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/Triple.h" 24 #include "llvm/CodeGen/FaultMaps.h" 25 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 26 #include "llvm/MC/MCAsmInfo.h" 27 #include "llvm/MC/MCContext.h" 28 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 29 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 30 #include "llvm/MC/MCInst.h" 31 #include "llvm/MC/MCInstPrinter.h" 32 #include "llvm/MC/MCInstrAnalysis.h" 33 #include "llvm/MC/MCInstrInfo.h" 34 #include "llvm/MC/MCObjectFileInfo.h" 35 #include "llvm/MC/MCRegisterInfo.h" 36 #include "llvm/MC/MCSubtargetInfo.h" 37 #include "llvm/Object/Archive.h" 38 #include "llvm/Object/COFF.h" 39 #include "llvm/Object/ELFObjectFile.h" 40 #include "llvm/Object/MachO.h" 41 #include "llvm/Object/ObjectFile.h" 42 #include "llvm/Support/Casting.h" 43 #include "llvm/Support/CommandLine.h" 44 #include "llvm/Support/Debug.h" 45 #include "llvm/Support/Errc.h" 46 #include "llvm/Support/FileSystem.h" 47 #include "llvm/Support/Format.h" 48 #include "llvm/Support/GraphWriter.h" 49 #include "llvm/Support/Host.h" 50 #include "llvm/Support/ManagedStatic.h" 51 #include "llvm/Support/MemoryBuffer.h" 52 #include "llvm/Support/PrettyStackTrace.h" 53 #include "llvm/Support/Signals.h" 54 #include "llvm/Support/SourceMgr.h" 55 #include "llvm/Support/TargetRegistry.h" 56 #include "llvm/Support/TargetSelect.h" 57 #include "llvm/Support/raw_ostream.h" 58 #include <algorithm> 59 #include <cctype> 60 #include <cstring> 61 #include <system_error> 62 63 using namespace llvm; 64 using namespace object; 65 66 static cl::list<std::string> 67 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 68 69 cl::opt<bool> 70 llvm::Disassemble("disassemble", 71 cl::desc("Display assembler mnemonics for the machine instructions")); 72 static cl::alias 73 Disassembled("d", cl::desc("Alias for --disassemble"), 74 cl::aliasopt(Disassemble)); 75 76 cl::opt<bool> 77 llvm::DisassembleAll("disassemble-all", 78 cl::desc("Display assembler mnemonics for the machine instructions")); 79 static cl::alias 80 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 81 cl::aliasopt(DisassembleAll)); 82 83 cl::opt<bool> 84 llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 85 86 cl::opt<bool> 87 llvm::SectionContents("s", cl::desc("Display the content of each section")); 88 89 cl::opt<bool> 90 llvm::SymbolTable("t", cl::desc("Display the symbol table")); 91 92 cl::opt<bool> 93 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 94 95 cl::opt<bool> 96 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 97 98 cl::opt<bool> 99 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 100 101 cl::opt<bool> 102 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 103 104 cl::opt<bool> 105 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 106 107 cl::opt<bool> 108 llvm::RawClangAST("raw-clang-ast", 109 cl::desc("Dump the raw binary contents of the clang AST section")); 110 111 static cl::opt<bool> 112 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 113 static cl::alias 114 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 115 116 cl::opt<std::string> 117 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 118 "see -version for available targets")); 119 120 cl::opt<std::string> 121 llvm::MCPU("mcpu", 122 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 123 cl::value_desc("cpu-name"), 124 cl::init("")); 125 126 cl::opt<std::string> 127 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 128 "see -version for available targets")); 129 130 cl::opt<bool> 131 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 132 "headers for each section.")); 133 static cl::alias 134 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 135 cl::aliasopt(SectionHeaders)); 136 static cl::alias 137 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 138 cl::aliasopt(SectionHeaders)); 139 140 cl::list<std::string> 141 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 142 "With -macho dump segment,section")); 143 cl::alias 144 static FilterSectionsj("j", cl::desc("Alias for --section"), 145 cl::aliasopt(llvm::FilterSections)); 146 147 cl::list<std::string> 148 llvm::MAttrs("mattr", 149 cl::CommaSeparated, 150 cl::desc("Target specific attributes"), 151 cl::value_desc("a1,+a2,-a3,...")); 152 153 cl::opt<bool> 154 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 155 "instructions, do not print " 156 "the instruction bytes.")); 157 158 cl::opt<bool> 159 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 160 161 static cl::alias 162 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 163 cl::aliasopt(UnwindInfo)); 164 165 cl::opt<bool> 166 llvm::PrivateHeaders("private-headers", 167 cl::desc("Display format specific file headers")); 168 169 cl::opt<bool> 170 llvm::FirstPrivateHeader("private-header", 171 cl::desc("Display only the first format specific file " 172 "header")); 173 174 static cl::alias 175 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 176 cl::aliasopt(PrivateHeaders)); 177 178 cl::opt<bool> 179 llvm::PrintImmHex("print-imm-hex", 180 cl::desc("Use hex format for immediate values")); 181 182 cl::opt<bool> PrintFaultMaps("fault-map-section", 183 cl::desc("Display contents of faultmap section")); 184 185 cl::opt<DIDumpType> llvm::DwarfDumpType( 186 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 187 cl::values(clEnumValN(DIDT_Frames, "frames", ".debug_frame"), 188 clEnumValEnd)); 189 190 static StringRef ToolName; 191 192 namespace { 193 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 194 195 class SectionFilterIterator { 196 public: 197 SectionFilterIterator(FilterPredicate P, 198 llvm::object::section_iterator const &I, 199 llvm::object::section_iterator const &E) 200 : Predicate(P), Iterator(I), End(E) { 201 ScanPredicate(); 202 } 203 const llvm::object::SectionRef &operator*() const { return *Iterator; } 204 SectionFilterIterator &operator++() { 205 ++Iterator; 206 ScanPredicate(); 207 return *this; 208 } 209 bool operator!=(SectionFilterIterator const &Other) const { 210 return Iterator != Other.Iterator; 211 } 212 213 private: 214 void ScanPredicate() { 215 while (Iterator != End && !Predicate(*Iterator)) { 216 ++Iterator; 217 } 218 } 219 FilterPredicate Predicate; 220 llvm::object::section_iterator Iterator; 221 llvm::object::section_iterator End; 222 }; 223 224 class SectionFilter { 225 public: 226 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 227 : Predicate(P), Object(O) {} 228 SectionFilterIterator begin() { 229 return SectionFilterIterator(Predicate, Object.section_begin(), 230 Object.section_end()); 231 } 232 SectionFilterIterator end() { 233 return SectionFilterIterator(Predicate, Object.section_end(), 234 Object.section_end()); 235 } 236 237 private: 238 FilterPredicate Predicate; 239 llvm::object::ObjectFile const &Object; 240 }; 241 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 242 return SectionFilter([](llvm::object::SectionRef const &S) { 243 if(FilterSections.empty()) 244 return true; 245 llvm::StringRef String; 246 std::error_code error = S.getName(String); 247 if (error) 248 return false; 249 return std::find(FilterSections.begin(), 250 FilterSections.end(), 251 String) != FilterSections.end(); 252 }, 253 O); 254 } 255 } 256 257 void llvm::error(std::error_code EC) { 258 if (!EC) 259 return; 260 261 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 262 errs().flush(); 263 exit(1); 264 } 265 266 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 267 std::error_code EC) { 268 assert(EC); 269 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 270 exit(1); 271 } 272 273 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 274 // Figure out the target triple. 275 llvm::Triple TheTriple("unknown-unknown-unknown"); 276 if (TripleName.empty()) { 277 if (Obj) { 278 TheTriple.setArch(Triple::ArchType(Obj->getArch())); 279 // TheTriple defaults to ELF, and COFF doesn't have an environment: 280 // the best we can do here is indicate that it is mach-o. 281 if (Obj->isMachO()) 282 TheTriple.setObjectFormat(Triple::MachO); 283 284 if (Obj->isCOFF()) { 285 const auto COFFObj = dyn_cast<COFFObjectFile>(Obj); 286 if (COFFObj->getArch() == Triple::thumb) 287 TheTriple.setTriple("thumbv7-windows"); 288 } 289 } 290 } else 291 TheTriple.setTriple(Triple::normalize(TripleName)); 292 293 // Get the target specific parser. 294 std::string Error; 295 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 296 Error); 297 if (!TheTarget) 298 report_fatal_error("can't find target: " + Error); 299 300 // Update the triple name and return the found target. 301 TripleName = TheTriple.getTriple(); 302 return TheTarget; 303 } 304 305 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 306 return a.getOffset() < b.getOffset(); 307 } 308 309 namespace { 310 class PrettyPrinter { 311 public: 312 virtual ~PrettyPrinter(){} 313 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 314 ArrayRef<uint8_t> Bytes, uint64_t Address, 315 raw_ostream &OS, StringRef Annot, 316 MCSubtargetInfo const &STI) { 317 OS << format("%8" PRIx64 ":", Address); 318 if (!NoShowRawInsn) { 319 OS << "\t"; 320 dumpBytes(Bytes, OS); 321 } 322 if (MI) 323 IP.printInst(MI, OS, "", STI); 324 else 325 OS << " <unknown>"; 326 } 327 }; 328 PrettyPrinter PrettyPrinterInst; 329 class HexagonPrettyPrinter : public PrettyPrinter { 330 public: 331 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 332 raw_ostream &OS) { 333 uint32_t opcode = 334 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 335 OS << format("%8" PRIx64 ":", Address); 336 if (!NoShowRawInsn) { 337 OS << "\t"; 338 dumpBytes(Bytes.slice(0, 4), OS); 339 OS << format("%08" PRIx32, opcode); 340 } 341 } 342 void printInst(MCInstPrinter &IP, const MCInst *MI, 343 ArrayRef<uint8_t> Bytes, uint64_t Address, 344 raw_ostream &OS, StringRef Annot, 345 MCSubtargetInfo const &STI) override { 346 if (!MI) { 347 printLead(Bytes, Address, OS); 348 OS << " <unknown>"; 349 return; 350 } 351 std::string Buffer; 352 { 353 raw_string_ostream TempStream(Buffer); 354 IP.printInst(MI, TempStream, "", STI); 355 } 356 StringRef Contents(Buffer); 357 // Split off bundle attributes 358 auto PacketBundle = Contents.rsplit('\n'); 359 // Split off first instruction from the rest 360 auto HeadTail = PacketBundle.first.split('\n'); 361 auto Preamble = " { "; 362 auto Separator = ""; 363 while(!HeadTail.first.empty()) { 364 OS << Separator; 365 Separator = "\n"; 366 printLead(Bytes, Address, OS); 367 OS << Preamble; 368 Preamble = " "; 369 StringRef Inst; 370 auto Duplex = HeadTail.first.split('\v'); 371 if(!Duplex.second.empty()){ 372 OS << Duplex.first; 373 OS << "; "; 374 Inst = Duplex.second; 375 } 376 else 377 Inst = HeadTail.first; 378 OS << Inst; 379 Bytes = Bytes.slice(4); 380 Address += 4; 381 HeadTail = HeadTail.second.split('\n'); 382 } 383 OS << " } " << PacketBundle.second; 384 } 385 }; 386 HexagonPrettyPrinter HexagonPrettyPrinterInst; 387 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 388 switch(Triple.getArch()) { 389 default: 390 return PrettyPrinterInst; 391 case Triple::hexagon: 392 return HexagonPrettyPrinterInst; 393 } 394 } 395 } 396 397 template <class ELFT> 398 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 399 const RelocationRef &RelRef, 400 SmallVectorImpl<char> &Result) { 401 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 402 403 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 404 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 405 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 406 407 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 408 409 ErrorOr<const Elf_Shdr *> SecOrErr = EF.getSection(Rel.d.a); 410 if (std::error_code EC = SecOrErr.getError()) 411 return EC; 412 const Elf_Shdr *Sec = *SecOrErr; 413 ErrorOr<const Elf_Shdr *> SymTabOrErr = EF.getSection(Sec->sh_link); 414 if (std::error_code EC = SymTabOrErr.getError()) 415 return EC; 416 const Elf_Shdr *SymTab = *SymTabOrErr; 417 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 418 SymTab->sh_type == ELF::SHT_DYNSYM); 419 ErrorOr<const Elf_Shdr *> StrTabSec = EF.getSection(SymTab->sh_link); 420 if (std::error_code EC = StrTabSec.getError()) 421 return EC; 422 ErrorOr<StringRef> StrTabOrErr = EF.getStringTable(*StrTabSec); 423 if (std::error_code EC = StrTabOrErr.getError()) 424 return EC; 425 StringRef StrTab = *StrTabOrErr; 426 uint8_t type = RelRef.getType(); 427 StringRef res; 428 int64_t addend = 0; 429 switch (Sec->sh_type) { 430 default: 431 return object_error::parse_failed; 432 case ELF::SHT_REL: { 433 // TODO: Read implicit addend from section data. 434 break; 435 } 436 case ELF::SHT_RELA: { 437 const Elf_Rela *ERela = Obj->getRela(Rel); 438 addend = ERela->r_addend; 439 break; 440 } 441 } 442 symbol_iterator SI = RelRef.getSymbol(); 443 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 444 StringRef Target; 445 if (symb->getType() == ELF::STT_SECTION) { 446 ErrorOr<section_iterator> SymSI = SI->getSection(); 447 if (std::error_code EC = SymSI.getError()) 448 return EC; 449 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 450 ErrorOr<StringRef> SecName = EF.getSectionName(SymSec); 451 if (std::error_code EC = SecName.getError()) 452 return EC; 453 Target = *SecName; 454 } else { 455 ErrorOr<StringRef> SymName = symb->getName(StrTab); 456 if (!SymName) 457 return SymName.getError(); 458 Target = *SymName; 459 } 460 switch (EF.getHeader()->e_machine) { 461 case ELF::EM_X86_64: 462 switch (type) { 463 case ELF::R_X86_64_PC8: 464 case ELF::R_X86_64_PC16: 465 case ELF::R_X86_64_PC32: { 466 std::string fmtbuf; 467 raw_string_ostream fmt(fmtbuf); 468 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 469 fmt.flush(); 470 Result.append(fmtbuf.begin(), fmtbuf.end()); 471 } break; 472 case ELF::R_X86_64_8: 473 case ELF::R_X86_64_16: 474 case ELF::R_X86_64_32: 475 case ELF::R_X86_64_32S: 476 case ELF::R_X86_64_64: { 477 std::string fmtbuf; 478 raw_string_ostream fmt(fmtbuf); 479 fmt << Target << (addend < 0 ? "" : "+") << addend; 480 fmt.flush(); 481 Result.append(fmtbuf.begin(), fmtbuf.end()); 482 } break; 483 default: 484 res = "Unknown"; 485 } 486 break; 487 case ELF::EM_LANAI: 488 case ELF::EM_AARCH64: { 489 std::string fmtbuf; 490 raw_string_ostream fmt(fmtbuf); 491 fmt << Target; 492 if (addend != 0) 493 fmt << (addend < 0 ? "" : "+") << addend; 494 fmt.flush(); 495 Result.append(fmtbuf.begin(), fmtbuf.end()); 496 break; 497 } 498 case ELF::EM_386: 499 case ELF::EM_IAMCU: 500 case ELF::EM_ARM: 501 case ELF::EM_HEXAGON: 502 case ELF::EM_MIPS: 503 res = Target; 504 break; 505 case ELF::EM_WEBASSEMBLY: 506 switch (type) { 507 case ELF::R_WEBASSEMBLY_DATA: { 508 std::string fmtbuf; 509 raw_string_ostream fmt(fmtbuf); 510 fmt << Target << (addend < 0 ? "" : "+") << addend; 511 fmt.flush(); 512 Result.append(fmtbuf.begin(), fmtbuf.end()); 513 break; 514 } 515 case ELF::R_WEBASSEMBLY_FUNCTION: 516 res = Target; 517 break; 518 default: 519 res = "Unknown"; 520 } 521 break; 522 default: 523 res = "Unknown"; 524 } 525 if (Result.empty()) 526 Result.append(res.begin(), res.end()); 527 return std::error_code(); 528 } 529 530 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 531 const RelocationRef &Rel, 532 SmallVectorImpl<char> &Result) { 533 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 534 return getRelocationValueString(ELF32LE, Rel, Result); 535 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 536 return getRelocationValueString(ELF64LE, Rel, Result); 537 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 538 return getRelocationValueString(ELF32BE, Rel, Result); 539 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 540 return getRelocationValueString(ELF64BE, Rel, Result); 541 } 542 543 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 544 const RelocationRef &Rel, 545 SmallVectorImpl<char> &Result) { 546 symbol_iterator SymI = Rel.getSymbol(); 547 ErrorOr<StringRef> SymNameOrErr = SymI->getName(); 548 if (std::error_code EC = SymNameOrErr.getError()) 549 return EC; 550 StringRef SymName = *SymNameOrErr; 551 Result.append(SymName.begin(), SymName.end()); 552 return std::error_code(); 553 } 554 555 static void printRelocationTargetName(const MachOObjectFile *O, 556 const MachO::any_relocation_info &RE, 557 raw_string_ostream &fmt) { 558 bool IsScattered = O->isRelocationScattered(RE); 559 560 // Target of a scattered relocation is an address. In the interest of 561 // generating pretty output, scan through the symbol table looking for a 562 // symbol that aligns with that address. If we find one, print it. 563 // Otherwise, we just print the hex address of the target. 564 if (IsScattered) { 565 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 566 567 for (const SymbolRef &Symbol : O->symbols()) { 568 std::error_code ec; 569 ErrorOr<uint64_t> Addr = Symbol.getAddress(); 570 if ((ec = Addr.getError())) 571 report_fatal_error(ec.message()); 572 if (*Addr != Val) 573 continue; 574 ErrorOr<StringRef> Name = Symbol.getName(); 575 if (std::error_code EC = Name.getError()) 576 report_fatal_error(EC.message()); 577 fmt << *Name; 578 return; 579 } 580 581 // If we couldn't find a symbol that this relocation refers to, try 582 // to find a section beginning instead. 583 for (const SectionRef &Section : ToolSectionFilter(*O)) { 584 std::error_code ec; 585 586 StringRef Name; 587 uint64_t Addr = Section.getAddress(); 588 if (Addr != Val) 589 continue; 590 if ((ec = Section.getName(Name))) 591 report_fatal_error(ec.message()); 592 fmt << Name; 593 return; 594 } 595 596 fmt << format("0x%x", Val); 597 return; 598 } 599 600 StringRef S; 601 bool isExtern = O->getPlainRelocationExternal(RE); 602 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 603 604 if (isExtern) { 605 symbol_iterator SI = O->symbol_begin(); 606 advance(SI, Val); 607 ErrorOr<StringRef> SOrErr = SI->getName(); 608 error(SOrErr.getError()); 609 S = *SOrErr; 610 } else { 611 section_iterator SI = O->section_begin(); 612 // Adjust for the fact that sections are 1-indexed. 613 advance(SI, Val - 1); 614 SI->getName(S); 615 } 616 617 fmt << S; 618 } 619 620 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 621 const RelocationRef &RelRef, 622 SmallVectorImpl<char> &Result) { 623 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 624 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 625 626 unsigned Arch = Obj->getArch(); 627 628 std::string fmtbuf; 629 raw_string_ostream fmt(fmtbuf); 630 unsigned Type = Obj->getAnyRelocationType(RE); 631 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 632 633 // Determine any addends that should be displayed with the relocation. 634 // These require decoding the relocation type, which is triple-specific. 635 636 // X86_64 has entirely custom relocation types. 637 if (Arch == Triple::x86_64) { 638 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 639 640 switch (Type) { 641 case MachO::X86_64_RELOC_GOT_LOAD: 642 case MachO::X86_64_RELOC_GOT: { 643 printRelocationTargetName(Obj, RE, fmt); 644 fmt << "@GOT"; 645 if (isPCRel) 646 fmt << "PCREL"; 647 break; 648 } 649 case MachO::X86_64_RELOC_SUBTRACTOR: { 650 DataRefImpl RelNext = Rel; 651 Obj->moveRelocationNext(RelNext); 652 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 653 654 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 655 // X86_64_RELOC_UNSIGNED. 656 // NOTE: Scattered relocations don't exist on x86_64. 657 unsigned RType = Obj->getAnyRelocationType(RENext); 658 if (RType != MachO::X86_64_RELOC_UNSIGNED) 659 report_fatal_error("Expected X86_64_RELOC_UNSIGNED after " 660 "X86_64_RELOC_SUBTRACTOR."); 661 662 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 663 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 664 printRelocationTargetName(Obj, RENext, fmt); 665 fmt << "-"; 666 printRelocationTargetName(Obj, RE, fmt); 667 break; 668 } 669 case MachO::X86_64_RELOC_TLV: 670 printRelocationTargetName(Obj, RE, fmt); 671 fmt << "@TLV"; 672 if (isPCRel) 673 fmt << "P"; 674 break; 675 case MachO::X86_64_RELOC_SIGNED_1: 676 printRelocationTargetName(Obj, RE, fmt); 677 fmt << "-1"; 678 break; 679 case MachO::X86_64_RELOC_SIGNED_2: 680 printRelocationTargetName(Obj, RE, fmt); 681 fmt << "-2"; 682 break; 683 case MachO::X86_64_RELOC_SIGNED_4: 684 printRelocationTargetName(Obj, RE, fmt); 685 fmt << "-4"; 686 break; 687 default: 688 printRelocationTargetName(Obj, RE, fmt); 689 break; 690 } 691 // X86 and ARM share some relocation types in common. 692 } else if (Arch == Triple::x86 || Arch == Triple::arm || 693 Arch == Triple::ppc) { 694 // Generic relocation types... 695 switch (Type) { 696 case MachO::GENERIC_RELOC_PAIR: // prints no info 697 return std::error_code(); 698 case MachO::GENERIC_RELOC_SECTDIFF: { 699 DataRefImpl RelNext = Rel; 700 Obj->moveRelocationNext(RelNext); 701 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 702 703 // X86 sect diff's must be followed by a relocation of type 704 // GENERIC_RELOC_PAIR. 705 unsigned RType = Obj->getAnyRelocationType(RENext); 706 707 if (RType != MachO::GENERIC_RELOC_PAIR) 708 report_fatal_error("Expected GENERIC_RELOC_PAIR after " 709 "GENERIC_RELOC_SECTDIFF."); 710 711 printRelocationTargetName(Obj, RE, fmt); 712 fmt << "-"; 713 printRelocationTargetName(Obj, RENext, fmt); 714 break; 715 } 716 } 717 718 if (Arch == Triple::x86 || Arch == Triple::ppc) { 719 switch (Type) { 720 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 721 DataRefImpl RelNext = Rel; 722 Obj->moveRelocationNext(RelNext); 723 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 724 725 // X86 sect diff's must be followed by a relocation of type 726 // GENERIC_RELOC_PAIR. 727 unsigned RType = Obj->getAnyRelocationType(RENext); 728 if (RType != MachO::GENERIC_RELOC_PAIR) 729 report_fatal_error("Expected GENERIC_RELOC_PAIR after " 730 "GENERIC_RELOC_LOCAL_SECTDIFF."); 731 732 printRelocationTargetName(Obj, RE, fmt); 733 fmt << "-"; 734 printRelocationTargetName(Obj, RENext, fmt); 735 break; 736 } 737 case MachO::GENERIC_RELOC_TLV: { 738 printRelocationTargetName(Obj, RE, fmt); 739 fmt << "@TLV"; 740 if (IsPCRel) 741 fmt << "P"; 742 break; 743 } 744 default: 745 printRelocationTargetName(Obj, RE, fmt); 746 } 747 } else { // ARM-specific relocations 748 switch (Type) { 749 case MachO::ARM_RELOC_HALF: 750 case MachO::ARM_RELOC_HALF_SECTDIFF: { 751 // Half relocations steal a bit from the length field to encode 752 // whether this is an upper16 or a lower16 relocation. 753 bool isUpper = Obj->getAnyRelocationLength(RE) >> 1; 754 755 if (isUpper) 756 fmt << ":upper16:("; 757 else 758 fmt << ":lower16:("; 759 printRelocationTargetName(Obj, RE, fmt); 760 761 DataRefImpl RelNext = Rel; 762 Obj->moveRelocationNext(RelNext); 763 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 764 765 // ARM half relocs must be followed by a relocation of type 766 // ARM_RELOC_PAIR. 767 unsigned RType = Obj->getAnyRelocationType(RENext); 768 if (RType != MachO::ARM_RELOC_PAIR) 769 report_fatal_error("Expected ARM_RELOC_PAIR after " 770 "ARM_RELOC_HALF"); 771 772 // NOTE: The half of the target virtual address is stashed in the 773 // address field of the secondary relocation, but we can't reverse 774 // engineer the constant offset from it without decoding the movw/movt 775 // instruction to find the other half in its immediate field. 776 777 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 778 // symbol/section pointer of the follow-on relocation. 779 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 780 fmt << "-"; 781 printRelocationTargetName(Obj, RENext, fmt); 782 } 783 784 fmt << ")"; 785 break; 786 } 787 default: { printRelocationTargetName(Obj, RE, fmt); } 788 } 789 } 790 } else 791 printRelocationTargetName(Obj, RE, fmt); 792 793 fmt.flush(); 794 Result.append(fmtbuf.begin(), fmtbuf.end()); 795 return std::error_code(); 796 } 797 798 static std::error_code getRelocationValueString(const RelocationRef &Rel, 799 SmallVectorImpl<char> &Result) { 800 const ObjectFile *Obj = Rel.getObject(); 801 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 802 return getRelocationValueString(ELF, Rel, Result); 803 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 804 return getRelocationValueString(COFF, Rel, Result); 805 auto *MachO = cast<MachOObjectFile>(Obj); 806 return getRelocationValueString(MachO, Rel, Result); 807 } 808 809 /// @brief Indicates whether this relocation should hidden when listing 810 /// relocations, usually because it is the trailing part of a multipart 811 /// relocation that will be printed as part of the leading relocation. 812 static bool getHidden(RelocationRef RelRef) { 813 const ObjectFile *Obj = RelRef.getObject(); 814 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 815 if (!MachO) 816 return false; 817 818 unsigned Arch = MachO->getArch(); 819 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 820 uint64_t Type = MachO->getRelocationType(Rel); 821 822 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 823 // is always hidden. 824 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 825 if (Type == MachO::GENERIC_RELOC_PAIR) 826 return true; 827 } else if (Arch == Triple::x86_64) { 828 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 829 // an X86_64_RELOC_SUBTRACTOR. 830 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 831 DataRefImpl RelPrev = Rel; 832 RelPrev.d.a--; 833 uint64_t PrevType = MachO->getRelocationType(RelPrev); 834 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 835 return true; 836 } 837 } 838 839 return false; 840 } 841 842 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 843 const Target *TheTarget = getTarget(Obj); 844 845 // Package up features to be passed to target/subtarget 846 std::string FeaturesStr; 847 if (MAttrs.size()) { 848 SubtargetFeatures Features; 849 for (unsigned i = 0; i != MAttrs.size(); ++i) 850 Features.AddFeature(MAttrs[i]); 851 FeaturesStr = Features.getString(); 852 } 853 854 std::unique_ptr<const MCRegisterInfo> MRI( 855 TheTarget->createMCRegInfo(TripleName)); 856 if (!MRI) 857 report_fatal_error("error: no register info for target " + TripleName); 858 859 // Set up disassembler. 860 std::unique_ptr<const MCAsmInfo> AsmInfo( 861 TheTarget->createMCAsmInfo(*MRI, TripleName)); 862 if (!AsmInfo) 863 report_fatal_error("error: no assembly info for target " + TripleName); 864 std::unique_ptr<const MCSubtargetInfo> STI( 865 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr)); 866 if (!STI) 867 report_fatal_error("error: no subtarget info for target " + TripleName); 868 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 869 if (!MII) 870 report_fatal_error("error: no instruction info for target " + TripleName); 871 std::unique_ptr<const MCObjectFileInfo> MOFI(new MCObjectFileInfo); 872 MCContext Ctx(AsmInfo.get(), MRI.get(), MOFI.get()); 873 874 std::unique_ptr<MCDisassembler> DisAsm( 875 TheTarget->createMCDisassembler(*STI, Ctx)); 876 if (!DisAsm) 877 report_fatal_error("error: no disassembler for target " + TripleName); 878 879 std::unique_ptr<const MCInstrAnalysis> MIA( 880 TheTarget->createMCInstrAnalysis(MII.get())); 881 882 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 883 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 884 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 885 if (!IP) 886 report_fatal_error("error: no instruction printer for target " + 887 TripleName); 888 IP->setPrintImmHex(PrintImmHex); 889 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 890 891 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 892 "\t\t\t%08" PRIx64 ": "; 893 894 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 895 // in RelocSecs contain the relocations for section S. 896 std::error_code EC; 897 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 898 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 899 section_iterator Sec2 = Section.getRelocatedSection(); 900 if (Sec2 != Obj->section_end()) 901 SectionRelocMap[*Sec2].push_back(Section); 902 } 903 904 // Create a mapping from virtual address to symbol name. This is used to 905 // pretty print the symbols while disassembling. 906 typedef std::vector<std::pair<uint64_t, StringRef>> SectionSymbolsTy; 907 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 908 for (const SymbolRef &Symbol : Obj->symbols()) { 909 ErrorOr<uint64_t> AddressOrErr = Symbol.getAddress(); 910 error(AddressOrErr.getError()); 911 uint64_t Address = *AddressOrErr; 912 913 ErrorOr<StringRef> Name = Symbol.getName(); 914 error(Name.getError()); 915 if (Name->empty()) 916 continue; 917 918 ErrorOr<section_iterator> SectionOrErr = Symbol.getSection(); 919 error(SectionOrErr.getError()); 920 section_iterator SecI = *SectionOrErr; 921 if (SecI == Obj->section_end()) 922 continue; 923 924 AllSymbols[*SecI].emplace_back(Address, *Name); 925 } 926 927 // Create a mapping from virtual address to section. 928 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 929 for (SectionRef Sec : Obj->sections()) 930 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 931 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 932 933 // Linked executables (.exe and .dll files) typically don't include a real 934 // symbol table but they might contain an export table. 935 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 936 for (const auto &ExportEntry : COFFObj->export_directories()) { 937 StringRef Name; 938 error(ExportEntry.getSymbolName(Name)); 939 if (Name.empty()) 940 continue; 941 uint32_t RVA; 942 error(ExportEntry.getExportRVA(RVA)); 943 944 uint64_t VA = COFFObj->getImageBase() + RVA; 945 auto Sec = std::upper_bound( 946 SectionAddresses.begin(), SectionAddresses.end(), VA, 947 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 948 return LHS < RHS.first; 949 }); 950 if (Sec != SectionAddresses.begin()) 951 --Sec; 952 else 953 Sec = SectionAddresses.end(); 954 955 if (Sec != SectionAddresses.end()) 956 AllSymbols[Sec->second].emplace_back(VA, Name); 957 } 958 } 959 960 // Sort all the symbols, this allows us to use a simple binary search to find 961 // a symbol near an address. 962 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 963 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 964 965 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 966 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 967 continue; 968 969 uint64_t SectionAddr = Section.getAddress(); 970 uint64_t SectSize = Section.getSize(); 971 if (!SectSize) 972 continue; 973 974 // Get the list of all the symbols in this section. 975 SectionSymbolsTy &Symbols = AllSymbols[Section]; 976 std::vector<uint64_t> DataMappingSymsAddr; 977 std::vector<uint64_t> TextMappingSymsAddr; 978 if (Obj->isELF() && Obj->getArch() == Triple::aarch64) { 979 for (const auto &Symb : Symbols) { 980 uint64_t Address = Symb.first; 981 StringRef Name = Symb.second; 982 if (Name.startswith("$d")) 983 DataMappingSymsAddr.push_back(Address - SectionAddr); 984 if (Name.startswith("$x")) 985 TextMappingSymsAddr.push_back(Address - SectionAddr); 986 } 987 } 988 989 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 990 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 991 992 // Make a list of all the relocations for this section. 993 std::vector<RelocationRef> Rels; 994 if (InlineRelocs) { 995 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 996 for (const RelocationRef &Reloc : RelocSec.relocations()) { 997 Rels.push_back(Reloc); 998 } 999 } 1000 } 1001 1002 // Sort relocations by address. 1003 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1004 1005 StringRef SegmentName = ""; 1006 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1007 DataRefImpl DR = Section.getRawDataRefImpl(); 1008 SegmentName = MachO->getSectionFinalSegmentName(DR); 1009 } 1010 StringRef name; 1011 error(Section.getName(name)); 1012 outs() << "Disassembly of section "; 1013 if (!SegmentName.empty()) 1014 outs() << SegmentName << ","; 1015 outs() << name << ':'; 1016 1017 // If the section has no symbol at the start, just insert a dummy one. 1018 if (Symbols.empty() || Symbols[0].first != 0) 1019 Symbols.insert(Symbols.begin(), std::make_pair(SectionAddr, name)); 1020 1021 SmallString<40> Comments; 1022 raw_svector_ostream CommentStream(Comments); 1023 1024 StringRef BytesStr; 1025 error(Section.getContents(BytesStr)); 1026 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1027 BytesStr.size()); 1028 1029 uint64_t Size; 1030 uint64_t Index; 1031 1032 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1033 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1034 // Disassemble symbol by symbol. 1035 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1036 1037 uint64_t Start = Symbols[si].first - SectionAddr; 1038 // The end is either the section end or the beginning of the next 1039 // symbol. 1040 uint64_t End = 1041 (si == se - 1) ? SectSize : Symbols[si + 1].first - SectionAddr; 1042 // Don't try to disassemble beyond the end of section contents. 1043 if (End > SectSize) 1044 End = SectSize; 1045 // If this symbol has the same address as the next symbol, then skip it. 1046 if (Start >= End) 1047 continue; 1048 1049 outs() << '\n' << Symbols[si].second << ":\n"; 1050 1051 #ifndef NDEBUG 1052 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1053 #else 1054 raw_ostream &DebugOut = nulls(); 1055 #endif 1056 1057 for (Index = Start; Index < End; Index += Size) { 1058 MCInst Inst; 1059 1060 // AArch64 ELF binaries can interleave data and text in the 1061 // same section. We rely on the markers introduced to 1062 // understand what we need to dump. 1063 if (Obj->isELF() && Obj->getArch() == Triple::aarch64) { 1064 uint64_t Stride = 0; 1065 1066 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1067 DataMappingSymsAddr.end(), Index); 1068 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1069 // Switch to data. 1070 while (Index < End) { 1071 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1072 outs() << "\t"; 1073 if (Index + 4 <= End) { 1074 Stride = 4; 1075 dumpBytes(Bytes.slice(Index, 4), outs()); 1076 outs() << "\t.word"; 1077 } else if (Index + 2 <= End) { 1078 Stride = 2; 1079 dumpBytes(Bytes.slice(Index, 2), outs()); 1080 outs() << "\t.short"; 1081 } else { 1082 Stride = 1; 1083 dumpBytes(Bytes.slice(Index, 1), outs()); 1084 outs() << "\t.byte"; 1085 } 1086 Index += Stride; 1087 outs() << "\n"; 1088 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1089 TextMappingSymsAddr.end(), Index); 1090 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1091 break; 1092 } 1093 } 1094 } 1095 1096 if (Index >= End) 1097 break; 1098 1099 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1100 SectionAddr + Index, DebugOut, 1101 CommentStream); 1102 if (Size == 0) 1103 Size = 1; 1104 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1105 Bytes.slice(Index, Size), 1106 SectionAddr + Index, outs(), "", *STI); 1107 outs() << CommentStream.str(); 1108 Comments.clear(); 1109 1110 // Try to resolve the target of a call, tail call, etc. to a specific 1111 // symbol. 1112 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1113 MIA->isConditionalBranch(Inst))) { 1114 uint64_t Target; 1115 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1116 // In a relocatable object, the target's section must reside in 1117 // the same section as the call instruction or it is accessed 1118 // through a relocation. 1119 // 1120 // In a non-relocatable object, the target may be in any section. 1121 // 1122 // N.B. We don't walk the relocations in the relocatable case yet. 1123 auto *TargetSectionSymbols = &Symbols; 1124 if (!Obj->isRelocatableObject()) { 1125 auto SectionAddress = std::upper_bound( 1126 SectionAddresses.begin(), SectionAddresses.end(), Target, 1127 [](uint64_t LHS, 1128 const std::pair<uint64_t, SectionRef> &RHS) { 1129 return LHS < RHS.first; 1130 }); 1131 if (SectionAddress != SectionAddresses.begin()) { 1132 --SectionAddress; 1133 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1134 } else { 1135 TargetSectionSymbols = nullptr; 1136 } 1137 } 1138 1139 // Find the first symbol in the section whose offset is less than 1140 // or equal to the target. 1141 if (TargetSectionSymbols) { 1142 auto TargetSym = std::upper_bound( 1143 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1144 Target, [](uint64_t LHS, 1145 const std::pair<uint64_t, StringRef> &RHS) { 1146 return LHS < RHS.first; 1147 }); 1148 if (TargetSym != TargetSectionSymbols->begin()) { 1149 --TargetSym; 1150 uint64_t TargetAddress = std::get<0>(*TargetSym); 1151 StringRef TargetName = std::get<1>(*TargetSym); 1152 outs() << " <" << TargetName; 1153 uint64_t Disp = Target - TargetAddress; 1154 if (Disp) 1155 outs() << "+0x" << utohexstr(Disp); 1156 outs() << '>'; 1157 } 1158 } 1159 } 1160 } 1161 outs() << "\n"; 1162 1163 // Print relocation for instruction. 1164 while (rel_cur != rel_end) { 1165 bool hidden = getHidden(*rel_cur); 1166 uint64_t addr = rel_cur->getOffset(); 1167 SmallString<16> name; 1168 SmallString<32> val; 1169 1170 // If this relocation is hidden, skip it. 1171 if (hidden) goto skip_print_rel; 1172 1173 // Stop when rel_cur's address is past the current instruction. 1174 if (addr >= Index + Size) break; 1175 rel_cur->getTypeName(name); 1176 error(getRelocationValueString(*rel_cur, val)); 1177 outs() << format(Fmt.data(), SectionAddr + addr) << name 1178 << "\t" << val << "\n"; 1179 1180 skip_print_rel: 1181 ++rel_cur; 1182 } 1183 } 1184 } 1185 } 1186 } 1187 1188 void llvm::PrintRelocations(const ObjectFile *Obj) { 1189 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1190 "%08" PRIx64; 1191 // Regular objdump doesn't print relocations in non-relocatable object 1192 // files. 1193 if (!Obj->isRelocatableObject()) 1194 return; 1195 1196 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1197 if (Section.relocation_begin() == Section.relocation_end()) 1198 continue; 1199 StringRef secname; 1200 error(Section.getName(secname)); 1201 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1202 for (const RelocationRef &Reloc : Section.relocations()) { 1203 bool hidden = getHidden(Reloc); 1204 uint64_t address = Reloc.getOffset(); 1205 SmallString<32> relocname; 1206 SmallString<32> valuestr; 1207 if (hidden) 1208 continue; 1209 Reloc.getTypeName(relocname); 1210 error(getRelocationValueString(Reloc, valuestr)); 1211 outs() << format(Fmt.data(), address) << " " << relocname << " " 1212 << valuestr << "\n"; 1213 } 1214 outs() << "\n"; 1215 } 1216 } 1217 1218 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1219 outs() << "Sections:\n" 1220 "Idx Name Size Address Type\n"; 1221 unsigned i = 0; 1222 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1223 StringRef Name; 1224 error(Section.getName(Name)); 1225 uint64_t Address = Section.getAddress(); 1226 uint64_t Size = Section.getSize(); 1227 bool Text = Section.isText(); 1228 bool Data = Section.isData(); 1229 bool BSS = Section.isBSS(); 1230 std::string Type = (std::string(Text ? "TEXT " : "") + 1231 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1232 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1233 Name.str().c_str(), Size, Address, Type.c_str()); 1234 ++i; 1235 } 1236 } 1237 1238 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1239 std::error_code EC; 1240 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1241 StringRef Name; 1242 StringRef Contents; 1243 error(Section.getName(Name)); 1244 uint64_t BaseAddr = Section.getAddress(); 1245 uint64_t Size = Section.getSize(); 1246 if (!Size) 1247 continue; 1248 1249 outs() << "Contents of section " << Name << ":\n"; 1250 if (Section.isBSS()) { 1251 outs() << format("<skipping contents of bss section at [%04" PRIx64 1252 ", %04" PRIx64 ")>\n", 1253 BaseAddr, BaseAddr + Size); 1254 continue; 1255 } 1256 1257 error(Section.getContents(Contents)); 1258 1259 // Dump out the content as hex and printable ascii characters. 1260 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1261 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1262 // Dump line of hex. 1263 for (std::size_t i = 0; i < 16; ++i) { 1264 if (i != 0 && i % 4 == 0) 1265 outs() << ' '; 1266 if (addr + i < end) 1267 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1268 << hexdigit(Contents[addr + i] & 0xF, true); 1269 else 1270 outs() << " "; 1271 } 1272 // Print ascii. 1273 outs() << " "; 1274 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1275 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1276 outs() << Contents[addr + i]; 1277 else 1278 outs() << "."; 1279 } 1280 outs() << "\n"; 1281 } 1282 } 1283 } 1284 1285 void llvm::PrintSymbolTable(const ObjectFile *o) { 1286 outs() << "SYMBOL TABLE:\n"; 1287 1288 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1289 printCOFFSymbolTable(coff); 1290 return; 1291 } 1292 for (const SymbolRef &Symbol : o->symbols()) { 1293 ErrorOr<uint64_t> AddressOrError = Symbol.getAddress(); 1294 error(AddressOrError.getError()); 1295 uint64_t Address = *AddressOrError; 1296 ErrorOr<SymbolRef::Type> TypeOrError = Symbol.getType(); 1297 error(TypeOrError.getError()); 1298 SymbolRef::Type Type = *TypeOrError; 1299 uint32_t Flags = Symbol.getFlags(); 1300 ErrorOr<section_iterator> SectionOrErr = Symbol.getSection(); 1301 error(SectionOrErr.getError()); 1302 section_iterator Section = *SectionOrErr; 1303 StringRef Name; 1304 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1305 Section->getName(Name); 1306 } else { 1307 ErrorOr<StringRef> NameOrErr = Symbol.getName(); 1308 error(NameOrErr.getError()); 1309 Name = *NameOrErr; 1310 } 1311 1312 bool Global = Flags & SymbolRef::SF_Global; 1313 bool Weak = Flags & SymbolRef::SF_Weak; 1314 bool Absolute = Flags & SymbolRef::SF_Absolute; 1315 bool Common = Flags & SymbolRef::SF_Common; 1316 bool Hidden = Flags & SymbolRef::SF_Hidden; 1317 1318 char GlobLoc = ' '; 1319 if (Type != SymbolRef::ST_Unknown) 1320 GlobLoc = Global ? 'g' : 'l'; 1321 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1322 ? 'd' : ' '; 1323 char FileFunc = ' '; 1324 if (Type == SymbolRef::ST_File) 1325 FileFunc = 'f'; 1326 else if (Type == SymbolRef::ST_Function) 1327 FileFunc = 'F'; 1328 1329 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1330 "%08" PRIx64; 1331 1332 outs() << format(Fmt, Address) << " " 1333 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1334 << (Weak ? 'w' : ' ') // Weak? 1335 << ' ' // Constructor. Not supported yet. 1336 << ' ' // Warning. Not supported yet. 1337 << ' ' // Indirect reference to another symbol. 1338 << Debug // Debugging (d) or dynamic (D) symbol. 1339 << FileFunc // Name of function (F), file (f) or object (O). 1340 << ' '; 1341 if (Absolute) { 1342 outs() << "*ABS*"; 1343 } else if (Common) { 1344 outs() << "*COM*"; 1345 } else if (Section == o->section_end()) { 1346 outs() << "*UND*"; 1347 } else { 1348 if (const MachOObjectFile *MachO = 1349 dyn_cast<const MachOObjectFile>(o)) { 1350 DataRefImpl DR = Section->getRawDataRefImpl(); 1351 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1352 outs() << SegmentName << ","; 1353 } 1354 StringRef SectionName; 1355 error(Section->getName(SectionName)); 1356 outs() << SectionName; 1357 } 1358 1359 outs() << '\t'; 1360 if (Common || isa<ELFObjectFileBase>(o)) { 1361 uint64_t Val = 1362 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1363 outs() << format("\t %08" PRIx64 " ", Val); 1364 } 1365 1366 if (Hidden) { 1367 outs() << ".hidden "; 1368 } 1369 outs() << Name 1370 << '\n'; 1371 } 1372 } 1373 1374 static void PrintUnwindInfo(const ObjectFile *o) { 1375 outs() << "Unwind info:\n\n"; 1376 1377 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1378 printCOFFUnwindInfo(coff); 1379 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1380 printMachOUnwindInfo(MachO); 1381 else { 1382 // TODO: Extract DWARF dump tool to objdump. 1383 errs() << "This operation is only currently supported " 1384 "for COFF and MachO object files.\n"; 1385 return; 1386 } 1387 } 1388 1389 void llvm::printExportsTrie(const ObjectFile *o) { 1390 outs() << "Exports trie:\n"; 1391 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1392 printMachOExportsTrie(MachO); 1393 else { 1394 errs() << "This operation is only currently supported " 1395 "for Mach-O executable files.\n"; 1396 return; 1397 } 1398 } 1399 1400 void llvm::printRebaseTable(const ObjectFile *o) { 1401 outs() << "Rebase table:\n"; 1402 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1403 printMachORebaseTable(MachO); 1404 else { 1405 errs() << "This operation is only currently supported " 1406 "for Mach-O executable files.\n"; 1407 return; 1408 } 1409 } 1410 1411 void llvm::printBindTable(const ObjectFile *o) { 1412 outs() << "Bind table:\n"; 1413 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1414 printMachOBindTable(MachO); 1415 else { 1416 errs() << "This operation is only currently supported " 1417 "for Mach-O executable files.\n"; 1418 return; 1419 } 1420 } 1421 1422 void llvm::printLazyBindTable(const ObjectFile *o) { 1423 outs() << "Lazy bind table:\n"; 1424 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1425 printMachOLazyBindTable(MachO); 1426 else { 1427 errs() << "This operation is only currently supported " 1428 "for Mach-O executable files.\n"; 1429 return; 1430 } 1431 } 1432 1433 void llvm::printWeakBindTable(const ObjectFile *o) { 1434 outs() << "Weak bind table:\n"; 1435 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1436 printMachOWeakBindTable(MachO); 1437 else { 1438 errs() << "This operation is only currently supported " 1439 "for Mach-O executable files.\n"; 1440 return; 1441 } 1442 } 1443 1444 /// Dump the raw contents of the __clangast section so the output can be piped 1445 /// into llvm-bcanalyzer. 1446 void llvm::printRawClangAST(const ObjectFile *Obj) { 1447 if (outs().is_displayed()) { 1448 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1449 "the clang ast section.\n" 1450 "Please redirect the output to a file or another program such as " 1451 "llvm-bcanalyzer.\n"; 1452 return; 1453 } 1454 1455 StringRef ClangASTSectionName("__clangast"); 1456 if (isa<COFFObjectFile>(Obj)) { 1457 ClangASTSectionName = "clangast"; 1458 } 1459 1460 Optional<object::SectionRef> ClangASTSection; 1461 for (auto Sec : ToolSectionFilter(*Obj)) { 1462 StringRef Name; 1463 Sec.getName(Name); 1464 if (Name == ClangASTSectionName) { 1465 ClangASTSection = Sec; 1466 break; 1467 } 1468 } 1469 if (!ClangASTSection) 1470 return; 1471 1472 StringRef ClangASTContents; 1473 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1474 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1475 } 1476 1477 static void printFaultMaps(const ObjectFile *Obj) { 1478 const char *FaultMapSectionName = nullptr; 1479 1480 if (isa<ELFObjectFileBase>(Obj)) { 1481 FaultMapSectionName = ".llvm_faultmaps"; 1482 } else if (isa<MachOObjectFile>(Obj)) { 1483 FaultMapSectionName = "__llvm_faultmaps"; 1484 } else { 1485 errs() << "This operation is only currently supported " 1486 "for ELF and Mach-O executable files.\n"; 1487 return; 1488 } 1489 1490 Optional<object::SectionRef> FaultMapSection; 1491 1492 for (auto Sec : ToolSectionFilter(*Obj)) { 1493 StringRef Name; 1494 Sec.getName(Name); 1495 if (Name == FaultMapSectionName) { 1496 FaultMapSection = Sec; 1497 break; 1498 } 1499 } 1500 1501 outs() << "FaultMap table:\n"; 1502 1503 if (!FaultMapSection.hasValue()) { 1504 outs() << "<not found>\n"; 1505 return; 1506 } 1507 1508 StringRef FaultMapContents; 1509 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1510 1511 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1512 FaultMapContents.bytes_end()); 1513 1514 outs() << FMP; 1515 } 1516 1517 static void printPrivateFileHeaders(const ObjectFile *o) { 1518 if (o->isELF()) 1519 printELFFileHeader(o); 1520 else if (o->isCOFF()) 1521 printCOFFFileHeader(o); 1522 else if (o->isMachO()) { 1523 printMachOFileHeader(o); 1524 printMachOLoadCommands(o); 1525 } else 1526 report_fatal_error("Invalid/Unsupported object file format"); 1527 } 1528 1529 static void printFirstPrivateFileHeader(const ObjectFile *o) { 1530 if (o->isELF()) 1531 printELFFileHeader(o); 1532 else if (o->isCOFF()) 1533 printCOFFFileHeader(o); 1534 else if (o->isMachO()) 1535 printMachOFileHeader(o); 1536 else 1537 report_fatal_error("Invalid/Unsupported object file format"); 1538 } 1539 1540 static void DumpObject(const ObjectFile *o) { 1541 // Avoid other output when using a raw option. 1542 if (!RawClangAST) { 1543 outs() << '\n'; 1544 outs() << o->getFileName() 1545 << ":\tfile format " << o->getFileFormatName() << "\n\n"; 1546 } 1547 1548 if (Disassemble) 1549 DisassembleObject(o, Relocations); 1550 if (Relocations && !Disassemble) 1551 PrintRelocations(o); 1552 if (SectionHeaders) 1553 PrintSectionHeaders(o); 1554 if (SectionContents) 1555 PrintSectionContents(o); 1556 if (SymbolTable) 1557 PrintSymbolTable(o); 1558 if (UnwindInfo) 1559 PrintUnwindInfo(o); 1560 if (PrivateHeaders) 1561 printPrivateFileHeaders(o); 1562 if (FirstPrivateHeader) 1563 printFirstPrivateFileHeader(o); 1564 if (ExportsTrie) 1565 printExportsTrie(o); 1566 if (Rebase) 1567 printRebaseTable(o); 1568 if (Bind) 1569 printBindTable(o); 1570 if (LazyBind) 1571 printLazyBindTable(o); 1572 if (WeakBind) 1573 printWeakBindTable(o); 1574 if (RawClangAST) 1575 printRawClangAST(o); 1576 if (PrintFaultMaps) 1577 printFaultMaps(o); 1578 if (DwarfDumpType != DIDT_Null) { 1579 std::unique_ptr<DIContext> DICtx(new DWARFContextInMemory(*o)); 1580 // Dump the complete DWARF structure. 1581 DICtx->dump(outs(), DwarfDumpType, true /* DumpEH */); 1582 } 1583 } 1584 1585 /// @brief Dump each object file in \a a; 1586 static void DumpArchive(const Archive *a) { 1587 for (auto &ErrorOrChild : a->children()) { 1588 if (std::error_code EC = ErrorOrChild.getError()) 1589 report_error(a->getFileName(), EC); 1590 const Archive::Child &C = *ErrorOrChild; 1591 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1592 if (std::error_code EC = ChildOrErr.getError()) 1593 if (EC != object_error::invalid_file_type) 1594 report_error(a->getFileName(), EC); 1595 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 1596 DumpObject(o); 1597 else 1598 report_error(a->getFileName(), object_error::invalid_file_type); 1599 } 1600 } 1601 1602 /// @brief Open file and figure out how to dump it. 1603 static void DumpInput(StringRef file) { 1604 1605 // If we are using the Mach-O specific object file parser, then let it parse 1606 // the file and process the command line options. So the -arch flags can 1607 // be used to select specific slices, etc. 1608 if (MachOOpt) { 1609 ParseInputMachO(file); 1610 return; 1611 } 1612 1613 // Attempt to open the binary. 1614 ErrorOr<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 1615 if (std::error_code EC = BinaryOrErr.getError()) 1616 report_error(file, EC); 1617 Binary &Binary = *BinaryOrErr.get().getBinary(); 1618 1619 if (Archive *a = dyn_cast<Archive>(&Binary)) 1620 DumpArchive(a); 1621 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 1622 DumpObject(o); 1623 else 1624 report_error(file, object_error::invalid_file_type); 1625 } 1626 1627 int main(int argc, char **argv) { 1628 // Print a stack trace if we signal out. 1629 sys::PrintStackTraceOnErrorSignal(); 1630 PrettyStackTraceProgram X(argc, argv); 1631 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 1632 1633 // Initialize targets and assembly printers/parsers. 1634 llvm::InitializeAllTargetInfos(); 1635 llvm::InitializeAllTargetMCs(); 1636 llvm::InitializeAllDisassemblers(); 1637 1638 // Register the target printer for --version. 1639 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 1640 1641 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 1642 TripleName = Triple::normalize(TripleName); 1643 1644 ToolName = argv[0]; 1645 1646 // Defaults to a.out if no filenames specified. 1647 if (InputFilenames.size() == 0) 1648 InputFilenames.push_back("a.out"); 1649 1650 if (DisassembleAll) 1651 Disassemble = true; 1652 if (!Disassemble 1653 && !Relocations 1654 && !SectionHeaders 1655 && !SectionContents 1656 && !SymbolTable 1657 && !UnwindInfo 1658 && !PrivateHeaders 1659 && !FirstPrivateHeader 1660 && !ExportsTrie 1661 && !Rebase 1662 && !Bind 1663 && !LazyBind 1664 && !WeakBind 1665 && !RawClangAST 1666 && !(UniversalHeaders && MachOOpt) 1667 && !(ArchiveHeaders && MachOOpt) 1668 && !(IndirectSymbols && MachOOpt) 1669 && !(DataInCode && MachOOpt) 1670 && !(LinkOptHints && MachOOpt) 1671 && !(InfoPlist && MachOOpt) 1672 && !(DylibsUsed && MachOOpt) 1673 && !(DylibId && MachOOpt) 1674 && !(ObjcMetaData && MachOOpt) 1675 && !(FilterSections.size() != 0 && MachOOpt) 1676 && !PrintFaultMaps 1677 && DwarfDumpType == DIDT_Null) { 1678 cl::PrintHelpMessage(); 1679 return 2; 1680 } 1681 1682 std::for_each(InputFilenames.begin(), InputFilenames.end(), 1683 DumpInput); 1684 1685 return EXIT_SUCCESS; 1686 } 1687