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 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 274 llvm::Error E) { 275 assert(E); 276 std::string Buf; 277 raw_string_ostream OS(Buf); 278 logAllUnhandledErrors(std::move(E), OS, ""); 279 OS.flush(); 280 errs() << ToolName << ": " << Buf; 281 exit(1); 282 } 283 284 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 285 // Figure out the target triple. 286 llvm::Triple TheTriple("unknown-unknown-unknown"); 287 if (TripleName.empty()) { 288 if (Obj) { 289 TheTriple.setArch(Triple::ArchType(Obj->getArch())); 290 // TheTriple defaults to ELF, and COFF doesn't have an environment: 291 // the best we can do here is indicate that it is mach-o. 292 if (Obj->isMachO()) 293 TheTriple.setObjectFormat(Triple::MachO); 294 295 if (Obj->isCOFF()) { 296 const auto COFFObj = dyn_cast<COFFObjectFile>(Obj); 297 if (COFFObj->getArch() == Triple::thumb) 298 TheTriple.setTriple("thumbv7-windows"); 299 } 300 } 301 } else 302 TheTriple.setTriple(Triple::normalize(TripleName)); 303 304 // Get the target specific parser. 305 std::string Error; 306 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 307 Error); 308 if (!TheTarget) 309 report_fatal_error("can't find target: " + Error); 310 311 // Update the triple name and return the found target. 312 TripleName = TheTriple.getTriple(); 313 return TheTarget; 314 } 315 316 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 317 return a.getOffset() < b.getOffset(); 318 } 319 320 namespace { 321 class PrettyPrinter { 322 public: 323 virtual ~PrettyPrinter(){} 324 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 325 ArrayRef<uint8_t> Bytes, uint64_t Address, 326 raw_ostream &OS, StringRef Annot, 327 MCSubtargetInfo const &STI) { 328 OS << format("%8" PRIx64 ":", Address); 329 if (!NoShowRawInsn) { 330 OS << "\t"; 331 dumpBytes(Bytes, OS); 332 } 333 if (MI) 334 IP.printInst(MI, OS, "", STI); 335 else 336 OS << " <unknown>"; 337 } 338 }; 339 PrettyPrinter PrettyPrinterInst; 340 class HexagonPrettyPrinter : public PrettyPrinter { 341 public: 342 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 343 raw_ostream &OS) { 344 uint32_t opcode = 345 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 346 OS << format("%8" PRIx64 ":", Address); 347 if (!NoShowRawInsn) { 348 OS << "\t"; 349 dumpBytes(Bytes.slice(0, 4), OS); 350 OS << format("%08" PRIx32, opcode); 351 } 352 } 353 void printInst(MCInstPrinter &IP, const MCInst *MI, 354 ArrayRef<uint8_t> Bytes, uint64_t Address, 355 raw_ostream &OS, StringRef Annot, 356 MCSubtargetInfo const &STI) override { 357 if (!MI) { 358 printLead(Bytes, Address, OS); 359 OS << " <unknown>"; 360 return; 361 } 362 std::string Buffer; 363 { 364 raw_string_ostream TempStream(Buffer); 365 IP.printInst(MI, TempStream, "", STI); 366 } 367 StringRef Contents(Buffer); 368 // Split off bundle attributes 369 auto PacketBundle = Contents.rsplit('\n'); 370 // Split off first instruction from the rest 371 auto HeadTail = PacketBundle.first.split('\n'); 372 auto Preamble = " { "; 373 auto Separator = ""; 374 while(!HeadTail.first.empty()) { 375 OS << Separator; 376 Separator = "\n"; 377 printLead(Bytes, Address, OS); 378 OS << Preamble; 379 Preamble = " "; 380 StringRef Inst; 381 auto Duplex = HeadTail.first.split('\v'); 382 if(!Duplex.second.empty()){ 383 OS << Duplex.first; 384 OS << "; "; 385 Inst = Duplex.second; 386 } 387 else 388 Inst = HeadTail.first; 389 OS << Inst; 390 Bytes = Bytes.slice(4); 391 Address += 4; 392 HeadTail = HeadTail.second.split('\n'); 393 } 394 OS << " } " << PacketBundle.second; 395 } 396 }; 397 HexagonPrettyPrinter HexagonPrettyPrinterInst; 398 399 class AMDGCNPrettyPrinter : public PrettyPrinter { 400 public: 401 void printInst(MCInstPrinter &IP, 402 const MCInst *MI, 403 ArrayRef<uint8_t> Bytes, 404 uint64_t Address, 405 raw_ostream &OS, 406 StringRef Annot, 407 MCSubtargetInfo const &STI) override { 408 SmallString<40> InstStr; 409 raw_svector_ostream IS(InstStr); 410 411 IP.printInst(MI, IS, "", STI); 412 413 OS << left_justify(IS.str(), 60) << format("// %012" PRIX64 ": ", Address); 414 typedef support::ulittle32_t U32; 415 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 416 Bytes.size() / sizeof(U32))) 417 // D should be explicitly casted to uint32_t here as it is passed 418 // by format to snprintf as vararg. 419 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 420 421 if (!Annot.empty()) 422 OS << "// " << Annot; 423 } 424 }; 425 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 426 427 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 428 switch(Triple.getArch()) { 429 default: 430 return PrettyPrinterInst; 431 case Triple::hexagon: 432 return HexagonPrettyPrinterInst; 433 case Triple::amdgcn: 434 return AMDGCNPrettyPrinterInst; 435 } 436 } 437 } 438 439 template <class ELFT> 440 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 441 const RelocationRef &RelRef, 442 SmallVectorImpl<char> &Result) { 443 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 444 445 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 446 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 447 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 448 449 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 450 451 ErrorOr<const Elf_Shdr *> SecOrErr = EF.getSection(Rel.d.a); 452 if (std::error_code EC = SecOrErr.getError()) 453 return EC; 454 const Elf_Shdr *Sec = *SecOrErr; 455 ErrorOr<const Elf_Shdr *> SymTabOrErr = EF.getSection(Sec->sh_link); 456 if (std::error_code EC = SymTabOrErr.getError()) 457 return EC; 458 const Elf_Shdr *SymTab = *SymTabOrErr; 459 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 460 SymTab->sh_type == ELF::SHT_DYNSYM); 461 ErrorOr<const Elf_Shdr *> StrTabSec = EF.getSection(SymTab->sh_link); 462 if (std::error_code EC = StrTabSec.getError()) 463 return EC; 464 ErrorOr<StringRef> StrTabOrErr = EF.getStringTable(*StrTabSec); 465 if (std::error_code EC = StrTabOrErr.getError()) 466 return EC; 467 StringRef StrTab = *StrTabOrErr; 468 uint8_t type = RelRef.getType(); 469 StringRef res; 470 int64_t addend = 0; 471 switch (Sec->sh_type) { 472 default: 473 return object_error::parse_failed; 474 case ELF::SHT_REL: { 475 // TODO: Read implicit addend from section data. 476 break; 477 } 478 case ELF::SHT_RELA: { 479 const Elf_Rela *ERela = Obj->getRela(Rel); 480 addend = ERela->r_addend; 481 break; 482 } 483 } 484 symbol_iterator SI = RelRef.getSymbol(); 485 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 486 StringRef Target; 487 if (symb->getType() == ELF::STT_SECTION) { 488 ErrorOr<section_iterator> SymSI = SI->getSection(); 489 if (std::error_code EC = SymSI.getError()) 490 return EC; 491 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 492 ErrorOr<StringRef> SecName = EF.getSectionName(SymSec); 493 if (std::error_code EC = SecName.getError()) 494 return EC; 495 Target = *SecName; 496 } else { 497 Expected<StringRef> SymName = symb->getName(StrTab); 498 if (!SymName) 499 return errorToErrorCode(SymName.takeError()); 500 Target = *SymName; 501 } 502 switch (EF.getHeader()->e_machine) { 503 case ELF::EM_X86_64: 504 switch (type) { 505 case ELF::R_X86_64_PC8: 506 case ELF::R_X86_64_PC16: 507 case ELF::R_X86_64_PC32: { 508 std::string fmtbuf; 509 raw_string_ostream fmt(fmtbuf); 510 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 511 fmt.flush(); 512 Result.append(fmtbuf.begin(), fmtbuf.end()); 513 } break; 514 case ELF::R_X86_64_8: 515 case ELF::R_X86_64_16: 516 case ELF::R_X86_64_32: 517 case ELF::R_X86_64_32S: 518 case ELF::R_X86_64_64: { 519 std::string fmtbuf; 520 raw_string_ostream fmt(fmtbuf); 521 fmt << Target << (addend < 0 ? "" : "+") << addend; 522 fmt.flush(); 523 Result.append(fmtbuf.begin(), fmtbuf.end()); 524 } break; 525 default: 526 res = "Unknown"; 527 } 528 break; 529 case ELF::EM_LANAI: 530 case ELF::EM_AARCH64: { 531 std::string fmtbuf; 532 raw_string_ostream fmt(fmtbuf); 533 fmt << Target; 534 if (addend != 0) 535 fmt << (addend < 0 ? "" : "+") << addend; 536 fmt.flush(); 537 Result.append(fmtbuf.begin(), fmtbuf.end()); 538 break; 539 } 540 case ELF::EM_386: 541 case ELF::EM_IAMCU: 542 case ELF::EM_ARM: 543 case ELF::EM_HEXAGON: 544 case ELF::EM_MIPS: 545 res = Target; 546 break; 547 case ELF::EM_WEBASSEMBLY: 548 switch (type) { 549 case ELF::R_WEBASSEMBLY_DATA: { 550 std::string fmtbuf; 551 raw_string_ostream fmt(fmtbuf); 552 fmt << Target << (addend < 0 ? "" : "+") << addend; 553 fmt.flush(); 554 Result.append(fmtbuf.begin(), fmtbuf.end()); 555 break; 556 } 557 case ELF::R_WEBASSEMBLY_FUNCTION: 558 res = Target; 559 break; 560 default: 561 res = "Unknown"; 562 } 563 break; 564 default: 565 res = "Unknown"; 566 } 567 if (Result.empty()) 568 Result.append(res.begin(), res.end()); 569 return std::error_code(); 570 } 571 572 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 573 const RelocationRef &Rel, 574 SmallVectorImpl<char> &Result) { 575 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 576 return getRelocationValueString(ELF32LE, Rel, Result); 577 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 578 return getRelocationValueString(ELF64LE, Rel, Result); 579 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 580 return getRelocationValueString(ELF32BE, Rel, Result); 581 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 582 return getRelocationValueString(ELF64BE, Rel, Result); 583 } 584 585 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 586 const RelocationRef &Rel, 587 SmallVectorImpl<char> &Result) { 588 symbol_iterator SymI = Rel.getSymbol(); 589 Expected<StringRef> SymNameOrErr = SymI->getName(); 590 if (!SymNameOrErr) 591 return errorToErrorCode(SymNameOrErr.takeError()); 592 StringRef SymName = *SymNameOrErr; 593 Result.append(SymName.begin(), SymName.end()); 594 return std::error_code(); 595 } 596 597 static void printRelocationTargetName(const MachOObjectFile *O, 598 const MachO::any_relocation_info &RE, 599 raw_string_ostream &fmt) { 600 bool IsScattered = O->isRelocationScattered(RE); 601 602 // Target of a scattered relocation is an address. In the interest of 603 // generating pretty output, scan through the symbol table looking for a 604 // symbol that aligns with that address. If we find one, print it. 605 // Otherwise, we just print the hex address of the target. 606 if (IsScattered) { 607 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 608 609 for (const SymbolRef &Symbol : O->symbols()) { 610 std::error_code ec; 611 ErrorOr<uint64_t> Addr = Symbol.getAddress(); 612 if ((ec = Addr.getError())) 613 report_fatal_error(ec.message()); 614 if (*Addr != Val) 615 continue; 616 Expected<StringRef> Name = Symbol.getName(); 617 if (!Name) { 618 std::string Buf; 619 raw_string_ostream OS(Buf); 620 logAllUnhandledErrors(Name.takeError(), OS, ""); 621 OS.flush(); 622 report_fatal_error(Buf); 623 } 624 fmt << *Name; 625 return; 626 } 627 628 // If we couldn't find a symbol that this relocation refers to, try 629 // to find a section beginning instead. 630 for (const SectionRef &Section : ToolSectionFilter(*O)) { 631 std::error_code ec; 632 633 StringRef Name; 634 uint64_t Addr = Section.getAddress(); 635 if (Addr != Val) 636 continue; 637 if ((ec = Section.getName(Name))) 638 report_fatal_error(ec.message()); 639 fmt << Name; 640 return; 641 } 642 643 fmt << format("0x%x", Val); 644 return; 645 } 646 647 StringRef S; 648 bool isExtern = O->getPlainRelocationExternal(RE); 649 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 650 651 if (isExtern) { 652 symbol_iterator SI = O->symbol_begin(); 653 advance(SI, Val); 654 Expected<StringRef> SOrErr = SI->getName(); 655 error(errorToErrorCode(SOrErr.takeError())); 656 S = *SOrErr; 657 } else { 658 section_iterator SI = O->section_begin(); 659 // Adjust for the fact that sections are 1-indexed. 660 advance(SI, Val - 1); 661 SI->getName(S); 662 } 663 664 fmt << S; 665 } 666 667 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 668 const RelocationRef &RelRef, 669 SmallVectorImpl<char> &Result) { 670 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 671 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 672 673 unsigned Arch = Obj->getArch(); 674 675 std::string fmtbuf; 676 raw_string_ostream fmt(fmtbuf); 677 unsigned Type = Obj->getAnyRelocationType(RE); 678 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 679 680 // Determine any addends that should be displayed with the relocation. 681 // These require decoding the relocation type, which is triple-specific. 682 683 // X86_64 has entirely custom relocation types. 684 if (Arch == Triple::x86_64) { 685 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 686 687 switch (Type) { 688 case MachO::X86_64_RELOC_GOT_LOAD: 689 case MachO::X86_64_RELOC_GOT: { 690 printRelocationTargetName(Obj, RE, fmt); 691 fmt << "@GOT"; 692 if (isPCRel) 693 fmt << "PCREL"; 694 break; 695 } 696 case MachO::X86_64_RELOC_SUBTRACTOR: { 697 DataRefImpl RelNext = Rel; 698 Obj->moveRelocationNext(RelNext); 699 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 700 701 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 702 // X86_64_RELOC_UNSIGNED. 703 // NOTE: Scattered relocations don't exist on x86_64. 704 unsigned RType = Obj->getAnyRelocationType(RENext); 705 if (RType != MachO::X86_64_RELOC_UNSIGNED) 706 report_fatal_error("Expected X86_64_RELOC_UNSIGNED after " 707 "X86_64_RELOC_SUBTRACTOR."); 708 709 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 710 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 711 printRelocationTargetName(Obj, RENext, fmt); 712 fmt << "-"; 713 printRelocationTargetName(Obj, RE, fmt); 714 break; 715 } 716 case MachO::X86_64_RELOC_TLV: 717 printRelocationTargetName(Obj, RE, fmt); 718 fmt << "@TLV"; 719 if (isPCRel) 720 fmt << "P"; 721 break; 722 case MachO::X86_64_RELOC_SIGNED_1: 723 printRelocationTargetName(Obj, RE, fmt); 724 fmt << "-1"; 725 break; 726 case MachO::X86_64_RELOC_SIGNED_2: 727 printRelocationTargetName(Obj, RE, fmt); 728 fmt << "-2"; 729 break; 730 case MachO::X86_64_RELOC_SIGNED_4: 731 printRelocationTargetName(Obj, RE, fmt); 732 fmt << "-4"; 733 break; 734 default: 735 printRelocationTargetName(Obj, RE, fmt); 736 break; 737 } 738 // X86 and ARM share some relocation types in common. 739 } else if (Arch == Triple::x86 || Arch == Triple::arm || 740 Arch == Triple::ppc) { 741 // Generic relocation types... 742 switch (Type) { 743 case MachO::GENERIC_RELOC_PAIR: // prints no info 744 return std::error_code(); 745 case MachO::GENERIC_RELOC_SECTDIFF: { 746 DataRefImpl RelNext = Rel; 747 Obj->moveRelocationNext(RelNext); 748 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 749 750 // X86 sect diff's must be followed by a relocation of type 751 // GENERIC_RELOC_PAIR. 752 unsigned RType = Obj->getAnyRelocationType(RENext); 753 754 if (RType != MachO::GENERIC_RELOC_PAIR) 755 report_fatal_error("Expected GENERIC_RELOC_PAIR after " 756 "GENERIC_RELOC_SECTDIFF."); 757 758 printRelocationTargetName(Obj, RE, fmt); 759 fmt << "-"; 760 printRelocationTargetName(Obj, RENext, fmt); 761 break; 762 } 763 } 764 765 if (Arch == Triple::x86 || Arch == Triple::ppc) { 766 switch (Type) { 767 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 768 DataRefImpl RelNext = Rel; 769 Obj->moveRelocationNext(RelNext); 770 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 771 772 // X86 sect diff's must be followed by a relocation of type 773 // GENERIC_RELOC_PAIR. 774 unsigned RType = Obj->getAnyRelocationType(RENext); 775 if (RType != MachO::GENERIC_RELOC_PAIR) 776 report_fatal_error("Expected GENERIC_RELOC_PAIR after " 777 "GENERIC_RELOC_LOCAL_SECTDIFF."); 778 779 printRelocationTargetName(Obj, RE, fmt); 780 fmt << "-"; 781 printRelocationTargetName(Obj, RENext, fmt); 782 break; 783 } 784 case MachO::GENERIC_RELOC_TLV: { 785 printRelocationTargetName(Obj, RE, fmt); 786 fmt << "@TLV"; 787 if (IsPCRel) 788 fmt << "P"; 789 break; 790 } 791 default: 792 printRelocationTargetName(Obj, RE, fmt); 793 } 794 } else { // ARM-specific relocations 795 switch (Type) { 796 case MachO::ARM_RELOC_HALF: 797 case MachO::ARM_RELOC_HALF_SECTDIFF: { 798 // Half relocations steal a bit from the length field to encode 799 // whether this is an upper16 or a lower16 relocation. 800 bool isUpper = Obj->getAnyRelocationLength(RE) >> 1; 801 802 if (isUpper) 803 fmt << ":upper16:("; 804 else 805 fmt << ":lower16:("; 806 printRelocationTargetName(Obj, RE, fmt); 807 808 DataRefImpl RelNext = Rel; 809 Obj->moveRelocationNext(RelNext); 810 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 811 812 // ARM half relocs must be followed by a relocation of type 813 // ARM_RELOC_PAIR. 814 unsigned RType = Obj->getAnyRelocationType(RENext); 815 if (RType != MachO::ARM_RELOC_PAIR) 816 report_fatal_error("Expected ARM_RELOC_PAIR after " 817 "ARM_RELOC_HALF"); 818 819 // NOTE: The half of the target virtual address is stashed in the 820 // address field of the secondary relocation, but we can't reverse 821 // engineer the constant offset from it without decoding the movw/movt 822 // instruction to find the other half in its immediate field. 823 824 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 825 // symbol/section pointer of the follow-on relocation. 826 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 827 fmt << "-"; 828 printRelocationTargetName(Obj, RENext, fmt); 829 } 830 831 fmt << ")"; 832 break; 833 } 834 default: { printRelocationTargetName(Obj, RE, fmt); } 835 } 836 } 837 } else 838 printRelocationTargetName(Obj, RE, fmt); 839 840 fmt.flush(); 841 Result.append(fmtbuf.begin(), fmtbuf.end()); 842 return std::error_code(); 843 } 844 845 static std::error_code getRelocationValueString(const RelocationRef &Rel, 846 SmallVectorImpl<char> &Result) { 847 const ObjectFile *Obj = Rel.getObject(); 848 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 849 return getRelocationValueString(ELF, Rel, Result); 850 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 851 return getRelocationValueString(COFF, Rel, Result); 852 auto *MachO = cast<MachOObjectFile>(Obj); 853 return getRelocationValueString(MachO, Rel, Result); 854 } 855 856 /// @brief Indicates whether this relocation should hidden when listing 857 /// relocations, usually because it is the trailing part of a multipart 858 /// relocation that will be printed as part of the leading relocation. 859 static bool getHidden(RelocationRef RelRef) { 860 const ObjectFile *Obj = RelRef.getObject(); 861 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 862 if (!MachO) 863 return false; 864 865 unsigned Arch = MachO->getArch(); 866 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 867 uint64_t Type = MachO->getRelocationType(Rel); 868 869 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 870 // is always hidden. 871 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 872 if (Type == MachO::GENERIC_RELOC_PAIR) 873 return true; 874 } else if (Arch == Triple::x86_64) { 875 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 876 // an X86_64_RELOC_SUBTRACTOR. 877 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 878 DataRefImpl RelPrev = Rel; 879 RelPrev.d.a--; 880 uint64_t PrevType = MachO->getRelocationType(RelPrev); 881 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 882 return true; 883 } 884 } 885 886 return false; 887 } 888 889 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 890 const Target *TheTarget = getTarget(Obj); 891 892 // Package up features to be passed to target/subtarget 893 std::string FeaturesStr; 894 if (MAttrs.size()) { 895 SubtargetFeatures Features; 896 for (unsigned i = 0; i != MAttrs.size(); ++i) 897 Features.AddFeature(MAttrs[i]); 898 FeaturesStr = Features.getString(); 899 } 900 901 std::unique_ptr<const MCRegisterInfo> MRI( 902 TheTarget->createMCRegInfo(TripleName)); 903 if (!MRI) 904 report_fatal_error("error: no register info for target " + TripleName); 905 906 // Set up disassembler. 907 std::unique_ptr<const MCAsmInfo> AsmInfo( 908 TheTarget->createMCAsmInfo(*MRI, TripleName)); 909 if (!AsmInfo) 910 report_fatal_error("error: no assembly info for target " + TripleName); 911 std::unique_ptr<const MCSubtargetInfo> STI( 912 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr)); 913 if (!STI) 914 report_fatal_error("error: no subtarget info for target " + TripleName); 915 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 916 if (!MII) 917 report_fatal_error("error: no instruction info for target " + TripleName); 918 std::unique_ptr<const MCObjectFileInfo> MOFI(new MCObjectFileInfo); 919 MCContext Ctx(AsmInfo.get(), MRI.get(), MOFI.get()); 920 921 std::unique_ptr<MCDisassembler> DisAsm( 922 TheTarget->createMCDisassembler(*STI, Ctx)); 923 if (!DisAsm) 924 report_fatal_error("error: no disassembler for target " + TripleName); 925 926 std::unique_ptr<const MCInstrAnalysis> MIA( 927 TheTarget->createMCInstrAnalysis(MII.get())); 928 929 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 930 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 931 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 932 if (!IP) 933 report_fatal_error("error: no instruction printer for target " + 934 TripleName); 935 IP->setPrintImmHex(PrintImmHex); 936 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 937 938 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 939 "\t\t\t%08" PRIx64 ": "; 940 941 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 942 // in RelocSecs contain the relocations for section S. 943 std::error_code EC; 944 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 945 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 946 section_iterator Sec2 = Section.getRelocatedSection(); 947 if (Sec2 != Obj->section_end()) 948 SectionRelocMap[*Sec2].push_back(Section); 949 } 950 951 // Create a mapping from virtual address to symbol name. This is used to 952 // pretty print the symbols while disassembling. 953 typedef std::vector<std::pair<uint64_t, StringRef>> SectionSymbolsTy; 954 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 955 for (const SymbolRef &Symbol : Obj->symbols()) { 956 ErrorOr<uint64_t> AddressOrErr = Symbol.getAddress(); 957 error(AddressOrErr.getError()); 958 uint64_t Address = *AddressOrErr; 959 960 Expected<StringRef> Name = Symbol.getName(); 961 error(errorToErrorCode(Name.takeError())); 962 if (Name->empty()) 963 continue; 964 965 ErrorOr<section_iterator> SectionOrErr = Symbol.getSection(); 966 error(SectionOrErr.getError()); 967 section_iterator SecI = *SectionOrErr; 968 if (SecI == Obj->section_end()) 969 continue; 970 971 AllSymbols[*SecI].emplace_back(Address, *Name); 972 } 973 974 // Create a mapping from virtual address to section. 975 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 976 for (SectionRef Sec : Obj->sections()) 977 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 978 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 979 980 // Linked executables (.exe and .dll files) typically don't include a real 981 // symbol table but they might contain an export table. 982 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 983 for (const auto &ExportEntry : COFFObj->export_directories()) { 984 StringRef Name; 985 error(ExportEntry.getSymbolName(Name)); 986 if (Name.empty()) 987 continue; 988 uint32_t RVA; 989 error(ExportEntry.getExportRVA(RVA)); 990 991 uint64_t VA = COFFObj->getImageBase() + RVA; 992 auto Sec = std::upper_bound( 993 SectionAddresses.begin(), SectionAddresses.end(), VA, 994 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 995 return LHS < RHS.first; 996 }); 997 if (Sec != SectionAddresses.begin()) 998 --Sec; 999 else 1000 Sec = SectionAddresses.end(); 1001 1002 if (Sec != SectionAddresses.end()) 1003 AllSymbols[Sec->second].emplace_back(VA, Name); 1004 } 1005 } 1006 1007 // Sort all the symbols, this allows us to use a simple binary search to find 1008 // a symbol near an address. 1009 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1010 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1011 1012 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1013 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1014 continue; 1015 1016 uint64_t SectionAddr = Section.getAddress(); 1017 uint64_t SectSize = Section.getSize(); 1018 if (!SectSize) 1019 continue; 1020 1021 // Get the list of all the symbols in this section. 1022 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1023 std::vector<uint64_t> DataMappingSymsAddr; 1024 std::vector<uint64_t> TextMappingSymsAddr; 1025 if (Obj->isELF() && Obj->getArch() == Triple::aarch64) { 1026 for (const auto &Symb : Symbols) { 1027 uint64_t Address = Symb.first; 1028 StringRef Name = Symb.second; 1029 if (Name.startswith("$d")) 1030 DataMappingSymsAddr.push_back(Address - SectionAddr); 1031 if (Name.startswith("$x")) 1032 TextMappingSymsAddr.push_back(Address - SectionAddr); 1033 } 1034 } 1035 1036 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1037 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1038 1039 // Make a list of all the relocations for this section. 1040 std::vector<RelocationRef> Rels; 1041 if (InlineRelocs) { 1042 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1043 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1044 Rels.push_back(Reloc); 1045 } 1046 } 1047 } 1048 1049 // Sort relocations by address. 1050 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1051 1052 StringRef SegmentName = ""; 1053 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1054 DataRefImpl DR = Section.getRawDataRefImpl(); 1055 SegmentName = MachO->getSectionFinalSegmentName(DR); 1056 } 1057 StringRef name; 1058 error(Section.getName(name)); 1059 outs() << "Disassembly of section "; 1060 if (!SegmentName.empty()) 1061 outs() << SegmentName << ","; 1062 outs() << name << ':'; 1063 1064 // If the section has no symbol at the start, just insert a dummy one. 1065 if (Symbols.empty() || Symbols[0].first != 0) 1066 Symbols.insert(Symbols.begin(), std::make_pair(SectionAddr, name)); 1067 1068 SmallString<40> Comments; 1069 raw_svector_ostream CommentStream(Comments); 1070 1071 StringRef BytesStr; 1072 error(Section.getContents(BytesStr)); 1073 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1074 BytesStr.size()); 1075 1076 uint64_t Size; 1077 uint64_t Index; 1078 1079 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1080 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1081 // Disassemble symbol by symbol. 1082 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1083 1084 uint64_t Start = Symbols[si].first - SectionAddr; 1085 // The end is either the section end or the beginning of the next 1086 // symbol. 1087 uint64_t End = 1088 (si == se - 1) ? SectSize : Symbols[si + 1].first - SectionAddr; 1089 // Don't try to disassemble beyond the end of section contents. 1090 if (End > SectSize) 1091 End = SectSize; 1092 // If this symbol has the same address as the next symbol, then skip it. 1093 if (Start >= End) 1094 continue; 1095 1096 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1097 // make size 4 bytes folded 1098 End = Start + ((End - Start) & ~0x3ull); 1099 Start += 256; // add sizeof(amd_kernel_code_t) 1100 // cut trailing zeroes - up to 256 bytes (align) 1101 const uint64_t EndAlign = 256; 1102 const auto Limit = End - (std::min)(EndAlign, End - Start); 1103 while (End > Limit && 1104 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1105 End -= 4; 1106 } 1107 1108 outs() << '\n' << Symbols[si].second << ":\n"; 1109 1110 #ifndef NDEBUG 1111 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1112 #else 1113 raw_ostream &DebugOut = nulls(); 1114 #endif 1115 1116 for (Index = Start; Index < End; Index += Size) { 1117 MCInst Inst; 1118 1119 // AArch64 ELF binaries can interleave data and text in the 1120 // same section. We rely on the markers introduced to 1121 // understand what we need to dump. 1122 if (Obj->isELF() && Obj->getArch() == Triple::aarch64) { 1123 uint64_t Stride = 0; 1124 1125 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1126 DataMappingSymsAddr.end(), Index); 1127 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1128 // Switch to data. 1129 while (Index < End) { 1130 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1131 outs() << "\t"; 1132 if (Index + 4 <= End) { 1133 Stride = 4; 1134 dumpBytes(Bytes.slice(Index, 4), outs()); 1135 outs() << "\t.word"; 1136 } else if (Index + 2 <= End) { 1137 Stride = 2; 1138 dumpBytes(Bytes.slice(Index, 2), outs()); 1139 outs() << "\t.short"; 1140 } else { 1141 Stride = 1; 1142 dumpBytes(Bytes.slice(Index, 1), outs()); 1143 outs() << "\t.byte"; 1144 } 1145 Index += Stride; 1146 outs() << "\n"; 1147 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1148 TextMappingSymsAddr.end(), Index); 1149 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1150 break; 1151 } 1152 } 1153 } 1154 1155 if (Index >= End) 1156 break; 1157 1158 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1159 SectionAddr + Index, DebugOut, 1160 CommentStream); 1161 if (Size == 0) 1162 Size = 1; 1163 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1164 Bytes.slice(Index, Size), 1165 SectionAddr + Index, outs(), "", *STI); 1166 outs() << CommentStream.str(); 1167 Comments.clear(); 1168 1169 // Try to resolve the target of a call, tail call, etc. to a specific 1170 // symbol. 1171 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1172 MIA->isConditionalBranch(Inst))) { 1173 uint64_t Target; 1174 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1175 // In a relocatable object, the target's section must reside in 1176 // the same section as the call instruction or it is accessed 1177 // through a relocation. 1178 // 1179 // In a non-relocatable object, the target may be in any section. 1180 // 1181 // N.B. We don't walk the relocations in the relocatable case yet. 1182 auto *TargetSectionSymbols = &Symbols; 1183 if (!Obj->isRelocatableObject()) { 1184 auto SectionAddress = std::upper_bound( 1185 SectionAddresses.begin(), SectionAddresses.end(), Target, 1186 [](uint64_t LHS, 1187 const std::pair<uint64_t, SectionRef> &RHS) { 1188 return LHS < RHS.first; 1189 }); 1190 if (SectionAddress != SectionAddresses.begin()) { 1191 --SectionAddress; 1192 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1193 } else { 1194 TargetSectionSymbols = nullptr; 1195 } 1196 } 1197 1198 // Find the first symbol in the section whose offset is less than 1199 // or equal to the target. 1200 if (TargetSectionSymbols) { 1201 auto TargetSym = std::upper_bound( 1202 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1203 Target, [](uint64_t LHS, 1204 const std::pair<uint64_t, StringRef> &RHS) { 1205 return LHS < RHS.first; 1206 }); 1207 if (TargetSym != TargetSectionSymbols->begin()) { 1208 --TargetSym; 1209 uint64_t TargetAddress = std::get<0>(*TargetSym); 1210 StringRef TargetName = std::get<1>(*TargetSym); 1211 outs() << " <" << TargetName; 1212 uint64_t Disp = Target - TargetAddress; 1213 if (Disp) 1214 outs() << "+0x" << utohexstr(Disp); 1215 outs() << '>'; 1216 } 1217 } 1218 } 1219 } 1220 outs() << "\n"; 1221 1222 // Print relocation for instruction. 1223 while (rel_cur != rel_end) { 1224 bool hidden = getHidden(*rel_cur); 1225 uint64_t addr = rel_cur->getOffset(); 1226 SmallString<16> name; 1227 SmallString<32> val; 1228 1229 // If this relocation is hidden, skip it. 1230 if (hidden) goto skip_print_rel; 1231 1232 // Stop when rel_cur's address is past the current instruction. 1233 if (addr >= Index + Size) break; 1234 rel_cur->getTypeName(name); 1235 error(getRelocationValueString(*rel_cur, val)); 1236 outs() << format(Fmt.data(), SectionAddr + addr) << name 1237 << "\t" << val << "\n"; 1238 1239 skip_print_rel: 1240 ++rel_cur; 1241 } 1242 } 1243 } 1244 } 1245 } 1246 1247 void llvm::PrintRelocations(const ObjectFile *Obj) { 1248 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1249 "%08" PRIx64; 1250 // Regular objdump doesn't print relocations in non-relocatable object 1251 // files. 1252 if (!Obj->isRelocatableObject()) 1253 return; 1254 1255 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1256 if (Section.relocation_begin() == Section.relocation_end()) 1257 continue; 1258 StringRef secname; 1259 error(Section.getName(secname)); 1260 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1261 for (const RelocationRef &Reloc : Section.relocations()) { 1262 bool hidden = getHidden(Reloc); 1263 uint64_t address = Reloc.getOffset(); 1264 SmallString<32> relocname; 1265 SmallString<32> valuestr; 1266 if (hidden) 1267 continue; 1268 Reloc.getTypeName(relocname); 1269 error(getRelocationValueString(Reloc, valuestr)); 1270 outs() << format(Fmt.data(), address) << " " << relocname << " " 1271 << valuestr << "\n"; 1272 } 1273 outs() << "\n"; 1274 } 1275 } 1276 1277 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1278 outs() << "Sections:\n" 1279 "Idx Name Size Address Type\n"; 1280 unsigned i = 0; 1281 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1282 StringRef Name; 1283 error(Section.getName(Name)); 1284 uint64_t Address = Section.getAddress(); 1285 uint64_t Size = Section.getSize(); 1286 bool Text = Section.isText(); 1287 bool Data = Section.isData(); 1288 bool BSS = Section.isBSS(); 1289 std::string Type = (std::string(Text ? "TEXT " : "") + 1290 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1291 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1292 Name.str().c_str(), Size, Address, Type.c_str()); 1293 ++i; 1294 } 1295 } 1296 1297 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1298 std::error_code EC; 1299 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1300 StringRef Name; 1301 StringRef Contents; 1302 error(Section.getName(Name)); 1303 uint64_t BaseAddr = Section.getAddress(); 1304 uint64_t Size = Section.getSize(); 1305 if (!Size) 1306 continue; 1307 1308 outs() << "Contents of section " << Name << ":\n"; 1309 if (Section.isBSS()) { 1310 outs() << format("<skipping contents of bss section at [%04" PRIx64 1311 ", %04" PRIx64 ")>\n", 1312 BaseAddr, BaseAddr + Size); 1313 continue; 1314 } 1315 1316 error(Section.getContents(Contents)); 1317 1318 // Dump out the content as hex and printable ascii characters. 1319 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1320 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1321 // Dump line of hex. 1322 for (std::size_t i = 0; i < 16; ++i) { 1323 if (i != 0 && i % 4 == 0) 1324 outs() << ' '; 1325 if (addr + i < end) 1326 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1327 << hexdigit(Contents[addr + i] & 0xF, true); 1328 else 1329 outs() << " "; 1330 } 1331 // Print ascii. 1332 outs() << " "; 1333 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1334 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1335 outs() << Contents[addr + i]; 1336 else 1337 outs() << "."; 1338 } 1339 outs() << "\n"; 1340 } 1341 } 1342 } 1343 1344 void llvm::PrintSymbolTable(const ObjectFile *o) { 1345 outs() << "SYMBOL TABLE:\n"; 1346 1347 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1348 printCOFFSymbolTable(coff); 1349 return; 1350 } 1351 for (const SymbolRef &Symbol : o->symbols()) { 1352 ErrorOr<uint64_t> AddressOrError = Symbol.getAddress(); 1353 error(AddressOrError.getError()); 1354 uint64_t Address = *AddressOrError; 1355 ErrorOr<SymbolRef::Type> TypeOrError = Symbol.getType(); 1356 error(TypeOrError.getError()); 1357 SymbolRef::Type Type = *TypeOrError; 1358 uint32_t Flags = Symbol.getFlags(); 1359 ErrorOr<section_iterator> SectionOrErr = Symbol.getSection(); 1360 error(SectionOrErr.getError()); 1361 section_iterator Section = *SectionOrErr; 1362 StringRef Name; 1363 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1364 Section->getName(Name); 1365 } else { 1366 Expected<StringRef> NameOrErr = Symbol.getName(); 1367 if (!NameOrErr) 1368 report_error(o->getFileName(), NameOrErr.takeError()); 1369 Name = *NameOrErr; 1370 } 1371 1372 bool Global = Flags & SymbolRef::SF_Global; 1373 bool Weak = Flags & SymbolRef::SF_Weak; 1374 bool Absolute = Flags & SymbolRef::SF_Absolute; 1375 bool Common = Flags & SymbolRef::SF_Common; 1376 bool Hidden = Flags & SymbolRef::SF_Hidden; 1377 1378 char GlobLoc = ' '; 1379 if (Type != SymbolRef::ST_Unknown) 1380 GlobLoc = Global ? 'g' : 'l'; 1381 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1382 ? 'd' : ' '; 1383 char FileFunc = ' '; 1384 if (Type == SymbolRef::ST_File) 1385 FileFunc = 'f'; 1386 else if (Type == SymbolRef::ST_Function) 1387 FileFunc = 'F'; 1388 1389 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1390 "%08" PRIx64; 1391 1392 outs() << format(Fmt, Address) << " " 1393 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1394 << (Weak ? 'w' : ' ') // Weak? 1395 << ' ' // Constructor. Not supported yet. 1396 << ' ' // Warning. Not supported yet. 1397 << ' ' // Indirect reference to another symbol. 1398 << Debug // Debugging (d) or dynamic (D) symbol. 1399 << FileFunc // Name of function (F), file (f) or object (O). 1400 << ' '; 1401 if (Absolute) { 1402 outs() << "*ABS*"; 1403 } else if (Common) { 1404 outs() << "*COM*"; 1405 } else if (Section == o->section_end()) { 1406 outs() << "*UND*"; 1407 } else { 1408 if (const MachOObjectFile *MachO = 1409 dyn_cast<const MachOObjectFile>(o)) { 1410 DataRefImpl DR = Section->getRawDataRefImpl(); 1411 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1412 outs() << SegmentName << ","; 1413 } 1414 StringRef SectionName; 1415 error(Section->getName(SectionName)); 1416 outs() << SectionName; 1417 } 1418 1419 outs() << '\t'; 1420 if (Common || isa<ELFObjectFileBase>(o)) { 1421 uint64_t Val = 1422 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1423 outs() << format("\t %08" PRIx64 " ", Val); 1424 } 1425 1426 if (Hidden) { 1427 outs() << ".hidden "; 1428 } 1429 outs() << Name 1430 << '\n'; 1431 } 1432 } 1433 1434 static void PrintUnwindInfo(const ObjectFile *o) { 1435 outs() << "Unwind info:\n\n"; 1436 1437 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1438 printCOFFUnwindInfo(coff); 1439 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1440 printMachOUnwindInfo(MachO); 1441 else { 1442 // TODO: Extract DWARF dump tool to objdump. 1443 errs() << "This operation is only currently supported " 1444 "for COFF and MachO object files.\n"; 1445 return; 1446 } 1447 } 1448 1449 void llvm::printExportsTrie(const ObjectFile *o) { 1450 outs() << "Exports trie:\n"; 1451 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1452 printMachOExportsTrie(MachO); 1453 else { 1454 errs() << "This operation is only currently supported " 1455 "for Mach-O executable files.\n"; 1456 return; 1457 } 1458 } 1459 1460 void llvm::printRebaseTable(const ObjectFile *o) { 1461 outs() << "Rebase table:\n"; 1462 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1463 printMachORebaseTable(MachO); 1464 else { 1465 errs() << "This operation is only currently supported " 1466 "for Mach-O executable files.\n"; 1467 return; 1468 } 1469 } 1470 1471 void llvm::printBindTable(const ObjectFile *o) { 1472 outs() << "Bind table:\n"; 1473 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1474 printMachOBindTable(MachO); 1475 else { 1476 errs() << "This operation is only currently supported " 1477 "for Mach-O executable files.\n"; 1478 return; 1479 } 1480 } 1481 1482 void llvm::printLazyBindTable(const ObjectFile *o) { 1483 outs() << "Lazy bind table:\n"; 1484 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1485 printMachOLazyBindTable(MachO); 1486 else { 1487 errs() << "This operation is only currently supported " 1488 "for Mach-O executable files.\n"; 1489 return; 1490 } 1491 } 1492 1493 void llvm::printWeakBindTable(const ObjectFile *o) { 1494 outs() << "Weak bind table:\n"; 1495 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1496 printMachOWeakBindTable(MachO); 1497 else { 1498 errs() << "This operation is only currently supported " 1499 "for Mach-O executable files.\n"; 1500 return; 1501 } 1502 } 1503 1504 /// Dump the raw contents of the __clangast section so the output can be piped 1505 /// into llvm-bcanalyzer. 1506 void llvm::printRawClangAST(const ObjectFile *Obj) { 1507 if (outs().is_displayed()) { 1508 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1509 "the clang ast section.\n" 1510 "Please redirect the output to a file or another program such as " 1511 "llvm-bcanalyzer.\n"; 1512 return; 1513 } 1514 1515 StringRef ClangASTSectionName("__clangast"); 1516 if (isa<COFFObjectFile>(Obj)) { 1517 ClangASTSectionName = "clangast"; 1518 } 1519 1520 Optional<object::SectionRef> ClangASTSection; 1521 for (auto Sec : ToolSectionFilter(*Obj)) { 1522 StringRef Name; 1523 Sec.getName(Name); 1524 if (Name == ClangASTSectionName) { 1525 ClangASTSection = Sec; 1526 break; 1527 } 1528 } 1529 if (!ClangASTSection) 1530 return; 1531 1532 StringRef ClangASTContents; 1533 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1534 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1535 } 1536 1537 static void printFaultMaps(const ObjectFile *Obj) { 1538 const char *FaultMapSectionName = nullptr; 1539 1540 if (isa<ELFObjectFileBase>(Obj)) { 1541 FaultMapSectionName = ".llvm_faultmaps"; 1542 } else if (isa<MachOObjectFile>(Obj)) { 1543 FaultMapSectionName = "__llvm_faultmaps"; 1544 } else { 1545 errs() << "This operation is only currently supported " 1546 "for ELF and Mach-O executable files.\n"; 1547 return; 1548 } 1549 1550 Optional<object::SectionRef> FaultMapSection; 1551 1552 for (auto Sec : ToolSectionFilter(*Obj)) { 1553 StringRef Name; 1554 Sec.getName(Name); 1555 if (Name == FaultMapSectionName) { 1556 FaultMapSection = Sec; 1557 break; 1558 } 1559 } 1560 1561 outs() << "FaultMap table:\n"; 1562 1563 if (!FaultMapSection.hasValue()) { 1564 outs() << "<not found>\n"; 1565 return; 1566 } 1567 1568 StringRef FaultMapContents; 1569 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1570 1571 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1572 FaultMapContents.bytes_end()); 1573 1574 outs() << FMP; 1575 } 1576 1577 static void printPrivateFileHeaders(const ObjectFile *o) { 1578 if (o->isELF()) 1579 printELFFileHeader(o); 1580 else if (o->isCOFF()) 1581 printCOFFFileHeader(o); 1582 else if (o->isMachO()) { 1583 printMachOFileHeader(o); 1584 printMachOLoadCommands(o); 1585 } else 1586 report_fatal_error("Invalid/Unsupported object file format"); 1587 } 1588 1589 static void printFirstPrivateFileHeader(const ObjectFile *o) { 1590 if (o->isELF()) 1591 printELFFileHeader(o); 1592 else if (o->isCOFF()) 1593 printCOFFFileHeader(o); 1594 else if (o->isMachO()) 1595 printMachOFileHeader(o); 1596 else 1597 report_fatal_error("Invalid/Unsupported object file format"); 1598 } 1599 1600 static void DumpObject(const ObjectFile *o) { 1601 // Avoid other output when using a raw option. 1602 if (!RawClangAST) { 1603 outs() << '\n'; 1604 outs() << o->getFileName() 1605 << ":\tfile format " << o->getFileFormatName() << "\n\n"; 1606 } 1607 1608 if (Disassemble) 1609 DisassembleObject(o, Relocations); 1610 if (Relocations && !Disassemble) 1611 PrintRelocations(o); 1612 if (SectionHeaders) 1613 PrintSectionHeaders(o); 1614 if (SectionContents) 1615 PrintSectionContents(o); 1616 if (SymbolTable) 1617 PrintSymbolTable(o); 1618 if (UnwindInfo) 1619 PrintUnwindInfo(o); 1620 if (PrivateHeaders) 1621 printPrivateFileHeaders(o); 1622 if (FirstPrivateHeader) 1623 printFirstPrivateFileHeader(o); 1624 if (ExportsTrie) 1625 printExportsTrie(o); 1626 if (Rebase) 1627 printRebaseTable(o); 1628 if (Bind) 1629 printBindTable(o); 1630 if (LazyBind) 1631 printLazyBindTable(o); 1632 if (WeakBind) 1633 printWeakBindTable(o); 1634 if (RawClangAST) 1635 printRawClangAST(o); 1636 if (PrintFaultMaps) 1637 printFaultMaps(o); 1638 if (DwarfDumpType != DIDT_Null) { 1639 std::unique_ptr<DIContext> DICtx(new DWARFContextInMemory(*o)); 1640 // Dump the complete DWARF structure. 1641 DICtx->dump(outs(), DwarfDumpType, true /* DumpEH */); 1642 } 1643 } 1644 1645 /// @brief Dump each object file in \a a; 1646 static void DumpArchive(const Archive *a) { 1647 for (auto &ErrorOrChild : a->children()) { 1648 if (std::error_code EC = ErrorOrChild.getError()) 1649 report_error(a->getFileName(), EC); 1650 const Archive::Child &C = *ErrorOrChild; 1651 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1652 if (std::error_code EC = ChildOrErr.getError()) 1653 if (EC != object_error::invalid_file_type) 1654 report_error(a->getFileName(), EC); 1655 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 1656 DumpObject(o); 1657 else 1658 report_error(a->getFileName(), object_error::invalid_file_type); 1659 } 1660 } 1661 1662 /// @brief Open file and figure out how to dump it. 1663 static void DumpInput(StringRef file) { 1664 1665 // If we are using the Mach-O specific object file parser, then let it parse 1666 // the file and process the command line options. So the -arch flags can 1667 // be used to select specific slices, etc. 1668 if (MachOOpt) { 1669 ParseInputMachO(file); 1670 return; 1671 } 1672 1673 // Attempt to open the binary. 1674 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 1675 if (!BinaryOrErr) 1676 report_error(file, errorToErrorCode(BinaryOrErr.takeError())); 1677 Binary &Binary = *BinaryOrErr.get().getBinary(); 1678 1679 if (Archive *a = dyn_cast<Archive>(&Binary)) 1680 DumpArchive(a); 1681 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 1682 DumpObject(o); 1683 else 1684 report_error(file, object_error::invalid_file_type); 1685 } 1686 1687 int main(int argc, char **argv) { 1688 // Print a stack trace if we signal out. 1689 sys::PrintStackTraceOnErrorSignal(); 1690 PrettyStackTraceProgram X(argc, argv); 1691 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 1692 1693 // Initialize targets and assembly printers/parsers. 1694 llvm::InitializeAllTargetInfos(); 1695 llvm::InitializeAllTargetMCs(); 1696 llvm::InitializeAllDisassemblers(); 1697 1698 // Register the target printer for --version. 1699 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 1700 1701 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 1702 TripleName = Triple::normalize(TripleName); 1703 1704 ToolName = argv[0]; 1705 1706 // Defaults to a.out if no filenames specified. 1707 if (InputFilenames.size() == 0) 1708 InputFilenames.push_back("a.out"); 1709 1710 if (DisassembleAll) 1711 Disassemble = true; 1712 if (!Disassemble 1713 && !Relocations 1714 && !SectionHeaders 1715 && !SectionContents 1716 && !SymbolTable 1717 && !UnwindInfo 1718 && !PrivateHeaders 1719 && !FirstPrivateHeader 1720 && !ExportsTrie 1721 && !Rebase 1722 && !Bind 1723 && !LazyBind 1724 && !WeakBind 1725 && !RawClangAST 1726 && !(UniversalHeaders && MachOOpt) 1727 && !(ArchiveHeaders && MachOOpt) 1728 && !(IndirectSymbols && MachOOpt) 1729 && !(DataInCode && MachOOpt) 1730 && !(LinkOptHints && MachOOpt) 1731 && !(InfoPlist && MachOOpt) 1732 && !(DylibsUsed && MachOOpt) 1733 && !(DylibId && MachOOpt) 1734 && !(ObjcMetaData && MachOOpt) 1735 && !(FilterSections.size() != 0 && MachOOpt) 1736 && !PrintFaultMaps 1737 && DwarfDumpType == DIDT_Null) { 1738 cl::PrintHelpMessage(); 1739 return 2; 1740 } 1741 1742 std::for_each(InputFilenames.begin(), InputFilenames.end(), 1743 DumpInput); 1744 1745 return EXIT_SUCCESS; 1746 } 1747