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