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