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 950 std::unique_ptr<MemoryBuffer> Buffer; 951 if (LineInfo.Source) { 952 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 953 } else { 954 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 955 if (!BufferOrError) 956 return false; 957 Buffer = std::move(*BufferOrError); 958 } 959 // Chomp the file to get lines 960 size_t BufferSize = Buffer->getBufferSize(); 961 const char *BufferStart = Buffer->getBufferStart(); 962 for (const char *Start = BufferStart, *End = BufferStart; 963 End < BufferStart + BufferSize; End++) 964 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 965 (*End == '\r' && *(End + 1) == '\n')) { 966 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 967 if (*End == '\r') 968 End++; 969 Start = End + 1; 970 } 971 SourceCache[LineInfo.FileName] = std::move(Buffer); 972 return true; 973 } 974 975 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 976 StringRef Delimiter) { 977 if (!Symbolizer) 978 return; 979 DILineInfo LineInfo = DILineInfo(); 980 auto ExpectecLineInfo = 981 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 982 if (!ExpectecLineInfo) 983 consumeError(ExpectecLineInfo.takeError()); 984 else 985 LineInfo = *ExpectecLineInfo; 986 987 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 988 LineInfo.Line == 0) 989 return; 990 991 if (PrintLines) 992 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 993 if (PrintSource) { 994 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 995 if (!cacheSource(LineInfo)) 996 return; 997 auto FileBuffer = SourceCache.find(LineInfo.FileName); 998 if (FileBuffer != SourceCache.end()) { 999 auto LineBuffer = LineCache.find(LineInfo.FileName); 1000 if (LineBuffer != LineCache.end()) { 1001 if (LineInfo.Line > LineBuffer->second.size()) 1002 return; 1003 // Vector begins at 0, line numbers are non-zero 1004 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 1005 << "\n"; 1006 } 1007 } 1008 } 1009 OldLineInfo = LineInfo; 1010 } 1011 1012 static bool isArmElf(const ObjectFile *Obj) { 1013 return (Obj->isELF() && 1014 (Obj->getArch() == Triple::aarch64 || 1015 Obj->getArch() == Triple::aarch64_be || 1016 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 1017 Obj->getArch() == Triple::thumb || 1018 Obj->getArch() == Triple::thumbeb)); 1019 } 1020 1021 class PrettyPrinter { 1022 public: 1023 virtual ~PrettyPrinter() = default; 1024 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 1025 ArrayRef<uint8_t> Bytes, uint64_t Address, 1026 raw_ostream &OS, StringRef Annot, 1027 MCSubtargetInfo const &STI, SourcePrinter *SP, 1028 std::vector<RelocationRef> *Rels = nullptr) { 1029 if (SP && (PrintSource || PrintLines)) 1030 SP->printSourceLine(OS, Address); 1031 if (!NoLeadingAddr) 1032 OS << format("%8" PRIx64 ":", Address); 1033 if (!NoShowRawInsn) { 1034 OS << "\t"; 1035 dumpBytes(Bytes, OS); 1036 } 1037 if (MI) 1038 IP.printInst(MI, OS, "", STI); 1039 else 1040 OS << " <unknown>"; 1041 } 1042 }; 1043 PrettyPrinter PrettyPrinterInst; 1044 class HexagonPrettyPrinter : public PrettyPrinter { 1045 public: 1046 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 1047 raw_ostream &OS) { 1048 uint32_t opcode = 1049 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 1050 if (!NoLeadingAddr) 1051 OS << format("%8" PRIx64 ":", Address); 1052 if (!NoShowRawInsn) { 1053 OS << "\t"; 1054 dumpBytes(Bytes.slice(0, 4), OS); 1055 OS << format("%08" PRIx32, opcode); 1056 } 1057 } 1058 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1059 uint64_t Address, raw_ostream &OS, StringRef Annot, 1060 MCSubtargetInfo const &STI, SourcePrinter *SP, 1061 std::vector<RelocationRef> *Rels) override { 1062 if (SP && (PrintSource || PrintLines)) 1063 SP->printSourceLine(OS, Address, ""); 1064 if (!MI) { 1065 printLead(Bytes, Address, OS); 1066 OS << " <unknown>"; 1067 return; 1068 } 1069 std::string Buffer; 1070 { 1071 raw_string_ostream TempStream(Buffer); 1072 IP.printInst(MI, TempStream, "", STI); 1073 } 1074 StringRef Contents(Buffer); 1075 // Split off bundle attributes 1076 auto PacketBundle = Contents.rsplit('\n'); 1077 // Split off first instruction from the rest 1078 auto HeadTail = PacketBundle.first.split('\n'); 1079 auto Preamble = " { "; 1080 auto Separator = ""; 1081 StringRef Fmt = "\t\t\t%08" PRIx64 ": "; 1082 std::vector<RelocationRef>::const_iterator rel_cur = Rels->begin(); 1083 std::vector<RelocationRef>::const_iterator rel_end = Rels->end(); 1084 1085 // Hexagon's packets require relocations to be inline rather than 1086 // clustered at the end of the packet. 1087 auto PrintReloc = [&]() -> void { 1088 while ((rel_cur != rel_end) && (rel_cur->getOffset() <= Address)) { 1089 if (rel_cur->getOffset() == Address) { 1090 SmallString<16> name; 1091 SmallString<32> val; 1092 rel_cur->getTypeName(name); 1093 error(getRelocationValueString(*rel_cur, val)); 1094 OS << Separator << format(Fmt.data(), Address) << name << "\t" << val 1095 << "\n"; 1096 return; 1097 } 1098 rel_cur++; 1099 } 1100 }; 1101 1102 while(!HeadTail.first.empty()) { 1103 OS << Separator; 1104 Separator = "\n"; 1105 if (SP && (PrintSource || PrintLines)) 1106 SP->printSourceLine(OS, Address, ""); 1107 printLead(Bytes, Address, OS); 1108 OS << Preamble; 1109 Preamble = " "; 1110 StringRef Inst; 1111 auto Duplex = HeadTail.first.split('\v'); 1112 if(!Duplex.second.empty()){ 1113 OS << Duplex.first; 1114 OS << "; "; 1115 Inst = Duplex.second; 1116 } 1117 else 1118 Inst = HeadTail.first; 1119 OS << Inst; 1120 HeadTail = HeadTail.second.split('\n'); 1121 if (HeadTail.first.empty()) 1122 OS << " } " << PacketBundle.second; 1123 PrintReloc(); 1124 Bytes = Bytes.slice(4); 1125 Address += 4; 1126 } 1127 } 1128 }; 1129 HexagonPrettyPrinter HexagonPrettyPrinterInst; 1130 1131 class AMDGCNPrettyPrinter : public PrettyPrinter { 1132 public: 1133 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1134 uint64_t Address, raw_ostream &OS, StringRef Annot, 1135 MCSubtargetInfo const &STI, SourcePrinter *SP, 1136 std::vector<RelocationRef> *Rels) override { 1137 if (SP && (PrintSource || PrintLines)) 1138 SP->printSourceLine(OS, Address); 1139 1140 typedef support::ulittle32_t U32; 1141 1142 if (MI) { 1143 SmallString<40> InstStr; 1144 raw_svector_ostream IS(InstStr); 1145 1146 IP.printInst(MI, IS, "", STI); 1147 1148 OS << left_justify(IS.str(), 60); 1149 } else { 1150 // an unrecognized encoding - this is probably data so represent it 1151 // using the .long directive, or .byte directive if fewer than 4 bytes 1152 // remaining 1153 if (Bytes.size() >= 4) { 1154 OS << format("\t.long 0x%08" PRIx32 " ", 1155 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 1156 OS.indent(42); 1157 } else { 1158 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 1159 for (unsigned int i = 1; i < Bytes.size(); i++) 1160 OS << format(", 0x%02" PRIx8, Bytes[i]); 1161 OS.indent(55 - (6 * Bytes.size())); 1162 } 1163 } 1164 1165 OS << format("// %012" PRIX64 ": ", Address); 1166 if (Bytes.size() >=4) { 1167 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 1168 Bytes.size() / sizeof(U32))) 1169 // D should be explicitly casted to uint32_t here as it is passed 1170 // by format to snprintf as vararg. 1171 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 1172 } else { 1173 for (unsigned int i = 0; i < Bytes.size(); i++) 1174 OS << format("%02" PRIX8 " ", Bytes[i]); 1175 } 1176 1177 if (!Annot.empty()) 1178 OS << "// " << Annot; 1179 } 1180 }; 1181 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 1182 1183 class BPFPrettyPrinter : public PrettyPrinter { 1184 public: 1185 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1186 uint64_t Address, raw_ostream &OS, StringRef Annot, 1187 MCSubtargetInfo const &STI, SourcePrinter *SP, 1188 std::vector<RelocationRef> *Rels) override { 1189 if (SP && (PrintSource || PrintLines)) 1190 SP->printSourceLine(OS, Address); 1191 if (!NoLeadingAddr) 1192 OS << format("%8" PRId64 ":", Address / 8); 1193 if (!NoShowRawInsn) { 1194 OS << "\t"; 1195 dumpBytes(Bytes, OS); 1196 } 1197 if (MI) 1198 IP.printInst(MI, OS, "", STI); 1199 else 1200 OS << " <unknown>"; 1201 } 1202 }; 1203 BPFPrettyPrinter BPFPrettyPrinterInst; 1204 1205 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 1206 switch(Triple.getArch()) { 1207 default: 1208 return PrettyPrinterInst; 1209 case Triple::hexagon: 1210 return HexagonPrettyPrinterInst; 1211 case Triple::amdgcn: 1212 return AMDGCNPrettyPrinterInst; 1213 case Triple::bpfel: 1214 case Triple::bpfeb: 1215 return BPFPrettyPrinterInst; 1216 } 1217 } 1218 } 1219 1220 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1221 assert(Obj->isELF()); 1222 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1223 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1224 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1225 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1226 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1227 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1228 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1229 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1230 llvm_unreachable("Unsupported binary format"); 1231 } 1232 1233 template <class ELFT> static void 1234 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1235 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1236 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1237 uint8_t SymbolType = Symbol.getELFType(); 1238 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1239 continue; 1240 1241 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1242 if (!AddressOrErr) 1243 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1244 uint64_t Address = *AddressOrErr; 1245 1246 Expected<StringRef> Name = Symbol.getName(); 1247 if (!Name) 1248 report_error(Obj->getFileName(), Name.takeError()); 1249 if (Name->empty()) 1250 continue; 1251 1252 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1253 if (!SectionOrErr) 1254 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1255 section_iterator SecI = *SectionOrErr; 1256 if (SecI == Obj->section_end()) 1257 continue; 1258 1259 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1260 } 1261 } 1262 1263 static void 1264 addDynamicElfSymbols(const ObjectFile *Obj, 1265 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1266 assert(Obj->isELF()); 1267 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1268 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1269 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1270 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1271 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1272 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1273 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1274 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1275 else 1276 llvm_unreachable("Unsupported binary format"); 1277 } 1278 1279 static void addPltEntries(const ObjectFile *Obj, 1280 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 1281 StringSaver &Saver) { 1282 Optional<SectionRef> Plt = None; 1283 for (const SectionRef &Section : Obj->sections()) { 1284 StringRef Name; 1285 if (Section.getName(Name)) 1286 continue; 1287 if (Name == ".plt") 1288 Plt = Section; 1289 } 1290 if (!Plt) 1291 return; 1292 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1293 for (auto PltEntry : ElfObj->getPltAddresses()) { 1294 SymbolRef Symbol(PltEntry.first, ElfObj); 1295 1296 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1297 1298 Expected<StringRef> NameOrErr = Symbol.getName(); 1299 if (!NameOrErr) 1300 report_error(Obj->getFileName(), NameOrErr.takeError()); 1301 if (NameOrErr->empty()) 1302 continue; 1303 StringRef Name = Saver.save((*NameOrErr + "@plt").str()); 1304 1305 AllSymbols[*Plt].emplace_back(PltEntry.second, Name, SymbolType); 1306 } 1307 } 1308 } 1309 1310 // Normally the disassembly output will skip blocks of zeroes. This function 1311 // returns the number of zero bytes that can be skipped when dumping the 1312 // disassembly of the instructions in Buf. 1313 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 1314 // When -z or --disassemble-zeroes are given we always dissasemble them. 1315 if (DisassembleZeroes) 1316 return 0; 1317 1318 // Find the number of leading zeroes. 1319 size_t N = 0; 1320 while (N < Buf.size() && !Buf[N]) 1321 ++N; 1322 1323 // We may want to skip blocks of zero bytes, but unless we see 1324 // at least 8 of them in a row. 1325 if (N < 8) 1326 return 0; 1327 1328 // We skip zeroes in multiples of 4 because do not want to truncate an 1329 // instruction if it starts with a zero byte. 1330 return N & ~0x3; 1331 } 1332 1333 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1334 if (StartAddress > StopAddress) 1335 error("Start address should be less than stop address"); 1336 1337 const Target *TheTarget = getTarget(Obj); 1338 1339 // Package up features to be passed to target/subtarget 1340 SubtargetFeatures Features = Obj->getFeatures(); 1341 if (!MAttrs.empty()) { 1342 for (unsigned i = 0; i != MAttrs.size(); ++i) 1343 Features.AddFeature(MAttrs[i]); 1344 } 1345 1346 std::unique_ptr<const MCRegisterInfo> MRI( 1347 TheTarget->createMCRegInfo(TripleName)); 1348 if (!MRI) 1349 report_error(Obj->getFileName(), "no register info for target " + 1350 TripleName); 1351 1352 // Set up disassembler. 1353 std::unique_ptr<const MCAsmInfo> AsmInfo( 1354 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1355 if (!AsmInfo) 1356 report_error(Obj->getFileName(), "no assembly info for target " + 1357 TripleName); 1358 std::unique_ptr<const MCSubtargetInfo> STI( 1359 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1360 if (!STI) 1361 report_error(Obj->getFileName(), "no subtarget info for target " + 1362 TripleName); 1363 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1364 if (!MII) 1365 report_error(Obj->getFileName(), "no instruction info for target " + 1366 TripleName); 1367 MCObjectFileInfo MOFI; 1368 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1369 // FIXME: for now initialize MCObjectFileInfo with default values 1370 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1371 1372 std::unique_ptr<MCDisassembler> DisAsm( 1373 TheTarget->createMCDisassembler(*STI, Ctx)); 1374 if (!DisAsm) 1375 report_error(Obj->getFileName(), "no disassembler for target " + 1376 TripleName); 1377 1378 std::unique_ptr<const MCInstrAnalysis> MIA( 1379 TheTarget->createMCInstrAnalysis(MII.get())); 1380 1381 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1382 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1383 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1384 if (!IP) 1385 report_error(Obj->getFileName(), "no instruction printer for target " + 1386 TripleName); 1387 IP->setPrintImmHex(PrintImmHex); 1388 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1389 1390 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1391 "\t\t\t%08" PRIx64 ": "; 1392 1393 SourcePrinter SP(Obj, TheTarget->getName()); 1394 1395 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1396 // in RelocSecs contain the relocations for section S. 1397 std::error_code EC; 1398 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1399 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1400 section_iterator Sec2 = Section.getRelocatedSection(); 1401 if (Sec2 != Obj->section_end()) 1402 SectionRelocMap[*Sec2].push_back(Section); 1403 } 1404 1405 // Create a mapping from virtual address to symbol name. This is used to 1406 // pretty print the symbols while disassembling. 1407 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1408 SectionSymbolsTy AbsoluteSymbols; 1409 for (const SymbolRef &Symbol : Obj->symbols()) { 1410 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1411 if (!AddressOrErr) 1412 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1413 uint64_t Address = *AddressOrErr; 1414 1415 Expected<StringRef> Name = Symbol.getName(); 1416 if (!Name) 1417 report_error(Obj->getFileName(), Name.takeError()); 1418 if (Name->empty()) 1419 continue; 1420 1421 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1422 if (!SectionOrErr) 1423 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1424 1425 uint8_t SymbolType = ELF::STT_NOTYPE; 1426 if (Obj->isELF()) 1427 SymbolType = getElfSymbolType(Obj, Symbol); 1428 1429 section_iterator SecI = *SectionOrErr; 1430 if (SecI != Obj->section_end()) 1431 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1432 else 1433 AbsoluteSymbols.emplace_back(Address, *Name, SymbolType); 1434 1435 1436 } 1437 if (AllSymbols.empty() && Obj->isELF()) 1438 addDynamicElfSymbols(Obj, AllSymbols); 1439 1440 BumpPtrAllocator A; 1441 StringSaver Saver(A); 1442 addPltEntries(Obj, AllSymbols, Saver); 1443 1444 // Create a mapping from virtual address to section. 1445 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1446 for (SectionRef Sec : Obj->sections()) 1447 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1448 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1449 1450 // Linked executables (.exe and .dll files) typically don't include a real 1451 // symbol table but they might contain an export table. 1452 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1453 for (const auto &ExportEntry : COFFObj->export_directories()) { 1454 StringRef Name; 1455 error(ExportEntry.getSymbolName(Name)); 1456 if (Name.empty()) 1457 continue; 1458 uint32_t RVA; 1459 error(ExportEntry.getExportRVA(RVA)); 1460 1461 uint64_t VA = COFFObj->getImageBase() + RVA; 1462 auto Sec = std::upper_bound( 1463 SectionAddresses.begin(), SectionAddresses.end(), VA, 1464 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1465 return LHS < RHS.first; 1466 }); 1467 if (Sec != SectionAddresses.begin()) 1468 --Sec; 1469 else 1470 Sec = SectionAddresses.end(); 1471 1472 if (Sec != SectionAddresses.end()) 1473 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1474 else 1475 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1476 } 1477 } 1478 1479 // Sort all the symbols, this allows us to use a simple binary search to find 1480 // a symbol near an address. 1481 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1482 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1483 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1484 1485 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1486 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1487 continue; 1488 1489 uint64_t SectionAddr = Section.getAddress(); 1490 uint64_t SectSize = Section.getSize(); 1491 if (!SectSize) 1492 continue; 1493 1494 // Get the list of all the symbols in this section. 1495 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1496 std::vector<uint64_t> DataMappingSymsAddr; 1497 std::vector<uint64_t> TextMappingSymsAddr; 1498 if (isArmElf(Obj)) { 1499 for (const auto &Symb : Symbols) { 1500 uint64_t Address = std::get<0>(Symb); 1501 StringRef Name = std::get<1>(Symb); 1502 if (Name.startswith("$d")) 1503 DataMappingSymsAddr.push_back(Address - SectionAddr); 1504 if (Name.startswith("$x")) 1505 TextMappingSymsAddr.push_back(Address - SectionAddr); 1506 if (Name.startswith("$a")) 1507 TextMappingSymsAddr.push_back(Address - SectionAddr); 1508 if (Name.startswith("$t")) 1509 TextMappingSymsAddr.push_back(Address - SectionAddr); 1510 } 1511 } 1512 1513 llvm::sort(DataMappingSymsAddr); 1514 llvm::sort(TextMappingSymsAddr); 1515 1516 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1517 // AMDGPU disassembler uses symbolizer for printing labels 1518 std::unique_ptr<MCRelocationInfo> RelInfo( 1519 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1520 if (RelInfo) { 1521 std::unique_ptr<MCSymbolizer> Symbolizer( 1522 TheTarget->createMCSymbolizer( 1523 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1524 DisAsm->setSymbolizer(std::move(Symbolizer)); 1525 } 1526 } 1527 1528 // Make a list of all the relocations for this section. 1529 std::vector<RelocationRef> Rels; 1530 if (InlineRelocs) { 1531 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1532 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1533 Rels.push_back(Reloc); 1534 } 1535 } 1536 } 1537 1538 // Sort relocations by address. 1539 llvm::sort(Rels, RelocAddressLess); 1540 1541 StringRef SegmentName = ""; 1542 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1543 DataRefImpl DR = Section.getRawDataRefImpl(); 1544 SegmentName = MachO->getSectionFinalSegmentName(DR); 1545 } 1546 StringRef SectionName; 1547 error(Section.getName(SectionName)); 1548 1549 // If the section has no symbol at the start, just insert a dummy one. 1550 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1551 Symbols.insert( 1552 Symbols.begin(), 1553 std::make_tuple(SectionAddr, SectionName, 1554 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1555 } 1556 1557 SmallString<40> Comments; 1558 raw_svector_ostream CommentStream(Comments); 1559 1560 StringRef BytesStr; 1561 error(Section.getContents(BytesStr)); 1562 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1563 BytesStr.size()); 1564 1565 uint64_t Size; 1566 uint64_t Index; 1567 bool PrintedSection = false; 1568 1569 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1570 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1571 // Disassemble symbol by symbol. 1572 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1573 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1574 // The end is either the section end or the beginning of the next 1575 // symbol. 1576 uint64_t End = 1577 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1578 // Don't try to disassemble beyond the end of section contents. 1579 if (End > SectSize) 1580 End = SectSize; 1581 // If this symbol has the same address as the next symbol, then skip it. 1582 if (Start >= End) 1583 continue; 1584 1585 // Check if we need to skip symbol 1586 // Skip if the symbol's data is not between StartAddress and StopAddress 1587 if (End + SectionAddr < StartAddress || 1588 Start + SectionAddr > StopAddress) { 1589 continue; 1590 } 1591 1592 /// Skip if user requested specific symbols and this is not in the list 1593 if (!DisasmFuncsSet.empty() && 1594 !DisasmFuncsSet.count(std::get<1>(Symbols[si]))) 1595 continue; 1596 1597 if (!PrintedSection) { 1598 PrintedSection = true; 1599 outs() << "Disassembly of section "; 1600 if (!SegmentName.empty()) 1601 outs() << SegmentName << ","; 1602 outs() << SectionName << ':'; 1603 } 1604 1605 // Stop disassembly at the stop address specified 1606 if (End + SectionAddr > StopAddress) 1607 End = StopAddress - SectionAddr; 1608 1609 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1610 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1611 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1612 Start += 256; 1613 } 1614 if (si == se - 1 || 1615 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1616 // cut trailing zeroes at the end of kernel 1617 // cut up to 256 bytes 1618 const uint64_t EndAlign = 256; 1619 const auto Limit = End - (std::min)(EndAlign, End - Start); 1620 while (End > Limit && 1621 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1622 End -= 4; 1623 } 1624 } 1625 1626 outs() << '\n'; 1627 if (!NoLeadingAddr) 1628 outs() << format("%016" PRIx64 " ", SectionAddr + Start); 1629 1630 StringRef SymbolName = std::get<1>(Symbols[si]); 1631 if (Demangle) 1632 outs() << demangle(SymbolName) << ":\n"; 1633 else 1634 outs() << SymbolName << ":\n"; 1635 1636 // Don't print raw contents of a virtual section. A virtual section 1637 // doesn't have any contents in the file. 1638 if (Section.isVirtual()) { 1639 outs() << "...\n"; 1640 continue; 1641 } 1642 1643 #ifndef NDEBUG 1644 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1645 #else 1646 raw_ostream &DebugOut = nulls(); 1647 #endif 1648 1649 for (Index = Start; Index < End; Index += Size) { 1650 MCInst Inst; 1651 1652 if (Index + SectionAddr < StartAddress || 1653 Index + SectionAddr > StopAddress) { 1654 // skip byte by byte till StartAddress is reached 1655 Size = 1; 1656 continue; 1657 } 1658 // AArch64 ELF binaries can interleave data and text in the 1659 // same section. We rely on the markers introduced to 1660 // understand what we need to dump. If the data marker is within a 1661 // function, it is denoted as a word/short etc 1662 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1663 !DisassembleAll) { 1664 uint64_t Stride = 0; 1665 1666 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1667 DataMappingSymsAddr.end(), Index); 1668 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1669 // Switch to data. 1670 while (Index < End) { 1671 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1672 outs() << "\t"; 1673 if (Index + 4 <= End) { 1674 Stride = 4; 1675 dumpBytes(Bytes.slice(Index, 4), outs()); 1676 outs() << "\t.word\t"; 1677 uint32_t Data = 0; 1678 if (Obj->isLittleEndian()) { 1679 const auto Word = 1680 reinterpret_cast<const support::ulittle32_t *>( 1681 Bytes.data() + Index); 1682 Data = *Word; 1683 } else { 1684 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1685 Bytes.data() + Index); 1686 Data = *Word; 1687 } 1688 outs() << "0x" << format("%08" PRIx32, Data); 1689 } else if (Index + 2 <= End) { 1690 Stride = 2; 1691 dumpBytes(Bytes.slice(Index, 2), outs()); 1692 outs() << "\t\t.short\t"; 1693 uint16_t Data = 0; 1694 if (Obj->isLittleEndian()) { 1695 const auto Short = 1696 reinterpret_cast<const support::ulittle16_t *>( 1697 Bytes.data() + Index); 1698 Data = *Short; 1699 } else { 1700 const auto Short = 1701 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1702 Index); 1703 Data = *Short; 1704 } 1705 outs() << "0x" << format("%04" PRIx16, Data); 1706 } else { 1707 Stride = 1; 1708 dumpBytes(Bytes.slice(Index, 1), outs()); 1709 outs() << "\t\t.byte\t"; 1710 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1711 } 1712 Index += Stride; 1713 outs() << "\n"; 1714 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1715 TextMappingSymsAddr.end(), Index); 1716 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1717 break; 1718 } 1719 } 1720 } 1721 1722 // If there is a data symbol inside an ELF text section and we are only 1723 // disassembling text (applicable all architectures), 1724 // we are in a situation where we must print the data and not 1725 // disassemble it. 1726 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1727 !DisassembleAll && Section.isText()) { 1728 // print out data up to 8 bytes at a time in hex and ascii 1729 uint8_t AsciiData[9] = {'\0'}; 1730 uint8_t Byte; 1731 int NumBytes = 0; 1732 1733 for (Index = Start; Index < End; Index += 1) { 1734 if (((SectionAddr + Index) < StartAddress) || 1735 ((SectionAddr + Index) > StopAddress)) 1736 continue; 1737 if (NumBytes == 0) { 1738 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1739 outs() << "\t"; 1740 } 1741 Byte = Bytes.slice(Index)[0]; 1742 outs() << format(" %02x", Byte); 1743 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1744 1745 uint8_t IndentOffset = 0; 1746 NumBytes++; 1747 if (Index == End - 1 || NumBytes > 8) { 1748 // Indent the space for less than 8 bytes data. 1749 // 2 spaces for byte and one for space between bytes 1750 IndentOffset = 3 * (8 - NumBytes); 1751 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1752 AsciiData[Excess] = '\0'; 1753 NumBytes = 8; 1754 } 1755 if (NumBytes == 8) { 1756 AsciiData[8] = '\0'; 1757 outs() << std::string(IndentOffset, ' ') << " "; 1758 outs() << reinterpret_cast<char *>(AsciiData); 1759 outs() << '\n'; 1760 NumBytes = 0; 1761 } 1762 } 1763 } 1764 if (Index >= End) 1765 break; 1766 1767 if (size_t N = 1768 countSkippableZeroBytes(Bytes.slice(Index, End - Index))) { 1769 outs() << "\t\t..." << '\n'; 1770 Index += N; 1771 if (Index >= End) 1772 break; 1773 } 1774 1775 // Disassemble a real instruction or a data when disassemble all is 1776 // provided 1777 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1778 SectionAddr + Index, DebugOut, 1779 CommentStream); 1780 if (Size == 0) 1781 Size = 1; 1782 1783 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1784 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1785 *STI, &SP, &Rels); 1786 outs() << CommentStream.str(); 1787 Comments.clear(); 1788 1789 // Try to resolve the target of a call, tail call, etc. to a specific 1790 // symbol. 1791 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1792 MIA->isConditionalBranch(Inst))) { 1793 uint64_t Target; 1794 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1795 // In a relocatable object, the target's section must reside in 1796 // the same section as the call instruction or it is accessed 1797 // through a relocation. 1798 // 1799 // In a non-relocatable object, the target may be in any section. 1800 // 1801 // N.B. We don't walk the relocations in the relocatable case yet. 1802 auto *TargetSectionSymbols = &Symbols; 1803 if (!Obj->isRelocatableObject()) { 1804 auto SectionAddress = std::upper_bound( 1805 SectionAddresses.begin(), SectionAddresses.end(), Target, 1806 [](uint64_t LHS, 1807 const std::pair<uint64_t, SectionRef> &RHS) { 1808 return LHS < RHS.first; 1809 }); 1810 if (SectionAddress != SectionAddresses.begin()) { 1811 --SectionAddress; 1812 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1813 } else { 1814 TargetSectionSymbols = &AbsoluteSymbols; 1815 } 1816 } 1817 1818 // Find the first symbol in the section whose offset is less than 1819 // or equal to the target. If there isn't a section that contains 1820 // the target, find the nearest preceding absolute symbol. 1821 auto TargetSym = std::upper_bound( 1822 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1823 Target, [](uint64_t LHS, 1824 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1825 return LHS < std::get<0>(RHS); 1826 }); 1827 if (TargetSym == TargetSectionSymbols->begin()) { 1828 TargetSectionSymbols = &AbsoluteSymbols; 1829 TargetSym = std::upper_bound( 1830 AbsoluteSymbols.begin(), AbsoluteSymbols.end(), 1831 Target, [](uint64_t LHS, 1832 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1833 return LHS < std::get<0>(RHS); 1834 }); 1835 } 1836 if (TargetSym != TargetSectionSymbols->begin()) { 1837 --TargetSym; 1838 uint64_t TargetAddress = std::get<0>(*TargetSym); 1839 StringRef TargetName = std::get<1>(*TargetSym); 1840 outs() << " <" << TargetName; 1841 uint64_t Disp = Target - TargetAddress; 1842 if (Disp) 1843 outs() << "+0x" << Twine::utohexstr(Disp); 1844 outs() << '>'; 1845 } 1846 } 1847 } 1848 outs() << "\n"; 1849 1850 // Hexagon does this in pretty printer 1851 if (Obj->getArch() != Triple::hexagon) 1852 // Print relocation for instruction. 1853 while (rel_cur != rel_end) { 1854 bool hidden = getHidden(*rel_cur); 1855 uint64_t addr = rel_cur->getOffset(); 1856 SmallString<16> name; 1857 SmallString<32> val; 1858 1859 // If this relocation is hidden, skip it. 1860 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1861 ++rel_cur; 1862 continue; 1863 } 1864 1865 // Stop when rel_cur's address is past the current instruction. 1866 if (addr >= Index + Size) break; 1867 rel_cur->getTypeName(name); 1868 error(getRelocationValueString(*rel_cur, val)); 1869 outs() << format(Fmt.data(), SectionAddr + addr) << name 1870 << "\t" << val << "\n"; 1871 ++rel_cur; 1872 } 1873 } 1874 } 1875 } 1876 } 1877 1878 void llvm::PrintRelocations(const ObjectFile *Obj) { 1879 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1880 "%08" PRIx64; 1881 // Regular objdump doesn't print relocations in non-relocatable object 1882 // files. 1883 if (!Obj->isRelocatableObject()) 1884 return; 1885 1886 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1887 if (Section.relocation_begin() == Section.relocation_end()) 1888 continue; 1889 StringRef secname; 1890 error(Section.getName(secname)); 1891 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1892 for (const RelocationRef &Reloc : Section.relocations()) { 1893 bool hidden = getHidden(Reloc); 1894 uint64_t address = Reloc.getOffset(); 1895 SmallString<32> relocname; 1896 SmallString<32> valuestr; 1897 if (address < StartAddress || address > StopAddress || hidden) 1898 continue; 1899 Reloc.getTypeName(relocname); 1900 error(getRelocationValueString(Reloc, valuestr)); 1901 outs() << format(Fmt.data(), address) << " " << relocname << " " 1902 << valuestr << "\n"; 1903 } 1904 outs() << "\n"; 1905 } 1906 } 1907 1908 void llvm::PrintDynamicRelocations(const ObjectFile *Obj) { 1909 1910 // For the moment, this option is for ELF only 1911 if (!Obj->isELF()) 1912 return; 1913 1914 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1915 1916 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1917 error("not a dynamic object"); 1918 return; 1919 } 1920 1921 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1922 1923 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1924 if (DynRelSec.empty()) 1925 return; 1926 1927 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1928 for (const SectionRef &Section : DynRelSec) { 1929 if (Section.relocation_begin() == Section.relocation_end()) 1930 continue; 1931 for (const RelocationRef &Reloc : Section.relocations()) { 1932 uint64_t address = Reloc.getOffset(); 1933 SmallString<32> relocname; 1934 SmallString<32> valuestr; 1935 Reloc.getTypeName(relocname); 1936 error(getRelocationValueString(Reloc, valuestr)); 1937 outs() << format(Fmt.data(), address) << " " << relocname << " " 1938 << valuestr << "\n"; 1939 } 1940 } 1941 } 1942 1943 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1944 outs() << "Sections:\n" 1945 "Idx Name Size Address Type\n"; 1946 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1947 StringRef Name; 1948 error(Section.getName(Name)); 1949 uint64_t Address = Section.getAddress(); 1950 uint64_t Size = Section.getSize(); 1951 bool Text = Section.isText(); 1952 bool Data = Section.isData(); 1953 bool BSS = Section.isBSS(); 1954 std::string Type = (std::string(Text ? "TEXT " : "") + 1955 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1956 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1957 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1958 Address, Type.c_str()); 1959 } 1960 outs() << "\n"; 1961 } 1962 1963 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1964 std::error_code EC; 1965 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1966 StringRef Name; 1967 StringRef Contents; 1968 error(Section.getName(Name)); 1969 uint64_t BaseAddr = Section.getAddress(); 1970 uint64_t Size = Section.getSize(); 1971 if (!Size) 1972 continue; 1973 1974 outs() << "Contents of section " << Name << ":\n"; 1975 if (Section.isBSS()) { 1976 outs() << format("<skipping contents of bss section at [%04" PRIx64 1977 ", %04" PRIx64 ")>\n", 1978 BaseAddr, BaseAddr + Size); 1979 continue; 1980 } 1981 1982 error(Section.getContents(Contents)); 1983 1984 // Dump out the content as hex and printable ascii characters. 1985 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1986 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1987 // Dump line of hex. 1988 for (std::size_t i = 0; i < 16; ++i) { 1989 if (i != 0 && i % 4 == 0) 1990 outs() << ' '; 1991 if (addr + i < end) 1992 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1993 << hexdigit(Contents[addr + i] & 0xF, true); 1994 else 1995 outs() << " "; 1996 } 1997 // Print ascii. 1998 outs() << " "; 1999 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 2000 if (isPrint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 2001 outs() << Contents[addr + i]; 2002 else 2003 outs() << "."; 2004 } 2005 outs() << "\n"; 2006 } 2007 } 2008 } 2009 2010 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 2011 StringRef ArchitectureName) { 2012 outs() << "SYMBOL TABLE:\n"; 2013 2014 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 2015 printCOFFSymbolTable(coff); 2016 return; 2017 } 2018 2019 for (auto I = o->symbol_begin(), E = o->symbol_end(); I != E; ++I) { 2020 // Skip printing the special zero symbol when dumping an ELF file. 2021 // This makes the output consistent with the GNU objdump. 2022 if (I == o->symbol_begin() && isa<ELFObjectFileBase>(o)) 2023 continue; 2024 2025 const SymbolRef &Symbol = *I; 2026 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 2027 if (!AddressOrError) 2028 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 2029 ArchitectureName); 2030 uint64_t Address = *AddressOrError; 2031 if ((Address < StartAddress) || (Address > StopAddress)) 2032 continue; 2033 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 2034 if (!TypeOrError) 2035 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 2036 ArchitectureName); 2037 SymbolRef::Type Type = *TypeOrError; 2038 uint32_t Flags = Symbol.getFlags(); 2039 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 2040 if (!SectionOrErr) 2041 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 2042 ArchitectureName); 2043 section_iterator Section = *SectionOrErr; 2044 StringRef Name; 2045 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 2046 Section->getName(Name); 2047 } else { 2048 Expected<StringRef> NameOrErr = Symbol.getName(); 2049 if (!NameOrErr) 2050 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 2051 ArchitectureName); 2052 Name = *NameOrErr; 2053 } 2054 2055 bool Global = Flags & SymbolRef::SF_Global; 2056 bool Weak = Flags & SymbolRef::SF_Weak; 2057 bool Absolute = Flags & SymbolRef::SF_Absolute; 2058 bool Common = Flags & SymbolRef::SF_Common; 2059 bool Hidden = Flags & SymbolRef::SF_Hidden; 2060 2061 char GlobLoc = ' '; 2062 if (Type != SymbolRef::ST_Unknown) 2063 GlobLoc = Global ? 'g' : 'l'; 2064 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2065 ? 'd' : ' '; 2066 char FileFunc = ' '; 2067 if (Type == SymbolRef::ST_File) 2068 FileFunc = 'f'; 2069 else if (Type == SymbolRef::ST_Function) 2070 FileFunc = 'F'; 2071 else if (Type == SymbolRef::ST_Data) 2072 FileFunc = 'O'; 2073 2074 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 2075 "%08" PRIx64; 2076 2077 outs() << format(Fmt, Address) << " " 2078 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2079 << (Weak ? 'w' : ' ') // Weak? 2080 << ' ' // Constructor. Not supported yet. 2081 << ' ' // Warning. Not supported yet. 2082 << ' ' // Indirect reference to another symbol. 2083 << Debug // Debugging (d) or dynamic (D) symbol. 2084 << FileFunc // Name of function (F), file (f) or object (O). 2085 << ' '; 2086 if (Absolute) { 2087 outs() << "*ABS*"; 2088 } else if (Common) { 2089 outs() << "*COM*"; 2090 } else if (Section == o->section_end()) { 2091 outs() << "*UND*"; 2092 } else { 2093 if (const MachOObjectFile *MachO = 2094 dyn_cast<const MachOObjectFile>(o)) { 2095 DataRefImpl DR = Section->getRawDataRefImpl(); 2096 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 2097 outs() << SegmentName << ","; 2098 } 2099 StringRef SectionName; 2100 error(Section->getName(SectionName)); 2101 outs() << SectionName; 2102 } 2103 2104 outs() << '\t'; 2105 if (Common || isa<ELFObjectFileBase>(o)) { 2106 uint64_t Val = 2107 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2108 outs() << format("\t %08" PRIx64 " ", Val); 2109 } 2110 2111 if (Hidden) { 2112 outs() << ".hidden "; 2113 } 2114 2115 if (Demangle) 2116 outs() << demangle(Name) << '\n'; 2117 else 2118 outs() << Name << '\n'; 2119 } 2120 } 2121 2122 static void PrintUnwindInfo(const ObjectFile *o) { 2123 outs() << "Unwind info:\n\n"; 2124 2125 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 2126 printCOFFUnwindInfo(coff); 2127 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2128 printMachOUnwindInfo(MachO); 2129 else { 2130 // TODO: Extract DWARF dump tool to objdump. 2131 WithColor::error(errs(), ToolName) 2132 << "This operation is only currently supported " 2133 "for COFF and MachO object files.\n"; 2134 return; 2135 } 2136 } 2137 2138 void llvm::printExportsTrie(const ObjectFile *o) { 2139 outs() << "Exports trie:\n"; 2140 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2141 printMachOExportsTrie(MachO); 2142 else { 2143 WithColor::error(errs(), ToolName) 2144 << "This operation is only currently supported " 2145 "for Mach-O executable files.\n"; 2146 return; 2147 } 2148 } 2149 2150 void llvm::printRebaseTable(ObjectFile *o) { 2151 outs() << "Rebase table:\n"; 2152 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2153 printMachORebaseTable(MachO); 2154 else { 2155 WithColor::error(errs(), ToolName) 2156 << "This operation is only currently supported " 2157 "for Mach-O executable files.\n"; 2158 return; 2159 } 2160 } 2161 2162 void llvm::printBindTable(ObjectFile *o) { 2163 outs() << "Bind table:\n"; 2164 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2165 printMachOBindTable(MachO); 2166 else { 2167 WithColor::error(errs(), ToolName) 2168 << "This operation is only currently supported " 2169 "for Mach-O executable files.\n"; 2170 return; 2171 } 2172 } 2173 2174 void llvm::printLazyBindTable(ObjectFile *o) { 2175 outs() << "Lazy bind table:\n"; 2176 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2177 printMachOLazyBindTable(MachO); 2178 else { 2179 WithColor::error(errs(), ToolName) 2180 << "This operation is only currently supported " 2181 "for Mach-O executable files.\n"; 2182 return; 2183 } 2184 } 2185 2186 void llvm::printWeakBindTable(ObjectFile *o) { 2187 outs() << "Weak bind table:\n"; 2188 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2189 printMachOWeakBindTable(MachO); 2190 else { 2191 WithColor::error(errs(), ToolName) 2192 << "This operation is only currently supported " 2193 "for Mach-O executable files.\n"; 2194 return; 2195 } 2196 } 2197 2198 /// Dump the raw contents of the __clangast section so the output can be piped 2199 /// into llvm-bcanalyzer. 2200 void llvm::printRawClangAST(const ObjectFile *Obj) { 2201 if (outs().is_displayed()) { 2202 WithColor::error(errs(), ToolName) 2203 << "The -raw-clang-ast option will dump the raw binary contents of " 2204 "the clang ast section.\n" 2205 "Please redirect the output to a file or another program such as " 2206 "llvm-bcanalyzer.\n"; 2207 return; 2208 } 2209 2210 StringRef ClangASTSectionName("__clangast"); 2211 if (isa<COFFObjectFile>(Obj)) { 2212 ClangASTSectionName = "clangast"; 2213 } 2214 2215 Optional<object::SectionRef> ClangASTSection; 2216 for (auto Sec : ToolSectionFilter(*Obj)) { 2217 StringRef Name; 2218 Sec.getName(Name); 2219 if (Name == ClangASTSectionName) { 2220 ClangASTSection = Sec; 2221 break; 2222 } 2223 } 2224 if (!ClangASTSection) 2225 return; 2226 2227 StringRef ClangASTContents; 2228 error(ClangASTSection.getValue().getContents(ClangASTContents)); 2229 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2230 } 2231 2232 static void printFaultMaps(const ObjectFile *Obj) { 2233 const char *FaultMapSectionName = nullptr; 2234 2235 if (isa<ELFObjectFileBase>(Obj)) { 2236 FaultMapSectionName = ".llvm_faultmaps"; 2237 } else if (isa<MachOObjectFile>(Obj)) { 2238 FaultMapSectionName = "__llvm_faultmaps"; 2239 } else { 2240 WithColor::error(errs(), ToolName) 2241 << "This operation is only currently supported " 2242 "for ELF and Mach-O executable files.\n"; 2243 return; 2244 } 2245 2246 Optional<object::SectionRef> FaultMapSection; 2247 2248 for (auto Sec : ToolSectionFilter(*Obj)) { 2249 StringRef Name; 2250 Sec.getName(Name); 2251 if (Name == FaultMapSectionName) { 2252 FaultMapSection = Sec; 2253 break; 2254 } 2255 } 2256 2257 outs() << "FaultMap table:\n"; 2258 2259 if (!FaultMapSection.hasValue()) { 2260 outs() << "<not found>\n"; 2261 return; 2262 } 2263 2264 StringRef FaultMapContents; 2265 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2266 2267 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2268 FaultMapContents.bytes_end()); 2269 2270 outs() << FMP; 2271 } 2272 2273 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2274 if (o->isELF()) { 2275 printELFFileHeader(o); 2276 return printELFDynamicSection(o); 2277 } 2278 if (o->isCOFF()) 2279 return printCOFFFileHeader(o); 2280 if (o->isWasm()) 2281 return printWasmFileHeader(o); 2282 if (o->isMachO()) { 2283 printMachOFileHeader(o); 2284 if (!onlyFirst) 2285 printMachOLoadCommands(o); 2286 return; 2287 } 2288 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2289 } 2290 2291 static void printFileHeaders(const ObjectFile *o) { 2292 if (!o->isELF() && !o->isCOFF()) 2293 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2294 2295 Triple::ArchType AT = o->getArch(); 2296 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2297 Expected<uint64_t> StartAddrOrErr = o->getStartAddress(); 2298 if (!StartAddrOrErr) 2299 report_error(o->getFileName(), StartAddrOrErr.takeError()); 2300 2301 StringRef Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2302 uint64_t Address = StartAddrOrErr.get(); 2303 outs() << "start address: " 2304 << "0x" << format(Fmt.data(), Address) 2305 << "\n"; 2306 } 2307 2308 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2309 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2310 if (!ModeOrErr) { 2311 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 2312 consumeError(ModeOrErr.takeError()); 2313 return; 2314 } 2315 sys::fs::perms Mode = ModeOrErr.get(); 2316 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2317 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2318 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2319 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2320 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2321 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2322 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2323 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2324 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2325 2326 outs() << " "; 2327 2328 Expected<unsigned> UIDOrErr = C.getUID(); 2329 if (!UIDOrErr) 2330 report_error(Filename, UIDOrErr.takeError()); 2331 unsigned UID = UIDOrErr.get(); 2332 outs() << format("%d/", UID); 2333 2334 Expected<unsigned> GIDOrErr = C.getGID(); 2335 if (!GIDOrErr) 2336 report_error(Filename, GIDOrErr.takeError()); 2337 unsigned GID = GIDOrErr.get(); 2338 outs() << format("%-d ", GID); 2339 2340 Expected<uint64_t> Size = C.getRawSize(); 2341 if (!Size) 2342 report_error(Filename, Size.takeError()); 2343 outs() << format("%6" PRId64, Size.get()) << " "; 2344 2345 StringRef RawLastModified = C.getRawLastModified(); 2346 unsigned Seconds; 2347 if (RawLastModified.getAsInteger(10, Seconds)) 2348 outs() << "(date: \"" << RawLastModified 2349 << "\" contains non-decimal chars) "; 2350 else { 2351 // Since ctime(3) returns a 26 character string of the form: 2352 // "Sun Sep 16 01:03:52 1973\n\0" 2353 // just print 24 characters. 2354 time_t t = Seconds; 2355 outs() << format("%.24s ", ctime(&t)); 2356 } 2357 2358 StringRef Name = ""; 2359 Expected<StringRef> NameOrErr = C.getName(); 2360 if (!NameOrErr) { 2361 consumeError(NameOrErr.takeError()); 2362 Expected<StringRef> RawNameOrErr = C.getRawName(); 2363 if (!RawNameOrErr) 2364 report_error(Filename, NameOrErr.takeError()); 2365 Name = RawNameOrErr.get(); 2366 } else { 2367 Name = NameOrErr.get(); 2368 } 2369 outs() << Name << "\n"; 2370 } 2371 2372 static void DumpObject(ObjectFile *o, const Archive *a = nullptr, 2373 const Archive::Child *c = nullptr) { 2374 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2375 // Avoid other output when using a raw option. 2376 if (!RawClangAST) { 2377 outs() << '\n'; 2378 if (a) 2379 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2380 else 2381 outs() << o->getFileName(); 2382 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2383 } 2384 2385 if (ArchiveHeaders && !MachOOpt && c) 2386 printArchiveChild(ArchiveName, *c); 2387 if (Disassemble) 2388 DisassembleObject(o, Relocations); 2389 if (Relocations && !Disassemble) 2390 PrintRelocations(o); 2391 if (DynamicRelocations) 2392 PrintDynamicRelocations(o); 2393 if (SectionHeaders) 2394 PrintSectionHeaders(o); 2395 if (SectionContents) 2396 PrintSectionContents(o); 2397 if (SymbolTable) 2398 PrintSymbolTable(o, ArchiveName); 2399 if (UnwindInfo) 2400 PrintUnwindInfo(o); 2401 if (PrivateHeaders || FirstPrivateHeader) 2402 printPrivateFileHeaders(o, FirstPrivateHeader); 2403 if (FileHeaders) 2404 printFileHeaders(o); 2405 if (ExportsTrie) 2406 printExportsTrie(o); 2407 if (Rebase) 2408 printRebaseTable(o); 2409 if (Bind) 2410 printBindTable(o); 2411 if (LazyBind) 2412 printLazyBindTable(o); 2413 if (WeakBind) 2414 printWeakBindTable(o); 2415 if (RawClangAST) 2416 printRawClangAST(o); 2417 if (PrintFaultMaps) 2418 printFaultMaps(o); 2419 if (DwarfDumpType != DIDT_Null) { 2420 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*o); 2421 // Dump the complete DWARF structure. 2422 DIDumpOptions DumpOpts; 2423 DumpOpts.DumpType = DwarfDumpType; 2424 DICtx->dump(outs(), DumpOpts); 2425 } 2426 } 2427 2428 static void DumpObject(const COFFImportFile *I, const Archive *A, 2429 const Archive::Child *C = nullptr) { 2430 StringRef ArchiveName = A ? A->getFileName() : ""; 2431 2432 // Avoid other output when using a raw option. 2433 if (!RawClangAST) 2434 outs() << '\n' 2435 << ArchiveName << "(" << I->getFileName() << ")" 2436 << ":\tfile format COFF-import-file" 2437 << "\n\n"; 2438 2439 if (ArchiveHeaders && !MachOOpt && C) 2440 printArchiveChild(ArchiveName, *C); 2441 if (SymbolTable) 2442 printCOFFSymbolTable(I); 2443 } 2444 2445 /// Dump each object file in \a a; 2446 static void DumpArchive(const Archive *a) { 2447 Error Err = Error::success(); 2448 for (auto &C : a->children(Err)) { 2449 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2450 if (!ChildOrErr) { 2451 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2452 report_error(a->getFileName(), C, std::move(E)); 2453 continue; 2454 } 2455 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2456 DumpObject(o, a, &C); 2457 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2458 DumpObject(I, a, &C); 2459 else 2460 report_error(a->getFileName(), object_error::invalid_file_type); 2461 } 2462 if (Err) 2463 report_error(a->getFileName(), std::move(Err)); 2464 } 2465 2466 /// Open file and figure out how to dump it. 2467 static void DumpInput(StringRef file) { 2468 2469 // If we are using the Mach-O specific object file parser, then let it parse 2470 // the file and process the command line options. So the -arch flags can 2471 // be used to select specific slices, etc. 2472 if (MachOOpt) { 2473 ParseInputMachO(file); 2474 return; 2475 } 2476 2477 // Attempt to open the binary. 2478 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2479 if (!BinaryOrErr) 2480 report_error(file, BinaryOrErr.takeError()); 2481 Binary &Binary = *BinaryOrErr.get().getBinary(); 2482 2483 if (Archive *a = dyn_cast<Archive>(&Binary)) 2484 DumpArchive(a); 2485 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2486 DumpObject(o); 2487 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2488 ParseInputMachO(UB); 2489 else 2490 report_error(file, object_error::invalid_file_type); 2491 } 2492 2493 int main(int argc, char **argv) { 2494 InitLLVM X(argc, argv); 2495 2496 // Initialize targets and assembly printers/parsers. 2497 llvm::InitializeAllTargetInfos(); 2498 llvm::InitializeAllTargetMCs(); 2499 llvm::InitializeAllDisassemblers(); 2500 2501 // Register the target printer for --version. 2502 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2503 2504 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2505 2506 ToolName = argv[0]; 2507 2508 // Defaults to a.out if no filenames specified. 2509 if (InputFilenames.empty()) 2510 InputFilenames.push_back("a.out"); 2511 2512 if (AllHeaders) 2513 PrivateHeaders = Relocations = SectionHeaders = SymbolTable = true; 2514 2515 if (DisassembleAll || PrintSource || PrintLines) 2516 Disassemble = true; 2517 2518 if (!Disassemble 2519 && !Relocations 2520 && !DynamicRelocations 2521 && !SectionHeaders 2522 && !SectionContents 2523 && !SymbolTable 2524 && !UnwindInfo 2525 && !PrivateHeaders 2526 && !FileHeaders 2527 && !FirstPrivateHeader 2528 && !ExportsTrie 2529 && !Rebase 2530 && !Bind 2531 && !LazyBind 2532 && !WeakBind 2533 && !RawClangAST 2534 && !(UniversalHeaders && MachOOpt) 2535 && !ArchiveHeaders 2536 && !(IndirectSymbols && MachOOpt) 2537 && !(DataInCode && MachOOpt) 2538 && !(LinkOptHints && MachOOpt) 2539 && !(InfoPlist && MachOOpt) 2540 && !(DylibsUsed && MachOOpt) 2541 && !(DylibId && MachOOpt) 2542 && !(ObjcMetaData && MachOOpt) 2543 && !(!FilterSections.empty() && MachOOpt) 2544 && !PrintFaultMaps 2545 && DwarfDumpType == DIDT_Null) { 2546 cl::PrintHelpMessage(); 2547 return 2; 2548 } 2549 2550 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2551 DisassembleFunctions.end()); 2552 2553 llvm::for_each(InputFilenames, DumpInput); 2554 2555 return EXIT_SUCCESS; 2556 } 2557